Target condition being diagnosed
Pharyngitis is defined as an acute inflammation of the pharynx, tonsils or both. Viruses are the most common cause of pharyngitis but the bacterium most frequently identified during acute pharyngitis is Streptococcus pyogenes (S. pyogenes), also known as group A β-haemolytic streptococcus (GAS). GAS is estimated to account for 20% to 40% of cases of pharyngitis in children and 5% to 15% in adults (Shaikh 2010; Wessels 2011). The estimated number of cases of GAS pharyngitis in children is 450 million/year worldwide (Carapetis 2005a). GAS pharyngitis is ubiquitous but is more frequent in low-income countries. Most cases are benign and self limiting within a week but suppurative complications (cervical lymphadenitis, retropharyngeal abscess, peritonsillar cellulitis or abscess (quinsy), sinusitis, acute otitis media and mastoiditis) or non-suppurative post-streptococcal diseases (acute rheumatic fever and rheumatic heart disease, acute glomerulonephritis, Sydenham’s chorea, scarlet fever, streptococcal toxic shock syndrome and paediatric autoimmune neuropsychiatric disorder associated with group A streptococci) can occur (Gerber 2005).
In low-income countries, rheumatic heart disease remains the most commonly acquired heart disease in children, adolescents and young adults: a recent estimate of the number of deaths from rheumatic heart disease is 233,000 per year worldwide (Carapetis 2005a). In high-income countries acute rheumatic fever and rheumatic heart disease are rare (e.g., ≤ 10 cases/year/100,000 children for acute rheumatic fever) (Carapetis 2005b; Seckeler 2011) because of improvements in living conditions, hygiene, increased antibiotic usage, increased access to primary care providers and changes in GAS epidemiology (Carapetis 2007). As a result, the public health goal is shifting from preventing rare GAS complications to minimising inappropriate use of antibiotics.
Many experts recommend the prescription of antibiotics for children with GAS-suspected or GAS-proven pharyngitis (Matthys 2007). The goal of antibiotic treatment is to reduce the individual risk of suppurative or non-suppurative complications, the duration of symptoms and the spread of the condition (Spinks 2011). Correct identification of GAS ensures against not missing GAS-positive cases that can lead to complications. The correct exclusion of GAS ensures against unnecessary use of antibiotics (thus reducing the incidence of adverse drug reactions, antibiotic resistance and associated costs).
There is a lack of consensus on the most suitable diagnostic method for GAS in children with pharyngitis and the 'standard' diagnostic practice varies greatly amongst countries. Because the signs and symptoms of GAS and viral pharyngitis overlap broadly (Shaikh 2011), most guidelines that recommend antibiotic treatment of GAS also recommend confirmation of the presence of GAS on the basis of a throat swab (Matthys 2007). However, throat swabs are explicitly not recommended in some countries (e.g., United Kingdom, Belgium and The Netherlands) (Matthys 2007). International discrepancies might be explained by academic reasons and 'clinical traditions', different targets of sensitivity and specificity because of local epidemiological differences (i.e., rheumatic fever and rheumatic heart disease prevalence), international differences in health systems and policies, and the sparseness of recent data on the incidence of GAS complications and the efficacy of antibiotic treatment for their prevention.
The standard criterion for the diagnosis of GAS in children with pharyngitis is throat culture on a blood agar plate in a microbiology laboratory (AAP 2009). The major advantage of laboratory throat culture is its detection of GAS from swabs with a very low number of bacteria, but the major limitation is the 48-hour delay in obtaining results. In addition, throat cultures cannot distinguish true GAS infection from GAS carriage with intercurrent viral pharyngitis. Asymptomatic pharyngeal GAS carriage is usually defined as positive throat culture results for GAS without a GAS-specific immune response (anti-streptolysin O and anti-DNase B antibodies) (Tanz 2007). Asymptomatic GAS carriage occurs in 10% to 15% of healthy children (Shaikh 2010) and does not require antibiotic treatment (Tanz 2007).
Agreement is lacking on the most suitable culture technique for diagnosing GAS in children with pharyngitis. Several parameters are likely to affect the sensitivity of the test (culture medium, atmosphere of incubation, duration of incubation, group A identification technique and the number of plates inoculated) (Kellogg 1990; Tanz 1997). These variables affect the diagnostic accuracy of the throat culture and thus the diagnostic accuracy of rapid antigen detection tests (RADTs) as compared to throat culture.
Simple RADTs were developed in the 1980s to provide an immediate indication for the clinician about the presence or absence of GAS in children with pharyngitis. RADTs do not require any special equipment and can be performed at the point of care with a throat swab (Cohen 2004). They can provide immediate results and are calibrated to produce binary results (positive or negative).
In children, the reported sensitivity of RADTs is about 85% (Gerber 2004) but varies greatly amongst studies (from 66% (Van Limbergen 2006) to 99% (Harbeck 1993)) and the specificity is high and stable, about 95% (Gerber 2004). Because of this high specificity, most experts agree on prescribing antibiotics with positive RADT results, even if RADTs cannot differentiate GAS true infection from GAS carriage. However, the consequences of a negative RADT result depend on national guidelines. North American guidelines recommend backing up negative RADT results with throat culture to avoid not treating RADT false-negative cases (Bisno 2002; Gerber 2009) but most recent European guidelines recommend relying on negative RADT results without culture confirmation (Pelucchi 2012). In low-income countries, the clinical consequences of RADT results might be the same as in high-income countries (treat RADT-positive cases only) but resources for testing might be limited and practices may vary from generalised empiric antibiotic treatment to selective antibiotic treatment or selective rapid testing based on clinical scoring systems (Joachim 2010; Steinhoff 2005; WHO 1995).
All available RADTs involve the detection of the Lancefield group A carbohydrate, a GAS-specific cell-wall antigen. Different immunologic techniques are available for carbohydrate detection (Gerber 2004), from older to most recent:
- Latex agglutination (LA) assay: the sample is placed in the presence of latex beads coupled with GAS-specific antibodies; the result is determined by observing the agglutination of the beads if they are related to the specific antigen in the sample. These first-generation tests are no longer used in clinical practice and will not be considered in this review.
- Enzyme immunoassay (EIA): the sample is placed at the end of a nitrocellulose strip and then migrates to an area where it forms an antigen-antibody complex. These second-generation tests are also known as immunochromatographic, sandwich or lateral-flow assays. They are the most widespread and most used RADTs in clinical practice.
- Optical immunoassay (OIA): the sample is placed on a silicon membrane in the presence of the reagent. The result is based on the change in optical properties of the inert membrane in the presence of an antigen-antibody complex. These third-generation tests seem to be more sensitive than EIAs but their use is limited because of their high cost.
Another test for the diagnosis of GAS in children with pharyngitis is a throat culture performed in the physician's office (office culture). Office culture has the same disadvantage as a laboratory culture (a 48-hour delay in obtaining results), with the major limitation being insufficient sensitivity (from 50% to 85%) (Battle 1971; Mondzac 1967; Rosenstein 1970; Tanz 2009; Wegner 1992). Office culture is almost completely abandoned and will not be considered in this review.
Streptococcal antibody tests
Assessment of GAS-specific antibodies is the traditional reference test to differentiate true GAS infection and GAS carriage. The most commonly used GAS-specific antibody assays tests are for anti-streptolysin O and anti-DNase B antibodies. Increased antibody titre assessment diagnoses true GAS infection better than a single absolute titre assessment (Gerber 1986; Johnson 2010). Streptococcal antibody tests are not used for the diagnosis of GAS in children with pharyngitis because of the need for repeat blood samples. Moreover, the information about the kinetics of the immune response to GAS in children with pharyngitis is very limited and the most recent data show that the interpretation of streptococcal antibody test results is not straightforward (Johnson 2010). Therefore, their use is usually limited to documenting recent GAS infection in patients suspected of having GAS non-suppurative complications or to epidemiologic studies (Gerber 1986; Johnson 2010).
Clinical scoring systems
Clinical scoring systems have been developed to diagnose GAS on the basis of clinical grounds. The most popular of these scores are the Centor score (Centor 1981) and the McIsaac score (McIsaac 1998). The scores are based on assessing simple clinical criteria (history of fever, cough, tonsillar swelling or exudate, tender cervical adenopathy and age). Their use is recommended in adults but might be inappropriate in children; several authors have reported a lack of diagnostic accuracy in this population (Cohen 2012; Fischer Walker 2006; Shaikh 2011). Clinical scoring systems will not be considered in this review.
Rapid molecular biology assays
Rapid molecular biology assays for GAS in children with pharyngitis have been recently developed (Group A Streptococcus Direct Test; GenProbe Inc., San Diego, CA; and LightCycler Strep-A assay; Roche Applied Science, Indianapolis, IN) (Chapin 2002; Heelan 1996; Pokorski 1994; Uhl 2003). These techniques, based on DNA-rRNA hybridisation or PCR, are highly sensitive but are not currently used widely because of their cost, the need for highly specialised equipment and personnel and the two-hour delay in results (Gerber 2004). Molecular assays are not antigen-detection tests and will not be considered in this review.
Childhood pharyngitis is a significant public health problem with, on the one hand, suppurative and non-suppurative complications of GAS pharyngitis (especially acute rheumatic fever and rheumatic heart disease) and, on the other, costly diagnostic tests and unnecessary antibiotics. RADTs for GAS are now widely available and their use in children with pharyngitis might increase accurate diagnosis and reduce antibiotic consumption.
According to local clinical guidelines, RADTs may be used as stand-alone diagnostic tests in replacement of throat culture (e.g., in contexts where throat culture is unavailable or not used), or as triage tests, with negative results being supported by a throat culture. These international discrepancies might be explained in part by persistent gaps in knowledge regarding the diagnostic accuracy of RADTs:
- What is the accuracy of RADTs for GAS in children with pharyngitis compared to the most consensual reference test (throat culture on a blood agar plate)?
- Are there significant differences in diagnostic accuracy between EIAs and OIAs?
- Which study-level factors could explain variations in diagnostic accuracy across clinical studies?
The questions of whether RADTs should be performed in all patients presenting with signs and symptoms of pharyngitis or only in selected patients on the basis of a clinical score (selective testing strategies), and whether clinical protocols that incorporate RADTs are sufficient to reduce antibiotic prescription, will not be addressed in this review. We aim to provide information to help clinicians and public health decision makers better define the precise role of RADTs in the diagnosis of GAS in children with pharyngitis on the basis of unbiased evidence.
The primary objective is to determine the diagnostic accuracy of RADTs for GAS in children with pharyngitis.
A secondary objective is to assess the relative diagnostic accuracy of EIA and OIA tests by indirect and direct comparison.
Criteria for considering studies for this review
Types of studies
We will include reports of cross-sectional studies reporting the diagnostic accuracy of one or more RADTs for the diagnosis of GAS in children with pharyngitis, with laboratory throat culture as the standard test. Reports of randomised controlled trials (RCTs) will also be included if we can extract 2 x 2 tables for children. Reports of studies involving the RADT result as part of the reference test will be included but we will investigate the effect of such studies by using this study characteristic (i.e., RADT as part of the reference test) as a potential criterion to carry out sensitivity analyses.
We will include reports of studies of children or young adults (age ≤ 21 years, according to the upper limit chosen by the American Academy of Pediatrics) seeking ambulatory medical care because of a sore throat or with a diagnosis of pharyngitis, who provided a throat swab for a RADT and laboratory throat culture. Ambulatory care settings will include private physicians’ offices, walk-in clinics, hospital outpatients, emergency departments and family medicine centres.
We will also include reports of studies with only a subgroup of participants eligible for inclusion in the review, provided that we can extract relevant data specific to that subgroup. Reports of studies will not be excluded on the basis of whether studies were performed in high-income or low-income countries because no data exist to support variations in the accuracy of RADTs according to this criterion.
We will include only studies of EIA or OIA RADTs for GAS in children with pharyngitis, including those no longer marketed.
GAS in children with pharyngitis.
Studies are required to diagnose GAS with throat culture on a blood agar plate in a microbiology laboratory used as the reference test. Several parameters may affect the accuracy of throat culture. For studies involving more than one throat culture technique (different medium, duration or atmosphere of incubation), we a priori chose to extract data related to the culture technique recommend by a panel of North American content experts, i.e., simple blood agar plate (versus selective or enriched media), incubation 48 hours total (versus 18 to 24 hours only), aerobic atmosphere (versus other) (Shulman 2000), in order to avoid data-driven approaches.
Search methods for identification of studies
We will search MEDLINE using the search strategy described in Appendix 1. The search strategy was developed in consultation with a Medical Librarian and the Trials Search Co-ordinator for the Acute Respiratory Infections Group and will be adapted to search EMBASE (Ovid) and Web of Science (which incorporates the Science Citation Index). We will not use any filter related to age because many RADT studies enrol adults and children and could provide extractable data for children. We will not use methodological filters to identify diagnostic studies because such filters may result in omission of relevant studies (Leeflang 2006; Whiting 2011b). The searches will be run from 1980 onwards because RADTs were not available prior to this date. We will search the Cochrane Register of Diagnostic Test Accuracy Studies for relevant studies.
We will search the following databases to identify potentially relevant studies referenced in reviews and guidelines:
- the Cochrane Database of Systematic Reviews;
- DARE (Database of Abstracts of Reviews of Effects);
- the MEDION database (for Systematic Reviews of Diagnostic Studies); and
- TRIP (Turning Research Into Practice).
The search strategy will also include names of some commercial kits from the most common manufacturers: Test Pack Strep A (Abbott), ICON Strep A (Beckman Coulter), Link 2 Strep A Rapid Test (Becton Dickinson), Acceava Strep A (Inverness Medical), OSOM Strep A (Genzyme), Poly Stat Strep A (Polymedco) and QuickVue Strep A (Quidel).
Searching other resources
We will handsearch reference lists of included articles and any relevant review articles identified through the search and the 'related articles' function in PubMed (20 first related articles of each included article) for eligible articles. We will use Science Citation Index and Google Scholar to search for reports that cite included articles.
We will electronically search the following databases for 'grey' literature (unpublished studies, including conference proceedings and reports):
- OpenSIGLE database; and
- OAISTER Database.
We will contact manufacturers of the most common RADTs to seek additional or unpublished studies. Manufacturers include Abbott, Beckman Coulter, Becton Dickinson, Genzyme, Inverness Medical, Polymedco and Quidel.
Data collection and analysis
Selection of studies
We will consider studies published in any language. One review author (JFC) will exclude studies that are not related to pharyngitis or RADT on the basis of the titles and abstracts identified by the search strategy. Two review authors (JFC, MC) will retrieve the full text of relevant articles and independently evaluate them for inclusion by using a pro forma as a guide. One review author (RC) will act as arbiter in case of discrepancies between two review authors (JFC, MC) who will discuss the inclusion of the studies. We will select the most recent or most complete report in cases of multiple reports for a given study or when we cannot exclude the possibility of overlapping populations. The study selection process will involve use of ReSyWeb, a web service developed at the French Cochrane Centre and INSERM U872. The tool handles the importation of references from multiple databases, the semi-automatic deletion of duplicate records of the same report, the selection of studies by independent reviewers, consensus procedures for disagreements and the linkage of multiple reports of the same study. Finally, the web service automatically produces a flowchart to report the search process. We will record and report reasons for excluding studies but we will not report their references.
Data extraction and management
Two authors (JFC, MC) will independently extract a standard set of data from each study by using a pre-specified data extraction form. We will extract the number of true positives, true negatives, false positives and false negatives for each index test evaluated in each study to construct 2 x 2 tables. If such data are not provided by the trial authors, we will attempt to contact them to construct the 2 x 2 table for the study population or the pre-specified subgroups. Otherwise, we will calculate the number of true positives, true negatives, false positives and false negatives from the summary estimates of sensitivity and specificity of the index test, if available. If reported, we will extract data on the number of undetermined or uninterpretable RADT results but they will be excluded from the analysis. For studies for which only a subgroup of patients will be included in the review, we will extract, analyse and present data for this subgroup only. If some data are unclear or missing, we will attempt to contact study authors to obtain additional data. The data extraction will be cross-checked and we will resolve discrepancies by discussion until a consensus is reached or we will consult the third review author (RC). See Table 1 for a description of which data will be extracted for each study.
Assessment of methodological quality
Two review authors (JFC, MC) will independently assess the methodological quality of each study using a four-domain tool adapted from QUADAS-2 (Whiting 2011a). We tailored the quality assessment tool to our review question. We developed review-specific guidance on how to assess each signalling question and how to use this information to judge the risk of bias and applicability. We refined the tool until satisfactory inter-rater agreement was achieved. We will summarise the methodological quality assessment in tables for each study. See Table 2.
Statistical analysis and data synthesis
We will enter data of the 2 x 2 tables into RevMan 2012 and we will plot estimates of sensitivity and specificity on forest plots and in the receiver-operating characteristic (ROC) space to represent the variability in diagnostic test accuracy within and between studies.
The hierarchical bivariate model described by Reitsma (Reitsma 2005) will be fitted by use of Stata/SE, which will allow for calculating summary estimates of sensitivity and specificity and the associated 95% confidence intervals (CIs). We will also report the estimate of correlation between sensitivity and specificity (rho). We will put the results from the bivariate model into RevMan 2012 to provide plots of the estimated curve(s) or summary point(s) and confidence region(s), superimposed on the study-specific estimates of sensitivity and specificity in the ROC space.
We will include the same study in the same meta-analysis more than once if needed, i.e., if one study reports different index tests. We will not present results in groups according to commercial test name because RADTs can be considered the same diagnostic test (all RADTs detect the same antigen using immunology-based techniques) and because there might be too many different commercial kits (> 50).
We will present results according to test type (EIA versus OIA) and will also attempt to directly compare the accuracy of EIA versus OIA RADTs using only studies that directly compare the two techniques.
Investigations of heterogeneity
We will initially visually inspect the forest plots and ROC space to check for heterogeneity between study results. To investigate sources of heterogeneity, we will incorporate covariates in the bivariate model, i.e., meta-regression. We will assess the significance of the difference in covariate by likelihood ratio test comparing the bivariate model with and without the covariate. A P value of less than 0.05 will be used to denote statistical significance. With a significant test result, we will assess effects of covariates on sensitivity and specificity separately by testing the significance of the change in -2 log-likelihood of the model with or without corresponding terms. If the bivariate model does not converge or produces unstable parameter estimates, we will simplify the bivariate model by assuming fixed-effect estimates.
If sufficient studies are available, we will address the five following sources of heterogeneity by adding variables to the meta-analysis model:
a. Effect of test type
Some authors suggested that OIA may be more sensitive than EIA tests (Gerber 2004). Therefore, we will try to indirectly compare the RADT tests by using test type as a categorical covariate in the models (EIA versus OIA).
b. Effect of the reference standard
In this review, the reference standard will be throat culture on a blood agar plate. However, the following parameters may affect the accuracy of throat culture on blood agar and therefore the accuracy of RADTs as compared with throat culture: culture medium (standard versus inhibitory), atmosphere of incubation (aerobic versus other), duration of incubation (18 to 24 hours versus 48 hours total), use of an enrichment broth before plating (yes/no), group A identification technique (latex agglutination versus bacitracin testing) and number of plates inoculated (one versus more than one). We will assess the effect of such parameters on the accuracy of the index test by adding categorical covariates in the models.
c. Effect of age
The sensitivity of RADTs is known to be higher in younger children than in older ones (Cohen 2012; Edmonson 2005). This might be explained by higher GAS prevalence in school-age children with pharyngitis than in older children. Therefore, we will try to explore age as a potential source of heterogeneity by using the mean age of patients in the study as a categorised covariate in the model (e.g., 0 to 4; 5 to 9; 10 to 14; 15 to 20 years).
d. Effect of disease severity
Spectrum effect has been demonstrated for RADTs, with increasing sensitivity with increasing disease severity, usually assessed by the McIsaac score (Cohen 2012; Edmonson 2005; Hall 2004; Tanz 2009). Therefore, disease severity might be a relevant source of heterogeneity to explore, for example by using the % of patients with a McIsaac score > two as a numerical covariate in the model.
e. Effect of GAS prevalence
Diagnostic accuracy may vary with disease prevalence (Leeflang 2009), usually with better performances in a population with higher disease prevalence. We will consider GAS prevalence as a dichotomised covariate to define low-risk versus high-risk study populations (i.e., below or above median of GAS prevalence across studies).
We will carry out the following sensitivity analyses to explore the robustness of the results:
- include studies for which patient selection was avoided;
- include studies for which patients were excluded on the basis of antibiotics use within seven days before inclusion;
- include studies for which GAS antibody response was used as the reference test;
- include only studies of high quality according to QUADAS-2.
Assessment of reporting bias
We will not try to assess reporting bias (Macaskill 2010).
We thank Philippe Ravaud and Ludovic Trinquart (French Cochrane Centre, Université Paris Descartes, Paris, France) and the members of the Cochrane ARI Group and the Cochrane DTA Group for their comments and support. We also thank the following people for commenting on the draft protocol: Noorin Bhimani, Samileh Noorbakhsh, Saleh Altamimi, Conor Teljeur and Jenny Doust.
Appendix 1. MEDLINE (Ovid) search strategy
5 (tonsillopharyngitis or pharyngotonsillitis).tw.
6 sore throat*.tw.
7 ((throat* or pharyn* or tonsil*) adj5 (infect* or inflam*)).tw.
8 Pharynx/mi [Microbiology]
9 Streptococcal Infections/
10 (strep* adj5 (throat* or pharyn* or tonsil*)).tw.
11 ("group a" adj5 streptococc*).tw.
13 (beta-hemoly* or beta-haemoly*).tw.
14 lancefield group a.tw.
15 Streptococcus pyogenes/
16 (streptococcus pyogenes or "s. pyogenes" or "s.pyogenes").tw.
19 exp Immunoenzyme Techniques/
20 (enzyme adj2 (immunoassay* or immuno-assay* or immunosorbent)).tw.
23 Immunosorbent Techniques/
24 exp Enzyme-Linked Immunosorbent Assay/
25 (elisa or elisas or eia or eias).tw.
26 (sandwich* adj2 assay*).tw.
27 (lateral flow adj2 assay).tw.
28 (optical adj2 (immunoassay* or immuno-assay*)).tw.
29 (oia or oias).tw.
30 Antigens, Bacterial/
31 Reagent Kits, Diagnostic/
32 Point-of-Care Systems/
33 ((rapid or "point of care" or "near patient" or poc or poct or bedside) adj5 (test or tests or testing or detect* or diagnos* or screen* or kit or kits or assay*)).tw.
34 (radt or radts or rdt or rdts).tw.
35 (antigen* adj3 detect*).tw.
37 17 and 36
38 exp animals/ not humans/
39 37 not 38
Contributions of authors
MC and JFC had the original idea for the review and wrote first draft of the protocol.
RC edited the protocol.
Declarations of interest
All authors declared no financial conflicts of interest.
All authors have been involved in studies that might be included in the review.
Sources of support
- No sources of support supplied
- Laboratoires Guigoz - Société Française de Pédiatrie - Groupe de Pédiatrie Générale – Groupe de Recherches Epidémiologiques en Pédiatrie, France.Educational Grant to JFC (2010)
- Agence Régionale de Santé d'Ile-de-France, France.Educational Grant to JFC (2011)