Homeopathy: meta-analyses of pooled clinical data

Abstract

In the first decade of the evidence-based era, which began in the mid-1990s, meta-analyses were used to scrutinize homeopathy for evidence of beneficial effects in medical conditions. In this review, meta-analyses including pooled data from placebo-controlled clinical trials of homeopathy and the aftermath in the form of debate articles were analyzed. In 1997 Klaus Linde and co-workers identified 89 clinical trials that showed an overall odds ratio of 2.45 in favor of homeopathy over placebo. There was a trend toward smaller benefit from studies of the highest quality, but the 10 trials with the highest Jadad score still showed homeopathy had a statistically significant effect. These results challenged academics to perform alternative analyses that, to demonstrate the lack of effect, relied on extensive exclusion of studies, often to the degree that conclusions were based on only 5-10% of the material, or on virtual data. The ultimate argument against homeopathy is the 'funnel plot' published by Aijing Shang's research group in 2005. However, the funnel plot is flawed when applied to a mixture of diseases, because studies with expected strong treatments effects are, for ethical reasons, powered lower than studies with expected weak or unclear treatment effects. To conclude that homeopathy lacks clinical effect, more than 90% of the available clinical trials had to be disregarded. Alternatively, flawed statistical methods had to be applied. Future meta-analyses should focus on the use of homeopathy in specific diseases or groups of diseases instead of pooling data from all clinical trials.