Abstract
ABSTRACT
Background
The relative benefits of carbohydrate‐restricted diets (CRDs) combined with exercise in adults with type 2 diabetes remain uncertain. We compared regimens and explored incremental benefits, dose–response associations and potential effect modifiers.
Methods
Four databases were searched from inception to 31 August 2026 for English‐language randomized controlled trials (RCTs). A Bayesian multilevel model synthesized arm‐level changes in glycated haemoglobin (HbA1c) across eligible post‐baseline assessments. An auxiliary frequentist network meta‐analysis evaluated network structure and inconsistency; Confidence in Network Meta‐Analysis (CINeMA) assessed 20 prespecified comparisons. Dose–response and moderator analyses were exploratory. Risk of bias was assessed using the revised Cochrane tool for randomized trials.
Results
Twenty‐four RCTs involving 2038 participants were included. Compared with pooled active/passive controls, aerobic training plus a moderately low‐carbohydrate diet was associated with lower HbA1c (posterior median mean difference, −0.53 percentage points; 95% credible interval, −0.76 to −0.28). Estimates for other combined regimens were less precise. The auxiliary network had two disconnected components, and confidence was very low for all 20 assessed comparisons. Under the specified priors, the Bayes factor favoured the model without treatment‐specific effects (BF10 = 0.034). Higher prescribed exercise doses were associated with larger reductions versus pooled controls, but an incremental benefit over exercise or CRD alone was not established. Matched carbohydrate‐dose contrasts were imprecise.
Conclusions
Limitations of the arm‐level model, sparse evidence and very low confidence in the assessed auxiliary‐network comparisons leave comparative benefits uncertain. The findings do not establish a preferred regimen, an incremental benefit over either component alone, or a specific dietary or exercise prescription.