Background. Rapid antigen detection testing (RADT) is often performed for diagnosis of group A β-hemolytic streptococcal (GABHS) pharyngitis among children. Among adults, the sensitivity of this test varies on the basis of disease severity (spectrum bias). A similar phenomenon may occur when this test is used in a pediatric population, which may affect the need for culture confirmation of all negative RADT results.
Objectives. To assess the performance of a clinical scoring system and to determine whether RADT spectrum bias is present among children who are evaluated for GABHS pharyngitis.
Methods. Laboratory and clinical records for a consecutive series of pediatric patients who underwent RADT at the Marshfield Clinic between January 2002 and March 2002 were reviewed retrospectively. Patients were stratified according to the number of clinical features present by using modified Centor criteria, ie, history of fever, absence of cough, presence of pharyngeal exudates, and cervical lymphadenopathy. The sensitivity of the RADT was defined as the number of patients with positive RADT results divided by the number of patients with either positive RADT results or negative RADT results but positive throat culture results.
Results. RADT results were positive for 117 of 561 children (21%), and culture results were positive for 35 of 444 children (8%) with negative RADT results. The overall prevalence of GABHS pharyngitis was 27% (95% confidence interval: 23–31%). The prevalence of GABHS pharyngitis was 18% among patients with 0 Centor criteria, 16% among those with 1 criterion, 32% among those with 2 criteria, and 50% among those with 3 or 4 criteria. Spectrum bias was present, inasmuch as RADT sensitivity increased with Centor scores, ie, 47% sensitivity among children with 0 Centor criteria, 65% among those with 1 criterion, 82% among those with 2 criteria, and 90% among those with 3 or 4 criteria.
Conclusions. The sensitivity of RADT for GABHS pharyngitis is not a fixed value but varies with the severity of disease. However, even among pediatric patients with ≥3 Centor criteria for GABHS pharyngitis, the sensitivity of RADT is still too low to support the use of RADT without culture confirmation of negative results.
Group A β-hemolytic streptococcal (GABHS) pharyngitis is an important cause of childhood morbidity and the cause of acute rheumatic fever.1 According to national guidelines, the diagnosis of GABHS pharyngitis among children should be based on a laboratory test in conjunction with clinical and epidemiologic findings.2 Culture isolation of GABHS organisms from the pharynx is the standard method, but rapid antigen detection testing (RADT) is now widely available.3–6 Rapid testing has many benefits, ie, early treatment within 48 hours after onset can provide symptomatic relief,7–9 spread to contacts may be limited, the need for follow-up management is lessened, and additional testing can be avoided. However, the RADT offers no advantage over culture for prevention of rheumatic fever, because treatment is effective when initiated as late as 7 to 10 days after onset.
The American Academy of Pediatrics and the Infectious Disease Society of America recommend confirmation of negative RADT results with a throat culture.2,10 Proponents of culture confirmation base this position on the reported sensitivity of RADT, which ranges between 80% and 90%.2,6,10 However, the sensitivity of RADT may vary in relation to disease severity, a property known as spectrum bias (or spectrum effect).11,12 This phenomenon has been demonstrated to occur when RADT is used to identify GABHS pharyngitis among adults.13 If spectrum bias also occurs when this test is used for children, then the sensitivity may be higher among children with a high pretest probability, based on clinical findings. If the sensitivity is sufficiently high among this subset of patients, then culture confirmation of negative RADT results may be unnecessary. The purpose of this study was to assess the occurrence and magnitude of spectrum bias for a licensed rapid antigen detection test used to diagnose streptococcal pharyngitis in a pediatric population.
Study Design and Setting
This was a cross-sectional survey of laboratory and clinical records for a consecutive sample of pediatric patients (age range: 2–17 years) who underwent RADT at the Marshfield Clinic between January 15, 2002, and March 15, 2002. The Marshfield Clinic is a >700-physician multispecialty group practice in central and northern Wisconsin. It is also a teaching institution, and pediatric residents provide supervised primary care with staff physicians. Patients treated in the departments of pediatrics, family medicine, urgent care, and emergency medicine and at primary care satellite centers were included in this study. Children who had been treated for sore throat in the previous 30 days were excluded, to prevent the inclusion of follow-up visits.
At this institution, all children with suspected GABHS pharyngitis undergo RADT for group A streptococci, performed at the point of care by a trained medical assistant or nurse. Medical assistants and nurses are provided with specific training in specimen collection techniques and performance of the RADT, including proper swabbing of the posterior pharyngeal wall and both tonsils (or tonsillar fossae) without touching other areas of the oropharynx. The RADT is performed and the results are recorded within minutes after the specimen is obtained. A second throat swab is obtained simultaneously and is sent for culture if the RADT results are negative.
A color immunochromatographic dipstick assay (Acceava Strep A; Thermo BioStar, Boulder, CO) was used for all RADT. According to the manufacturer, this test has a sensitivity of 96.0% and a specificity of 97.8%. The test detects both viable and nonviable organisms, on the basis of chemical extraction of a carbohydrate antigen that is unique to group A streptococci.
Swabs for culture were obtained with rayon-tip swabs and were sent in liquid Stuart transport medium to the clinical microbiology laboratory (Marshfield Laboratories). The specimens were promptly transported to the laboratory, and cultures were generally prepared within 1 hour. Specimens were plated onto both sheep blood agar medium plates with a bacitracin susceptibility disk and selective agar medium plates. The blood agar plates were incubated aerobically at 35°C, and the selective medium plates were incubated in 5% carbon dioxide. The plates were examined 24 hours after inoculation, and negative cultures were reexamined at 48 hours. Suspected colonies of β-hemolytic streptococci were confirmed as Streptococcus pyogenes with latex agglutination.
Relevant clinical information was abstracted by trained research coordinators, using a standardized form. Demographic and symptom data included date of visit, date of birth, gender, location of visit, presence of cough, and presence of fever (subjective). Physical examination variables included temperature and route of measurement, presence of tonsillar/pharyngeal exudate, and presence of cervical lymphadenopathy. Each finding was classified as present, absent, or not reported. Laboratory records were used to record the RADT result and the culture result (when indicated). Ten percent of the records were randomly selected for duplicate abstraction, for quality assurance purposes.
Determination of Clinical Scores
A modified version of the criteria described by Centor et al14 was used to stratify children on the basis of clinical symptoms. The 4 Centor criteria include 1) history of fever, 2) absence of cough, 3) presence of pharyngeal or tonsillar exudates, and 4) presence of tender anterior cervical lymphadenopathy. These criteria have been validated as predictors of GABHS pharyngitis in adult populations.15 Most expert reviews include the Centor variables as clinical findings consistent with streptococcal pharyngitis.16–18
For this study, the Centor criteria were modified to accept any documentation of cervical lymphadenopathy as a positive finding, because the specific lymph nodes and the presence of tenderness were not consistently reported in the medical records. The Centor score was thus defined as the number of modified Centor criteria (history of fever, absence of cough, presence of pharyngeal or tonsillar exudates, and presence of cervical lymphadenopathy) present. For determination of the Centor scores, clinical factors were considered absent if not reported in the medical records. The study protocol was reviewed and approved by the Marshfield Clinic institutional review board.
All analyses were conducted by using SAS 8.2 software (SAS Institute, Cary, NC). Children were stratified into 4 groups on the basis of Centor scores. Patients with scores of 3 or 4 were combined into a single group, because of sample size considerations.
Because cultures were performed only for children with negative RADT findings, we assumed that all children with positive RADT results would have been culture-positive. Therefore, the sensitivity estimates presented in our analysis represent the maximal sensitivity that could be observed in the study population. Previous studies found that the false-positive rate (1 − specificity) is very low for the immunochromatographic GABHS RADT system,19–21 and the false-positive rate for the test used in this study was <3% in prelicensure studies.
To evaluate the association between individual modified Centor criteria and GABHS pharyngitis, we used a model similar to a logistic regression model but with a logarithmic (rather than logistic) link. The model permits direct estimation of prevalence ratios. Because GABHS pharyngitis is not a rare outcome, the odds ratio generated from a logistic regression model would not be a good approximation of the prevalence ratio. The model included terms for each modified Centor criterion and age group (categorized as 2–6, 7–12, or 13–17 years).
To evaluate spectrum bias, we first fit models similar to that described above (binomial distribution with logarithmic link). By restricting the analysis to patients with positive GABHS results (ie, positive RADT or culture results) and using the RADT results as the outcome in the model, we were able to model the sensitivities directly. Age was evaluated as a potential effect modifier and confounder of the Centor score effect. On the basis of the results of these models, we conducted a trend test and a dose-response analysis. First, the Cochran-Armitage trend test was performed to assess generally whether the sensitivity of RADT increased with increasing Centor scores. Second, we computed incremental sensitivity ratios,22 to assess whether there was a significant increase in the RADT sensitivity for each incremental increase in Centor scores. Incremental sensitivity ratios are the ratios of the sensitivities for adjacent categories of Centor scores. Values exceeding 1.0 for all adjacent categories would be consistent with a dose-response relationship.
A total of 589 children underwent RADT during the study period; 28 (4.8%) of those children were excluded because the test was performed at a follow-up visit. For the remaining 561 children, the median age was 9 years and the gender distribution was approximately equal (Table 1). Other than sore throat, the most common clinical findings were history of fever and lymphadenopathy. The presence or absence of ≥3 Centor criteria was fully documented in the medical records for 73% of the children; 21% had documented information for 2 criteria, and 6% had complete documentation for 1 or 0 criteria.
One hundred fifty-two of the 561 children (27%) demonstrated laboratory evidence of GABHS pharyngitis, ie, either RADT or culture findings. The RADT results were positive for 117 children (21%), and the throat culture results were positive for 35 of 444 children (8%) with negative RADT findings. The presence of pharyngeal exudate, cervical lymphadenopathy, and the absence of cough were independently associated with GABHS recovery, after adjustment for age (Table 2). Fever was not associated with an increased risk of GABHS.
The association between Centor scores and RADT sensitivity did not differ according to age (P = .8). Because the age-adjusted association was nearly identical to the unadjusted association, the subsequent analyses are reported without adjustment for age.
The overall sensitivity of RADT was 77% (95% confidence interval [CI]: 70–84%). The association between Centor scores and RADT sensitivity was highly significant (P = .0006). RADT sensitivity increased with increasing Centor scores, ranging from 47% when none of the modified Centor criteria were present to 90% when 3 or 4 criteria were present (P < .0001, Cochran-Armitage test for trend) (Table 3). All incremental sensitivity ratios exceeded a value of 1.0; however, the 95% CI for each ratio overlapped 1.0. These results are consistent with a dose-response relationship, although a statistically significant dose-response relationship was not observed.
This analysis was repeated for the subset of children with full medical record documentation regarding the presence or absence of ≥3 criteria. As in the original analysis, the RADT sensitivity increased with increasing Centor scores, ranging from 44% when 0 criteria were present to 88% when 3 or 4 criteria were present. The incremental sensitivity ratios were essentially unchanged when the results for this subset were compared with those for the entire group.
Among children with negative RADT findings, there was no association between Centor scores and the presence of GABHS organisms in throat cultures (P = .7). The negative predictive value of RADT was unrelated to the modified Centor scores for this population (Table 3).
Performance of a laboratory test can vary across subgroups within a population, a phenomenon referred to as spectrum bias or, more recently, spectrum effect.11,12 Recognition of spectrum bias has often occurred in relation to the severity of disease.13,23,24 We hypothesized that the sensitivity of RADT for group A streptococci within a pediatric sample would demonstrate spectrum bias, resulting in a higher sensitivity for the subgroup of patients with manifestations of more severe disease. If this is true and the sensitivity is sufficiently high, then culture confirmation of negative RADT findings may be unnecessary for some children.
The results of this study demonstrate that spectrum bias may occur when RADT is used for diagnosis of GABHS pharyngitis among children. The sensitivity of the RADT was lowest (47%) for children with 0 Centor criteria and highest (90%) for those with 3 or 4 criteria. The sensitivities observed in this study, even among children with a high pretest probability of GABHS pharyngitis, were substantially lower than the sensitivity reported by the manufacturer. Other investigators found spectrum bias when RADT was used to diagnose GABHS pharyngitis in adult populations. In a study by DiMatteo et al,13 the sensitivity increased from 61% for patients with 0 or 1 Centor criteria to 97% for those with all 4 criteria. Spectrum bias was also observed by Dagnelie et al25 in a population of patients ranging in age from 4 to 60 years. In that study, the sensitivity of the RADT among children 4 to 14 years of age was 59% when 0 to 2 criteria were present and 83% when 3 or 4 criteria were present.
Why RADT displays spectrum bias is unknown. One contributing factor may be the quantity of bacteria in the pharynx, which may be higher among patients with more severe clinical manifestations. This might increase the probability of positive test results for patients with severe symptoms and high bacterial loads. Our laboratory does not report quantitative data on cultures; therefore, such information is not available for this study. However, the package insert for the commercial test used in this study does provide data indicating that the sensitivity of the test increases as the numbers of colonies present in the culture increase.
Current pediatric guidelines recommend culture confirmation for all children with negative RADT results.2,10 In this pediatric population, the negative predictive value of the RADT was moderately high and did not vary on the basis of Centor scores; it was not sufficiently high to allow abandonment of the confirmatory cultures for any of the Centor score groups. At our institution, 5% to 10% of GABHS pharyngitis cases would have been undiagnosed and untreated if confirmatory cultures had not been performed. The children would have been at risk for developing acute rheumatic fever, although the incidence of this disease has declined dramatically in the past 4 decades. The negative predictive value and GABHS pharyngitis prevalence in this study were similar to findings reported by DiMatteo et al13
The prevalence of GABHS pharyngitis generally increased as the number of clinical criteria indicative of the illness increased. This confirms the ability of the 4 clinical factors used in this study (history of fever, absence of cough, presence of pharyngeal exudates, and cervical lymphadenopathy) to stratify children into groups with increasing likelihoods of GABHS pharyngitis. A similar clinical predictive model for GABHS pharyngitis has been validated in the pediatric population.16 The pediatric model is based on tonsillar swelling, lymphadenopathy, lack of coryza, and scarlatina; fever and lack of cough are not included.
The patients in this study were representative of children presenting to primary care or urgent care settings for evaluation of sore throat or related symptoms. The 27% prevalence of GABHS pharyngitis is consistent with the prevalence documented for children presenting with sore throat during the winter and early spring.18 However, we selected all children for whom RADT was performed, rather than only those with a chief complaint of sore throat. Therefore, the study population may not represent all children with pharyngitis. Patients with an obvious alternative diagnosis, such as influenza, would not have undergone RADT, and neither would those the clinicians considered to have no clinical probability of having GABHS pharyngitis. Finally, sore throat may not be the primary complaint for children with GABHS pharyngitis,18 and clinicians might have performed RADT because of other complaints (such as headache or abdominal pain). Therefore, the findings of this study are based on usual care in typical clinical settings.
The prevalence of GABHS infection among children with Centor scores of 0 or 1 was similar to the prevalence among asymptomatic carriers. This is not surprising, because children with low Centor scores are likely to have a viral process causing their clinical symptoms. For children with low Centor scores, it is challenging to distinguish the minority with GABHS pharyngitis from the majority with a viral syndrome. Detecting GABHS organisms in the pharynx of these children may not be helpful, because there is no way to distinguish a chronic carrier with a viral syndrome from a child with mild or atypical GABHS pharyngitis. Rapid identification of the viral pathogen may be helpful.
A limitation of this study is that the data were obtained through retrospective chart review. Documentation of Centor criteria was limited by the accuracy and completeness of the medical records. We classified undocumented criteria as absent, on the basis of the expectation that physicians would consistently record pertinent positive findings, whereas negative findings might be omitted. This was confirmed by our secondary analysis of data for patients with medical record documentation of the presence or absence of ≥3 criteria. The findings on spectrum bias were unchanged for this group, indicating that no bias was introduced with the classification of criteria without medical record documentation. The validity of the Centor score classification method was also supported by the observation that GABHS prevalence increased as the modified Centor scores increased. The use of throat culture as the reference test was also a potential limitation, because it is recognized that culture is an imperfect standard test, despite sensitivities in the range of 90% to 95%.26 Reference test bias may occur when an imperfect standard test is used.27
In this study, we assumed that all cases with positive RADT findings would be culture-positive (100% specificity). The package insert for the test used in this study indicated that the test has a specificity of >97%. Clinical studies with the same RADT technology confirmed this level of specificity.19–21 Because the true specificity is likely to be <100%, the estimates of sensitivity reported here may overestimate the actual sensitivity of the test for each stratum of Centor scores. The implications of the aforementioned assumption are conveyed in Fig 1, where we display all possible sensitivity/specificity combinations for the observed data. The results from our primary analysis are presented at the right margin, where the specificity is 100%. If the RADT has the same specificity (<100%) in each Centor score stratum, then there will be larger differences in the sensitivities for the Centor score strata (ie, more severe spectrum bias) than presented in the primary analysis. Our conclusions regarding spectrum bias should hold true in realistic situations in which the specificities for the Centor score strata differ. For example, specificities of 100%, 95%, 90%, and 85% for the Centor score strata of 0, 1, 2, and 3 or 4 yield sensitivities of 47%, 50%, 77%, and 88%, respectively.
Spectrum bias should be considered when pediatric providers interpret RADT results. However, the sensitivity did not exceed 90% even among children with the highest pretest probability of GABHS infection, and the negative predictive value of the RADT was <95% for all groups of children, regardless of Centor scores. The data from this study do not support discontinuation of culture confirmation for children with negative RADT findings.
Funding for this study was provided by a cooperative agreement with the US Centers for Disease Control and Prevention (U50/CCU513299-01).
We thank the following individuals who contributed to this study: Debra Kempf, Carol Beyer, Theresa Esser, Sarah Grambsch, and Deborah Hilgemann. We also thank Richard Besser, MD, for critical review of the manuscript.
- ↵Webb KH. Does culture confirmation of high-sensitivity rapid streptococcal tests make sense? A medical decision analysis. Pediatrics.1998;101(2) . Available at: www.pediatrics.org/cgi/content/full/101/2/e2
- ↵Bisno AL, Gerber MA, Gwaltney JM Jr, Kaplan EL, Schwartz RH. Practice guidelines for the diagnosis and management of group A streptococcal pharyngitis. Clin Infect Dis.2002;35 :113– 125
- ↵Centor RM, Witherspoon JM, Dalton AP, et al. The diagnosis of strep throat in adults in the emergency room. Med Decis Making.1981;1 :239– 246
- Dajani A, Taubert K, Ferrieri P, Peter G, Shulman S. Treatment of acute streptococcal pharyngitis and prevention of rheumatic fever: a statement for health professionals. Pediatrics.1995;96 :758– 764
- ↵Maclure M, Greenland S. Tests for trend and dose response: misinterpretations and alternatives. Am J Epidemiol.1992;135 :96– 104
- ↵Yzerman EP, den Boer JW, Lettinga KD, Schellekens J, Dankert J, Peeters M. Sensitivity of three urinary antigen tests associated with clinical severity in a large outbreak of Legionnaires' disease in the Netherlands. J Clin Microbiol.2002;40 :3232– 3236
- ↵Dagnelie CF, Bartelink ML, Van der Graaf Y, Goessens W, De Melker RA. Towards a better diagnosis of throat infections (with group A β-haemolytic streptococcus) in general practice. Br J Gen Pract.1998;48 :959– 962
- ↵Kellogg JA. Suitability of throat culture procedures for detection of group A streptococci and as reference standards for evaluation of streptococcal antigen detection kits. J Clin Microbiol.1990;28 :165– 169
- ↵Deneef P. Evaluating rapid tests for streptococcal pharyngitis: the apparent accuracy of a diagnostic test when there are errors in the standard of comparison. Med Decis Making.1987;7 :92– 96
- Copyright © 2004 by the American Academy of Pediatrics