Does stimulation at Spleen-6 reduce pain in primary dysmenorrhoea?
The review reports a pain-relief signal for acupressure, but does not establish benefit from acupuncture overall. Across nine acupressure reports, the pooled standardised difference was −0.52 (95% confidence interval −0.71 to −0.33), favouring acupressure over varied controls. Across ten acupuncture reports, the estimate was −0.15 (−0.30 to 0.01), leaving benefit uncertain rather than proving no effect. These are standardised effects, not changes on a familiar pain scale. Treatment schedules and assessment times varied, and the review does not clearly identify a common pooled time point. Consequently, it does not establish how much relief patients would notice or how long it would last.
Does this review establish that acupressure is more effective than acupuncture?
No. The superiority claim comes from comparing separate groups of studies, not a pooled direct comparison in which participants were allocated to acupressure or acupuncture. Differences in participants, treatment schedules and controls could explain the different estimates. Controls included rest, light touch, other acupoints and alternative needling techniques, which answer different clinical questions and are not necessarily inactive. The reported −0.52 estimate measures acupressure against its controls; it is not the advantage of acupressure over acupuncture. Likewise, researcher-administered acupressure showed a favourable estimate while self-acupressure remained inconclusive, but that indirect contrast does not establish that professional administration is superior.
How dependable are the pooled estimates and their apparent precision?
The estimates require substantial caution. Results differed markedly between studies: the overall heterogeneity statistic was 90%, indicating extensive variation beyond sampling uncertainty. Nevertheless, the authors used a fixed-effect analysis, which does not incorporate between-study variation into its uncertainty estimate; the narrow confidence interval may therefore give an overly reassuring impression. The abstract also names a method usually associated with random-effects analysis, leaving the statistical description inconsistent. Possible overlap between two Hu reports with identical participant counts needs checking against the original studies. These issues limit reliance on the pooled conclusions, while the absence of reported adverse events does not establish safety.
Assessed 21 Sept 2026.
