Retropharyngeal abscess due to Gemella morbillorum.

provided in the eligible studies) for each indicator [1]. Because their meta-analysis included a large number of subjects, these authors obtained narrow confidence intervals for virtually all variables; as these confidence intervals differed from the unit, the reader is led to believe that each one is an independent indicator of AOM. However, this might not be the case, because in pooling crude risk ratios, Uhari and co-workers failed to consider the possible role of confounding [2]. Confounding is not a big issue in randomized trials because randomization usually results in distribution of the confounders equally between the intervention groups, which allows unbiased pooling of the risk ratios. However, in nonexperimental (cohort and case-control) studies the crude risk ratio can be highly affected by other variables that are simultaneously associated with exposure and outcome and that are unevenly distributed in the two comparison groups [2, 3]. For example, in the meta-analysis of Uhari et al., breastfeeding, a protective factor against AOM, is likely to be inversely correlated with attendance at a day care center, a risk factor for AOM. If this is so, crude risk ratios overestimate the real magnitude of these indicators, which should more properly be assessed by the calculation of adjusted risk ratios. One can also speculate whether the variable "positive family history of AOM" is correlated with the variable "at least one sibling" (the reasoning being that the larger the family, the higher the probability of someone having AOM) and wonder whether the effect of the latter risk factor would not disappear when adjusted by the former risk factor. In meta-analyses of nonexperimental studies that are performed to assess associations where several risk factors play a role (and confounding is therefore expected), adjusted risk ratios, rather than crude risk ratios, should be pooled; this can be achieved by applying general variance-based methods [3]. I would advise readers to be cautious in the interpretation of crude figures such as those provided by Uhari and co-workers.