Methodological aspects of a meta-analysis of placebo-controlled trials of homeopathic treatment strategies

Tags: Oral
Linde K, Clausius N, Ramirez G, Hedges L, Eitel F, Melchart D, Jonas W

Introduction/Objective: We performed a meta-analysis of all available randomized, placebo-controlled clinical trials of homeopathic treatment strategies to investigate whether the effects of homeopathic treatment can be attributed to placebo. Within this meta-analysis we tried to investigate whether indicators of study quality had an influence on study results.

Methods: Randomized, placebo-controlled clinical trials of interventions including the application of homeopathic remedies for treatment or prevention of pathological conditions were eligible if sufficient data was presented to allow for inclusion into quantitative meta-analysis. Study quality was assessed with two score systems and by assessing single components (adequacy of randomization concealment, double-blinding, acceptability of loss to follow-up). Pooled odds ratios (with values > 1 indicating a superiority of homeopathic treatment vs. placebo) were calculated for all studies and various subsets and a cumulative meta-analysis according to study quality was performed.

Results: 89 trials met the inclusion criteria. The pooled random effects odds ratio for all trials was 2.45 (95%CI 2.05 to 2.93). Pooled odds ratios were lower for the subsets of trials exceeding a predefined minimum score (1.66), having adequate concealment (1.93), being double-blind (2.17), or having a predefined main outcome measure (2.27), while studies with little loss to follow-up showed larger effect sizes (3.18). Cumulative meta-analysis showed that effect sizes tended to increase with addition of studies with lower quality scores.

Discussion: The results of the actually available studies are not compatible with the hypothesis that all clinical effects of homeopathy are due to placebo. Consistent with most other empirical investigations on the impact of quality aspects on outcome our results indicate that studies with lower quality tend to inflate effect sizes.