Initial Antibiotic Choice in the Treatment of Group A Streptococcal Pharyngitis and Return Visit Rates

Abstract

Background

Our objectives were to describe the incidence of return visits for children with Group A Streptococcal (GAS) pharyngitis (ie, clinical treatment failure) and to assess whether initial treatment with amoxicillin or penicillin was associated with the rate of retreatment for GAS pharyngitis.

Methods

This analysis was a retrospective cohort study of 5533 children 0–17 years from a multisite practice. Eligible visits (n = 6585) were associated with a positive test for GAS, receipt of antibiotics within 7 days, no allergies to penicillins or cephalosporins, and no codiagnoses requiring antibiotic treatment. Retreatment for GAS pharyngitis was defined as an index visit followed by another visit within 1–4 weeks. Five hundred episodes (250 treatment failures and 250 controls) were randomly selected for chart review to validate cases.

Results

Amoxicillin or penicillin was the initial antibiotic treatment at 76.1% of visits, and retreatment for GAS pharyngitis occurred after 5.8% of initial visits. Children initially prescribed amoxicillin or penicillin had higher odds of retreatment of GAS pharyngitis even after adjusting for age, sex, symptoms, and community-level covariates such as race, income, and education (odds ratio, 1.51; 95% confidence interval, 1.07–2.13).

Conclusions

Retreatment for GAS pharyngitis was uncommon and associated with receipt of amoxicillin or penicillin, although the impact of GAS carriage is unknown. Recommendations for initial treatment of GAS pharyngitis should reflect both individual and societal considerations, including the potential impact on antibiotic resistance in the community.

BACKGROUND

Pharyngitis accounts for 10% of total antibiotic prescriptions among children in the United States [1, 2]. Only 15%–36% of children will have a bacterial cause such as group A Streptococcus (GAS) [3]. According to the National Ambulatory Medical Care Survey in 1998, 8.7 million antibiotic prescriptions were given to outpatients for pharyngitis at an estimated cost of $215 million each year [4].

Optimal treatment of GAS has become more controversial in recent years. Penicillin is the first-line agent recommended by the American Academy of Pediatrics [5], the Infectious Diseases Society of America [6], and the Centers for Disease Control [7]. However, multiple studies have shown that cephalosporins are superior to penicillin in eradication of GAS from the pharynx [8–11]. Moreover, there has been some suggestion that failure of penicillin in bacteriologic cure is increasing [9, 12]. Based on this evidence, some experts are calling for first-generation cephalosporins to be the new agent of choice [13]. Such a change could have a significant impact, because such agents are more expensive and unnecessarily broad compared with penicillin [5].

Moreover, bacteriologic cure may not be the ideal outcome measure for GAS pharyngitis. Between 5% and 25% of children go on to become asymptomatic carriers of GAS after appropriate treatment, and they are at no higher risk of invasive disease than children with bacteriologic cure [14, 15]. Most tellingly, the incidence of rheumatic fever has not changed over the last decade despite the increasing failure of penicillin in bacteriologic clearance of GAS pharyngitis [16].

Given the uncertain significance of bacteriologic cure, clinical cure of GAS pharyngitis is arguably a more appropriate endpoint. Persistent symptoms due to pharyngitis leading to return visits and retreatment are associated with missed parental days of work and missed days of school [17, 18]. Multiple antibiotic courses result in costs to both families and the healthcare system and lead to frustration on the part of parents and physicians. The reasons for treatment failure remain unclear, despite the many proposed explanations [19], including poor compliance, carrier state with a viral upper respiratory infection misdiagnosed as new GAS infection, recurrent exposure to GAS, inadequate antibiotic penetration, and in vivo eradication of normal protective flora.

In light of the health and economic burden of GAS pharyngitis, the optimal choice of antibiotic to minimize return visits and retreatment of GAS pharyngitis has important implications. The purpose of this study is to assess whether initial treatment with amoxicillin or penicillin is associated with return visits and retreatment for GAS pharyngitis, as well as to describe prescribing patterns for GAS pharyngitis in a large multicenter practice.

METHODS

Setting and Patients

A retrospective cohort of children aged 0–17 years was assembled from Harvard Vanguard Medical Associates, a multispecialty multisite practice in Boston with ∼70 000 children enrolled. Data were available from 13 practice sites that treated children. Visits made between May 1, 2006 and October 31, 2007 were identified using an electronic health record. Visits were eligible for inclusion if associated with (1) a positive test for GAS (rapid strep test or throat culture) and (2) antibiotic treatment (both oral and intramuscular) within 7 days of the visit. Visits were excluded if there was (1) a documented allergy to penicillins or cephalosporins or (2) the presence of ICD-9 codes representing codiagnoses requiring antibiotics (pneumonia, otitis media, urinary tract infection, Lyme disease, sinusitis, skin and soft tissue infections, mastoiditis, and lymphadenitis).

The outcome of interest was retreatment for GAS pharyngitis, defined as an index visit that was followed within 1–4 weeks by another visit associated with a positive test for GAS and antibiotic retreatment within 7 days (to avoid inclusion of children perceived to be asymptomatic carriers by their healthcare provider and thus not treated). Previous studies have excluded children with a repeat positive test within the first week of treatment [14]. Children who returned for another episode of GAS pharyngitis were included in the cohort as long as the episode occurred more than 4 weeks after the last visit.

From this cohort, we conducted medical record review on a subset of visits to validate our classification of retreatment for GAS (Figure 1). We also performed a nested, unmatched case-control study to evaluate whether initial clinical presentation was associated with retreatment for GAS pharyngitis. Cases were those visits that resulted in a return visit and retreatment for GAS, and control visits were those from the cohort that did not require retreatment for GAS pharyngitis. A total of 500 visits were selected for the nested case-control study at a ratio of 1:1 to yield a power of 80% to detect a difference of 10% between proportions of cases and controls receiving penicillin or amoxicillin.

Figure 1.

Identification of visits for the cohort and case-control study. Abbreviation: GAS, group A Streptococcus.

Data Analysis

In the cohort study, data were available on age, sex, date of visit, type of test performed, oral antibiotic(s) prescribed (but not dose), type of provider, and practice site. Age was classified into <6 years, 6–14 years, and 15–17 years [20]. Season of presentation was categorized into fall (September–November), winter (December–February), spring (March–May), and summer (June–August) [5]. The type of test that was positive at the index visit was dichotomized into rapid strep test positive versus culture positive only. The primary predictor of interest was initial antibiotic prescribed, which was dichotomized into amoxicillin or penicillin versus all other antibiotics. For patients who had census tract information available (85%), community-level predictors including race, median income, and education level were included in analyses.

For patients in the case-control study, presenting signs and symptoms were available, including presence or absence of sore throat, fever (reported by patient or measured in the office), headache, abdominal pain, cough, rhinorrhea, and physical exam findings, including adenopathy, tonsillar exudate, tonsillar swelling, and rash. Select symptoms were amalgamated into a modified Centor score [21]. The retrospective nature of the chart review made it necessary to slightly modify this score (Table 1). We chose the highest possible cutoff score of 4 or higher, because this predicts an approximately 50% risk of streptococcal infection, with some experts even suggesting empirical use of antibiotics in such patients [21].

Table 1.

Use of Modified Centor Score

Criteria		Points
Modified Centor score [21]	Score as used in this study
Temperature >38˚C	Any fever	1
Absence of cough	Absence of cough	1
Tender, anterior, cervical adenopathy	Cervical adenopathy	1
Tonsillar swelling or exudate	Tonsillar swelling or exudate	1
Age	Age
3–14 years	0–14 years	1
15–44 years	15–17 years	0
≥ 45 years	Not applicable	−1

Bivariate analyses were conducted to assess the relationship between retreatment for GAS pharyngitis and potential predictors using the χ² test of association. Multivariable logistic regression analyses were conducted to assess independent predictors of clinical treatment failure in children. Potential predictors in the cohort model included age, sex, amoxicillin/penicillin as initial antibiotic, and type of test as well as community-level predictors such as race, income, and education. For the case-control study, modified Centor score was also included. Because some children could contribute more than 1 visit, generalized estimating equations were used to produce robust standard errors that would adjust for clustering by child and by practice site. Predicted probabilities for retreatment for GAS pharyngitis were generated from the cohort multivariate model. All analyses were performed with Stata 9.0 (Stata, College Station, TX).

RESULTS

Study Population

After excluding visits with a codiagnosis requiring antibiotics or an allergy to penicillins or cephalosporins, there were 6585 index visits in the cohort associated with a positive test for GAS phyarngitis that were treated with antibiotics, made by 5533 children between May 1, 2006 and October 31, 2007 (Table 2). There was an average of 1.2 index visits per child for GAS during this 18-month period. Cases were similar to controls in terms of age, sex, and type of provider (nurse practitioner vs medical doctor) seen at the index visit. However, cases were significantly more likely to have had a positive rapid strep test than controls.

Table 2.

Characteristics of All Index Visits for Group A Streptococcus Pharyngitis

Variable	All Index Visits (N = 6585)	Treatment Successes (N = 6203)	Treatment Failures (N = 382)	P Value
Age	8.4 ± 3.8 y	8.4 ± 3.8 y	8.2 ± 3.3 y	.13
Male sex	3415 (51.9%)	3211 (51.8%)	204 (53.4%)	.53
Initial antibiotic prescribed
Amoxicillin/penicillin	5017 (76.2%)	4694 (75.7%)	323 (84.6%)	.002
Cephalexin	1226 (18.6%)	1179 (19.0%)	47 (12.3%)
Amoxicillin-clavulanate	173 (2.6%)	168 (2.7%)	5 (1.3%)
Azithromycin/erythromycin	160 (2.4%)	153 (2.5%)	7 (1.8%)
Other	9 (0.2%)	9 (0.2%)	0 (0%)
Type of test positive at index visit
Rapid strep	5345 (81.2%)	4994 (80.6%)	351 (91.9%)	<.001
Throat culture only	1240 (18.8%)	1209 (19.4%)	31 (8.1%)	<.001
Provider type
MD	3964 (60.2%)	3730 (60.1%)	234 (61.2%)	.87
NP	2226 (33.8%)	2105 (33.9%)	121 (31.7%)
Other (PA, DO, Unknown)	395 (6.0%)	368 (5.9%)	27 (7.1%)

Abbreviations: DO, doctor of osteopathy; MD, medical doctor; NP, nurse practitioner; PA, physician's assistant.

Charts were available for review for 248 of 250 case visits (ie, index visits that required subsequent retreatment for GAS pharyngitis) and 248 of 250 control visits (ie, index visits with no retreatment). There were no significant differences in initial antibiotic received, age, sex, race, income, or education among selected cases compared or among selected controls. Chart review revealed that 6 of the 248 controls were misclassified and were in fact cases. Of these misclassified controls, information for both index visit and return visit were available for 5 of the 6 visits. No cases were misclassified as controls. All cases presented for the return visit due to new symptoms, as opposed to test of cure for GAS pharyngitis.

Return Visits for Retreatment (Cohort Study)

Clinicians prescribed amoxicillin or penicillin at 76.1% of all index visits and a cephalosporin at 18.6% of all index visits (Table 2). The index visit resulted in a return visit and retreatment for GAS pharyngitis 5.8% of the time, accounting for an additional 382 visits. Approximately half of the return visits (47%) occurred within 2 weeks of the index visits, with the majority (82%) occurring by 3 weeks. The mean timing of the return visit was 16.5 days (± 5.1 days).

The rate of retreatment for GAS pharyngitis was 6.4% among children who initially received amoxicillin or penicillin, and 3.8% among children who received antibiotics other than amoxicillin or penicillin (3.8% for cephalexin, 2.9% for amoxicillin-clavulanate, 4.4% for azithromycin/erythromycin). To prevent 1 return visit associated with retreatment of pharyngitis, the number needed to treat with a nonamoxicillin or nonpenicillin would be 38. At return visits associated with retreatment for GAS pharyngitis (N = 382), cephalosporins were prescribed more commonly (62.9%) than amoxicillin or penicillin (15.4%).

Symptoms and Modified Centor Score (Case-Control)

Case visits and unmatched control visits were associated with similar frequencies of symptoms at the index visit (Table 3). However, cough was significantly more common at control visits (24.4% vs 17.4%, P = .046). Case visits were somewhat more likely to be associated with a modified Centor score of 4 or higher at the index visit compared with controls (40.7% vs 33.5%; P = .096).

Table 3.

Presenting Signs and Symptoms for Controls and for Cases at Initial Visits for Group A Streptococcus Pharyngitis

	Cases (N = 253)	Controls (N = 242)	P Value
Sore throat	87.4% (221)	81.0% (196)	.052
Fever	58.5% (148)	63.2% (153)	.28
Headache	27.3% (69)	24.8% (60)	.53
Abdominal pain	21.7% (55)	21.9% (53)	.96
Cough	17.4% (44)	24.4% (59)	.046
Rhinorrhea	21.0% (53)	26.4% (64)	.15
Cervical adenopathy	52.6% (133)	47.9% (116)	.30
Tonsillar exudate or swelling	28.5% (72)	23.6% (57)	.21
Mean Centor score (±1 standard deviation)	3.2 ± 1.0	3.0 ± 1.1	.15
Centor score ≥ 4	40.7% (103)	33.5% (81)	.096

Predictors of Return Visits for Retreatment

In bivariate analyses of the full cohort, those who were retreated for GAS pharyngitis were significantly more likely to have received amoxicillin or penicillin as the initial antibiotic, more likely to be aged 6–14 years, more likely to have tested positive by rapid strep at the index visit, and least likely to present in the summer. In a multivariate model for the cohort that controlled for clustering by patient, site of visit, and race, income, and education, receipt of amoxicillin or penicillin remained significantly associated with retreatment, with an odds ratio (OR) of 1.56 (95% confidence interval, 1.11–2.20). Testing positive by rapid strep at the initial visit was also significantly associated with retreatment, as was presenting in the fall, winter, and spring compared to summer (Table 4).

Table 4.

Multivariate Models Predicting Clinical Treatment Failure of Group A Streptococcus Pharyngitis^a

	Cohort (N = 5831)	Case-Control Study (N = 419)
	OR [95% CI]
Initial antibiotic
Amoxicillin or  penicillin	1.51 [1.07–2.13]	1.32 [.77–2.28]
Cephalexin or other	1.0	1.0
Male sex	1.01 [.80–1.27]	0.86 [.56–1.32]
Age
<6 years	1.31 [.74–2.31]	1.67 [.63–4.43]
6–14 years	1.51 [.88–2.57]	1.68 [.67–4.18]
15–17 years	1.0	1.0
Type of test positive at index visit
Rapid strep	2.64 [1.76–3.96]	1.62 [.87, 3.03]
Throat culture only	1.0	1.0
Season
September–  November	1.56 [1.07–2.27]	1.52 [.74–3.11]
December–February	1.58 [1.10–2.26]	1.15 [.62–2.16]
March–May	1.44 [1.04–2.01]	1.17 [.63–2.16]
June–August	1.0	1.0
Centor scoreb
1		1.09 [.35, 3.39]
2		1.26 [.55, 2.86]
3		1.15 [.54, 2.42]
4		1.79 [.85, 3.77]
5		1.0

Abbreviations: CI, confidence interval; OR, odds ratio.

^aAdjusted for race, income, and education at the census tract level, and site of visit.

^bCentor score of 0 not included because it perfectly predicted probability of failure.

In the case-control study (N = 419 because of attrition due to missing demographic information), none of the variables examined were significantly associated with retreatment in the bivariate analyses, including initial antibiotic received, age, sex, type of test, season, symptoms, and modified Centor score. The multivariate model for the case-control study was identical to that for the cohort, but it also included modified Centor score as an independent predictor. No predictors were significantly associated with retreatment (Table 4). Leaving the Centor score out of the multivariate model did not change the significance of the remaining predictors.

Estimating Population Risk for Treatment Failure

Predicted incidence of repeat visits associated with retreatment for GAS pharyngitis if the entire study population were to be given amoxicillin or penicillin would be 6.1% (± 2.6%). If the entire study population were to be given cephalexin or other antibiotic, the predicted incidence would be 4.0% (± 1.8%).

DISCUSSION

The results from the cohort portion of our study suggest that receipt of amoxicillin or penicillin as the initial antibiotic and season increased the risk of return visits associated with retreatment for GAS pharyngitis. However, retreatment was still fairly uncommon in this cohort, with an overall prevalence of 5.8%. In our case-control study, with validation of outcomes and adjusting for potential confounding by Centor score, receipt of amoxicillin or penicillin as initial antibiotic was not significantly associated with retreatment for GAS pharyngitis, although our power may have been limited.

Overall, the rate of retreatment for GAS pharyngitis was quite low in our cohort. Other studies that have followed children out to at least 20 days beyond completion of treatment have found varying rates of clinical treatment failures. These rates range from as low as 6.3% [22] to 16% [23, 24] depending on the initial treatment administered. Because all of the sites in our study are concentrated in 1 metropolitan area, it is possible that the local treatment failure rates differ from those in other locations. Variation in bacteriologic failure rates around the United States has been documented, and clinical treatment failure rates (for which return visits and retreatment is a reasonable proxy) may have similar geographic variation [24, 25]. In addition, we only have information about follow-up that patients would have received within the provider network. If patients sought medical care at other facilities, such as emergency rooms or hospital clinics, we could have underestimated the rate of treatment failure.

The finding that initial treatment with amoxicillin or penicillin is associated with clinical treatment failure resulting in retreatment—and not just bacteriologic failure—has been suggested previously [9, 23, 24]. However, both the overall risk of clinical treatment failure and its potential predictors may be biased. Children with chronic GAS carriage who actually have viral pharyngitis may not always be accurately identified by clinicians, and our study likely included a substantial portion of carriers. Indeed, we observed a high incidence of both cough and rhinorrhea at the index visit associated with treated GAS pharyngitis. Because amoxicillin or penicillin do not eradicate throat carriage as well as other agents (such as cephalosporins, clindamycin, penicillin together with rifampin, or amoxicillin-clavulanate), carriers who were treated with amoxicillin or penicillin would have been more likely to continue to be positive for GAS at return visits. Thus, a small beneficial effect of alternative agents may be observed simply because distinguishing between treatment failure and chronic carriage can be difficult.

Nonetheless, if we assume that our findings represent the upper bound of (1) the rate of return visits and retreatment for GAS pharyngitis and (2) the relative contribution of the choice of antibiotics for treatment, at least 38 children would need to be treated with a cephalosporin to prevent 1 return visit with retreatment. The advantages of preventing such visits and retreatment through the use of a cephalosporin must be balanced against the very real disadvantages. First-generation cephalosporins cost more than penicillin and given the volume of prescriptions for pharyngitis [4], even small incremental differences in cost could be amplified. Perhaps more important is the unknown effect of widespread and unnecessary cephalosporin use on antibiotic resistance. The costs associated with potential antibiotic resistance are difficult to quantify, but they would likely be very high [26]. The cephalosporin use in our study at the index visit for GAS was quite high at approximately 19% overall, especially given that we had excluded children with allergies from the cohort. The high rate of cephalosporin use may have been partially accounted for by the fact that we could not exclude children with a more remote history of recurrent GAS (ie, earlier than 4 weeks before the index visit).

Finally, presenting in the fall, winter, or spring (compared to summer) was identified as a significant predictor of return visits resulting in retreatment in our cohort. Group A Streptococcal pharyngitis is more common during the late fall, winter, and spring in temperate climates, presumably to be due to close person-to-person contact in the school setting [5]. Thus, the children in our study may have been more likely to be reinfected with GAS (or alternatively, recolonized) due to higher risk of exposure during these seasons. Of note, we did oversample the late spring, summer, and early fall months in our sample to gain adequate power. However, we accounted for seasonal variability in confounders or effect modifiers by controlling for season in the multivariate analysis.

Our study does have some additional limitations. Given the study's retrospective nature, it was impossible to assess compliance with antibiotic therapy, recurrent exposure resulting in reinfection, or carrier status of the children. It is possible that the children with positive GAS tests were actually carriers and returning for care because of coincident infections such as viral upper respiratory tract infections. In fact, the presence of cough was high among both cases and controls, indicating this very possibility. However, our chart review revealed that children returning for visits for retreatment were at least as likely as those who did not to have symptoms associated with GAS pharyngitis at both the index and follow-up visits. In addition, cases were significantly more likely than controls to present with a positive rapid strep test at the index visit, and this remained a positive predictor of treatment failure in the cohort model. This finding could indicate a higher initial inoculum, meaning that cases may have been more likely to have true infection at the index visit, as opposed to viral infection in a GAS carrier.

In the case-control portion of the study, we identified 6 children (of 248) who were misclassified as cases when they were in fact controls. Assuming that this is likely to have occurred in the overall cohort study as well, this would make the controls more similar to the cases and thus bias our OR in the cohort study towards the null.

Children were not randomized to receive antibiotic treatment, which could result in selection bias. Instead, providers used their usual clinical judgment to choose an initial antibiotic to treat GAS pharyngitis. This could lead to confounding by indication, meaning that clinicians might be more likely to prescribe a nonpenicillin to children who are sicker at presentation. Our data suggest that those who returned for retreatment were sicker at presentation. If this confounding existed it would, if anything, bias our OR towards 1.0, because the sicker children who received nonpenicillins would be less likely to fail treatment and return. The other possible issue related to confounding by indication is that children who had multiple recurrences of GAS pharyngitis in the past may have been more likely to receive a nonpenicillin for the index episode in our study. Again, this could bias our risk estimates towards the null as such children may be more prone to recurrent GAS pharyngitis and more likely to receive a nonpenicillin initially because of their history. In making decisions about optimal initial antibiotic treatment for GAS pharyngitis, return visits—whether for true treatment failure or not—may still be a relevant outcome for patients, providers, and policymakers.

CONCLUSIONS

Retreatment of GAS pharyngitis was associated with receipt of amoxicillin or penicillin as the initial antibiotic and season in our cohort, although this risk may be due in part to misclassification of patients with chronic GAS carriage. However, the overall rate of return visits associated with retreatment of GAS pharyngitis was still quite low. Any analysis to determine the benefit-risk balance of changing the first-line antibiotic for the treatment of GAS pharyngitis from penicillin should take into account both individual and societal considerations, including the hidden costs of potential impact on antibiotic resistance in the community.

Acknowledgments

We thank Yury Vilk at the Research Data Services Center of Harvard Pilgrim Health Care and Ken Kleinman at the Department of Population Medicine at Harvard Medical School and Harvard Pilgrim Health Care. In addition, we thank Qiaoli Chen and Susan McDermott at the Clinical Research Program of Children's Hospital Boston.

Financial support. This work was supported by Agency for Healthcare Research and Quality (Grant Numbers T32 HS000063-13 [to C. A. G. through the Harvard Pediatric Health Services Research Fellowship Program] and 5 K08 HS013808-04 [to G. M. L.]).

Disclaimer. Neither sponsor had a role in any of the following processes: design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. Dr. Gidengil had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Potential conflicts of interest. All authors: No reported conflicts.

All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.