Skip navigation

Homeopathy trials and quality

In Bandolier 45 we reported on a meta-analysis that concluded that homeopathy was effective. There were a whole range of problems:

Eighty-nine trials could be analysed. They broke down like this:

The other problem was that although studies had to be randomised, and controlled, there was no formal sensitivity testing according to trial quality. There are enough examples now that even randomised studies can be of poor quality, and still be biased, so this is important.

In a follow-up paper [1], the authors claimed that there was no difference in outcome according to quality, but a letter [2] shows how important qualit was to the overall estimate of efficacy.

What they did was to analyse the placebo-controlled trials according to a well used quality scoring system (Jadad et al, 1996). The quality score is out of five points, and we know that studies with scores of 2 or less are more likely to give a positive result than those with a score of 3 or more.

Result

The results by quality score are shown in the Table.

Quality score (of 5)

Number of trials

Odds ratio

0

2

6.9 (1.5 to 31.4)

1

15

3.6 (2.1 to 6.1)

2

32

3.2 (2.3 to 4.6)

3

19

1.8 (1.4 to 2.3)

4

11

1.4 (1.0 to 2.0)

5

10

2.0 (1.4 to 2.9)

Comment

As the quality score increases (better studies) the odds ratio (for homeopathy versus placebo) goes down, and goes down to about 1-2. The authors of the letter comment that:

"some (but by no means all) methodologically astute and highly convioncing homeopaths have published results that look convincing but are, in fact, not credible. Viewed in this way, the reanalysis.... can be seen as the ultimate epidemiological proof that homeopathic remedies are, in fact, placebos."

Quod erat demonstrandum, perhaps, because quality is but one issue in clinical trials. The other is validity. Trials have to look at sensible and useful outcomes, or be performed in situations that make clinical and scientific sense. There is still a point that even this sensitivity analysis mixes up all kinds of different trials in all kinds of different clinical situations. Pooling them makes no logical sense. Even Linde and his co-authors [1] admit that many of the more recent high quality studies are quite negative.

References:

1 K Linde et al. Impact of study quality on outcome in placebo controlled trials of homeopathy. Journal of Clinical Epidemiology 1999 52: 631-636.

2 E Ernst & MH Pittler. Re-analysis of previous meta-analysis of clinical trials of homeopathy. Journal of Clinical Epidemiology 2000 53: 1188.