To the Editor—Several studies have demonstrated a negative association between carriage of Streptococcus pneumoniae (specifically, carriage of the vaccine types [VTs]) and Staphylococcus aureus in children [1–6]. This has raised concern that, similar to the occurrence of serotype replacement with non-VT S. pneumoniae [7–11], the widespread use of 7-valent pneumococcal conjugate vaccine (PCV7) and the reduction in both VT S. pneumoniae and total S. pneumoniae carriage will be followed by an increase in S. aureus carriage [12, 13].
Cohen et al.  have tried to address this issue in their recently published study, in which they conclude that PCV7 does not appear to influence S. aureus carriage in young children with otitis media. However, their study suffers from several flaws, and we believe that their conclusions may be unjustified.
Our main concern is methodological. Although the authors performed multivariate analysis, critical variables were not included in the analysis, thus rendering the results largely irrelevant. The choice of variables to include in a multivariate analysis is highly dependent on the question being asked. If the purpose of the study was to evaluate whether receipt of PCV7 modified the risk of staphylococcal colonization, a multivariate analysis to address this question must include this predictor. If, on the other hand, the purpose was to reassess whether S. pneumoniae carriage is associated with S. aureus carriage (as several independent studies have shown), the predictor “S. pneumoniae carriage” should have been included in the multivariate analysis. Because neither of these analyses was performed, the authors' 2 conclusions, that PCV7 does not affect S. aureus colonization and that S. pneumoniae carriage is not associated with S. aureus carriage, are not proven. Finally, age is known to be one of the most important variables associated with colonization by S. pneumoniae and S. aureus [2, 15, 16], and children who were colonized with S. aureus in the study by Cohen et al.  were younger than uncolonized children (P = .06). Therefore, we believe that the variable age should have also been included in the multivariate analysis. The multivariate analysis presented here is thus flawed in 2 important ways: it does not test the primary hypothesis, and it does not include probable confounding factors. Naturally, we cannot be certain what results would be obtained in a proper analysis, but the present analysis is uninformative because of these flaws.
Another concern is the method of swab transfer. The authors state that swabs were transferred to the lab within 48 h after sample collection. This is a long delay, but more worrisome is the fact that the length of the delay was not consistent. Some swabs could have been transferred within hours after sample collection, with full recovery of the bacteria, whereas others could have been transferred after a delay of 48 h, resulting in a low recovery rate. This could be a cause of misclassification bias. An additional concern is why the authors chose to perform a detailed analysis on colonization data (presented in table 2 ) from children >1 year of age only, despite the fact that 41% of their study population was <1 year of age. The authors give no explicit reason for this partial analysis.
Finally, the article contains numerous obvious errors or inconsistencies that make it difficult to follow and, at times, impossible to interpret. In table 2 , the total number of children in the subgroups exceeds the number of children in the total group (for example, 9.4% of the children carried both S. aureus and VT S. pneumoniae, whereas only 7.8% of the children carried S. aureus). In addition, the weighted averages are miscalculated for some of the rows (VT S. pneumoniae and S. aureus plus S. pneumoniae). Overall, it is difficult to determine the number of PCV7-vaccinated children in the study. At one point, the authors write that 17% were vaccinated; later, they refer to much higher numbers, whereas in table 1, they state that 82.5% of the children (1470 of 1781 children) were vaccinated .
It is clear that PCV7 has dramatically reduced invasive pneumococcal disease, both among vaccinated children and among nonvaccinated individuals, via indirect effects on transmission . However, the growing concerns about the emergence of non-VT disease [7, 8, 10, 11] in vaccinated populations and the possible effects of this vaccine on the epidemiology of other upper respiratory pathogens, including S. aureus, warrant careful investigation in well-designed and adequately reported studies.
Potential conflicts of interest. All authors: no conflicts.