Identifying patients with chronic pain who respond to acupuncture: results from an individual patient data meta-analysis

Abstract

Background In a recent individual patient data meta-analysis, acupuncture was found to be superior to sham and non-sham controls in patients with chronic pain. It has been suggested that a subgroup of patients has an exceptional response to acupuncture. We hypothesized the presence of exceptional acupuncture responders would lead to a different distribution of pain scores in acupuncture versus control groups, with the former being skewed to the right. Methods This individual patient data meta-analysis included 39 high-quality randomized trials of acupuncture for chronic headache, migraine, osteoarthritis, low back pain, neck pain and shoulder pain published before December 2015 (n = 20,827). In all, 25 involved sham acupuncture controls (n = 7097) and 25 non-acupuncture controls (n = 16,041). We analyzed the distribution of change scores and calculated the difference in the skewness statistic—which assesses asymmetry in the data distribution—between acupuncture and either sham or non-acupuncture control groups. We then entered the difference in skewness along with standard error into a meta-analysis. Findings Control groups were more right-skewed than acupuncture groups, although this difference was very small. The difference in skew was 0.124 for non-acupuncture-controlled trials (p = 0.047) and 0.141 for sham-controlled trials (p = 0.029). In a pre-specified sensitivity analysis excluding three trials with outlying results known a priori, the difference in skew between acupuncture and sham was no longer statistically significant (p = 0.2). Conclusion We did not find evidence to support the notion that there are exceptional acupuncture responders. The challenge remains to identify features of chronic pain patients that can be used to distinguish those that have a good response to acupuncture treatment.