GLP-1 RAs, GIP/GLP-1 RAs, and TZDs demonstrate favorable glucose and weight control in prediabetes

Methodology

This was a systematic review and Bayesian network meta-analysis registered in PROSPERO (CRD42025636991) and conducted according to PRISMA-NMA guidelines. The investigators systematically searched PubMed, Web of Science, Embase, Cochrane Library, and ClinicalTrials.gov from November 1996 through March 2025 using MeSH terms and keywords including 'prediabetes,' 'reverse,' and 'delay.' Eligible studies were randomized controlled trials enrolling adults aged 18-75 years with prediabetes (HbA1c 5.7-6.4%, FPG 5.6-6.9 mmol/L, or 2h-PPG 7.8-11.0 mmol/L) and comparing at least one of nine classes of anti-prediabetic drugs against placebo or another intervention, with minimum 12-week follow-up.

From 6,663 initially identified records, 55 RCTs involving 16,610 participants were ultimately included after removing duplicates and applying inclusion/exclusion criteria through title/abstract and full-text screening by two independent reviewers. Risk of bias was assessed using the Cochrane Risk of Bias Tool version 2.0 across five domains. Evidence quality was graded using the CINeMA framework. Eight efficacy outcomes were analyzed (weight loss, BMI, fasting plasma glucose, HbA1c, total cholesterol, triglycerides, HDL, LDL) along with four safety outcomes (adverse events, serious adverse events, gastrointestinal disorders, renal/urinary disorders).

Statistical analysis employed Bayesian network meta-analysis using R software (version 4.4.1) with the BUGSnet package (version 1.1.2). Both random-effects and fixed-effects models were constructed with 5,000 adaptation iterations, 3 burn-in iterations, and 20,000 simulation iterations. Mean differences were calculated for continuous outcomes and odds ratios for dichotomous outcomes. The surface under the cumulative ranking curve (SUCRA) was used to rank interventions. Global consistency was assessed using deviance information criterion (DIC) differences between consistency and inconsistency models (threshold >5), while local consistency was evaluated using node-splitting methods. Heterogeneity was quantified using I-squared statistics (≥50% or p<0.05 indicating high heterogeneity). Sensitivity analyses excluded high-risk studies and those with duration exceeding 56 weeks. Publication bias was assessed using funnel plots.

Interventions

The network meta-analysis evaluated 37 specific intervention-dose combinations across nine major drug classes compared to placebo and lifestyle intervention. The pharmacological interventions included: (1) Metformin at various doses (500 mg, 850 mg, 1500 mg, 2000 mg), (2) GLP-1 receptor agonists including semaglutide 2.4 mg subcutaneous weekly, liraglutide at 1.2 mg, 1.8 mg, and 3.0 mg, and exenatide 2 mg weekly, (3) GIP/GLP-1 dual agonist tirzepatide at 5 mg, 10 mg, and 15 mg, (4) SGLT-2 inhibitors including dapagliflozin 10 mg alone and in combination with exenatide 2 mg, (5) DPP-4 inhibitors including sitagliptin 100 mg, vildagliptin 50 mg, saxagliptin 5 mg, linagliptin 5 mg, and cofrogliptin 10 mg, with some used in combination with metformin, (6) Thiazolidinediones including pioglitazone 30 mg and 45 mg, rosiglitazone 4 mg, and the withdrawn agent troglitazone, (7) Alpha-glucosidase inhibitor acarbose 100 mg, (8) Lipase inhibitor orlistat 120 mg, and (9) Vitamin D supplementation at various doses (40 µg, 75 µg, 80 µg, 500 µg). Lifestyle intervention was also included as a non-pharmacological comparator.

The comparison groups varied across the 55 included trials, with most employing placebo controls rather than active comparators. Study durations ranged from 12 to 120 weeks (median follow-up approximately 24-52 weeks depending on the specific trial). Participants had mean baseline characteristics of: age 50.6 years, 53% female, body weight 89.1 kg, BMI 30.8 kg/m², fasting plasma glucose 106.4 mg/dL, and HbA1c 5.88%. The interventions were administered according to standard dosing schedules for each medication class, with most studies employing once-daily or once-weekly administration depending on the specific agent.

Key Findings

Among the 37 interventions analyzed, GLP-1 receptor agonists and GIP/GLP-1 dual agonists demonstrated the most favorable combined efficacy for weight reduction and glycemic control in adults with prediabetes. Semaglutide 2.4 mg subcutaneous weekly achieved the greatest weight loss (mean difference -13.59 kg, 95% CI -17.30 to -9.91 compared to placebo; SUCRA 97.47%) and showed strong HbA1c reduction (MD -0.39%, 95% CI -0.55 to -0.25). Tirzepatide 15 mg demonstrated optimal BMI reduction (MD -4.50 kg/m², 95% CI -8.45 to -0.51; SUCRA 97.47%), excellent fasting glucose control (MD -9.58 mg/dL, 95% CI -12.00 to -7.15; SUCRA 90.60%), and superior lipid-lowering effects with total cholesterol reduction (MD -17.34 mg/dL, 95% CI -27.76 to -7.10). Liraglutide 1.2 mg showed the strongest HbA1c reduction among all interventions (MD -0.94%, 95% CI -1.34 to -0.54; SUCRA 97.41%). Thiazolidinediones also demonstrated potent glucose-lowering efficacy, with pioglitazone 30 mg achieving the greatest fasting glucose reduction (MD -26.42 mg/dL, 95% CI -36.99 to -16.73; SUCRA 99.98%) and substantial triglyceride lowering (MD -6.16 mg/dL, 95% CI -11.73 to -1.03). However, thiazolidinediones were associated with weight gain concerns. Metformin 500 mg showed significant HbA1c reduction (MD -0.75%, 95% CI -1.34 to -0.14) but demonstrated less pronounced effects on weight and lipids compared to newer agents. For safety outcomes, no intervention showed significantly higher incidence of overall adverse events compared to placebo, though sitagliptin 100 mg demonstrated elevated serious adverse events (OR 1.70×10⁷, 95% CI 17.63 to 2.53×10¹⁸). Gastrointestinal adverse events were more common with GLP-1 RAs and GIP/GLP-1 RAs compared to placebo, consistent with the known side effect profile of these medication classes.

Comparison with other Studies

The findings of this network meta-analysis align well with previous systematic reviews and meta-analyses of pharmacological interventions for prediabetes. The landmark Diabetes Prevention Program (DPP) demonstrated that metformin reduced progression to type 2 diabetes by 31% compared to placebo, though lifestyle intervention was more effective (58% reduction). The current network meta-analysis confirms metformin's moderate efficacy in glycemic control but highlights that newer agents, particularly GLP-1 RAs and GIP/GLP-1 RAs, demonstrate superior combined benefits for weight and glucose management. A 2019 Cochrane review by Madsen et al. on metformin for diabetes prevention found similar HbA1c reductions to those reported in this analysis, validating the consistency of metformin's effects across multiple evidence syntheses.

The superior weight loss efficacy of semaglutide 2.4 mg observed in this analysis is consistent with findings from the STEP (Semaglutide Treatment Effect in People with obesity) trial program, which demonstrated substantial weight reductions in populations with obesity and prediabetes. The current analysis extends these findings by providing comparative effectiveness evidence across multiple drug classes through network meta-analysis methodology. Similarly, the strong performance of tirzepatide aligns with recent phase 3 trials showing dual GIP/GLP-1 agonism produces greater metabolic benefits than GLP-1 agonism alone, though the current analysis represents the first comprehensive network comparison in the prediabetes population specifically.

Regarding thiazolidinediones, previous systematic reviews including a 2020 Cochrane review by Ipsen et al. on pioglitazone for diabetes prevention demonstrated similar glucose-lowering efficacy but raised concerns about weight gain and other adverse effects including heart failure and fracture risk, which limited enthusiasm for these agents despite their metabolic benefits. The current analysis confirms the potent glucose and lipid effects of thiazolidinediones while contextualizing them within a broader treatment landscape that includes newer agents with more favorable weight profiles. For DPP-4 inhibitors, the modest effects observed in this analysis are consistent with a 2017 Cochrane review by Hemmingsen et al., which found limited evidence for DPP-4 inhibitors in diabetes prevention compared to metformin or lifestyle intervention. The heterogeneity noted in some outcomes, particularly fasting plasma glucose, is consistent with previous network meta-analyses in diabetes and likely reflects differences in study populations, intervention durations, and baseline characteristics across the included trials.

Journal Reference

Wu Y, Wang Z, Tuersun A, et al. Efficacy and safety of anti-prediabetic drugs in patients with prediabetes: a Bayesian network meta-analysis. BMC Med. 2026;24:174. doi:10.1186/s12916-026-04705-2