It is estimated that 2% to 5% of all general practice consultations are eye related (Dart 1986; McDonnell 1988; McCormick 1995). Data from Norway suggest that acute infective conjunctivitis is suspected in approximately 3% of patients seen in general medical practice, with this diagnosis being correct in two-thirds of patients (Hovding 1991; Hovding 2008). Acute infective conjunctivitis is therefore one of the most frequently encountered ocular disorders in primary care. Such infection is usually viral or bacterial in aetiology. Infection of the conjunctiva produces a number of local symptoms including red eyes, discharge and discomfort.
Viral conjunctivitis is usually caused by adenovirus infection, which is contagious and is responsible for many of the epidemics that occur in school-aged children, but is in most cases self limiting; a small proportion of patients may, however, develop longer-lasting visual problems.
Bacterial conjunctivitis is commonly due to infection with Haemophilus influenzae, Streptococcus pneumoniae or Staphylococcus aureus. Bacterial conjunctivitis is also regarded as self limiting. However, antibiotics are generally considered desirable on the grounds that they seem to speed recovery, reduce relapse and may prevent important sight-threatening complications such as orbital cellulitis, keratitis and panophthalmitis.
As bacterial and viral conjunctivitis may be difficult to differentiate on clinical grounds, and eye swabs are not considered practical (on grounds of delay and cost), many doctors will treat all presumed cases of infective conjunctivitis with a broad-spectrum antibiotic.
Topical antibiotic treatments are most commonly used; some of these also contain topical steroid therapy. Systemic antibiotic therapy has been advocated by some (Wald 1997) in order to prevent the development of 'conjunctivitis-otitis syndrome' (i.e. conjunctivitis followed by acute otitis media).
There has been significant public and professional concern regarding the use of chloramphenicol eye drops because of the associated rare risk of bone marrow aplasia. This concern led to a considerable reduction in its use in the United States and there have been calls for its use in the United Kingdom (UK) to also be curtailed (Doona 1995). More recent work, however, suggests that these risks of bone marrow aplasia may have been over-stated (Lancaster 1998). The widespread use of broad-spectrum antibiotics has also led to some concerns that antibiotic resistance may become a significant problem.
The management of common infections encountered in primary care has undergone a radical transformation over the past 15 years. Previously it would have been regarded as heretical to question the use of antibiotics for infections such as sinusitis, otitis media and sore throat (pharyngitis/tonsillitis), which was near universally routine. Randomised controlled trials and systematic reviews have since cast doubt on the effectiveness and cost-effectiveness of antibiotic therapy for these conditions (Del Mar 2004; Glasziou 2004).
In view of these developments, it is timely to assess whether antibiotic therapy confers significant benefit in the treatment of acute bacterial conjunctivitis. This review reports an updated assessment of the question, incorporating data on newer available treatments (McLean 2010) and in so doing updates previous versions of this review (Hurwitz 2005; New Reference; New Reference; New Reference).
To analyze evidence from double-masked, randomised, placebo-controlled trials to ascertain if antibiotic therapy (topical or systemic) confers any benefit in the management of acute bacterial conjunctivitis.
Criteria for considering studies for this review
Types of studies
We included double-masked, randomised, placebo-controlled trials.
Types of participants
Participants were people with acute bacterial conjunctivitis, aged one month or older. The diagnosis of bacterial conjunctivitis may have been on clinical or microbiological grounds. 'Acute' was defined as symptoms of less than four weeks duration.
Types of interventions
We included studies in which any form of antibiotic treatment was compared with placebo in the management of acute bacterial conjunctivitis; this included topical, systemic and combination treatments (for example, one or more antibiotics, antibiotics and steroids).
Types of outcome measures
We considered the following outcome measures.
1. Time to clinical cure
2. Time to microbiological cure
1. Recurrence of infection within four weeks
2. Cost-effectiveness of treatment
3. Compliance with treatment and number of drop-outs
4. Number of participants that experience complications of acute bacterial conjunctivitis
5. Adverse outcomes as reported in trials
Search methods for identification of studies
We searched the Cochrane Central Register of Controlled Trials (CENTRAL) 2012, Issue 7, part of The Cochrane Library. www.thecochranelibrary.com (accessed 18 July 2012), MEDLINE (January 1950 to July 2012), EMBASE (January 1980 to July 2012), OpenGrey (System for Information on Grey Literature in Europe) (www.opengrey.eu/), the metaRegister of Controlled Trials (mRCT) (www.controlled-trials.com), ClinicalTrials.gov (www.clinicaltrials.gov) and the WHO International Clinical Trials Registry Platform (ICTRP) (www.who.int/ictrp/search/en). We did not use any date or language restrictions in the electronic searches for trials. We last searched the electronic databases on 18 July 2012.
See: Appendices for details of search strategies for CENTRAL (Appendix 1), MEDLINE (Appendix 2), EMBASE (Appendix 3), OpenGrey (Appendix 4), mRCT (Appendix 5), ClinicalTrials.gov (Appendix 6) and the ICTRP (Appendix 7).
Searching other resources
We searched the reference lists of identified trial reports to find additional trials and we contacted the first author of identified trials to ask for information on further trials. We used the Science Citation Index to find studies which had cited the identified trials.
Correspondence with Alcon Laboratories helped to clarify the origin and publication of three trials referred to in a French study (Kodjikian 2010).
We contacted investigators and pharmaceutical companies to identify additional published and unpublished studies. Using MIMS (http://www.emims.net/) we identified the following pharmaceutical companies as producers of relevant ophthalmic preparations and contacted them for information about additional relevant trials: Chauvin Pharmaceuticals; Goldshield Pharmaceuticals; Leo Laboratories; Wyeth Laboratories; Hoechst Roussel; Schering-Plough; Roche Products; Dominion Pharma; Alcon Laboratories; Florizel; Allergan; Rhone-Poulenc Rorer; Typharm.
Data collection and analysis
Selection of studies
In New Reference and New Reference two authors (AS and BH) independently checked the titles and abstracts resulting from the searches.
In this update (2012) we downloaded the search results into Endnote (a citation manager) and de-duplicated. Relevant titles and abstracts were then selected by UN, AS and SM. UN then obtained the full text of any report referring to possibly or definitely relevant trials. UN and SM assessed all full-text articles using the above-described 'Criteria for considering studies for this review'. Only trials meeting these criteria were then assessed for methodological quality.
Data extraction and management
One author extracted data using a template developed for use by the Cochrane Eyes and Vision Group and entered the data in to RevMan (Review Manager 2011). A second author compared the extraction to the original reports. Due to the paucity of appropriate data it was only possible to consider the following outcomes:
- clinical remission rate:
- early (days two to five post-intervention);
- late (days six to 10).
- microbiological remission rate:
- early (days two to five post-intervention);
- late (days six to 10).
Assessment of risk of bias in included studies
We assessed trial methodological quality according to methods set out in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
These are the six parameters/questions used to assess risk of bias.
- Was the allocation sequence adequately generated?
- Was allocation adequately concealed?
- Were the patients and researcher masked (blinded) to the allocated intervention?
- Were incomplete outcome data adequately addressed?
- Were study reports free of any suggestion of selective reporting?
- Was the study apparently free of other problems than might cause bias?
We deemed questions that were answered 'Yes' to indicate a low risk of bias and considered those answered 'No' at high risk of bias. Sometimes, insufficient information was available to allow a parameter to be assessed in which case a question would be answered as 'Unclear'.
From this assessment, we assigned an overall grade to each study: A – low risk of bias; B - moderate risk of bias; C – high risk of bias.
Two authors independently assessed trial quality and resolved any disagreements by discussion. Authors were not masked to any trial details. We documented authors' agreement on the quality assessment of included trials.
Unit of analysis issues
Acute bacterial conjunctivitis is usually a bilateral problem and all outcomes were assessed at the patient level.
We undertook our quantitative meta-analyses of outcomes on an intention-to-treat basis using random-effects modelling. We expressed the results as risk ratios (RR with 95% confidence intervals (CI)) and risk difference (RD with 95% CI) for dichotomous outcomes. We performed tests for statistical heterogeneity.
We assessed clinical remission with a denominator based on the total number of participants who had been randomised and this was therefore an intention-to-treat analysis. In contrast, we assessed microbiological remission with a denominator based on those patients who had demonstrated microbiological growth in the cultures from their eyes: i.e. 'culture-confirmed' bacterial conjunctivitis. The analysis of clinical cure therefore included patients who were not confirmed to have bacterial conjunctivitis and may have had viral conjunctivitis. This reflects day-to-day clinical practice as many clinicians will choose to start treatment in the absence of culture results.
In our meta-analyses, we included all randomised participants in the reports of clinical outcomes. However, in tests of microbiological cure we used the denominator of those who had had culture-proven bacterial conjunctivitis at the beginning of the study.
Description of studies
Results of the search
The initial electronic searches identified 155 reports of possible trials comparing antibiotics versus placebo in the management of acute bacterial conjunctivitis. We obtained full-text copies of five studies for potential inclusion in the review. Studies by Gigliotti 1984, Leibowitz 1991 and Miller 1992 met the inclusion criteria and were included in the review. Studies by Leibowitz 1976 and Mitsui 1986 were not eligible and were excluded from the review.
Contact with the first authors of identified trials and searching the reference lists of these study reports failed to identify any additional trials.
The searches for the 2002 update identified 211 reports of trials but no new studies were found. The updated searches which covered the period 2002 to January 2006 identified 111 studies and from these, two trials met the inclusion criteria (Rietveld 2005; Rose 2005).
An update search was done in October 2007 which yielded a further 124 reports of studies. The Trials Search Co-ordinator (TSC) scanned the search results and removed any references which were not relevant to the scope of the review. One trial (Everitt 2006) was identified but was excluded as it was not placebo-controlled.
An update search done in July 2012 yielded a further 361 reports of studies. The TSC scanned the search results and removed 168 references which were not relevant to the scope of the review. We screened 193 records and excluded 144 records for not meeting the review criteria. Two review authors (UN and SM) reviewed the remaining 49 abstracts and excluded 45 papers as they did not meet the inclusion criteria for the review. Four new trials were identified and included in the review (Karpecki 2009; Silverstein 2011; Tauber 2011; Tepedino 2009).
Examination of the studies identified in a recent review (McLean 2010) revealed two additional RCTs (Abelson 2008; Gross 2003). We also found a report (Kodjikian 2010) providing outline details of three additional multi-centred, double-masked RCTs (C-00-02; C-00-55; C-01-66) comparing moxifloxacin 0.5% versus vehicle. We contacted Alcon Research Ltd, the pharmaceutical company responsible for producing this report, to obtain further details of peer-reviewed publications and they revealed that the detailed data from these trials had not been published in the peer-reviewed literature (personal communication Notivol 2011).
This review therefore included a total of 11 trials satisfying the inclusion criteria (Abelson 2008; Gigliotti 1984; Gross 2003; Karpecki 2009; Leibowitz 1991; Miller 1992; Rietveld 2005; Rose 2005; Silverstein 2011; Tauber 2011; Tepedino 2009) (see the 'Characteristics of included studies' table for detailed information about the individual trials).
This review was based on trials which had randomised a total of 3673 participants. The breakdown of participants was as follows:
Abelson 2008: 279 participants; Gigliotti 1984: 78 participants; Gross 2003: 73 participants; Karpecki 2009: 269 participants; Leibowitz 1991: 177 participants; Miller 1992: 284 participants; Rietveld 2005: 181 participants; Rose 2005: 326 participants; Silverstein 2011: 202 participants; Tauber 2011: 847 participants; Tepedino 2009: 957 participants.
Abelson 2008 included participants aged from one to 96 years; the trials were heterogeneous in terms of the age groups studied with Gigliotti 1984 including children aged between one month and 18 years and Rose 2005 including children aged six months to 12 years. Karpecki 2009 included participants from one to 92 years; Tepedino 2009 from one to 98 years and Gross 2003 from one to 89 years old. Miller 1992 and Rietveld 2005 included only adults. In Leibowitz 1991 no information was provided regarding the age of participants. In Tauber 2011, participants were aged from one month to 92 years and in Silverstein 2011, participants were aged from one to 86 years old.
The trials differed in the diagnostic inclusion criteria used. Gigliotti 1984 included only those that were swab-positive for Haemophilus influenzae or Streptococcus pneumoniae; Leibowitz 1991 included only those with microbiologically proven bacterial conjunctivitis; Miller 1992 and Karpecki 2009 included those with a clinical diagnosis of acute bacterial conjunctivitis or blepharoconjunctivitis. Abelson 2008, Gross 2003, Rietveld 2005, Rose 2005 and Tepedino 2009 included those with a clinical diagnosis of infective conjunctivitis. Tauber 2011 included participants with a clinical diagnosis of conjunctivitis independent of a positive bacterial culture result at day one. Silverstein 2011 included participants with a clinical diagnosis of conjunctivitis.
In Abelson 2008, a best-corrected visual acuity (BCVA) score of 20/100 or better in each eye was also required for inclusion. Karpecki 2009, Tepedino 2009 and Silverstein 2011 required a pinhole visual acuity ≥ 20/200 in both eyes.
Abelson 2008 examined 1% azithromycin in DuraSite. Gigliotti 1984 examined an ophthalmic ointment containing 10,000 U/gm polymyxin and 500 U/gm bacitracin. Gross 2003 investigated 0.5% moxifloxacin. Leibowitz 1991 studied ciprofloxacin 0.3% eye drops. Miller 1992 examined 0.3% norfloxacin eye drops. Rietveld 2005 studied fusidic acid gel 1%. Rose 2005 examined 0.5% chloramphenicol eye drops. Karpecki 2009, Tepedino 2009 and Silverstein 2011 studied 0.6% besifloxacin ophthalmic suspension. Tauber 2011 studied a new formulation of moxifloxacin containing xanthum gum so as to be retained in the eye.
All of the trials were 'vehicle-controlled', i.e. they used control preparations that did not contain any active antibiotic-related ingredient. However, in some trials the vehicle drop mixture contained an antiseptic, which may have had an impact on outcomes in the placebo arm. The active and placebo ingredients are detailed in Table 1. In order to quantify the effect size of the antiseptic on acute bacterial conjunctivitis it would be necessary to investigate these drops separately against saline controls or no treatment arms. A no treatment arm would be impossible to mask and so would be open to bias. The results of the trials we present are the results measuring the additional effect of the antibiotic over the vehicle drops.
Abelson 2008 used the outcome of clinical resolution which was evaluated at day six or seven. Clinical resolution was considered to be the absence of the three signs of bulbar conjunctival injection, palpebral conjunctival injection and ocular discharge. The secondary outcome was microbiological eradication, also at day six or seven, and was scored categorically from zero (eradicated) to three (worsening) compared with baseline, following gram-staining and other tests.
The trials used different combinations of outcome measures. Gigliotti 1984 used clinical cure and microbiological eradication which were each assessed as a dichotomous outcome: cured or not cured, at two points in time after treatment was commenced: 'early' days three to five, and 'late' days eight to 10. This trial considered a participant to be clinically cured "if the eye was normal by physical examination"; participants' symptoms were not included in this assessment.
Gross 2003 used a culture test-of-cure taken at the visit on day seven. This indicated the outcome of bacterial eradication. The secondary outcome of 'clinical cure' was assessed from bulbar conjunctival injection and conjunctival discharge.
Karpecki 2009 defined the primary outcome as clinical resolution, i.e. the absence of conjunctival discharge and bulbar conjunctival injection at day eight. All signs and symptoms of acute bacterial conjunctivitis were rated on a four-point scale.
Leibowitz 1991 classified 'early' as day three and measured microbiological outcome as one of four categories: pathogen eradication, pathogen reduction, pathogen persistence or pathogen proliferation. The eradication and reduction group were grouped together as signifying a positive outcome, and persistence and proliferation were grouped together to represent a negative outcome.
Miller 1992 classified 'early' as days two to three and assessed clinical outcome categorically as either: cured, improved, no change or worsened: the cured and improved categories were combined together as a positive outcome. This trial categorised clinical outcome on the basis of a combination of physical signs and participant symptoms. Microbiological outcome was also recorded at days two to three as: pathogen eradication, pathogen suppression or pathogen persistence; these measures were also measured at a 'late' stage in the trial at days five to seven. To facilitate analysis the eradicated and suppressed group were combined together as reflecting a positive outcome.
Rietveld 2005 used clinical remission at day seven, defined as the absence of any symptoms or signs of conjunctivitis, which was confirmed by the general practitioner. Microbiological eradication rates were also assessed at day seven.
Rose 2005 used clinical cure at seven days as the primary outcome measure, this being assessed from a parent-held diary. Three categories were used for microbiological outcomes: reduced, unchanged and increased colony counts of three organisms: Haemophilus influenzae, Streptococcus pneumoniae or Moraxella catarrhalis.
Silverstein 2011 examined clinical resolution in all randomised patients at day four or five and also eradication of the baseline bacterial infection in those with culture-confirmed bacterial conjunctivitis. Secondary outcomes included clinical and microbiological resolution at day seven ± one day. Individual clinical outcomes were graded on a four-point scale with the aid of standardised photographs.
Tauber 2011 concentrated on microbiology outcomes at day four ± one day. Microbiological success was determined in the "worst eye" if all bacterial isolates which had been recovered at day one were eradicated by day four.
Tepedino 2009 assessed early clinical remission and microbiological eradication at day five ± one day and late clinical remission and microbiological eradication at day eight or nine. All clinical signs and symptoms of acute bacterial conjunctivitis and their assessment were graded in the same way as the Karpecki 2009 study: on a four-point scale.
Three potentially eligible studies were excluded from the review. One was excluded because it was single-masked (Leibowitz 1976). One study, published in Japanese, was excluded following translation of the abstract because it did not include a placebo group (Mitsui 1986). The third study (Everitt 2006) was not placebo-controlled. These three studies randomised 823 participants. See 'Characteristics of excluded studies' for further details.
Seven studies were excluded from the update search in January 2012 as they were not placebo-controlled trials (Bremond-Gignac 2010; Denis 2008; Granet 2008; McDonald 2009; Robert 2010; Szaflik 2009; Ta 2007).
Studies awaiting assessment
One study, which randomised 132 participants, has been published as an abstract only (Ofloxacin 1990) and we have written to the authors for further information.
We also wrote to the Medical Product Information, Global Medical Affairs Alcon Research Ltd. Dr. Ricardo Notivol, 17 June 2011) for further information about three trials (C-00-02; C-00-55; C-01-66). In a response letter they confirmed that all trials have been conducted but have not been published in peer-reviewed journals
Risk of bias in included studies
|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.|
Six trials (Gigliotti 1984; Leibowitz 1991; Miller 1992, Gross 2003, Silverstein 2011; Tauber 2011) did not record information on a priori sample size calculations. It is therefore not possible to make any unequivocal comment on the adequacy of the size of the individual trials. Abelson 2008, Tepedino 2009 and Karpecki 2009 were all large, multi-centre trials and had been planned so as to be adequately powered. The two community-based trials (Rietveld 2005; Rose 2005) were adequately powered.
Random sequence generation
Gigliotti 1984, Gross 2003, Leibowitz 1991, Miller 1992 and Tauber 2011 did not provide sufficient information regarding randomisation. Abelson 2008, Karpecki 2009, Rietveld 2005, Rose 2005, Silverstein 2011 and Tepedino 2009 all provided details of an adequate randomisation procedure. However, although adequately described there were some mis-randomisations reported by the Tepedino 2009 study. Fourteen patients randomised to besifloxacin received vehicle and 12 patients randomised to vehicle received besifloxacin.
Abelson 2008, Rietveld 2005, Rose 2005, Silverstein 2011 and Tepedino 2009 all stated that they had used a central centre for randomisation. The information for allocation concealment for Gigliotti 1984, Leibowitz 1991, Tauber 2011 and Tepedino 2009 was unclear. Gross 2003 described the use of sealed envelopes and Miller 1992 used identical plastic dropper vials. At no point did any study state that the envelopes were opaque.
Abelson 2008 described how formulations of vehicle with and without the active drug were identically packaged. In Karpecki 2009 the intervention was coded by its packaging site. Rietveld 2005 and Rose 2005 described the use of identical dropping bottles, as did Silverstein 2011.
Incomplete outcome data
Leibowitz 1991 reported why patient data were missing in a table, however no provision was made to impute these data into the results calculations. Abelson 2008 reported that more than twice as many patients in the vehicle group discontinued the treatment early, because of lack of efficacy, than patients in the treatment group, however, it was unclear as to whether outcome data for such patients were somehow incorporated into the results calculations. In Gigliotti 1984, patients who failed to return for a final visit were not included in the intention-to-treat analysis and there was no mechanism for dealing with these missing data. In Gross 2003, it was unclear as to how data were analysed for patients who were not bacterial culture-positive. In Karpecki 2009 and Miller 1992, there were significantly different withdrawal rates between groups, which were not taken into account in the analysis. However, in Rietveld 2005 they provided a table of all patients lost to follow-up and undertook additional analysis with the explicit assumption that these patients had not recovered by seven days to produce the most conservative estimate for efficacy. Similarly in Rose 2005, an intention-to-treat analysis was performed as if those lost to follow-up had been treatment failures. Silverstein 2011 documented all losses to follow-up and reasons for other missing data and all missing data were imputed by the last-observation-carried-forward method. The missing data from Tauber 2011 were not relevant to the calculations for microbial resolution. Tepedino 2009 reported results using a different denominator for the number of patients randomised to each arm of the trial. Their 2009 study randomised total of 957 patients (besifloxacin = 475 and vehicle = 482, however, for clinical resolution early and late they used only the limited culture-proven population (besifloxacin = 199 and vehicle = 191) instead of the clinical intention-to–treat population, which should have included all randomised patients.
None of the papers had an a priori trial protocol published detailing outcomes that would be reported on. Only Tauber 2011 reported microbiological, 'early' outcomes - it may be that a future publication will detail clinical outcomes.
Other potential sources of bias
In terms of other possible biases, there may be evidence of some selection bias where studies are conducted in a specialist setting (Gigliotti 1984; Gross 2003; Leibowitz 1991). Furthermore, Aziz Sheikh (AS) (lead author on this review) was a co-author on the Rose trial, and in order to minimise any risk of bias he was not involved in the selection, quality assessment or data extraction discussions in relation to this study.
We created funnel plots to help the assessment of publication bias. Though their interpretation is complex, 'by eye' only and they have been criticised (Terrin 2005) as misleading even with expert assessment, and they offer only limited information in the context of meta-analyses with smaller numbers of trials, they revealed a relative paucity of small negative studies, which indicates the possibility of publication bias, as small studies may only have been published if they yielded positive results (Stern 2011).
Overall we graded the studies as follows:
A: low risk of bias: Rietveld 2005; Rose 2005;
C: high risk of bias: Abelson 2008; Gigliotti 1984; Gross 2003; Karpecki 2009; Leibowitz 1991; Miller 1992; Silverstein 2011; Tauber 2011; Tepedino 2009.
Effects of interventions
Despite relatively high heterogeneity between different studies (discussed below), given that the trial effects were all in the same direction, we considered it appropriate to pool data across studies for 'early' and 'late' outcomes for rates of both clinical and microbiological remission.
'Early' clinical remission: days two to five
Pooling the data from six trials (Gigliotti 1984; Karpecki 2009; Miller 1992; Rose 2005; Silverstein 2011; Tepedino 2009), we calculated a risk ratio (RR) of 1.36 (95% confidence interval (CI) 1.15 to 1.61) for 'early' (day two to five) clinical remission with antibiotics when compared with placebo. There was high heterogeneity I
|Figure 3. Forest plot of comparison: 1 Antibiotics versus placebo, outcome: 1.1 Clinical remission (early).|
'Early' microbiological remission: day two to five
Meta-analysis of data from seven trials (Gigliotti 1984; Karpecki 2009; Leibowitz 1991; Miller 1992; Silverstein 2011; Tauber 2011; Tepedino 2009) revealed a RR of 1.55 (95% CI 1.37 to 1.76) for 'early' microbiological remission, i.e. by days two to five. There was high heterogeneity among these studies I
|Figure 4. Forest plot of comparison: 1 Antibiotics versus placebo, outcome: 1.2 Microbiological remission (early).|
'Late' clinical remission: day six to 10
We calculated a RR of 1.21 (95% CI 1.10 to 1.33) from eight trials (Abelson 2008; Gigliotti 1984; Gross 2003; Karpecki 2009; Rietveld 2005; Rose 2005; Silverstein 2011; Tepedino 2009) for 'late' clinical remission over days six to 10 with antibiotics as opposed to with placebo. This is very similar to the proportion in 'early' clinical remission. There was modest heterogeneity: I
|Figure 5. Forest plot of comparison: 1 Antibiotics versus placebo, outcome: 1.3 Clinical remission (late).|
'Late' microbiological remission: day six to 10
The RR of 'late' microbiological remission was 1.37 (95% CI 1.24 to 1.52) across nine trials (Abelson 2008; Gigliotti 1984; Gross 2003; Karpecki 2009; Miller 1992; Rietveld 2005; Rose 2005; Silverstein 2011; Tepedino 2009) with high heterogeneity I
|Figure 6. Forest plot of comparison: 1 Antibiotics versus placebo, outcome: 1.4 Microbiological remission (late).|
- Recurrence of infection within four weeks. This outcome was not reported on in any of the trials.
- Cost-effectiveness of treatment. This outcome was not reported on in any of the trials.
- Compliance with treatment, and number of drop-outs. This outcome is incorporated into the 'Risk of bias' tables under the 'Incomplete reporting' section.
- Number of participants that experience complications of acute bacterial conjunctivitis. This outcome was not reported on in any of the trials.
- Adverse outcomes as reported in trials. No serious adverse outcomes were reported in either the active or placebo arms of the trials, indicating that important sight-threatening complications such as bacterial keratitis and orbital cellulitis are an infrequent occurrence in patients with acute bacterial conjunctivitis. The whole spectrum of all other sight-threatening complications were not reported.
Summary of main results
The increased absolute risk difference due to treatment with antibiotic eye drops was 11% for 'early' clinical remission and 9% for 'late' clinical remission, in comparison to treatment with placebo drops. The baseline remission rate of 30% with placebo by day five could be considered to be quite high. There may be some effect of administering antiseptic vehicle-placebo eye drops three or four times a day which may have contributed to this. However, there may also be some bias, as in Abelson 2008 twice as many participants discontinued in the placebo arm as in the treatment arm due to lack of efficacy. This would increase the measured effectiveness of the placebo arm as these data were not imputed in any way.
Antibiotic treatment was associated with a risk ratio (RR) of 1.36 (95% confidence interval (CI) 1.15 to 1.61) for 'early' clinical remission and a RR of 1.21 (95% CI 1.10 to 1.33) for 'late' clinical remission. It was also associated with a RR of 'early' microbiological cure of 1.55 (95% CI 1.37 to 1.76) and 'late' microbiological cure of 1.37 (95% CI 1.24 to 1.52). It appears that there is a higher rate for 'early' microbiological cure than for 'early' clinical cure, suggesting that clinical cure lags behind microbiological eradication. This lag may also be in evidence at the 'late' time point.
Overall completeness and applicability of evidence
There are several limitations to this review. Acute bacterial conjunctivitis is frequently cited to be a 'self limiting' condition. However, its natural history cannot be inferred from the trials included in this review because there is likely to be some effect of applying placebo eye drops containing antiseptic three or four times a day. This is perhaps reflected in the relatively high proportion (30%) of placebo arm patients who achieved clinical cure by day five (and 41% by days six to 10).
All the studies used different antibiotics, so no recommendations can be made on the basis of this review as to which of these is superior to any other or indeed which may be particularly useful for specific organisms.
There was no attempt in any of the studies to determine the cost-effectiveness of treating acute bacterial conjunctivitis. This would be a very useful study in the context of a self limiting condition.
As all of the trials show some benefit of antibiotics over placebo it would seem reasonable to conclude that these results are generalisable to the use of similar antibiotics for acute bacterial conjunctivitis in similar primary and secondary care settings.
Quality of the evidence
We clearly assessed the two primary care studies (Rietveld 2005; Rose 2005) as being of higher quality than the remaining studies. We assessed the remaining nine trials (Abelson 2008; Gigliotti 1984; Gross 2003; Karpecki 2009; Leibowitz 1991; Miller 1992; Silverstein 2011; Tauber 2011; Tepedino 2009) as being at high risk of bias and thus lower quality.
In three of the four meta-analyses we found high heterogeneity (I
Another factor possibly contributing to this heterogeneity was the differing time periods of administration of antibiotics. For example, although studies included in the 'early' groups administered antibiotics for approximately five days, this may have actually been for a total of three to six days of antibiotic drops. Therefore this is a highly heterogeneous group and it is difficult to give clear recommendations on the appropriate time period for antibiotic administration.
Potential biases in the review process
As AS was an author of Rose 2005, he did not participate in selection, quality assessment, data extraction and interpretation of this trial; instead, this was done by UN and SM.
This review process should have minimum bias. The broad search strategy, performed in conjunction with The Cochrane Library, should have captured as many relevant studies as possible. This means that both completed and ongoing trials which have been appropriately coded by the Cochrane Eyes and Vision Group were included. Two authors then independently chose the relevant articles from this list. We included trials published in any language, thus minimising bias.
Agreements and disagreements with other studies or reviews
This latest review confirms the previous evidence that treatment of acute bacterial conjunctivitis with antibiotics offers some benefits in speeding up early resolution and more clearly establishes that more modest benefits are also apparent in relation to later resolution (Hurwitz 2005). Jefferis 2011 conducted an individual patient data meta-analysis of these trials, which also found that antibiotics confer modest benefits in improving symptom resolution.
Implications for practice
There remains only a small evidence base of two trials directly relevant to primary care (Rietveld 2005; Rose 2005) where the majority of cases are seen and managed. However, the remaining nine trials may be generalised to some extent and add to this evidence. Early clinical and microbiological remission rates by day five are improved following administration of a topical broad-spectrum antibiotic and these benefits persist but are more modest for later resolution. Given the self limiting nature of conjunctivitis, operating a wait and see policy to see if symptoms/infection spontaneously resolve appears reasonable. It is also appropriate to consider use of antibiotic eye drops as these increase the speed of resolution of symptoms associated with acute bacterial conjunctivitis.
Implications for research
Further research is required in primary care assessment and treatment to understand better the natural history of conjunctivitis in these settings, to ascertain whether with no active treatment the resolution rates seen in the placebo arm are indeed due to spontaneous resolution or whether they reflect the impact of the vehicle of the placebo. In addition, further research is needed to assess cost-effectiveness in this setting in terms of improving quality of life and reducing time off work and education.
- Our thanks to the Cochrane Eyes and Vision Group (CEVG) for their support throughout this review, and especially to Iris Gordon, Trials Search Co-ordinator for CEVG and Anupa Shah, Managing Editor for CEVG.
- We thank Andrew Tullo for peer reviewing a previous version of this review and Steve Tuft and Catey Bunce for reviewing the 2012 update.
Richard Wormald (Co-ordinating Editor for CEVG) acknowledges financial support for his CEVG research sessions from the Department of Health through the award made by the National Institute for Health Research to Moorfields Eye Hospital NHS Foundation Trust and UCL Institute of Ophthalmology for a Specialist Biomedical Research Centre for Ophthalmology. The views expressed in this publication are those of the authors and not necessarily those of the Department of Health.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. CENTRAL search strategy
1 MeSH descriptor Conjunctivitis, Bacterial
#2 conjunctiv* near (acute or infect* or bacteria*)
#3 (#1 OR #2)
#4 MeSH descriptor Anti-Bacterial Agents
#6 (#4 OR #5)
#7 (#3 AND #6)
Appendix 2. MEDLINE (OvidSP) search strategy
1. randomized controlled trial.pt.
2. (randomized or randomised).ab,ti.
9. exp animals/
10. exp humans/
11. 9 not (9 and 10)
12. 8 not 11
13. exp conjunctivitis,bacterial/
14. ((acute or infect$ or bacteria$) adj4 conjunctiv$).tw.
16. exp anti-bacterial agent/
19. 15 and 18
20. 12 and 19
The search filter for trials at the beginning of the MEDLINE strategy is from the published paper by Glanville (Glanville 2006).
Appendix 3. EMBASE (OvidSP) search strategy
1. exp randomized controlled trial/
2. exp randomization/
3. exp double blind procedure/
4. exp single blind procedure/
7. (animal or animal experiment).sh.
9. 7 and 8
10. 7 not 9
11. 6 not 10
12. exp clinical trial/
13. (clin$ adj3 trial$).tw.
14. ((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$)).tw.
15. exp placebo/
18. exp experimental design/
19. exp crossover procedure/
20. exp control group/
21. exp latin square design/
23. 22 not 10
24. 23 not 11
25. exp comparative study/
26. exp evaluation/
27. exp prospective study/
28. (control$ or prospectiv$ or volunteer$).tw.
30. 29 not 10
31. 30 not (11 or 23)
32. 11 or 24 or 31
33. exp conjunctivitis/
35. 33 and 34
36. ((acute or infect$ or bacteria$) adj4 conjunctiv$).tw.
38. exp antibiotic agent/
41. 35 and 40
42. 32 and 41
Appendix 4. Open Grey search strategy
Appendix 5. metaRegister of Controlled Trials search strategy
(bacterial conjunctivitis) and antibiotic
Appendix 6. ClinicalTrials.gov search strategy
Bacterial Conjunctivitis AND Antibiotic
Appendix 7. ICTRP search strategy
Bacterial Conjunctivitis = Condition AND Antibiotic = Intervention
Last assessed as up-to-date: 18 July 2012.
Protocol first published: Issue 3, 1998
Review first published: Issue 3, 1999
Contributions of authors
Conceiving the study: AS
Formulating the protocol: AS
Searching databases: IG
Selecting trials: UN, SM, AS
Extracting data: UN, SM
Checking extracted data: AS
Performing data synthesis: UN, SM
Writing the review: UN, SM, CS, AS, BH
Declarations of interest
AS was a co-author on the Rose 2005 trial. James Cave jointly conceived the study and formulated the protocol for this review.
Sources of support
- University of Edinburgh, UK.
- King's College London, UK.
- Maastricht University, Netherlands.
Differences between protocol and review
In previous versions of the review we have not separated primary and secondary outcomes of interest. In this updated review we have divided all outcome measures into two: primary and secondary outcomes.
Medical Subject Headings (MeSH)
MeSH check words