Heart and metabolism effects of oral versus injected GLP-1 drugs in adults with overweight or obesity

This systematic review and NMA followed PRISMA 2020 guidelines and was prospectively registered in PROSPERO (CRD420251238152). PubMed, Embase and the Cochrane Central Register of Controlled Trials (CENTRAL) were searched from January 2014 through November 2025 for RCTs evaluating GLP‐1 receptor agonists for weight management (search strategies in Table). S1

Eligible RCTs enrolled adults with overweight (BMI ≥ 27 kg/m² with comorbidity) or obesity (BMI ≥ 30 kg/m²), included ≥ 200 participants and ≥ 44 weeks of follow‐up, compared a GLP‐1 receptor mono‐agonist with placebo or an active comparator, and reported at least one prespecified outcome. Only RCTs were included to minimize confounding and preserve internal validity and the transitivity assumption for NMA [6, 7].

To reflect the current RCT evidence base and support the transitivity assumption for NMA [8], analyses were restricted to GLP‐1 receptor mono‐agonists that were FDA‐approved or under regulatory review for weight management, reflecting established dosing regimens rather than exact pharmacologic equivalence: subcutaneous semaglutide (1.7, 2.4, 7.2 mg), liraglutide (1.8, 3.0 mg), oral semaglutide (25 mg) and orforglipron (6, 12, 36 mg). The ATTAIN clinical trials evaluated orforglipron using capsule formulations; the FDA‐approved tablet formulation (Foundayo) has bioequivalent strengths of 5.5, 9, and 17.2 mg, respectively [9, 10]. Dual agonists (e.g., tirzepatide), crossover trials and post hoc or secondary analyses were excluded. Analyses used the efficacy estimand to reflect treatment effects under continued intervention exposure before treatment discontinuation or use of prohibited medications [11].

The primary outcome was the CEI, a ranking‐based composite measure (0–1) summarizing treatment performance across seven cardiometabolic endpoints: total body weight loss percentage (TBWL%), triglycerides (TG), HDL‐C, LDL‐C and absolute changes in waist circumference (WC), haemoglobin A1c (HbA1c) and systolic blood pressure (SBP). Higher CEI values indicate more favourable multidimensional cardiometabolic profiles.

Absolute body weight was not included because of its high correlation with TBWL% and WC. High‐sensitivity C‐reactive protein (hsCRP) and clinical cardiovascular or kidney outcomes were excluded due to inconsistent reporting, low‐risk population and limited follow‐up across obesity trials. The CEI integrates outcome‐specific rankings derived from NMA and should be interpreted as a descriptive summary of relative multidimensional performance rather than a calibrated measure of clinical benefit. Small numerical differences therefore reflect relative patterns rather than clinically meaningful superiority. Secondary outcomes included treatment effects for each individual CEI component.

Two reviewers (J.C. and Y.G.) independently extracted data and assessed risk of bias via the Cochrane Risk of Bias tool, with disagreements resolved by a third reviewer (Y.L.).

Given observed heterogeneity and outcome‐specific inconsistency within the network, random‐effects NMA models synthesized direct and indirect evidence, estimating placebo‐adjusted mean differences with 95% confidence intervals, presented in forest plots. Network inconsistency was assessed using node‐splitting and design‐by‐treatment models. Publication bias was evaluated with Egger's test. Lipid outcomes reported as treatment ratios or percent changes were log‐transformed for analysis and back‐transformed for interpretation. Subgroup analyses were performed by T2D status.

The CEI was constructed using surface under the cumulative ranking curve (SUCRA) values [12] derived from the NMA for seven cardiometabolic risk factors: TBWL%, WC, HbA1c, SBP, TG, HDL‐C and LDL‐C. SUCRA values represent the probability that a treatment ranks among the most effective for each outcome. These values were displayed in radar (origami) plots, with each axis representing one risk factor. The enclosed polygon area was normalized to a 0–1 scale to generate the CEI, with higher values indicating more favourable cardiometabolic profiles. Certainty of evidence was assessed using the GRADE framework (high, moderate, low or very low) [13]. Analyses used R (version 4.5.2, netmeta package) [14].

Of 5022 citations, 19 RCTs (13 117 participants) met the inclusion criteria, enrolling adults with overweight or obesity, with or without T2D (Figure S1). Trials evaluated subcutaneous GLP‐1 receptor mono‐agonists (semaglutide 1.7, 2.4, 7.2 mg; liraglutide 1.8, 3.0 mg) and oral agents (oral semaglutide 25 mg; orforglipron 6, 12 and 36 mg). Baseline characteristics are summarized in Table S2. Mean age was 51.6 years, 61.6% female, BMI 31.6–40.5 kg/m², and 33.0% had T2D. HbA1c ranged 5.5%–5.9% in non‐T2D and 6.7%–8.2% in T2D populations. Trial duration ranged from 44 to 104 weeks.

Between‐study heterogeneity was moderate to high overall and lower in T2D populations, except for HDL‐C (Table). Network inconsistency was observed for TBWL%, WC, HbA1c and LDL‐C (Table). Overall risk of bias was low (Table), and Egger's test did not suggest publication bias (Table). S3 S4 S5 S6

All treatments formed a coherent evidence network with placebo as the common comparator, with up to 18 RCTs contributing to individual endpoint analyses (Figure 1). Seven treatments with complete data across all endpoints were included in the origami plots: subcutaneous semaglutide 1.7, 2.4 and 7.2 mg; subcutaneous liraglutide 3.0 mg; and oral orforglipron 6, 12 and 36 mg.

CEI values (Table 1A, and Figure S2) indicated that higher‐dose GLP‐1 receptor mono‐agonists produced broader improvements across cardiometabolic domains. Semaglutide 7.2 mg had the highest CEI (0.86), followed by orforglipron 36 mg (0.68) and semaglutide 2.4 mg (0.66). Orforglipron 36 mg and semaglutide 2.4 mg showed comparable CEI values but differed in domain contributions: orforglipron showed stronger effects on glycemic control, HDL‐C and SBP, whereas semaglutide demonstrated larger effects on body weight, WC and LDL‐C. Dose–response patterns were evident for both semaglutide and orforglipron.

GLP‐1 receptor mono‐agonists consistently improved cardiometabolic endpoints versus placebo (Figure 1, Table S7). Semaglutide 7.2 mg achieved the largest reduction in TBWL% (MD −14.91%), followed by semaglutide 2.4 mg (−12.10%). Orforglipron showed a dose–response pattern (36 mg: −9.91%; 12 mg: −7.00%; 6 mg: −5.10%). WC reductions paralleled weight loss (semaglutide 7.2 mg: −11.56 cm; 2.4 mg: −9.38 cm; orforglipron 36 mg: −7.77 cm). HbA1c reductions were greatest with semaglutide 7.2 mg (−1.03%) and orforglipron 36 mg (−0.98%) across the overall population, regardless of T2D status. SBP reductions were also largest with semaglutide 7.2 mg (−6.30 mmHg), semaglutide 1.7 mg (−6.02 mmHg) and orforglipron 36 mg (−5.30 mmHg). Lipid outcomes improved modestly and dose‐dependently: TG decreased most with semaglutide 7.2 mg (−26.19%), HDL‐C increased modestly across therapies, with the largest increase observed with orforglipron 36 mg (5.8%) and LDL‐C reductions were largest with semaglutide 1.7 mg (−7.00%), with smaller or non‐significant effects for other agents. Certainty of evidence was rated as moderate to low across endpoints, generally higher for higher‐dose treatments (Table S8).

Across subgroups (non‐T2D versus T2D), CEI rankings were similar to the overall population, with semaglutide 7.2 mg ranking highest, followed by orforglipron 36 mg and semaglutide 2.4 mg (Table 1A–C, Figures S2–S6). Oral semaglutide 25 mg was included only in the non‐T2D subgroup, given the absence of T2D data; its CEI value (0.63) ranked fourth among all agents (Table 1B).

Data are derived from public domain resources.