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

This study represents one of the largest analyses ever performed on response patterns to acupuncture in patients with chronic pain. The Acupuncture Trialists' Collaboration analyzed individual data from 20,827 patients drawn from 39 high-quality randomized clinical trials, all published through December 2015. The primary objective was to investigate whether a subgroup of patients can be classified as "exceptional responders" to acupuncture.

The motivation for this study arose from clinical observations that some patients appear to have dramatically better responses to acupuncture than others, and from comments in the literature suggesting the existence of "deep responders." This question has important implications for the development of stratified care, in which patients would be directed to treatments based on characteristics that predict better response.

The methodology was innovative in focusing on the statistical distribution of outcomes rather than simply examining group means. The researchers analyzed the skewness in the distribution of pain change scores between acupuncture and control groups. The hypothesis was that, if exceptional responders existed, the distribution in the acupuncture group would be more right-skewed, indicating a "tail" of patients with improvements far above the mean.

The included studies covered various chronic pain conditions: nonspecific low back pain, neck pain, shoulder pain, chronic headache, migraine, and osteoarthritis. All trials had adequately determined allocation concealment, ensuring methodological quality. Twenty-five studies included sham acupuncture controls (7,097 patients) and 25 studies included no-acupuncture controls (16,041 patients), with some studies having a three-arm design.

The results were surprising and contrary to the initial hypothesis. Rather than finding more exceptional responders in the acupuncture group, the researchers discovered that the control groups were slightly more right-skewed than the acupuncture groups. The skewness difference was 0.124 for no-acupuncture-controlled studies (p = 0.047) and 0.141 for sham-controlled studies (p = 0.029). Importantly, these differences were very small and of questionable clinical significance.

In a prespecified sensitivity analysis, when three studies with known extreme results (studies from the Vas group) were excluded, the skewness difference between acupuncture and sham controls was no longer statistically significant (p = 0.2). This suggests that the initial findings may have been influenced by these outlier studies.

Individual-study analysis revealed that only a few studies showed significant differences in skewness between groups. Interestingly, where there were significant differences, it was often the control groups that showed greater right-skewness, especially in studies in which the control group received active interventions such as supervised exercise or prophylactic medications.

These findings have important implications for understanding the mechanisms of response to acupuncture and for the development of personalized care strategies. The approximately normal distribution of outcomes suggests that the benefits of acupuncture follow an additive model rather than a distinct-subgroup model. This means that patient characteristics likely contribute continuously and cumulatively to treatment response, rather than there being discrete categories of "good" and "poor" responders.

The study also highlights the importance of examining not only group means but also distribution patterns of outcomes in clinical research. This methodological approach can be applied to other interventions for chronic pain, helping clarify whether different treatments truly benefit specific subgroups of patients or whether benefits are distributed more uniformly.

Limitations include the heterogeneity of acupuncture protocols across studies, different chronic pain conditions grouped in the analysis, and the possibility that biological characteristics not measured in the clinical trials could still identify responder subgroups. The authors suggest that future research should focus on additive models involving continuous variables, potentially including biomarkers such as β-endorphin and met-enkephalin.

This work represents an important milestone in acupuncture research by challenging widely accepted assumptions about response patterns and providing a solid methodological basis for future investigations into the personalization of treatments for chronic pain.