Acupuncture trials and quality [Oct 2000; 80-3]

Acupuncture trials and quality
Search Outcomes Results Comment Acupuncture is commonly believed to be effective for the treatment of chronic pain, despite growing evidence from systematic reviews ( Bandolier 60 ) that it is not. Another systematic review across all of chronic pain draws attention to the fact that the number of clinical trials of high quality that showed acupuncture to be effective is risibly small. There are only three, at best.

Search

The search strategy was heroic, using standard databases and at least four specialist databases for complementary therapies as well as dissertation abstracts and conference proceedings. For inclusion trials had to be randomised, have a comparison group, have studied patients with pain longer than three months, used needles, were in English, and measured pain relief. Authors were contacted for details if necessary.

Outcomes

Trials were defined as positive if acupuncture was found to be significantly better than control, neutral if there was no significant difference between acupuncture and control, and negative when acupuncture was significantly worse than control. A P value of 0.05 was used to define statistical significance.

Results

There were 50 trials with 2,394 patients. Thirty-four trials (68%) had a quality score of 2 or less on a five point scale. Controls included waiting lists, inert controls, sham acupuncture and active controls, usually transcutaneous electrical nerve stimulation.

The main results are shown in the Table. Most high quality studies either showed no benefit or that acupuncture was worse than control. Forty to fifty percent of the trials or patients showed acupuncture to be better than control. Studies of low methodological quality showed significantly higher treatment effect than those of high quality.

Table: Effect of quality of trial reporting on whether trials of acupuncture in chronic pain are better, the same, or worse than control

	Number of trials	Percent	Number of patients	Percent
Quality score 3 or more
Acupuncture better than control	3	19	111	12
Acupuncture same as control	12	75	715	79
Acupuncture worse than control	1	6	77	9
	16	100	903	100
Quality score 2 or less
Acupuncture better than control	16	47	643	43
Acupuncture same as control	16	47	736	49
Acupuncture worse than control	2	6	112	8
	34	100	1491	100

Comment

Without labouring the point about poor quality studies over-estimating effects of treatment, or that evidence for acupuncture is thin on the ground, this study demonstrates both with some clarity. It is worth commenting that for chronic pain where acupuncture is much used, the absence of significant effect is all too apparent. Only three studies showed a benefit, and they contained only 12% of the total patients studied in high quality trials. As a balance, one trial with 9% of the total patients studied in high quality trials showed less effect than an active treatment. The remainder of the high quality studies showed no difference at all between acupuncture and control.

What about bias? Well, the search was only for English language papers, but most non-English papers would be from countries where only positive results are published, so that this strategy may have avoided bias. Many of the studies were small. Of the 19 positive studies, 14 enrolled fewer than 50 patients, and the smallest number was 12. Overall, positive studies were smaller than neutral studies, which were smaller than negative studies. We might conclude that there is some residual bias in this, which would result in an even more negative conclusion.

There are two bottom lines. The first is to emphasise again the importance of using quality information for decision-making. Use poor quality information is likely to result in poor quality decision-making. The second bottom line is that the use of acupuncture for chronic pain is unsupported by any evidence of quality. Consumers and providers should beware.

Reference:

J Ezzo et al. Is acupuncture effective for the treatment of chronic pain? A systematic review. Pain 2000 86: 217-225.

previous or next story in this issue