Comprehensive evaluation of GLP-1 receptor agonists: an umbrella review of clinical outcomes across multiple diseases

Abstract

Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) were initially approved for type 2 diabetes mellitus. Recent evidence suggests their therapeutic potential in various diseases, though their efficacy and safety remain incompletely understood. Here, we conduct an umbrella review including 123 meta-analyses covering 464 outcomes from 5617 articles to comprehensively assess the effectiveness and adverse events (AEs) of GLP-1 RAs across diverse outcomes. GLP-1 RAs showed trends toward improvements in endocrine and metabolic, cardiovascular, renal, and respiratory outcomes, cognitive function, with a potential reduction in fracture risk and all-cause mortality in certain populations. However, increased risks of certain AEs, including diabetic retinopathy, ketoacidosis, gastrointestinal events, and treatment discontinuation, were also observed. AMSTAR 2 assessments indicate that the existing evidence is limited by methodological shortcomings, including incomplete reporting of excluded studies, suboptimal literature search strategies, and insufficient evaluation of how bias in primary studies may influence meta-analytic estimates. Some outcomes did not reach statistical significance in all populations, highlighting the need for further high-quality clinical research. Careful consideration of potential benefits and risks is essential for optimizing treatment outcomes and ensuring patient safety when using GLP-1 RAs in clinical practice.

This umbrella review synthesizes the evidence for safety and efficacy GLP-1 receptor agonists on multiple outcomes including but not limited to metabolic, cardiovascular and renal outcomes.

Introduction

Glucagon-like peptide-1 receptor agonists (GLP-1 RAs) are a class of drugs that mimic the effects of endogenous gut-derived incretin hormone GLP-1¹. These agents primarily exert their effects through specific binding to GLP-1 receptors, thereby stimulating G protein Gs-mediated signaling pathways². The currently approved GLP-1 RAs include exenatide, liraglutide, lixisenatide, dulaglutide, and semaglutide³. Additionally, several novel GLP-1 RAs, including efpeglenatide, ecnoglutide, and orforglipron, are under development and have entered preclinical and clinical trials. GLP-1 RAs were initially approved for the treatment of type 2 diabetes mellitus (T2DM). However, further research has revealed their therapeutic potential across various diseases and health conditions. Exenatide, the first FDA-approved GLP-1 RA in 2005, was derived from Gila monster venom⁴. In 2010, liraglutide—97% homologous to human GLP-1—was approved first for diabetes and later for obesity in 2014, while semaglutide, also a modified GLP-1 analogue, now treats both conditions⁵. Multiple clinical studies have demonstrated that semaglutide effectively induces weight loss, establishing it as an effective anti-obesity treatment for individuals with overweight or obesity^6,7. The weight reduction effect of GLP-1 RAs has gained global attention, prompting researchers to explore their potential in managing other diseases.

In 2023, a pooled analysis of two clinical trials investigated the impact of semaglutide on individuals with obesity and heart failure (HF) with preserved ejection fraction. The results showed that semaglutide significantly improved both weight and HF-related clinical symptoms⁸. Moreover, a large-scale randomized controlled trial (RCTs) involving more than 17,000 individuals with obesity and cardiovascular disease demonstrated that semaglutide reduced the risk of cardiovascular death, myocardial infarction (MI), and stroke compared to the placebo⁹. In addition to cardiovascular outcomes, GLP-1 RAs have been studied in various fields, including endocrine disorders, renal diseases, and neurological conditions, generating encouraging findings in these areas. Consequently, GLP-1 RAs were recognized by Science as the top scientific breakthrough in 2023, highlighting their current and future significance as a research hotspot¹⁰.

Despite significant progress in GLP-1 RA research, their adverse events (AEs) should not be ignored. Gastrointestinal side effects, such as nausea and vomiting, have been reported, along with concerns about an increased risk of intestinal obstruction and pancreatitis^11,12. However, some clinical evidences have presented conflicting evidence, suggesting that the AEs of GLP-1 RAs may not be as severe or widespread as initially thought. A growing body of meta-analyses examines the overall or specific therapeutic efficacy of GLP-1 RAs across various diseases. Yet overlapping analyses of variable quality complicate drawing definitive, comprehensive conclusions.

In this work, we systematically evaluate the therapeutic efficacy and AEs of GLP-1 RAs across multiple disease areas through an umbrella review of meta-analyses. We show that GLP-1 RAs improve endocrine and metabolic, cardiovascular, renal, respiratory, and cognitive outcomes, and reduce all-cause mortality in certain populations. However, their use is associated with increased risks of specific AEs, and some outcomes do not reach statistical significance across all populations, although the direction of effects remains largely consistent. Overall, GLP-1 RAs demonstrate promising clinical benefits across multiple domains, while further large-scale and high-quality studies are warranted to confirm these findings and elucidate the underlying mechanisms.

Result

Characteristics of included meta-analyses

A total of 5617 articles were retrieved from five databases. After de-duplication and screening based on inclusion and exclusion criteria, 123 meta-analyses (115 based on RCTs and 8 based on observational studies) with 464 outcomes were included in this umbrella review (Fig. 1, Supplementary data 1). These outcomes were categorized into endocrine and metabolic outcomes, cardiovascular outcomes, cancer outcomes, renal outcomes, respiratory outcomes, mortality and AEs, and other outcomes. Only statistically significant results are presented in the main text, while non-significant findings are reported in Supplementary information.

Fig. 1. PRISMA flow diagram of the study selection process.

Flow diagram illustrating the identification, screening, eligibility assessment, and inclusion of studies in the meta-analyses, in accordance with the PRISMA guidelines. PRISMA Preferred Reporting Items for Systematic Reviews and Meta-Analyses, GLP-1 RAs glucagon-like peptide-1 receptor agonists, PICOs Participants, Interventions, Comparisons, Outcomes, and Study design.

Detailed AMSTAR 2 and GRADE assessments are provided in supplementary data 2 and 3. The median AMSTAR 2 score for the included meta-analyses was 23 (range: 7–32; interquartile range: 19–26). Several critical domains showed consistently low ratings. Item 7 (whether a list of excluded primary studies with justification was provided) received the lowest scores, with only eight reviews explicitly presenting such a list. While most reviews reported the total number of excluded primary studies and general exclusion criteria, few provided detailed justifications for individual exclusions. This omission, possibly due to space limitations, compromises the transparency and reproducibility of the evidence synthesis process. Item 4 (whether a comprehensive literature search strategy was used) was also frequently rated as partially met, as most meta-analyses did not clarify whether grey literature was searched when appropriate or whether experts in the field were consulted. These omissions may have led to incomplete retrieval of relevant primary studies and introduced selection bias. Additionally, Item 12 (whether the impact of risk of bias in individual primary studies on the meta-analysis results was assessed) also received relatively low scores. Many meta-analyses failed to adequately explore how bias in the primary studies may have influenced the overall findings, thereby undermining the robustness and credibility of the synthesized results. To further evaluate the reliability of the evidence, we assessed the degree of overlap among systematic reviews using the CCA method (Supplementary data 4). Although some primary studies were included in multiple meta-analyses, the overall level of overlap was low, indicating a limited impact of duplicate data. However, overlapping primary studies were predominantly concentrated in a limited number of large-scale, frequently cited RCTs in the field of GLP-1 RAs-such as LEADER, SUSTAIN, EXSCEL, HARMONY Outcomes, REWIND, ELIXA, SELECT, PIONEER, AWARD, AMPLITUDE, LEAD, SCALE, and STEP.

Endocrine and metabolic outcomes

A total of 41 meta-analyses (118 outcomes) evaluated GLP-1 RAs for endocrine and metabolic outcomes (Fig. 2, supplementary table 1, supplementary data 5). Evidence classification showed that one outcome was Class I (0.8%), one outcome was Class III (0.8%), 86 were Class IV (73%), and the rest were NS. The GRADE assessment rated 24 outcomes (20%) as High, 17 (14%) as Moderate, 25 (21%) as Low, and 52 (44%) as Very Low. Among 41 meta-analyses, 6 did not report primary studies. The CCA for the remaining meta-analyses was 1.32%, indicating a slight level of overlap.

Fig. 2. Overall effects of GLP-1 RAs on endocrine and metabolic outcomes.

All included meta-analyses were based on RCTs and analyzed using random-effects models. All statistical tests were two-sided, and no adjustments were made for multiple comparisons. Effect estimates are shown as pooled effect sizes with 95% confidence intervals. NR not reported, IG intervention group, CG control group, VL Very Low, L Low, M Moderate, H High, NAFLD nonalcoholic fatty liver disease, NASH nonalcoholic steatohepatitis, MAFLD metabolic dysfunction–associated fatty liver disease, T2DM type 2 diabetes mellitus, PCOS polycystic ovary syndrome, AD Alzheimer’s disease, MCI mild cognitive impairment, DKA diabetic ketoacidosis, ALT alanine aminotransferase, AST aspartate aminotransferase, BMI body mass index, HbA1c hemoglobin A1c, IHA intrahepatic fat accumulation, LFC liver fat content.

Liver enzyme profiles: For clarity, we synthesize legacy labels (NAFLD/NASH and MAFLD) under the current steatotic liver disease (SLD) framework—MASLD (formerly NAFLD) and MASH (formerly NASH). Across the SLD spectrum, GLP-1 RAs demonstrated modest but directionally consistent improvements in liver enzyme profiles (alanine aminotransferase [ALT] and aspartate aminotransferase [AST]), with 4 of the 7 ALT outcome estimates showing significant reductions, and 4 of the 9 AST outcome estimates showing significant reductions. For transparency, the following results are presented according to the original terminology used in the included meta-analyses. In drug-specific analyses, exenatide was associated with reductions in ALT (Class IV; High) and AST (Class IV; Low) among persons labeled NAFLD/NASH. At the drug-class level, ALT decreased in NAFLD, NAFLD or NASH, and T2DM with NAFLD (all Class IV; Very Low), while AST decreased in MAFLD, NAFLD, and T2DM with NAFLD (all Class IV; Very Low).

BMI and Body Weight: GLP-1 RAs were associated with notable reductions in both body mass index (BMI) and body weight across a wide range of populations. Of the 24 reported BMI outcomes, 18 showed significant reductions, and 30 of the 33 reported body-weight outcomes were significant. At the class level, GLP-1 RAs reduced BMI in individuals with Alzheimer’s disease (AD) or mild cognitive impairment (MCI) (Class IV; High), those overweight/obesity and psychotic disorders (Class IV; High), individuals with polycystic ovary syndrome (PCOS) (Class IV; High), PCOS with overweight (Class IV; High), T2DM with NAFLD (Class IV; Moderate), adolescents with overweight/obesity (Class IV; Low), MAFLD (Class IV; Very Low), and T2DM (Class IV; Very Low). For body weight, significant reductions were observed in individuals with overweight/obesity (Class I; Low), youth with T2DM (Class IV; High), premenopausal women with PCOS and overweight/obesity (Class IV; Moderate), adolescents with overweight/obesity (Class IV; Low), individuals with psychotic disorders receiving antipsychotic medications and having overweight/obesity (Class IV; Low), T1DM (Class IV; Low), NAFLD or NASH (Class IV; Very Low), NAFLD (Class IV; Very Low), T2DM with NAFLD (Class IV; Very Low), and T2DM with obesity (Class IV; Very Low). Detailed single-drug findings are provided in Supplementary Materials 2.

Hemoglobin A1c (HbA1c): Among 27 outcome estimates, 24 showed significant reductions in HbA1c. At the class level, GLP-1 RAs significantly reduced HbA1c in individuals with T2DM and severe chronic kidney disease (CKD) (Class IV; High), youth with T2DM (Class IV; High and Moderate), individuals with T1DM (Class IV; Low), individuals with T2DM and NAFLD (Class IV; Low), individuals with NAFLD or NASH (Class IV; Very Low), and individuals with T2DM and obesity (Class IV; Very Low). Detailed single-drug findings are provided in Supplementary Materials 2.

Hepatic Fat Accumulation: GLP-1 RAs were associated with reductions in hepatic steatosis, as reflected by decreases in intrahepatic fat accumulation (IHA) and liver fat content (LFC), both of which are measures of hepatic lipid deposition. Among 1 IHA outcome estimate, 1 showed a significant reduction; among 6 LFC outcome estimates, 5 showed significant reductions. GLP-1 RAs significantly reduced IHA in individuals with T2DM and NAFLD (Class IV; Very Low). For LFC, significant reductions were observed in individuals with MAFLD (Class IV; High), NAFLD (Class IV; Moderate), T2DM (Class IV; Very Low), and NAFLD or NASH (Class IV; Very Low). In analyses of semaglutide as a single agent, LFC was significantly reduced in individuals with NAFLD (Class IV; Very Low).

Diabetic Ketoacidosis and Ketosis: In individuals with T1DM, GLP-1 RAs were associated with an increased risk of ketosis, including diabetic ketoacidosis (DKA) (Class IV; Low). However, analyses of single-agent interventions found that neither exenatide nor liraglutide significantly affected the risk of DKA in individuals with T1DM, indicating insufficient evidence to support a definitive protective or harmful effect at the individual drug level.

Diabetic Retinopathy: Evidence from eight outcomes assessing diabetic retinopathy risk yielded predominantly non-significant findings. One meta-analysis reported a statistically significant increase in risk, whereas the remaining seven showed no significant associations. Across drug-specific analyses, liraglutide, lixisenatide, semaglutide, and albiglutide generally showed no significant associations. In one RCT, albiglutide 30 mg was associated with an increased risk of retinopathy (Class IV; High), while albiglutide 50 mg showed no significant effect.

Cardiovascular outcomes

A total of 52 meta-analyses (covering 141 outcomes) evaluated GLP-1 RAs for cardiovascular outcomes (Fig. 3, supplementary table 2, supplementary data 6). Evidence classification indicated that ten outcomes were Class III (7%), 50 were Class IV (35%), and the rest were NS. The GRADE assessment rated 36 outcomes (26%) as High, 34 (24%) as Moderate, 31 (22%) as Low, and 40 (28%) as Very Low. Among 52 meta-analyses, 7 did not report primary studies. The CCA for the remaining meta-analyses was 2%, indicating a slight level of overlap.

Fig. 3. Overall effects of GLP-1 RAs on cardiovascular outcomes.

All included meta-analyses were based on RCTs and analyzed using random-effects models. All statistical tests were two-sided, and no adjustments were made for multiple comparisons. Effect estimates are shown as pooled effect sizes with 95% confidence intervals. Symbols (*) and (F) within the figure refer to additional details explained directly in the panel. (NR not reported, IG intervention group, CG control group, VL very low, L low, M moderate, H high, CVD cardiovascular disease, CAD coronary artery disease, CKD chronic kidney disease, T1DM type 1 diabetes mellitus, T2DM type 2 diabetes mellitus, HF heart failure, MACE major adverse cardiovascular events, MI myocardial infarction, LVEF left ventricular ejection fraction, SBP systolic blood pressure, LDL low-density lipoprotein cholesterol.

Stroke-related outcomes: GLP-1 RAs were associated with a reduced risk of stroke, including non-fatal and fatal subtypes. Among 17 outcome estimates, 6 reported significant reductions. Significant reductions were observed in individuals with cardiovascular disease or at high cardiovascular risk (Class IV; Low), and in individuals with T2DM (Class IV; Low-Very Low). Specifically, non-fatal stroke risk was reduced in individuals with cardiovascular disease/high-risk populations (Class IV; Low), and in individuals with T2DM and established atherosclerotic cardiovascular disease (ASCVD) or high cardiovascular risk (Class IV; Moderate). GLP-1 RAs also significantly lowered fatal stroke risk in individuals with T2DM (Class IV; High).

MI: GLP-1 RAs showed a trend toward a reduced risk of myocardial infarction (MI), including acute subtypes. Among 18 outcome estimates, 4 reported significant reductions. Significant reductions in MI risk were observed in individuals with T2DM (Class IV; High), individuals with overweight or obesity without diabetes (Class III; High), and those without cardiovascular disease or at high cardiovascular risk (Class IV; Very Low). Additionally, liraglutide significantly reduced the risk of acute myocardial infarction (AMI) in individuals with T2DM (Class IV; High).

HF-related outcomes: GLP-1 RAs showed a trend toward a reduced risk of hospitalization for heart failure (HF). Among 20 outcome estimates, 4 reported significant reductions. Significant effects were observed in individuals with diabetes excluding T1DM/GDM (Class IV; High), individuals with T2DM (Class IV; High), and individuals without a history of HF (Class IV; Moderate). In meta-analyses of observational studies, GLP-1 RAs also reduced the risk of HF hospitalization individuals with T2DM without serious cardiovascular events (Class IV; Very Low).

Atherosclerotic/Ischemic outcomes: GLP-1 RAs were associated with a reduced risk of revascularization in individuals with overweight or obesity without diabetes (Class III; High), and a reduced risk of peripheral artery disease in individuals with T2DM (Class IV; Low).

Composite cardiovascular outcomes: GLP-1 RAs were associated with overall cardiovascular benefits across multiple populations. Among the outcome estimates, significant reductions were reported in 13 of 17 for major adverse cardiovascular events (MACE), 5 of 8 for cardiovascular mortality, 3 of 4 for cardiovascular disease (CVD) events, 1 of 3 for cardiovascular death, and 1 of 1 for composite cardiovascular outcomes. These benefits were observed in individuals with T2DM, those with established cardiovascular disease or at high cardiovascular risk, and in individuals with or without a history of HF. Collectively, GLP-1 RAs reduced the risks of MACE, composite cardiovascular outcomes, CVD events, and cardiovascular mortality (Class III–IV; Very Low to High). Full stratified findings are provided in Supplementary Materials 2.

Myocardial injury and function: Among the outcome estimates, significant benefits were reported in 4 of 8 for left ventricular ejection fraction (LVEF), 10 of 17 for systolic blood pressure (SBP), 5 of 12 for LDL cholesterol, and 1 of 1 for final infarct size. GLP-1 RAs improved LVEF in individuals with acute myocardial infarction (AMI) (Class IV; Very Low). Liraglutide, as an individual agent, significantly improved LVEF in individuals with AMI (Class IV; Moderate), MI (Class IV; Moderate), and T2DM with coronary artery disease (CAD) (Class IV; High). SBP reductions were observed in individuals with overweight/obesity, T2DM, and T1DM. Exenatide, liraglutide, and semaglutide were effective in individuals with overweight or obesity (Class IV–III; Moderate to High), while dulaglutide, liraglutide, and semaglutide were effective in those with T2DM (Class IV; Low to High), including those with concomitant obesity (Class IV; Low). Liraglutide also lowered SBP in individuals with T1DM (Class IV; High). GLP-1 RAs reduced LDL cholesterol in multiple populations with metabolic disorders, including individuals with overweight or obesity and antipsychotic-treated psychotic disorders (Class IV; High), non-diabetic individuals treated with semaglutide (Class IV–III; High to Moderate), and individuals with NASH receiving liraglutide (Class IV; Low). In individuals with T2DM and CAD, liraglutide also showed LDL-lowering effects, though with very low evidence certainty (Class IV; Very Low).

Cancer outcomes

Eight meta-analyses (covering 31 outcomes) evaluated GLP-1 RAs for cancer outcomes (Fig. 4A, supplementary data 7). Evidence classification indicated that one outcome was Class IV (3%), and the rest were NS. The GRADE assessment rated three outcomes (10%) as High, 15 (48%) as Moderate, seven (23%) as Low, and six (19%) as Very Low. Among 8 meta-analyses, 4 did not report primary studies. The CCA for the remaining meta-analyses was 8.93%, indicating a moderate level of overlap.

Fig. 4. Overall effects of GLP-1 RAs.

Overall effects of A cancer outcomes; B renal outcomes; C respiratory outcomes; D other outcomes. All included meta-analyses were based on RCTs and analyzed using random-effects models. All statistical tests were two-sided, and no adjustments were made for multiple comparisons. Effect estimates are shown as pooled effect sizes with 95% confidence intervals. Symbols (*) and (F) within the figure refer to additional details explained directly in the panel. NR not reported, IG intervention group, CG control group, VL very low, L low, M moderate, H high, AD Alzheimer’s disease, MCI mild cognitive impairment, T2DM type 2 diabetes mellitus, CAD coronary artery disease, CKD chronic kidney disease, MDRS Mattis Dementia Rating Scale, UPDRS Unified Parkinson’s Disease Rating Scale.

Most outcomes were classified as NS, indicating that GLP-1 RAs have no overall effect on cancer risk. However, compared to placebo or active drugs, GLP-1 RAs were associated with a reduced risk of prostate neoplasm in individuals with T2DM (Class IV, High).

Renal outcomes

Eleven meta-analyses (17 outcomes) evaluated GLP-1 RAs for renal outcomes (Fig. 4B, supplementary data 8). Evidence classification indicated that one outcome was Class II (6%), three outcome was Class III (18%), five were Class IV (29%), and the rest were NS. The GRADE assessment rated seven outcomes (41%) as High, three (18%) as Moderate, five (29%) as Low, and two (12%) as Very Low. Among 11 meta-analyses, 3 did not report primary studies. The CCA for the remaining meta-analyses was 10.58%, indicating a high level of overlap. GLP-1 RAs showed a trend toward improvements in renal damage markers and albuminuria indicators, and toward reduced risks of renal failure and composite renal outcomes across various populations. Detailed outcome-specific results, including effect estimates and stratification by population, are provided in Supplementary Materials 2.

Respiratory outcomes

11 meta-analyses (covering 24 outcomes) evaluated GLP-1 RAs for respiratory outcomes (Fig. 4C, supplementary data 9). Evidence classification indicated that five were Class IV (21%), and the rest were NS. The GRADE assessment rated five outcomes (21%) as High, four (17%) as Moderate, 13 (54%) as Low, and two (8%) as Very Low. Among 11 meta-analyses, 4 did not report primary studies. The CCA for the remaining meta-analyses was 20.83%, indicating a very high level of overlap.

In individuals with T2DM and high cardiovascular risk, GLP-1 RAs reduced the risk of pneumonia (Class IV; Low). In individuals with COVID-19 and T2DM, use of GLP-1 RAs was linked to a lower risk of hospitalization (Class IV; High), risk of severe manifestations (Class IV; High), and mortality (Class IV; Very Low). Similarly, in individuals with diabetes and SARS-CoV-2 infection, GLP-1 RAs were associated with decreased mortality risk (Class IV; Moderate).

Mortality and AEs

In all, 48 meta-analyses (covering 101 outcomes) evaluated GLP-1 RAs for mortality and AEs (Fig. 5, supplementary Table 3, supplementary data 10). Evidence classification indicated that three outcomes were Class I (3%), one outcome was Class II (1%), 13 outcomes were Class III (13%), 23 were Class IV (23%), and the rest were NS. The GRADE assessment rated 29 outcomes (29%) as High, 14 (14%) as Moderate, 18 (18%) as Low, and 40 (40%) as Very Low. Non-significant results were primarily observed for pancreatitis, severe hypoglycemia, and serious adverse events (SAEs). Among 48 meta-analyses, 2 did not report primary studies. The CCA for the remaining meta-analyses was 1.39%, indicating a slight level of overlap.

Fig. 5. Overall effects of GLP-1 RAs on mortality and AEs.

All included meta-analyses were based on RCTs and analyzed using random-effects models. All statistical tests were two-sided, and no adjustments were made for multiple comparisons. Effect estimates are shown as pooled effect sizes with 95% confidence intervals. Symbols (*) and (F) within the figure refer to additional details explained directly in the panel. NR not reported, IG intervention group, CG control group, VL very low, L low, M moderate, H high, AE adverse event, CV cardiovascular, PCOS polycystic ovary syndrome, T1DM type 1 diabetes mellitus, T2DM type 2 diabetes mellitus, COVID-19 coronavirus disease 2019, SARS-CoV-2 severe acute respiratory syndrome coronavirus 2.

AEs: GLP-1 RAs showed a trend toward an increased risk of AEs across multiple populations. Among 16 outcome estimates, 9 reported significant associations. Significant associations were observed in individuals with T1DM (Class III; Very Low), individuals with T2DM (Class III; Low), and women with PCOS and overweight (Class IV; High). Individual agent analyses showed a consistent increase in AE risk with liraglutide, semaglutide, lixisenatide and dulaglutide across individuals with diabetes and obesity.

All-Cause Mortality: GLP-1 RAs showed a trend toward reduced all-cause mortality across diverse populations. Among 15 outcome estimates, 8 reported significant reductions. Significant associations were observed in individuals with T2DM (Class III–IV; High), including those with ASCVD or high cardiovascular risk (Class III; Moderate), and those without a history of HF (Class IV; Low). Similar reductions were reported in individuals with cardiovascular disease (Class IV; Moderate) or at high cardiovascular risk (Class IV; Low). Among individual agents, liraglutide demonstrated a consistent mortality benefit in individuals with T2DM (Class IV; Very Low), consistent with the class effect.

Discontinuation Due to AEs: GLP-1 RAs were associated with an increased risk of treatment discontinuation due to AEs. Among 17 outcome estimates, 13 reported significant associations. Significant increases were observed in individuals with overweight/obesity (Class IV; High), those with obesity and obesity-related comorbidities without diabetes (Class III; Very Low), and individuals with T1DM (Class III; Very Low). Among individual agents, semaglutide, liraglutide, dulaglutide, exenatide, efpeglenatide, and lixisenatide all demonstrated similar trends of increased discontinuation risk in various diabetic and non-diabetic individuals with overweight.

Gastrointestinal AEs: GLP-1 RAs were associated with an increased risk of gastrointestinal AEs in both diabetic and non-diabetic populations. Among 10 outcome estimates, 6 reported significant associations. Elevated risks were observed in individuals with overweight or obesity (Class I–IV; Low to High), individuals with T1DM (Class I; High), and individuals with T2DM (Class II–III; Very Low to High). Among individual agents, semaglutide showed consistently high risk of gastrointestinal AEs in individuals with T2DM (Class II; High), non-diabetic individuals with overweight or obesity (Class I; High), and individuals with overweight/obesity without T2DM (Class III; High). Liraglutide also increased gastrointestinal AEs in individuals with overweight or obesity (Class IV; Low).

Hypoglycemia: GLP-1 RAs showed a trend toward an increased risk of hypoglycemia across several populations. Among 12 outcome estimates, 4 reported significant associations. Elevated risks were observed in individuals with T1DM (Class IV; Low to High), and in youths with T2DM aged 10–18 years (Class IV; High). Among individual agents, liraglutide significantly increased the risk of hypoglycemia in individuals with T1DM (Class IV; High), while semaglutide was associated with a higher risk of hypoglycemia in individuals with overweight/obesity (Class III; High).

Other outcomes

Twenty meta-analysis (covering 32 outcomes) evaluated GLP-1 RAs for other outcomes (Fig. 4D, supplementary table 4, supplementary data 11). Evidence classification indicated that one outcome was Class I (3%), eight were Class IV (25%), and the rest were NS. The GRADE assessment rated 15 outcomes (47%) as High, six (19%) as Moderate, three (9%) as Low, and eight (25%) as Very Low. GLP-1 RAs showed a trend toward favorable neurological outcomes, including improved cognitive function in individuals with AD/MCI, reduced dementia risk in individuals with T2DM, and improvements in Parkinson’s disease scores. However, they were linked to an increased risk of headache in women with PCOS. Additionally, GLP-1 RAs reduced the risks of genital infections and bone fractures in individuals with T2DM. Detailed outcome-specific results across individual groups and agents are presented in Supplementary Materials 2.

Discussion

This umbrella review systematically reviewed 123 meta-analyses encompassing 464 outcomes on GLP-1 RAs across various populations. The outcomes were grouped into seven categories: endocrine and metabolic, cardiovascular, cancer, renal, respiratory, mortality and AEs, and other outcomes. Overall, GLP-1 RAs showed a trend toward improvements across most disease outcomes. However, potential AEs were also highlighted, emphasizing the need for careful evaluation in clinical practice to ensure proper usage.

GLP-1 RAs have demonstrated favorable trends toward improving endocrine and metabolic outcomes, including reductions in ALT, AST, HbA1c, body weight, BMI, LFC, and IHA. The glucose-lowering effect of GLP-1 RAs primarily operates through stimulating insulin secretion, inhibiting glucagon release, and reducing hepatic glucose production. Upon binding to GLP-1 receptors, GLP-1 RAs activate G-protein signaling pathways, leading to increased cyclic AMP and activation of protein kinase A, which in turn enhances insulin biosynthesis and suppresses glucagon secretion^13–15. Weight reduction is likely mediated by delayed gastric emptying, enhanced satiety, and reduced caloric intake¹⁶. Moreover, GLP-1 RAs can cross the blood–brain barrier and target key hypothalamic nuclei involved in energy homeostasis^17–19. By activating peripheral and central GLP-1 receptors, they modulate appetite-related neurohormonal signals—including insulin and leptin—and regulate adipocyte metabolism, thereby contributing to sustained weight loss^20,21. GLP-1 RAs may also improve liver function, this effect is thought to involve activation of AMP-activated protein kinase (AMPK), a key metabolic regulator that inhibits fatty acid synthesis and promotes fatty acid β-oxidation, leading to decreased lipid accumulation in hepatocytes^22–25.

Despite these benefits, our findings also suggest an increased risk of ketosis in individuals with T1DM treated with GLP-1 RAs²⁶. This might be due to reduced or interrupted insulin doses, gastrointestinal AEs (e.g., vomiting, dehydration), and decreased appetite leading to reduced food intake²⁷. Although meta-analytical data on ketosis in non-T1DM populations are lacking, pharmacovigilance analyses from the FDA Adverse Event Reporting System (FAERS) have indicated a potential signal of ketosis risk in individuals without T1DM receiving GLP-1 RAs without concurrent SGLT2 inhibitors or insulin therapy²⁷. Furthermore, our review identified an elevated risk of diabetic retinopathy with albiglutide at a 30 mg dose in individuals with T2DM²⁸. However, this association was not observed at the 50 mg dose, and the analysis was based on only two trials with limited sample sizes. No consistent association was identified for other GLP-1 RAs^28–31, and therefore, this conclusion should be interpreted cautiously.

GLP-1 RAs have shown a trend toward cardioprotective effects, including reductions in the risk of major cardiovascular events—such as stroke, MI, HF, peripheral artery disease, MACE, CVD, and cardiovascular mortality. In addition, potential improvements have been reported in several indicators of cardiac function, including revascularization, LVEF, SBP, and LDL. The cardioprotective mechanisms of GLP-1 RAs are thought to involve a combination of anti-inflammatory and antioxidant actions, reduced secretion of atherogenic lipoproteins, improved vascular endothelial function, and blood pressure regulation^32,33. GLP-1 receptors are widely expressed in myocardial and vascular systems³⁴, and activation of these receptors may reduce cardiovascular risk by enhancing endothelial nitric oxide production^35,36, modulating vascular tone, and reducing epicardial adipose tissue volume^37,38. Furthermore, GLP-1 RAs mitigate hemodynamic burden associated with individuals with obesity and hypertension, contributing to their overall cardiovascular benefit³². In terms of blood pressure management, GLP-1 RAs modulate the renin-angiotensin-aldosterone system (RAAS), reduce vascular resistance, and improve vascular relaxation, effectively lowering SBP^39–41. Additionally, GLP-1 RAs may exert antihypertensive effects through reduced carotid body excitability and inhibition of sympathetic nerve activation⁴². Notably, the cardiovascular functional improvement induced by GLP-1 RAs appears to be independent of weight reduction. A recent perspective article highlighted that semaglutide demonstrated cardiovascular protective effects in individuals with obesity even before significant weight loss, suggesting that GLP-1 RAs may exert cardioprotection through weight-independent mechanisms³³.

Beyond endocrine and cardiovascular systems, GLP-1 RAs showed potential beneficial trends across multiple disease systems. In oncology, although most meta-analyses reported no significant association between GLP-1 RAs and overall cancer risk, a possible reduced risk of prostate neoplasm was observed in individuals with T2DM. Regarding renal outcomes, GLP-1 RAs showed trends toward improvements in both renal damage indicators and renal composite endpoints, suggesting potential renoprotective effects across the continuum of chronic kidney disease. Regarding respiratory outcomes, while most findings were not statistically significant, GLP-1 RAs were associated with benefits in specific clinical scenarios, such as reducing the risk of pneumonia and improving outcomes in individuals with COVID-19, possibly mediated by anti-inflammatory mechanisms. Evidence also supports their neuroprotective potential, with improved cognitive function in individuals with AD or MCI, and motor symptom relief in individuals with Parkinson’s disease. Notably, GLP-1 RAs were also associated with a lower risk of bone fractures and an increased incidence of headache in women with PCOS. The potential biological mechanisms underlying these associations are discussed in Supplementary Materials 3.

Overall, GLP-1 RAs showed a trend toward a reduction in all-cause mortality. However, a potential increase in the risk of AEs was also noted, including a higher incidence of overall AEs, discontinuation due to AEs, hypoglycemia, and gastrointestinal complications. The most frequently reported AEs were gastrointestinal in nature, with nausea, vomiting, and diarrhea being the most common. These symptoms are a leading cause of treatment discontinuation and are likely related to the delayed gastric emptying induced by GLP-1 RAs^43,44. While delayed gastric emptying increases satiety and reduces caloric intake, excessive delay can result in gastric content retention, leading to upper abdominal discomfort. Prolonged gastric transit time may contribute to bile stasis, thereby increasing the risk of gallbladder-related disorders⁴⁵. In addition, nausea is believed to result mainly from central mechanisms, in which activation of GLP-1R neurons in the area postrema plays a predominant role rather than from delayed gastric emptying alone⁴⁶. As glucose-lowering agents, GLP-1 RAs stimulate insulin secretion in a glucose-dependent manner, theoretically minimizing the risk of hypoglycemia. However, our findings indicated a potential increase in hypoglycemic events. Given that some included studies involved populations with diabetes, this increased risk may be related to prior or concomitant use of other glucose-lowering medications rather than an intrinsic effect of GLP-1 RAs. Therefore, in clinical practice, close monitoring of blood glucose levels is recommended for individuals with diabetes, particularly those receiving combination antidiabetic therapy or during the early phase of dose adjustment. Additional discussions, including comparisons with FDA drug labels and major clinical guidelines, are presented in Supplementary Materials 4.

Beyond statistical significance on isolated endpoints, findings from this umbrella review indicate a shift in the management of metabolic diseases from single-metric control to an integrated, multi-organ and multi-risk-factor strategy. The therapeutic role of GLP-1RAs is extending from glucose lowering alone to a broader, multi-organ intervention. Their clinical value lies not in any single endpoint, but in the capacity to concurrently address the cardiovascular, renal, and metabolic risk factors that commonly coexist in complex individuals. A longstanding unmet need in the care of T2DM and obesity has been to identify therapies that deliver effective glycemic control and clinically meaningful weight loss without increasing cardiovascular risk or severe hypoglycemia. Our umbrella evidence synthesis and grading suggest that GLP-1RAs may partially meet this need. For glycemic control, the mean reduction in HbA1c across most T2DM subgroups approaches or reaches the commonly referenced threshold for clinical relevance of 0.5%⁴⁷, and in key target populations (e.g., T2DM with obesity) can reach 0.8–1.0%, indicating clinically perceptible improvements. For weight management, the proportion of individuals with overweight or obesity achieving ≥5% weight loss increases (approximately equivalent to 2–5 kg, depending on baseline weight), a threshold associated with improvements in blood pressure, blood lipids, functional capacity, and health-related quality of life⁴⁸. Concomitantly, mean systolic blood pressure declines by 3–5 mmHg, with modest LDL improvements, providing biologically plausible support for downstream cardiovascular risk reduction.

Regarding organ outcomes, our review indicates that, in high-risk populations, GLP-1RAs are associated with a 12-15% relative risk reduction in MACE and favorable effects on composite renal endpoints. In selected settings (e.g., AMI or T2DM with CAD), an increase of 4–5% in LVEF has also been observed. Collectively, these threshold-attaining and multidimensionally consistent signals provide a foundation for translating biomarker improvements into clinically perceptible benefits in specific populations. Safety and treatment persistence must be considered in parallel. Although our synthesis shows increased gastrointestinal AEs and treatment discontinuation related to AEs, these challenges can generally be managed in practice through low-dose initiation, gradual up-titration, and regular monitoring⁴⁹. In line with current ADA/EASD guidance, GLP-1 RAs may be prioritized for people with T2DM who have established ASCVD or are at high renal risk; among individuals with overweight or obesity without diabetes, the combination of ≥5% weight loss with modest systolic blood pressure reduction suggests potential multidimensional net benefit, provided that tolerability management and discontinuation-risk counseling are undertaken concurrently. Given that the overall body of evidence in this review is predominantly of low-to-moderate quality and includes a proportion of non-significant findings, we recommend cautious clinical application, with stratified decision-making based on population characteristics and concomitant therapies, to avoid extrapolation beyond the evidentiary boundaries.

The methodological limitations identified by AMSTAR 2 indicate that some pooled effect-size estimates may be vulnerable to selection and reporting bias. In particular, incomplete retrieval of grey literature and insufficient documentation of study-exclusion decisions could lead to overestimation of benefits or underdetection of harms, while the infrequent propagation of primary-study bias into meta-analytic models reduces confidence in the magnitude of pooled effects. Future meta-analyses should report excluded studies with justifications, incorporate comprehensive grey-literature and expert searches, preregister protocols, and explicitly propagate primary-study risk of bias into quantitative synthesis to strengthen the reliability and generalisability of conclusions. Although our CCA showed generally low overlap across reviews—mitigating double-counting—evidence aggregation was disproportionately driven by a small number of large, frequently cited cardiovascular outcome trials (CVOTs; e.g., LEADER, SUSTAIN-6, EXSCEL, HARMONY Outcomes, REWIND, ELIXA, AMPLITUDE-O, SELECT) and other large-scale clinical programs (e.g., PIONEER, AWARD, LEAD, SCALE, STEP). Such dominance of a few large-scale trials may constrain the external validity of pooled estimates and obscure underlying heterogeneity across smaller studies. Accordingly, while the direction of effects appears broadly consistent across systems, the precision of the estimated magnitudes should be interpreted with caution.

Although this umbrella review attempted to comprehensively summarize the current published meta-analyses and analyze the related health outcomes, some emerging research topics, due to being in the experimental phase and lacking support from multiple primary studies, may have been overlooked. For instance, a recent cohort study⁵⁰ on the use of GLP-1 RAs prior to upper gastrointestinal endoscopy showed that while GLP-1 RAs do not increase the risk of aspiration pneumonia, they may be associated with an increased risk of endoscopy interruption. In a related context, a large retrospective cohort study using real-world surgical data found no significant association between preoperative GLP-1 RA use and 30-day postoperative aspiration pneumonia⁵¹. These findings suggest that while GLP-1 RAs may influence gastric motility and procedural tolerance, their impact on clinically meaningful respiratory complications remains unconfirmed. Further research is warranted to clarify perioperative management guidelines in this setting. The potential association between GLP-1 RAs and suicidality has raised increasing concern following regulatory safety warnings. While clinical evidence^52,53 has indicated a possible elevated risk, causality has not been established. A disproportionality analysis using global pharmacovigilance data identified a significant signal for semaglutide-associated suicidal ideation, particularly among individuals concurrently using antidepressants or benzodiazepines⁵⁴. In contrast, a recent meta-analysis⁵⁵ synthesizing randomized and observational data did not find a significant association between GLP-1 RAs and suicide or self-harm outcomes. Similarly, a target trial emulation using U.S. real-world data in older adults with T2DM reported no clear excess risk compared to other antidiabetic agents. However, the available clinical evidence was limited by substantial heterogeneity, low event rates, and imprecise estimates⁵⁶. Notably, a recent large-scale cohort study by Xie and colleagues using U.S. Department of Veterans Affairs data mapped associations between GLP-1 RA use and 175 health outcomes, providing comprehensive real-world insights to inform future investigations⁵⁷. Overall, the conclusions of this umbrella review are largely consistent with those reported in that study, though differences exist; for example, their analysis identified a significant increase in the risk of drug-induced pancreatitis, whereas our review did not find a statistically significant association. Such discrepancies may reflect differences in populations, outcome definitions, or statistical power.

Despite the comprehensive analysis of the effects of GLP-1 RAs on various health outcomes and AEs, this umbrella review has several limitations. First, differences in design, individual characteristics, and follow-up duration among the included meta-analyses may contribute to heterogeneity in the results, particularly in the context of patients’ prior medication history. Although individual primary studies indicated that baseline characteristics do not significantly affect the results, combining outcomes from multiple primary studies in meta-analyses could be influenced by these differences. Additionally, this umbrella review focused only on comparisons between GLP-1 RAs and placebos or non-GLP-1 RA drugs, without directly comparing the efficacy among different GLP-1 RAs. Although this umbrella review primarily focused on statistically significant outcomes, several evaluated outcomes did not reach statistical significance in certain individual populations. For outcomes with significant findings in some groups, the direction of effects was consistent, although statistical significance was not achieved in all populations. However, for outcomes that were non-significant across all analyses, current evidence remains inconclusive. These findings should be interpreted cautiously, and further high-quality clinical research is needed to confirm these associations in diverse populations.

Method

Umbrella review methods

This umbrella review follows the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and is prospectively registered with PROSPERO (CRD42024585231).

Literature search

Data were systematically searched across five databases: PubMed, Embase, Web of Science, the Cochrane Database of Systematic Reviews, and Epistemonikos, with the search conducted up to August 26, 2024. The search strategy combined subject headings and free-text terms related to GLP-1 RAs. Representative terms included: (“GLP-1 RAs” OR “liraglutide” OR “semaglutide” OR “exenatide” OR “dulaglutide” OR “albiglutide”) AND (“systematic review” OR “meta-analys*”). The complete search strategies tailored for each database are provided in Supplementary Materials 1. All retrieved articles were imported into EndNote X9 for deduplication and subsequent screening. F.K. and Y.Z. independently reviewed the titles, abstracts, and full texts of the articles according to the inclusion criteria, and any disagreements were resolved through discussion with W.Z. and X.W. Additionally, the reference lists of all included meta-analyses were manually searched to ensure that no relevant meta-analyses were overlooked.

Eligibility criteria

Inclusion criteria were summarized according to the PICOS principle: (1) P (Participants): No restriction on diseases; but participants in each meta-analysis must have the same condition; (2) I (Intervention): GLP-1 RAs as a whole or as a single drug; (3) C (Comparison): Comparators included placebo alone, active non-GLP-1 RA drugs alone, or both placebo and active comparators; (4) O (Outcomes): Dichotomous and/or continuous variables; (5) S (Study Design): Meta-analyses of RCTs or observational studies. Network meta-analyses were also eligible if they provided direct pairwise comparisons between GLP-1 RAs and relevant control groups.

For meta-analyses reporting multiple drugs or outcomes, data were extracted separately. If meta-analyses were identical in type, intervention, participants, and outcomes, selection was based on: (1) publication year (most recent if >24 months apart); (2) number of included trials (highest if ≤24 months apart); (3) AMSTAR 2 score (highest if trials were identical); (4) evidence quality and population size if AMSTAR 2 scores were identical⁵⁸.

Exclusion criteria were: (1) non-English publications; (2) non-clinical research; (3) meta-analyses with heterogeneous participant conditions or GLP-1 RAs in the control group; (4) systematic reviews without meta-analyses; (5) network meta-analyses reporting only indirect comparisons.

Data extraction

F.K., Y.Z., and W.Z. independently extracted the following information from the included meta-analyses: first author, publication year, intervention type, study population, study design, number of included trials, number of participants, outcomes, effect model, effect size, heterogeneity, and publication bias. Any discrepancies were resolved by discussion between X.W. and T.W.

Evaluation of primary study overlap

To assess the potential risk of bias due to overlapping primary studies across the included meta-analyses, outcome indicators were first categorized by clinical system, and each outcome was assigned to a single system category.

For each system, a citation matrix was constructed, with rows representing unique primary studies and columns representing the included meta-analyses. Based on these matrices, the Corrected Covered Area (CCA) was calculated using the following formula:

$$CCA=\frac{\mathrm{N}-r}{\mathrm{r}\left(c-1\right)}$$

Where N is the total number of primary study occurrences across all meta-analyses, r is the number of unique primary studies, and c is the number of meta-analyses. The degree of overlap was interpreted as slight (≤5%), moderate (6-10%), high (11-15%), or very high (>15%). All calculations were performed using Microsoft Excel 2019.

Quality assessment of methods and evidence

The methodological quality of the included meta-analyses was assessed using the AMSTAR 2 tool⁵⁹. The GRADE approach was employed to systematically assess the certainty of evidence for each outcome. GRADE categorizes evidence quality into four levels: high, moderate, low, and very low⁶⁰. Finally, the certainty of the outcomes was categorized into four classes based on evidence classification criteria: Class I (convincing evidence), Class II (highly suggestive evidence), Class III (suggestive evidence), Class IV (weak evidence), and NS (non-significant)^58,61.

Data analysis

Given the broad range of conditions and outcomes assessed, all outcome indicators were categorized a priori by clinical system to facilitate a structured synthesis and interpretation. To further account for heterogeneity in the populations, interventions, and outcome definitions reported in the included meta-analyses, we performed a stratified synthesis. Specifically, outcomes were grouped and presented by clinical domain, individual population, and intervention type, as reported in the original meta-analyses.

Where meta-analysis-level data were available, we extracted effect sizes and 95% confidence intervals (CIs) and recalculated pooled estimates using a random-effects model. If unavailable, original pooled results were retained. Statistical significance was set at p < 0.05. Heterogeneity was assessed using I² and Cochran’s Q test and interpreted as low (I² ≤ 40%), moderate (30% <I² ≤ 60%), or high (I² ≥ 50%). All analyses and visualizations were conducted using R (v4.4.1) and RevMan (v5.3).

Ethics statement

Not needed.

Supplementary Information

The supplementary materials can be found in “Supplementary information.pdf”.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

F.J.K., Y.R.Z., W.M.Z., X.Y.W., and Z.Y.Z. conducted study selection. F.J.K., Y.R.Z., W.M.Z., X.Y.W., and T.Y.W. performed data extraction and analysis. Y.R.Z., W.M.Z., X.Y.W., and T.Y.W. prepared statistical analysis and quality assessment. F.J.K., Y.R.Z., W.M.Z., X.Y.W., and T.Y.W. drafted and revised the manuscript. F.J.K., Y.X., L.N.X., and T.S. designed the study, supervised the project, and revised the manuscript. All authors reviewed and approved the final version of the manuscript. T.S. is the guarantor.

Peer review

Peer review information

Nature Communications thanks Sheyu Li, Dario Giugliano, Sophie Beese and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The authors declare that the data supporting the findings of this study are available within the paper and its Supplementary Information files.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-67701-9.