Researchers in Germany and the United States analysed the results of 89 controlled trials of homoeopathy published in 13 countries since 1943. The homoeopathic remedies were used to treat everything from warts to strokes.
The findings, published in the Lancet medical journal, present a challenge to orthodox science.
Homoeopathy is based on the belief that it is possible to cure diseases by giving patients substances which cause similar symptoms - such as fever - in healthy individuals. These remedies are said to retain their potency even when repeatedly diluted to the point where no molecules of the original substance remain. How the solution "remembers" information from the original substance is unknown.
Orthodox doctors say that the theory of homoeopathy runs counter to what is known about physics and chemistry, and would require the basic laws of science to be rewritten.
Their argument has been that the claimed success of homoeopathic treatments can be ascribed to the placebo effect, a well-known phenomenon, in which people given a pharmacologically inactive substance, such as a sugar pill, report that their condition has improved,just because they think they are getting proper treatment.
The survey is a blow to that argument. It found that homoeopathic remedies were about two-and-a-half times more effective than placebos, although the researchers could find no convincing evidence that any single homoeopathic approach was effective for any single condition. They say "a serious effort to research homoeopathy is clearly warranted, despite its implausibility".
The journal carries two sceptical commentaries, by Professor Michael Langman of the University of Birmingham and Professor Jan Vandenbroucke of the University of Leiden, Netherlands.
Professor Langman questions the quality of the trials - only 26 of the 89 are judged to be of high standard - and argues that the results may be skewed in favour of homoeopathy because those with positive results are more likely to published than those with negative results.
Of 100 trials of homoeopathy conducted, if 20 show a positive effect, of which 15 are published ,and 80 show no effect and five are published, a survey of the published trials will suggest positive trials outnumber negative by three to one. In fact, taking all 100 trials, negative outnumber positive by four to one.