Antibiotics are not effective for otitis media with effusion: Reanalysis of meta-analyses

Objective: To determine the effectiveness of antibiotic therapy for otitis media with effusion (OME) and to examine the validity of meta-analysis for combining the results of published randomized controlled trials (RCTs) to reach generalized efficacy conclusions. Data Sources: English-language publications in refereed journals and proceedings published be tween 1980 and 1997, inclusive, producing 16 usable RCTs involving 2,067 children. Study Selection: RCTs evaluating the efficacy of antimicrobials for the treatment of OME with or without placebo controls. Data Synthesis: We used the technique of Der Simonian and Laird to determine heterogeneity and estimated the mean rate difference (RD). We also used weighted regression analysis to evaluate efficacy trends in light of changing quality of the included randomized trials, and sensitivity analysis to investigate the effects of quality differences among the clinical trials on estimated RDs. Results: The placebo-controlled trials had RDs significantly lower than non-placebo-controlled trials (RD of 0.04 vs. RD of 0.32; p < 0.001). When conventional techniques of meta-analysis were used for data synthesis, the presence of significant heterogeneity (p < 0.0001) precludes reaching any efficacy conclusions. Using weighted regression analysis to account for the effects of study quality on efficacy, we found that antimicrobial efficacy for OME is insignificant, less than 2% for an ideal clinical trial (p = 0.660). Conclusions: We conclude that meta-analysis as applied conventionally is inadequate to detect selection bias and is not useful for estimating true RDs when the results of the RCTs vary systematically with study quality. Our meta-analysis fails to support the continued use of antibiotics in the treatment of OME.