Analysis of the efficacy and safety of liraglutide, semaglutide, and tirzepatide for the treatment of overweight and obesity: a systematic review and network meta-analysis

Abstract

Background/Objectives

The rapid rise in overweight and obesity worldwide underscores the need for comparative evidence on long-term pharmacotherapy. The recent introduction of a 7.2 mg maintenance dose of semaglutide warrants particular attention. We assessed the efficacy and safety of glucagon-like peptide-1 (GLP-1) and GLP-1/glucose-dependent insulinotropic polypeptide (GIP) receptor agonists—liraglutide, semaglutide, and tirzepatide—focusing on the incremental value of semaglutide 7.2 mg.

Subjects/Methods

We conducted a systematic review and frequentist network meta-analysis (NMA) of phase 3 randomized controlled trials (RCTs) in adults (≥ 18 years) with body mass index (BMI) ≥ 25 kg/m², duration ≥ 52 weeks, comparing the above agents at approved weight management doses vs. placebo or each other (PROSPERO: CRD420251014401). Random effects models using the netmeta package estimated relative effects, ranking used P-scores.

Results

Twenty-four RCTs were included in the analysis, of which 23 were eligible for inclusion in the NMA. All active treatments reduced body weight, waist circumference, and BMI vs. placebo. Tirzepatide 15 mg ranked highest for weight and waist reduction across analyses based on P-scores. Semaglutide 7.2 mg ranked above semaglutide 2.4 mg for all efficacy endpoints based on P-scores; but the significant advantage in percentage bodyweight change was found only in the diabetes subgroup. Liraglutide 3.0 mg yielded the smallest weight loss. No significant between treatment differences were observed for serious adverse events (SAEs); however, tirzepatide was associated with higher risk of injection site reactions, and liraglutide with higher risks of any AEs and discontinuation due to AEs. Considerable heterogeneity was present for efficacy outcomes, with limited evidence of networkwide inconsistency; safety networks showed low to moderate heterogeneity.

Conclusions

GLP-1 and GLP-1/GIP receptor agonists produce clinically meaningful weight loss vs. placebo over ≥ 1 year. Semaglutide 7.2 mg consistently ranks above 2.4 mg, but the average incremental benefit (~ 2–3% points) appears modest and should be interpreted in the context of limited statistically significant differences between doses, as well as balanced against considerations of tolerability and patient preferences. Further head-to-head trials and real world studies are warranted to refine dose selection and long-term safety.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s40618-026-03006-y.

Introduction

Overweight and obesity are recognized as global epidemics of the 21st century and currently represent one of the greatest public health challenges [1]. According to the World Health Organization (WHO), in 2022, approximately 2.5 billion adults aged 18 years and older were classified as having overweight (body mass index [BMI] ≥ 25 kg/m²). Furthermore, 1 in 8 individuals worldwide were affected by obesity (BMI ≥ 30 kg/m²) [2]. This growing trend is expected to continue, with projections indicating that by 2030, more than one billion adults will have obesity [3].

Obesity is a chronic, relapsing, and complex disease with a strong biological basis and limited likelihood of sustained remission without long-term treatment. As a result, healthy lifestyle habits alone are often insufficient to achieve and sustain long-term treatment effects [1, 4, 5]. Therefore, the use of pharmacotherapy in the treatment of overweight and obesity may be necessary and should be considered a key component of a comprehensive weight management strategy [5].

Currently, the most widely used pharmacological treatments for overweight and obesity are incretin mimetics, which include human glucagon-like peptide-1 (GLP-1) receptor agonists (RAs) such as liraglutide [6, 7], semaglutide [8, 9], and tirzepatide—the latter being a dual agonist of GLP-1 and glucose-dependent insulinotropic polypeptide (GIP) receptors [10, 11]. GLP-1 and GLP-1/GIP RAs exert multidirectional effects by activating appetite-regulating pathways that promote a sustained energy deficit [12].

Semaglutide 2.4 mg, administered subcutaneously once weekly, is currently regarded as an effective and safe pharmacological treatment for overweight and obesity. However, not all patients achieve meaningful therapeutic benefits at this dose. Consequently, following the European Commission’s approval on February 17, 2026, of a new 7.2 mg maintenance dose of semaglutide, this threefold higher dose may offer a valuable alternative for these individuals [13].

Despite substantial scientific evidence from randomized controlled trials (RCTs) and meta-analyses confirming the efficacy and acceptable safety profile of GLP-1 and GLP-1/GIP RAs, many individuals remain reluctant to receive pharmacological treatment for overweight and obesity due to concerns about adverse events (AEs) and potential long-term side effects. This highlights the need for a comprehensive network meta-analysis (NMA). Therefore, the aim of this systematic review with NMA was to evaluate the efficacy and safety of liraglutide, injectable semaglutide, and tirzepatide—approved for use in patients with overweight and obesity—with particular focus on semaglutide at the new 7.2-mg dose.

Materials and methods

General principles

This systematic review was conducted in accordance with the PRISMA Extension Statement for Reporting of Systematic Reviews Incorporating Network Meta-Analysis of Health Care Interventions [14, 15], the Cochrane Handbook for Systematic Reviews of Interventions [16], and the guidelines for conducting and interpreting the NMA developed by the International Society for Pharmacoeconomics and Outcomes Research Task Force [17] and Cipriani et al. [18]. The review was registered in the PROSPERO database (registration number: CRD420251014401) [19].

Data sources and search

Eligible studies were identified through a systematic search of 3 major databases: MEDLINE (via PubMed), EMBASE, and the Cochrane Library. The search was conducted in December 2025, and only studies published in English were included. The full search strategy is described in Supplementary Tables 1–3. Additionally, the reference lists of the included studies were screened to identify potentially relevant publications.

Selection criteria

Studies were considered eligible if they were prospective, phase 3 RCTs published in English and conducted in adults (≥ 18 years) with overweight or obesity (BMI ≥ 25 kg/m²), with or without diabetes or other comorbidities. All included trials were required to assess GLP-1 or GLP-1/GIP RAs approved by the European Medicines Agency and/or the US Food and Drug Administration for the treatment of overweight and obesity, including liraglutide, injectable semaglutide, and tirzepatide, compared with each other or with placebo. To evaluate the long-term efficacy of GLP-1 and GLP-1/GIP RAs in weight management, trials with durations shorter than 52 weeks and studies with withdrawal or maintenance designs were excluded. For comparability across studies and to assess the maximum therapeutic potential of each medication, analyses were limited to the following recommended maintenance doses used for weight loss management: liraglutide 3.0 mg, tirzepatide 5 mg, 10 mg, or 15 mg, semaglutide 2.4 mg and the newly approved 7.2 mg-dose of semaglutide. The efficacy outcomes of interest were changes from baseline in body weight (percentage), waist circumference, and BMI. The safety analysis was conducted for the key safety outcomes which included: any AEs, serious AEs (SAEs), AEs leading to discontinuation of the drug or placebo, and injection site reactions. Nonrandomized studies, unpublished studies, and studies published only as conference abstracts were excluded due to concerns regarding methodological quality and the reliability of reported results.

Trials were selected in accordance with PRISMA recommendations [14, 15]. Titles and abstracts identified during database search were screened, and a preliminary list of studies meeting the inclusion criteria was compiled. Full-text articles were then assessed to determine eligibility based on all predefined inclusion and exclusion criteria. This process yielded the final set of studies, which were subsequently evaluated for risk of bias and reported outcomes. Study selection was performed independently by 2 reviewers (W.P., K.Ś.). Any discrepancies at any stage of the review were resolved through discussion, consultation with a third reviewer (P.M.), and ultimately by consensus. Reviewer agreement was high, with approximately 98% concordance.

Data extraction and quality assessment

Data from the included studies were extracted independently by 2 reviewers (W.P., K.Ś., or P.M.) using a predefined data extraction form. The following information was collected and analyzed to assess the homogeneity of the trials: study design (methodology), patient characteristics, treatment regimen and previous therapies, duration of follow-up or treatment, and sample size of study arms. The quality of eligible RCTs was assessed using the Cochrane risk-of-bias 2.0 tool, which evaluates bias arising from the randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Judgements were classified as “low risk of bias,” “some concerns,” or “high risk of bias.” Each domain was rated as having “low” (“+”), “high” (“–”), or “some concerns” (“!”) risk of bias [20, 21]. Microsoft Excel was used to graphically present the results of the risk-of-bias assessment for individual trials.

Data analysis and synthesis

The NMA was conducted using the “netmeta” package in R software [22], which applies a graph-theoretical approach (with vertices representing treatments and edges representing randomized comparisons). This method provides point estimates from the network along with 95% confidence intervals (CIs). This frequentist approach used in “netmeta” serves as an alternative to traditional Bayesian NMA methods [23].

In the NMA, consistency and random-effects models were used, with adjustments made for multi-arm studies. All eligible treatments and their respective dosing regimens, including variations in dose or dosing interval, were included in the network. Each treatment at a specific dosage regimen was represented as a separate node (vertex) in the network graph. However, only the treatments of interest at their approved dosing regimens for the treatment of overweight and obesity were presented in the manuscript (Supplementary Table 4).

All comparisons evaluated in the analyzed trials, including suboptimal or experimental dosing regimens and comparator treatments not assessed in the review, were presented in the Supplementary file. The heterogeneity of evidence was assessed using the Q test, I² statistic, and tau values, while consistency was assessed using the node-splitting approach and comparison with direct evidence [24]. Publication bias was examined by visually inspecting funnel plots for evidence of small-study effects.

Treatment ranking was performed using the P-score, a frequentist equivalent of the surface under the cumulative ranking curve. A higher P-score indicates a higher ranking in terms of safety (i.e., lower risk of AEs) or efficacy (i.e., greater reduction in body weight) [25]. However, caution is warranted when interpreting treatment rankings alone, as they reflect only the probability of a treatment being the best and do not directly account for the magnitude of differences between treatments. Therefore, treatment rankings should be interpreted alongside average event probabilities and relative effect estimates from the NMA [26, 27].

For the NMA, odds ratios (ORs) were used to assess binary outcomes, while mean differences (MDs) were used for continuous outcomes. The results were reported as ORs or MDs with corresponding 95% CIs to indicate the precision of the estimates. Separate networks were constructed for each clinical outcome, provided that outcome definitions were comparable across studies and sufficient data were available for inclusion.

Results

Search results and included studies

The database search yielded several RCTs that met the inclusion criteria for the systematic review and NMA (Fig. 1). A total of 3345 potentially relevant publications were identified, of which 1482 were duplicates and 1808 were excluded after screening titles and abstracts. Of the 55 full-text articles assessed for eligibility, 31 were excluded, resulting in a final sample of 24 trials included in the review [28–51]. The included trials were as follows: 4 RCTs for liraglutide—SCALE Obesity and Prediabetes [28], SCALE Diabetes [29], SCALE IBT [30], and SCALE Insulin [31]; 13 RCTs for semaglutide—STEP 1 [32], STEP 2 [33], STEP 3 [34], STEP 5 [35], STEP 6 [36], STEP 9 [37], STEP 10 [38], STEP-HFpEF [39], STEP-HFpEF DM [40], SELECT [41], STEP UP [42], STEP UP T2D [43], and REDEFINE 1 [44]; 5 RCTs for tirzepatide—SURMOUNT-1 [45], SURMOUNT-2 [46], SURMOUNT-CN [47], SURMOUNT-J [48], and SUMMIT [49]. Additionally, 2 head-to-head RCTs were identified: STEP 8 [50] (semaglutide vs. liraglutide) and SURMOUNT-5 [51] (tirzepatide vs. semaglutide). The SURMOUNT-5 trial was excluded from the NMA because the doses of semaglutide and tirzepatide were not clearly reported and there were no common comparator prevented network connectivity. Therefore, its results were used for a descriptive comparison with the results obtained from our NMA in the discussion. The NMA was conducted using data from all 23 eligible studies [28–50], as well as within the following subgroups: (1) patients with overweight or obesity without diabetes (15 RCTs: [28, 30, 32, 34, 35, 37–39, 41, 42, 44, 45, 47, 48, 50]); (2) patients with overweight or obesity with diabetes (6 RCTs: [29, 31, 33, 40, 43, 46]); and (3) patients with overweight or obesity, with or without diabetes (18 RCTs: [28–36, 38, 42–48, 50]). The characteristics of the included studies are presented in Table 1, and the baseline characteristics of the study populations are provided in Supplementary Table 5.

Fig. 1.

Search flow diagram

Table 1.

Methodological characteristics of trials included in the review and/or network meta-analysis

Acronym [trial number]	References	Trial design	Population	Intervention	Dosing	N	n	Follow-up period	Included in NMA (Yes/No)
SCALE Obesity and Prediabetes [NCT01272219]	Pi-Sunyer 2015 [28]	RCT, phase 3, multicenter (191), double-blind, placebo-controlled	Overweight/obesity ± prediabetes	Liraglutide	3.0 mg	3662	2437	56 weeks	Yes
				Placebo	-		1225
SCALE Diabetes [NCT01272232]	Davies 2015 [29]	RCT, phase 3, multicenter (126), double-blind, placebo-controlled, parallel-group	Overweight/obesity + diabetes	Liraglutide	3.0 mg	623	412	56 weeks	Yes
				Placebo	-		211
SCALE IBT [NCT02963935]	Wadden 2020 [30]	RCT, phase 3b, multicenter (17), double-blind, placebo-controlled	Overweight/obesity	Liraglutide	3.0 mg	282	142	56 weeks	Yes
				Placebo	-		140
SCALE Insulin [NCT02963922]	Garvey 2020 [31]	RCT, phase 3, multicenter (53), double-blind, placebo-controlled	Overweight/obesity + diabetes	Liraglutide	3.0 mg	396	198	56 weeks	Yes
				Placebo	-		198
STEP 1 [NCT03548935]	Wilding 2021 [32]	RCT, phase 3, multicenter (129), double-blind, placebo-controlled	Overweight/obesity	Semaglutide	2.4 mg	1961	1306	68 weeks	Yes
				Placebo	-		655
STEP 2 [NCT03552757]	Davies 2021 [33]	RCT, phase 3, multicenter (149), double-blind, double-dummy, placebo-controlled	Overweight/obesity + diabetes	Semaglutide	2.4 mg	807	404	68 weeks	Yes
				Placebo	-		403
STEP 3 [NCT03611582]	Wadden 2021 [34]	RCT, phase 3a, multicenter (41), double-blind, placebo-controlled	Overweight/obesity	Semaglutide	2.4 mg	611	407	68 weeks	Yes
				Placebo	-		204
STEP 5 [NCT03693430]	Garvey 2022 [35]	RCT, phase 3, multicenter (41), double-blind, placebo-controlled	Overweight/obesity	Semaglutide	2.4 mg	304	152	104 weeks	Yes
				Placebo	-		152
STEP 6 [NCT03811574]	Kadowaki 2022 [36]	RCT, phase 3a, multicenter (28), double-blind, double-dummy, placebo-controlled	Overweight/obesity ± diabetes	Semaglutide	2.4 mg	300	199	68 weeks	Yes
				Placebo	-		101
STEP 8 [NCT04074161]	Rubino 2022 [50]	RCT, phase 3b, multicenter (19), open-label	Overweight/obesity	Semaglutide	2.4 mg	338	126	68 weeks	Yes
				Liraglutide	3.0 mg		127
				Placebo	-		85
STEP 9 [NCT05064735]	Bliddal 2024 [37]	RCT, phase 3, multicenter (61), double-blind, placebo-controlled	Overweight/obesity + knee osteoarthritis	Semaglutide	2.4 mg	407	271	68 weeks	Yes
				Placebo	-		136
STEP 10 [NCT05040971]	McGowan 2024 [38]	RCT, phase 3, multicenter (30), double-blind, parallel-group	Overweight/obesity + prediabetes	Semaglutide	2.4 mg	207	138	52 weeks	Yes
				Placebo	-		69
STEP-HFpEF [NCT04788511]	Kosiborod 2023 [39]	RCT, phase 3, multicenter (96), double-blind, placebo-controlled	Overweight/obesity + Heart Failure with Preserved Ejection Fraction	Semaglutide	2.4 mg	529	263	52 weeks	Yes
				Placebo	-		266
STEP-HFpEF DM [NCT04916470]	Kosiborod 2024 [40]	RCT, phase 3, multicenter (108), double-blind, placebo-controlled	Overweight/obesity + Heart Failure with Preserved Ejection Fraction + diabetes	Semaglutide	2.4 mg	616	310	52 weeks	Yes
				Placebo	-		306
SELECT [NCT03574597]	Lincoff 2023 [41]	RCT, phase 3, multicenter (804), double-blind, placebo-controlled	Overweight/obesity + cardiovascular diseases	Semaglutide	2.4 mg	17,604	8803	104 weeks	Yes
				Placebo	-		8801
STEP UP [NCT05646706]	Wharton 2025 [42]	RCT, phase 3b, multicenter (95), double-blind, placebo-controlled, active-controlled	Overweight/obesity	Semaglutide	7.2 mg	1407	1005	72 weeks	Yes
				Semaglutide	2.4 mg		201
				Placebo	-		201
STEP UP T2D [NCT05649137]	Lingvay 2025 [43]	RCT, phase 3b, multicenter (68), double-blind, placebo-controlled, active-controlled	Overweight/obesity + diabetes	Semaglutide	7.2 mg	512	307	72 weeks	Yes
				Semaglutide	2.4 mg		103
				Placebo	-		102
REDEFINE 1 [NCT05567796]	Garvey 2025 [44]	RCT, phase 3a, multicenter, double-blind, placebo-controlled, active- controlled	Overweight/obesity	Semaglutide	2.4 mg	1007	302	68 weeks	Yes
				Placebo	-		705
SURMOUNT-1 [NCT04184622]	Jastreboff 2022 [45]	RCT, phase 3, multicenter (119), double-blind, placebo-controlled	Overweight/obesity	Tirzepatide	5 mg	2539	630	72 weeks	Yes
				Tirzepatide	10 mg		636
				Tirzepatide	15 mg		630
				Placebo	-		643
SURMOUNT-2 [NCT04657003]	Garvey 2023 [46]	RCT, phase 3, multicenter (77), double-blind, placebo-controlled	Overweight/obesity + diabetes	Tirzepatide	10 mg	938	312	72 weeks	Yes
				Tirzepatide	15 mg		311
				Placebo	-		315
SURMOUNT-5 [NCT05822830]	Aronne 2025 [51]	RCT, phase 3b, multicenter (32), open-label, parallel-arm, controlled trial	Overweight/obesity	Tirzepatide	10 mg or 15 mg (MTD)	750	374	72 weeks	No
				Semaglutide	1.7 mg or 2.4 mg (MTD)		376
SURMOUNT-CN [NCT05024032]	Zhao 2024 [47]	RCT, phase 3, multicenter (29), double-blind, placebo-controlled	Overweight/obesity	Tirzepatide	10 mg	210	70	52 weeks	Yes
				Tirzepatide	15 mg		71
				Placebo	-		69
SURMOUNT-J [NCT04844918]	Kadowaki 2025 [48]	RCT, phase 3, multicenter (17), double-blind, placebo-controlled, parallel-group	Overweight/obesity	Tirzepatide	10 mg	225	73	72 weeks	Yes
				Tirzepatide	15 mg		77
				Placebo	-		75
SUMMIT [NCT04847557]	Packer 2025 [49]	RCT, phase 3, multicenter (129), double-blind, placebo-controlled	Overweight/obesity + Heart Failure with Preserved Ejection Fraction ± diabetes	Tirzepatide	15 mg	731	364	52 weeks	Yes
				Placebo	-		367

MTD – maximum tolerated dose; NMA – Network Meta-Analysis; RCT – randomized controlled trial

All included studies were multicenter, randomized phase 3 trials. Baseline characteristics were comparable between active-treatment and placebo groups across all trials. Most studies were double-blinded, with the exception of STEP 8 [50] and SURMOUNT-5 [51]. Networks were constructed only for outcomes with similar definitions in all trials.

The risk-of-bias assessment is presented in Fig. 2. Most of the included studies were assessed as having a “low” risk of bias. However, 4 trials—SCALE Obesity and Prediabetes [28], STEP 8 [50], SELECT [41], and SURMOUNT-5 [51]—were identified as having “some concerns”. All trials reported data on the intention-to-treat analysis, except for SCALE Diabetes [29], SURMOUNT-5 [51], and SURMOUNT-J [48], which used a modified intention-to-treat analysis. In the STEP 8 [50] trial, randomization to semaglutide or liraglutide was not blinded due to differences in dosing regimens. However, each active treatment group was double-blinded against its corresponding placebo to minimize potential bias. Still, the absence of blinding between the 2 active comparators may have influenced participants or investigator behavior. Consequently, this aspect was judged as raising “some concerns” in the “deviations from intended interventions” domain. Due to its open-label design, the SURMOUNT-5 [51] trial also received a rating of “some concerns” in the same domain. The SCALE Obesity and Prediabetes [28] trial was assessed as having “some concerns” in the “missing outcome data” domain due to a high number of participant withdrawals and dropouts. In the SELECT [41] trial, some participants (36 in the semaglutide group and 121 in the placebo group) received open-label GLP-1 RA treatment, which constituted a protocol violation. As a result, this trial also received a rating of “some concerns” rating in the “deviations from intended interventions” domain.

Fig. 2.

Risk-of-bias assessment

The input data for the NMA are presented in the supplementary materials (Supplementary Tables 6 and 7).

NMA results: change in body weight

A total of 23 RCTs were included in the NMA for the percentage change in body weight endpoint. Graphical representations of the treatment networks for the primary analysis and subgroup analyses are provided in Supplementary Fig. 1. In our analysis, all evaluated drugs showed a significantly greater relative reduction in body weight% compared with placebo, regardless of the analysis performed (Supplementary Fig. 5). However, subgroup analyses suggest a potential effect modification by diabetes status (Supplementary Tables 8–11).

Tirzepatide 15 mg resulted in a significantly greater percentage reduction in body weight compared with all other interventions, except semaglutide 7.2 mg in the main analysis. However, in the subgroup of patients with diabetes, this difference was also statistically significant. Based on the P-score, the highest dose of tirzepatide ranked as the most effective therapeutic option for body weight% reduction (ranked first) across both the primary and all subgroup analyses (Supplementary Tables 8–11 and 23–26).

Semaglutide 7.2 mg ranked second among the studied treatments in the main analysis. No statistically significant difference was observed between the 7.2-mg and 2.4-mg doses of semaglutide overall; however, based on the P-score, the higher dose was associated with a potentially greater percentage reduction in body weight. Findings in the subgroups of patients without diabetes, and those with or without diabetes, were consistent with those obtained in the main analysis. In the subgroup of patients with diabetes, semaglutide 7.2 mg ranked third according to the P-score. Notably, in this subgroup, the difference between the 7.2-mg and 2.4-mg doses was statistically significant, favoring the higher dose. Clinically, this suggests that while the difference between semaglutide 7.2 mg and 2.4 mg is modest, it may be particularly relevant in patients with diabetes (Supplementary Tables 8–11 and 23–26). Overall, semaglutide 7.2 mg consistently ranked higher than the 2.4-mg dose, although statistically significant differences between the doses were limited and appeared to depend on the patient population.

Across all analyses, liraglutide 3.0 mg was associated with a lower percentage reduction in body weight compared with the other drugs evaluated (Supplementary Tables 8–11).

NMA results: change in waist circumference

For the endpoint of change in waist circumference, 21 RCTs were included in the NMA. The corresponding network diagrams for the primary analysis and subgroup analyses are shown in Supplementary Fig. 2. Our NMA revealed that all analyzed active interventions led to significantly greater reductions in waist circumference compared with placebo (Supplementary Fig. 5). According to the P-score rankings, tirzepatide 15 mg had the highest probability of being in the first ranking position, regardless of the analysis performed. In the subpopulation with diabetes, tirzepatide 15 mg demonstrated statistically significant superiority in all pairwise comparisons (Supplementary Tables 12–15 and 23–26).

Semaglutide 7.2 mg ranked third in the main analysis and in the subgroup of patients with diabetes, while it ranked second in the subgroups without diabetes and with or without diabetes. Although no statistically significant differences were observed between the 7.2-mg and 2.4-mg doses, all analyses showed greater reductions in waist circumference in favor of the higher dose (Supplementary Tables 12–15 and 23–26).

Subgroup analyses produced results consistent with the main analysis. In the subgroup without diabetes, all evaluated drugs demonstrated greater reductions in waist circumference relative to placebo compared with the subgroups including individuals with diabetes and the mixed population (Supplementary Tables 12–15).

NMA results: change in body mass index

A total of 11 RCTs were included in the NMA for the change in BMI endpoint. Graphical representations of the treatment networks for the primary analysis and subgroup analyses are presented in Supplementary Fig. 3. The study demonstrated significant superiority of GLP-1 and GLP-1/GIP RAs over placebo in reducing BMI (Supplementary Fig. 5). In the primary analysis, semaglutide 7.2 mg ranked first based on the P-score, suggesting it may be the most effective treatment; however, its differences compared with the treatments ranked second to fourth were not statistically significant. This top ranking for semaglutide 7.2 mg was also observed in the subgroups without diabetes and in the mixed population. However, in the subgroup with diabetes, semaglutide 7.2 mg was ranked below tirzepatide 15 mg, although the difference between these treatments was not statistically significant in either analysis (Supplementary Tables 16–18 and 23–26).

The difference in BMI reduction between semaglutide 7.2 mg and 2.4 mg favored the higher dose and reached statistical significance in both subgroups of individuals with and without diabetes (Supplementary Tables 16–18). Compared with liraglutide 3.0 mg, all other evaluated treatments were associated with significantly greater reductions in BMI across all analyses (Supplementary Tables 16–18).

The results of the subgroup analyses in patients with or without diabetes were consistent with those of the main analysis, as all were based on the same set of study data. In the subgroup of individuals with diabetes, all evaluated treatments showed smaller reductions in BMI relative to placebo compared with the reductions observed in the other subgroups (Supplementary Tables 16–18).

NMA results: safety outcomes

For the safety endpoints, 18 RCTs were included in the NMA of any AEs, 23 for SAEs, 23 for AEs leading to discontinuation of the drug or placebo, and 12 for injection site reactions. The corresponding network diagrams are shown in Supplementary Fig. 4. Safety analysis was conducted only for the main analysis and focused on key outcomes: any AEs, SAEs, AEs leading to discontinuation of the drug or placebo, and injection site reactions. All analysed drugs were associated with a statistically significant higher risk of any AEs compared with placebo (Supplementary Fig. 6). Tirzepatide 15 mg ranked second after placebo based on the P-score, suggesting it may be associated with the lowest risk of any AEs among the active treatments (Supplementary Table 27). Liraglutide 3.0 mg was associated with a significantly higher risk of any AEs compared with both tirzepatide 15 mg and placebo (Supplementary Table 19). No significant differences in the incidence of SAEs were observed between treatments (Supplementary Table 20). For AEs leading to discontinuation of the drug or placebo, no statistically significant differences were observed between most of the studied treatments (Supplementary Table 21). However, liraglutide 3.0 mg was associated with a significantly higher risk of AEs leading to treatment discontinuation compared with tirzepatide 5 mg, semaglutide 2.4 mg and placebo (Supplementary Table 21). P-score analysis indicated that semaglutide 2.4 mg may be associated with the lowest rate of injection site reactions among the studied drugs (rank 1) (Supplementary Table 27). In contrast, tirzepatide at all 3 doses (5, 10, and 15 mg) was associated with a higher frequency of injection site reactions compared with both doses of semaglutide (2.4 and 7.2 mg), liraglutide 3.0 mg, and placebo (Supplementary Table 22). Across all safety outcomes, no significant differences were found between semaglutide 7.2 mg and 2.4 mg. Nonetheless, the trend suggests that semaglutide 7.2 mg may be associated with a higher risk of AEs than the lower dose of 2.4 mg, despite the lack of statistical significance (Supplementary Tables 19–22).

Assessment of the network

The study revealed substantial heterogeneity in effect sizes across all efficacy outcomes (Supplementary Table 28) in networks including all trials. This was evident from the total Q and I² statistics for percentage change in body weight (I²=92%), change in waist circumference (I²=86%), and change in BMI (I²=92.4%). For all these outcomes, there was strong evidence of overall heterogeneity (p < 0.0001) and considerable within-design heterogeneity (p < 0.0001), indicating notable variability between trials, even within the same design or comparison structure. The comparison between placebo and semaglutide 2.4 mg was identified as the main contributor to between-design heterogeneity. Standard decomposition suggested between-design inconsistency (p < 0.0001); however, under a full design-by-treatment interaction random-effects framework, the remaining inconsistency was not statistically significant (p > 0.05), despite substantial within-design heterogeneity (τ²_within ≈ 4.08 for percentage change in body weight and τ²_within ≈ 4.24 for change in waist circumference). These findings support the conclusion that heterogeneity, rather than clear network-wide inconsistency, is the primary issue. Although the standard Q decomposition indicated possible between-design inconsistency, neither the design-by-treatment interaction model nor the SIDE (separate indirect from direct evidence) checks provided convincing evidence of inconsistency for most comparisons. This supports the adequacy of the random-effects consistency model for the main analysis. No substantial disagreement between direct and indirect evidence was observed.

Similarly, substantial heterogeneity in effect sizes was observed in the subgroup of patients without diabetes, as well as in the subgroup of patients with overweight or obesity and with or without diabetes, after excluding trials involving cardiovascular disease or osteoarthritis. This heterogeneity was evident in the networks assessing percentage change in body weight and change in waist circumference. In contrast, the network for BMI change in the subpopulation without diabetes showed low heterogeneity, with an I² value of 15.5%. In the subgroup of patients with diabetes, low heterogeneity was observed in the network assessing changes in waist circumference, while for the remaining outcomes, I² could not be calculated due to the small number of available trials.

A low level of heterogeneity in the effect sizes (Supplementary Table 29) was found in the networks assessing any AEs (I²=31.6%), AEs leading to discontinuation of the drug or placebo (I²=0.9%), and injection site reactions (I²=0%), with no significant heterogeneity within or between designs (p > 0.5 for both). Moderate heterogeneity (I²=54.8%) was found in the SAEs network, with substantial within-design heterogeneity (p < 0.0001), primarily concentrated in the placebo vs. semaglutide 2.4 mg comparison (p < 0.0001); other designs showed no meaningful within-design heterogeneity. A notably high proportion of patients with SAEs observed in the placebo arms (> 20%) of the SELECT [41], STEP-HFpEF [39], and STEP-HFpEF DM [40] trials differed significantly from the rates observed in the remaining studies. This may be attributed to the fact that the patient populations in these studies were more burdened—mean patient age was significantly higher than in the other studies, exceeding 60 years—and inclusion criteria specifically required a diagnosis of cardiovascular diseases. Most SIDE tests were nonsignificant, indicating no substantial disagreement between direct and indirect evidence, and supporting the appropriateness of the random-effects consistency model.

Discussion

Our study was designed to evaluate the efficacy and safety profiles of liraglutide, injectable semaglutide, and tirzepatide in the treatment of overweight and obesity, with particular emphasis on injectable semaglutide, the most extensively studied GLP-1 analogue. This is particularly relevant in the context of the European Commission’s approval of a higher 7.2 mg dose on 17 February 2026 [13]. The NMA results demonstrated statistically significant reductions in body weight and a favorable safety profile with no significant increase in the risk of SAEs for any of the investigated drugs.

The 7.2-mg dose of semaglutide showed numerically greater reductions across all efficacy endpoints compared with the 2.4 mg dose. On the other hand, a possible trend towards a higher risk of any AEs and a higher rate of treatment discontinuation was observed for the 7.2-mg dose compared with semaglutide 2.4 mg. However, none of these differences reached statistical significance in the primary analysis. In practice, the observed discrepancy of approximately 2–3% points in weight reduction between 2.4 mg and 7.2 mg is unlikely to translate into a clinically meaningful improvement for most patients. This modest numerical advantage remains substantially lower than the considerably larger treatment effects consistently observed for tirzepatide 10–15 mg. Although the P-score indicated a higher probability of greater weight reduction with the 7.2-mg dose, ranking metrics should be interpreted separately from the magnitude of clinically meaningful treatment effects. Consequently, while the direction of effect favours the higher dose, its practical relevance should be interpreted with caution, particularly in the context of non-significant statistical differences and the limited number of available studies evaluating the higher dose.

As expected, tirzepatide emerged as an effective and safe therapeutic option in terms of both efficacy and safety. Its lower risk of AEs is likely related to its dual agonist activity, including action at the GIP receptor, which may help mitigate gastrointestinal side effects such as nausea and vomiting [12]. However, its use may be associated with a higher risk of injection site reactions. It is important to recognize that patients with overweight and obesity represent a heterogeneous population, and individual responses to therapy may vary. Therefore, treatment should be tailored to each patient’s needs to optimize clinical outcomes.

In December 2025, in response to the rapidly growing global burden of overweight and obesity, the WHO issued its first global guidance on the long-term use of GLP-1 and GLP-1/GIP RAs in the treatment of overweight and obesity in adults [52]. These recommendations underscore the need for current evidence on the efficacy and safety profiles of liraglutide, semaglutide, and tirzepatide in the long-term management of overweight and obesity and their associated comorbidities. Our systematic review with NMA provides cumulative evidence that supports the implementation of updated standards of care for patients with obesity.

A systematic review with NMA by Sinha et al. [53] included 19 studies assessing GLP-1 RAs (liraglutide, semaglutide, orforglipron), dual agonists targeting GLP-1/GIP or GLP-1/glucagon receptor pathways (tirzepatide, survodutide), and a triple agonist combining GLP-1, GIP, and glucagon receptor agonism (retatrutide) for weight loss in adults with overweight or obesity. However, the systematic review included RCTs from phases 2 to 4, which may complicate the interpretation of results due to variability in trial objectives, designs, and populations. Additionally, a methodological limitation was identified: the authors limited their search to the Cochrane database, which may have reduced the number of eligible studies and impacted the overall credibility of the review. Another key difference lies in the inclusion criteria. While our analysis focused exclusively on drugs approved for the treatment of obesity and overweight, Sinha et al. [53] also included investigational agents currently undergoing clinical evaluation. This fundamental distinction reflects the differing objectives of the 2 analyses.

The results of our NMA comparing the 2 most widely used medications for the treatment of overweight and obesity—tirzepatide and semaglutide—are consistent with findings from the head-to-head SURMOUNT-5 trial [51]. In that study, patients received their maximum tolerated dose of tirzepatide (10 mg or 15 mg) or semaglutide (1.7 mg or 2.4 mg). Notably, 89.3% of participants in the tirzepatide group received at least one 15-mg dose, while 92.8% in the semaglutide group received at least one 2.4-mg dose. These data allow for meaningful comparison with our NMA findings, in which the included studies also permitted dose reductions in cases of unacceptable side effects. For efficacy outcomes, our analysis in the subpopulation of patients without diabetes confirmed significantly greater reductions in body weight% and waist circumference with tirzepatide 15 mg compared with semaglutide 2.4 mg. A similar directional effect favoring tirzepatide was observed for BMI reduction, although this difference did not reach statistical significance. The safety analysis of the NMA also aligned with the findings of the head-to-head SURMOUNT-5 study, in which no significant differences were observed between tirzepatide 15 mg and semaglutide 2.4 mg in the incidence of any AEs, SAEs, and AEs leading to treatment discontinuation. An increased risk of injection site reactions in the tirzepatide group compared with the semaglutide group was also consistent with the results of the direct comparison. However, this RCT may carry a risk of bias due to its open-label design and the lack of precise specification of the administered drug doses. Therefore, its findings should be interpreted with caution.

One of the key strengths of our study is its high methodological and substantive quality supported by a systematic review and NMA conducted in accordance with widely accepted methodological standards. We applied a rigorous approach based on the PRISMA Extension Statement [14, 15], the Cochrane Handbook for Systematic Reviews of Interventions [16], and ISPOR guidelines for NMAs [17, 18]. A major advantage of NMA, as reflected in our analysis, is its ability to compare multiple therapeutic options (3 or more) simultaneously, integrating both direct and indirect evidence. In contrast, conventional methods such as classic meta-analyses or indirect comparisons are limited to evaluating no more than 2 interventions at a time. This limitation poses a challenge for clinical decision-making when multiple effective treatment options are available for a given condition. Furthermore, our analysis included only RCTs, which are considered the highest standard of evidence in clinical research.

Other important strengths of the NMA include the large sample size and the high methodological quality of the included studies. The study methods and procedures were prespecified and described in detail in a protocol registered in PROSPERO [19]. Data collection, extraction, and calculation were performed independently by 2 authors. Finally, this NMA provided a comprehensive evaluation of key clinical efficacy and safety outcomes.

Patients with overweight or obesity often present with additional comorbidities, such as diabetes and cardiovascular diseases. Our main analysis included all studies evaluating the use of the selected drugs in individuals with overweight or obesity, regardless of the presence or absence of diabetes or cardiac comorbidities. This pragmatic approach was adopted to enhance the external validity and clinical applicability of the findings. Another strength of this NMA was the inclusion of subgroup analyses, which allowed for verification and more nuanced interpretation of the overall findings. This approach is particularly important in this context, as treatment outcomes may be influenced not only by the pharmacological intervention itself but also by concurrent behavioral therapies, lifestyle changes, and other patient-specific factors.

Despite the strengths outlined above, our findings should be interpreted with caution due to potential methodological limitations of the NMA and the risk of bias. The interpretation of absolute treatment effects in this NMA should be approached with caution due to the considerable heterogeneity observed across the included trials. As demonstrated by the high I² values—reaching 86–92% for the principal efficacy endpoints—substantial variability exists in the underlying evidence base, which limits the robustness of pooled absolute estimates. This heterogeneity is likely to be indicative of a number of structural differences between the studies. Firstly, the populations enrolled exhibited significant heterogeneity, encompassing individuals with obesity alone, those with type 2 diabetes, and those with additional comorbid conditions such as heart failure with preserved ejection fraction or knee osteoarthritis. It is important to note that each of these factors influences achievable weight loss and treatment responsiveness [54]. A lower treatment effect was observed in participants with diabetes compared with those without diabetes; however, substantial heterogeneity remained in the non-diabetic subgroup analysis, indicating that diabetes status alone does not fully explain the observed variability. Potential effect modifiers, including baseline BMI, age, and sex distribution, were not systematically different across treatment comparisons based on reported baseline characteristics. Although meta-regression could theoretically be used to explore these factors, the limited number of studies per comparison raised concerns regarding statistical stability and interpretability; therefore, subgroup analyses were considered more appropriate. Secondly, the intensity and type of behavioural interventions accompanying pharmacotherapy differed notably across trials, ranging from standard lifestyle counselling to structured intensive behavioural therapy. This modification of the magnitude of observed effects is therefore of significance. Thirdly, studies utilised divergent titration schedules and maximum achievable doses, particularly for semaglutide and tirzepatide, which may result in disparities in early tolerability, treatment continuity, and weight-loss trajectories. Finally, variability in response to GLP-1–based therapies is increasingly recognised, suggesting that unmeasured biological and behavioural factors may also underlie the remaining heterogeneity. Collectively, these factors serve to reduce the comparability of absolute outcomes, suggesting that relative treatment effects may offer a more reliable basis for inference than pooled absolute reductions in body weight, waist circumference, or BMI. To address heterogeneity, we applied a standard random-effects model, which is appropriate when between-study variability is present. Another potential limitation is the risk of bias due to incomplete data, as not all studies reported every outcome of interest. Missing data were particularly notable for the efficacy outcome of change in BMI; therefore, these results should be interpreted with caution. An additional limitation was the identification of only 2 head-to-head trials, of which only one—STEP-8—was eligible for inclusion in the NMA [50]. The second head-to-head study, the SURMOUNT-5 [51], was excluded from the NMA due to the absence of a common comparator arm, which prevented the study from contributing to a connected evidence network. In contrast to the other included trials, SURMOUNT-5 did not incorporate a placebo group or any intervention that directly overlapped with the comparator structure of the remaining studies. The study elected to compare the administration of tirzepatide at the maximum tolerated dose (10 mg or 15 mg) with that of semaglutide at an individualized maximum tolerated dose (1.7 mg or 2.4 mg), without using fixed, uniform dosing regimens across participants. The absence of a shared comparator, the non-standardised dosing which is incompatible with dose-specific nodes, and the risk of creating a disconnected or unstable evidence structure – SURMOUNT-5 could not be integrated into the quantitative network meta-analysis. Consequently, it was summarised narratively instead. Moreover, due to the considerable heterogeneity observed, the P-score ranking cannot be interpreted as unequivocal evidence that a given intervention is clinically superior, including NMA, which exhibited a high level of heterogeneity for efficacy endpoints (I²=86–92%).

Given the limitations inherent to indirect comparisons, further research is warranted to confirm our findings, ideally through head-to-head RCTs and large-scale observational studies. National and international registries, along with other real-world data sources, could provide valuable complementary evidence on the short- and long-term efficacy and safety of GLP-1 and GLP-1/GIP RAs.

Conclusion

The NMA indicated that the treatment with GLP-1 and GLP-1/GIP RAs was associated with significant reductions in body weight among patients with overweight or obesity and accompanying comorbidities. Our findings suggest that semaglutide 7.2 mg may represent a valuable therapeutic option and could serve as an effective alternative for individuals who do not achieve adequate therapeutic outcomes with the lower dose of semaglutide (2.4 mg). The safety analysis did not reveal a significantly increased risk of AEs with the higher dose of semaglutide compared to other assessed drugs. However, further studies—especially those evaluating long-term use of semaglutide at this dose—are necessary.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

WP and PM conceived the concept and design of the study; WP and KŚ performed systematic review and data extraction; PH conducted network meta-analysis, validated the models, and visualized the results; WP, PM, KŚ, and PK drafted the manuscript; PM and PK critically revised and edited the manuscript. All authors approved the final version submitted for publication.

Funding

The manuscript was self-financed by the authors.

Data availability

All data are from original publications; extraction sheets and R scripts (netmeta) will be made available in the repository.

Declarations

Conflict of interest

The authors declare that they have no conflict of interest.

Ethical approval

Not required (analysis of published, anonymized data).