Effect of yoga on glycemic control in adults with type 2 diabetes mellitus: a Bayesian three-level meta-analysis of randomized controlled trials.

Background: Type 2 diabetes mellitus (T2DM) imposes a growing global burden, and glycemic control remains central to the prevention of long-term microvascular and cardiovascular complications. Yoga has been proposed as an adjunctive mind-body intervention, but existing meta-analytic evidence has been constrained by pooling approaches that ignore within-study dependency, dichotomized dose comparisons, and limited reporting of prediction intervals. Methods: PubMed, Embase, Web of Science, and the Cochrane Central Register of Controlled Trials were systematically searched through May 2026 for randomized controlled trials (RCTs) comparing structured yoga interventions with usual care or waiting-list control in adults with T2DM. The primary outcomes were glycated hemoglobin (HbA1c) and fasting plasma glucose (FBG). A Bayesian three-level random-effects model was fitted to accommodate within-study dependency. Categorical subgroup analyses, spline-based meta-regression, sensitivity analyses, publication bias assessments, and GRADE certainty ratings were also performed. Results: Twenty-eight RCTs comprising 2, 241 adults were included. Yoga reduced HbA1c by 0.64% (95% credible interval [CrI] -0.87 to -0.41) and FBG by 1.36 mmol/L (95% CrI -1.75 to -1.00), with direction probabilities of 100% and extreme Bayes factor evidence against the null hypothesis. The 95% prediction intervals remained below zero for both outcomes, although the upper bound for FBG approached the null. No clear evidence of effect modification was found by country, supervision modality, or the WHO threshold of 600 MET-min per week, although these subgroup comparisons were underpowered and evidence from outside India was limited. GRADE certainty was low for both HbA1c and FBG. Conclusions: In adults with T2DM, yoga was associated with clinically meaningful reductions in HbA1c and fasting glucose, with the largest benefits in patients with poorer baseline control, although the certainty of this evidence was low. These findings support consideration of yoga as a feasible adjunct to standard diabetes care rather than a fixed prescription. Systematic review registration: https://www.crd.york.ac.uk/PROSPERO/view/CRD420261391309, identifier CRD420261391309.