This is the first large-scale study to investigate the effectiveness of antimicrobial treatment strategies for children with both drained and nondrained SSTIs. In this population-based cohort of nearly 50 000 children from Tennessee, use of trimethoprim-sulfamethoxazole or β-lactams for treatment of incident SSTIs was significantly associated with both treatment failure and recurrence, compared with use of clindamycin.
Among children who received drainage, use of trimethoprim-sulfamethoxazole or a β-lactam doubled the odds of treatment failure, compared with use of clindamycin. In the only other study comparing the effectiveness of clindamycin and trimethoprim-sulfamethoxazole for treatment of MRSA SSTIs among children after drainage, Hyun et al13
demonstrated no difference in the occurrence of treatment failure for >400 hospitalized children discharged with either orally administered clindamycin or trimethoprim-sulfamethoxazole. In contrast to our study, all of the children in that study were hospitalized initially and nearly 90% received parenterally administered clindamycin, which makes it difficult to compare the 2 treatment groups.
In a study examining the epidemiological features of purulent SSTIs in Tennessee, Talbot et al22
reported that, of 182 drained abscesses from a pediatric emergency department that were cultured, 142 (79.7%) were attributable to MRSA. Similarly, data from across the United States demonstrated the predominant organism in pediatric skin abscesses to be MRSA.4,23,24
Therefore, it is likely that the majority of purulent infections in our cohort also were attributable to MRSA. This would explain the increased risk of treatment failure for the β-lactams, which uniformly lack antimicrobial activity against MRSA; however, the mechanism for increased treatment failures in the trimethoprim-sulfamethoxazole group is unknown.
Despite high rates of in vitro susceptibility to trimethoprim-sulfamethoxazole among CA-MRSA isolates,25
failure rates of nearly 20% among adults treated with trimethoprim-sulfamethoxazole for SSTIs in a clinical trial setting16
and as high as 50% among adults with significant comorbidities26,27
have been reported. Moreover, 2 recent randomized trials, although they likely were underpowered to detect small differences, failed to demonstrate superiority of trimethoprim-sulfamethoxazole over placebo for treatment of SSTIs after drainage among both children15
(neither study included an active comparator). In the pediatric trial by Duong et al,15
4% of children who received trimethoprim-sulfamethoxazole experienced treatment failure and an additional 13% developed new lesions within 10 days. Trimethoprim-sulfamethoxazole inhibits bacterial growth by blocking key steps in the biosynthesis of thymidine, halting bacterial DNA replication. However, S aureus
, through its ability to liberate and to acquire free thymidine from DNA fragments (which are present in high concentrations in abscess fluid), may bypass the antimicrobial effects of trimethoprim-sulfamethoxazole.26
This mechanism might explain the increased risk of treatment failure associated with trimethoprim-sulfamethoxazole.
Although recurrences were common among all children who received drainage, recurrence risk increased by nearly 30% for those who received trimethoprim-sulfamethoxazole and by >40% for those who received β-lactams, compared with clindamycin. Because persistent colonization with S aureus
(including CA-MRSA) has been linked to the development of clinical infection,28,29
it is possible that clindamycin is superior at preventing recurrence because it decreases colonization more effectively.30
Another potential mechanism for the increased recurrence risk associated with trimethoprim-sulfamethoxazole may be the development of staphylococcal small-colony variants, slow-growing, intracellular organisms that have been isolated among persons receiving chronic trimethoprim-sulfamethoxazole therapy and patients with recurrent SSTIs.26,31
However, the development of this phenotype among a large population of children with incident SSTIs seems unlikely.
Among children with nondrained SSTIs, the risk of treatment failure remained significantly higher for those treated with either trimethoprim-sulfamethoxazole or β-lactams, although the magnitude of the effect was decreased, compared with children who underwent drainage. Risk of recurrence also was significantly increased among trimethoprim-sulfamethoxazole users but not among users of β-lactams. In contrast to SSTIs that underwent drainage (for which CA-MRSA likely was the predominant organism), SSTIs that did not undergo drainage likely were more heterogeneous. Some children might have had small abscesses that were not amenable to drainage or that drained spontaneously, both likely attributable to CA-MRSA. Nonpurulent SSTIs, however, likely were attributable to a mixture of Streptococcus pyogenes
and S aureus
. Consistent with this hypothesis, nearly 95% of children who received drainage had an ICD-9-CM–coded diagnosis of abscess/cellulitis, whereas only 60% of those without a drainage procedure had this diagnosis. Trimethoprim-sulfamethoxazole is largely ineffective for streptococcal infections, whereas both clindamycin and β-lactams possess potent activity against S pyogenes
. In a recent, nested, case-control study with >2000 children with nondrained, noncultured SSTIs, trimethoprim-sulfamethoxazole doubled the risk of treatment failure, compared with β-lactams, whereas clindamycin conferred no benefit over β-lactams.14
Therefore, the use of β-lactams for nondrained, nonpurulent SSTIs may still be appropriate even in areas with high CA-MRSA prevalence.
Although several prospective treatment trials are ongoing, results likely will not be available for several years; therefore, we selected a retrospective, population-based cohort to investigate the effectiveness of these commonly used antimicrobial agents for pediatric SSTIs. Observational studies that use previously validated administrative data offer several advantages over randomized trials, including larger sample sizes, long-term follow-up monitoring and outcome assessment (in a relatively short period of time), and the ability to assess treatment effectiveness (versus efficacy). The TennCare administrative data files have been used for pharmacoepidemiological research for >30 years and have been validated extensively. In addition, a validation sample confirmed the predictive value of our SSTI definitions.
Despite the study's strengths, several limitations should be discussed, including residual confounding, the lack of microbiologic data, and potential misclassification of antibiotic exposure and outcomes. To control for potential confounding, several patient-level characteristics, as well as study year and diagnosis, were included in the multivariate models. We did find evidence for confounding according to race, because the change in the adjusted models was attributable primarily to this variable. Black individuals were least likely to receive a β-lactam and were most likely to receive clindamycin. The reasons for treatment differences according to race are unclear; however, they might stem from a perceived increased risk for MRSA among black individuals, as well as regional variations in treatment preferences. Although comorbidities were not assessed, there is no indication that children with comorbidities would be more likely to receive one antibiotic over another.
Data on microbiologic causes are important for understanding the mechanisms governing treatment failure and recurrence; even if cultures are obtained, however, results are not immediately available and initial treatment decisions are nearly always empiric. Compared with children who received clindamycin or trimethoprim-sulfamethoxazole, those who received a β-lactam might have had less severe disease, because β-lactams are not recommended for suspected CA-MRSA infections. In contrast, both clindamycin and trimethoprim-sulfamethoxazole are first-line, outpatient, antimicrobial agents when MRSA is a consideration, and both are used frequently in clinical practice. Although we were unable to control for lesion size, location, or character, it is unlikely that these characteristics differed among children who received either clindamycin or trimethoprim-sulfamethoxazole. Confounding according to indication was further limited through restriction of the cohort to outpatient visits for children who received a single antibiotic and stratification according to drainage status. Although it represents a question of great interest, we did not intend to evaluate the effectiveness of drainage alone, because of concerns regarding systematic differences between this group and the group of children who received antimicrobial treatment. However, a posthoc analysis that included children who received drainage alone, compared with clindamycin treatment, revealed an increased risk of treatment failure but not recurrence for this group (Supplemental Table 9). Because of the likely differences mentioned above, however, these results should be interpreted with caution.
We were unable to account for medication noncompliance or antibiotic changes without an additional SSTI visit; however, antibiotic changes without an additional visit should be rare, and prescription filling has been shown to predict medication adherence reliably.32,33
Palatability also is likely a major determinant of drug adherence in children, although clindamycin is notoriously unpalatable, compared with other antibiotics34,35
; therefore, it would not be expected that different levels of adherence would explain the decreased risk of treatment failure and recurrence associated with clindamycin. Occasionally, antibiotic changes at subsequent follow-up visits (within 14 days) might have been based on culture results and not a true treatment failure. This would most likely affect children who received a β-lactam, because CA-MRSA causes the vast majority of purulent SSTIs in our region and both clindamycin and trimethoprim-sulfamethoxazole demonstrate good in vitro activity against local CA-MRSA isolates.
It is inevitable that a small proportion of secondary SSTIs that occurred ~14 days after the incident SSTI were misclassified (treatment failures classified as recurrences and vice versa), although any misclassification would be nondifferential. Among children who did not receive a drainage procedure, it was impossible to determine whether children had purulent infections (ie, abscesses) or nonpurulent cellulitis, because these diagnoses fall under the same ICD-9-CM code. This likely contributed to heterogeneity among these children and may have affected outcome assessment for this group. Similarly, a small number of children who received drainage might have been assigned to the group without drainage if a drainage procedure claim was not filed.