The review was conducted according to PRISMA 2020 reporting standards and prospectively registered in PROSPERO under registration number CRD420261400459, with the registration dated 20 May 2026. The research team searched six bibliographic databases along with the ClinicalTrials.gov registry through 20 August 2026, applying no language restriction, and supplemented the electronic searches with backward citation searching of the included studies. Two reviewers independently screened records and full texts, extracted data, assessed risk of bias at the result level, and rated the certainty of the evidence, with disagreements resolved through discussion and, where necessary, adjudication by a third reviewer. This dual-reviewer architecture matters because systematic reviews live or die by the reproducibility of their selection and appraisal decisions; single-reviewer pipelines are a well-recognised source of error in the evidence-synthesis literature.

The scale of the underlying evidence base is the first striking finding. The searches identified 3,197 database records and 68 ClinicalTrials.gov entries, yet only eight reports describing six independent randomised trials met the inclusion criteria, and just five of those trials contributed quantitative data to the syntheses. For a condition as prevalent and disabling as endometriosis, that attrition rate from thousands of records to a handful of usable trials is itself a message about how little high-quality experimental work has been done. The trials that did qualify were small and heterogeneous, which constrained every downstream analysis the team attempted.

The statistical machinery the authors deployed deserves attention because it was chosen specifically to guard against the overconfidence that small samples can breed. Rather than relying on a single pooling approach, the team ran random-effects meta-analyses using restricted maximum likelihood estimation with Hartung-Knapp inference, a method that widens confidence intervals when the number of studies is small and between-study variance is uncertain. They also ran alternative small-k models as sensitivity analyses and computed prediction intervals, which estimate the range of effects a new trial in a comparable setting might plausibly show. The GRADE framework was applied using review-stage clinical decision thresholds to grade certainty for each outcome, and the result was unambiguous: certainty was rated as very low for all key outcomes.

The headline quantitative result concerns overall endometriosis-related pain, pooled across four trials with a combined 271 participants. The standardised mean difference was -0.36, with a 95 percent confidence interval spanning -0.95 to 0.24, meaning the interval crosses the null and cannot exclude either a substantial benefit or a modest harm. Heterogeneity was moderate, with an I-squared of 57.2 percent, and the prediction interval stretched from -1.74 to 1.03, an extraordinarily wide band that captures effects ranging from large improvement to no benefit at all. In plain terms, even the point estimate, which modestly favoured exercise, is compatible with almost any clinical reality a future trial might reveal.

The multidimensional symptom picture fared no better. Psychological distress, synthesised from three trials with 243 participants, yielded a standardised mean difference of -0.29 with a confidence interval of -1.43 to 0.85, a range so wide it is essentially uninformative. Sexual function, pooled from three trials totalling 144 participants, produced a mean difference of -2.23 with a confidence interval of -5.96 to 1.51 on the relevant instrument scale. Perhaps most tellingly, the authors report that statistical significance differed across random-effects specifications for five of their syntheses, meaning the conclusions were method-dependent: whether exercise appeared statistically significant depended on which defensible statistical model the analysts chose. Findings that flip with the choice of estimator are, by convention, treated as unstable rather than as evidence of a real effect.

Safety data added another layer of concern. Harms reporting across the included trials was inconsistent, making it impossible to characterise the adverse-effect profile of structured exercise in this population with any confidence. This is a recurring blind spot in exercise-intervention research generally, where physical activity is assumed to be benign and trialists often neglect systematic adverse-event collection. For women with endometriosis, whose symptoms can include deep dyspareunia and activity-provoked pelvic pain, the assumption that exercise is universally tolerable is not self-evident, and the absence of standardised harms data leaves clinicians without the information needed to counsel patients responsibly.

What does this mean for practice? The authors are careful to distinguish between the absence of evidence and evidence of absence. The point estimates across outcomes generally favoured exercise, and nothing in the analysis suggests structured training is harmful. Their conclusion is deliberately nuanced: decisions to use structured exercise should be individualised according to patient preference, general physical-activity guidance, comorbidity, and tolerance, rather than justified by an assumed endometriosis-specific benefit. In other words, a woman with endometriosis who enjoys swimming or yoga has every reason to continue for the same reasons exercise benefits anyone else, including cardiovascular and mental health, but she should not be told that the published trial evidence proves it will shrink her pain scores or her symptom burden.

The path forward, the review makes clear, runs through better trials. The evidence base needs larger, adequately powered randomised studies with standardised exercise protocols, validated multidimensional outcome measures such as the Endometriosis Health Profile-30 and the Female Sexual Function Index, pre-registered primary outcomes, and systematic harms reporting. Until such trials exist, the review stands as a corrective to the enthusiasm that has outpaced the data, a reminder that in evidence-based medicine, plausibility is not proof, and that even the most intuitively sensible interventions must earn their clinical recommendations the hard way, one well-designed trial at a time.

Subject of Research: Effects of structured exercise training on pain and symptom burden in women with endometriosis

Article Title: Effects of structured exercise training on pain and multidimensional symptom burden in women with endometriosis: an updated systematic review and meta-analysis

Article References: Sun, M., Zhu, Y., Wang, L., & Liang, W. (2026). Effects of structured exercise training on pain and multidimensional symptom burden in women with endometriosis: an updated systematic review and meta-analysis. BMC Psychology. https://doi.org/10.1186/s40359-026-05724-7

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05724-7

Keywords: endometriosis, structured exercise, systematic review, meta-analysis, chronic pelvic pain, psychological distress, sexual function, quality of life, GRADE, randomised controlled trials, PROSPERO, BMC Psychology