Antibiotics for acute maxillary sinusitis in adults

  • Review
  • Intervention

Authors


Abstract

Background

Sinusitis is one of the most common diagnoses among adults in ambulatory care, accounting for 15% to 21% of all adult outpatient antibiotic prescriptions. However, the role of antibiotics for sinusitis is controversial.

Objectives

To assess the effects of antibiotics in adults with acute maxillary sinusitis by comparing antibiotics with placebo, antibiotics from different classes and the side effects of different treatments.

Search methods

We searched CENTRAL 2013, Issue 2, MEDLINE (1946 to March week 3, 2013), EMBASE (1974 to March 2013), SIGLE (OpenSIGLE, later OpenGrey (accessed 15 January 2013)), reference lists of the identified trials and systematic reviews of placebo-controlled studies. We also searched for ongoing trials via ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform (ICTRP). We imposed no language or publication restrictions.

Selection criteria

Randomised controlled trials (RCTs) comparing antibiotics with placebo or antibiotics from different classes for acute maxillary sinusitis in adults. We included trials with clinically diagnosed acute sinusitis, confirmed or not by imaging or bacterial culture.

Data collection and analysis

Two review authors independently screened search results, extracted data and assessed trial quality. We calculated risk ratios (RRs) for differences between intervention and control groups in whether the treatment failed or not. All measures are presented with 95% confidence intervals (CIs). We conducted the meta-analyses using either the fixed-effect or random-effects model. In meta-analyses of the placebo-controlled studies, we combined data across antibiotic classes. Primary outcomes were clinical failure rates at 7 to 15 days and 16 to 60 days follow-up. We used GRADEpro to assess the quality of the evidence.

Main results

We included 63 studies in this updated review; nine placebo-controlled studies involving 1915 participants (seven of the studies clearly conducted in primary care settings) and 54 studies comparing different classes of antibiotics (10 different comparisons). Five studies at low risk of bias comparing penicillin or amoxicillin to placebo provided information on the main outcome: clinical failure rate at 7 to 15 days follow-up, defined as a lack of full recovery or improvement, for participants with symptoms lasting at least seven days. In these studies antibiotics decreased the risk of clinical failure (pooled RR of 0.66, 95% CI 0.47 to 0.94, 1084 participants randomised, 1058 evaluated, moderate quality evidence). However, the clinical benefit was small. Cure or improvement rates were high in both the placebo group (86%) and the antibiotic group (91%) in these five studies. When clinical failure was defined as a lack of full recovery (n = five studies), results were similar: antibiotics decreased the risk of failure (pooled RR of 0.73, 95% CI 0.63 to 0.85, high quality evidence) at 7 to 15 days follow-up.

Adverse effects in seven of the nine placebo-controlled studies (comparing penicillin, amoxicillin, azithromycin or moxicillin to placebo) were more common in antibiotic than in placebo groups (median of difference between groups 10.5%, range 2% to 23%). However, drop-outs due to adverse effects were rare in both groups: 1.5% in antibiotic groups and 1% in control groups.

In the 10 head-to-head comparisons, none of the antibiotic preparations were superior to another. However, amoxicillin-clavulanate had significantly more drop-outs due to adverse effects than cephalosporins and macrolides.

Authors' conclusions

There is moderate evidence that antibiotics provide a small benefit for clinical outcomes in immunocompetent primary care patients with uncomplicated acute sinusitis. However, about 80% of participants treated without antibiotics improved within two weeks. Clinicians need to weigh the small benefits of antibiotic treatment against the potential for adverse effects at both the individual and general population levels.

Résumé scientifique

Antibiotiques pour la sinusite maxillaire aiguë chez l'adulte

Contexte

La sinusite constitue l'un des diagnostics les plus fréquents chez les adultes en soins ambulatoires, représentant 15 % à 21 % des prescriptions d'antibiotiques pour les patients adultes en ambulatoire. Cependant, le rôle des antibiotiques dans le traitement de la sinusite est controversé.

Objectifs

Évaluer les effets des antibiotiques chez les adultes atteints de sinusite maxillaire aiguë en comparant les antibiotiques à un placebo, les antibiotiques de différentes classes et les effets secondaires des différents traitements.

Stratégie de recherche documentaire

Nous avons effectué des recherches dans CENTRAL 2013, numéro 2, MEDLINE (de 1946 à la 3ème semaine de mars 2013), EMBASE (de 1974 à mars 2013), SIGLE (OpenSIGLE, plus tard OpenGrey (accès le 15 janvier 2013)), les références bibliographiques des essais identifiés et les revues systématiques sur des études contrôlées par placebo. Nous avons également recherché des essais en cours via ClinicalTrials.gov et le système d'enregistrement international des essais cliniques (ICTRP) de l'OMS. Nous n'avons imposé aucune restriction concernant la langue ou la publication.

Critères de sélection

Essais contrôlés randomisés (ECR) comparant des antibiotiques à un placebo, ou des antibiotiques de différentes classes, pour la sinusite maxillaire aiguë chez l'adulte. Nous avons inclus les essais portant sur des patients souffrant d'une sinusite aiguë diagnostiquée cliniquement, confirmée ou non par imagerie médicale ou culture bactérienne.

Recueil et analyse des données

Deux auteurs de la revue ont indépendamment passé au crible les résultats de recherche, extrait les données et évalué la qualité des essais. Les risques relatifs (RR) ont été calculés pour les différences entre les groupes d'intervention et témoin dans l'échec ou non du traitement. Toutes les mesures sont présentées avec des intervalles de confiance (IC) à 95 %. Nous avons effectué les méta-analyses en utilisant le modèle à effets fixes ou à effets aléatoires. Dans les méta-analyses des études contrôlées par placebo, nous avons combiné les données de plusieurs classes d'antibiotiques. Les critères de jugement principaux étaient les taux d'échec clinique entre 7 et 15 jours de suivi et entre 16 et 60 jours de suivi. Nous avons utilisé GRADEpro pour évaluer la qualité des preuves.

Résultats principaux

Nous avons inclus 63 études dans cette revue mise à jour ; neuf études contrôlées par placebo portant sur 1 915 participants (sept études clairement réalisées dans des environnements de soins primaires) et 54 études comparant différentes classes d'antibiotiques (10 comparaisons différentes). Cinq études à faible risque de biais comparant la pénicilline ou l'amoxicilline à un placebo ont fourni des informations sur le principal critère de jugement, le taux d'échec clinique au bout de 7 à 15 jours de suivi, défini comme l'absence de guérison complète ou d'amélioration, pour les participants présentant des symptômes d'une durée d'au moins sept jours. Dans ces études, les antibiotiques réduisaient le risque d'échec clinique (RR combiné de 0,66, IC à 95 % 0,47 à 0,94, 1 084 participants randomisés, 1 058 évalués, preuves de qualité modérée). Toutefois, l'avantage clinique était faible. Les taux de guérison ou d'amélioration étaient élevés tant dans les groupes sous placebo (86 %) que les groupes sous antibiotiques (91 %) dans ces cinq études. Lorsque l'échec clinique était défini comme l'absence de guérison complète (n = cinq études), les résultats étaient similaires : les antibiotiques réduisaient le risque d'échec (RR combiné de 0,73, IC à 95 % 0,63 à 0,85, preuves de haute qualité) au bout de 7 à 15 jours de suivi.

Les effets indésirables dans sept des neuf études contrôlées par placebo (comparant la pénicilline, l'amoxicilline, l'azithromycine ou la moxicilline à un placebo) ont été plus fréquents dans les groupes sous antibiotique que dans les groupes sous placebo (médiane de différence entre les groupes de 10,5 %, fourchette de 2 % à 23 %). Cependant, les sorties d'étude en raison d'effets indésirables étaient rares dans les deux groupes : 1,5 % dans les groupes d'antibiotiques et 1 % dans les groupes de contrôle.

Dans les 10 comparaisons en face à face, aucune des préparations antibiotiques n'était supérieure à une autre. Cependant, le groupe sous amoxicilline-clavulanate avait significativement plus de sorties d'étude en raison d'effets indésirables que les groupes traités par les céphalosporines et les macrolides.

Conclusions des auteurs

Des preuves modérées indiquent que les antibiotiques procurent un léger avantage pour les résultats cliniques chez les patients immunocompétents, en milieu de soins primaires, souffrant d'une sinusite aiguë sans complications. Cependant, environ 80 % des participants traités sans antibiotiques allaient mieux dans un délai de deux semaines. Les cliniciens doivent mettre en balance les faibles bénéfices du traitement antibiotique avec les effets indésirables potentiels sur le plan individuel et au niveau de la population générale.

Plain language summary

Antibiotics for acute maxillary sinusitis

Sinusitis is one of the most common reasons for visiting a doctor and an estimated 20 million cases of acute sinusitis occur every year in the USA alone. There are four pairs of sinuses linked to the bony structures around the nose: maxillary, frontal, ethmoidal and sphenoidal sinuses. In sinusitis, these membrane-lined air spaces become infected, which causes pain and discharge from the nose. Treatment options include decongestants, steroid drops or sprays, mucus-clearing drugs (mucolytics), antihistamines and antibiotics and sometimes sinus puncture and irrigation for removal of purulent secretions.

In most cases sinusitis occurs during viral infections where antibiotics are ineffective, but the few cases that also have a bacterial infection (one or two out of every 100 patients with sinus symptoms) could benefit. Unfortunately it is difficult to distinguish between those who have a bacterial infection and those who do not. It is important to avoid unnecessary use of antibiotics and limit the potential for antibiotic resistance.

This updated review compiled data from 63 separate studies that used a variety of antibiotics for simple maxillary sinus infection, i.e. non-complicated acute sinusitis in a person with a healthy immune system. Nine of the studies compared antibiotics to placebo (1915 participants; seven of the studies conducted in primary care), and 54 studies compared different classes of antibiotics. Five of the nine placebo-controlled studies involving 1058 participants found that most participants got better within two weeks, regardless of whether they received the antibiotic or not (roughly 9 out of 10 participants in antibiotic groups and 8 out of 10 in placebo groups). Although all the five studies in this main outcome were assessed as having low risk of bias, the overall evidence was assessed only as being of moderate quality (it is possible that a new large study can change the estimate). In the remaining 54 studies comparing different antibiotics (10 different comparisons), no antibiotic was found to be superior to another. The evidence is current to March 2013.

The small benefit gained by antibiotics may be overridden by the negative effects of the drugs. In addition to patient-related adverse effects (like skin rash and gastrointestinal problems such as diarrhea, abdominal pain and vomiting), side effects include the risk of increased resistance to antibiotics amongst community-acquired pathogens (bacteria).

This review found that antibiotics will help some people a bit, but do not make a major difference to most people with acute maxillary sinusitis being treated in primary care.

Résumé simplifié

Les antibiotiques pour le traitement de la sinusite maxillaire aiguë

La sinusite constitue l'un des motifs de consultation médicale les plus courants et environ 20 millions de cas de sinusite aiguë se produisent tous les ans aux seuls États-Unis. Il existe quatre paires de sinus liés aux structures osseuses autour du nez : les sinus maxillaires, frontaux, ethmoïdaux et sphénoïdaux. Dans la sinusite, ces cavités aérées tapissées d'une membrane s'infectent, provoquant ainsi des douleurs et des écoulements nasaux. Les options de traitement comprennent des décongestionnants, des stéroïdes sous forme de gouttes ou de pulvérisateur, des agents permettant l'élimination des mucosités (mucolytiques), des antihistaminiques et des antibiotiques, et parfois la ponction et le lavage des sinus pour en retirer les sécrétions purulentes.

Dans la plupart des cas, la sinusite accompagne les infections virales pour lesquelles les antibiotiques sont inefficaces, mais ils pourraient être bénéfiques dans les rares cas présentant également une infection bactérienne (un ou deux patients sur 100 avec des symptômes des sinus). Il est malheureusement difficile de savoir si l'infection est d'origine bactérienne ou non, et il est important d'éviter l'utilisation inutile d'antibiotiques afin de limiter le risque d'antibiorésistance.

Cette revue mise à jour a recueilli des données de 63 études distinctes ayant utilisé une variété d'antibiotiques pour l'infection des sinus maxillaires simple, c.-à-d. une sinusite aiguë non compliquée chez une personne avec un système immunitaire en bonne santé. Neuf de ces études comparaient les antibiotiques à un placebo (1 915 participants ; sept des études menées en soins primaires), et 54 études comparaient différentes classes d'antibiotiques. Cinq des neuf études contrôlées par placebo portant sur 1 058 participants ont constaté que la plupart des participants allaient mieux dans un délai de deux semaines, qu'ils aient reçu l'antibiotique ou non (environ 9 sur 10 participants dans les groupes d'antibiotiques et 8 sur 10 dans les groupes sous placebo). Bien que les cinq études sur ce critère de jugement principal aient toutes été évaluées comme présentant un faible risque de biais, dans l'ensemble les preuves ont été évaluées seulement comme étant de qualité modérée (il est possible qu'une nouvelle étude à grande échelle puisse modifier l'estimation). Dans les 54 autres études comparant différents antibiotiques (10 comparaisons différentes), aucun antibiotique ne s'est avéré supérieur à un autre. Les preuves sont à jour en mars 2013.

Le léger avantage que procuraient les antibiotiques peut être annulé par les effets négatifs des médicaments. En plus des effets indésirables relatifs au patient (tels que les éruptions cutanées et les troubles gastro-intestinaux comme la diarrhée, les douleurs abdominales et les vomissements), les effets secondaires comprennent le risque de résistance accrue aux antibiotiques des agents pathogènes (bactéries) communautaires.

Selon cette revue, les antibiotiques permettent une légère amélioration chez certaines personnes, mais n'apportent aucune différence majeure pour la plupart des patients souffrant d'une sinusite maxillaire aiguë traités en soins primaires.

Notes de traduction

Traduit par: French Cochrane Centre 22nd June, 2014
Traduction financée par: Financeurs pour le Canada : Instituts de Recherche en Santé du Canada, Ministère de la Santé et des Services Sociaux du Québec, Fonds de recherche du Québec-Santé et Institut National d'Excellence en Santé et en Services Sociaux; pour la France : Ministère en charge de la Santé

Summary of findings(Explanation)

Summary of findings for the main comparison. Antibiotics compared to placebo for uncomplicated acute maxillary sinusitis
  1. 1Five studies at 'low' risk of bias for the five key domains of allocation concealment, blinding, selective outcome reporting, incomplete outcome data and baseline comparability.
    2Because the effect size seems to be small and the number of evaluated participants so far is 1058, it is not unreasonable to anticipate that a new large study could change this estimate.
    3Five studies at 'low' risk of bias for the five key domains of allocation concealment, blinding, selective outcome reporting, incomplete outcome data and baseline comparability.
    4When including one study at 'unclear' risk of bias in the analysis, the results did not change (RR of 0.73, 95% CI 0.63 to 0.85).
    5Only one small study at 'low' risk of bias (Haye 1998).
    6Also the study of Hadley 2010 at 'unclear' risk of bias (comparing moxifloxacin to placebo) did not find a significant difference between antibiotic and placebo either (RR value of 0.62, 95% CI 0.37 to 1.03). These two studies represent almost the utmost ends of the clinical variation; Haye 1998 excluded participants with radiographic findings more than mild and Hadley 2010 included only participants with bacteriologically confirmed sinusitis.
    7Two studies at 'unclear' risk of bias, published in 2010 and 2012 in the USA.
    8Variation of reported adverse effects in eight studies.

Antibiotics compared to placebo for uncomplicated acute maxillary sinusitis
Patient or population: patients with uncomplicated acute maxillary sinusitis
Settings: primary care
Intervention: antibiotics: penicillin V, amoxicillin and azithromycin
Comparison: placebo
OutcomesIllustrative comparative risks* (95% CI)Relative effect
(95% CI)
No of participants
(studies)
Quality of the evidence
(GRADE)
Comments
Assumed riskCorresponding risk
Placebo Antibiotics
Clinical failure rate
Follow-up:
7 to 15 days
Failure defined as a lack of full recovery or improvement

136 per 1000
Failure defined as a lack of full recovery or improvement

90 per 1000
(64 to 128)
RR 0.66
(0.47 to 0.94)
1084 patients randomised and 1058 evaluated
(5 studies1)
⊕⊕⊕⊝
moderate 2

Adverse effects8:
antibiotics 8% to 59%;

placebo 6% to 38%

Failure defined as a lack of full recovery

614 per 1000
Failure defined as a lack of full recovery

448 per 1000
(387 to 522)
RR 0.73
(0.63 to 0.85)
716 patients randomised and 680 evaluated
(5 studies3)
⊕⊕⊕⊕
high 4
Clinical failure rate
Follow-up: 16 to 60 days
Failure defined as a lack of full recovery or improvement

122 per 1000
Failure defined as a lack of full recovery or improvement

104 per 1000
(44 to 241)
RR 0.85
(0.36 to 1.98)
169 patients randomised and 169 evaluated
(1 study)

⊕⊕⊝⊝

low 5,6

Failure defined as a lack of full recovery

329 per 1000
Failure defined as a lack of full recovery

207 per 1000
(125 to 346)
RR 0.63
(0.38 to 1.05)
169 patients randomised and 169 evaluated
(1 study)
⊕⊕⊝⊝
low 5
 
Quality of life and activity impairment
Follow-ups:
3, 6 to 8, 10 days
Impact Number of participants
(studies)
⊕⊝⊝⊝
very low
 
Inconsistent effect. There were differences in: outcome measures, follow-up times and results270 patients evaluated
(2 studies7)
 
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: confidence interval; RR: risk ratio
GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

Summary of findings 2 Antibiotics compared to other antibiotics for uncomplicated acute maxillary sinusitis

Summary of findings 2. Antibiotics compared to other antibiotics for uncomplicated acute maxillary sinusitis
  1. 1Many of the 54 studies were non-inferiority studies.
    2Four studies were at 'low' risk of bias and the other 50 at 'unclear' or 'high' risk of bias.
    3Amoxicillin-clavulanate had significantly more drop-outs due to adverse effects than cephalosporins and macrolides.

Antibiotics compared to other antibiotics for uncomplicated acute maxillary sinusitis
Patient or population: patients with uncomplicated acute maxillary sinusitis
Settings: primary care
Comparisons: non-penicillin antibiotic versus beta-lactamase sensitive penicillins (n = 8), non-tetracycline versus tetracycline (n = 5, of which one study was also included in the comparison of non-penicillin antibiotic versus beta-lactamase sensitive penicillins), macrolides versus amoxicillin-clavulanate (n = 11), cephalosporins versus amoxicillin-clavulanate (n = 10) and miscellaneous comparisons (n = 21)
OutcomesImpactNo of participants
(studies)
Quality of the evidence
(GRADE)
Comments
Clinical failure rate (failure defined as a lack of full recovery or improvement or a lack of full recovery)
Follow-ups:
7 to 15 days and 16 to 60 days
No difference in failure rates54 studies representing 10 different comparisons1⊕⊕⊕⊝
moderate 2
Adverse effects3
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: confidence interval; RR: risk ratio
GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.

Background

Description of the condition

Sinusitis is a prevalent and important cause of ill health in adults. In the US alone, an estimated 20 million cases of acute sinusitis occur each year (Gwaltney 2005). Sinusitis is the third to fifth most common diagnosis for which an antibiotic is prescribed within primary care settings in Nordic countries (Andre 2002; Rautakorpi 1999) and the US (SAHP 2004). Sinusitis accounts for 15% to 21% of all antibiotic prescriptions for adults in outpatient care.

Acute bacterial maxillary sinusitis is often preceded by an acute viral upper respiratory tract infection (URTI). Up to 90% of patients with acute URTIs have symptoms of rhinosinusitis (Gwaltney 1994). Following a common cold lasting for one week, up to 39% of adults (Puhakka 1998) have reversible abnormalities in the sinus cavity which show up in magnetic resonance imaging or X-ray. It has been estimated that 0.5% to 2% of patients with a common cold have complications in the form of acute bacterial infection of the sinuses (Berg 1986; Gwaltney 1996). However, distinguishing patients with bacterial infection from those with symptoms of rhinosinusitis with a viral origin is challenging.

A bacterial sinus infection can be caused by one or more bacterial species. Commonly isolated bacteria include Streptococcus pneumoniae (S. pneumoniae),Haemophilus influenzae (H. influenzae) and Moraxella catarrhalis (M. catarrhalis) (Gwaltney 1992; Low 1997). In approximately a third of suspected bacterial sinusitis cases, bacterial cultures from the sinus cavity come back negative (Axelsson 1972; Gwaltney 1992; Jousimies-Somer 1988).

Formally sinusitis is classified as acute or chronic, depending on the pathological findings and duration of symptoms (Fokkens 2007; ICD-10). In this classification, acute sinusitis is defined as lasting less than 12 weeks. However, clinically sinusitis can be classified to more than two classes: acute, subacute or chronic, depending on the duration of the symptoms (Lanza 1997; Meltzer 2004; Slavin 2005). In this case acute maxillary sinusitis is usually defined as lasting less than four weeks and subacute as four to 12 weeks (Ah-See 2007; Lanza 1997). Subacute maxillary sinusitis, although not a distinct disease entity from acute sinusitis, may differ histopathologically and require different therapy from acute or chronic maxillary sinusitis (Hickner 2001; Lanza 1997). Many cases of chronic sinusitis again represent a separate entity with underlying problems, for example, mechanical obstruction of sinus drainage, abnormalities in mucociliary clearance or immunology (Gwaltney 2005). For research purposes it is recommended that acute sinusitis be defined as lasting less than four weeks (Meltzer 2006).

Typical signs and symptoms of acute sinusitis include purulent nasal discharge, postnasal drip, sinus pain at palpation, nasal obstruction with poor response to decongestants, unilateral facial pain and maxillary toothache (Axelsson 1972; Williams 1993), but none of the signs or symptoms is diagnostic when presenting alone. Acute bacterial sinusitis is more likely if the symptoms have lasted for more than one week (Gwaltney 2005).

Description of the intervention

Clinical guidelines for acute sinusitis currently recommend antibiotic treatment for patients with severe or persistent moderate symptoms and specific findings of bacterial sinusitis (Fokkens 2007; Slavin 2005; Snow 2001). Unnecessary antibiotic prescriptions should be avoided for several reasons: in addition to patient-related adverse effects, side effects are associated with resistance to antibiotics among community-acquired pathogens. The correlation between resistance and community antibiotic use has been seen in many countries (Albrich 2004; Arason 1996; Bronzwaer 2002; Goossens 2005; Seppälä 1995; Steinke 2001). Not only the volume of antibiotic use but also the selection of broad-spectrum drugs, low doses and long duration of antibiotic treatment increase antibiotic resistance (Guillemot 1998; Hay 2005; Odenholt 2003). The European Antimicrobial Resistance Surveillance System (EARSS) has revealed significant geographical differences in resistance rates within Europe (Goossens 2005); high rates of antibiotic resistance were seen more often in southern and eastern European countries with high antibiotic use. For example, there are remarkable differences between European countries in the prevalence of resistance to penicillin and macrolide antibiotics in treating S. pneumoniae (EARSS 2008). In recent years, an increasing number of published studies have also suggested that antibiotics might have other harmful effects on health through the disturbance of human microbiota (Kilkkinen 2002; Maxwell 2002; Velicer 2004) and perhaps predispose to new infections (Arason 2005; Howard 1976; Joki-Erkkilä 2000; Margolis 2005; Smith 1997).

How the intervention might work

The purpose of antibiotics is, by eradicating bacteria from the sinuses, to decrease the signs and symptoms of infections, restore the normal function of the sinuses and to prevent complications and the development of chronic sinusitis.

Why it is important to do this review

The purpose of this systematic review was to quantify the effectiveness of antibiotic therapy for patients diagnosed with acute sinusitis and treated in ambulatory settings. The word 'antibiotic' is used as a general term referring to all antibacterials.

Objectives

To compare the effects of antibiotics versus placebo on clinical failure rates for acute maxillary sinusitis.
To compare different classes of antibiotics for treatment of acute maxillary sinusitis.
To compare the effect of short-course versus long-course antibiotics for acute maxillary sinusitis.
To compare the side effects of different treatments.

Methods

Criteria for considering studies for this review

Types of studies

Randomised controlled trials (RCTs) evaluating and comparing antibiotics to a placebo, or different classes of antibiotics for acute sinusitis, and reported in full-text.

We included trials having a sample size of at least 30 participants with acute maxillary sinusitis. This is to guarantee that data in individual studies are as unbiased as possible. Also in very small samples many estimators are known to be sensitive to variation.

We excluded studies reported only as abstracts because there is evidence that there are discrepancies between data reported in the abstract and the final published full report and that information on trial quality indicators is often lacking (Chokkalingam 1998; Hopewell 2006). Thus we required full-text reports to ensure reliable data extraction and assessment of risk of bias. To diminish the risk of publication bias, we attempted to contact authors of potential abstracts to obtain information as to whether a full-text report of the study (unpublished or published) was available.

Types of participants

We included trials with adults or trials that separately reported data on subgroups of adults. We accepted adolescents (at least 12 years old) if less than 20% of participants were under 18 years of age.

Acute maxillary sinusitis was defined by:

  1. a history of URTI lasting seven to 30 days, with at least two clinical signs or symptoms (sinus pain at palpation, postnasal drip, purulent nasal discharge, nasal obstruction, unilateral facial pain, maxillary toothache, impaired sense of smell); or

  2. radiography, ultrasound or other imaging; or

  3. bacterial culture from a sinus secretion obtained by puncture or endoscopy and irrigation or aspiration.

In studies where the clinical diagnosis was not clearly described, diagnosis of acute maxillary sinusitis should be confirmed in at least of 80% of participants by imaging or culture.

We included trials with a mixed population of acute (symptoms less than 30 days) and non-acute sinusitis or acute exacerbations of chronic sinusitis if they separately reported data on the subgroup with acute sinusitis, or if at least 80% of participants had acute sinusitis.

We excluded trials that focused on patients with complicated sinusitis such as pansinusitis or frontal sinusitis (or solely ethmoidal or sphenoidal sinusitis), or infections of dental origin.

Types of interventions

We reviewed the following drug therapies:

  1. antibiotics versus control; and

  2. comparisons between different antibiotic classes.

We systematically recorded co-interventions such as decongestants, antihistamines, mucolytics, non-steroidal anti-inflammatory drugs and corticosteroids.

Types of outcome measures

Primary outcomes
  1. Clinical failure rate at 7 to 15 days after the start of treatment. Failure is defined as lack of full recovery or improvement of participants with acute maxillary sinusitis at follow-up.

  2. Clinical failure rate at 16 to 60 days after the start of treatment. Failure is defined as lack of full recovery or improvement of participants with acute maxillary sinusitis at follow-up.

As spontaneous cure rate or cure plus improvement rate in maxillary sinusitis is expected to also be fairly high in the placebo group, we decided after careful consideration to select the negative outcome 'failure' instead of 'success'. This will enable clinicians to consider the risks of not prescribing antibiotics as opposed to the benefits of prescribing them, compared to possible adverse effects of antibiotics.

Secondary outcomes
  1. Clinical failure rate at 7 to 15 days after the start of treatment. Failure is defined as lack of full recovery of participants with acute maxillary sinusitis at follow-up.

  2. Clinical failure rate at 16 to 60 days after the start of treatment. Failure is defined as lack of full recovery of participants with acute maxillary sinusitis at follow-up.

  3. Bacteriological failure.

  4. Radiographic failure.

  5. Relapse rates: new acute episodes of sinusitis after 60 days from the start of the treatment.

  6. Drop-outs due to adverse effects.

  7. Quality of life.

  8. Activity impairment and ability to work.

Search methods for identification of studies

Electronic searches

For this 2013 update we searched the Cochrane Central Register of Controlled Trials (CENTRAL) 2013, Issue 2, part of The Cochrane Library, www.thecochranelibrary.com (accessed 20 March 2013), which contains the Cochrane Acute Respiratory Infections Group's Specialised Register, MEDLINE (1946 to March week 3, 2013) and EMBASE (1974 to March 2013).

We searched CENTRAL and MEDLINE using the search strategy described in Appendix 1. The MEDLINE search was combined with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity- and precision-maximising version; Ovid format (Lefebvre 2011). We adapted the search strategy for EMBASE (Appendix 2).

We also searched:

  1. the System for Information on Grey Literature in Europe (SIGLE) to locate grey literature. SIGLE is currently known as OpenGrey. The search strategy was adapted to the new system's search language. The search was run on the Exalead search engine at http://www.opengrey.eu/ (accessed 15 January 2013) (Appendix 3);

  2. ClinicalTrials.gov (to 4 February 2013) to locate ongoing trials (Appendix 4);

  3. WHO International Clinical Trials Registry Platform (ICTRP) http://www.who.int/ictrp (to 4 February 2013) to locate ongoing trials (Appendix 4).

Searching other resources

We used the function for finding related articles in PubMed for the already identified placebo-controlled trials to track down additional, relevant articles. We reviewed reference lists of the identified trials and 14 systematic reviews (13 previously identified systematic reviews Appendix 5, and the new Cochrane review Lemiengre 2012) considering placebo-controlled study designs for additional, appropriate studies.

For a previous review version (Williams 2003), pharmaceutical companies that manufacture antibiotics used in the treatment of acute sinusitis were contacted and data and references from all published and unpublished trials on acute sinusitis were requested. Further, three experts in the field had been contacted and asked to review the bibliography of the initial review (Williams 1999) for completeness.

Details of previous searches are in Appendix 5.

Data collection and analysis

Selection of studies

Two review authors (AAS and UMR) independently selected papers on the basis of the title, keywords and abstract, and decided on eligibility. We obtained the full text of every article considered for inclusion. If the information relevant to the inclusion criteria was unavailable in the abstract or if the title was relevant but the abstract was unavailable, we obtained the full text of the report. We resolved any disagreements by consensus between these two review authors.

We contacted trial authors of the placebo-controlled studies to obtain additional information if the study seemed to fulfil the inclusion criteria for this review but the information in the report was insufficient to make the final assessment of inclusion or exclusion.

Data extraction and management

Two review authors (AAS, UMR) extracted data from all included studies. We also extracted data presented only in graphs and figures whenever possible.

Extracted information relating to study methodology or quality included: randomisation concealment, blinding, time of follow-up, percentage of missing data during follow-up and baseline comparability of the groups. Characteristics relating to methods that we extracted included: criteria for accepting participants into the study (diagnosis on clinical/radiography/culture basis), definition of cure/failure and treatment compliance.

Characteristics relating to participants that we extracted included the number of participants in treatment and control groups at start, the mean age of participants (plus age range), the number of men and women, the year the study was published, the location where the study was conducted (country and setting where participants were recruited). Characteristics of the intervention that we extracted included: intervention comparisons, courses of antibiotics and information about co-interventions.

We extracted information on clinical, bacteriological and radiographic outcomes and adverse event rates. We mainly extracted outcome information as numbers of participants with clinical failure/non-failure by study group. We classified relapse within 60 days from the onset of the treatment as a clinical failure.

In some studies the results were given at more than one follow-up point. We extracted all data for preselected times, which were 7 to 15 days, 16 to 60 days and over 60 days of follow-up. We carried out meta-analyses at these preselected times based on the available data.

Two review authors (AAS, UMR) agreed on final results and reached decisions by consensus.

Assessment of risk of bias in included studies

Two authors (AAS, UMR) independently assessed the risk of bias of the included studies. We resolved disagreements by consensus. Attempts were made to contact trial authors of the included placebo-controlled studies if the information in the report was insufficient for assessment. We presented the questions as open-ended, for example, "how was randomisation performed in practice?" As recommended by the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a), we assessed the following six domains. Within each domain, we judged each study as having a 'low', 'high' or 'unclear' risk of bias, the latter indicating either lack of information or uncertainty over the potential for bias.

Random sequence generation
Was the method used to generate the allocation sequence appropriate to produce comparable groups?

We graded this domain 'low' risk of bias if authors described a random component in the sequence generation process (for example, random number table, coin tossing, drawing of lots). If there was no or insufficient information about the random sequence generation process, we graded this domain 'unclear' risk. In this review we excluded studies which were identified as quasi-randomised (for example, if allocation method was alternation, use of case record numbers, dates of birth or day of the week).

Allocation concealment
Was the method used to conceal the allocation sequence appropriate to prevent the allocation being known in advance of, or during, enrolment?

We graded this domain 'low' risk of bias if the authors described adequate concealment (for example, by means of central randomisation or sequentially numbered, opaque and sealed envelopes) and 'high' risk of bias if inadequate concealment was documented or allocation concealment was not used. If no information about allocation concealment was provided or the information was unclearly reported, we graded this domain 'unclear' risk of bias.

Blinding
Were participants, personnel and investigators (outcome assessors) blinded about the intervention a participant had received?

We graded this domain 'low' risk of bias if at least the participant and the investigator were blinded. If the study only reported a double-blind design without detailed description of the methods of blinding, we graded this domain 'unclear' risk of bias.

Incomplete outcome data
How complete were the outcome data for the clinical outcomes? Were drop-out rates and reasons for withdrawals reported? Were missing data imputed appropriately?

The judgements were restricted to clinical efficacy outcomes at 7 to 15 or at 16 to 60 days after the start of treatment. If the clinical efficacy outcomes in a trial were reported at both follow-up periods, the judgement of this domain was based on the 7 to 15 days after the start of treatment but the risk of bias was separately assessed and reported in the 'Risk of bias' table also for the second follow-up period in case it was used in the review analyses. We decided to grade this domain to 'low' risk of bias if the total proportion of the missing outcome data was marginal (less than 5%); or when the proportion of the missing outcome data was less than 20% and the groups were balanced in numbers and reasons for missing data (regardless of reason or population used in the analyses); or if missing data were imputed using appropriate methods. Also, if reasons for exclusion of the participants after randomisation were justifiable and not seen to affect the randomisation, a higher proportion of missing data could be accepted. For example, if a positive culture of sinus excretion was defined as inclusion criterion, a higher number of exclusions due to absence of pathogen in pretreatment cultures is justifiable because it is anticipated and the results of the cultures are available only several days after the start of the treatment. If there was no information on reasons for missing data across intervention groups or proportions of missing data were documented as total proportion (5% to 20%), not by groups, the judgement was 'unclear' risk of bias. Classifying the total proportion of the missing data over 20% as 'high' risk of bias (besides the exception described above) was a pragmatic approach to this domain to make the judgement uniform and transparent.

Selective reporting
Were appropriate outcomes reported and were key outcomes missing?

To be included in this review, clinical outcomes had to be reported. However, studies could also report other (secondary) outcomes, for example radiological and bacteriological responses, relapse rates, chronic evolution and adverse effects. In this review, we graded freedom from selective reporting 'low' risk of bias if the study’s prespecified outcomes (those of interest in this review) had been reported in the prespecified way.

Other sources of bias

This domain included information on the baseline comparability of the intervention and control groups, on possible use of the co-interventions by group and on funding. We assessed each of these features in their own entries.

  1. Balance in baseline characteristics between the study groups is an essential feature representing the methodological quality of the trial. We graded this entry 'high' risk of bias if there were important differences in demographic characteristics or sinusitis severity at baseline between the study groups. If baseline characteristics were not described clearly enough we graded this entry as 'unclear' risk of bias.

  2. We graded the co-intervention entry 'low' risk of bias if the groups were balanced in amount and quality of co-interventions or no co-interventions were included in the protocol, and 'high' risk of bias if the groups received a different amount or quality of co-interventions during the trial. If no information was provided on co-interventions we graded this entry 'unclear' risk of bias.

  3. We graded the funding entry 'low' risk of bias if the funding was only from academic or public sources, and 'high' risk of bias if the study was funded by a pharmaceutical company, since there is evidence that industry-sponsored trials may overestimate the treatment effect (Bekelman 2003; Bhandari 2004; Lexchin 2003). The judgement was 'unclear' risk if there was pharmaceutical industry funding but the researchers had full rights to the data and manuscript or one or more of the authors had affiliations to pharmaceutical company; or when the funding source was unclear.

We completed a 'Risk of bias' table for each included study (Characteristics of included studies). Results are presented graphically by domain over all studies (Figure 1) and by study (Figure 2).

Figure 1.

'Risk of bias' graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.

Figure 2.

'Risk of bias' summary: review authors' judgements about each risk of bias item for each included study.

Summary assessments of risk of bias

To draw conclusions about the overall risk of bias for the main clinical outcomes within a study, we decided to classify the studies into three categories: studies with 'low', 'unclear' or 'high' risk of bias. The five main clinical outcomes were: clinical failure rate at 7 to 15 days after the start of treatment and failure defined as 1) a lack of full recovery or improvement of participants, or 2) a lack of full recovery of participants; clinical failure rate at 16 to 60 days after the start of treatment and failure defined as 3) a lack of full recovery or improvement of participants, or 4) a lack of full recovery of participants; 5) quality of life and activity impairment.

Our classification was based on five domains which we considered as most fundamental in assessing the quality of studies: allocation concealment, blinding, selective outcome reporting, incomplete outcome data and baseline comparability.

The 'Risk of bias' categories were defined as follows.

  1. Low risk of bias (plausible bias unlikely to seriously alter the results) if all five fundamental domains defined above were graded as 'low' risk of bias.

  2. Unclear risk of bias (plausible bias that raises some doubt about results) if one or more of the domains were graded as 'unclear'.

  3. High risk of bias (plausible bias that seriously weakens confidence in the results) if one or more domains were graded as 'high' risk of bias.

We used the GRADEprofiler software version 3.2 (GRADEpro 2008) to provide the overall grading of the quality of evidence for the clinical outcomes in the following comparisons: antibiotics compared to placebo (Summary of findings for the main comparison) and antibiotic versus antibiotic (Summary of findings 2).

Measures of treatment effect

All studies presented the outcome results in dichotomous form.

We calculated risk ratios (RR) for differences between intervention and control groups for whether the treatment failed or not, along with appropriate standard errors and 95% confidence intervals (CI), using RevMan 2012. In calculating drop-out rates due to adverse effects, we used the Peto odds ratio (OR) as the outcome measure. In placebo-controlled studies with multiple antibiotic arms, the data from the antibiotic arms were combined and compared to placebo or control group as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b).

Unit of analysis issues

The unit of analysis was an individual because all RCTs included in this review were simple parallel-group trials in which participants were randomly allocated to two or several groups and a single result for each outcome from each individual was collected and analyzed.

Dealing with missing data

We performed the analyses using an available case data analysis as presented in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011b). This approach for calculating the response rate uses the total number of participants who had data recorded for the particular outcome in question as a denominator.

If the missing outcome data were reported to be over 35% or if there was a significant imbalance in missing data rates between the intervention and control groups, we excluded the study from the meta-analyses. Excluding these studies in the analyses was seen as a quality issue, especially when there are a lack of universal strategies for handling missing data (Unnebrink 2001). In this review, limiting meta-analyses to studies with a missing data rate of less than 35% was a pragmatic approach to make comparisons without compromising the reliability of the overall results. 

In assessing the influence of missing data (less than 35%) on the overall results of the meta-analyses of primary outcomes in studies comparing antibiotic with placebo, we used two ways of imputing data for sensitivity analyses: assuming the outcomes of participants for whom no outcome was recorded as a) failures, or b) non-failures. The analyses comparing antibiotic versus antibiotic were undertaken by using available case data without imputations. When a study provided available case data for the first follow-up but not for the second one, we included the study in the second follow-up meta-analyses if the missing data rate (compared to the first follow-up) was marginal (about 5% at most).

Assessment of heterogeneity

We assessed the significance of any discrepancies in the estimates of the treatment effects from the different trials by means of Cochran's test for heterogeneity and by a measure of the I2 statistic. The I2 statistic describes the percentage of the variability in effect estimates that is due to heterogeneity rather than sampling error. A value greater than 50% may be considered to represent substantial heterogeneity (Higgins 2003).

Assessment of reporting biases

If there had been sufficient numbers of trials (more than 10) in any meta-analysis, publication bias would have been assessed according to the recommendations on testing for funnel plot asymmetry as described in the Cochrane Handbook for Systematic Reviews of Interventions (Sterne 2011). If asymmetry was identified we would have examined possible causes.

Data synthesis

The data consisted of comparisons for antibiotic versus placebo and antibiotic versus antibiotic. The antibiotic versus antibiotic comparisons were macrolide/cephalosporins versus amoxicillin-clavulanate, non-penicillin antibiotics versus beta-lactamase sensitive penicillins and tetracyclines versus mixed classes of antibiotics.

We conducted the meta-analyses in RevMan 2012, using either the fixed-effect or the random-effects model. In meta-analyses including two or three studies, we used the fixed-effect model, and in those including four or more studies, the random-effects model. We pooled the data for the studies in each comparison regardless of whether the studies could represent somewhat different kinds of populations (depending on the definition of the diagnosis of acute maxillary sinusitis; clinical diagnosis only versus radiological/bacteriological confirmation). In meta-analyses of the placebo-controlled studies, we combined data across antibiotic classes. These decisions were made because the main interest was to estimate the average intervention effects. We pooled the Peto ORs of individual studies, estimating the drop-out rate due to adverse effects, using the fixed-effect model.

To incorporate the 'Risk of bias' assessments in analyses, we decided to pool data from the studies in each comparison by restricting the analyses to studies at 'low' risk of bias. We performed sensitivity analyses to see how conclusions were affected by including studies at 'unclear' risk or 'high' risk of bias in the analyses, as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011a). However, if all studies in a comparison were graded as 'unclear' or 'high' risk of bias, we pooled the results of these studies in a meta-analysis. The methodological classifications were taken into consideration when interpreting the results.

Subgroup analysis and investigation of heterogeneity

We had planned to conduct further analyses to identify subgroups of participants that might get more benefit from antibiotic treatment (for example, regarding type of signs or symptoms, their severity or duration). Due to insufficient information from studies, it was not possible to undertake such subgroup analyses.

Sensitivity analysis

To test the robustness of the results, we undertook sensitivity analyses to explore the effect of missing data and overall risk of bias in the studies on the clinical outcomes. We also planned to carry out sensitivity analyses to examine the effect of different diagnostic criteria for acute sinusitis (clinical or radiograph/bacteriological) but the number of trials was not sufficient for such analysis.

Results

Description of studies

Results of the search

In this 2013 update, we retrieved a total of 138 new records, after duplicates were removed from searches in MEDLINE (60 records), EMBASE (137 records), CENTRAL (20 records) and OpenGrey (three records). One of the 138 records was identified by checking the reference list of a new systematic review by Lemiengre 2012. The electronic searches did not identify any potential ongoing trials for this review.

The total number of records considered from all updates was 2347, including 2209 records from the earlier searches and 138 new records from this 2013 update search. Detailed descriptions of the earlier search results are in Appendix 6. Of these 2347 records, we rejected 2052 records as definitely not meeting the inclusion criteria simply on the basis of title or abstract or because they were duplicates (in the earlier searches). From all updates we obtained altogether 295 full-text reports. From these 295 reports, 131 were clearly irrelevant for this review leaving 164 reports for final assessment. The main reasons for irrelevance were: trials without control group, studies with treatments other than antibiotic, studies comparing the same classes of antibiotics and studies regarding children.

We identified 164 reports to be evaluated in detail. We considered 68 reports representing 63 individual studies to be eligible for inclusion in the review. Compared to the previous version of this review, we included three new antibiotic versus placebo studies (Garbutt 2012; Hadley 2010; Meltzer 2005) and one new antibiotic versus antibiotic study (Lari 2010).

The reasons for exclusion of the 89 studies with 96 reports are explained in the Characteristics of excluded studies table.

Included studies

Nine of the 63 included studies compared antibiotic treatment to placebo. The remaining 54 studies compared different classes of antibiotics. The studies of Lindbaek 1996 and Lindbaek 1998 were conducted as the same trial but used different randomised study populations with different inclusion criteria, and they are handled as separate studies in the study descriptions and analyses.

Description of the studies comparing antibiotics versus placebo

Nine trials with 1915 randomised participants evaluated antibiotic treatments compared with non-antibiotic controls for acute maxillary sinusitis (Axelsson 1970; Garbutt 2012; Hadley 2010; Haye 1998; Lindbaek 1996; Lindbaek 1998; Meltzer 2005; Merenstein 2005; van Buchem 1997). Four studies compared amoxicillin versus placebo (Garbutt 2012; Meltzer 2005; Merenstein 2005; van Buchem 1997), two studies compared penicillin V and amoxicillin to placebo (Lindbaek 1996; Lindbaek 1998), one study compared azithromycin to placebo (Haye 1998), and one study compared moxifloxacin to placebo (Hadley 2010). The study by Axelsson 1970 did not have a placebo group but the comparison of penicillin V plus oxymetazoline versus oxymetazoline alone was included in this review.

In four studies diagnosis was based on clinical signs and symptoms lasting at least seven days (Garbutt 2012; Haye 1998; Meltzer 2005; Merenstein 2005). In four studies, diagnosis was confirmed by a radiograph (criteria: secretion in the study of Axelsson 1970; and mucosal thickening more than 5 mm, presence of air-fluid level or total opacification in the van Buchem 1997 study); or by a computer tomography (criteria: presence of air-fluid level or total opacification in the study by Lindbaek 1996; and mucosal thickening of 5 mm or more without fluid level or total opacification in the study by Lindbaek 1998). In one study (Hadley 2010) the diagnosis was confirmed by a radiograph (criteria: presence of air-fluid levels and/or opacification on a radiographic paranasal sinus film) and a sinus puncture for bacterial culture of the sinuses; only culture-positive patients with target pathogen (Streptococcus pneumonia (S. pneumoniae), Haemophilus influenzae (H. influenzae), Moraxella catarrhalis (M. catarrhalis), Streptococcus pyogenes (S. pyogenes) or Staphylococcus aureus (S. aureus)) were included in the analyses. In the study by Haye 1998 radiography was used to exclude patients with empyema.

Nasal decongestants and/or analgesics were allowed but not prescribed in three studies (Hadley 2010; Lindbaek 1996; Lindbaek 1998). Three studies used decongestants, cough medication and/or analgesics as a prescribed supplementary therapy in all groups (Axelsson 1970; Garbutt 2012; van Buchem 1997). Over-the-counter medications without further specification were allowed in one study (Merenstein 2005) and one study did not provide any information on co-interventions (Haye 1998). One study allowed nasal or broncho-inhaled corticosteroids if they had been in stable use for over four weeks prior to enrolment (Hadley 2010). In one study all concomitant medications were prohibited (Meltzer 2005) (Table 1).

Table 1. Collected information from placebo-controlled studies
  1. 1FR = full recovery.
    2FR + I = full recovery + improvement.

StudyComparisonDiagnostic methodDay assessedMissing dataFR1 controlFR1 antibioticsFR + I2 controlFR + I2
antibiotics
Side effects
Axelsson 1970 (Sweden)Penicillin V (400 mg 3 times daily for 10 days) + nasal decongestant (oxymetazoline) versus oxymetazolinePatients with clinically suspected sinusitis were radiologically examined; only participants with secretion included107%31%46%72%83%Penicillin: 8%
Control: 6%
Garbutt 2012 (USA)

Amoxicillin (500 mg 3 times daily for 10 days) versus placebo
Acetaminophen, guaifenesin, dextromethorphan hydrobromide and pseudoephedrine

prescribed to be used as needed

Clinical symptoms > 7 days, purulent nasal discharge, facial pain10;
28

8%;

4%

Information not availableInformation not available

80%;

92%

78%;

85%

Amoxicillin: 16%

Placebo: 14%

Hadley 2010 (USA)Moxifloxacin (400 mg once daily for 5 days) versus placebo
Oral or nasal decongestants or antihistamines, corticosteroid nasal sprays and broncho-inhalants (if in stable use) allowed
Isolation of pathogenic bacteria from the sinus secretion was a prerequisite

6 to 8;

20 to 26

6%;

6%

Information not availableInformation not available

67%;

58%

78%,
74%

Moxifloxacin: 14%

Placebo: 7%

Haye 1998 (Norway)Azithromycin (500 mg once daily for 3 days) versus placeboClinical symptoms lasting 11 to 29 days (symptoms: purulent secretion, maxillary sinus tenderness and/or pain); radiograph taken to exclude patients with more than 6 mm mucosal thickening, complete opacity or air-fluid level10 to 12; 23 to 27

0.6%;

0%

33%;
67%
58%;
79%
89%;
88%
93%;
90%
Azithromycin: 28% Placebo: 18%
Lindbaek 1996 (Norway)Penicillin V (1320 mg 3 times daily for 10 days), or amoxicillin (500 mg 3 times daily for 10 days) versus placebo. Nasal decongestants and analgesics were allowed but not prescribed in all groupsClinical symptoms lasting at least 8 days; computer tomography showing opacity or fluid-level102%11%Penicillin: 31%,
Amoxicillin:
45%
89%Penicillin:
97%,
Amoxicillin:
98%
Penicillin: 59%
Amoxicillin: 56%
Placebo: 36%
Lindbaek 1998 (Norway)Penicillin V (1320 mg 3 times daily for 10 days), or amoxicillin (500 mg 3 times daily for 10 days) versus placebo. Nasal decongestants and analgesics were allowed but not prescribed in all groupsClinical symptoms lasting at least 8 days; computer tomography showing mucosal thickening of 5 mm or more in any sinus without opacity or fluid-level107%43%Penicillin:
30%
Amoxicillin:
41%
86%Penicillin:
91%
Amoxicillin:
86%
Information not available

Meltzer 2005

(14 countries, not specified)

Amoxicillin (500 mg 3 times daily for 10 days) versus placebo
Concomitant medications not allowed

Clinical signs and symptoms lasting at least 7 but less than 29 days

(rhinorrhoea, postnasal drip, nasal congestion/stuffiness, sinus headache, and facial pain/pressure/tenderness on palpation over the paranasal sinuses)

During the 15-day treatment phase1.8%Information not availableInformation not available89%93%

Treatment-emergent

adverse effects
Amoxicillin: 34%
Placebo: 38%

Merenstein 2005 (USA)

Amoxicillin (500 mg twice daily for 10 days) versus placebo

Over-the-counter medications were allowed in both groups

Clinical symptoms > 7 days, purulent nasal discharge, facial pain1414%42%57%Information not availableInformation not availableAmoxicillin: 23% Placebo: 12%
van Buchem 1997 (the Netherlands)

Amoxicillin (750 mg 3 times daily for 7 days) versus placebo

Adjuvant therapy with oxymetazoline steam inhalation (duration not specified). Paracetamol as needed

The mean duration of the symptomatic period before treatment 2.2 weeks; radiograph showing > 5 mm mucosal thickening, opacity or air-fluid level144%52%65%77%83%Amoxicillin: 28% Placebo: 9%

Of the nine studies, three were conducted in Norway (Haye 1998; Lindbaek 1996; Lindbaek 1998), three in the USA (Garbutt 2012; Hadley 2010; Merenstein 2005), one in Sweden (Axelsson 1970) and one in the Netherlands (van Buchem 1997). One study was multinational (Meltzer 2005).

Participants were clearly recruited from community-based general practices in seven of the nine studies. The study by Axelsson 1970 did not describe the treatment setting, and in the multinational study by Meltzer 2005 the nature of participating medical centres was not described. The average age of the participants was approximately 36 years and some 65% were women. Ear, nose and throat (ENT) co-morbidity was assessed in three studies (Garbutt 2012; Meltzer 2005; van Buchem 1997). The study of van Buchem 1997 reported that 12% of participants had an allergic disease. The study of Garbutt 2012 reported allergic rhinitis in 32%, asthma in 11% and a history of sinus disease in 74% of participants in each group; in this study 13% were smokers in the amoxicillin group and 26% in placebo group. The study by Meltzer 2005 reported that 42% of participants had either seasonal or perennial allergic rhinitis in the amoxicillin and placebo groups. In the study of Axelsson 1970 patients with a history of nasal allergy were excluded.

In addition to clinical outcomes, radiological outcomes were reported in three studies (Axelsson 1970; Lindbaek 1996; van Buchem 1997). No studies reported bacteriological outcomes.

Description of the studies comparing antibiotic versus antibiotic

Fifty-four studies comparing different classes of antibiotics were included in the review. Treatment comparisons were: non-penicillin antibiotic versus beta-lactamase sensitive penicillins (n = 8); non-tetracycline versus tetracycline (n = 5) (one study of these was also included in the comparison of non-penicillin antibiotic versus beta-lactamase sensitive penicillin); macrolides versus amoxicillin-clavulanate (n = 11); cephalosporins versus amoxicillin-clavulanate (n = 10); and miscellaneous comparisons (n = 21).

Participants were recruited from otolaryngology speciality settings in 17 studies, primary care settings in 12 studies and mixed settings in three studies. The recruitment or treatment setting was not described or unclear in 22 studies. The average age of the participants was approximately 38 years; three studies did not report the mean age of the participants. In 37 studies, the male to female ratio was mostly about 1:1 or 1:1.5. Fourteen studies assessed ENT co-morbidity.

The diagnosis was based only on clinical signs and symptoms in three studies. In 38 studies diagnosis was confirmed by a radiograph and in 13 studies by radiograph and/or culture. In addition to clinical outcomes, bacteriological outcomes were reported in 28 studies and radiological outcomes in 11 studies. Three studies reported the effects on function or quality of life.

Topical decongestants and/or antihistamines were prescribed in seven studies. They were allowed but not prescribed in 21 studies, prohibited in one study and not described in 25 studies. One trial prescribed nasal corticosteroids as part of the intervention (Pessey 1996).

For more detailed information on the included studies see the Characteristics of included studies table.

Excluded studies

The Characteristics of excluded studies table presents the reasons for exclusion of controlled clinical studies. Only those controlled clinical studies which compared antibiotic treatment with placebo or antibiotics of different classes for acute maxillary sinusitis are included in the table. Studies without a control group were excluded.

Of the 89 excluded studies, 13 studies included comparison of antibiotic versus placebo and 76 studies compared antibiotic versus antibiotic. The reasons for exclusion were varied and in many studies there were several reasons for exclusion. In 19 out of the 89 studies there was no mention of random allocation or the study design was clearly not randomised; two of these studies were placebo-controlled. In the other 11 randomised, placebo-controlled studies, the diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review.

Risk of bias in included studies

We contacted the authors of the included placebo-controlled studies to obtain additional information to assess the methodological quality issues if the information in the report was insufficient to make final decisions. We requested additional information from seven studies (Garbutt 2012; Hadley 2010; Haye 1998; Lindbaek 1996; Lindbaek 1998; Meltzer 2005; Merenstein 2005). Authors of four of those seven studies could deliver additional information on quality issues (Garbutt 2012; Lindbaek 1996; Lindbaek 1998; Merenstein 2005) and information was not available for this update from three studies (Hadley 2010; Haye 1998; Meltzer 2005). We did not contact the authors of the Axelsson 1970 study because the report was published 40 years ago.

Quality assessments for each individual study are presented in 'Risk of bias' tables included in the Characteristics of included studies, and the results are presented graphically by study (Figure 2) and by domain over all studies (Figure 1).

Allocation

Antibiotics versus placebo

Random sequence generation was adequate, indicating 'low' risk of bias in seven of the nine placebo-controlled studies (Garbutt 2012; Haye 1998; Lindbaek 1996; Lindbaek 1998; Meltzer 2005; Merenstein 2005; van Buchem 1997). In five studies the random sequence generation was made by computer (Garbutt 2012; Haye 1998; Meltzer 2005; Merenstein 2005; van Buchem 1997) and in the other two studies by dice (Lindbaek 1996; Lindbaek 1998). Two studies did not report the randomisation method (Axelsson 1970; Hadley 2010) and the judgement was 'unclear' risk of bias.

Allocation concealment was graded 'low' risk of bias in the same seven studies as those assessed to have adequate random sequence generation. Allocation concealment was performed centrally in the study of van Buchem 1997 (hospital pharmacy-controlled randomisation), by an investigational pharmacist in the study of Garbutt 2012 and by a statistician in the studies of Lindbaek 1996, Lindbaek 1998 and Merenstein 2005. Although the studies of Haye 1998 and Meltzer 2005 did not give exact information on how the randomisation result was concealed, the method of random sequence generation (computer-randomised code) and information on double-dummy design gave the impression that concealment had been fulfilled (additional information not received from the trial authors). Two studies did not report the allocation concealment method (Axelsson 1970; Hadley 2010) and the judgement was 'unclear' risk of bias.

Antibiotics versus antibiotics

Random sequence generation was adequate, indicating 'low' risk of bias in 19 of the 54 studies (35%) comparing different classes of antibiotics. In most of these studies the sequence generation was computer-based. Thirty-five studies did not report the random sequence generation method and the judgement was 'unclear' risk. However, although random sequence generation was reported in 19 studies, the allocation concealment was adequately described in only 13 out of 54 studies (24%) to allow this domain be graded 'low' risk of bias. In these studies, the allocation concealment was mostly executed by a third party (for example, in pharmaceutical company or by a statistician not involved with the personnel in the studies). Forty-one studies did not report the random allocation concealment method at all or the description was not clear enough to make the assessment and the judgement was 'unclear' risk of bias.

Blinding

Antibiotics versus placebo

Seven of the nine placebo-controlled studies documented adequate blinding of participants and investigators (Garbutt 2012; Haye 1998; Lindbaek 1996; Lindbaek 1998; Meltzer 2005; Merenstein 2005; van Buchem 1997). The study medications were delivered by a third party and the authors did not know which medicine each participant had received. The participants were not aware which treatment they received, and most of the studies stated that treatment and placebo products were indistinguishable from each other. Two studies were justified as 'unclear' risk of bias: the study of Axelsson 1970 did not report anything about the blinding, and Hadley 2010 only stated the study to be double-blinded with matching placebos but did not give any other information on the procedures.

Antibiotics versus antibiotics

We documented and assessed blinding of participants and investigators as adequate in 19 out of the 54 studies (35%) comparing different classes of antibiotics. In most cases study medications were delivered by a third party and matched placebo tablets were used to ensure blinding of participants. Twenty of the 54 studies were open-labelled (37%) and 10 studies (19%) single-blinded and we graded these as having 'high' risk of bias. In five studies the description was not clear enough to assess this domain and the judgement was 'unclear' risk of bias.

Incomplete outcome data

Antibiotics versus placebo

We assessed eight placebo-controlled studies to have 'low' risk of bias in this domain (Axelsson 1970; Hadley 2010; Haye 1998; Lindbaek 1996; Lindbaek 1998; Meltzer 2005; Merenstein 2005; van Buchem 1997). The reported total missing data rates were between 0.6% and 14% at 7 to 15 days follow-up. All of those eight studies except one (Lindbaek 1998) adequately reported missing data rates by study group. In the study by Lindbaek 1998 the total missing data rate was 5/68 (7.4%). However, we obtained additional information from the trial author that there were drop-outs in each of the three groups. Thus we assessed the study of Lindbaek 1998 as having a 'low' risk of bias in this domain.

The ninth placebo-controlled study by Garbutt 2012 was graded as 'unclear' risk of bias at day 10. The study reported different drop-out rates: in the amoxicillin group 4/85 (4.7%) and in the placebo group 10/81 (12.3%) on day 10. Although there probably was not any systematic difference in reasons for drop-outs between the groups (the authors gave additional information that participants simply missed the interviews), the difference in proportions of missing outcomes (7.6%) compared to the difference in anticipated failure risks (10% to 20%) may be big enough to cause bias.

Two studies reported results at 16 to 60 days follow-up and had total missing data rates of 0% (Haye 1998) and 6% (Hadley 2010), indicating 'low' risk of bias.

Antibiotics versus antibiotics

In this domain we assessed 31 out of 54 studies (57%) comparing antibiotics with antibiotics to have 'low' risk of bias, 12 studies (22%) to have 'high' risk of bias, and in 11 studies (20%) the judgement was 'unclear' risk of bias. The way of reporting the results made it difficult to assess this domain in many of the studies. Several of the 12 studies graded as having 'high' risk of bias reported responses only for per-protocol populations and the proportion of missing data was high. Even when studies reported clinical responses for both intention-to-treat (ITT) and per-protocol populations, often only the clinical cure rates were reported while failure rates were not. The indeterminate/missing data were then usually categorised as failures in ITT analyses and it remained unclear how many of the responses were true failures in each group. Three studies had to be downgraded from 'unclear' to 'high' risk of bias because of this particular problem. In the 11 studies with the judgement 'unclear' risk, the description was not clear enough to assess this domain (especially regarding reasons for missing data across groups).

We excluded three studies from the meta-analyses because the proportion of missing data was too high (38% in Marple 2010, 44% in Luterman 2003 and 52% in Steurer 2000).

Selective reporting

Antibiotics versus placebo

All nine placebo-controlled studies reported all prespecified outcomes adequately and were graded as 'low' risk of bias in this domain (Axelsson 1970; Garbutt 2012; Hadley 2010; Haye 1998; Lindbaek 1996; Lindbaek 1998; Meltzer 2005; Merenstein 2005; van Buchem 1997).

Antibiotics versus antibiotics

Prespecified outcomes were adequately reported in 47 out of 54 studies (87%) comparing different classes of antibiotics. In two studies the reporting was deficient and they were graded as having 'high' risk of bias. In five studies the description was too vague to make an assessment and the judgement was 'unclear' risk of bias.

Other potential sources of bias

Antibiotics versus placebo
Baseline comparability

Demographic characteristics and disease severity ratings were described and assessed to be balanced across the groups in eight placebo-controlled studies. In the study of Hadley 2010 this domain was graded as 'unclear' risk of bias because in the placebo group there were 9% more bilateral sinus infections than in the moxifloxacin group (38% versus 29%) and no by-group description was provided on allowed concomitant medications (especially stable-use nasal corticosteroids).

Co-interventions

Seven of the nine placebo-controlled studies reported co-interventions during the trial (Axelsson 1970; Garbutt 2012; Hadley 2010; Lindbaek 1996; Lindbaek 1998; Merenstein 2005; van Buchem 1997). In these studies co-interventions were mainly symptom-relieving over-the-counter medications which we judged not to cause bias. The study of Meltzer 2005 prohibited all concomitant medications, and the study of Haye 1998 did not provide any information on co-interventions.

We graded seven out of the nine studies as having a 'low' risk of bias in this domain (Axelsson 1970; Garbutt 2012; Lindbaek 1996; Lindbaek 1998; Meltzer 2005; Merenstein 2005; van Buchem 1997. The study of Hadley 2010 did not give a detailed description of the use of allowed co-interventions (especially nasal corticosteroids which also may have a curative effect) and was judged as having 'unclear' risk of bias, as well as Haye 1998 which had no information on co-interventions.

Funding

Four of the nine placebo-controlled studies were funded by governmental or academic sources, or independent research foundations (Garbutt 2012; Lindbaek 1996; Lindbaek 1998; Merenstein 2005) and were judged as having a 'low' risk of bias. Two studies did not state the funding source (Axelsson 1970; van Buchem 1997) and in one study (Haye 1998) one of the trial authors was affiliated with a pharmaceutical company. In these three studies the funding domain was graded as 'unclear' risk. Two studies were supported by a pharmaceutical company and were classified as having a 'high' risk of bias in this domain (Hadley 2010; Meltzer 2005).

Antibiotics versus antibiotics
Baseline comparability

Demographic characteristics and disease severity ratings were described and assessed to be balanced across the groups in 34 out of 54 studies (63%) comparing different classes of antibiotics. In two studies we assessed the groups to be imbalanced and they were graded as having 'high' risk of bias. In 18 studies the information was insufficient to assess this domain and the judgement was 'unclear' risk.

Co-interventions

In 30 out of the 54 studies (56%) we graded this domain as having 'low' risk of bias. In most cases the co-interventions were symptom-relieving over-the-counter medications and they were seen not to cause bias. In 24 studies the judgement was 'unclear' risk because no information was provided.

Funding

Thirty out of the 54 studies (56%) were supported by a pharmaceutical company and they were graded as having 'high' risk of bias. In 24 studies one or more of the authors were affiliated to a pharmaceutical company or the funding source was not stated, and this domain was graded as 'unclear' risk.

Effects of interventions

See: Summary of findings for the main comparison Antibiotics compared to placebo for uncomplicated acute maxillary sinusitis; Summary of findings 2 Antibiotics compared to other antibiotics for uncomplicated acute maxillary sinusitis

Clinical outcomes were reported in all trials and evaluated quantitatively. We evaluated radiological and bacteriological outcomes only for placebo-controlled studies. However, none of the placebo-controlled studies reported bacteriological outcomes. Radiographic outcomes are described qualitatively due to the small number of trials reporting these data.

Primary outcomes

Antibiotics versus placebo
1. Clinical failure defined as a lack of full recovery or improvement at 7 to 15 days follow-up

Eight studies provided data for comparison of antibiotic treatment to placebo when clinical failure was defined as a lack of full recovery or improvement at 7 to 15 days follow-up (Axelsson 1970; Garbutt 2012; Hadley 2010; Haye 1998; Lindbaek 1996; Lindbaek 1998; Meltzer 2005; van Buchem 1997). In two studies unclearness of data (Garbutt 2012; Hadley 2010) and in one study the difference in the design (Hadley 2010) were seen to cause too much uncertainty, which prevented including them in the meta-analysis. Five of the remaining six studies were assessed as 'low' risk of bias and were included in the main analysis (Haye 1998; Lindbaek 1996; Lindbaek 1998; Meltzer 2005; van Buchem 1997) and one study at 'unclear' risk of bias was used in the sensitivity analysis (Axelsson 1970).

In each of the five individual studies of the main analysis the effect estimate of risk ratio (RR) indicated a small benefit for antibiotics but the confidence interval in all studies crossed the no difference line 1.0. When pooling the estimates of effects of these studies in the meta-analysis, antibiotics were found to be statistically significantly beneficial with a RR of 0.66 (95% confidence interval (CI) 0.47 to 0.94), based on a random-effects model using available case data (1058 evaluated participants; Analysis 1.1). The failure rates were 8.7% and 13.6% in antibiotic and placebo groups, respectively. This is the main result presented in the Summary of findings for the main comparison. There was no statistically significant heterogeneity between the studies in the analysis (I2 statistic = 0%). When double-checking the result using a fixed-effect model the result was similar (RR 0.65, 95% CI 0.46 to 0.92). Also, when using odds ratio (OR) as the effect measure the inference of the result was similar (OR 0.62, 95% CI 0.41 to 0.92) and there was no heterogeneity (data not shown).

When including Axelsson 1970 (at 'unclear' risk of bias) in the main analysis (with the five other studies) the results did not change (RR 0.66, 95% CI 0.47 to 0.91) (data not shown). The Axelsson 1970 study was graded as 'unclear' risk of bias because no information on randomisation, allocation concealment and blinding was given. Lack of information may be related to the way studies were reported 40 years ago but it was not possible to get additional information.

The amount of missing data in each of the five studies was balanced between the antibiotic and placebo groups, ranging from 0% to 5% by group (actual numbers zero to five per treatment group). In sensitivity analyses assuming missing outcomes as failures or full recoveries/improvements did not significantly affect the result of the meta-analysis: RR 0.68, 95% CI 0.50 to 0.93 and RR 0.67, 95% CI 0.48 to 0.95, respectively (data not shown).

The meta-analyses, including sensitivity analyses, slightly favoured antibiotics. However, in the individual six studies the difference between groups in the full recovery or improvement rates was small, being on average 90% (range 83% to 98%) in antibiotic groups and 84% (range 72% to 89%) in placebo groups.

The results of the studies by Hadley 2010 and Garbutt 2012

The results by Hadley 2010 (at 'unclear' risk of bias) comparing moxifloxacin 400 mg once daily to placebo for five days and (unlike the other placebo-controlled trials) recruiting only patients with isolation of pathogenic bacteria from the sinus secretion, were in line with the results of the individual studies included in the main analysis. Hadley 2010 reported that on day six to eight, clinical cure rates were numerically greater in the moxifloxacin arm (57/73, 78.1%) than the placebo arm (30/45; 66.7%) but this difference was not statistically significant (P = 0.189). Hadley 2010 measured clinical response at days six to eight after the start of therapy while other studies included in the meta-analysis did so at days 10 to 14.

The study by Garbutt 2012 (graded as having an 'unclear' risk of bias) comparing amoxicillin 500 mg three times daily to placebo reported there was no statistically significant difference in symptom improvement at day 10 (78% in amoxicillin group versus 80% in placebo group; P = 0.71). The study was not considered in the main analysis because it remained unclear whether the patients who received rescue therapy were classified as non-improved or not.

2. Clinical failure defined as a lack of full recovery or improvement at 16 to 60 days follow-up

Two studies provided data for an antibiotic versus placebo comparison at 16 to 60 days follow-up (Hadley 2010; Haye 1998). Neither of these studies found a significant difference between antibiotic treatment and placebo at this follow-up, Hadley 2010 with a RR value of 0.62 (95% CI 0.37 to 1.03) and Haye 1998 with a RR value of 0.85 (95% CI 0.36 to 1.98) (Analysis 1.2). The study of Haye 1998 was assessed to be at 'low' risk of bias at this follow-up, and the study of Hadley 2010 at 'unclear' risk of bias, and thus data from these two studies were not pooled in the meta-analysis. Further, the populations of these studies were very different; Haye 1998 excluded participants with radiographic findings of more than a mild sinusitis and Hadley 2010 included only participants with bacteriologically confirmed sinusitis.

At this follow-up, the drop-out rate was 6% in the study of Hadley 2010 and 0% in the study of Haye 1998.

Secondary outcomes

1. Clinical failure defined as a lack of full recovery at 7 to 15 days follow-up

Five studies were included in the meta-analysis (Haye 1998; Lindbaek 1996; Lindbaek 1998; Merenstein 2005; van Buchem 1997) and one was used in the sensitivity analysis (Axelsson 1970). These studies found a significant benefit for antibiotics with a RR of 0.73 (95% CI 0.63 to 0.85) (Analysis 1.3). The failure rates were 47% and 61% in the antibiotic and placebo groups, respectively. There was no statistically significant heterogeneity between studies. When including the study by Axelsson 1970 at 'unclear' risk of bias in the analysis, the results did not change: RR of 0.73 (95% CI 0.63 to 0.85) (data now shown).

2. Clinical failure defined as a lack of full recovery at 16 to 60 days follow-up

One study at 'low' risk of bias provided data for an antibiotic versus placebo comparison at 16 to 60 days follow-up (Haye 1998). The difference failed to reach statistical significance with a RR of 0.63 (95% CI 0.38 to 1.05) (Analysis 1.4). The failure rates were 20.7% and 33% in the antibiotic and placebo groups, respectively.

3. Bacteriological failure

None of the studies reported this outcome.

4. Radiographic failure

Radiographic outcomes were reported in three studies (Axelsson 1970; Lindbaek 1996; van Buchem 1997) and the results were consistent with the clinical outcomes in the individual studies. In the study by Axelsson 1970, the nasal decongestant group was less improved than the other groups, measured by the mean number of radiological severity points per sinus. In the study by Lindbaek 1996, computer tomography scores improved significantly more for individuals treated with antibiotics. In the study by van Buchem 1997, more participants treated with antibiotics showed radiographic resolution (74% versus 60%).

5. Relapse rate after 60 days

Only one study reported long-term relapse rates (van Buchem 1997). During a one-year follow-up, relapse and recurrence rates were not significantly different between antibiotic (21%) and placebo (17%) groups, with a RR of 1.25 (95% CI 0.72 to 2.19) (Analysis 1.5).

6. Drop-outs due to adverse effects

Drop-outs due to adverse effects were rare in both groups: 15 out of 1013 (1.5%) in antibiotic groups and 8 out of 805 (1%) in control groups (Analysis 1.6). Three out of the nine placebo-controlled studies reported no drop-outs due to side effects in the antibiotic or control groups.

7. Quality of life

Two studies (Garbutt 2012; Hadley 2010) documented patient-reported symptom improvements evaluated by the SNOT-16 questionnaire (Sinonasal Outcome Test-16, a validated disease-specific quality of life instrument, where lower scores indicate less severe symptoms (scale from 0 to 3)). The study of Garbutt 2012 reported the results as mean scores (baseline value 1.71 in amoxicillin and 1.70 in placebo group). Similar quality of life in both groups was documented at day three and 10 but a difference at day seven favoured amoxicillin (P = 0.02). The mean change in SNOT-16 scores was at three days: amoxicillin group 0.59 (95% CI 0.47 to 0.71) versus placebo group 0.54 (95% CI 0.41 to 0.67); at seven days 1.06 (95% CI 0.93 to 1.20) versus 0.86 (95% CI 0.71 to 1.02); and at 10 days 1.23 (95% CI 1.08 to 1.37) versus 1.20 (95% CI 1.07 to 1.32), respectively.

The study of Hadley 2010 reported the results as SNOT-16 total scores (scale from 0 to 48). Moxifloxacin-treated patients were stated to have experienced a significantly greater mean reduction in total score compared to placebo-treated patients at day six to eight (-17.54 versus -12.83 (P = 0.032) from baseline values of about 28 in both groups).

8. Activity impairment and ability to work

Only one study gave information on patients' ability to work (Garbutt 2012). The mean period missed from work was stated to be the same, 0.55 days in both groups (amoxicillin and placebo).

In addition, two studies provided information on patients' activity impairment (Garbutt 2012; Hadley 2010). Hadley 2010 stated that from day three moxifloxacin-treated patients experienced a greater improvement in activity impairment than placebo-treated patients evaluated by a five-item questionnaire (scale from 0 (no impairment) to 20 (maximum impairment)). At day six to eight, the mean changes in the scores for activity impairment were: -6.1 (standard deviation (SD) ± 5.9) in the moxifloxacin group and -3.7 (± 5.8) in the placebo group.

The study by Garbutt 2012 reported a non-significant difference between the groups in the period of being unable to do usual non-work activities (mean 1.15 days (95% CI 0.76 to 1.54) in the amoxicillin group and 1.67 days (95% CI 1.08 to 2.26) in the placebo group).

Antibiotics versus antibiotics

In comparisons of antibiotic versus antibiotic, we only analyzed data for the primary outcomes measures: clinical failure rate at seven to 15 days follow-up and at 16 to 60 days follow-up (clinical failure defined as a lack of full recovery or improvement) and for two secondary outcome measures: relapse rates and drop-outs due to adverse effects.

Cephalosporins and macrolides versus amoxicillin with clavulanate

Twenty-one studies compared a cephalosporin or a macrolide (cephalosporin, n = 10; macrolide, n = 11) to amoxicillin-clavulanate. Seventeen studies provided data for meta-analyses; nine studies compared cephalosporin (other than first generation) to amoxicillin-clavulanate and eight studies compared macrolides (azithromycin, clarithromycin, roxithromycin or telithromycin) to amoxicillin-clavulanate. We assessed all the studies to be at 'unclear' or 'high' risk of bias regardless of the follow-up time.

Six studies, involving 1887 participants, provided data for comparison of cephalosporins to amoxicillin-clavulanate at seven to 15 days follow-up when clinical failure was defined as a lack of full recovery or improvement. These studies found a statistically significant difference for amoxicillin-clavulanate with a RR of 1.37 (95% CI 1.04 to 1.80) (Analysis 2.1). The average failure rate was 12% in the cephalosporin group and 8% in the amoxicillin-clavulanate group. The missing data rates ranged from 2% to 21%, and the missing data rates were approximately equal across the treatment groups in the studies. None of the studies were graded as having 'low' risk of bias and the two studies having highest weight (70% altogether) in the meta-analysis were graded as having 'high' risk of bias. Confidence in this finding is limited by the low methodological quality.

Seven studies provided data for comparison of cephalosporins to amoxicillin-clavulanate at 16 to 60 days follow-up. These studies did not find any significant difference between the antibiotics, with a RR of 1.08 (95% CI 0.85 to 1.37) (Analysis 2.2).

Seven studies, involving 1807 participants, provided data for comparison of macrolides to amoxicillin-clavulanate at seven to 15 days follow-up. These studies failed to reach significant difference between the antibiotics with a risk ratio value of 0.83 (95% CI 0.62 to 1.13) (Analysis 2.4). The missing data rates ranged from 3.4% to 15%.

Four studies provided data for comparison of macrolides to amoxicillin-clavulanate at 16 to 60 days follow-up. These studies did not find any significant difference between the antibiotics, with a RR of 0.85 (95% CI 0.57 to 1.27) (Analysis 2.5).

None of the studies comparing cephalosporins or macrolides to amoxicillin-clavulanate reported long-term relapse rates after 60 days.

Drop-outs due to adverse effects occurred significantly less often in the macrolide group (2.1%) compared to the amoxicillin-clavulanate group (4.8%), with a Peto OR of 0.47 (95% CI 0.30 to 0.72) (the macrolide group ranged from 0% to 3.4%; the amoxicillin-clavulanate group ranged from 0% to 10.3%) (Analysis 2.6). Participants dropped out significantly more often also in the amoxicillin-clavulanate group (4.4%) compared to the cephalosporin group (1.3%) with a Peto OR of 0.32 (95% CI 0.21 to 0.49) (in the cephalosporin group range 0% to 2.1%; in the amoxicillin-clavulanate group 0% to 11%) (Analysis 2.3).

Non-penicillins versus beta-lactamase sensitive penicillins

Eight studies compared a non-penicillin antibiotic (cephalosporins, n = 3; macrolides, n = 3; minocycline, n = 1; folate inhibitor, n = 1) to a penicillin-class antibiotic (amoxicillin, n = 7; penicillin V, n = 1), where n is the number of studies. Seven of the eight studies were classified as having 'unclear' or 'high' risk of bias, and one as 'low' risk of bias (Haye 1996). Because there was only one study at 'low' risk of bias, we combined the results of all studies in the meta-analyses.

Seven of these eight studies (involving 1083 participants) provided data for clinical outcome at 7 to 15 days follow-up and we combined the data from these studies regardless of the antibiotic used. The studies failed to find a significant difference between antibiotics, with a RR of 0.70 (95% CI 0.47 to 1.06) (Analysis 3.1). When excluding the study of Haye 1996(at 'low' risk of bias) from the analyses, the result of the meta-analysis did not change the conclusions (RR of 0.75, 95% CI 0.48 to 1.17). The missing data rates of these seven studies ranged from 0% to 33%, and the missing data rates were approximately equal across the treatment groups in the studies.

One study (Haye 1996) provided data at 16 to 60 days follow-up and did not find a significant difference between the treatment groups with a risk ratio value of 0.67 (95% CI 0.37 to 1.20) (Analysis 3.2). This study was assessed to be at 'low' risk of bias also at this second follow-up.

None of the studies reported long-term relapse rates after 60 days.

Drop-outs due to adverse effects were infrequent, occurring in 1.3% and 2.3% of the non-penicillin and penicillin-treated groups, respectively (Analysis 3.3).

Tetracycline antibiotics versus mixed classes of antibiotics

Five studies, involving 807 participants, compared a tetracycline (doxycycline, n = 3; tetracycline, n = 1; minocycline, n = 1) to a heterogenous mix of antibiotics (folate inhibitor, n = 2; cephalosporin, n = 1; macrolide, n = 1; amoxicillin, n = 1). All these studies provided data for clinical failure defined as a lack of full recovery or improvement at 7 to 15 days follow-up, and we combined the data from these studies, despite the antibiotic used. The studies failed to find a significant difference between antibiotics, with a RR of 1.09 (95% CI 0.70 to 1.71) (Analysis 4.1). The missing data rates of these five studies ranged from 0% to 20%, and the missing data rates were approximately equal across the treatment groups in the studies. All the studies in this comparison were graded as 'unclear' or 'high' risk of bias.

None of the studies reported long-term relapse rates after 60 days.

Drop-outs due to adverse effects were infrequent, occurring in 2.6% and 3.5% of the tetracycline and mixed classes of antibiotics groups, respectively (Analysis 4.2).

Miscellaneous comparisons

Twenty-one studies made the following comparisons: macrolide to fluoroquinolone (n = 5), fluoroquinolone to cephalosporin (n = 4), macrolide to cephalosporin (n = 3), fluoroquinolone to amoxicillin-clavulanate (n = 5), streptogramin to cephalosporin (n = 2) and faropenem to cephalosporin (n = 2). None of these studies reported a statistically significant difference in the clinical outcomes.

Comparison of the effect of short versus long courses of antibiotics

We did not perform this comparison because there were no placebo-controlled studies available to undertake it.

Discussion

Summary of main results

Antibiotics versus placebo

In this review, the risk of treatment failure (defined as lack of full recovery or improvement) was statistically slightly smaller in antibiotic groups than in placebo groups, at 7 to 15 days in participants with acute sinusitis diagnosed either clinically or by radiography (main analysis, five studies; Summary of findings for the main comparison). However, the clinical benefit was small. The difference between groups in the full recovery or improvement rates was 10% at most in all of the eight studies available for this comparison. The average cure or improvement rate in the antibiotic groups was 87% (range 78% to 98%) and in the placebo groups 81% (range 67% to 89%) (Table 1).

On the other hand, stronger statistical significance favouring antibiotics in terms of clinical failure rates, defined as lack of full recovery (at 7 to 15 days; five studies; Summary of findings for the main comparison), might indicate a faster cure rate with antibiotics than without. The average full recovery rate in the antibiotic groups was about 50% (range 30% to 65%) and in the control groups 35% (range 11% to 52%) (all six available studies for this comparison, Table 1).

At day 16 to 60 follow-up there was no significant difference between the antibiotic and placebo groups in relieving signs and symptoms regardless of the definition of failure (full recovery or improvement (Hadley 2010; Haye 1998); full recovery (Haye 1998)) (Summary of findings for the main comparison). Although not statistically significant, Haye 1998 reported higher full recovery rates in the antibiotic group (69 out of 87 patients (79%)) than in the placebo group (55 out of 82 of patients (67%)).

None of the nine placebo-controlled studies reported severe complications.

Even longer follow-up times would be needed to assess the efficacy of antibiotic treatment in the longer term and to assess potential adverse effects. There is some evidence from other respiratory tract infections that the use of antibiotics increases relapse rates and may end up increasing antibiotic consumption (Arason 2005; Joki-Erkkilä 2000). Therefore, in optimal circumstances, the sinusitis trials should monitor signs, infection rates and drug use over one year. Only one study in this review reported long-term relapse rates (van Buchem 1997) during a one-year follow-up. The relapse and recurrence rates were not significantly different between groups.

Patient perspective on recovery

Quality of life and activity impairment were assessed in two studies (Garbutt 2012; Hadley 2010) (Summary of findings for the main comparison). Both of these studies reported somewhat better quality of life in the antibiotic than in the placebo group at days six to eight. However, the Garbutt 2012 trial documented that quality of life was similar in the antibiotic and placebo groups at day three and day 10 follow-ups. Garbutt 2012 also reported equal mean periods missed from work in the antibiotic and placebo groups (0.55 days) and a non-significant difference in the period where participants were unable to do usual non-work activities (mean 1.15 days in amoxicillin and 1.67 days in the placebo group; P = 0.14).

Hadley 2010, on the other hand, reported that from day three to days six to eight, patients in the antibiotic group experienced a greater improvement in activity impairment than patients in the placebo group. The differences in change in mean activity impairment scores were 1 at day three and 2.4 at days six to eight in favour of the antibiotic group. The significance of the results remains unclear because baseline scores (in the scale 0 to 20) were not reported, and even the five items of the questionnaire were not specified.

In both studies it remained unclear whether data for participants who received rescue therapy were included in these results.

In this review the primary outcome was failure at 7 to 15 days after the start of the treatment because recovery from acute maxillary sinusitis takes generally more than one week (Gwaltney 2005). However, from the patient's perspective, fast symptom relief is important and data are also needed on short-term improvement. Six studies (Axelsson 1970; Garbutt 2012; Hadley 2010; Haye 1998; Lindbaek 1996; Merenstein 2005) also reported cure or improvement rates prior to one-week follow-up. Three of the six studies reported higher improvement rates for antibiotic than placebo (Hadley 2010: improvement rate 85% versus 73% at day three; Lindbaek 1996: improvement rate 60% versus 36% at day three; Merenstein 2005: the average number of days to entire improvement was consistently two to two and a half days shorter in the antibiotic group among those participants who had entirely improved by day 14). Two of these studies represented patients with a more severe form of disease (Hadley 2010; Lindbaek 1996). The other three studies reported only minor differences between the groups at day three to five (Axelsson 1970: full recovery or improvement rate was 69% in the antibiotic group versus 65% in the placebo group; Garbutt 2012: 37% versus 34%; Haye 1998: 80% versus 79%, respectively). The data do not make it possible to draw strong conclusions on the short-term improvement of patients' symptoms.

Although we found slight statistical significance and potentially faster resolution of symptoms favouring antibiotics, the clinical significance of the result is questionable because of the considerable improvement rate in the placebo group. Also, the possible benefit needs to be weighed against the potential for adverse effects at both the individual and population levels. The results were based on studies of penicillin, amoxicillin, azithromycin and moxifloxacin performed in middle and northern Europe in five studies (Axelsson 1970; Haye 1998; Lindbaek 1996; Lindbaek 1998; van Buchem 1997), in the USA in three studies (Garbutt 2012; Hadley 2010; Merenstein 2005) and one study was multinational (Meltzer 2005).

Key information about placebo-controlled studies is collated in Table 1.

Different classes of antibiotics

In the study designs comparing different classes of antibiotics, the antibiotics had a similar efficacy to each other (Summary of findings 2). However, at 7 to 15 days follow-up, the risk of clinical failure was statistically significantly lower for amoxicillin-clavulanate than for cephalosporins, but the significance of the difference disappeared at longer follow-up.

Amoxicillin and penicillin have the advantage of low cost but are often recommended at a three times a day dose, a dosing schedule associated with decreased compliance compared to once or twice daily regimens. A more serious concern is the rising prevalence of beta-lactamase-producing organisms and penicillin and macrolide-resistant pneumococci. In eight studies comparing non-penicillin antibiotics to a beta-lactamase sensitive penicillin, clinical outcomes were virtually identical. These studies were conducted mostly in the 1990s, four of them in the USA, three in Nordic countries and one in Switzerland.

Among the newer, extended-spectrum antibiotics, the long-term efficacy (at 16 to 60 days follow-up) was similar but amoxicillin-clavulanate had significantly more adverse effects than cephalosporins and macrolides. At 7 to 15 days follow-up, however, the risk of clinical failure was statistically significantly lower for amoxicillin-clavulanate than for cephalosporins, but the significance of the difference disappeared at longer follow-up. The studies were conducted in the 1990s and 2000s.

None of the antibiotic preparations were superior to each other. Given the similar efficacy, the differences in the adverse effects, costs and risk of promoting bacterial resistance should be considered when choosing antibiotic treatment for acute sinusitis.

Adverse effects

In general, adverse effects were more common in the antibiotic than in the placebo groups (median of difference between groups 10.5% in seven of the nine studies, range 2% to 23%) (Table 1). The adverse effects rates of penicillin and amoxicillin differed among the placebo-controlled studies. In the study by Axelsson 1970, the rate for penicillin was 8% (in the control group 6%) and in the study by Lindbaek 1996 59% (in the placebo group 36%). The mean adverse effect rate based on five studies for amoxicillin was 31% (range 16% to 56%), while the corresponding rate for placebo was 22% (9% to 38%) (Garbutt 2012; Lindbaek 1996; Meltzer 2005; Merenstein 2005; van Buchem 1997). The study by Haye 1998 reported 28% adverse effects for azithromycin and 18% for placebo; and the study by Hadley 2010 reported 14% for moxifloxacin and 7% for placebo. The most commonly reported adverse effects in antibiotic groups were: gastrointestinal problems (for example, diarrhea, abdominal pain, vomiting) and skin rash. However, drop-outs due to adverse effects were rare (1.5%) for penicillin, amoxicillin, azithromycin and moxifloxacin in the placebo-controlled studies.

In the antibiotic versus antibiotic comparisons, the drop-out rates due to adverse effects were significantly different only in the comparisons of amoxicillin-clavulanate versus macrolides (Analysis 2.6) and cephalosporins (Analysis 2.3). There were 17 studies comparing amoxicillin-clavulanate to macrolides or cephalosporins. The drop-out rate due to adverse effects varied from 0% to 11% in the amoxicillin-clavulanate groups (median 3.4% in 17 studies), and from 0% to 3.4% in macrolide and cephalosporin groups (medians 2% in eight studies and 1.4% in nine studies, respectively).

One problem in the sinusitis trials is that standardised information on side effects is available in fewer than half of the cases (Ioannidis 2002). It may be that the true rate of side effects equals or even exceeds the marginal benefits of the treatments. On the other hand, both severe adverse effects and severe complications (for example, anaphylaxis and meningitis) are such rare events that they are seldom manifested in randomised controlled trials with limited numbers of patients.

Overall completeness and applicability of evidence

Setting and diagnostic methods

This review is restricted to adults. This is because sinusitis in children differs from that in adults, for example regarding anatomy, diagnosis, epidemiology and microbiology (Hsin 2010). Newborns have maxillary sinuses the size of a bean. The sinuses start to widen at the age of three or four, the development takes a long time and only by the age of 15 to 18 is the anatomy of the sinuses fully developed (Lawson 2008). It is difficult to find figures, but a common expert opinion is that although purulent nasal discharge is a common symptom in children, sinusitis appears to be more rare than among adults and often chronic, concentrating in special groups like asthmatic and immunocompromised children and those with cystic fibrosis (Wu 2009). On the other hand, complications of acute sinusitis appear to be about three times more common among children than adults (Hansen 2012) and these develop during the first days of acute respiratory infection (Kristo 2009). Because sinusitis in children seems to be quite different in many aspects compared to that in adults, we chose to limit our review to studies of adults.

As the role of radiographs in diagnosing acute sinusitis is controversial, we decided also to include trials that included participants based on clinical criteria only. The diagnosis of acute maxillary sinusitis was based solely on clinical diagnosis in four studies (Garbutt 2012; Haye 1998; Meltzer 2005; Merenstein 2005), confirmed by radiograph in three studies (Axelsson 1970; Hadley 2010; van Buchem 1997) and confirmed by computer tomography (CT) scan in two studies (Lindbaek 1996; Lindbaek 1998). Haye 1998 used X-rays to exclude patients who had evidence of empyema. The studies with radiographic confirmation of diagnosis accepted participants with air-fluid level or opacification (Axelsson 1970; Hadley 2010) and van Buchem 1997 also those with more than 5 mm mucosal thickening. In the study by Lindbaek 1996 the acute maxillary sinusitis was confirmed by a CT scan showing opacity or air-fluid level. Lindbaek 1998 included only patients with mucosal thickening in a CT scan.

Three studies reported that bacteriological samples were taken. Hadley 2010 included only culture-positive patients with specified target pathogens in the analyses. Lindbaek 1996 reported that in 58% of participants bacteriological specimens obtained from the nasopharynx grew bacteria connected with sinusitis. Haye 1996 reported that only in some samples taken from the posterior part of the nasal cavity was growth of pathogenic bacteria obtained.

On the whole, the proportion of participants with true bacterial sinusitis remained unclear in seven out of the nine placebo-controlled studies, as is the case in real clinical practice. Participants in seven of the nine studies were clearly recruited from community-based general practices, where these kinds of patients are frequently treated. Thus the results in this review are valid in adult patients in general practice settings.

Strengths and weaknesses of diagnostic options

The diagnosis of acute sinusitis is challenging in primary care, especially when trying to differentiate between a viral or bacterial origin of disease. The high proportion of participants that improved in the non-antibiotic group indicates that only some of the participants defined (by these criteria) as having acute sinusitis benefit from antibiotic treatment. Clinical examination is sensitive in ruling out sinusitis (Williams 1993) but not in identifying bacterial disease. Specific methods would be needed to find the subgroups of patients that might benefit from antibiotics. Some official opinions support the practice of diagnosing acute maxillary sinusitis by clinical examination. For example, the expert panel of five national societies in the USA has stated that sinusitis can be diagnosed, in the majority of patients, by using only the history and physical examination (Meltzer 2004). In the referral guideline for imaging of the European Commission, radiograph is not indicated as a routine diagnostic method (ECRP 2000). Further, it states that thickened mucosa is a non-specific finding and may occur in asymptomatic patients. CT scan has a role in cases of treatment failure with suspicion of complications or malignancy or when surgery is considered (ECRP 2000; Meltzer 2004). In Nordic countries, the diagnosis of sinusitis is often confirmed by means of an ultrasound device that is suitable for primary care but its use requires training and experience. Sinusitis can be ruled out in patients with no fluid retention in sinuses with this method, so it may reduce the use of antibiotics (Puhakka 2000; Varonen 2003b).

Duration of signs and symptoms

Bacterial sinusitis is more likely if the signs and symptoms have lasted at least seven days (Meltzer 2004) and therefore studies with shorter illness duration, without confirmed diagnosis by radiograph or culture, were excluded. Acute exacerbations of chronic sinusitis were considered as part of the chronic form of the disease, which is a separate and more complicated entity, often related to background factors that have an effect on recovery. Therefore, studies with acute exacerbations of chronic sinusitis were included only if they reported separately data on the subgroup with acute sinusitis, or if not reported separately at least 80% of participants had acute sinusitis.

Different classes of antibiotics

We combined data from all placebo-controlled studies, regardless of the type of antibiotic used. We anticipated that there would be only some placebo-controlled studies available for the analyses, and the main interest would be whether antibiotic treatment in general is beneficial. We presumed that researchers in different countries had local reasons for selecting particular antibiotics for their trials, for example taking into account local resistance rates to antibiotics among community-acquired pathogens.

Duration of treatment

One of our aims was to evaluate the effect of short and long courses of antibiotics. Our view is that, for this purpose, antibiotic versus antibiotic study designs without a non-antibiotic group are not valid enough. Unfortunately, there were no placebo-controlled studies available to undertake this comparison. Antibiotics also have side effects; therefore more placebo-controlled trials are needed to find the optimal duration of treatment and those subgroups that might benefit from the treatment. The same applies to other respiratory tract infections in primary health care.

Quality of the evidence

Antibiotics versus placebo

The body of evidence comparing antibiotics with placebo at one to two weeks comprised five randomised studies. When a failure was defined as a lack of full recovery or improvement (the main analysis), we assessed the body of evidence as being of moderate quality according to the GRADE assessment criteria (Summary of findings for the main comparison). This implies that further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. The pooled estimate of the five studies slightly favoured antibiotics with a RR of 0.66 (95% CI 0.47 to 0.94). Because the effect size seems to be small and the number of evaluated participants so far is 1058, it is not unreasonable, however, to anticipate that a new large study could change this estimate.

When a failure was defined as a lack of full recovery, we assessed the body of evidence as being of high quality according to the GRADE assessment criteria (the RR for failure was 0.73 (95% CI 0.63 to 0.85). This implies that further research is very unlikely to change our confidence in the estimate of effect.

The body of evidence comparing antibiotics with a placebo at two to nine weeks comprised only one randomised study which was assessed as being of low quality according to the GRADE assessment criteria (Summary of findings for the main comparison). This implies that further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. The RR for failure was 0.85 (95% CI 0.36 to 1.98) when a failure was defined as a lack full recovery or improvement, and 0.63 (95% CI 0.38 to 1.05) when a failure was defined as a lack of full recovery.

Antibiotics versus antibiotics

We assessed the quality of the body of evidence for comparing one class of antibiotic against another as being of moderate quality according to the GRADE assessment criteria (Summary of findings 2). Although the number of the studies was large (54 studies) most of them were non-inferiority studies and often supported by pharmaceutical companies. Because most patients with sinusitis will improve without antibiotics and the margin of getting added clinical benefit from antibiotics on the whole, compared to placebo, is small (about 10%), it is easy to show non-inferiority between antibiotics.

There were 54 studies comparing an antibiotic to another antibiotic. In general, we assessed the methodological quality of the studies to be modest, mostly due to incomplete blinding and/or incomplete reporting of other methodological issues, especially of randomisation (Figure 1; Figure 2). We graded only four of the 54 studies as having 'low' risk of bias based on the methodological criteria used in this review, and graded the other 50 as 'unclear' or 'high' risk of bias. In the only comparison in this review where a slight statistical difference was found between antibiotics, we graded none of the studies as having 'low' risk of bias and the two studies having highest weight (70% altogether) in the meta-analysis also had a 'high' risk of bias. This comparison was for cephalosporins versus amoxicillin-clavulanate at seven to 15 days follow-up with clinical failure defined as a lack of full recovery or improvement (Analysis 2.1). Confidence in this finding is limited by the low methodological quality and the disappearance of significance at longer follow-up.

Potential biases in the review process

Diagnosis

We expected that the diagnostic method in an individual placebo-controlled study (clinical diagnosis alone versus radiological/bacteriological confirmation) might have had an influence on the results. In the meta-analyses, however, no statistically significant heterogeneity between individual study results was observed. Clinical diversity was plausible in the studies of Lindbaek 1996 and Hadley 2010 compared to the other studies. The study of Lindbaek 1996 differed in only accepting participants with opacity or fluid-level in the sinuses, not with mucosal thickening alone, by using CT scan. The study also reported pathogenic bacteria connected with acute sinusitis in 58% of participants. The study of Hadley 2010, which was not included in the meta-analyses, differed by including only patients with isolation of pathogenic bacteria from the sinus secretion. It is obvious that the participants selected for these two studies represented people with a more severe form of the disease.

Definition of cure

The definitions of full recovery, improvement and failure in individual studies can influence the data extraction and the results, both at study and meta-analysis level, especially in cases where only a few studies are available for analysis. In studies with a dichotomous classification of the outcomes, the criteria can vary between studies, as well as compared to studies with multi-level classifications. In this review, four of the nine placebo-controlled studies reported success dichotomously. The highest number of categories was used in the studies by Lindbaek 1996 and Lindbaek 1998. The outcomes were classified into five categories: restored, much better, somewhat better, unimproved and worse. Two other studies reported outcomes in three categories and one in four categories. Dichotomising different outcome classifications for the analyses causes uncertainty in the results.

Agreements and disagreements with other studies or reviews

Three other recent systematic reviews have considered comparison of antibiotics with placebo for acute maxillary sinusitis. These reviews have used more permissive eligibility criteria than our review and would therefore include a more heterogeneous sample of individuals with and without acute bacterial rhinosinusitis. Overall, these reviews found similar antibiotic treatment effects, despite differing eligibility criteria.

The study by Rosenfeld 2007 found that antimicrobials, compared to placebo control, increased clinical improvement rates at two weeks for 800 patients by an absolute difference of 7% (95% CI 2 to 13%). Diarrhoea (1583 patients, risk difference (RD) of 0.05 (95% CI 0.01 to 0.09)) and any adverse events (1853 patients, RD of 0.11 (95% CI 0.05 to 0.16)) were more common in the group treated with antimicrobials. A patient-level meta-analysis (Young 2008) of 10 placebo-controlled trials involving 2640 patients specifically excluded trials that enrolled patients on the basis of laboratory, imaging or bacterial culture results. This study found a small increase in clinical cures with antibiotics within two weeks (OR of 1.35, 95% CI 1.15 to 1.59) and a number needed to treat to benefit (NNTB) of 15 (95% CI 7 to infinity). The authors conclude that "common clinical signs and symptoms cannot identify patients with rhinosinusitis for whom antibiotic treatment is clearly justified, not even if a patient reports symptoms for longer than 7 to 10 days". In the third meta-analysis by Falagas 2008a, including 16 randomised, double-blind trials with 2648 patients, antibiotics were associated with a higher rate of cure or improvement within two weeks (OR 1.64, 95% CI 1.35 to 2.00) but with more adverse effects (12 trials with 1963 patients, OR 1.87 (95% CI 1.21 to 2.90)). The authors conclude that "use of antibiotics for acute sinusitis confers a small therapeutic benefit over placebo with a raise in risk of adverse effects, and that antibiotics should be reserved for carefully selected patients with a higher probability for bacterial disease". 

Another review by Falagas 2008b evaluated the effects of short (up to seven days) versus longer-course (at least two days longer than short-course) antibiotic treatment for acute rhinosinusitis. Twelve trials compared the same antibiotic for varying durations. Rates of clinical success did not vary significantly for short- versus longer-term treatment (OR 0.95, 95% CI 0.81 to 1.12). Our review did not include studies comparing different durations of the same antibiotic because there were no placebo-controlled studies measuring this.

Authors' conclusions

Implications for practice

Antibiotics have a small benefit for clinical outcomes in patients with symptoms of uncomplicated acute sinusitis lasting for more than seven days in a primary care setting. However, about 80% of patients treated with a placebo also improve within two weeks.

The patients in the placebo-controlled studies were enrolled from unselected primary care populations (except perhaps Axelsson 1970; Meltzer 2005). The individual studies represented clinical conditions ranging from mild to more severe, depending on the inclusion criteria, as in a real general practice setting. All but one of the placebo-controlled studies favoured antibiotics to some extent. We see that the overall result of the meta-analysis for the main outcome corresponds fairly well to the mixture of patients in primary care. However, compared to situations where no imaging is used, the benefit of antibiotics may be somewhat overestimated. It seems that even when special effort is put into selecting patients from primary care settings who most probably have bacterial sinusitis, the actual benefit of antibiotics is modest. It is obvious that there are individual patients who benefit from antibiotics but we do not have proper means to identify them in primary care settings.

Clinicians and people responsible for producing guidelines need to weigh the moderate benefits of antibiotic treatment against the patient's general condition and the potential for adverse effects at both individual and general population level (for example, local bacterial resistance).

Implications for research

Given the small number of trials, additional high-quality, placebo-controlled trials are needed to identify the potential subgroups of sinusitis patients that may benefit from antibiotics. Diagnosis should be made on prespecified recommended clinical criteria which also allow subgroup analyses according to severity of the disease. The benefit of adding a nasopharyngeal swab in the diagnosis of patients with sufficient clinical criteria is also unclear. The data in this review did not allow such subgroup analyses.

Special attention should be paid to identifying the prognostic factors of sinusitis and finding factors that potentially modify the treatment effect, such as co-morbid conditions (for example, asthma, allergies), lifestyle and other individual factors (for example, smoking). Effects on functional status, work performance and quality of life may provide important additional information. The trials should be double-blinded, with adequate allocation and concealment procedures, and should report clinical outcomes and time to clinical response. Large trials may be needed to identify clinical predictors of the need for antibiotics. Different treatment durations should also be studied in placebo-controlled settings.

Acknowledgements

We would like to thank the Cochrane Acute Respiratory Infections Group's Trials Search Co-ordinator Sarah Thorning for help in searching the literature and the Managing Editor, Elizabeth Dooley, for all her help. We warmly thank Information Specialist Jaana Isojärvi at National Institute for Health and Welfare/THL, Finland for her help in the supplementary search of the literature. We would also like to thank Johannes Hang Guo, Ryuki Kassai, Sachiyo Yoshida, Jaana Elberkennou, Tiina Lehmussaari and Helena Laasonen for translating trial reports published in Chinese, Japanese, French, Spanish and Portuguese. We are also grateful to Systems Manager Marko Ekqvist, English Language Editor Mark Phillips, and MSC Niina Kovanen for their inputs in the previous version of this review. We would also like to thank Dr An de Sutter, Dr Jane M. Garbutt, Dr Morten Lindbaek, Dr Eli O. Meltzer, Dr Dan Merenstein and Dr Ian G Williamson for providing additional information about their trials, and Schering-Plough Research Institute for providing information on their unpublished study. We are grateful to Bruce Arroll, Sree Nair, Dilip Raghavan and Meenu Singh for valuable referee comments in previous versions, and to Bruce Arroll, Abimbola Farinde, José Luis Ferrero Albert, Terry Neeman and Devada Singh-Franco in the current version.

Data and analyses

Download statistical data

Comparison 1. Antibiotics versus placebo
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Clinical failure defined as a lack of full recovery or improvement at 7 to 15 days of follow-up51058Risk Ratio (M-H, Random, 95% CI)0.66 [0.47, 0.94]
2 Clinical failure defined as a lack of full recovery or improvement at 16 to 60 days of follow-up2 Risk Ratio (M-H, Fixed, 95% CI)Totals not selected
3 Clinical failure defined as a lack of full recovery at 7 to 15 days of follow-up5680Risk Ratio (M-H, Random, 95% CI)0.73 [0.63, 0.85]
4 Clinical failure defined as a lack of full recovery at 16 to 60 days of follow-up1169Risk Ratio (M-H, Fixed, 95% CI)0.63 [0.38, 1.05]
5 Relapse rates after 60 days1214Risk Ratio (M-H, Fixed, 95% CI)1.25 [0.72, 2.19]
6 Drop-outs due to adverse effects91818Peto Odds Ratio (Peto, Fixed, 95% CI)1.40 [0.60, 3.25]
Analysis 1.1.

Comparison 1 Antibiotics versus placebo, Outcome 1 Clinical failure defined as a lack of full recovery or improvement at 7 to 15 days of follow-up.

Analysis 1.2.

Comparison 1 Antibiotics versus placebo, Outcome 2 Clinical failure defined as a lack of full recovery or improvement at 16 to 60 days of follow-up.

Analysis 1.3.

Comparison 1 Antibiotics versus placebo, Outcome 3 Clinical failure defined as a lack of full recovery at 7 to 15 days of follow-up.

Analysis 1.4.

Comparison 1 Antibiotics versus placebo, Outcome 4 Clinical failure defined as a lack of full recovery at 16 to 60 days of follow-up.

Analysis 1.5.

Comparison 1 Antibiotics versus placebo, Outcome 5 Relapse rates after 60 days.

Analysis 1.6.

Comparison 1 Antibiotics versus placebo, Outcome 6 Drop-outs due to adverse effects.

Comparison 2. Cephalosporin/macrolide versus amoxicillin-clavulanate
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Ceph versus amox-clav; clinical failure defined as a lack of full recovery or improvement at 7 to 15 days of follow-up61887Risk Ratio (M-H, Random, 95% CI)1.37 [1.04, 1.80]
2 Ceph versus amox-clav; clinical failure defined as lack of full recovery or improvement at 16 to 60 days of follow-up71415Risk Ratio (M-H, Random, 95% CI)1.08 [0.85, 1.37]
3 Drop-outs due to adverse effects (cephalosporins)92973Peto Odds Ratio (Peto, Fixed, 95% CI)0.32 [0.21, 0.49]
4 Macrolides versus amox-clav; clinical failure defined as a lack of full recovery or improvement at 7 to 15 days of follow-up71807Risk Ratio (M-H, Random, 95% CI)0.83 [0.62, 1.13]
5 Macrolides versus amox-clav; clinical failure defined as a lack of full recovery or improvement at 16 to 60 days of follow-up4908Risk Ratio (M-H, Random, 95% CI)0.85 [0.57, 1.27]
6 Drop-outs due to adverse effects (macrolides)82550Peto Odds Ratio (Peto, Fixed, 95% CI)0.47 [0.30, 0.72]
Analysis 2.1.

Comparison 2 Cephalosporin/macrolide versus amoxicillin-clavulanate, Outcome 1 Ceph versus amox-clav; clinical failure defined as a lack of full recovery or improvement at 7 to 15 days of follow-up.

Analysis 2.2.

Comparison 2 Cephalosporin/macrolide versus amoxicillin-clavulanate, Outcome 2 Ceph versus amox-clav; clinical failure defined as lack of full recovery or improvement at 16 to 60 days of follow-up.

Analysis 2.3.

Comparison 2 Cephalosporin/macrolide versus amoxicillin-clavulanate, Outcome 3 Drop-outs due to adverse effects (cephalosporins).

Analysis 2.4.

Comparison 2 Cephalosporin/macrolide versus amoxicillin-clavulanate, Outcome 4 Macrolides versus amox-clav; clinical failure defined as a lack of full recovery or improvement at 7 to 15 days of follow-up.

Analysis 2.5.

Comparison 2 Cephalosporin/macrolide versus amoxicillin-clavulanate, Outcome 5 Macrolides versus amox-clav; clinical failure defined as a lack of full recovery or improvement at 16 to 60 days of follow-up.

Analysis 2.6.

Comparison 2 Cephalosporin/macrolide versus amoxicillin-clavulanate, Outcome 6 Drop-outs due to adverse effects (macrolides).

Comparison 3. Non-penicillin antibiotics versus beta-lactamase sensitive penicillins
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Clinical failure defined as a lack of full recovery or improvement at 7 to 15 days of follow-up71083Risk Ratio (M-H, Random, 95% CI)0.70 [0.47, 1.06]
2 Clinical failure defined as a lack of full recovery or improvement at 16 to 60 days of follow-up1436Risk Ratio (M-H, Fixed, 95% CI)0.67 [0.37, 1.20]
3 Drop-outs due to adverse effects71208Peto Odds Ratio (Peto, Fixed, 95% CI)0.58 [0.25, 1.35]
Analysis 3.1.

Comparison 3 Non-penicillin antibiotics versus beta-lactamase sensitive penicillins, Outcome 1 Clinical failure defined as a lack of full recovery or improvement at 7 to 15 days of follow-up.

Analysis 3.2.

Comparison 3 Non-penicillin antibiotics versus beta-lactamase sensitive penicillins, Outcome 2 Clinical failure defined as a lack of full recovery or improvement at 16 to 60 days of follow-up.

Analysis 3.3.

Comparison 3 Non-penicillin antibiotics versus beta-lactamase sensitive penicillins, Outcome 3 Drop-outs due to adverse effects.

Comparison 4. Tetracyclines versus mixed classes of antibiotics
Outcome or subgroup titleNo. of studiesNo. of participantsStatistical methodEffect size
1 Clinical failure defined as a lack of full recovery or improvement at 7 to 15 days of follow-up5807Risk Ratio (M-H, Random, 95% CI)1.09 [0.70, 1.71]
2 Drop-outs due to adverse effects5854Peto Odds Ratio (Peto, Fixed, 95% CI)0.73 [0.33, 1.60]
Analysis 4.1.

Comparison 4 Tetracyclines versus mixed classes of antibiotics, Outcome 1 Clinical failure defined as a lack of full recovery or improvement at 7 to 15 days of follow-up.

Analysis 4.2.

Comparison 4 Tetracyclines versus mixed classes of antibiotics, Outcome 2 Drop-outs due to adverse effects.

Appendices

Appendix 1. MEDLINE (Ovid) search strategy

  1. exp Sinusitis/

  2. sinusit*.tw.

  3. or/1-2

  4. exp Anti-Bacterial Agents/

  5. antibiotic*.tw,nm.

  6. or/4-5

  7. 3 and 6

  8. randomized controlled trial.pt.

  9. controlled clinical trial.pt.

  10.  randomized.ab.

  11.  placebo.ab.

  12.  clinical trials as topic.sh.

  13.  randomly.ab.

  14.  trial.ti.

  15.  8 or 9 or 10 or 11 or 12 or 13 or 14

  16.  (animals not (humans and animals)).sh.

  17.  15 not 16

  18.  7 and 17

Appendix 2. Embase.com search strategy

#11. #7 AND #10
#10. #8 OR #9
#9. random*:ab,ti OR placebo*:ab,ti OR crossover*:ab,ti OR 'cross over':ab,ti OR 'cross-over':ab,ti OR factorial*:ab,ti OR (doubl* NEAR/1 blind*):ab,ti OR (singl* NEAR/1 blind*):ab,ti OR assign*:ab,ti OR allocat*:ab,ti OR volunteer*:ab,ti
#8. 'randomized controlled trial'/exp OR 'single blind procedure'/exp OR 'double blind procedure'/exp OR 'crossover procedure'/exp
#7. #3 AND #6
#6. #4 OR #5
#5. antibiotic*:ab,ti
#4. 'antibiotic agent'/exp
#3. #1 OR #2
#2. sinusit*:ab,ti
#1. 'sinusitis'/exp

Appendix 3. SIGLE search strategy

keyword: sinusit* AND Subject collection: 06 - Biological and medical sciences

Appendix 4. Search strategy for ongoing trials

acute maxillary sinusitis AND antibiotics

Appendix 5. Details of previous searches

Review update 2011

In the 2011 review update, we searched the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2010, Issue 1), which contains the Cochrane Acute Respiratory Infections Group's Specialised Register, MEDLINE (2007 to March Week 2, 2010) and EMBASE (2007 to March 2010) and the System for Information on Grey Literature in Europe (SIGLE) (1976 to August 2009). We used the following search terms to search MEDLINE and CENTRAL. We combined the MEDLINE search with the Cochrane Highly Sensitive Search Strategy for identifying randomised trials in MEDLINE: sensitivity- and precision- maximising version (2008 revision); Ovid format (Lefebvre 2011). The search terms were adapted to search EMBASE (Appendix 2). To locate unpublished literature, the search of SIGLE was carried out on the OpenSIGLE (http://opensigle.inist.fr/) (Appendix 3).

MEDLINE (Ovid)

  1. exp Sinusitis/

  2. sinusit*.tw.

  3. or/1-2

  4. exp Anti-Bacterial Agents/

  5. antibiotic*.tw,nm.

  6. or/4-5

  7. 3 and 6

  8. randomized controlled trial.pt.

  9. controlled clinical trial.pt.

  10.  randomized.ab.

  11.  placebo.ab.

  12.  clinical trials as topic.sh.

  13.  randomly.ab.

  14.  trial.ti.

  15.  8 or 9 or 10 or 11 or 12 or 13 or 14

  16.  (animals not (humans and animals)).sh.

  17.  15 not 16

  18.  7 and 17

We used the function for finding related articles in PubMed for those already identified placebo-controlled trials to track down additional, relevant articles. We reviewed reference lists of the identified new trials and four systematic reviews considering placebo-controlled study designs (Falagas 2008a; Gwaltney 2004; Rosenfeld 2007; Young 2008) for additional, appropriate studies.

There were no language or publication restrictions.

Review update 2008

In the review version 2008, we used a revised, more specific strategy (compared to review version 2003) to search for trials in the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2007, Issue 3); MEDLINE (1950 to May 2007) and EMBASE (1974 to June 2007). In the revised strategy "antibiotics" was used as an additional keyword and "anti-bacterial agents" as an exploded MeSH term in addition to "sinusitis". The MEDLINE search terms were run over CENTRAL and adapted for EMBASE.

MEDLINE (OVID)
1 exp SINUSITIS/
2 sinusitis.mp.
3 or/1-2
4 exp Anti-Bacterial Agents/
5 antibiotic$.mp.
6 or/4-5
7 3 and 6
8 RANDOMIZED CONTROLLED TRIAL.pt.
9 CONTROLLED CLINICAL TRIAL.pt.
10 RANDOMIZED CONTROLLED TRIALS.sh.
11 RANDOM ALLOCATION.sh.
12 DOUBLE BLIND METHOD.sh.
13 SINGLE-BLIND METHOD.sh.
14 or/8-13
15 Animals/
16 human.sh.
17 15 not 16
18 14 not 17
19 CLINICAL TRIAL.pt.
20 exp Clinical Trials/
21 (clin$ adj25 trial$).ti,ab.
22 ((singl$ or doubl$ or trebl$ or tripl$) adj25 (blind$ or mask$)).ti,ab.
23 PLACEBOS.sh.
24 placebo$.ti,ab.
25 random$.ti,ab.
26 or/19-25
27 26 not 17
28 18 or 27
29 7 and 28

The function for finding related articles in PubMed was used for those already identified placebo-controlled trials so as to track down additional, relevant articles. Reference lists from the already identified trials and nine systematic reviews considering placebo-controlled study designs (Benninger 2000; de Bock 1997; de Ferranti 1998; Ioannidis 2001; Ioannidis 2002; Ip 2005; Linder 2003; Low 1997; Stalman 1997b) were reviewed for additional, appropriate studies. Included and excluded trials in the previous version were rechecked by using the revised inclusion criteria for this review. A search of the database 'System for Information on Grey Literature in Europe (SIGLE)' was intended, but the database was not available via the Internet in November 2005.

For the review version 2003 pharmaceutical companies that manufacture antibiotics used in the treatment of acute sinusitis had been contacted and data and references from all published and unpublished trials on acute sinusitis were requested. Further, three experts in the field had been contacted and asked to review the bibliography of the initial review (1999) for completeness.

Appendix 6. Results of the earlier searches

Review update 2011

In the 2011 update, the total number of the records considered from all updates was 2209 including 2036 records from the earlier searches and 173 records from the 2010 update search. Of the total 2209 records, we rejected 1923 records as definitely not meeting the inclusion criteria simply on the basis of title or abstract or because they were duplicates. Altogether 286 full-text reports were obtained. From these 286 reports, 128 were clearly irrelevant for this review leaving 158 reports for final assessment. The main reasons for irrelevance were: trials without control group, studies with treatments other than antibiotic, studies comparing same classes of antibiotics and studies regarding children.

Over all updates, there were in total 158 reports to be considered in detail. Additional information was needed from four studies with antibiotic versus placebo comparisons to assess whether they meet the inclusion criteria for this review (Meltzer 2005; Schering-Plough 2003; Stalman 1997a; Williamson 2007). Finally we considered 62 reports representing 59 individual studies eligible for inclusion in the review. Compared to the 2008 update, two new antibiotic versus antibiotic studies were included. Eighty-nine studies with 96 reports were excluded.

Review update 2008

The electronic search based on the revised search strategy in the update 2008 produced 2030 records, many of which were duplicates. Of these, 1761 records were rejected as definitely not meeting the inclusion criteria simply on the basis of title or abstract. Duplicates were omitted. Checking the lists of the included and excluded studies in the previous version of the review (version 2003), reviewing reference lists from the already identified trials and systematic reviews and using the function for finding related articles in PubMed yielded six more appropriate trials. Altogether 269 full-text reports were obtained. All non-English language reports were translated to assess the studies. The review authors could read reports in German, Russian and Scandinavian languages. Outside translators were consulted to identify and assess the reports in Portuguese, Spanish, French, Chinese and Japanese. From these 269 reports, 117 were clearly irrelevant for this review. The main reasons for exclusion were: trials without control group, studies with other treatment than antibiotic, studies comparing same classes of antibiotics and studies regarding children.

In total there were 152 reports to be considered in detail. From these, 60 reports representing 57 individual studies were considered eligible for inclusion in the review. Six of the 57 included studies compared antibiotic treatment to placebo. The remaining 51 studies compared different classes of antibiotics.

Feedback

Antibiotics for acute maxillary sinusitis

Summary

There are 4 primary questions to be answered by this systematic review:

1) Is antibiotic therapy effective for acute sinusitis?
2) Which antibiotic is most effective?
3) What is the best duration for antibiotic therapy?
4) Is there a change in the efficacy of antibiotics over time?

Question 1

Firstly, all 4 questions are clinically important. But it is paramount to recognize the interrelationships of all 4 questions, such that, if the answer to the first question is negative, the value of the other answers is significantly diminished. It is therefore essential to include all the evidence unless the exclusion is completely justified.

On the efficacy of antibiotics for acute sinusitis, there are only 6 published randomised clinical trials with control groups: Axelsson 1970, Gananca 1973, Wald 1986, Lindbaek 1996, van Buchen 1997, and Stalman 1997.

I agree that the Wald study involving children and the Gananca study from Brazil should be excluded. Gananca trial had only 50 subjects, with 50% more patients in the antibiotic group after randomisation. It also had extreme differences in cure rates between antibiotic and placebo groups (15/30 versus 0/20).

The last 3 trials on this list, according to the recent meta-analysis by de Ferranti et al (BMJ 1998;317:632-7), received a quality score of 5 by the Jadad scale (a maximum score indicating highest quality). The Axelsson study had a score of 1. The Williams et al answered this first question, "Is antibiotic therapy effective?", by excluding Stalman 1997 (with quality score of 5) and by doubling the control group populations in Axelsson 1970 (n=34) and in Lindbaek 1996 (n=44). This artificially created population of 156 controls from 78 real controls would inflate the contributions of those 2 trials on the outcome. The exclusion of those true placebo-controls in Stalman 1997 (n=94), while adding 78 phantom controls, is not a reasonable approach to answer the antibiotic efficacy question for sinusitis.

The Stalman study was excluded, according to the author's, because it did not use radiographic images for diagnosis. The same is true however for the included Axelsson study. Though Axelsson et al took several different radiographic views of sinuses, they did not describe a radiographic criteria for diagnosis.

Moreover, the relationship between radiograms, even the more advanced images of paranasal sinuses obtained by computed tomography (CT scan), and the clinical description of sinusitis is extremely poor. The 1997 study of Bhattaharyya et al involving 586 patients (the largest published data set) failed to find any correlation between CT scan results and patient described symptoms of sinusitis (See Arch. Otolaryngol. 1997;123:1189-92). Also note that Williams et al argue that the cure rate differences between Lindbaek and van Buchem were probably related to the diagnostic criteria. That is, a diagnosis confirmed by CT scan is more likely associated with bacterial sinusitis than a diagnosis by radiography. This new theory of differential diagnosis of bacterial sinusitis however is not supported by any published data.

To be consistent, either both Axelsson and Stalman should be included, or both should be excluded. If they are both excluded, then the meta-analysis of Williams et al on the efficacy of antibiotics for sinusitis would depend on only 2 placebo controlled trials - Lindbaek 1996 and van Buchem 1997 - with opposite conclusions.

The Lindbaek study was terminated early with 130 patients instead of the original sample size of 180 subjects. This was done without correction and without predefined stopping rules. Also, the randomisation was in a most unusual fashion for a modern clinical trial using dice. Lindbeak et al do not provide a table, as customary, showing the baseline patient characteristics at entry indicating their randomisation was effective. In contrast to those deficits, van Buchem reports this baseline information about randomisation and an analysis of prognostic factors on the outcome. It shows that initial severity of sinusitis does not change the results. The van Buchem study has a much higher quality and reliability than the Lindbaek trial. If the entire evidence has to be based on combining or contrasting only those 2 trials, then conclusion has to follow the van Buchem results.

Most important, the Stalman trial which was published 9 months later completely supports the van Buchem results. Among the 3 trials with Jadad score of 5, reasonable people can easily conclude that van Buchem and Stalman clearly show that antibiotics for acute sinusitis are not effective; and, Lindbaek et al is an outlier. Lindbaek's efficacy claim based on one endpoint (cure) is not sustained by his second endpoint (improvement). This inconsistent outcome and those basic problems associated with early termination and poor randomisation perhaps is why Lindbaek results are not replicated by the last 2 trials (van Buchem and Stalman).

I suggest however that we include all 4 clinical trials to combine results and estimate a pooled efficacy. As such, even this would yield still a meager sample size, a total of 648 subjects with only 278 controls. Consequently, this is the entire dataset we have to establish the scientific evidence-base in justifying 15 million antibiotic prescriptions in the U.S. for sinusitis. Instead of finding excuses to exclude the few controls in this limited database, we should savor them.

The answer to the first question, then, is a simple pooling of the results from those 4 clinical trials: Axelsson 1970 (n=112), Lindbaek 1996 (n=130), van Buchem 1997 (n=214), and Stalman 1997(n=192).

In all sinusitis trials the primary outcome measures of at 10-14 days are subjective. Outcomes are not based on some "Gold Standard" but study-variant definitions of cure and success (cure/improvement) by each investigator. Those primary data unfortunately are not based on objective endpoints. A well defined diagnostic criteria, such as combining radiographs with symptoms, validated against a standard is lacking.

I argue that if the entry includes radiograms to make the diagnosis of sinusitis, then to be consistent for at least in defining cure, the primary endpoint should also include both measures (e.g.. combination of normal radiograph with symptom free status). If both measures are not negative, then improvement (or scaled improvement) is the only logical alternative. Without a normal radiogram, cured diagnosis would be in contradiction with the entry criteria.

Regardless of these problems associated with an objective and study invariant outcome, the combined results from 4 trials are as follows:

Cure

Controls: 10/34 + 5/44 + 53/106 + 56/94 = 124/278 or 44.6%

Antibiotics: 27/78 + 32/86 + 68/108 + 59/98 = 186/370 or 50.3%

Success

Controls: 23/34 + 39/44 + 78/106 + 80/94 = 220/278 or 79.1%

Antibiotics: 51/78 + 81/86 + 87/108 + 83/98 = 302/370 or 81.6%

Antibiotic-placebo rate difference in cure is 5.7% (OR=1.26) and the rate difference in success is 2.5% (OR=1.17). These marginal differences, neither statistically nor clinically significant, indicate the lack of efficacy of antibiotic therapy for the treatment of acute paranasal sinusitis.

The methods used by Williams et al in their meta-analysis therefore must be re-examined. Relying on magical statistical packages, with interesting inclusions and exclusions, can sometimes lead to positively misleading results. As in this case, claiming antibiotic efficacy for sinusitis is not a reasonable conclusion but perhaps an artifact of meta-analysis.

The answer to this first question, "Are antibiotics for acute sinusitis effective?", therefore is negative.

Question 2

Now the second question, which antibiotic is most effective, is a less important scientific issue. Notwithstanding, in real life clinical practice, this question continues to be a primary issue because of the ongoing campaign of promoting newer, wider spectrum antibiotics for sinusitis with higher costs.

Though there is no clinical trial which exclusively has focused on bacterial sinusitis caused by resistant bacteria, and which has shown the benefit of such drugs, the emergence of multi-drug resistant bacteria is always cited to advocate those wide spectrum drugs. Therefore, the second part of the Williams meta-analysis is clinically most significant.

Using a large sample size, Williams et al (n=7330 patients in 32 trials) show that there is no superior antibiotic. This result supports the earlier findings of de Bock 1997 (n=3358 patients in 16 trials) and de Ferranti 1998 (n=2717 patients in 27 trials) meta-analyses. The ineffectiveness of wide spectrum drugs are based on a wide collection of clinical trials. In fact, among those 3 meta-analyses, the inclusion of the same trials are not common. The number of trials commonly inclusive between meta-analyses are: Williams - de Ferranti (13/47), Williams - de Bock (3/45), and de Ferranti - de Bock (5/39). This diversity in inclusion/exclusion indicates the arbitrary nature of selection of trials included in meta-analyses while supporting the robustness of conclusions.

Note also that this result, ineffectiveness of wide spectrum antibiotics for sinusitis, is consistent with the meta-analysis findings in acute otitis media. It is a logical result since acute otitis media in general has the same pathogenic bacterial spectrum as acute sinusitis.

There is however an interesting point to consider. If the antibiotics are not better than placebo, then it is logical not to find a superior antibiotic for sinusitis. Unless each new antibiotic is concurrently matched with a placebo control in the primary clinical trial, it would be impossible to show the usefulness of new, wide spectrum antibiotics. Without such data, I think it is irresponsible to advocate wide spectrum antibiotics for sinusitis.

I hope that Williams et al would reduce the complexity of their presentation to make this point by decreasing the number of figures and by selecting a single primary endpoint to display. I suggest success (cure/improvement) as the measure of choice as I discussed above. The remaining measures can be simply stated as not being different with some odds ratios or rate differences. Also, those redundant figures with confidence intervals all over the place, are they necessary? It only illustrates that poor evidence is being included in the meta-analysis.

I suggest that their Summary figure should also include sample sizes, exclude tetracycline's, include duration of therapy analysis (short versus long), and finally include some cumulative meta-analysis to indicate change in efficacy over time.

Question 3

Authors conclude in recommending 10-days of amoxicillin therapy for acute sinusitis but do not provide evidence-based arguments for selecting this particular duration of therapy. This is not scientific and somewhat odd, because a few years ago, Dr.Williams published arguably the best study showing that 3-days of antibiotic therapy is as good as the standard 10-days of therapy (See JAMA 1995;273:1015-21).

A recent review by Pichichero and Cohen (Pediatr Infect Dis J 1997;16:680-95) indicates that there are at least 5 published studies involving adult patients which show that short duration therapy is as effective as long duration therapy. I suggest Williams et al provide a meta-analysis to answer this question or explain why such combination of data is not possible. This part on duration of therapy I think is extremely important.

Question 4

The recent increase in the prevalence of multi-drug resistant bacteria is clearly shown to be changing rapidly and is caused by the misuse or overuse of antibiotics. However, the research about the change of efficacy of antibiotics in vivo for common respiratory diseases is lacking. This is an important question. I am not sure how it can be answered. The approach taken by Williams et al seems a reasonable one, but I doubt that it can really answer the question of changing antibiotic efficacy for sinusitis.

The change in efficacy over time perhaps can be best approached by using a cumulative meta-analysis (See N Eng J Med 1992;327:248-54, JAMA 1992;268:240-8). Williams et al state that they have performed such a cumulative meta-analysis as well as a meta-regression. However, they do not show those results. I think the cumulative meta-analysis involving 32 trials in some chronological order (by the date of the trial, and not the publication date) would be an excellent figure to make the point.

We must also keep in mind another possibility. The lack of change in efficacy over time, suggesting that the emergence of drug resistant bacteria had no effect, could be simply untrue but related to the ineffectiveness of antibiotics for sinusitis in general.

With $2 billion over the counter medications, 15 million antibiotic prescriptions, 17 million office visits, 300,000 endoscopic sinus surgeries, and 37 million Americans suffering from chronic sinus disease, this is not a trivial healthcare issue. If the Cochrane is going to enter a systematic review in it's library, it should be based on the best evidence and analysis. William et al require many changes and much improvement before it is suitable for the Cochrane database.

I certify that I have no affiliations with or involvement in any organisation or entity with a direct financial interest in the subject matter of my criticisms.

Reply

Dr. Cantekin identifies the question – "are antibiotics more effective than placebo for acute sinusitis" as critically important but disagrees with a number of our methods for answering this question. We will address each point.

First, it is stated that the control populations are double counted. This is an important and difficult methodological issue that occurs when trials have more than two treatment arms. It was only problematic for the Lindbaek trial, which had amoxicillin, penicillin and placebo arms. To avoid "double counting," we analysed the amoxicillin and penicillin trials separately, thereby including the control populations only once in each of these separate analyses.

Second, it is argued that the Stalman trial should be included in the analysis. We agree that the Stalman study is high quality, but it was excluded because it lacked a diagnostic radiographic or sinus puncture. Our review was restricted to studies enrolling adult subjects with sinusitis as determined by clinical symptoms and either an abnormal radiographic or sinus aspiration (the Stalman study used neither). Because clinical symptoms alone lead to relatively high rates of misclassification, we thought the requirement for an objective measure (x-ray or aspiration) was important. Studies with high rates of misclassification may include large proportions of subjects without bacterial sinusitis, leading to a negative result because antibiotics are unlikely to help allergic or viral rhinosinusitis. These criteria were applied to all studies considered for inclusion. Although we excluded studies without a diagnostic criterion standard, authors of other literature syntheses on this topic have used more liberal inclusion criteria by dropping the requirement for an objective measure of sinusitis. Their results were consistent with our findings (risk ratio to prevent clinical failure 0.54, 95% CI 0.37 to 0.79)1.

Third, simple pooling is suggested as an appropriate method for combining trial results. We disagree. Simple pooling (as opposed to meta-analytic techniques), can lead to biased results for a variety of reasons. These include bias if there is imbalance in the number of subjects in treatment arms, underestimation of the variance, and inappropriate weighting of small and large trials. Further, it is not possible to evaluate heterogeneity through simple pooling.

Next, the reviewer reaches the conclusion that antibiotics are no more effective than placebo. Although the trials are few, and the results limited to patients with maxillary sinusitis confirmed radiographically or by aspiration, we interpret the current evidence as showing that penicillin is more effective than placebo and that there are similar, but more heterogeneous findings for amoxicillin.

The reviewer asks for a single endpoint to simplify the presentation. We agree that it increases the complexity of the presentation to consider both clinical cure and clinical cure or improvement. However, we think that these two clinical outcomes give information that is worth the additional complexity. As clinicians, we think it is useful to be able to tell patients that on average 50% will be well and 75% will feel substantially better within the next 10 days.

The issue of cumulative meta-analysis is important. We report the findings of these results without showing the forest plots because of limitations of the software. For newer non-penicillins versus penicillins, there was no effect over time for clinical cure (P = 0.17) or for clinical cure or improvement (P = 0.60). It is difficult to draw robust conclusions from this analysis because clinically important changes (other than bacterial resistance) may affect the results.

The reviewer asks for an analysis addressing the duration of treatment. We agree that treatment duration is a clinically important issue. In fact we completed a study showing therapeutic equivalency for three versus 10 days of TMP/SMX2. This study was excluded from the current literature synthesis because it was a within-class antibiotic comparison (not eligible). Four other studies compared two antibiotics with differing duration of treatment. Three of these compared two different antibiotics for differing duration making it difficult to isolate treatment duration as the key variable. Only one study compared two durations for the same antibiotic (cefixime four versus 10 days) and found similar efficacy. Because of differing pharmacokinetics and tissue penetration, we do not think it is wise to generalise results about TMP/SMX and cefixime to antibiotics in general. Readers should incorporate these findings cautiously. Because study duration was not one of our study questions, we have not done a systematic search to identify relevant literature. We will consider adding this question to future reviews.

We strongly endorse the reviewer's call for more clinical trials in this area, particularly with placebo control and look forward to other readers' comments.

Reference list

1de Ferranti SD, Ioannidis JPA, Lau j, Anninger WV, Barza M. Are amoxycillin and folate inhibitors as effective as other antibiotics for acute sinusitis? A meta-analysis. BMJ 1998;317:632-637

2Williams JW, Jr., Holleman DR, Jr., Samsa GP, Simel DL. Randomized controlled trial of 3 vs 10 days of trimethoprim/sulfamethoxazole for acute maxillary sinusitis. JAMA 1995; 273:1015-1021.

I certify that I have no affiliations with or involvement in any organisation or entity with a direct financial interest in the subject matter of my criticisms.
John W. Williams Jr

Contributors

Erdem Cantekin
This response was originally added to the Cochrane Review in 2000.

Amoxicillin results for cure

Summary

The heterogeneity in this outcome may be explained by the choice of cure as your outcome measure. You are pooling relative benefit instead of relative risk. If the outcome is entered as failure to cure the heterogeneity disappears and the pooled RR is significant even with a random model. RR = 0.68 (95% CI 0.55 to 0.83).

In my view it is therefore wrong to conclude that Amoxicillin does not have significant benefit in maxillary sinusitis.

I certify that I have no affiliations with or involvement in any organisation or entity with a direct financial interest in the subject matter of my criticisms.

Reply

The current author group has analysed the data differently to the earlier authors group. Unlike the previous version, in the meta-analyses the data of all placebo-controlled studies were combined across antibiotic classes. In the new analyses pooled risk ratios using the clinical failure rate as an outcome measure were calculated.

The new meta-analyses results do not have any statistically significant heterogeneity between the individual study results.

The data of all placebo-controlled studies were combined despite which antibiotic was used because it was anticipated that there would be only a few placebo-controlled studies available in the analyses. Further, it was presumed that researchers in different countries had had local reasons for selecting particular antibiotics for their trials, taking into account for example, local resistance rates to antibiotics among community-acquired pathogens.

This reply was added 9 October 2007.

Contributors

Christopher Cates
This feedback comment was posted 1999.

Antibiotics for acute maxillary sinusitis, 27 September 2010

Summary

Dear Authors,

The Cochrane Report on "acute maxillary rhinosinusitis" is not very helpful for GPs because it is misleading in some points: 

A) Most of the relevant studies are ruled out by not fulfilling the inclusion criterion "Duration of symptoms > 1 week". But  "Acute maxillary sinusitis as defined by:  a history of URTI lasting 7 to 30 days...". But there are important issues besides diagnostic via duration of sickness and CT findings. 

B) For 5 of the remaining 6 placebo-controlled RCTs used for the meta-analyses, the external validity  is weak. For practical applicability it is not so important to know if there might be any subgroups in acute rhinosinusitis that profit from antibiotics. Rather, we need results from trials with completely realistic practice conditions. Mixed inclusion criteria for meta-analyses are therefore questionable.    

*  X-ray or CT is not routinely accessible in general practice nor indicated for diagnosis of acute rhinosinusitis. As stated in the report "In the referral guidelines for imaging of the European Commission, radiograph is not indicated as a routine diagnostic method (ECRP 2007)".  The results of the following studies are therefore not applicable: Axelsson 1970, Lindbaek 1996a+b, Lindbaek 1998a+b, van Buchem 1997.   

* The Haye-trial used x-ray to exclude those with positive findings. As there has been yet no general-practice-based-trial to prove that these findings would be a good reason for antibiotics anyway, this filter criterion seems not only artificial but also deletes the practical use.  

C) Most of the Cochrane-report is concerned with the comparative trials. But very few of these studies applied any of the successful diagnostic criteria (see below) - so unequivocal results were likely to result.  

In fact, my main point is that for us GPs it is very important to keep in mind the special prevalences, supplies and priorities of primary care.Thus we focused our german DEGAM (German Society of General Practitioners) guideline "akute rhinosinusitis" (http://www.degam-leitlinien.de/)  on primary care.  We grouped the evidence from diagnostic and therapeutic clinical trials in the categories "predominantly primary care/secondary care/unknown setting". (The guideline included literature on sinusitis until April 2007 - the trials of Haye and Merenstein were not included).

Evaluating literature on acute rhinosinusitis in adults, this focus leads to interesting discoveries:   

1. In primary care, antibiotics proved only helpful when a technical diagnostic test was added to clinical evaluation: CT (Lindbaek 1996) / nasal swabs + culture (Kaiser 2001) / CRP (Hansen 2000); trials with clinical evaluation only (Stalman 1997, Bucher 2003, Merenstein 2005) or added plain-x-ray (van Buchem 1997, Rantanen 1973, Norrelund 1978, Kristo 2005) or added ultrasound (Laine 1998, Varonen 2000, Varonen 2003) were negative.  
2. Strikingly, in secondary care (ENT) studies with CE plus plain x-ray were significantly positive (Axelsson 1970, Wald 1976)  
3. Comparative studies showed no significant difference between various types of antibiotics (Stalman 1997, de Ferranti 1998, de Bock 1997, Williams 2000, Piccirillo 2001)  
4. Nearly all trials with positive effect of corticoid nasal spray for acute rhinosinusitis were done by allergologists (Sinusitis Study Group) - thus making  confounding by allergies likely. 

The above findings might be explained by combining the following observations:   

1. In cases of acute maxillary sinusitis the prevalence of pathogenic bacteria in sinus puncture differs by setting: 34% in primary care (Hansen 1995, van Buchem 1995) / ENT 45% (Axelsson 1973) / soldiers 72-78% (Savolainen 1997, Penttil? 1997) - the latter probably due to endemic spreads of hemophilus influenzae.  
2. Diagnostic tests can moderately "raise" the prevalence of bacterial vs viral infections. Positive Likelihood-ratios for diagnostic test are approximately 4 for CT (Hansen 1995) / plain X-ray 3,4 (Vaaronen 2000) / ultrasound 2,8 (Vaaronen 2000) / clinical test 1-2 (Hansen 1995) and 2.1-3.3 for CRP (Lindbaek 1996)  
3. Calculating post-test prevalence by applying LR to pre-test prevalence suggests that antibiotic trials were only significantly positive when post-test bacterial prevalence was > 60%  (probably due to relatively small trials) 

Comment: In a health system with immediate direct accessibility to specialists (like in Germany), the study results suggest that primary care guidelines should be applied for example by ENT-specialists when involved directly. Referred patients would probably merit from secondary care guideline.
Because of interpretation of mixed data, we discarded the use of the results of the preceding Cochrane-report of Williams "Antibiotics for acute maxillary sinusitis". 

Conclusion: after reading your interesting report, I would summarize the primary-care-related evidence on antibiotics for acute rhinosinusitis as follows    
* Antibiotics + referral if complications are suspected   
* Antibiotics with small proven effect if symptoms >7days (Merenstein 2005, ?Haye?)   
* Antibiotics with small proven effect if symptoms <7days + severe pain + CRP and/or ESR is markedly raised (Hansen 2000) 

I would like to suggest that future Cochrane reports reflect the difference of primary and secondary (+maybe tertiary) care. 

With best greetings from Germany and thanks for your work.
Uwe Popert  

Submitter agrees with default conflict of interest statement: I certify that I have no affiliations with or involvement in any organization or entity with a financial interest in the subject matter of my feedback.

Reply

Thank you for these comments. We understand the need to have guidance in primary care; however, the review only can use the available studies. Cochrane Reviews typically do not provide a strong recommendation on how to apply the results in clinical practice but leave these conclusions to be done locally, as practice also depends on many other factors than evidence of effectiveness.

The purpose of this review was to quantify the effectiveness of antibiotic therapy for adult patients diagnosed with acute sinusitis and treated in ambulatory settings (primary or secondary care). In this review only randomised controlled studies with adults were included and thus studies with children, e.g. Kristo 2005 and Wald 1976, were excluded.

We aimed to find trials which focused on patients who most probably had bacterial sinusitis. In primary care, diagnosis of adult acute sinusitis is often symptom-based alone, without e.g. any imaging (e.g. Meltzer 2004; Fokkens 2007). Many clinical guidelines in US and Europe state that patients with URTI and sinus symptoms should not be treated with antibiotics during the first seven to 10 days in case the signs and symptoms are not severe. Bacterial sinusitis is more probable if the signs and symptoms have lasted at least seven days (Meltzer 2004). Therefore we defined that if study inclusion criteria were clinical only, symptoms had to have lasted at least seven days. Although CT or endoscopy or culture are not investigations of primary care, studies which used these diagnostic methods were not excluded, however, from the analyses of studies conducted with  primary care patients, although it may be that the proportion of the patients with bacterial infection is higher.

Five of the six placebo-controlled studies (which fulfilled our definition of acute maxillary sinusitis) enrolled study participants from unselected primary care patients, even though the study procedures were conducted in secondary care. The study by Axelsson 1970 did not define the setting where the patients were enrolled. None of the included studies reported patients to be enrolled from secondary care.

Results and conclusions

The main outcome in this review was clinical failure defined as a lack of cure or improvement at seven to 15 days of follow-up. On average 80% of participants in placebo group were found to be cured or improved at seven to 15 days after the start of treatment, in antibiotic group the figure was 90% (difference of 10% in the number of failures). One study (Lindbaek 1996+Lindbaek 1996b) accepted solely participants with opacity or fluid-level by using CT scan. In this study pathogenic bacteria connected with acute sinusitis were documented in 58% of participants. It is obvious that the participants selected for this study represented patients with a more severe form of disease. Another study (Haye 1998) included patients with clinical symptoms and signs of acute maxillary sinusitis but excluded those with radiological evidence of empyema thereby selecting probably the mildest end of the diseased. Even in these two studies representing extreme ends of the severity spectrum of the disease, the differences between antibiotic and placebo groups regarding the numbers of failures were similar (10% and 4%, respectively).

The patients in the placebo-controlled studies were enrolled from unselected primary care populations (except perhaps Axelsson 1970). The individual studies represented clinical conditions ranging from mild to more severe, depending on the inclusion criteria, like in real GP practice setting. All these studies favoured antibiotics to some extent or were inconclusive. After pooling the results of the individual studies we have made several sensitivity analyses for the main outcome and have discussed the uncertainty issues related to the results.

In conclusion, we see that the overall result of the meta-analysis of this outcome corresponds to the mixture of patients in GP practice. However, compared to the situation where no imaging is used, the benefit of antibiotics may be somewhat overestimated. Even when special effort is put to select those patients from primary care who most probable have bacterial sinusitis, the benefit of antibiotics seems to be modest. It is obvious that there are patients who benefit from antibiotics but we do not have proper means to identify them so far.

This reply was added 20 December 2010.

Contributors

Uwe Popert

What's new

DateEventDescription
20 March 2013New citation required but conclusions have not changedWe amended the secondary outcome 'Ability to work' to include activity impairment in general.
Our conclusions remain unchanged.
20 March 2013New search has been performedSearches conducted. We included three new studies comparing antibiotic against placebo (Garbutt 2012; Hadley 2010; Meltzer 2005) and one new study comparing antibiotic against antibiotic (Lari 2010) in this updated review. We excluded one new study comparing antibiotic against placebo (Gananca 1977).

History

Protocol first published: Issue 1, 1997
Review first published: Issue 3, 1999

DateEventDescription
2 November 2010Feedback has been incorporatedFeedback comment added.
19 March 2010New search has been performedSearches conducted. We included two new studies comparing antibiotic against antibiotic (Desrosiers 2008; Marple 2010) in this updated review. We excluded six new studies (Miyamoto 2005; Sabater 1995; Schering-Plough 2003; Serra 2007; Stalman 1997a; Williamson 2007). The conclusions remain unchanged.
20 December 2007AmendedConverted to new review format.
9 October 2007Feedback has been incorporatedResponse to Feedback added.
5 October 2007New citation required and conclusions have changedSubstantive amendment.
29 May 2007New search has been performedSearches conducted.
5 October 2005Feedback has been incorporated Feedback added.
1 December 2001New search has been performedSearches conducted.
6 October 1998New search has been performedSearches conducted. Review first published Issue 3, 1999.

Contributions of authors

Anneli Ahovuo-Saloranta (AAS) was responsible for study selection, data extraction, analysis and writing the review.
Ulla-Maija Rautakorpi (UMR) was responsible for study selection, data extraction and writing the review.
Oleg Borisenko (OB) was responsible for writing the review (and study selection and data extraction in the previous version).
Marjukka Mäkelä (MM), Helena Liira (HL) and John Williams (JW) were responsible for writing the review.

Declarations of interest

None known.

Sources of support

Internal sources

  • Finnish Office for Health Technology Assessment/FinOHTA, National Institute for Health and Welfare/THL, Finland.

External sources

  • No sources of support supplied

Notes

The 2008 update contained substantial changes to the previous version published in 2003 (Williams 2003). There was a near total change in the authors from the previous version. Only two of the previous authors (John W. Williams, Jr. and Marjukka Mäkelä) participated in the 2008 revision.

1. We updated the objectives and added two new objectives:
a) to compare the effect of short versus long courses of antibiotics for acute maxillary sinusitis; and
b) to compare the side effects of different treatments.
We discarded the previous objective of studying the effects of microbial resistance on the efficacy of antibiotics, as it was not feasible to find reliable resistance data on population levels.

2. Several changes were made to the inclusion criteria. This version now also includes participants with a clinical diagnosis of acute maxillary sinusitis without imaging or culture. This is because many of these patients are seen in primary care settings where imaging is not always available and the culture does not provide help to the clinician at the point of decision on whether to prescribe antibiotics or not. Studies with adolescents (at least 12 years old) were also included provided that less than 20% of participants were under 18 years of age. Trials including patients with acute exacerbations of chronic sinusitis were included only if they reported separately the data on the subgroup with acute sinusitis or if at least 80% of participants had acute sinusitis (the acute exacerbation was considered as part of the chronic form of the disease, which is a separate and more complicated entity often related to background factors that have an effect on recovery).

3. The search strategy was amended.

4. Unlike the previous version, in meta-analyses the data from all placebo-controlled studies were combined across antibiotic classes. In the new analyses the clinical failure rate was used as an outcome measure.

5. Risk of bias was assessed for each included study as an additional quality assessment tool.

Compared to the 2003 version of the review the overall clinical conclusions were unchanged.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Adelglass 1998a

MethodsRandomised, single-blind study (investigator). Diagnosis of acute sinusitis confirmed radiographically by at least 5 mm mucosal thickening, opacification or air-fluid level. Outcomes assessed by an investigator; cure defined as disappearance of clinical signs and symptoms without need for further antimicrobial therapy. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 12% on day 16 to 19
ParticipantsThe study setting was not described. Participants were 40 years old on average (range 18 to 83) and included 122 men and 69 women. ENT co-morbidity was assessed; about 50% of patients had hay fever. Country - USA
Interventions2 treatment arms:
Group 1: levofloxacin 500 mg daily for 14 days
Group 2: clarithromycin 500 mg BID for 14 days
Acetaminophen and diphenhydramine were allowed
OutcomesClinical outcomes were assessed in 190 out of 216 randomised patients (in 101 out of 108 patients in levofloxacin group and in 89 out of 108 patients in clarithromycin group). Outcomes were assessed on day 16 to 19. Second follow-up on 42 to 46 days
NotesStudy supported by a pharmaceutical company. 5 of the authors were affiliated to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "The subject randomization list was computer-generated and stratified by site. Equal numbers of subjects were assigned to each treatment group"
Allocation concealment (selection bias)Low riskQuote: "To reduce potential bias during data collection and evaluation of clinical end points, a qualified third party assumed responsibility for all procedures related to the study drugs including dispensing"
Selective reporting (reporting bias)Unclear risk

Outcomes reported: clinical response, adverse effects and quality of life

Comment: in the methods section it is stated that repeat sinus X-rays were taken at 16 to 19 days giving the impression that this was an outcome. However, the results were not reported

Blinding? (Participant and investigator)High riskQuote: "Single-blinded (investigator)"
Incomplete outcome data addressed? (Clinical efficacy outcomes)High risk

Missing data: 7/108 (6%) in levofloxacin group and 19/108 (18%) in clarithromycin group on day 16 to 19

Comment: imbalance in numbers of missing data across groups

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingHigh riskStudy supported by a pharmaceutical company. 5 of the authors were affiliated to a pharmaceutical company

Adelglass 1998b

MethodsRandomised, unblinded trial. Diagnosis by clinical signs and symptoms (mean duration of symptoms prior to enrolment 9 days) and radiograph showing at least 4 mm mucosal thickening. Cure defined as absence of symptoms or a score of 1 or less on a 10-point symptom scale. At least 80% of assigned doses were taken by 94% of participants. Proportion of the participants without known or reported clinical outcome 21% on day 11 to 15
ParticipantsThe study setting was not described. Mean age approximately 37; 128 men and 150 women. About 40% of patients had ENT or eye diseases. Country - USA
Interventions2 treatment arms
Group 1: cefprozil 500 mg twice daily for 10 days
Group 2: amoxicillin-clavulanate (500 mg/125 mg) 3 times daily for 10 days
Mucolytics and decongestants allowed and taken by < 50%
OutcomesClinical outcomes assessed at day 11 to 15 in 219 out of 278 randomised patients (in 108 out of 140 patients in cefprozil group and in 111 out of 138 patients in amoxicillin-clavulanate group). Second follow-up 2 weeks post-treatment. Radiographic and bacteriological outcomes not reported
NotesStudy was supported by a grant from a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open-label study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)High risk

Missing data: 32/140 (22.9%) in cefprozil group and 27/138 (19.6%) in amoxicillin-clavulanate group on day 11 to 15. Total missing data rate 21% on day 11 to 15 and 28% on day 24

Comment: total missing data rate over 20%

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingHigh riskStudy was supported by a grant from a pharmaceutical company

Adelglass 1999

MethodsRandomised, unblinded trial. Diagnosis by clinical symptoms and radiograph showing at least 4 mm mucosal thickening, opacification or air-fluid level. Cure defined as disappearance of symptoms and signs, and stabilisation or improvement of radiographic findings. Treatment compliance not reported. Proportion of the participants included in the clinically evaluable population was 87% on day 12 to 19
ParticipantsParticipants were recruited from community-based primary care and otolaryngology clinics. Mean age 39 (range 18 to 85); 225 men and 390 women. Country - USA
Interventions2 treatment arms
Group 1: levofloxacin 500 mg daily for 10 to 14 days
Group 2: amoxicillin-clavulanate (500 mg/125 mg) 3 times daily for 10 to 14 days
The use of antihistamines and decongestants was encouraged
OutcomesStudy reported clinical responses for clinically evaluable patients only. Clinical outcomes for clinically evaluable population were assessed on day 12 to 19 in 535 out of 615 randomised patients (in 267 out of 307 patients in levofloxacin group and in 268 out of 308 patients in amoxicillin-clavulanate group). (To be clinically evaluable, the patient had to be evaluable for safety, have a confirmed clinical diagnosis of acute sinusitis, have received at least 7 days of therapy, have followed the protocol, and have a clinical evaluation within 2 to 10 days after ending therapy.) Radiographic outcomes not separately reported. Long-term follow-up at 28 to 32 days after completion of therapy to assess relapse
NotesStudy was supported by a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Unclear risk

Outcomes reported: overall clinical response (clinical response + radiographic response) and adverse effects

Comment: prespecified outcomes (in methods) were unclearly reported (3 analyses groups for treatment comparison defined in methods section but the role of mITT and ITT populations in results remains unclear)

Blinding? (Participant and investigator)High riskQuote: "Open-label study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 40/307 (13%) in levofloxacin group and 40/308 (13%) in amoxicillin-clavulanate group on day 12 to 19

Comment: the groups were balanced in numbers and reasons for missing data

Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability at baseline was insufficient to assess whether the groups were comparable at baseline or not (no information on severity rating)
Free of other bias? Co-interventionsLow risk

Quote: "The use of antihistamines and decongestants was encouraged"

Comment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)

Free of other bias? FundingHigh riskStudy was supported by a pharmaceutical company

Arndt 1994

MethodsRandomised, unblinded study. Diagnosis = clinical evidence, an abnormal radiograph and nasal swab. Radiographic criteria not described. Cure defined as "complete resolution of clinical signs/symptoms of infection"; improvement as "clear regression of clinical signs within 4 to 5 days without having entirely disappeared on completion of therapy." Treatment compliance was not reported. Proportion of the participants without known or reported clinical outcome 20%
ParticipantsParticipants were recruited from academic, outpatient otolaryngology clinics. Participants were 33 years old on average (range 18 to 65) and included 25 men and 31 women. Country - Germany
Interventions2 treatment arms
Group 1: brodimoprim 200 mg once daily for 8 to 12 days
Group 2: doxycycline 100 mg once daily for 8 to 12 days
Only non-specific medications were allowed (most patients received concomitant medications, mainly nasal decongestants)
OutcomesClinical and radiographic outcomes were assessed in 56 out of 70 randomised patients. Outcomes were assessed on day 8 to 12 after randomisation. Bacteriological outcomes were reported in 50 patients
Notes1 of the authors was affiliated to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low riskOutcomes reported: clinical, radiological and bacteriological responses; and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)High riskQuote: "Open study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear risk

Quote: "In total 14 of 70 randomized participants had to be excluded shortly after the trial began due to negative culture at baseline or because some of them did not come back after the first visit"

Comment: the missing data rates not reported by study group. It remains unclear how many participants were excluded because of negative culture in each group

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingUnclear risk1 of the authors was affiliated to a pharmaceutical company

Arrieta 2007

MethodsRandomised, unblinded design. Diagnosis by clinical signs and symptoms lasting 7 to 28 days, confirmed by computed tomography or X-ray (criteria not specified). Culture by sinus puncture or endoscopy of middle meatus, or both (51% of participants had at least 1 causative organism identified before study entry). Clinical resolution was defined as total resolution of signs and symptoms related to acute sinusitis and, at least, improvement in the radiographic or computer tomographic scan appearance of the sinuses to the extent that no additional or alternative antimicrobial therapy was necessary. Treatment compliance unclearly reported. Proportion of the participants included in the per-protocol population was 80% on day 17 to 24 (there was no description of missing data)
ParticipantsMulticentre (31 centres), multinational trial (Argentina, Brazil, Chile and Mexico). Demographic data reported for per-protocol population at baseline: 459 outpatients, 172 men and 287 women; mean age was 37 years. ENT co-morbidity was not assessed
Interventions2 treatment arms:
Group 1: moxifloxacin 400 mg once daily for 7 days
Group 2: amoxicillin-clavulanate 500/125 mg 3 times daily for 10 days
Concomitant therapies were not reported
OutcomesStudy reported clinical resolution rates for intention-to-treat and per-protocol populations. However, only per-protocol population could be used as basis for clinical outcome calculating in this review because failure rates were not reported and there was no description of the reasons and amounts of missing data. Clinical outcomes for per-protocol population (patients who completed the course of treatment) were assessed on day 17 to 24 in 459 out of 575 randomised participants (in 226 out of 289 participants in moxifloxacin group and in 233 out of 286 participants in amoxicillin/clavulanate group). Bacteriological outcomes for per-protocol population were assessed in 234 participants
NotesStudy supported by a pharmaceutical company but the authors stated that although they received a grant from the company to conduct the study, results were in no way modified in favour of the sponsor's product. 4 of the authors were affiliated to the same pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low riskOutcomes reported: clinical response (including radiological findings), bacteriological response in 234 participants and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)High riskQuote: "Open-label study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear risk

Missing data: 63/289 (22%) in moxifloxacin group and 53/286 (19%) in amoxicillin/clavulanate group on day 17 to 24

Comment: no description of reasons for missing data

Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingUnclear riskStudy supported by a pharmaceutical company but the authors stated that although they received a grant from the company to conduct the study, results were in no way modified in favour of the sponsor's product. 4 of the authors were affiliated to the same pharmaceutical company

Axelsson 1970

MethodsRandomised trial. Blinding was not reported. Patients with clinically suspected sinusitis were radiologically examined; only patients with secretion were included. Secretion was confirmed by a single diagnostic irrigation for those with completely opaque maxillary sinuses. Clinical outcomes assessed by an investigator but "cure" and "improvement" were not defined. Compliance with treatment was not reported. Proportion of the participants without known or reported clinical outcome 7% on day 10
ParticipantsThe study setting was not described. It is unclear if participants were a convenience sample or a consecutive series of eligible patients. Participants were 33 years old on average (range 13 to 80) and included 62 men and 94 women. Patients with a history of nasal allergy were excluded. Country - Sweden
Interventions4 treatment arms: (only groups 1 and 3 used in the analysis)
Group 1: oxymetazoline only (no placebo control)
Group 2: oxymetazoline plus sinus irrigation
Group 3: oxymetazoline plus penicillin V 400 mg 3 times daily for 10 days
Group 4: oxymetazoline plus lincomycin 500 mg 3 times daily for 8 days
OutcomesClinical outcomes were assessed in 142 out of 156 randomised participants (in 35 out of 38 patients in penicillin group and in 32 out of 34 patients in oxymetazoline group). Outcomes were assessed on day 10 after randomisation. Radiographic outcomes were reported as mean number of severity points per sinus (not by patient)
NotesThere was no identified funding source
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response, radiological response and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Unclear riskNo information provided
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 8% (3/38) in penicillin group and 6% (2/34) in oxymetazoline group

Comment: although the reasons for missing data were not stated, we decided to judge this domain 'low' risk of bias because the missing data rates were equal and marginal in both groups

Free of other bias? Baseline comparabilityLow riskComment: detailed description of radiological sinusitis severity rating at baseline. It appears that the groups were balanced although the information on demographic characteristics was scarce
Free of other bias? Co-interventionsLow riskComment: both groups received the same co-intervention during the trial
Free of other bias? FundingUnclear riskNo information provided

Boezeman 1988

MethodsRandomised, but unblinded trial. Diagnosis by clinical symptoms, confirmed by radiograph showing sinus opacity and positive culture taken near the ostium of, or by aspiration from, maxillary sinus. Clinical outcomes assessed by an investigator but a definition of clinical cure was not given. Treatment compliance was not reported. Proportion of the participants without known or reported clinical outcome 18%
ParticipantsThe study setting was not described. Outpatients were 30 years old on average (range 13 to 76) and included 23 men and 10 women. Country - The Netherlands
Interventions2 treatment arms
Group 1: spiramycin 1000 mg twice daily for 10 days
Group 2: doxycycline 200 mg on day 1 then 100 mg each day for 10 days
Nasal decongestants were allowed but were not given as part of the study treatment
OutcomesClinical outcomes were assessed in 27 out of 33 randomised participants (in 15 out of 18 patients in doxycycline group and in 12 out of 15 patients in spiramycin group). Outcomes were assessed on day 11 after randomisation. Radiographic and bacteriological outcomes were reported for 27 out of 33 patients
NotesThere was no identified funding source
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical, radiological and bacteriological outcomes, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Exclusions from analyses: absence of pathogen in pretreatment cultures 3/15 in spiramycin group and 3/18 in doxycycline group; other missing data 0%

Comment: the absence of pathogen was seen not to cause bias

Free of other bias? Baseline comparabilityHigh riskComment: the severity rating at baseline was different across the groups
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingUnclear riskThere was no identified funding source

Buchanan 2003

MethodsRandomised, double-blind design. Diagnosis by clinical signs and symptoms lasting at least 7 days, confirmed by radiograph showing at least 10 mm mucosal thickening, opacification or air fluid level in a sinus radiograph or CT. Bacteriological specimens were obtained by sinus puncture (in US sites) or endoscopy collection (non-US sites). Clinical cure assessed by investigators and defined as dichotomous - either cure or failure. Cure defined according to the following criteria: return to preinfection state, with no ABMS-related signs and symptoms present, as determined on a scale of 0 to 3 points in which 0 = absent, 1 = mild, and 3 = severe, supplemented by a sinus X-ray/CT scan confirming no worsening of infection or the presence of only those residual symptoms indicative of a normal course of clearance in the infection process, with no requirement for additional antibiotic treatment. Compliance among available patients was in average 95% (assessed by unused capsule count). Proportion of the participants included in the per-protocol clinically evaluable population was 72% on day 16 to 24
ParticipantsMulticentre, multinational trial (USA, Argentina, France, South Africa). 148 men and 208 women, mean age about 40 years (range from 14 to 84)
ENT co-morbidity was not assessed
Interventions2 treatment arms
Group 1: telithromycin 800 mg once daily for 5 days
Group 2: cefuroxime axetil 250 mg twice daily for 10 days
Concomitant treatments were not described
OutcomesStudy reported clinical cure rates for modified intention-to-treat and per-protocol evaluable populations. However, only per-protocol evaluable population could be used as basis for clinical failure rate calculating in this review because true failures were not reported and indeterminate responses were included in failures in mITT analyses. Clinical outcomes for per-protocol evaluable population assessed on day 16 to 24 in 278 out of 385 randomised patients (in 189 out of 260 participants in telithromycin group and in 89 out of 125 participants in cefuroxime axetil group). Per-protocol clinically evaluable population defined as patients who met the diagnostic criteria, and had no major protocol violation. Bacteriologic outcomes assessed on day 16 to 24 in 149 patients. Radiologic outcomes were not reported
NotesStudy supported by a grant from a pharmaceutical company. 1 of the authors was affiliated to pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response (including radiological findings) and bacteriological response, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low risk

Quote: "Double-blind". "To ensure blinding matched placebo capsules were used in telithromycin group"

Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described

Incomplete outcome data addressed? (Clinical efficacy outcomes)High risk

Missing data: 71/260 (27%) in telithromycin group and 36/125 (29%) in cefuroxime axetil group on day 16 to 24

Comment: total missing data rate 28%

Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability of intervention and control groups at baseline reported only for modified intention-to-treat population, not for per-protocol evaluable population which was used in the analyses
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy supported by a grant from a pharmaceutical company. 1 of the authors was affiliated to pharmaceutical company

Burke 1999

MethodsRandomised, double-blind trial. Diagnosis by symptoms lasting 1 to 4 weeks and radiograph showing air-fluid level, opacity or at least 6 mm mucosal thickening. Clinical resolution was defined resolution or improvement of clinical symptoms and radiographic findings and no need for additional antimicrobial treatment. To be included in the efficacy-valid population the compliance had to be at least 80%. Proportion of the participants without known or reported clinical outcome 16%
ParticipantsMixed otolaryngology and primary care. Mean age 40; 178 men, 279 women. Country - USA
Interventions

2 treatment arms
Group 1: cefuroxime axetil 250 mg twice daily for 10 days
Group 2: moxifloxacin 400 mg daily for 10 days

Decongestants, antihistamines and phenylephrine drops were allowed; systemic or topical corticosteroids not allowed

OutcomesStudy reported clinical cure rates for intention-to-treat and efficacy-valid populations but clinical failure rates only for efficacy-valid population. Clinical outcomes were assessed on day 17 to 24 in 457 out of 542 randomised patients (in 234 out of 275 participants in cefuroxime axetil group and in 223 out of 267 participants in moxifloxacin group). The efficacy-valid population defined as patients who met diagnostic criteria and were without major protocol violations. Radiographic outcomes not reported. Long-term follow-up on 37 to 41 days
NotesStudy supported by a research grant from a pharmaceutical company. 3 of the authors were affiliated to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Block-design random code"
Comment: although the description of the sequence generation was incomplete, this domain was graded 'low' risk of bias because description gives an impression of computer-based method
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low riskOutcomes reported: clinical response (including radiological findings) and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)Low risk

Quote: "Double-blind study". "Medications were encapsulated in gelatin for masking purposes. Placebo capsules used in moxifloxacin group"

Comment: although the description of the blinding was incomplete (regarding blinding of investigator), this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described

Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 41/275 (15%) in cefuroxime axetil group and 44/267 (16%) in moxifloxacin group on day 17 to 24

Comment: the groups were balanced in numbers and reasons for missing data

Free of other bias? Baseline comparabilityLow riskComment: although the proportion of patients with severe sinusitis was 47/223 (21%) in moxifloxacin group and 36/234 (15%) in cefuroxime axetil group, this domain was graded 'low' risk of bias
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingHigh riskStudy supported by a research grant from a pharmaceutical company. 3 of the authors were affiliated to a pharmaceutical company

Calhoun 1993

MethodsRandomised, single-blind trial (investigator). Diagnosis established by clinical symptoms and abnormal radiograph (criteria not reported). Clinical outcomes were assessed by an investigator. Clinical cure defined as signs and symptoms of infection resolved. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 18%
ParticipantsConsecutive patients were recruited from academic, primary care clinics. Mean age was 37 (range 14 to 77); 84 women and 58 men. ENT co-morbidity was not assessed. Country - USA
Interventions2 treatment arms
Group 1: clarithromycin 500 mg twice daily for 7 to 14 days
Group 2: amoxicillin 500 mg 3 times daily for 7 to 14 days
Both groups received oxymetazoline 2 sprays twice daily for 3 days
OutcomesClinical outcomes were assessed in 116 out of 142 randomised participants (in 55 out of 70 patients in clarithromycin group and in 61 out of 72 patients in amoxicillin group). Outcomes were assessed on day 7 to 16 after randomisation. Radiographic outcomes were assessed in 116 out of 142 patients randomised. Bacteriological outcomes were not reported
NotesStudy was supported in part by a grant from a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low riskOutcomes reported: clinical and radiological responses, and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)High riskQuote: "A third party dispensed the antibiotics and assessed compliance, ensuring the single-blind (investigator-blind) nature of the study. The single-blind design was necessary because of the difficulty in packaging the two drugs in identical capsules. Patients were instructed not to discuss the dosing schedule with their physicians"
Comment: no blinding of the participants
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear risk

Missing data: 15/70 (21.4%) in clarithromycin group and 11/72 (15.3%) in amoxicillin group

Comment: all reasons for missing data were not reported by group

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: both groups received the same co-intervention during the trial
Free of other bias? FundingHigh riskStudy was supported in part by a grant from a pharmaceutical company

Camacho 1992

MethodsRandomised, single-blind design (investigator). Clinical diagnosis of acute sinusitis and an abnormal radiograph (mucosal thickening, opacity or air-fluid level). Sinus puncture performed. Clinical outcomes assessed by an investigator; cure defined as "resolution of clinical symptoms with radiographic evidence of sinus decongestion with no further therapy required". Improvement defined as "clinical symptoms substantially reduced but radiographic evidence of residual sinus congestion". Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 25% on day 11 to 38
ParticipantsPatients were recruited from academic and community settings; probably otolaryngology practices. Mean age = 35 (range 18 to 79); 171 men and 146 women. ENT co-morbidity was not assessed. Countries - Brazil, Chile, Columbia, USA
Interventions2 treatment arms
Group 1: cefuroxime axetil 250 mg twice daily for 10 days
Group 2: amoxicillin-clavulanate 500 mg 3 times daily for 10 days
Concomitant therapies were not described
OutcomesClinical outcomes were assessed on day 11 to 38 in 239 out of 317 randomised participants (in 115 out of 157 participants in cefuroxime axetil group and in 124 out of 160 participants in amoxicillin-clavulanate group). Radiographic outcomes were not reported. Bacteriological outcomes were assessed in 76 participants
NotesStudy was supported in part by a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Computer-generated randomized code"
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low riskOutcomes reported: clinical response, bacteriological response for a proportion of participants, adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)High riskQuote: "Investigator-blinded"
Incomplete outcome data addressed? (Clinical efficacy outcomes)High riskMissing data: 42/157 (27%) in cefuroxime axetil group and 36/160 (23%) in amoxicillin-clavulanate group on day 11 to 38
Comment: total missing data over 20%
Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability was insufficient to assess whether the groups were comparable at baseline or not
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy was supported in part by a pharmaceutical company

Chatzimanolis 1998

MethodsRandomised, unblinded design. Diagnosis of acute or recurrent sinusitis documented by clinical and endoscopic findings, bacteriology (sinus puncture) and X-ray. Clinical cure not defined. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 7%
ParticipantsStudy setting not described. Mean age 38; 34 men and 26 women. ENT co-morbidity not assessed. Country - Greece
Interventions2 treatment arms
Group 1: roxithromycin 150 mg twice daily for 10 to 14 days
Group 2: amoxicillin-clavulanate (500 mg/ 125 mg) 3 times daily for 10 to 15 days
Concomitant treatments not reported
OutcomesClinical outcomes reported on day 10 to 17 for 56 out of 60 randomised participants (for 29 out of 31 participants in roxithromycin group and for 27 out of 29 participants in amoxicillin-clavulanate group). Radiographic outcomes not reported by group
NotesStudy funded by a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low riskMissing data: 2/31 (6.5%) in roxithromycin group and 2/29 (7%) in amoxicillin-clavulanate group
Comment: the groups were balanced in numbers and reasons for missing data
Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy funded by a pharmaceutical company

Clement 1998

MethodsRandomised, unblinded design. Clinical diagnosis of ethmoid or maxillary sinusitis confirmed by fibre-optic examination (purulent discharge from ostium of affected sinus) and computed tomography (criteria not specified). Pus samples were collected endoscopically for bacteriological culture. Clinical outcomes assessed by an investigator but were not defined. Compliance with treatment was not reported. Proportion of the participants without known or reported clinical outcome 8% on day 10 to 14
ParticipantsParticipants were recruited from otolaryngology clinics. Mean age 41; 103 men, 151 women. ENT co-morbidity not assessed. Country - Belgium
Interventions2 treatment arms
Group 1: azithromycin 500 mg daily for 3 days
Group 2: amoxicillin 500 mg/clavulanic acid 125 mg 3 times daily for 10 days
Concomitant therapies (decongestants, corticosteroids, mucolytics) were allowed but not prescribed. 133 participants used a concomitant treatment
OutcomesClinical outcomes were assessed on day 10 to 14 in 233 of 254 participants (in 151 out of 165 participants in azithromycin group and in 82 out of 89 participants in amoxicillin-clavulanate group) and on day 21 to 28 in 210 of 254 participants. Bacteriological outcomes were assessed in 100 participants
NotesStudy funding provided by a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response, bacteriological response for a proportion of participants and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low riskMissing data: 14/165 (8.5%) in azithromycin group and 7/89 (7.9%) in amoxicillin-clavulanate group on day 10 to 14. Total missing data 8% on day 10 to 14 and 17% on day 21 to 28
Comment: the groups appeared to be balanced in numbers and reasons for missing data. Available case data given for both follow-ups
Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsUnclear riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed. Corticosteroids may have effect on recovery but their use was not reported
Free of other bias? FundingHigh riskStudy funding provided by a pharmaceutical company

Clifford 1999

MethodsRandomised, double-blind trial. Participants with acute sinusitis and acute exacerbation of chronic sinusitis included. Proportion of the participants with acute sinusitis 75% and their results reported separately. Clinical diagnosis, confirmed by radiograph showing air-fluid level, opacification or at least 6 mm mucosal thickening. Cure defined as symptom resolution and radiographic improvement. Proportion of the participants without known or reported clinical outcome 18% for the whole study population including participants with acute sinusitis and participants with acute exacerbation of chronic sinusitis
ParticipantsStudy setting not described. Mean age approximately 41 (range 18 to 76); 187 men and 270 women. 25% of participants with acute exacerbation of chronic sinusitis in efficacy-valid population. About 26% of participants had allergic rhinitis. Country - USA
Interventions2 treatment arms
Group 1: ciprofloxacin 500 mg twice daily for 10 days
Group 2: clarithromycin 500 mg twice daily for 14 days
Decongestants allowed. Nasal corticosteroids prohibited
OutcomesClinical outcomes for participants with acute maxillary sinusitis were separately reported only for efficacy-valid population (342 participants at day 14). Efficacy-valid population was defined as participants who completed the course of treatment and did not use other antimicrobial agents concomitantly with the study drug. Radiographic outcomes not reported
NotesStudy supported by a research grant from a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "The patients were randomly assigned to 1 of 2 treatment groups by means of a block design random code computer-generated"
Allocation concealment (selection bias)Low riskComment: central allocation (randomisation made at a pharmaceutical company)
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response (including radiographic findings) and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low risk

Quote: "Double-blind". "The study drugs were encapsulated in gelatin capsules for blinding purposes. Matched placebos used"

Comment: although the description of the blinding was incomplete (in respect of investigator-blinding), this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described

Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear riskComment: missing data not described separately for the participants with acute sinusitis and participants with acute exacerbation of chronic sinusitis
Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingHigh riskStudy supported by a research grant from a pharmaceutical company

Desrosiers 2008

MethodsRandomised, unblinded trial. Diagnosis by clinical signs and symptoms lasting for 7 to 28 days (purulent nasal discharge; 1 additional major sign and symptom (facial pain/pressure/tightness over the maxillary sinuses; nasal congestion/obstruction, hyposmia/anosmia; fever) or 2 minor signs and symptoms (headache, halitosis, dental pain, ear pressure/fullness, cough, fatigue)), confirmed by abnormal maxillary sinus X-rays or limited sinus computed tomography scans or sinus ultrasound (at least 1 of the following: air/fluid level, total opacification, mucosal thickening at least 10 mm). Samples for microbiological analysis were taken by using rhinoscopic aspiration or micro-swabbing of the maxillary sinus (middle meatus). Clinical outcome was defined as "success", "failure" or "indeterminate" (success was defined as cured or improved). Clinical failure was recorded if at least on the following criteria was met: all signs/symptoms remained unchanged or had worsened, at least 1 additional antibiotic had to be prescribed and/or added to the study treatment for sinusitis, sinusitis-related complications occurred, sinus puncture or drainage had been performed. Treatment compliance at least 80% for those patients who were included in the per-protocol population. Proportion of the participants included in the per-protocol population was 83% on day 17 to 21
ParticipantsMulticentre (41 centres), multinational trial (Canada, Germany, Greece, Portugal and Turkey). Demographic data reported for modified intention-to-treat population at baseline: participants' mean age was 39 years; 109 men and 181 women. ENT co-morbidity was assessed; 1 participant had allergic rhinitis, 2 participants had chronic sinusitis and 5 participants had nasal septal deviation
Interventions2 treatment arms
Group 1: telithromycin 800 mg once daily for 5 days
Group 2: amoxicillin-clavulanate 875/125 mg twice daily for 10 days
Concomitant therapies were not clearly reported
OutcomesStudy reported clinical cure rates for modified intention-to-treat and per-protocol populations. However, only per-protocol population could be used as basis for clinical failure rate calculating in this review because true failures were not reported and it seemed that indeterminate and missing responses were included in failures in mITT analyses. Clinical outcomes for per-protocol population were assessed in 248 of 298 randomised participants on day 17 to 21 and on day 41 to 49. Per-protocol population was defined as all randomised participants with clinically and radiographically confirmed sinusitis, and were without major protocol violations. Bacteriological outcomes reported for 103 participants on day 17 to 21. Health-related quality of life measured by SF-36 questionnaire completed by 278 participants at baseline, and on day 41 to 49
NotesStudy funding provided by a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Centrally generated randomization list"
Allocation concealment (selection bias)Low riskQuote: "Centrally generated randomization list"
Selective reporting (reporting bias)Low riskOutcomes reported: clinical response, bacteriological response for a proportion of participants, quality of life and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)High riskQuote: "Open-label study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

8 out of 298 randomised participants were excluded at the beginning of the study because of the unconfirmed sinusitis. No information provided how these participants were divided across the 2 groups. Calculating missing data by group was therefore based on the modified intention-to-treat (mITT) population (290 participants) which was defined as all randomised participants with clinically and radiographically confirmed sinusitis, and who received at least 1 dose of study medication

Missing data: 21/144 (14.6%) in telithromycin group and 21/146 (14.4%) in amoxicillin-clavulanate group on day 17 to 21
Comment: groups appeared to be balanced in numbers and reasons for missing data

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsUnclear riskConcomitant therapies were not clearly reported
Free of other bias? FundingHigh riskStudy funding provided by a pharmaceutical company

Dubois 1993

MethodsRandomised, single-blind (investigator) trial. Diagnosis established by symptoms, an abnormal maxillary radiograph (minimum criteria were not given) and a positive culture of sinus fluid (collected by puncture or endoscopically). Clinical outcomes assessed by an investigator; "cure" defined as pretreatment signs and symptoms resolved; "improvement" defined as improved but not resolved." Treatment compliance not reported but participants had to take at least 70% of prescribed treatment to be included in the efficacy analysis. Total proportion of the participants without known clinical outcome 48%. Patients were excluded mostly due to negative culture; exclusions because of other reasons marginal 3.4%
ParticipantsParticipants were recruited from outpatient ENT and primary care settings. It is unclear if the study sample represents a consecutive series of participants or a convenience sample. Country - Canada
Interventions2 treatment arms
Group 1: clarithromycin 500 mg every 12 hours for a maximum of 14 days
Group 2: amoxicillin-clavulanate 500 mg every 8 hours for a maximum of 14 days
Both groups received oxymetazoline 0.05% nasal sprays twice daily for the first 3 days
OutcomesClinical, radiographic and bacteriological outcomes were assessed in 260 out of 497 randomise participants (in 132 out of 246 participants in clarithromycin group and in 128 out of 251 participants in amoxicillin-clavulanate group). Outcomes were assessed on about day 15 after randomisation
NotesStudy was supported in part by a grant from a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical, radiological and bacteriological responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High risk

Quote: "Investigators were blind to the drug being administered"

Comment: no description of blinding of the participants

Incomplete outcome data addressed? (Clinical efficacy outcomes)Low riskExclusions from analyses: absence of pathogen in pretreatment cultures 103/246 in clarithromycin group and 117/251 in amoxicillin-clavulanate group; other exclusions 10 (4.1%) and 7 (2.8%) of participants in clarithromycin group and amoxicillin-clavulanate group, respectively
Comment: the absence of pathogen was seen not to cause bias. Other missing data marginal
Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: both groups received the same co-intervention during the first 3 days
Free of other bias? FundingHigh riskStudy was supported in part by a grant from a pharmaceutical company

Ferguson 2004

MethodsRandomised, double-blind study. Diagnosis by clinical signs and symptoms, confirmed by radiograph showing at least 10 mm mucosal thickening, total sinus opacity or air-fluid level in a sinus radiograph. Pretreatment cultures of rhinoscopic aspirations or deep nasal Calgiswab samples were compared with cultures taken at 17 to 24 days of follow-up. Clinical success was defined: the patient was either cured or showed improvement relative to baseline (including no worsening of radiological findings) without addition of a new antibiotic. Treatment compliance reported as over 90% assessed by capsule counts. Proportion of the participants included in the per-protocol population was 78% on day 17 to 24
ParticipantsMulticentre study, 41 sites in USA. Median age approximately 45 (range 18 to 85); 123 men and 199 women. ENT co-morbidity not described
Interventions2 treatment arms
Group 1: telithromycin 800 mg once daily for 5 days
Group 2: moxifloxacin 400 mg once daily for 10 days
OutcomesStudy reported clinical success rates for modified intention-to-treat and per-protocol populations. However, only per-protocol population could be used as basis for clinical response calculating in this review because true failures were not reported and all instances of not completing the study were included in failures in mITT analyses. Clinical outcomes for per-protocol population were assessed at days 17 to 24 in 272 of 349 randomised participants (in 135 out of 173 participants in telithromycin group and in 137 out of 176 participants in moxifloxacin group). The per-protocol population was defined as all randomised participants who received at least 1 dose of study medication and had signs and symptoms of acute maxillary sinusitis and radiologic findings supporting the diagnosis and were without major protocol violations. Long-term follow-up at 31 to 36 days. Bacteriological outcomes were assessed in 67 participants
NotesStudy supported by a grant from a pharmaceutical company. 4 of the authors had received research support/grants from and/or acted as consultants for the same pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response (including radiologic findings), bacteriological response in 67 participants and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low risk

Quote: "Double-blind study". "Placebo-capsules used"

Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described

Incomplete outcome data addressed? (Clinical efficacy outcomes)High risk

Missing data: 38/173 (22%) in telithromycin group and 39/176 (22%) in moxifloxacin group on day 17 to 24

Comment: missing data over 20%, insufficient description of reasons for missing data

Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability was insufficient to assess whether the groups were comparable at baseline or not (information given for mITT population, not for per-protocol population)
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy supported by a grant from a pharmaceutical company. 4 of the authors had received research support/grants from and/or acted as consultants for the same pharmaceutical company

Garbutt 2012

Methods

Randomised, placebo-controlled trial. Clinical diagnosis of acute sinusitis was based on clinical findings, with history of maxillary pain or tenderness in the face or teeth, purulent nasal secretions, and rhinosinusitis symptoms for 7 days or more and 28 days or less that were not improving or worsening, or rhinosinusitis symptoms lasting for less than 7 days that had significantly worsened after initial improvement. The symptoms had to be moderate, severe or very severe. Mean duration of symptoms was 11 days in both groups (median 10.0, interquartile range (25th to 75th percentile) 7.0 to 14.0). Because the proportion of the participants with symptoms less than 7 days was small (14%), and because the reason to include those patients was that patients’ symptoms had significantly worsened, this study was included in this review.

Clinical outcomes were assessed by telephone interview using a structured questionnaire. Participants used a 6-point scale (a lot or a little worse or better, the same or no symptoms) to retrospectively assess symptom change since enrolment. Those reporting their symptoms as a lot better or absent were categorised as significantly improved. In total, 23 participants (14%) (11 in the amoxicillin group and 12 in placebo group) did not complete the 10-day course of treatment. Sixteen were treated with another antimicrobial (5 in the amoxicillin group and 11 in the placebo group). Proportion of the participants without known or reported clinical outcome 7% on day 7; and 8% on day 10; and 4% on day 28.

Follow-up time chosen for analyses of this review was 10 days. The 7-day follow-up was not used because the treatments lasted for 10 days. The 28 day follow-up was not considered because the outcome measure at this time point was recurrence or relapse and did not include all failures, and thus the outcome is not in accordance with the outcome definition of this review.

The study was not considered in the primary analysis of this review because it remained unclear whether the patients who received rescue therapy were classified as non-improved or not.

The results, as reported in the original article, are presented in the Results section.

ParticipantsParticipants recruited from 10 offices of primary care physicians. 60 men and 106 women; mean age 32 (range 18 to 69) years. Allergic rhinitis in 32%, asthma in 11% and history of sinus disease in 74% of participants in each group. Smokers 13% in amoxicillin and 26% in placebo group. Country - USA
Interventions2 treatment arms
Group 1: amoxicillin 500 mg 3 times daily for 10 days
Group 2: placebo 3 times daily for 10 days

Concomitant therapies
All patients received a 5 to 7-day supply of the following symptomatic treatments to be used as needed: acetaminophen for pain or fever at a dose of 500 mg every 6 hours, guaifenesin to thin secretions at a dose of 600 mg every 12 hours, 10 mg/5ml of dextromethorphan hydrobromide and 100 mg/5 ml of guaifenesin for cough at a dose of 10 ml every 4 to 6 hours, pseudoephedrine-sustained action for nasal congestion at a dose of 120 mg every 12 hours, and 0.65% saline spray using 2 puffs per nostril as needed
Concurrent use of symptomatic treatments was 92% and did not vary by study group. No new nasal steroid use was reported
Outcomes

Clinical outcomes were assessed on day 7 in 155 of 166 randomised participants (in 80 out of 85 participants in amoxicillin group and in 75 out of 81 participants in placebo group), on day 10 in 152 of 166 randomised participants (in 81 out of 85 participants in amoxicillin group and in 71 out of 81 participants in placebo group), and on day 28 in 159 of 166 randomised participants (in 82 out of 85 participants in amoxicillin group and in 77 out of 81 participants in placebo group)

Outcomes were assessed by telephone interview using a structured questionnaire. Participants used a 6-point scale (a lot or a little worse or better, the same or no symptoms) to retrospectively assess symptom change since enrolment
Analyses were based on the intention-to-treat principle (all of the eligible participants as randomised)

NotesStudy supported by a grant from the National Institute of Allergy and Infectious Diseases, which did not have a role in the design and conduct of the study, management and analysis of data, or in the preparation, review or approval of the manuscript
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Using a blocked randomization scheme, computer-generated random numbers were used to determine how the 2 study drugs were allocated to the consecutively numbered study treatment packages"
Allocation concealment (selection bias)Low riskQuote: "Randomisation was performed in advance by the investigational pharmacist who did not participate in patient enrolment or outcome assessment. Randomisation occurred when the research assistant assigned the treatment package"
Selective reporting (reporting bias)Low riskOutcomes reported: clinical responses, quality of life and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)Low riskQuote: "Placebo was similar in appearance and taste and dispensed in the same fashion than the antibiotic. Telephone interviews were conducted by trained research assistants blinded to group assignment"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear riskMissing data: in amoxicillin group 5/85 (5.9%) and in placebo group 6/81 (7.4%) on day 7; 4/85 (4.7%) and 10/81(12.3%) on day 10; and 3/85 (3.5%) and 4/81 (4.9%) on day 28, respectively
Total missing data rates 6.6% on day 7; and 8.4% on day 10; and 4.2% on day 28

Comment: although the authors gave additional information that participants simply missed the interviews and there was probably not any systematic difference in reasons for drop-outs between the groups, we graded this domain 'unclear' risk of bias at day 10. Although the difference in drop-out rates between groups was not statistically significant (P value 0.077), difference in proportions of missing outcomes (7.6%) compared to the difference of anticipated failure risks (10% to 20%) may be big enough to cause bias
Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Concurrent use of symptomatic treatments was 92% and did not vary by study group. No new nasal steroid use was reported
Free of other bias? FundingLow riskComment: no pharmaceutical company funding

Gehanno 1996a

MethodsRandomised, unblinded study. Diagnosis established by clinical symptoms and radiograph showing sinus opacity or air-fluid level. Culture based on the sample from the middle meatus; 79% of participants had positive cultures. Clinical outcomes assessed by investigator; cure defined as resolution of facial pain, purulent discharge, air-fluid level by radiograph and no recurrence of symptoms. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 7%
ParticipantsMulticentre study, 284 outpatients; 136 men and 148 women. Mean age was 40. Fewer than 25% of participants had concurrent allergic disease. Country - France
Interventions2 treatment arms
Group 1: clarithromycin 500 mg daily for 8 days
Group 2: amoxicillin-clavulanic acid 500 mg 3 times daily for 8 days
Intranasal decongestants were allowed
OutcomesClinical outcomes were assessed on about day 10 in 263 out of 284 randomised participants (in 134 out of 145 participants in clarithromycin group and in 129 out of 139 participants in amoxicillin-clavulanate group). Radiographic and bacteriological outcomes were not reported
Notes2 of the authors were affiliated to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskCentral randomisation (computer-generated?)
Allocation concealment (selection bias)Low riskCentral random allocation
Selective reporting (reporting bias)Low riskOutcomes reported: clinical response and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)High riskQuote: "Open study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 11/145 (7.6%) in clarithromycin group and 10/139 (7.2%) in amoxicillin-clavulanate group on day 10

Comment: the groups appeared to be balanced

Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingUnclear risk2 of the authors were affiliated to a pharmaceutical company

Gehanno 1996b

MethodsRandomised, double-blind study. Diagnosis by clinical signs and symptoms, confirmed by radiograph (criteria not reported) and/or bacteriological culture of the middle meatus aspirate (76% of participants had a positive culture). Clinical success defined as resolution of all signs and symptoms. Treatment compliance was not reported. Total proportion of the participants without known or reported clinical outcome 21%. Patients were excluded mostly due to normal sinus X-ray and a negative bacteriological culture at inclusion; exclusions because of other reasons 6%
ParticipantsMulticentre study. 376 outpatients; 160 men and 216 women, mean age about 41 years. ENT co-morbidity was not assessed. Country - France
Interventions2 treatment arms
Group 1: sparfloxacin 200 mg for 5 days
Group 2: cefuroxime axetil 250 mg twice daily for 8 days
OutcomesClinical outcomes were assessed on days 10 to 12 in 302 of 382 randomised participants (in 153 out of 193 participants in the sparfloxacin group and in 149 out of 189 participants in the cefuroxime axetil group)
NotesThere was no identified funding source. However, 2 pharmaceutical companies were mentioned in the methods section
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Low riskCentral allocation by a pharmaceutical company
Selective reporting (reporting bias)Low riskOutcomes reported: clinical response, overall response (included clinical, radiological and bacteriological efficacy) and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)Low risk

Quote: "Double-blind study". "Blinding was prepared by a pharmaceutical company and maintained throughout the study. Placebo-tablets used"

Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described

Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Exclusions from analyses on day 10 to 12: normal sinus X-ray and a negative bacteriological culture at inclusion 28/193 in the sparfloxacin group and 29/189 in the cefuroxime axetil group; other exclusions 12 (6%) and 11 (6%) of participants in the sparfloxacin and cefuroxime axetil group, respectively

Comment: exclusions because of lack of inclusion criteria were seen not to cause bias. The groups appeared to be balanced in numbers and reasons for missing data

Free of other bias? Baseline comparabilityUnclear riskComment: detailed description of demographic characteristics and sinusitis severity rating at baseline to assess the comparability of the groups for whole randomised population, not separately for the efficacy evaluable population
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingUnclear riskThere was no identified funding source. However, 2 pharmaceutical companies were mentioned in the methods section

Gehanno 1998

MethodsRandomised, unblinded trial. Diagnosis by symptoms and radiograph showing > 5 mm mucosal thickening, sinus opacity or air-fluid level. A causative pathogen was isolated in the pus sample at study entry, in 51% of the 241 participants enrolled. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 2% on day 10 to 12
ParticipantsParticipants recruited from community-based otolaryngology clinics. Mean age 42; 105 men and 136 women. Country - France
Interventions2 treatment arms
Group 1: cefatrizine 500 mg twice daily for 10 days
Group 2: amoxicillin-clavulanate 500 mg 3 times daily for 10 days
Vasoconstrictors and paracetamol allowed
OutcomesClinical outcomes assessed on days 10 to 12 in 236 out of 241 randomised participants (in 121 out of 123 participants in cefatrizine group and in 115 out of 118 participants in amoxicillin-clavulanate group). Second follow-up on day 30. Radiographic outcomes assessed in 214. Bacteriologic outcomes not reported
Notes1 of the authors was affiliated to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Central random allocation" (computer-generated?)
Allocation concealment (selection bias)Low riskQuote: "Central random allocation"
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and radiographic responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open-label study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 2/123 (1.6%) in cefatrizine group and 3/118 (2.5%) in amoxicillin-clavulanate group on days 10 to 12. Total missing data rate 2% on day 10 to 12 and 4.5% at day 30

Comment: marginal missing data rate

Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingUnclear risk1 of the authors was affiliated to a pharmaceutical company

Gehanno 2004

MethodsRandomised, double-blind design. Diagnosis by clinical signs and symptoms, confirmed by radiograph showing at least 5 mm mucosal thickening, opacity or air-fluid level. A rhinoscopic bacteriologic sampling was made on the middle nasal meatus (cultures were positive in 46% of participants). Clinical cure based on resolution of symptoms. Treatment compliance for per-protocol population 98%. Proportion of the participants without known or reported clinical outcome 11%
ParticipantsParticipants were recruited by otorhinolaryngologists. The mean age of the participants was 43 years; women 60% and men 40%. Multinational trial (France, Tunisia, Poland, Argentina)
Interventions

2 treatment arms
Group 1: pristinamycin 1 g BID for 4 days
Group 2: cefuroxime axetil 250 mg BID for 5 days

Topical vasoconstrictors prescribed for 3 days. Anti-inflammatory analgesic allowed. Corticosteroids and mucolytics prohibited

OutcomesStudy reported clinical cure rates for modified intention-to-treat and per-protocol populations. However, only per-protocol population could be used as basis for clinical failure rate calculating in this review because true failures were not reported in mITT population. Clinical outcomes were assessed on day 12 to 19 in 434 out of 485 randomised participants (in 220 out of 250 participants in pristinamycin group and in 214 out of participants 235 in cefuroxime axetil group). The per-protocol population was defined as participants who met the diagnostic criteria (radiologically confirmed) and had no major protocol violation. Bacteriological outcomes reported in 199
NotesStudy was supported by a pharmaceutical company. 3 of the authors were affiliated with the same pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskComment: although the description of the sequence generation was incomplete, this domain was graded 'low' risk of bias because description gives an impression of computer-based method
Allocation concealment (selection bias)Low riskQuote: "Central allocation in blocks of 4"
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical responses, bacteriological response in 199 and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind study". "Placebo-tablets used"
Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 30/250 (12%) in pristinamycin group and 21/235 (9%) in cefuroxime axetil group on day 12 to 19

Comment: the groups appeared to be balanced in numbers and reasons for missing data

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingHigh riskStudy was supported by a pharmaceutical company. 3 of the authors were affiliated with the same pharmaceutical company

Gwaltney 1997

MethodsRandomised, investigator-blind design. Clinical diagnosis confirmed by radiograph (criteria not specified). Bacterial cultures obtained by sinus aspirations. Cure and improvement reported as combined outcome. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 19% on day 7 to 14
ParticipantsStudy setting not described. Median age 35; 789 men and 1009 women. ENT co-morbidity not described. Countries - USA and Europe
Interventions3 treatment arms
Group 1: cefdinir 600 mg once daily for 10 days
Group 2: cefdinir 300 mg twice daily for 10 days
Group 3: amoxicillin-clavulanate (500 mg) 3 times daily for 10 days
Concomitant treatments not reported
Group 1 and Group 3 used in the analysis
OutcomesClinical outcomes reported at day 7 to 14 in 1446 out of 1798 people randomised (in 474 out of 585 participants in cefdinir 600 mg once daily group and in 491 out of 603 participants in amoxicillin-clavulanate group). Long-term follow-up at 3 to 5 weeks for participants cured or improved at the first follow-up. Radiographic outcomes not reported. Bacteriological outcomes not reported by group
Notes3 of the authors were affiliated with a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Low riskQuote: "Medications were dispensed by a third party, and all records regarding study medication were kept at a separate site"
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and bacteriological responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Investigator-blind study". "Patients were instructed to with-hold details of study medication appearance and dosing schedule"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 111/585 (19%) in cefdinir 600 mg once daily group and 112/603 (19%) in amoxicillin-clavulanate group on day 7 to 14

Quote: "The 3 groups were comparable in the reasons for their non-evaluability"
Comment: the groups appeared to be balanced in numbers and reasons for missing data on day 7 to 14

Additional missing data at 3 to 5 weeks compared to the first follow-up about 12%. Data not used in the long-term follow-up meta-analysis

Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability was insufficient to assess whether the groups were comparable at baseline or not
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingUnclear risk3 of the authors were affiliated with a pharmaceutical company

Hadley 2010

Methods

Randomised (2:1), double-blind, multicentre, phase III trial. Diagnosis by clinical signs and symptoms present for at least 7 days (but less than 28 days) as defined by radiographic and clinical criteria. Radiographic criteria included the presence of air-fluid levels and/or opacification on a radiographic paranasal sinus film. Eligible patients had to be at least 18 years of age and had to have 2 major symptoms (purulent anterior or posterior nasal discharge and unilateral facial pain or malar tenderness) or at least 1 major and 1 minor symptom (frontal headache or fever). A sinus puncture for bacterial culture of the sinuses was performed; only culture positive patients with target pathogen (S. pneumoniae, H. influenzae, M. catarrhalis, Streptococcus pyogenes or Staphylococcus aureus) were included in the analyses.

Clinical cure was defined as resolution or improvement in the signs and symptoms such that no further therapy (antimicrobial, corticosteroid or irrigation) was required. Patients who received rescue therapy were deemed clinical failures. Compliance not reported for modified intention-to-treat population (which was used in the primary analyses) but in the per-protocol population 80% compliance was required. Total proportion of the participants without known clinical outcome 68%. Patients were excluded mostly due to negative culture or non-target pathogen in it; exclusions because of other reasons 6% on day 6 to 8

The study was not considered in the primary analysis because the way of reporting left the exact numbers of failures somewhat unclear. Further, the data were not seen to be comparable with the data from the other trials (the clinical response was measured already at days 6 to 8 after start of therapy while other studies included in the meta-analyses did it at days 10 to 14; unlike the other placebo-controlled trials, in this study isolation of pathogenic bacteria from the sinus secretion was a prerequisite for inclusion)

The results at day 6 to 8, as reported in the original article, are presented in the Results section

The study was, however, included in the analysis of "clinical failure defined as a lack of full recovery or improvement at 16 to 60 days follow-up" (but not pooled)

ParticipantsParticipants recruited from 37 centres (ear, nose and throat practices, family practitioners and general medical clinics). Mean age 39 years; 45 men and 73 women (in modified intention-to-treat population; 118 out of 374 randomised). ENT co-morbidity was not assessed. Country - USA
Interventions2 treatment arms
Group 1: moxifloxacin 400 mg once daily for 5 days
Group 2: placebo once daily for 5 days

The following concomitant medications were allowed: oral or nasal decongestants or antihistamines, nasal sprays and nasal inhalants containing corticosteroids or corticosteroids used as broncho-inhalants if the patient had been on a stable dose for over 4 weeks prior to enrolment and there was no increase of their dose during the study period
OutcomesThe modified intention-to-treat population was the main analysis population and consisted of all patients who received at least 1 dose of study drug and whose initial quantitative culture was positive. Clinical outcomes were assessed on day 6 to 8 and on day 20 to 26 in 118 of 374 randomised participants (in 73 out of 251 participants in moxifloxacin group and in 45 out of 123 participants in placebo group). No radiological or bacteriological outcomes
NotesStudy supported by a pharmaceutical company. 3 of the 6 authors were employees of the same pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskQuote: "multicenter, placebo-controlled, 2:1 randomized, double-blind, phase IIIb clinical trial"

Comment: no detailed information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response during therapy and at follow-up, patient-reported symptom improvement, questionnaire scores, assessment of activity impairment via questionnaire and safety

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Unclear riskQuote: "Double-blinded trial, matching placebos"

Comment: no detailed information provided
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Exclusions from analyses: negative culture or non-target pathogen 161/251 (64%) in moxifloxacin group and 72/123 (58.5%) in placebo group; other exclusions 17 (6.8%) and 6 (4.9%) of participants in moxifloxacin group and placebo group, respectively, on day 6 to 8 and on day 20 to 26

Comment: non-target pathogen or absence of pathogen was seen not to cause bias. Other missing data were balanced in numbers and reasons across the groups

Free of other bias? Baseline comparabilityUnclear riskComment: in the placebo group there were 9% more bilateral sinus infections than in the moxifloxacin group (38% versus 29%) and no by-group description was provided on allowed concomitant medications (especially stable-use nasal corticosteroids)
Free of other bias? Co-interventionsUnclear riskComment: no detailed description of the use of co-interventions (especially nasal corticosteroids which may have also curative effect)
Free of other bias? FundingHigh riskStudy supported by a pharmaceutical company. 3 of the 6 authors were employees of the same pharmaceutical company

Haye 1996

MethodsRandomised, double-blind design. Diagnosis established clinically (symptoms lasting for 10 to 30 days) and radiograph showing at least 6 mm mucosal thickening. Culture was not done. Clinical outcomes assessed by investigator; cure defined as "disappearance of all pretreatment symptoms relevant to infection." Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 0.9% on day 10 to 12
ParticipantsPatients recruited from general practices in Norway; 150 men and 288 women. Mean age 40 (range 19 to 71). ENT co-morbidity was not assessed. Country - Norway
Interventions2 treatment arms
Group 1: azithromycin 500 mg daily for 3 days
Group 2: phenoxymethylpenicillin 1.3 g 3 times daily for 10 days
OutcomesClinical outcomes were assessed on day 10 to 12 in 434 out of 438 randomised participants and on day 23 to 27 in 436 out of 438 participants. Radiographic and bacteriological outcomes were not reported
Notes1 of the authors had affiliation to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Patients were computer-randomised in blocks of six to one of two treatment groups"
Allocation concealment (selection bias)Low riskComment: although there is no exact information on how the randomisation result was concealed, the method of sequence generation and information that placebo tablets were used give the impression that the concealment was real
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical responses and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low risk

Quote: "Double-blind, double-dummy study. Placebo-tablets were used"

Comment: although the description of the blinding was incomplete (in respect of investigator-blinding), this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described

Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 4/438 on day 10 to 12 and 2/438 participants on day 23 to 27

Comment: marginal missing data rates

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskQuote: "In addition to the clinical assessment, the amount of analgesics taken, compliance, concomitant medication, and adverse events were recorded"
Comment: although there was no information on use of analgesics or concomitant medication, this domain was graded 'low' risk of bias, because the way of reporting gives an impression that use of analgesics and concomitant medication was allowed for both groups but of such quality that they would not cause bias
Free of other bias? FundingUnclear risk1 of the authors had affiliation to a pharmaceutical company

Haye 1998

MethodsRandomised, double-blind, double-dummy design. Clinical diagnosis of acute sinusitis was based on clinical findings, which had to include 1 or both of the following signs: presence of nasal secretion (purulent at the time of examination) for > 10 and < 30 days, and maxillary sinus tenderness and/or pain of < 30 days duration (87% had both signs). Radiograph was taken to exclude participants with more than 6 mm mucosal thickening, complete opacity or an air-fluid level. Samples of nasal secretions (cotton-tipped swabs) were taken from the posterior part of the nasal cavity in 75 participants (only some samples had the growth of pathogenic bacteria). Clinical failure defined as no change or a worsening of the pretreatment symptoms and relapse if there was initial improvement or disappearance of pretreatment symptoms followed by worsening (failure and relapse figures are combined in the analyses in this review). Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 0.6% on day 10 to 12
ParticipantsPatients recruited from Norwegian General Practices. 44 men and 125 women; mean age 40 (range 21 to 70) years in the azithromycin group and 43 (range 18 to 68) years in placebo group. ENT co-morbidity was not assessed. Country - Norway
Interventions2 treatment arms
Group 1: azithromycin 500 mg once daily for 3 days
Group 2: placebo once daily for 3 days
OutcomesClinical outcomes were assessed on day 10 to 12 in 168 out of 169 participants randomised (in 86 out of 87 patients in antibiotic group and in 82 out of 82 participants in placebo group) and on day 23 to 27 in 169 participants
Notes1 of the authors had affiliation to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Patients were computer-randomised in blocks of six to either of the two treatment groups"
Allocation concealment (selection bias)Low riskComment: although there is no exact information on how the randomisation result was concealed, the method of sequence generation and information that the tablets were indistinguishable from each other give the impression that the concealment was real
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low risk

Quote: "Double-blind, double-dummy study. The azithromycin and placebo tablets were indistinguishable"

Comment: although the description of the blinding was incomplete (in respect of investigator-blinding), this domain was graded 'low' risk of bias because the study was reported to be double-blind and double-dummy, and blinding method was partly described

Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 0.6% at 10 to 12 days, and 0% at 23 to 27 days

Comment: minimal missing data rates

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskQuote: "In addition to the clinical assessment, the amount of analgesics taken, compliance, concomitant medication, and adverse events were recorded"
Comment: although there was no information on use of analgesics or concomitant medication, this domain was graded 'low' risk of bias, because the way of reporting gives an impression that use of analgesics and concomitant medication was allowed for both groups but of such quality that they would not cause bias
Free of other bias? FundingUnclear risk1 of the authors had affiliation to a pharmaceutical company

Henry 2003

MethodsRandomised, double-blind study. Diagnosis by clinical signs and symptoms (presence of either purulent nasal discharge or facial pain and/or pressure and/or tightness for 7 to 28 days), confirmed radiologically by presence of opacity or air-fluid level or mucosal thickening of more than 6 mm. Mean duration of symptoms prior to enrolment about 13 days. Clinical response was assessed by investigator; cure was defined as resolution of signs and symptoms to the level that existed prior to the occurrence of the acute illness; improvement defined as partial but incomplete resolution of the signs and symptoms. Compliance was measured by investigators reporting exact doses taken, reasons for missed doses and the amounts of study medication returned by participants at the end of therapy visit. Compliance was reported as 99% in the 3-day azithromycin group and 82% in the amoxicillin-clavulanate group. Proportion of the participants included in the per-protocol population was 85% on day 8 to 15
ParticipantsMulticentre study. 936 outpatients, 381 men and 555 women; mean age was 41 years (range 18 to 84). Allergic rhinitis was reported in 39% of participants. Country - USA
Interventions3 treatment arms: (only groups 1 and 3 used in the analysis)
Group 1: azithromycin 500 mg once daily for 3 days
Group 2: azithromycin 500 mg once daily for 6 days
Group 3: amoxicillin-clavulanate 500/125 mg 3 times daily for 10 days
Concomitant treatments not reported
OutcomesStudy reported clinical cure rates for intention-to-treat and per-protocol populations. However, only per-protocol population could be used as basis for clinical failure rate calculating in this review because true failures were not reported and there was unclear description of missing data in the ITT population. Clinical outcomes for per-protocol population were assessed at days 8 to 15 in 799 out of 941 randomised participants (in 269 out of 316 patient in azithromycin for 3 days - group and in 259 out of 314 participants in co-amoxiclav group). Second follow-up at days 22 to 36. Radiographic outcomes were not reported
NotesStudy was supported by a grant from a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Unclear riskQuote: "Double-blind, double-dummy study"
Comment: no description of the blinding procedure
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear riskComment: there was not a clear statement about missing data
Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy was supported by a grant from a pharmaceutical company

Henry 2004

MethodsRandomised, double-blind study. Clinical diagnosis (presence of facial pain/pressure/tightness over 1 or both maxillary sinus areas, facial swelling or toothache, as well as purulent discharge for > 7 and < 21 days) was confirmed by radiograph or CT (opacification or air/fluid level). Clinical response was assessed by investigator and cure was defined as resolution of at least 1 of the acute signs and symptoms of the original infection, with no worsening in the remaining symptoms or the radiographic appearance of the sinus. Compliance was assessed by inspection of returned blister cards and by counting unused study medication; to be clinically evaluable the patient had to have received at least 80% of study medication. Proportion of the participants without known or reported clinical outcome 11% on 19 to 24 days
ParticipantsMulticentre, multinational study (USA, Poland). 271 ambulatory participants, 146 women and 95 men, mean age was about 41 years. ENT co-morbidity was not assessed
Interventions2 treatment arms
Group 1: cefdinir 600 mg once daily for 10 days
Group 2: levofloxacin 500 mg once daily for 10 days
There was no restriction on the use of medications for symptom relief (decongestants, antihistamines)
OutcomesClinical outcomes were reported for clinically evaluable population only and assessed in 241 out of 271 randomised participants on 19 to 24 days (in 123 out of 138 patients in cefdinir group and in 118 out of 133 participants in levofloxacin group). Clinically evaluable population defined as participants who met the study selection criteria, received at least 80% of the intended amount of study drug, underwent an assessment at 19 to 24 days, and did not receive any other antibiotic before the assessment. Radiographic outcomes were assessed in 239 participants at 19 to 24 days
Notes3 of the authors had affiliation to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Computer-generated randomization schedule that determined each patient's treatment assignment by patient number"
Allocation concealment (selection bias)Low risk

Centralised randomisation

Quote: "Randomisation made by the statistics department of a pharmaceutical company. Study personnel called a centralized interactive voice response system to receive a code/randomization number and the corresponding drug"

Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and radiological responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind study". "Drugs were over-encapsulated to achieve an identical appearance. The drugs were prepackaged and dispensed in blister cards in such a way that each patient received the appropriate quantity of active drug for 10 days of therapy"
Comment: although the description of the blinding was incomplete (in respect of investigator blinding), this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 15/138 (11%) in cefdinir group and 15/133 (11%) in levofloxacin group on 19 to 24 days

Comment: the groups were balanced in numbers and reasons for missing data

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingUnclear risk3 of the authors had affiliation to a pharmaceutical company

Huck 1993

MethodsRandomised, double-blind trial. Diagnosis established by symptoms of acute, recurrent or chronic sinusitis with a positive radiograph showing, at a minimum, mucosal thickening. Sinus puncture was performed. Clinical outcomes assessed by investigator; improvement defined as "the patient's signs and symptoms had improved or resolved during therapy." No separate categories for "cured" and "improved." Treatment compliance reported as about 97% assessed by pills taken by participants. Proportion of the participants without known or reported clinical outcome 11% for the whole study population
ParticipantsPatients were recruited from community-based otolaryngology clinics. Participants were 38.4 years old on average (range 16 to 73) and included 50 men and 58 women. ENT co-morbidity was assessed: fewer than 25% of participants had chronic sinusitis. Country - USA
Interventions2 treatment arms
Group 1: cefaclor 500 mg twice daily for 10 days
Group 2: amoxicillin 500 mg 3 times daily for 10 days
Concomitant treatments not reported
OutcomesClinical outcomes were assessed in 96 out of 108 randomised participants overall. 56 participants with acute sinusitis and 25 participants with recurrent acute sinusitis with more than 1 episode per year with clinical improvement between episodes were considered in this review (15 participants with chronic sinusitis were not included). Outcomes were assessed on day 10 to 12 after randomisation. Radiographic and bacteriological outcomes were not measured
NotesStudy funding provided by a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical responses and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind". "To maintain blinding, drugs and placebo were placed in identical capsules and given to participants in identical bottles"
Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear risk

Total amount of missing data 11%

Comment: no description of missing data across intervention groups (when chronic participants are excluded)

Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability at baseline was insufficient to assess whether the groups were comparable at baseline or not (no information on severity rating)
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy funding provided by a pharmaceutical company

Jareoncharsri 2004

MethodsRandomised, unblinded design. 80% of participants had acute maxillary sinusitis and 20% had acute exacerbation of chronic sinusitis. Diagnosis by clinical symptoms and signs confirmed by the finding of mucopurulent discharge in the middle meatus or maxillary ostium as seen by nasal endoscopy and abnormal radiograph (criteria not specified); about 88% of patients had air-fluid level or opacity on X-ray. Bacteriological specimens were collected by sinus aspiration. Clinical cure was assessed by investigators using a 4-point scale; cure was defined as complete resolution of signs and symptoms with no radiological evidence of remaining disease; improvement defined as incomplete resolution of signs and symptoms and improvement of radiological findings. Treatment compliance was not reported. Proportion of participants without known or reported clinical outcome 2%
ParticipantsParticipants (n = 60) were recruited from 2 otolaryngologic centres. Mean age 36 (range 17 to 68); 23 men and 37 women. 20% of participants had acute exacerbation of chronic sinusitis (n = 12). About 28% of participants had allergic rhinitis. Country - Thailand
Interventions2 treatment arms
Group 1: levofloxacin 300 mg once daily for 14 days
Group 2: co-amoxiclav 625 mg 3 times a day for 14 days
Concomitant treatments were not described
OutcomesClinical outcomes were assessed in 47 out of 48 randomised participants with acute maxillary sinusitis. Outcomes were assessed on day 21 after the start of treatment. Bacteriological outcomes were assessed in 37 participants with acute maxillary sinusitis on day 14. Radiological findings were not reported separately for participants with acute maxillary sinusitis
NotesStudy was financially supported by a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low riskOutcomes reported for participants with acute maxillary sinusitis: clinical response (including radiological findings), bacteriological response and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)High riskQuote: "Open study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 2%

Comment: marginal missing data rate

Free of other bias? Baseline comparabilityLow risk

Comment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline

Concomitant allergy was more frequent in levofloxacin group than in co-amoxiclav group, but this was seen not to cause bias

Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy was financially supported by a pharmaceutical company

Karma 1991

MethodsRandomised, single-blinded trial. Diagnosis was established by at least 1 symptom of sinusitis plus an antral puncture showing fluid (study protocol required that pathogens cultured from sinus fluid be susceptible to both the study antibiotics). Radiographs showed findings "suggestive" of sinusitis. Clinical outcomes were assessed by an investigator. Clinical cure was defined as "pretreatment signs and symptoms of infection resolved." Improvement was defined as "improvement but not complete resolution of the infection." Compliance with treatment was not reported. Total proportion of the participants without known or reported clinical outcome 33% on day 11 to 13. Participants were excluded mostly due to absence of appropriate pathogens in pretreatment cultures; exclusions because of other reasons marginal (3%)
ParticipantsParticipants were recruited from academic, outpatient otolaryngology clinics. Mean age was approximately 30 (range 17 to 69); 89 men and 11 women. Country - Finland and Sweden
Interventions2 treatment arms
Group 1: clarithromycin 500 mg twice daily for 9 to 11 days
Group 2: amoxacillin 500 mg every 8 hours for 9 to 11 days
Both groups received oxymetazoline 0.05% twice daily for the first 3 days
OutcomesClinical outcomes were assessed in 67 out of 100 randomised participants (in 32 out of 50 participants in clarithromycin group and in 35 out of 50 participants in amoxicillin group) on day 11 to 13. Radiographic and bacteriological outcomes were assessed in 68 participants
NotesThere was no identified funding source
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical, radiological and bacteriological responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Single-blinded study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Exclusions from analyses: absence of appropriate pathogens in pretreatment cultures 16/50 in clarithromycin group and 14/50 in amoxicillin group; other exclusions 2 (4%) and 1 (2%) of participants in clarithromycin group and amoxicillin group, respectively

Comment: the absence of pathogen was seen not to cause bias. Other missing data marginal

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: both groups received the same co-intervention during the trial
Free of other bias? FundingUnclear riskThere was no identified funding source

Klapan 1999

MethodsRandomised, unblinded design. Diagnosis based on symptoms (lasting on average 7 days) and radiograph showing at least 4 mm mucosal thickening, opacity or air-fluid level. Sinus puncture performed in a subset of participants. Cure defined as complete disappearance of signs and symptoms; improvement as partial disappearance of symptoms without need for further antibiotics. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 6% on day 10 to 12
ParticipantsThe study setting not described. Mean age approximately 33 years old (range 15 to 50); 77 men and 23 women. Coexisting allergic rhinitis in 30%. Country - Croatia
Interventions2 treatment arms
Group 1: azithromycin 500 mg daily for 3 days
Group 2: amoxicillin-clavulanate (500 mg/125 mg) 3 times daily for 10 days
Concomitant treatments not reported
OutcomesClinical outcomes assessed on day 10 to 12 in 94 out of 100 randomised participants (in 47 out of 50 participants in azithromycin group and in 47 out of 50 participants in amoxicillin-clavulanate group); and 4 weeks after the initiation of treatment in 89 participants. Bacteriological cure assessed in 47 participants. Radiographic outcomes not reported
Notes1 of the authors had affiliation to a pharmaceutical company. Copyright by pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and bacteriological responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 3/50 (6%) in azithromycin group and 3/50 (6%) in amoxicillin-clavulanate group on day 10 to 12. Missing data 6% on day 10 to 12 and 11% at 4 weeks

Comment: the groups were balanced in numbers and reasons for missing data. Available case data given for both follow-ups

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh risk1 of the authors had affiliation to a pharmaceutical company. Copyright by pharmaceutical company

Lari 2010

MethodsRandomised, single-blinded design. Outpatients, both genders, 18 years or older. Diagnosis of acute sinusitis was based on clinical symptoms of postnasal discharge and facial pain and tenderness in maxillary sinus area for at least 7 and at last 28 days. Cure defined as complete disappearance of signs and symptoms to the level that existed prior to the occurrence of the acute illness and without requirement for other antibiotics; improvement (applicable only at the end of the treatment) as partial disappearance of symptoms without need for further antibiotics. Treatment compliance not reported. Proportion of the participants without known clinical outcome not reported (results give, however, an impression that at least at 10 days there are drop-outs)

Data not used in the meta-analysis in this review because exact data and participant numbers remain unclear at follow-ups (results given only as percentages)
ParticipantsSetting not described. In total 76 patients; detailed demographic characteristics not given. Patients with chronic sinusitis were excluded (chronic sinusitis was defined as sinusitis 3 or more times during the past 6 months). Country - Iran
Interventions2 treatment arms
Group 1: azithromycin 250 mg (500 mg single dose on the first day and 250 mg once daily for 4 days)
Group 2: co-amoxiclav 625 mg, 3 times daily for 10 days
Concomitant therapies were not reported
OutcomesClinical outcomes assessed on day 10 and 30 but the numbers of participants are missing and the results are given only as percentages
NotesThere was no identified funding source. 2 authors were employees of a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskQuote: "This was a single blind randomized clinical trial"
No further information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)High risk

Outcomes reported: clinical responses

Comment: in conclusions it was stated that "results revealed that azithromycin has less side effects, and patients were able to tolerate the medications better with a higher compliance and less economic cost than co-amoxiclav regimen".
However, any results for adverse effects were not reported

Blinding? (Participant and investigator)High riskQuote: "This was a single blind randomized clinical trial"
Comment: single-blinded trial
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear riskComment: the numbers of participants at follow-ups are missing and the results give an impression that at least at 10 days there are drop-outs
Free of other bias? Baseline comparabilityUnclear riskQuote: "Signs and symptoms of patients in both groups were more or less similar. Symptoms and signs prevalence and duration of symptoms before treatment were similar in both therapy groups. Participants also had similar age distribution"
Comment: no detailed information on the severity of symptoms and demographic characteristics
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingUnclear riskThere was no identified funding source. 2 of the authors had affiliation to a pharmaceutical company

Lasko 1998

MethodsRandomised, double-blind design. Clinical diagnosis confirmed by radiograph showing > 5 mm mucosal thickening, opacity or air-fluid level. Cure defined by symptom resolution. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 19%
ParticipantsSetting not described. Mean age 40; 102 men and 134 women. Country - Canada
Interventions2 treatment arms
Group 1: levofloxacin 500 mg daily for 10 to 14 days
Group 2: clarithromycin 500 mg twice daily for 10 to 14 days
Antihistamines and decongestants prohibited
OutcomesClinical outcomes reported at day 12 to 19 in 191 out of 236 participants randomised (in 98 out of 119 participants in levofloxacin group and in 93 out of 117 participants in clarithromycin group). Radiographic and bacteriological outcomes not reported
NotesStudy funding provided by a pharmaceutical company. 2 of the authors affiliated to the same pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Patients were assigned either to levofloxacin 500 mg once daily or to clarithromycin 500 mg twice daily, based on a computer-generated randomization schedule"
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Unclear risk

Outcomes reported: clinical response and adverse effects

Comment: in the methods section it is stated that repeat sinus X-rays were taken at 42 to 46 days giving the impression that this was an outcome. However, the results were not reported

Blinding? (Participant and investigator)Low risk

Quote: "Double-blinding was accomplished by encapsulation of tablets. The size of the capsule necessitated the use of the 250 mg tablet and each dose was therefore 2 capsules morning and evening". "Identical morning and evening bottles". "Placebo-tablets used"

Comment: although the description of the blinding was incomplete (in respect of investigator), this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described

Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 21/119 (18%) in levofloxacin group and in 24/117 (21%) in clarithromycin group

Comment: the groups were balanced in numbers and reasons for missing data

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskNo co-interventions allowed
Free of other bias? FundingHigh riskStudy funding provided by a pharmaceutical company. 2 of the authors affiliated to the same pharmaceutical company

Lindbaek 1996

MethodsRandomised, double-blind design. Clinical diagnosis of acute sinusitis (> 7 and < 30 days), confirmed by computed tomography (the criteria for confirming the diagnosis were presence of fluid level or total opacification in any sinus). Bacteriological specimens were obtained from the nasopharynx of 125 participants (42% had "normal nasal flora"). Participants assessed response as "restored", "much better", "somewhat better", "unimproved" or "worse"
Treatment compliance was not reported. In total, 12 participants (9%) (5 in the amoxicillin group and 7 in placebo group) did not complete the 10-day course of treatment. 9 were treated with another antimicrobial (2 in the amoxicillin group and 7 in the placebo group)
Proportion of the participants without known or reported clinical outcome 2% on day 10
ParticipantsParticipants recruited from Norwegian General Practices. 45 men and 85 women; mean age 39 (range 16 to 74) years. ENT co-morbidity was not assessed. Country - Norway
Interventions3 treatment arms
Group 1: penicillin V 1320 mg 3 times daily for 10 days
Group 2: amoxicillin 500 mg 3 times daily for 10 days
Group 3: placebo 3 times daily for 10 days
Groups 1 and 2 were combined in the analyses

Nasal decongestants and analgesics were allowed but not prescribed
OutcomesClinical and radiographic outcomes were assessed on day 10 in 127 out of 130 participants randomised (in 83 out of 86 participants in antibiotic groups and in 44 out of 44 participants in placebo group)
NotesStudy was funded by governmental sources
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low risk

Quote: "The randomization was performed in blocks of three within each of the six subgroups (3 subgroups by CT and 2 subgroups by a severity scale) by using a dice to generate the random allocations"

Comment: randomisation was done by a statistician (information obtained from the author)

Allocation concealment (selection bias)Low risk

The statistician sent the randomisation list to the company who produced the medication boxes with numbers according to the list. The author received the numbered boxes for each of the subgroups from the company. The randomisation codes were broken after the whole study was finished

Comment: information obtained from the author

Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and radiological responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "The trial was double blind; neither the patients, the general practitioner, nor the radiologists were aware of the allocation of treatment"
Comment: the randomisation codes were broken after the whole study was finished (information obtained from the author)
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 2% on day 10

Comment: marginal missing data rate

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating at baseline but actual numbers are not given
Free of other bias? Co-interventionsLow riskQuote: "Nasal decongestants and mild analgesics allowed in both groups"
Comment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingLow riskStudy was funded by governmental sources

Lindbaek 1998

MethodsSame study as Lindbaek 1996 but separate randomised population.
Randomised, double-blind design. Clinical diagnosis of acute sinusitis (> 7 and < 30 days), confirmed by computed tomography showing mucosal thickening of 5 mm or more in any sinus without fluid level or total opacification. Patients assessed response as "restored", "much better", "somewhat better", "unimproved" or "worse". Treatment compliance was not reported. Proportion of the participants without known or reported clinical outcome 7.4% on day 10
ParticipantsPatients recruited from Norwegian General Practices. 27 men and 43 women; mean age 40 (range 16 to 83) years. ENT co-morbidity was not assessed. Country - Norway
Interventions3 treatment arms
Group 1: penicillin V 1320 mg 3 times daily for 10 days
Group 2: amoxicillin 500 mg 3 times daily for 10 days
Group 3: placebo 3 times daily for 10 days
Groups 1 and 2 combined in the analyses.

Nasal decongestants and analgesics were allowed but not prescribed
OutcomesClinical outcomes (based on patient diary) were assessed on day 10 in 63 out of 68 participants randomised (no information for drop-outs per group)
NotesStudy was funded by governmental sources
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low risk

A dice

Comment: information obtained from the author

Allocation concealment (selection bias)Low risk

The statistician sent the randomisation results to the company who produced the medication boxes. The author received the numbered boxes from the company. The randomisation codes were broken after the whole study was finished

Comment: information obtained from the author

Selective reporting (reporting bias)Low risk

Outcomes reported: clinical responses and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "The trial was double-blind: Neither the patients, the general practitioner, nor the radiologists were aware of the treatment allocation"
Comment: the randomisation codes were broken after the whole study was finished (information obtained from the author)
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Total drop-out rate 5/68 (7.4%)

Comment: although there were no exact numbers on the drop-out rates across the 3 groups, this domain was graded 'low' risk of bias. The total drop-out rate was seen as marginal. Further, the author gave information that there were drop-outs in all groups

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskQuote: "Nasal decongestants and mild analgesics allowed in both groups"
Comment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingLow riskStudy was funded by governmental sources

Luterman 2003

MethodsRandomised, double-blind trial. Diagnosis by clinical symptoms and by radiograph showing air-fluid level, opacity or at least 6 mm mucosal thickening. Sinus puncture performed in a small subset of participants. About 60% of participants had had symptoms for longer than 7 days before enrolment to the study. Clinical cure defined as disappearance of infection, clinical improvement or a return to the preinfection state with regard to all signs and symptoms and sinus X-ray findings that were normal, improved or at least not worse. Treatment compliance was reported with over 90% of participants in each group taking at least 70% of their doses. Proportion of the participants without known or reported clinical outcome 44% on day 17 to 24 and therefore data are not used in the meta-analyses in this review
ParticipantsMulticentre, multinational trial (USA, Canada, South Africa, Argentina and Chile). Demographic data reported for modified intention-to-treat population at baseline: participants' mean age was 39 years (range 16 to 84); 248 men and 359 women. ENT co-morbidity was not assessed
Interventions3 treatment arms
Group 1: telithromycin 800 mg once daily for 5 days
Group 2: telithromycin 800 mg once daily for 10 days
Group 3: amoxicillin-clavulanate 500/125 mg 3 times daily for 10 days
Concomitant therapies were not reported
OutcomesResults reported for per-protocol population only. Clinical outcomes were assessed on day 17 to 24 in 423 out of 754 randomised participants; and on day 31 to 45 in 399 participants. The per-protocol population was defined as participants who met the diagnostic criteria and had no major protocol violation. Bacteriological outcomes were assessed in 24 participants
NotesStudy funded by an unrestricted grant from a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response (including radiological findings) and bacteriological response for a small proportion of the participants, adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind study". "The identity of all drugs was concealed by placing them in identical opaque capsules. Placebo-capsules were also used"
Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described
Incomplete outcome data addressed? (Clinical efficacy outcomes)High riskTotal missing data 44% on day 17 to 24
Free of other bias? Baseline comparabilityUnclear riskComment: the demographic and severity rating reported only on modified intention-to-treat population, not on the per-protocol population used in analysis of the study
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingUnclear riskStudy funded by an unrestricted grant from a pharmaceutical company

Marple 2010

MethodsRandomised, open-label study. Clinical diagnosis of acute sinusitis was based on signs and symptoms lasting for between 7 and 30 days (pain, pressure and/or tightness associated with 1 or both maxillary sinuses that worsened with movement or percussion, and discoloured (yellow-green) nasal discharge or discoloured drainage in the posterior pharynx or from the maxillary sinus orifice, together with 2 or more of fever, frequent coughing, nasal congestion or postnasal drainage). Treatment success by day 28 (the only time point eligible in this review) was defined as resolution of symptoms within the study period, no unscheduled sinusitis visits and no additional antibiotics prescribed for sinusitis. Treatment compliance not reported. Proportion of the participants included in the per-protocol population was 62% on day 28; data not used in the meta-analyses of this review
ParticipantsMulticentre study, participants were recruited in primary care settings by 70 study investigators. Demographic data for per-protocol population at baseline: participants' mean age was 46 years; 151 men and 323 women. ENT co-morbidity was assessed; about 37% of participants had history of allergic rhinitis. Country - USA
Interventions2 treatment arms
Group 1: azithromycin extended release (2 g), single dose
Group 2: amoxicillin-clavulanate potassium 875 mg/125 mg 2 times daily for 10 days
Concomitant therapies were not reported
OutcomesClinical responses on day 28 were reported for per-protocol population only. Clinical responses (based on follow-up telephone interviews) were assessed on day 28 in 462 out of 751 participants randomised (in 231 out of 380 participants in azithromycin group and in 231 out of 371 participants in amoxicillin-clavulanate group). The per-protocol population was defined as all eligible randomised participants who completed the diary and first telephone interview (at day 5) and reported taking at least 1 dose of the study medication
NotesThe study was sponsored by a pharmaceutical company. 3 of the authors had affiliation to, and other 3 of the authors were paid consultants of, the same pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical responses, quality of life, satisfaction with treatment and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open-label study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)High risk

Missing data: 149/380 (39%) in azithromycin group and 140/371 (38%) in amoxicillin-clavulanate group on day 28

Comment: total missing data over 20%

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating for per-protocol population to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskThe study was sponsored by a pharmaceutical company. 3 of the authors had affiliation to, and other 3 of the authors were paid consultants of, the same pharmaceutical company

Matthews 1997

MethodsRandomised, unblinded design. Diagnosis by clinical symptoms and signs, confirmed by radiograph showing opacity, clouding or air-fluid level. Bacteriological specimens were obtained by antral puncture. Acute and chronic sinusitis included (acute reported separately). Success defined as cure or improvement. Proportion of the participants without known or reported clinical outcome 24% among participants with acute sinusitis. Data not used in the meta-analyses of this review because time points for outcome measure were not clearly reported
ParticipantsParticipants recruited from otolaryngology clinics. Mean age of 42 (range 18 to 78). 93 of 182 participants had acute sinusitis. Country - USA
Interventions2 treatment groups
Group 1: cefixime 400 mg once daily for 10 to 14 days
Group 2: amoxicillin 500 mg 3 times daily for 10 to 14 days
Aspirin prohibited. Acetaminophen, antacids and nasal decongestants were permitted
OutcomesClinical outcomes reported in 71 out of 93 randomised participants with acute sinusitis (in 37 out of 49 participants in cefixime group and in 34 out of 44 participants in amoxicillin group). Bacteriological outcomes assessed in 32 participants with acute sinusitis. Drop-outs due to adverse events reported for acute and chronic sinusitis combined
NotesStudy funding not stated
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Patients were randomized by a computer-generated schedule to receive oral doses of either 400 mg of cefixime once daily or 500 mg of amoxicillin every 8 hours"
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and bacteriologic responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open label study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)High risk

Missing data: 12/49 (24%) in cefixime group and 10/44 (23%) in amoxicillin group

Comment: missing data over 20%

Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability was insufficient to assess whether the intervention groups with acute sinusitis were comparable at baseline or not
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingUnclear riskStudy funding not stated

Mattucci 1986

MethodsRandomised trial; unblinded. Diagnosis was established by a clinical diagnosis of acute sinusitis plus a radiograph showing "clouding" of the nasal sinus and a positive culture from or near the sinus ostium. Clinical outcomes were assessed by an investigator; clinical cure was not defined. Compliance with treatment was not reported. Proportion of the participants without known or reported clinical outcome 19%
ParticipantsPatients were recruited from outpatient otolaryngology clinics. Participants were 40 years old on average (range 12 to 76) and included 23 men and 35 women. ENT co-morbidity was not assessed. Country - USA
Interventions2 treatment arms
Group 1: minocycline 100 mg twice daily for 10 to 20 days (mean 11.6)
Group 2: amoxicillin 250 mg every 8 hours for 10 to 20 days (mean 11.4)
Both groups were allowed to take decongestants or analgesics but these were not prescribed by the investigators (all of the participants took decongestants during the trial)
OutcomesClinical outcomes were assessed in 47 out of 58 randomised participants (in 25 out of 29 participants in minocycline group and in 22 out of 29 participants in amoxicillin group). Outcomes were assessed on day 10 to 20 after randomisation. Radiographic outcomes were assessed in 42 participants and bacteriological in 46
NotesThere was no identified funding source
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low riskOutcomes reported: clinical response, radiological response for 42 participants, bacteriological response for 46 participants and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)High riskQuote: "Open-label study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear riskExclusions from analyses: absence of pathogen in pretreatment cultures 3/29 in minocycline group and 3/29 in amoxicillin group; other exclusions 1 (3.4%) and 4 (13.8%) of participants in minocycline group and amoxicillin group, respectively
Comment: the absence of pathogen was seen not to cause bias. Other missing data in numbers not balanced between the groups
Free of other bias? Baseline comparabilityHigh riskQuote: "The proportion of participants with severe sinusitis 68% in minocycline group and 43% in amoxicillin group"
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingUnclear riskThere was no identified funding source

Meltzer 2005

MethodsRandomised, double-blind, double-dummy, placebo-controlled study. Clinical diagnosis of acute sinusitis was based on clinical findings with signs and symptoms of acute rhinosinusitis lasting at least 7 but less than 29 days. In addition, the major symptom score (MSS) had to be at least 5 but less than 13 at screening and baseline visits with no more than 3 of the 5 symptoms rated severe. MSS was defined as sum of the scores for rhinorrhoea, postnasal drip, nasal congestion/stuffiness, sinus headache and facial pain/pressure/tenderness on palpation over the paranasal sinuses (each scaled from 0 (none) to 3 (severe)). Duration of symptoms at enrolment 6 to 14 days in 61% of participants in amoxicillin group and 58% in placebo group (out of the total 981 randomised patients 9 were enrolled with 6 days’ duration of symptoms instead of the originally specified minimum of 7 days). Participants were excluded if they had signs or symptoms suggestive of fulminant bacterial rhinosinusitis.
Clinical failure defined by worsening or no improvement of symptoms by day 3 to 4 or thereafter jointly evaluated by the investigator and the participant. Protocol noncompliance 1% in amoxicillin and placebo groups. Proportion of the participants without known or reported clinical outcome 1.8% by day 15
ParticipantsParticipants recruited from 71 medical centres in 14 countries. Study population consisted of a total of 981 randomised patients in 4 groups: 243 in mometasone furoate nasal spray (MFNS) 200 µg once daily, 235 in MFNS 200 µg twice daily, 251 in amoxicillin and 252 in placebo. 338 men and 643 women; mean age 35 years (4% of participants were 12 to 17 years of age in amoxicillin group and 6% in placebo group). ENT co-morbidity: 42% of participants had either seasonal or perennial allergic rhinitis in amoxicillin and placebo groups
Interventions

4 treatment arms
Group 1: MFNS 200 µg in the morning with placebo spray in the evening for 15 days plus placebo capsules 3 times daily for 10 days
Group 2: MFNS 200 µg in the morning and evening for 15 days plus placebo capsules 3 times daily for 10 days
Group 3: Amoxicillin 500 mg 3 times daily for 10 days plus placebo nasal spray twice daily for 15 days
Group 4: Placebo capsules 3 times daily for 10 days plus placebo nasal spray twice daily for 15 days
Groups 3 and 4 considered in this review.

Concomitant medications that would interfere with study evaluations were not permitted, including nasal saline, nasal cromolyn sodium, ipratropium bromide, corticosteroids (excluding oral inhaled corticosteroids for mild to moderate persistent asthma), antihistamines, decongestants and leukotriene pathway modifiers. Analgesics and nonsteroidal anti-inflammatory drugs were prohibited for the treatment of acute rhinosinusitis

OutcomesThe original article reported the total failure rates during the 15-day treatment phase per group but exact numbers of the assessed participants were not given (only numbers and reasons for discontinued participants per group). It remains unclear whether all other discontinued participants than those lost to follow-up were assessed or not. The original article used intention-to-treat principle in the analyses. We decided to exclude in our analyses those randomised participants who were reported to be lost to follow-up (1 in antibiotic group and 4 in placebo group) and those 4 participants in placebo group who did not meet the protocol criteria for entry. Thus clinical failures were assessed during the 15-day treatment phase in 494 out of 503 randomised participants (in 250 out of 251 participants in amoxicillin group and in 244 out of 252 participants in placebo group). Figures for failures at the follow-up phase (days 16 to 29) not given
NotesStudy supported by a grant from a pharmaceutical company. 1 of the authors was affiliated to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Randomisation was performed according to a computer-generated code, stratified on the basis of duration of rhinosinusitis symptoms before baseline (7-14 days and 15-28 days)"
Allocation concealment (selection bias)Low riskComment: although there is no exact information on how the randomisation result was concealed, the method of sequence generation of this multicentre study (71 centres in 14 countries) and information that placebo nasal sprays and placebo capsules were used, give the impression that the concealment was real
Selective reporting (reporting bias)Low riskOutcomes reported: clinical responses and adverse effects
Comment: prespecified outcomes (in methods) were reported in prespecified way
Blinding? (Participant and investigator)Low riskQuote: "Double-blind, double-dummy study. Those receiving amoxicillin a matching placebo nasal spray twice daily; patients randomized to placebo received placebo capsules 3 times daily and placebo nasal spray twice daily"
Comment: although the description of the blinding was incomplete (in respect of investigator blinding), this domain was graded 'low' risk of bias because the study was reported to be double-blind and double-dummy, and blinding method was partly described
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 1/251 (0.4%) in amoxicillin group and in 8/252 (3.2%) in placebo group

Comment: marginal missing data rate

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskQuote: "Concomitant medications that would interfere with study evaluations were not permitted"
Comment: medications that might cause bias were prohibited
Free of other bias? FundingHigh riskStudy supported by a grant from a pharmaceutical company. 1 of the authors was affiliated to a pharmaceutical company

Merenstein 2005

MethodsRandomised, double-blind design. Clinical diagnosis of acute sinusitis was based on clinical findings, which had to include at least 1 of the following: purulent nasal discharge predominating on 1 side, facial pain on 1 side, purulent nasal discharge on both sides, pus in the nasal cavity, lasting for at least 7 days. About 50% of participants had at least 2 of these signs at baseline (the study did not report any subjective symptoms); mean duration of symptoms prior to enrolment 11 days. Clinical cure defined as dichotomous - either "completely improved" or "not improved". Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 14% on day 14
ParticipantsParticipants recruited from a suburban primary care office. 42 men and 93 women; mean age 34 years. ENT co-morbidity was not assessed. Country - USA
Interventions

2 treatment arms
Group 1: amoxicillin 500 mg twice daily for 10 days
Group 2: placebo twice daily for 10 days

Over-the-counter medications were allowed in both groups (information obtained from the authors)

OutcomesClinical outcomes (based on follow-up telephone interviews of participants) were assessed on day 14 in 116 out of 135 participants randomised (in 56 out of 67 participants in antibiotic group and in 60 out of 68 participants in placebo group)
NotesStudy was funded mostly by academic sources
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Permuted block randomization stratified for the 3 participating clinicians was used to determine treatment assignment"
Comment: computer-generated by a statistician who gave the authors a list to follow (information obtained from the authors)
Allocation concealment (selection bias)Low riskComment: a statistician made the randomisation and gave the authors a list that they followed (information obtained from the authors)
Selective reporting (reporting bias)Low risk

Outcomes reported: several clinical responses and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low risk

Quote: "Patients were given an envelope containing either a placebo medicine taken twice daily for 10 days or antibiotic medicine taken twice daily for 10 days. The envelopes were opaque, and each had 40 identical-appearing pills"

Comment: the authors followed the randomisation list without knowing which medicine was given to each patient (information obtained from the authors)

Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 16.4% (11/67) in antibiotic group and 11.8% (8/68) in placebo group at 14 days

Comment: the groups appeared to be balanced in numbers and reasons for missing data

Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline. The groups appeared to be balanced at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingLow riskFunded mostly by academic sources

Murray 2005

MethodsRandomised, double-blind trial. Clinical diagnosis of acute sinusitis (> 6 and < 29 days), confirmed by radiograph (the criteria for confirming the diagnosis were presence of air-fluid level or total or partial opacification). Sinus puncture for all participants. Clinical outcomes were assessed by an investigator. Clinical failure was defined as the persistence or worsening of signs and symptoms requiring additional antibiotics or development of new signs and symptoms requiring antibiotics. Treatment compliance reported over 96%. Proportion of the participants without known or reported clinical outcome 6%
ParticipantsMulticentre study; 81 outpatient centres in the United States, India, Europe and Latin America. 225 men and 313 women; mean age 38 (range 18 to 88) years. ENT co-morbidity was assessed; about one-third of participants suffered from allergic rhinitis
Interventions2 treatment arms
Group 1: azithromycin-microspheres 2 g single dose
Group 2: levofloxacin 500 mg once daily for 10 days
OutcomesClinical outcomes were assessed at days 17 to 24 in 507 out of 541 participants randomised. Bacteriologic outcomes assessed in 213 participants
NotesStudy funded by a pharmaceutical company. 2 of the authors were affiliated to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and bacteriologic responses, clinical response by baseline pathogen and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind, double-dummy trial". "Placebo medication used"
Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 14/270 (5%) in azithromycin group and 17/268 (6%) in levofloxacin group

The information on number of randomised participants across groups was insufficiently reported. In calculating the missing data, the population who received at least 1 dose of study medication was used as the baseline population (missing only 3 participants from the total randomised population)

Comment: although reasons for missing data across groups were not reported, this domain was graded 'low' risk of bias because missing data were marginal and balanced in numbers

Free of other bias? Baseline comparabilityLow risk

Comment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline

Although the information was provided only for the whole treated population (not for clinically evaluable population which was the primary efficacy endpoint) this domain was graded 'low' risk of bias because the missing data rate was small

Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy funded by a pharmaceutical company. 2 of the authors were affiliated to a pharmaceutical company

Nyffenegger 1991

MethodsRandomised, unblinded study. Diagnosis was established by a clinical diagnosis of acute sinusitis, confirmed by a radiograph (criteria were not specified), transillumination and a positive bacterial culture obtained from samples of discharge from the sinuses. Clinical outcomes were assessed by an investigator. Clinical cure was defined as complete disappearance of clinical symptoms at completion of treatment; improvement was defined as a clear regression of the clinical symptomatology. Compliance with treatment was not reported. Proportion of the participants without known or reported clinical outcome 0%
ParticipantsParticipants were recruited from an outpatient otolaryngology clinic. Participants represented a convenience sample, not a consecutive series. Participants were 37.8 years old on average (range 18 to 75) and included 30 men and 50 women. ENT co-morbidity was not assessed. Country - Switzerland
Interventions2 treatment arms
Group 1: brodimoprim 200 mg once daily for 7 to 10 days
Group 2: amoxicillin 750 mg 3 times daily for 7 to 10 days
Mucolytic nose drops and steam inhalations were allowed but not prescribed as part of the intervention
OutcomesClinical outcomes were assessed in 80 out of 80 randomised participants. Outcomes were assessed on day 7 to 10 after randomisation. Radiographic outcomes were not reported. Bacteriological outcomes were reported for 40 participants
Notes2 of the authors were affiliated to pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "According to a computer-generated list, the patients were randomly assigned to receive 200 mg of brodimoprim once daily or 750 mg of amoxicillin three times daily for seven to 10 days"
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Unclear risk

Outcomes reported: clinical response, bacteriological response for 40 patients and adverse effects

Comment: it remains unclear why only a proportion of the population was included in the bacteriological response and how were they selected

Blinding? (Participant and investigator)High riskQuote: "Unblinded study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low riskMissing data rate 0%
Free of other bias? Baseline comparabilityUnclear riskComment: detailed description of demographic characteristics. However, no information on the severity rating across the groups
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingUnclear risk2 of the authors were affiliated to pharmaceutical company

O'Doherty 1996

MethodsRandomised, unblinded design. Participants with otitis media, pharyngitis/tonsillitis or sinusitis included. Sinusitis diagnosis was established clinically and confirmed by a radiograph (criteria not given). Throat swabs, nasopharyngeal swabs or exudates, as appropriate, were collected for bacterial culture. Clinical outcomes were assessed by an investigator; clinical cure defined as "disappearance of all signs and symptoms of infection". Compliance with treatment was not reported. Proportion of the participants without known or reported clinical outcome 14%
ParticipantsRecruitment settings were not described. Mean age was 35 for all diagnoses. ENT co-morbidity was not assessed. Countries - UK, Ireland
Interventions2 treatment arms
Group 1: azithromycin 500 mg daily for 3 days
Group 2: cefaclor 250 mg 3 times daily for 10 days
Adjuvant treatments were not described
OutcomesClinical outcomes were assessed on day 11 to 15 and reported separately for 78 out of 91 randomised participants with sinusitis (in 41 out of 48 participants in azithromycin group and in 37 out of 43 participants in cefaclor group). Radiographic outcomes were not reported. Bacteriological outcomes were assessed in 18 participants
NotesThere was no identified funding source
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response, bacteriological response for 18 participants and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear risk

Missing data: 7/48 (15%) in azithromycin group and 6/43 (14%) in cefaclor group on day 11 to 15

Comment: no information on reasons for missing data separately for sinusitis participants across groups

Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability was insufficient to assess whether the groups were comparable at baseline or not (information not stated separately for participants with sinusitis)
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingUnclear riskThere was no identified funding source

Olmo 1994

MethodsRandomised, obviously unblinded design. Diagnosis was established by clinical symptoms and a radiograph showing sinus opacity or air-fluid level. 66% of participants had positive culture. Clinical outcomes were assessed by an investigator; clinical cure was "remission of symptoms within the treatment time". Compliance with treatment was not reported. Proportion of the participants without known or reported clinical outcome 0% on day 24 to 26
ParticipantsParticipants were recruited from emergency rooms. Mean age was not reported (>= 20 for inclusion); 28 men and 19 women. ENT co-morbidity was not assessed. Country - Spain
Interventions2 treatment arms
Group 1: cefuroxime axetil 500 mg twice daily for 8 days
Group 2: amoxicillin-clavulanate 500 mg 3 times daily for 8 days
Adjuvant therapies were not described
OutcomesOverall clinical outcomes (including also radiological response and bacteriological response for those participants with positive culture at baseline) were assessed in 48 out of 48 randomised participants on day 24 to 26. Bacteriological outcomes were also reported separately for the 31 participants with positive culture at baseline
NotesThere was no identified funding source
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Random numbers table"
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response, bacteriological response and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Unclear riskNo information provided
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low riskNo missing data
Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability was insufficient to assess whether the groups were comparable at baseline or not
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingUnclear riskThere was no identified funding source

Otte 1983

MethodsRandomised, double-blind design. Diagnosis based on symptoms and abnormal radiograph (criteria not given). Bacteriological culture was performed. Clinical cure based on resolution of symptoms and normal radiograph. Treatment compliance was not reported. Proportion of the participants without known or reported clinical outcome 0% on day 10
ParticipantsPatients recruited from outpatient otolaryngology clinics. Mean age was 32 (range 17 to 56); 12 men and 19 women. ENT co-morbidity was not assessed. Country - Chile
Interventions2 treatment arms
Group 1: sulfamethopyrazin 200 mg/ trimethoprim 250 mg, 2 doses on day 1, then 1 dose per day for 10 days
Group 2: tetracycline 500 mg 4 times daily for 10 days
Adjuvant therapies were not described
OutcomesClinical outcomes were assessed on day 10 in all 31 participants randomised. Radiological outcomes reported on day 20 in 31 participants, bacteriological outcomes for 27 participants
NotesThere was no identified funding source
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and radiological responses for all participants, bacteriological response for a proportion of the participants and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind study". "Placebo-tables were used; antibiotics and placebo tablets were identical. Tablets were delivered in numbered envelopes"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low riskMissing data rate 0%
Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability was insufficient to assess whether the groups were comparable at baseline or not
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingUnclear riskThere was no identified funding source

Pessey 1996

MethodsRandomised, open design. Diagnosis established clinically and by a radiograph showing at least 5 mm mucosal thickening, sinus opacity or fluid-level. Clinical outcomes were assessed by an investigator and clinical cure was defined as resolution of clinical symptoms and radiographic cure or improvement. Compliance with treatment was not reported. Proportion of the participants without known or reported clinical outcome 4% on day 10 to 15
ParticipantsParticipants were recruited from community, otolaryngology clinics. Mean age = 43; 99 men and 143 women. Country - France
Interventions3 treatment arms: (only groups 2 and 3 used in the analysis)
Group 1: cefixime 200 mg twice daily for 4 days (+ placebo for 6 days)
Group 2: cefixime 200 mg twice daily for 10 days
Group 3: amoxicillin clavulanic acid 500 mg 3 times daily for 10 days
All groups received prednisolone 40 mg daily and oxymetazoline 3 sprays daily for 4 days
OutcomesClinical outcomes were assessed on day 10 to 15 in 242 out of 253 randomised participants (in 85 out of 87 participants in cefixime for 10 days group and in 77 out of 82 participants in amoxicillin clavulanic acid group). Second follow-up on day 33. Radiographic outcome was not reported separately
NotesThere was no identified funding source
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: overall response including clinical and radiological responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open study" (when comparing group 2 to group 3)
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 2/87 (2.3%) in cefixime group and 5/82 (6.1%) in amoxicillin clavulanic acid group on day 10 to 15. Total missing data 4% on day 10 to 15 and 9% on day 33

Comment: missing data rate marginal on day 10 to 15. The judgement on day 33 was 'unclear' because the study did not report the reasons for missing data by study groups

Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: groups received the same co-interventions during the trial
Free of other bias? FundingUnclear risk2 of the authors were affiliated to a pharmaceutical company(?)

Pessey 2001

MethodsRandomised, double-blind design. Diagnosis based on clinical signs and symptoms confirmed by radiograph (total opacity of sinus or air-fluid level). Culture based on an aspirate from the middle meatus; 52% of participants had positive cultures. Clinical cure based on resolution of symptoms. Treatment compliance over 90%. Proportion of the participants without known or reported clinical outcome 0.6% on day 10 to 12
ParticipantsMulticentre study; 47 community-based ENT specialists in 10 regions. Mean age = 43 years; men 141 and 167 women. ENT co-morbidity was not assessed. Country - France
Interventions2 treatment arms
Group 1: pristinamycin 1000 mg 2 times daily for 8 days
Group 2: cefuroxime axetil 500 mg twice daily for 8 days
Nasal decongestants prescribed for 3 days. Corticosteroids prohibited
OutcomesClinical outcomes were assessed on day 10 to 12 in 308 out of 310 randomised participants (in 160 out of 161 participants in pristinamycin group and in 148 out of 149 participants in cefuroxime axetil group). Follow-up (29 to 36 days) for those participants who were cured at the end of treatment. Participant-related radiographic and bacteriological outcomes were not reported
NotesThere was no identified funding source
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low risk

Quote: "Treatment was allocated by means of a randomization code. The randomization was balanced in each center by blocks of 4 participants"

Comment: although the description of the sequence generation was incomplete, this domain was graded 'low' risk of bias because description gives an impression of computer-based method

Allocation concealment (selection bias)Low riskQuote: "Sealed individual envelopes containing the identification of the treatment to be administered to each patient were handed to each investigator"
Comment: although the description of the allocation concealment was incomplete, this domain was graded 'low' risk of bias because the way of reporting the blinding also gives an impression of adequate allocation concealment
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind, double-dummy study". "Sealed individual envelopes containing the identification of the treatment to be administered to each patient were handed to each investigator. The code number of each patient was noted on each of the envelopes. The treatments had to be assigned by the investigator in ascending order of the numbers on the envelopes"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 1/161 (0.6%) in pristinamycin group and 1/149 (0.7%) in cefuroxime axetil group on day 10 to 12

Comment: marginal missing data rate

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: in both groups the same concomitant medication was prescribed
Free of other bias? FundingUnclear riskThere was no identified funding source

Rakkar 2001

MethodsRandomised, unblinded design. Diagnosis of acute sinusitis was based on clinical findings (purulent nasal discharge, nasal congestion, postnasal drainage, frequent coughing or throat clearing, frontal headache, malar tenderness/pain); lasting for at least 7 days. Clinical response was assessed by investigator and it was defined as disappearance of acute signs and symptoms related to the infection or sufficient improvement such that additional or alternative antimicrobial therapy was not required. Treatment compliance not reported but to be included in the per-protocol population the patient had to have received at least 80% of study medication. Proportion of the participants included in the per-protocol population was 72% on day 24 to 31
ParticipantsMulticentre study; primary care participants. 156 men and 315 women; mean age approximately 42 (range 18 to 87) years. ENT co-morbidity was not assessed. Country - USA
Interventions2 treatment arms
Group 1: moxifloxacin 400 mg once daily for 10 days
Group 2: amoxicillin clavulanate 875 mg twice daily for 10 days
Use of oral or nasal decongestants or antihistamines was permitted; systemic or topical corticosteroids were not allowed
OutcomesAvailable case data were basically possible to extract for per-protocol population only. However, the way of reporting does not enable identification of clinical failure rates because of inconsistent description of clinical response in results (it remains unclear in which outcome group improvements are included). Clinical outcomes for per-protocol population on day 24 to 31 were reported in 341 out of 475 randomised participants (in 170 out of 238 participants in moxifloxacin group and in 171 out of 237 participants in amoxicillin clavulanate group). (The per-protocol population included all participants with no major protocol violation who had received a study antibiotic for at least 72 hours if a clinical failure, had completed the evaluation at 14 to 21 days post-treatment, and had adequate compliance documented with at least 80% of study medication administered)
NotesStudy supported by a research grant from a pharmaceutical company. 2 of the authors were affiliated to the same pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low risk

Quote: "Block design random code"

Comment: although the description of the sequence generation was incomplete, this domain was graded 'low' risk of bias because description gives an impression of computer-based method

Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical responses and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Non-blinded"
Incomplete outcome data addressed? (Clinical efficacy outcomes)High risk

Missing data: 68/238 (29%) in moxifloxacin group and 66/237 (28%) in amoxicillin clavulanate group

Comment: missing data rate over 20%

Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingHigh riskStudy supported by a research grant from a pharmaceutical company. 2 of the authors were affiliated to the same pharmaceutical company

Riffer 2005

MethodsRandomised, investigator-blinded design. Diagnosis by clinical symptoms and opacification or an air/fluid level in a sinus radiograph or CT. Sinus fluid samples were obtained by aspiration or fibreoptic endoscopy. Clinical cure defined as resolution or improvement in purulent nasal discharge and at least 1 additional sinusitis sign or symptom observed at baseline. Treatment compliance at least 80% for those participants who were included in the clinically evaluable population. Proportion of the participants without known or reported clinical outcome 16% on day 16 to 18
ParticipantsMulticentre, multi-country trial. Mean age was 37 (range 13 to 79); 194 men and 243 women. ENT co-morbidity was not assessed
Interventions2 treatment arms
Group 1: clarithromycin ER 1000 mg once daily for 14 days
Group 2: amoxicillin/clavulanate 875/125 mg twice daily for 14 days
Oral sympathomimetic agents, nasal decongestants and antihistamines were used by 23.8%, 8.9% and 6.9% of participants, respectively
OutcomesStudy reported clinical cure rates for intention-to-treat and clinically evaluable populations. However, only clinically evaluable population could be used as basis for clinical failure rate calculation in this review because true failures were not reported and it seemed that indeterminate and missing responses were included in failures in ITT analyses. Clinical outcomes were assessed on day 16 to 18 in 373 of 437 randomised participants (in 188 out of 221 participants in clarithromycin group and in 185 out of 216 participants in amoxicillin-clavulanate group) and on day 24 to 31 in 361 participants. The clinically evaluable population was defined as all randomised participants with clinically and radiographically confirmed sinusitis, and were without major protocol violations. Bacteriological outcomes assessed on day 16 to 18 in 109 of 218 participants. Radiographic outcomes assessed on day 24 to 31 in 366 of 423 participants. Quality of life measured by symptom assessment and the SNOT-16 test
NotesThe study was supported by a grant from a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical, radiological and bacteriological responses, adverse effects and quality of life

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High risk

Quote: investigator-blinded study. To maintain the study blind, participants were instructed not to discuss the appearance of study drug or any aspect of dosing with the investigator

Comment: no blinding of the participants

Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 33/188 (17.5%) in clarithromycin group and 31/216 (14.4%) in amoxicillin-clavulanate group on day 16 to 18. Total missing data 16% on day 16 to 18 and 17% day 24 to 31

Comment: the groups appeared to be balanced in numbers and reasons for missing data

Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingHigh riskThe study was supported by a grant from a pharmaceutical company

Russell 1997

MethodsRandomised trial. Blinding not stated. Diagnosis by clinical signs and symptoms, confirmed by X-ray or CT (opacification, air/fluid level or mucosal thickening). Clinical outcomes were assessed by an investigator; clinical cure defined as all clinical signs and symptoms were resolved and no new clinical signs and symptoms were present and/or the follow-up radiograph showed resolution/improvement. Approximately 80% of participants received the prescribed antibiotic course. Proportion of the participants without known or reported clinical outcome 12% on day 20 to 35
ParticipantsStudy setting not described. Mean age was approximately 38 (range 13 to 79); 172 men and 256 women. 84% of participants had a history of ENT or eye-related medical conditions. Country - USA
Interventions

3 treatment arms: (groups 1 and 3 used in the analysis)
Group 1: cefprozil 250 mg twice daily for 10 to 15 days
Group 2: cefprozil 500 mg twice daily for 10 to 15 days
Group 3: amoxicillin-clavulanate (500 mg/125 mg) 3 times daily for 10 to 15 days

Adjuvant therapies permitted but not described

OutcomesClinical outcomes were assessed on day 20 to 35 in 428 out of 479 participants (in 146 out of 158 participants in cefprozil 250 mg group, in 147 out of 160 participants in cefprozil 500 mg group and in 135 out of 161 participants in amoxicillin-clavulanate group). Radiographic and bacteriological outcomes not reported
NotesStudy was supported by a pharmaceutical company. 3 of the authors were affiliated to pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Unclear riskNo information provided
Incomplete outcome data addressed? (Clinical efficacy outcomes)High risk

Missing data: 12/158 (8%) in cefprozil 250 mg group and 26/161 (16%) in amoxicillin-clavulanate group on day 20 to 35

Comment: the groups were imbalanced in numbers of missing data

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy was supported by a pharmaceutical company. 3 of the authors were affiliated to pharmaceutical company

Sher 2002

MethodsRandomised, investigator-blinded design. Diagnosis by clinical signs and symptoms lasting at least 7 days, confirmed by radiograph showing at least 5 mm mucosal thickening, opacification or air fluid level in a sinus radiograph. Clinical outcomes were assessed by an investigator and clinical cure was defined as improvement in or resolution of all acute signs and symptoms of the original infection, with no need for further antimicrobial therapy. Treatment compliance for clinically evaluable population over 80% by counting unused study medication. Proportion of the participants without known or reported clinical outcome 9%
ParticipantsMulticentre study; 30 centres. Participants were 42 years old on average (range 18 to 89); 159 men and 286 women. ENT co-morbidity was assessed; for example, 60% of participants had a history of allergic rhinitis. Country - USA
Interventions3 treatments arms
Group 1: gatifloxacin 400 mg once daily for 5 days
Group 2: gatifloxacin 400 mg once daily for 10 days
Group 3: amoxicillin/clavulanate 875 mg twice daily for 10 days
Medications for symptom relief (for example, decongestants and antihistamines) were allowed
Outcomes

Study reported results for clinically evaluable participants only. Clinical outcomes were assessed on day 10 to 13 and on day 17 to 24 in 405 out of 445 randomised participants (in 137 out of 149 participants in 5-day gatifloxacin group, in 127 out of 141 participants in 10-day gatifloxacin group and in 141 out of 155 participants in amoxicillin-clavulanate group). Clinically evaluable participants were those who met the inclusion criteria and had received at least 80% of the intended amount of study drug, had undergone the test-of-cure assessment (7 to 14 days after the completion of treatment), and had not received any other antibiotic during the study or before the test-of-cure visit for an infection other than sinusitis. Long-term follow-up on day 31 to 38 by telephone

Radiographic and bacteriological outcomes not reported

Notes2 of the authors were affiliated with a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Patients were randomized to treatment by means of a centralized telephone system. A permuted block design was used to minimize imbalance in treatment arms at each site and in the study overall"
Allocation concealment (selection bias)Low riskQuote: "Patients were randomized to treatment by means of a centralized telephone system. A permuted block design was used to minimize imbalance in treatment arms at each site and in the study overall"
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Unclear riskQuote: "Investigator-blinded trial". "Matching placebo-tablets were used, and drugs and placebos were supplied by the respective manufacturers"
Comment: although blinding of participants was also described, this domain was graded 'unclear' because the study was reported to be only investigator-blinded
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 12/149 (8%) in 5-day gatifloxacin group, 14/141 (10%) in 10-day gatifloxacin group and 14/155 (9%) in amoxicillin/clavulanate group

Quote: "The 40 unevaluable participants were distributed evenly between groups and between reasons for non-evaluability"
Comment: although there was no detailed description of the reasons for missing data, this domain was graded 'low' risk of bias because the study stated that the groups were balanced

Free of other bias? Baseline comparabilityUnclear riskComment: detailed description of demographic characteristics and sinusitis severity rating at baseline to assess the comparability of the groups for whole randomised population, not separately for clinically evaluable population
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and of such quality that it does not have effect on recovery (thus not cause bias)
Free of other bias? FundingUnclear risk2 of the authors were affiliated with a pharmaceutical company

Siegert 2000

MethodsRandomised, double-blind trial. Diagnosis by clinical signs and symptoms, confirmed by radiograph (mucosal thickening, opacification or air-fluid level) or bacteriologically (either by a swab collected under endoscopic view, cannulation of the middle meatus or a sinus puncture; positive cultures were obtained in 52% of participants). Clinical cure was not defined. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 5%
ParticipantsMulticentre (60 centres), multinational trial (Finland, France, Germany, Greece, Israel, Spain, Sweden). Mean age 40; 220 men, 273 women. ENT co-morbidity not reported
Interventions

2 treatment arms
Group 1: cefuroxime axetil 250 mg twice daily, 10 days
Group 2: moxifloxacin 400 mg daily, 7 days

Nasal decongestants and mucolytic drugs allowed; systemic or topical corticosteroids not allowed

OutcomesClinical outcomes assessed on day 14 in 468 out of 493 randomised (in 241 out of 251 participants in cefuroxime axetil group and in 227 out of 242 participants in moxifloxacin group). Bacteriological outcomes in 224; radiographic outcomes not assessed
Notes2 of the authors were affiliated to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response, bacteriological response for 224 participants and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind study". "Placebo-capsules used"
Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 10/251 (4%) in cefuroxime axetil group and 15/242 (6%) in moxifloxacin group on day 14

Comment: marginal missing data rate

Free of other bias? Baseline comparabilityUnclear riskComment: it appears that the groups were imbalanced in respect of severity of the sinusitis (more participants with severe disease in moxifloxacin group)
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingUnclear risk2 of the authors were affiliated to a pharmaceutical company

Siegert 2003

MethodsRandomised, double-blind trial. Diagnosis by clinical symptoms and radiograph showing at least 6 mm mucosal thickening, opacification or air-fluid level. Sinus material for bacteriological culture by puncture or by middle meatus swab under endoscopic guidance. Clinical cure defined as disappearance of signs and symptoms or significant improvement and no further therapy required. Treatment compliance was not reported. Proportion of the participants without known or reported clinical outcome 19% on day 14 to 23
ParticipantsMulticentre, multinational trial. Participants were enrolled by otolaryngologists at 43 centres in 7 countries: France, Germany, Greece, Israel, Lithuania, Spain and Sweden. Participants were 42 years old on average and included 193 men and 259 women. ENT co-morbidity was not assessed
Interventions2 treatment arms
Group 1: faropenem daloxate 300 mg twice daily for 7 days
Group 2: cefuroxime axetil 250 mg twice daily for 7 days
Decongestants, antihistamines and mucolytics were allowed. Intranasal or oral corticosteroids were not permitted
OutcomesStudy reported clinical responses for clinically evaluable patients only. Clinical outcomes assessed on day 14 to 23 in 452 out of 558 randomised participants (in 228 out of 279 participants in faropenem daloxate group and in 224 out of 279 participants in cefuroxime axetil group). Clinically evaluable population was defined as patients who had clinically and radiologically confirmed diagnosis, did not receive any concomitant antimicrobials and/or corticosteroids, and were assessed on day 14 to 23. Bacteriological outcomes were assessed in 136 participants
NotesStudy was supported by a grant from a pharmaceutical company. 2 of the authors were affiliated to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Patients were randomly assigned 1:1 to one of two treatment groups using a block design computer-generated random code"
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and bacteriological responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind study". "For blinding purposes, matched placebos were used in one group and encapsulated tablets in the other group"
Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear riskMissing data: 51/279 (18%) in faropenem daloxate group and 55/279 (20%) in cefuroxime axetil group on day 14 to 23
Comment: reasons for missing data across groups insufficiently described
Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingHigh riskStudy was supported by a grant from a pharmaceutical company. 2 of the authors were affiliated to a pharmaceutical company

SSG 1993

MethodsRandomised, double-blind trial. Diagnosis was established by purulent rhinorrhoea, a radiograph consistent with sinusitis or a sinus puncture and that isolated a bacterial pathogen (about 50% of participants had a pathogen). Clinical outcomes were assessed by an investigator. Clinical cure was defined as resolution of pretreatment signs and symptoms and no recurrence within 72 hours post-treatment; improvement was defined as a reduction in the number or severity of signs and symptoms but without complete resolution. Compliance with treatment was not reported. Proportion of the participants without known or reported clinical outcome 2%
ParticipantsParticipants were recruited from outpatient otolaryngology clinics. Demographic data reported only for evaluable participants (defined as participants with isolated bacterial pathogen plus no protocol violations): mean age approximately 35 (range 16 to 73); 186 men and 152 women. ENT co-morbidity was not assessed. Countries - Sweden, Finland, Iceland
Interventions2 treatment arms
Group 1: loracarbef 400 mg twice daily for 10 days
Group 2: doxycycline 200 mg first dose followed by 100 mg daily for 10 days
Adjuvant therapies were not described
OutcomesClinical outcomes were assessed in 648 out of 662 randomised participants (in 323 out of 330 participants in doxycycline group and in 325 out of 332 participants in loracarbef group). Outcomes were assessed on day 7 to 13 after randomisation. Radiographic outcomes were not reported. Bacteriological outcomes were assessed in 324 participants
NotesStudy was financially supported by a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response, bacteriological response for those participants who had positive culture at baseline and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind". "Participants in the doxycycline group also received a capsule containing a placebo in order to ensure blinding of both the participants and observers"
Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 7/332 (2%) in loracarbef group and 7/330 (2%) in doxycycline group

Comment: minimal missing data rate

Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability of intervention and control groups at baseline reported only from evaluable participants which was half of the enrolled population
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy was financially supported by a pharmaceutical company

Stefansson 1998

MethodsRandomised, double-blind trial. Diagnosis by symptoms and sinus opacity or air-fluid level on radiograph. Cure defined as clinical signs and symptoms improved or resolved at post-treatment (11 to 13 days) and radiographically confirmed absence at follow-up (38 to 45 days). Treatment compliance was not reported. Proportion of the participants without known or reported clinical outcome 4% on day 11 to 13
ParticipantsStudy setting not described. Mean age 37; 157 men and 213 women. ENT co-morbidity was not assessed. Countries - 8 European, Middle Eastern and African
Interventions2 treatment arms
Group 1: cefuroxime axetil 250 mg twice daily for 10 days
Group 2: clarithromycin 250 mg twice daily for 10 days
Nasal corticosteroids prohibited; other adjuvant treatments not described
OutcomesClinical outcomes were assessed on day 11 to 13 in 354 out of 370 randomised participants. Second follow-up on day 38 to 45 in 310 participants
Notes2 of the authors were affiliated to a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Unclear risk

Outcomes reported: clinical response (including also radiological findings at second follow-up) and adverse effects

Comment: the radiological responses are reported also separately. It remains unclear whether radiological response was an independent outcome or not

Blinding? (Participant and investigator)Low risk

Quote: "Double-blind study". "Placebo-tablets used"

Comment: although the description of the blinding was incomplete, this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described

Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 10/185 (5%) in cefuroxime axetil group and 6/185 (3%) in clarithromycin group at 11 to 13 days (primary outcome)

Comment: this assessment has been based on the numbers given in a table. Although the corresponding exact numbers given in the text were slightly different, this domain was graded 'low' risk of bias because the missing data rate is marginal

Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingUnclear risk2 of the authors was affiliated to a pharmaceutical company

Sterkers 1997

MethodsRandomised, unblinded design. Diagnosis by symptoms (duration not described) and fluid or opacity on radiograph. Clinical cure defined as disappearance of purulent nasal discharge. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 13%
ParticipantsMulticentre, but settings not described. Mean age approximately 41; 178 men and 280 women. Country - France
Interventions3 treatment arms: (only group 1 - the standard dosing and group 3 used in the analysis)
Group 1: ceftibuten 400 mg once daily for 8 days
Group 2: ceftibuten 200 mg twice daily for 8 days
Group 3: amoxicillin-clavulanate (500 mg/125 mg) 3 times daily
Adjuvant treatments not described
OutcomesOverall, clinical and radiographic outcomes reported on day 10 in 400 out of 458 randomised participants (in 134 participants out of 152 in ceftibuten 400 mg once daily group; 138 out of 157 in ceftibuten 200 mg twice daily group and 128 out of 149 participants in amoxicillin-clavulanate group); 50 excluded post-randomisation due to negative radiograph. Second follow-up on day 40 after randomisation
NotesFunding source not identified
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and radiographic responses, and adverse affects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Open study"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Unclear risk

Total missing data 13% on day 10 and 14% on day 40

Comment: no description of reasons for missing data across groups

Free of other bias? Baseline comparabilityUnclear riskComment: information on comparability was insufficient to assess whether the groups were comparable at baseline or not
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingUnclear riskFunding source not identified

Steurer 2000

MethodsRandomised, investigator-blinded trial. Diagnosis by clinical signs and symptoms, confirmed by a radiograph (criteria not specified). Sinus aspirate and culture (1 or more pathogens were isolated from the maxillary sinus aspirates of 66% participants). Clinical cure defined dichotomously - "cure" or "failure". Compliance not reported. Proportion of randomised participants without known clinical outcome was not reported. Study reported only clinical cure rates for intention-to-treat population (all randomised participants) and for "evaluable" population (participants all of whose clinical and microbial data were available). The data from this study were not used in the meta-analyses of this review because all non-cures in the ITT population (11%) were categorised as failures in analyses (it remains unclear how many responses were true failures in each group and how much was missing data); the missing data in the "evaluable" population was high (52%)
ParticipantsParticipants recruited from 16 medical centres throughout Europe. Setting not described. Mean age 31; 336 men, 233 women. ENT co-morbidity not reported
Interventions3 treatment arms
Group 1: cefdinir 600 mg daily, 10 days
Group 2: cefdinir 300 mg twice daily, 10 days
Group 3: amoxicillin-clavulanate (500 mg/125 mg) 3 times daily, 10 days
Topical nasal decongestants with xylomethazoline, oxymethazoline or naphazoline were permitted but not prescribed
OutcomesStudy did not report description of missing data. Study reported only clinical cure rates for intention-to-treat population and all clinical responses were known only for "evaluable" populations. Clinical outcomes for "evaluable population" were reported on day 17 to 25 and on day 31 to 35 in 295 out of 569 randomised participants (in 93 participants out of 182 in cefdinir 600 mg once daily group, in 96 out of 198 in cefdinir 300 mg twice daily group and in 106 out of 189 participants in amoxicillin-clavulanate group). Radiographic outcomes reported on day 17 to 25 for 295. Bacteriological outcomes reported in 248
NotesStudy sponsored by a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskQuote: "For each study centre an independent randomized schedule was prepared"
Allocation concealment (selection bias)Low riskQuote: "Study monitored by a pharmaceutical company. Medication was dispensed by a third person and all records concerning medications were kept in a separate location"
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical, radiological and bacteriological responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Participants were given consecutive patient numbers after screening. As this was investigator-blinded study, medication was dispensed by a third person and all records concerning medications were kept in a separate location. The participants were instructed not to reveal the type of medication to the investigator"
Comment: no blinding of the participants
Incomplete outcome data addressed? (Clinical efficacy outcomes)High risk

Missing data: 89/182 (49%) in cefdinir 600 mg once daily group, 102/198 (52%) in cefdinir 300 mg twice daily group, and 83/189 (44%) in amoxicillin-clavulanate group on day 17 to 25

Comment: missing data rate over 20%

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingHigh riskStudy sponsored by a pharmaceutical company

Sydnor 1992

MethodsRandomised, single-blind (investigator). Diagnosis was confirmed by a clinical diagnosis of acute sinusitis and a radiograph consistent with sinusitis (showing mucosal thickening at a minimum). A bacteriologic culture of aspirated maxillary sinus fluid; evaluable patients were required to have positive cultures of pathogens susceptible to both study antibiotics. Clinical cure was not defined. Compliance with therapy was not reported. Total proportion of the participants without known clinical outcome 62%. Participants were excluded mostly due to bacteriologic reasons; exclusions because of other reasons marginal 2%
ParticipantsThe study setting was not well described. Participants' mean age was approximately 38 (range 18 to 78); 39 men and 74 women. Country - USA
Interventions2 treatment arms
Group 1: loracarbef 400 mg twice daily for 7 to 10 days
Group 2: amoxicillin clavulanate 500 mg/125 mg 3 times daily for 7 to 10 days
No information provided on symptomatic medication for sinusitis
OutcomesClinical outcomes were assessed on day 7 to 13 in 43 of 113 randomised participants (in 22 out of 59 participants in loracarbef group and in 21 out of 54 participants in amoxicillin-clavulanate group). Second follow-up at 1 to 2 weeks after the conclusion of therapy for those participants with favourable symptomatic responses at the first follow-up. Radiographic outcomes were not reported; bacteriological outcomes were reported for 28 participants
NotesStudy supported by grants from a pharmaceutical company. 1 of the authors was affiliated to the same pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and bacteriologic responses, and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)High riskQuote: "Investigator-blinded"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Exclusions from analyses: bacteriologic reasons 36/59 in loracarbef group and 32/54 in amoxicillin clavulanate group; other exclusions 1 (2%) and 1 (2%) of participants, respectively. (Bacteriologic reasons included resistant causative organisms, negative pre-therapy cultures for pathogens, and no or low colony counts on pre-therapy isolates)

Comment: the bacteriologic reasons were seen not to cause bias. Other missing data marginal

Free of other bias? Baseline comparabilityLow riskComment: detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsUnclear riskNo information provided
Free of other bias? FundingHigh riskStudy supported by grants from a pharmaceutical company. 1 of the authors was affiliated to the same pharmaceutical company

Upchurch 2006

MethodsRandomised, double-blind design. Diagnosis by clinical signs and symptoms > 7 days but < 28 days, confirmed by radiograph showing air fluid level, opacification and/or at least 6 mm mucosal thickening. Clinical outcomes were assessed by an investigator and clinical cure was defined as resolution or improvement of clinical symptoms and signs such that no additional antimicrobial treatment was necessary. Treatment compliance 91%. Proportion of the participants included in the efficacy-valid population was 78% on day 17 to 31
ParticipantsMulticentre trial (57 centres) in USA and Canada. 1099 outpatients, 481 male and 618 women. Participants were 43 years old on average (range 15 to 90). 37% of participants had a history of allergic rhinitis
Interventions3 treatment arms:
Group 1: faropenem medoxomil 300 mg twice daily for 7 days
Group 2: faropenem medoxomil 300 mg twice daily for 10 days
Group 3: cefuroxime axetil 250 mg twice daily for 10 days
Oral or nasal decongestants or antihistamines were allowed; systemic or topical corticosteroids not permitted
OutcomesStudy reported clinical cure rates for intention-to-treat and efficacy-valid populations. However, only efficacy-valid population could be used as basis for clinical failure rate calculating in this review because true failures were not reported and both indeterminate and missing responses were included in failures in ITT analyses. Clinical outcomes for efficacy-valid population were assessed on day 17 to 31 in 861 out of 1106 randomised participants (in 295 out of 370 participants in faropenem medoxomil for 7 days group, in 280 out of 365 participants in faropenem medoxomil for 10 days group, and in 286 out of 371 participants in cefuroxime axetil group). Efficacy-valid population was defined as participants who met inclusion criteria, had radiologically confirmed sinusitis, study drug given for a minimum of 72 hours for participants considered treatment failures and a minimum of 5 days to be a treatment success, adequate adherence documented with over 80% of medication, absence of any protocol violation influencing treatment efficacy, and no essential data missing or indeterminate. Second follow-up 28 to 38 days post-treatment. Radiographic outcomes not reported
NotesFunding for trial provided by a pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "Participants were randomly assigned in a 1:1:1, double-blind fashion by using a computer-generated random code to one of three oral-treatment groups"
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical response and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Double-blind study". "Because faropenem medoxomil could not be encapsulated, a double-dummy design was used to maintain blinding". "Matching placebo-capsules used"
Comment: although the description of blinding was incomplete (no description of investigator blinding), this domain was graded 'low' risk of bias because the study was reported to be double-blind and blinding method was partly described
Incomplete outcome data addressed? (Clinical efficacy outcomes)High risk

Missing data: 75/370 (20%) in faropenem medoxomil for 7 days group, 85/365 (23%) in faropenem medoxomil for 10 days group and 85/371 (23%) in cefuroxime axetil group on day 17 to 31

Comment: missing data rate over 20%

Free of other bias? Baseline comparabilityLow riskQuote: "Demographic and baseline medical characteristics for the efficacy-valid population were generally consistent with those of the intent-to-treat population"
Comment: although data on baseline comparability were only shown for the ITT population (not separately for efficacy-valid population), this domain was graded 'low' risk of bias because the authors stated that also the efficacy-valid groups were comparable
Free of other bias? Co-interventionsLow riskComment: no co-interventions included in the protocol. In both groups the same concomitant medication was allowed and was of such quality that it did not have an effect on recovery (thus did not cause bias)
Free of other bias? FundingHigh riskFunding for trial provided by a pharmaceutical company

van Buchem 1997

MethodsRandomised, double-blind design. Diagnosis made clinically (the mean duration of the symptomatic period before treatment was 2.2 weeks), confirmed by radiograph showing > 5 mm mucosal thickening, opacity or air-fluid level. Clinical cure was not defined. Treatment compliance reported as 98% assessed by pills taking by participants. Proportion of the participants without known or reported clinical outcome 4% on day 14
ParticipantsParticipants were recruited from community-based, general medical practices. Mean age was 34; 79 men and 135 women. ENT co-morbidity was assessed; approximately 12% had allergic disease. Country - The Netherlands
Interventions2 treatment arms
Group 1: placebo 3 times daily for 7 days
Group 2: amoxicillin 750 mg 3 times daily for 7 days
Adjuvant therapy with oxymetazoline steam inhalation (duration not specified). Paracetamol as needed
OutcomesClinical and radiographic outcomes were assessed on day 14 for 206 out of 214 participants randomised (in 105 out of 108 participants in antibiotic group and in 101 out of 106 participants in placebo group). Long-term relapse rates during 1 year. Bacteriological outcomes were not assessed
NotesThere was no identified funding source
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Low riskQuote: "The randomization of allocation of the amoxycillin or placebo (distributed in identical bottles) was computer generated"
Allocation concealment (selection bias)Low riskQuote: "Randomisation of patients was carried out and capsules were provided by the hospital pharmacy in the hospital were patients were to see the ENT specialist. The code of the allocation schedule was kept in the office of the head of the hospital pharmacy, and was broken prematurely only if a severe clinical development or severe adverse effects occurred"
Comment: central allocation
Selective reporting (reporting bias)Low risk

Outcomes reported: clinical and radiological responses, relapses, chronic evolution and adverse effects

Comment: prespecified outcomes (in methods) were reported in prespecified way

Blinding? (Participant and investigator)Low riskQuote: "Neither the pharmacist's assistant who distributed the bottles (identical bottles for both groups), the patient, nor the ENT specialist was aware of the nature of the medication"
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 2.7% (3/108) in antibiotic group and 4.7% (5/106) in placebo group at 14 days

Comment: marginal missing data rate

Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: both groups received the same co-intervention during the trial
Free of other bias? FundingUnclear riskNo information provided

von Sydow 1995

  1. a

    ABMS: acute bacterial maxillary sinusitis
    BID: twice daily
    CT: computerised tomography (scan)
    ENT: ear, nose and throat
    ITT: intention-to-treat
    mITT: modified intention-to-treat

MethodsRandomised, double-blind design. Diagnosis based on clinical symptoms and radiograph showing air-fluid level or opacity (verified by sinus aspiration). Failure defined as clinical signs and sinus X-ray points unchanged or worsened. Treatment compliance not reported. Proportion of the participants without known or reported clinical outcome 10%
ParticipantsParticipants recruited from academic affiliated, otolaryngology practices. Mean age 33 (range 18 to 78); 149 men and 137 women. ENT co-morbidity was not assessed. Countries - Sweden, Finland, Norway
Interventions2 treatment arms
Group 1: cefpodoxime proxetil 200 mg twice daily for 10 days
Group 2: amoxicillin 750 mg twice daily for 10 days
All participants received xylometazoline (nasal decongestant) for up to 7 days
OutcomesClinical outcomes were assessed on day 11 to 20 in 258 out of 286 participants randomised (in 130 out of 143 participants in cefpodoxime proxetil group and in 128 out of 143 participants in amoxicillin group). Bacteriological outcomes reported in 150 participants
NotesStudy was financially supported by a pharmaceutical company. 1 of the authors had affiliation to pharmaceutical company
Risk of bias
BiasAuthors' judgementSupport for judgement
Random sequence generation (selection bias)Unclear riskNo information provided
Allocation concealment (selection bias)Unclear riskNo information provided
Selective reporting (reporting bias)High risk

Outcomes reported: clinical response (including radiological findings), bacteriological response for 150 participants and adverse effects

Comment: in the protocol it is stated that the second follow-up would be at days 28 to 35. Those results are, however, not reported

Blinding? (Participant and investigator)Unclear riskQuote: "Double-blind, double-dummy study"
Comment: no description of the blinding procedure
Incomplete outcome data addressed? (Clinical efficacy outcomes)Low risk

Missing data: 13/143 (9.1%) in cefpodoxime proxetil group and 15/143 (10.5%) in amoxicillin group

Comment: although description of the reasons for the missing data by group was unclear this domain was graded 'low' risk of bias because the total missing data rate was moderate and balanced

Free of other bias? Baseline comparabilityLow riskComment: very detailed description of demographic characteristics and sinusitis severity rating with which to assess the comparability of the groups at baseline
Free of other bias? Co-interventionsLow riskComment: both groups received the same co-intervention during the trial
Free of other bias? FundingHigh riskStudy was financially supported by a pharmaceutical company. 1 of the authors had affiliation to a pharmaceutical company

Characteristics of excluded studies [ordered by study ID]

StudyReason for exclusion
  1. a

    RCT: randomised controlled trial
    URTI: upper respiratory tract infection

Agbim 1974Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis by clinical signs and symptoms but the minimum duration of those before study entry not reported). Results not reported separately for participants with acute sinusitis
Alvart 1992Antibiotic versus antibiotic - study design. Sample size for acute sinusitis < 30
Axelsson 1971Antibiotic versus antibiotic - study design. Not randomised study. No clinical outcomes
Axelsson 1973Antibiotic versus antibiotic - study design. Not randomised study
Axelsson 1975Antibiotic versus antibiotic - study design. Not randomised study. No clinical outcomes, only radiological outcomes
Axelsson 1981Antibiotic versus antibiotic - study design. Not randomised study
Bandak 1999Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis by clinical signs and symptoms but the minimum duration of those before study entry not reported)
Beatson 1985Antibiotic versus antibiotic - study design. Inclusion criteria for this review not fulfilled (no definition reported for acute sinusitis)
Bockmeyer 1994Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis by clinical signs and symptoms but the minimum duration of those before study entry not reported; bacteriological specimens taken from nasal discharge)
Brodie 1989Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis mainly by clinical signs and symptoms but the minimum duration of those before study entry not reported)
Brook 2006Antibiotic versus antibiotic - study design. Not randomised study. No clinical outcomes
Bucher 2003Antibiotic versus placebo - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis based on clinical findings; only 32% of participants had symptoms for 7 days or longer before study entry)
Carenfelt 1975Antibiotic versus antibiotic - study design. Random allocation not stated
Casiano 1991Antibiotic versus antibiotic - study design. Insufficient information on diagnostic criteria for acute sinusitis
Correa 1986Antibiotic versus antibiotic - study design. Clinical outcomes measured are clinical plus bacteriological; they are not separated individually
De Abate 1992Antibiotic versus antibiotic - study design. Inclusion criteria for this review not fulfilled (no definition reported for acute sinusitis)
De Sutter 2002Antibiotic versus placebo - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis based on clinical findings; about 50% of participants with symptoms less than 7 days at study entry). The information obtained from the authors
Edelstein 1993Antibiotic versus antibiotic - study design. Approximately 50% of participants with acute exacerbation of chronic sinusitis. Results not reported separately for acute maxillary sinusitis
Ekedahl 1987Antibiotic versus antibiotic - study design. No clinical outcomes
Insufficient information on intervention groups
Elies 2001Antibiotic versus antibiotic - study design. Inclusion criteria for this review not fulfilled; proportion of included complicated cases (frontal sinusitis or pansinusitis) was not reported
Falser 1988Antibiotic versus antibiotic - study design including participants with otitis and sinusitis. Results not reported separately for participants with sinusitis
Federspil 1983Antibiotic versus antibiotic - study design. Mixed ENT infections. Sample size for acute sinusitis < 30
Felstead 1991Antibiotic versus antibiotic - study design. The criteria for diagnosis of acute maxillary sinusitis not clearly stated
Fiscella 1991Antibiotic versus antibiotic - study design. Sample size for acute maxillary sinusitis < 30 (the total number of the participants was > 30, but about 50% of participants had acute exacerbation of chronic sinusitis)
Freche 1988Antibiotic versus antibiotic - study design. The definition of acute maxillary sinusitis unclear
Gananca 1973Antibiotic versus placebo - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (definition of sinusitis by clinical signs and symptoms or by X-ray not reported; bacteriological specimens taken from purulent nasal discharge, site not specified)
Gananca 1977Antibiotic versus placebo - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis based on clinical findings without detailed description of signs and symptoms and duration)
Garcia 1998Antibiotic versus antibiotic - study design. Sample size for acute sinusitis < 30
Gladkov 1979Antibiotic versus dexamethazone and formaldehyde - study design. Randomisation not stated. Sample size for acute sinusitis < 30
Goumas 1997Antibiotic versus antibiotic - study design. Quasi-randomised design
Gurdogan 2001Antibiotic versus antibiotic - study design. Sample size for acute sinusitis < 30
Gwaltney 1981Antibiotic versus antibiotic - study design. Randomised trial for some of the participants. Out of 113 participants randomised only 32 had outcomes reported. Uncertain if outcomes relate only to those randomised
Hansen 2000Antibiotic versus placebo - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis based on clinical findings; at least 50% of participants had symptoms lasting for 6 days or less at study entry)
Hebblethwaite 1987Antibiotic versus antibiotic - study design. Diagnostic criteria for acute maxillary sinusitis insufficiently stated
Henry 1999aAntibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (the duration of symptoms before study entry not reported; the diagnosis in 30% of participants not confirmed by X-ray)
Henry 1999bAntibiotic versus antibiotic - study design. Unclear data, contradictory results in table and text
Husfeldt 1993Antibiotic versus antibiotic - study design. Random allocation not stated
Jeppesen 1970Antibiotic versus placebo - study design. Quasi-randomised design. Puncture and irrigation (every second day) used as co-intervention. Sample size for acute maxillary sinusitis < 30
Johnson 1999Antibiotic versus antibiotic - study design. Approximately 30% of participants had acute exacerbation of chronic sinusitis. Results not reported separately for acute maxillary sinusitis
Karpov 1998Antibiotic versus antibiotic - study design
Puncture used as co-intervention
Klein 1998Antibiotic versus antibiotic - study design. Approximately 60% of participants had acute exacerbation of chronic sinusitis. Results not reported separately for acute maxillary sinusitis
Lacroix 2002Study mainly designed to identify signs to predict the presence of bacteria in acute rhinosinusitis (including common cold and acute sinusitis). Inclusion criteria for the study did not fulfil the inclusion criteria for this review (69% of patients had other URTIs than sinusitis confirmed by X-ray; median duration of clinical symptoms 4 days for all randomised patients; data not reported separately for patients with acute sinusitis)
Li 2000Antibiotic versus antibiotic - study design. Approximately 67% of participants had chronic sinusitis. Randomisation unclear
Lopez 1975Sample size for acute sinusitis < 30 (7 participants)
Luchikhin 2005Antibiotic versus antibiotic - study design. Sample size for acute maxillary sinusitis < 30
Mannhardt 1980Antibiotic versus antibiotic - study design. Mixed sample of acute (n = 30) and chronic (n = 22) sinusitis. Only combined results are reported
Manzini 1993Antibiotic versus antibiotic - study design. Insufficient information on diagnostic criteria for acute maxillary sinusitis
Marchi 1990Antibiotic versus antibiotic - study design. Random allocation not stated
Marcolino 1999Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (clinical symptoms lasting at least 4 weeks)
Mesure 1973Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (no definition of acute maxillary sinusitis). Sample size for acute sinusitis < 30
Mira 2001Antibiotic versus antibiotic - study design. Sample size for acute sinusitis < 30. Insufficient information on diagnostic criteria for acute maxillary sinusitis
Miyamoto 2005Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis on clinical basis; the duration of symptoms before study entry not stated)
Moorhouse 1985Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis on clinical basis; the duration of symptoms before study entry not stated)
Norrelund 1978Antibiotic versus placebo - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis on clinical basis; the duration of symptoms before study entry not stated)
Olsson 1976Antibiotic versus antibiotic - study design. Questionable random allocation. Sample size for acute sinusitis < 30
Osman 1983Antibiotic versus antibiotic - study design. Diagnostic criteria for acute maxillary sinusitis not stated
Pallestrini 1995Antibiotic versus antibiotic - study design. Approximately 27% of participants had acute exacerbation of chronic sinusitis. Results not reported separately for acute maxillary sinusitis
Panosetti 1992Antibiotic versus antibiotic - study design. Hospitalised participants, ENT ward
Peyramond 1991Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis on clinical basis; duration of symptoms before study entry not stated)
Piragine 1988Antibiotic versus antibiotic - study design. Sample size for acute sinusitis < 30 (7 participants)
Podvinec 1982Antibiotic versus antibiotic - study design. Mixed sample of acute (< 80%) and chronic sinusitis. Results not reported separately
Polonovski 2006Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis on clinical basis, median duration of symptoms 4 days before study entry)
Quick 1973Antibiotic versus antibiotic - study design. Random allocation not stated. Sample size for acute sinusitis < 30
Quick 1975Antibiotic versus antibiotic - study design. Random allocation not stated. Sample size for acute sinusitis < 30
Rantanen 1973Antibiotic versus no antibiotic - study design. Study design not suitable for this review because weekly irrigation was used as co-intervention. Random allocation not stated
Rechtweg 2004Antibiotic versus antibiotic - study design. Sample size for acute sinusitis < 30
Riedel 1987Study design comparing antibiotic + serrapeptase versus serrapeptase. Not a randomised study
Rimmer 1997Antibiotic versus antibiotic - study design. Insufficient information on diagnostic criteria for sinusitis
Roenning 1987Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review
Overall clinical outcomes not reported
Sabater 1995Antibiotic versus antibiotic - study design. Assessment of the study was based on the abstract. Additional information was requested from the authors to assess the adequacy of the study for this review but the authors were not able to deliver the necessary information
Salmi 1986Antibiotic versus antibiotic - study design. Sample size for acute sinusitis < 30
Salmi 1993Antibiotic versus antibiotic - study design. Mixed respiratory tract infections. Sample size for acute sinusitis < 30. Insufficient information on diagnostic criteria for acute maxillary sinusitis
Schering-Plough 2003Unpublished study (assessment of the study based on the study synopsis which was obtained from the pharmaceutical company). The study was a randomised clinical trial comparing intranasal corticosteroid with antibiotic and placebo. In the synopsis, definition of outcome and data were not in a usable form for this review
Serra 2007Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (sinusitis not defined by clinical signs and symptoms; microbiological evaluation reported to be based on specimens taken by swab without specification of where they were taken from)
Spindler 1985Antibiotic versus antibiotic - study design. Inclusion criteria for this review not fulfilled (criteria for acute sinusitis not reported)
Stahl 1989Antibiotic versus antibiotic - study design. Included also other URTI than sinusitis. Inclusion criteria for this review not fulfilled: criteria for acute sinusitis not reported, the proportion of participants with chronic sinusitis > 20% and clinical outcomes for acute sinusitis not reported separately
Stalman 1997aAntibiotic versus placebo - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis based on clinical findings; duration of symptoms reported to have lasted at least 5 days). Additional information was requested from the authors about the proportion of the participants with symptoms for less than 7 days but the authors were not able to deliver the information
Strachunskii 1993Antibiotic versus antibiotic - study design. Not randomised study
Sydnor 1989Antibiotic versus antibiotic - study design. No clinical outcomes
Söderström 1991Antibiotic versus antibiotic - study design. Clinical outcomes not measured for sinusitis (included also other URTI)
Taub 1967RCT comparing bromelains + antibiotics to antibiotics
Diagnostic criteria for sinusitis not stated. Sample size for acute maxillary sinusitis < 30
Varonen 2003aAntibiotics versus placebo - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (about 27% of participants with symptoms for less than 6 days at study entry. Ultrasound positive about 50% of participants)
von Sydow 1984Antibiotic versus antibiotic - study design. Insufficient information (for example, the follow-up time and duration of the therapy were not stated)
Wallace 1985Antibiotic versus antibiotic - study design. Mixed respiratory tract infections. Insufficient information on diagnostic criteria for acute maxillary sinusitis
Weis 1998Antibiotic versus antibiotic - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis by clinical signs and symptoms but the minimum duration of those before study entry not reported)
Westerman 1975Antibiotic versus antibiotic - study design. Insufficient information on diagnostic criteria for acute maxillary sinusitis
Williamson 2007Antibiotic versus placebo - study design. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis based on clinical findings; only 57% of participants had symptoms for 7 days or longer at study entry). The information was obtained from the authors
Young 2003The study design mainly intended to examine the diagnostic indicators for acute bacterial rhinosinusitis than comparing antibiotic versus placebo. Diagnostic criteria for acute sinusitis did not fulfil the inclusion criteria for this review (diagnosis by clinical signs and symptoms; median duration of symptoms 4 days)
Zbären 1983Antibiotic versus antibiotic - study design. Sample size for acute maxillary sinusitis < 30