Efficacy of homoeopathic treatment: Systematic review of meta-analyses of randomised placebo-controlled homoeopathy trials for any indication.

The short version

This systematic review examined six earlier meta-analyses (statistical pooling studies) of placebo-controlled randomized trials of homeopathy for any health condition in humans, asking whether homeopathy shows an effect beyond placebo. Homeopathy is built on drug provings (recording symptoms in healthy people), the principle that a substance causing a symptom pattern is used to treat a similar one, and potentization, which means serial dilution with vigorous shaking. Because the six meta-analyses, published between 1997 and 2017, used different methods and reached different conclusions, the authors reviewed them together under a published protocol and judged their risk of bias and the certainty of the overall evidence. The authors searched eight databases plus a private collection and contacted experts, identifying 16 reports covering six meta-analyses, with searches running to December 2020 and an update to April 2023. Two reviewers screened and extracted data independently. The meta-analyses together covered 310 trials or trial comparisons (182 distinct trials published between 1943 and 2014), most often on respiratory conditions such as allergic rhinitis and influenza. Each meta-analysis was rated for risk of bias with ROBIS, and the combined evidence was graded with a GRADE-based approach. All five meta-analyses with reported pooled estimates found a significant effect favouring homeopathy across all trials, for example an odds ratio of 2.45 (95% CI 2.05-2.93) across 89 trials in the 1997 analysis and a standardized mean difference of 0.33 (0.21-0.44) across 54 non-individualised trials in 2017. When only high-quality trials were kept, three of four analyses stayed significant, but the 2017 analysis did not (SMD 0.18, -0.09 to 0.46, 3 trials). Only 6-29% of trials met each analysis's high-quality criteria. Three analyses showed significant heterogeneity and funnel plot asymmetry, a pattern that can signal publication or small-study bias; individualised homeopathy showed neither. Two analyses that combined quality and trial-size restrictions found no significant difference from placebo. Three meta-analyses were rated low and three high risk of bias. The authors rated evidence certainty as high for individualised homeopathy and moderate for the other categories. The authors conclude that the data do not support the hypothesis of no difference from placebo and that placebo-controlled evidence gives no justification for regulatory action against homeopathy. The review only covers trials already in six meta-analyses, so trials after 2011 (individualised) and 2014 (non-individualised) are missing. It also reports no indication-specific results, and its use of GRADE tested whether any effect exists rather than how large it is. The authors recommend updating the individualised homeopathy analysis and studying specific non-individualised treatments for specific conditions.

Key findings

• This systematic review (a structured review that gathers and appraises earlier research) examined six meta-analyses (statistical combinations of trial results) covering 310 trials or trial comparisons of homeopathy versus placebo for any condition.
• All five meta-analyses with pooled estimates found a significant overall effect favouring homeopathy, for example an odds ratio of 2.45 (95% CI 2.05-2.93) across 89 trials and a standardized mean difference of 0.33 across 54 trials.
• In analyses limited to high-quality trials, three of four stayed significant, but one did not (standardized mean difference 0.18, 95% CI -0.09 to 0.46), and only 6-29% of trials met each analysis's high-quality criteria.
• Three meta-analyses showed significant heterogeneity and funnel plot asymmetry (a pattern that can signal publication or small-study bias), and combining quality with larger trial size removed significance in both analyses that tested it.
• The authors rate the certainty of evidence as high for individualised homeopathy and moderate for the other categories, but the review only covers trials inside six meta-analyses, reports no results by condition, and rates three of the six meta-analyses at high risk of bias.

Common questions