A network meta-analysis of 378 randomised trials covering 113,122 adults with prediabetes has ranked what actually prevents type 2 diabetes, and incretin drugs finished first: dual GLP-1/GIP receptor agonists cut the risk of developing diabetes by around 90% against control, and single GLP-1 receptor agonists by around 74%. Metformin plus intensive lifestyle change came third. On blood sugar markers rather than on diabetes itself, a supplement topped the table instead, and that result is worth reading slowly rather than quickly.
What the numbers say
The review, published in Clinical and Translational Medicine and registered in advance on PROSPERO, searched MEDLINE, Embase and the Cochrane Library from inception to 19 December 2025 for randomised trials of any intervention aimed at dysglycaemia in adults with prediabetes. For diabetes incidence, dual GLP-1/GIP receptor agonists returned a relative risk of 0.10 (95% CI 0.05 to 0.21) against control. A relative risk of 0.10 means one-tenth the risk, a 90% reduction, and the interval says the data are consistent with anything between a 79% and a 95% reduction.
Single GLP-1 receptor agonists came in at 0.26 (0.19 to 0.40), a 74% reduction with a plausible range of 60% to 81%. Metformin combined with intensive lifestyle intervention was 0.27 (0.12 to 0.64), nominally third, but that interval runs from an 88% reduction down to a 36% one and overlaps the GLP-1 range along almost its whole length. Second and third place are therefore an artefact of ranking rather than a demonstrated difference between the two. Only the dual agonist separates from the field with an interval that clears the others outright.
The HbA1c table is where this gets interesting for anyone buying supplements. At one year, four interventions beat a 0.3% reduction: curcumin at −0.49% (CI −0.70 to −0.28), dual GLP-1/GIP agonists at −0.46%, single GLP-1 agonists at −0.41% and physical activity at −0.35%. Sulphonylureas and physical activity led on one-year fasting plasma glucose, while zinc supplementation and sulphonylureas produced the largest reductions in two-hour glucose after an oral tolerance test.
What is not yet known
Curcumin ranking above the incretins on HbA1c did not carry through to the outcome that matters. It does not appear among the leaders for actually preventing diabetes, and the two tables come apart precisely where a shopper would want them to agree. The honest reading is that a 0.49% HbA1c difference assembled from small supplement trials is a weaker piece of evidence than a diabetes-incidence figure assembled from large drug trials, even though it sits higher in one of the rankings. The same caution applies to zinc on two-hour glucose. Our page on GLP-1 supplements is unchanged by this analysis: nothing in it shows a supplement preventing diabetes.
Network meta-analyses also compare interventions that were never tested head to head, by routing through shared comparators, and across 378 trials the durations, populations and definitions of prediabetes vary a great deal. The abstract does not report a certainty-of-evidence grading, and rankings of this type are known to be unstable where the underlying network is sparse, which it will be for the newer incretin comparisons with the fewest trials behind them. None of this establishes that a person with prediabetes should be prescribed an incretin drug. That decision involves cost, side effects and how long someone would have to stay on it, and this analysis addresses none of the three.
Sources: Maria B, Orsolya E, Iulian PR, et al. “Evidence-based ranking of diabetes prevention in prediabetes: A comprehensive network meta-analysis of 378 randomised controlled trials.” Clinical and Translational Medicine 2026;16(9):e70829, doi:10.1002/ctm2.70829. Protocol registered on PROSPERO as CRD42023473875.
