Objective: To reconcile conflicting reports of antibiotic efficacy for otitis media with effusion in children.
Data sources: English-language MEDLINE search ("antibiotics" and "otitis" media with effusion") from January 1980 through December 1990. Current Contents 1990, consultation with experts, and references from review articles, textbook chapters, and retrieved reports.
Study selection: Randomized clinical trials with concurrent controls (placebo or no drug), and children with at least one ear not violated by tympanocentesis. Ten of the initial 82 articles were selected after blind review of the methods sections.
Data extraction: We independently evaluated each trial using 20 measures of internal and external validity, then extracted treatment and control responses for an end point of all affected ears free of effusion at the first posttreatment assessment.
Data synthesis: Pooled analysis of 1325 children yielded a rate difference of 22.8% (95% Cl, 10.5 to 35.1) that was minimally affected by interstudy quality differences, and was unlikely to represent publication bias. Variations in trial outcomes were not attributable to chance, study design, or choice of drug, but were inversely related to the control group natural cure rate. Children with chronic bilateral effusions not related to a recent episode of acute otitis media tended to have lower natural cure rates, and a more favorable response to therapy.
Conclusions: Antibiotics have a clinically and statistically significant impact on the resolution of otitis media with effusion. The association between outcome and natural cure rate has important implications for the design and interpretation of future trials.