Skip navigation

Acupuncture for Recurrent Headache

Clinical bottom line: There is no evidence from high quality trials that acupuncture is effective for the treatment of migraine and other forms of headache. The trials showing a significant benefit of acupuncture were of dubious methodological quality. Overall, the trials were of poor methodological quality.

Recurrent headaches are a major source of morbidity and represent a significant economic burden. Acupuncture is widely used for the treatment of headache.

Systematic review

Melchart, D.; Linde, K., and Fischer, P. et al. Acupuncture for recurrent headaches: a systematic review of randomized controlled trials. Cephalalgia. 1999; 19(9):779-86.

Date review completed: 1998

Number of trials included: 22

Number of patients: 1042 in active and control groups

Control group: placebo and active controls

Main outcomes: clinical outcomes related to headache (pain intensity, frequency, global assessment etc.)

Inclusion criteria were randomised or quasi-randomised trials; migraine, tension or cluster headache; acupuncture needles inserted at pain, trigger or traditional points with manual, electo- or laser stimulation; treatment phase greater than four weeks; published and unpublished trials.

Reviewers conducted comprehensive searches including the main databases and references of retrieved reports. Methodological quality of trials were assessed using the Oxford rating scale, maximum score is five (Jadad et al., 1996). A second methodological quality scoring method was also used, the Internal Validity Scale, maximum score six. Additional information was sought from trial authors but only yielded further data on one trial. Reviewers provide a descriptive summary of all included reports and a vote count of clinical outcome. A quantitative analysis on a sub-set of trials (sham-controlled studies) was performed by reviewers even though trials were heterogeneous with respect to quality, type of acupuncture given, outcome measures etc. A rate ratio (proportion of responders in acupuncture group/proportion of responders in sham-control group) was calculated with 95% confidence intervals using random effects.

Findings

Acupuncture treatment varied considerably across trials. Treatment periods ranged from two to 17 weeks. Methodological quality of the trials was variable, scores ranged from 1 to 5, (median=1, max=5) using the Oxford scale, and 1 to 5 (median=2.5, max=6). Individual scores for each report was not given by the reviewers, but based on the range of scores above, many trials would have had inadequacies in randomisation and blinding leading to over-estimation of treatment effects. Trials also had small group sizes (<30 patients), inadequately defined outcomes and diagnostic criteria, and inadequate reporting of results and statistical testing.

Placebo-controlled trials

There were fourteen sham-controlled trials. Based on a vote-counting exercise of trials comparing acupuncture to sham-acupuncture, six trials reported a statistically significant result, two trials reported no statistical significant difference between groups, and three reported a positive trend in favour of acupuncture that were not supported by statistical tests. Individual quality scores of these trials were not presented in the review. However, based on a description of the trials by reviewers, it would seem that the trials that reported positive effects had methodological flaws (not adequately randomised and not adequately blinded).

A sub-group analysis of ten of the 14 trials with dichotomous outcomes showed a statistically significant benefit of acupuncture over placebo. We disagree with this analysis as it is not valid to pool data for meta-analysis when trials are so heterogeneous.

Active controlled trials

Seven trials compared acupuncture with active control groups, five with standard drug treatment and two with physiotherapy. Based on a vote-counting exercise, two trials reported a statistically significant result, one reported no statistical difference between acupuncture and placebo, and four trials were negative but not supported by statistical tests. The reviewers described the two positive trials as being of doubtful validity. There was no description given about the sensitivity of these trials.

Adverse effects not reported in the review

Comment

Reviewers were not helpful with this review. First they did not provide the quality scores of the included trials after going to the trouble of measuring it using two quality scales. Had they done, it would have been interesting to determine if the trials reporting positive results were those with low quality scores. The reviewers imply this in the text and tables. Secondly, a quantitative analysis should only be performed when data are similar. In this case they were not. Trials are heterogeneous with respect to type of acupuncture given, disease definition, outcome measures, and study architecture. Inadequately randomised and blinded trials have been lumped together with those that were of higher quality which has led to an erroneous conclusion about the effectiveness of acupuncture by the reviewers.

The later Cochrane review informed us that the median quality score was 1.5 (out of a range of 1.5). This is important because there is good evidence that studies with scores of 2/5 or below are likely to be biased and show greater effects of treatment than better conducted trials. The later Cocharne review did not pool data, but commented that studies with a higher quality score the results seemed "less positive for migraine".

Given the very small size of most trials, and their inadequate quality, the best conclusion is that there is no evidence for any effect of acupuncture in headache, unless and until someone does a large high quality trial that proves the opposite.

Further reading

This review is substantially the same as a Cochrane review edited in 2000. The Cochrane review will, of course, be updated:

D Melchart, K Linde, P Fisher, B Berman, A White, A Vickers, G Allias. Acupuncture for idiopathic headache. The Cochrane Library. Update Softeware.

Related topics