Abstract
Aims
The association between glucagon‐like peptide‐1 receptor agonists (GLP‐1 RAs) and depression remains uncertain due to contradictory evidence. We compared the risk of incident depression between GLP‐1 RAs and sodium‐glucose cotransporter‐2 inhibitors (SGLT2is) in overweight or obese adults with type 2 diabetes.
Materials and Methods
We conducted a new‐user, active‐comparator cohort study using a deidentified electronic health record network from January 2016 to July 2024. After 1:1 propensity score matching, we compared 25 704 new GLP‐1 RA users to 25 704 SGLT2i users with newly diagnosed type 2 diabetes and overweight/obesity, excluding those with prior mood disorders. The primary outcome was a composite of incident depression diagnosis or antidepressant initiation, assessed from 1 month to 1 year post‐initiation using Cox models and time‐varying analyses.
Results
In 51 408 patients (mean age 56.8 years, 48.9% male), GLP‐1 RA use was associated with higher depression incidence versus SGLT2i use (17.0% vs. 14.8%; hazard ratio 1.09, 95% CI 1.04–1.14; p < 0.001), with an absolute risk difference of 2.2%. The association was stronger in adults ≥65 years (HR 1.15) and plateaued after approximately 6 months. In secondary analysis, GLP‐1 RA use was associated with a lower rate of all‐cause mortality (HR 0.74, 95% CI 0.63–0.88).
Conclusions
GLP‐1 RA initiation was associated with a statistically significant increase in depression risk compared to SGLT2i use (9% relative increase, 2.2% absolute risk difference over 1 year), particularly during the subacute period and in older adults. This observed association must be balanced against substantial mortality benefits. Enhanced monitoring and shared decision‐making are warranted.
Keywords: depression, GLP‐1 receptor agonists, neuropsychiatric adverse effects, pharmacovigilance, SGLT2 inhibitors, type 2 diabetes
1. INTRODUCTION
Glucagon‐like peptide‐1 receptor agonists (GLP‐1 RAs) have become a cornerstone in treating type 2 diabetes (T2D) and obesity, 1 with their application rapidly expanding. 2 , 3 , 4 , 5 However, this clinical success is shadowed by persistent concerns over potential neuropsychiatric side effects, 6 particularly depression and suicidality. 7 , 8 Current evidence from case series and large‐scale studies is contradictory, 9 , 10 , 11 leaving clinicians uncertain about how to assess this risk when counselling patients.
The inconsistency in prior research likely results from key methodological issues. Some studies suffer from confounding by indication, comparing GLP‐1 RA users to non‐users who inherently have different risk profiles. 7 Others are limited by using claims data lacking crucial confounders like BMI or laboratory data, 10 restricting populations to specific age groups, 12 or using inconsistent definitions of psychiatric outcomes. 13 The decision to focus on incident depression, despite its high comorbidity with anxiety, 14 is driven by a specific and biologically plausible rationale. While regulatory discussions have often centred on suicidality, 15 , 16 emerging reports of anhedonia suggest a more focused avenue for inquiry. 17 This is because anhedonia is a cardinal criterion for major depression and has a direct neurobiological link to the brain's reward circuits, which is precisely where the GLP‐1 receptors under investigation are expressed. 18 Furthermore, from a clinical standpoint, it serves as a more definitive endpoint, especially since persistent anxiety can often be a precursor to a depressive episode. 19
To resolve this clinical uncertainty, we employed a rigorous new‐user, active‐comparator design, comparing new users of GLP‐1 RAs to those initiating sodium‐glucose cotransporter‐2 inhibitor (SGLT2i). Both drug classes are indicated for similar populations of adults with type 2 diabetes and overweight or obesity, 20 , 21 and preclinical animal studies have suggested that each may influence mood. 22 , 23 Moreover, even small increases in risk from these widely used therapies could have substantial public health implications. Therefore, the primary goal of this study is to rigorously quantify the comparative risk of incident depression between new users of GLP‐1 RAs and SGLT2is, and to translate these findings into clinically relevant guidance. We aim to provide evidence that can inform monitoring approaches and patient‐provider discussions about risk–benefit considerations in contemporary cardiometabolic care.
2. MATERIALS AND METHODS
2.1. Study design and data source
This retrospective cohort study was conducted using the TriNetX research network, a federated platform that provides access to de‐identified electronic health records (EHRs) from a broad range of healthcare organizations. These organizations are predominantly hospital systems in the United States. A key advantage of this database is its inclusion of comprehensive laboratory data, which is crucial for detailed covariate adjustment. The study protocol received approval from the Institutional Review Board of Chung Shan Medical University (Case No. CS2‐25111). We employed a new‐user, active‐comparator design to minimize selection bias and confounding by indication. This study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (Method S1).
2.2. Patient selection and cohort definition
We identified a cohort of patients from the TriNetX database between January 1, 2016, and July 1, 2024. The study population included adults (aged ≥18 years) with a new‐onset diagnosis of T2D and concurrent overweight or obesity. To maximize sensitivity in identifying potential cases, a new‐onset T2D diagnosis was defined as the first occurrence of either an International Classification of Diseases, Tenth Revision, Clinical Modification (ICD‐10‐CM) code for T2D (E11) or a haemoglobin A1c (HbA1c) value of 6.5% or greater. Similarly, overweight or obesity was defined by the presence of an ICD‐10‐CM code for obesity (E66) or a recorded body mass index (BMI) of 25 kg/m2 or higher. Patients were excluded if they had a prior history of mood disorders (defined by ICD‐10‐CM codes F30–F39) or any record of prescription for antidepressants.
The exposure group comprised new users of GLP‐1 receptor agonists (GLP‐1 RAs), and the active‐comparator group comprised new users of sodium‐glucose cotransporter‐2 inhibitors (SGLT2is). To ensure that cohorts consisted of treatment‐naive patients, a key criterion for inclusion was that the initiating prescription must be the patient's first‐ever recorded prescription for any glucose‐lowering medication. The date of this first prescription was defined as the index date. The specific qualifying medications included tirzepatide, semaglutide, or liraglutide for the GLP‐1 RA cohort, and canagliflozin, dapagliflozin, or empagliflozin for the SGLT2i cohort. This initial prescription must have been issued within a 3‐month grace period following the patient's qualifying T2D diagnosis. Finally, to ensure cohort purity at baseline, a mutual exclusivity criterion was enforced: patients in the GLP‐1 RA cohort could not have received a prescription for an SGLT2i, and vice versa, within the first 3 months of follow‐up.
2.3. Covariate assessment
To reduce confounding and approximate the effects of randomization, we conducted 1:1 propensity score matching (PSM). A comprehensive set of baseline covariates, measured in the year prior to the index date, was included in the matching model. These covariates encompassed demographics (age, sex, race, ethnicity), proxies for social status and medication compliance, pre‐existing psychiatric and physical comorbidities (e.g., chronic kidney disease, cardiovascular disease, liver disease), concomitant medication use (e.g., non‐antidepressant psychotropics, antiarrhythmics, corticosteroids), healthcare utilization patterns, and key laboratory values, including stratified BMI, HbA1c, triglycerides, and LDL cholesterol. Details of the covariates and corresponding codes are provided in Table S1.
2.4. Outcomes
The primary outcome was a composite of incident depression, defined by a new diagnosis with specific ICD‐10‐CM codes (F32, F33.0, F33.1, F33.2, F33.8, excluding codes for remission), or the initiation of a new antidepressant medication (ATC code N06A). This composite definition was chosen to enhance sensitivity by capturing both formally diagnosed depression and cases where depression was treated without a formal diagnosis code being entered. The outcome was assessed in the follow‐up period from 1 month to 1 year after the index date. A one‐month lag period was implemented to reduce the risk of protopathic bias, as the diagnostic criteria for a major depressive episode in the DSM‐5‐TR require symptoms to be present for at least 2 weeks. Due to the limitations of performing a formal competing risk analysis for mortality on the TriNetX platform, all‐cause mortality was analysed separately, as well as in combination with incident depression as a composite secondary outcome. Additionally, depression diagnosis and antidepressant use were also analysed separately as secondary outcomes.
2.5. Statistical analysis
After PSM, baseline characteristics of the matched cohorts were compared to assess balance. The primary analysis compared the one‐year risk of incident depression between the two groups. Patients were censored at their last recorded fact. We calculated hazard ratios (HRs) with 95% confidence intervals (CIs) using Cox proportional hazards models and visualized probabilities with cumulative incidence curves, with differences assessed using the log‐rank test. The proportional hazards assumption was evaluated. We also calculated the E‐value to estimate the potential impact of unmeasured confounding. 24 An E‐value quantifies the minimum strength of association that an unmeasured confounder would need to have with both the exposure and the outcome to fully explain away the observed effect. To obtain an interpretable effect measure from aggregated survival data, we reconstructed individual patient data (IPD) with the IPDfromKM package from Kaplan–Meier curves, 25 from which we estimated the restricted mean survival time (RMST) and summarized between‐group effects using the RMST difference. 26 To visualize departures from proportional hazards, we fitted flexible spline‐based models (Royston–Parmar) to derive time‐varying hazard ratios (details in Methods S2). 27 Missing data were handled by TriNetX without imputation.
A series of prespecified sensitivity analyses were conducted to assess the robustness of our findings. One key sensitivity analysis, designed to more closely emulate a target trial, relaxed the mutual exclusivity requirement. In this analysis, patients were assigned to a cohort based on their initial prescription and were analysed within that cohort regardless of subsequent medication changes in the first 3 months. Other sensitivity analyses included: using landmark analyses with different follow‐up windows; requiring a prior healthcare visit; altering the cohort entry date to 2014; shortening the grace period to 2 months; and including patients with T2D regardless of weight status. We also performed analyses in the unmatched cohort and implemented negative‐control outcome analyses. Finally, we conducted additional analyses on September 12, 2025: one excluding patients with rare in our cohort but recognized contraindications, including history of end‐stage renal disease (ESRD), dialysis, or pregnancy within 1 year before the T2D and/or obesity diagnosis (ICD‐10‐CM codes N18.6, Z99.2, Z49, O00–O9A, Z3A, Z33, Z34, Z36; Current Procedural Terminology code 1012740). Another included a commonly prescribed comparator group with a therapeutic profile similar to that of SGLT2 inhibitors: patients initiated on dipeptidyl peptidase‐4 (DPP‐4) inhibitors (ATC code A10BH). Subgroup analyses were performed based on age (≥65 vs. <65), sex, prior psychiatric history, and specific GLP‐1 RA agent, with interaction p‐values estimated from aggregate data. All analyses were conducted using R software with the Evalue, survival, rstpm2, survRM2, and metafor packages.
3. RESULTS
3.1. Baseline characteristics
From an initial pool of 148,793,412 patients in the TriNetX research network, we identified 5,557,771 adults (age ≥ 18 years) with a new diagnosis of type 2 diabetes and overweight or obesity between January 1, 2016, and July 1, 2024. After applying exclusion criteria, including prior use of diabetes medications and a history of mood disorders or antidepressant use, we identified 31,056 new users of GLP‐1 RAs and 50,847 new users of SGLT2is who initiated treatment within 3 months of their qualifying diagnosis. Following 1:1 propensity score matching, two well‐balanced cohorts of 25 704 patients each were established for the primary analysis (Figure 1).
FIGURE 1.
Flowchart of patient selection for the study cohort. GLP‐1 RA, Glucagon‐like peptide‐1 receptor agonist; SGLT2i, sodium‐glucose cotransporter‐2 inhibitor; T2D, Type 2 diabetes mellitus.
Prior to matching, significant differences were observed between the two groups. GLP‐1 RA users were older on average (mean age 60.1 vs. 54.6 years), more likely to be male (62.1% vs. 43.4%), and had a substantially higher baseline BMI (37.1 ± 8.12 vs. 32.7 ± 6.71). Furthermore, they had a higher prevalence of numerous cardiometabolic comorbidities. After propensity score matching, the baseline demographic and clinical characteristics were well balanced between the GLP‐1 RA and SGLT2i cohorts. All absolute standardized mean differences (aSMDs) for the measured covariates were well below the conventional 0.1 threshold (Table 1 and love plot in Figure S1). In the matched cohort, the mean age was approximately 56.8 years, and the groups were nearly evenly split by sex (48.6% male in GLP‐1 RA vs. 49.2% in SGLT2i). The follow‐up duration was comparable, with both groups having a median follow‐up of 365 days. The mean follow‐up was also similar, at 291.8 days for the GLP‐1 RA cohort and 274.3 days for the SGLT2i cohort (Table S2).
TABLE 1.
Baseline demographic, clinical, and laboratory characteristics of new users of GLP‐1 RAs and SGLT2is, before and after 1:1 propensity score matching.
| Before matching | After matching | |||||
|---|---|---|---|---|---|---|
| SGLT2i (%) | GLP1 RA (%) | aSMD | SGLT2i (%) | GLP1 RA (%) | aSMD | |
| Demographics | 30 156 (100) | 50 874 (100) | 25 704 (100) | 25 704 (100) | ||
| Age at Index | 54.6 ± 13.5 | 60.1 ± 12.5 | 0.420 | 56.8 ± 12.7 | 56.7 ± 12.5 | 0.002 |
| Male | 13 491 (43.4) | 31 594 (62.1) | 0.381 | 12 653 (49.2) | 12 491 (48.6) | 0.013 |
| Female | 16 554 (53.3) | 17 849 (35.1) | 0.373 | 12 222 (47.5) | 12 364 (48.1) | 0.011 |
| White | 18 557 (59.8) | 28 610 (56.2) | 0.071 | 15 433 (60.0) | 15 314 (59.6) | 0.009 |
| American Indian or Alaska Native | 259 (0.8) | 225 (0.4) | 0.049 | 155 (0.6) | 171 (0.7) | 0.008 |
| Native Hawaiian or Other Pacific Islander | 270 (0.9) | 419 (0.8) | 0.005 | 206 (0.8) | 213 (0.8) | 0.003 |
| Black or African American | 5600 (18.0) | 7637 (15.0) | 0.081 | 4281 (16.7) | 4346 (16.9) | 0.007 |
| Asian | 1172 (3.8) | 4649 (9.1) | 0.22 | 1136 (4.4) | 1142 (4.4) | 0.001 |
| Not Hispanic or Latino | 18 625 (60.0) | 31 579 (62.1) | 0.043 | 15 309 (59.6) | 15 304 (59.5) | <0.001 |
| Hispanic or Latino | 2502 (8.1) | 4686 (9.2) | 0.041 | 2189 (8.5) | 2229 (8.7) | 0.006 |
| Diagnosis | ||||||
| F40‐F48 (Anxiety, dissociative, stress‐related, somatoform and other nonpsychotic mental disorders) | 1863 (6.0) | 2719 (5.3) | 0.028 | 1237 (4.8) | 1257 (4.9) | 0.004 |
| F90‐F98 (Behavioural and emotional disorders with onset usually occurring in childhood and adolescence) | 201 (0.6) | 156 (0.3) | 0.049 | 103 (0.4) | 94 (0.4) | 0.006 |
| F10‐F19 (Mental and behavioural disorders due to psychoactive substance use) | 1388 (4.5) | 4039 (7.9) | 0.144 | 1188 (4.6) | 1198 (4.7) | 0.002 |
| F60‐F69 (Disorders of adult personality and behaviour) | 63 (0.2) | 88 (0.2) | 0.007 | 44 (0.2) | 40 (0.2) | 0.004 |
| F20‐F29 (Schizophrenia, schizotypal, delusional, and other non‐mood psychotic disorders) | 97 (0.3) | 286 (0.6) | 0.038 | 86 (0.3) | 83 (0.3) | 0.002 |
| F01‐F09 (Mental disorders due to known physiological conditions) | 153 (0.5) | 678 (1.3) | 0.088 | 138 (0.5) | 151 (0.6) | 0.007 |
| F50‐F59 (Behavioural syndromes associated with physiological disturbances and physical factors) | 228 (0.7) | 208 (0.4) | 0.043 | 134 (0.5) | 137 (0.5) | 0.002 |
| F80‐F89 (Pervasive and specific developmental disorders) | 36 (0.1) | 64 (0.1) | 0.003 | 26 (0.1) | 22 (0.1) | 0.005 |
| Z55‐Z65 (Persons with potential health hazards related to socioeconomic and psychosocial circumstances) | 222 (0.7) | 768 (1.5) | 0.076 | 182 (0.7) | 176 (0.7) | 0.003 |
| I20‐I25 (Ischemic heart diseases) | 2083 (6.7) | 10 319 (20.3) | 0.405 | 1993 (7.8) | 1967 (7.7) | 0.004 |
| I11 (Hypertensive heart disease) | 525 (1.7) | 4850 (9.5) | 0.346 | 513 (2.0) | 502 (2.0) | 0.003 |
| I60‐I69 (Cerebrovascular diseases) | 807 (2.6) | 3027 (5.9) | 0.166 | 774 (3.0) | 790 (3.1) | 0.004 |
| J40‐J4A (Chronic lower respiratory diseases) | 1822 (5.9) | 3878 (7.6) | 0.07 | 1387 (5.4) | 1373 (5.3) | 0.002 |
| R45 (Symptoms and signs involving emotional state) | 135 (0.4) | 382 (0.8) | 0.041 | 110 (0.4) | 107 (0.4) | 0.002 |
| E78 (Disorders of lipoprotein metabolism and other lipidemias) | 7593 (24.4) | 15 625 (30.7) | 0.14 | 6181 (24.0) | 6126 (23.8) | 0.005 |
| K70‐K77 (Diseases of liver) | 1016 (3.3) | 2223 (4.4) | 0.057 | 790 (3.1) | 788 (3.1) | <0.001 |
| Z91.1 (Patient's noncompliance with medical treatment and regimen) | 285 (0.9) | 1267 (2.5) | 0.122 | 260 (1.0) | 254 (1.0) | 0.002 |
| I42 (Cardiomyopathy) | 234 (0.8) | 3247 (6.4) | 0.307 | 232 (0.9) | 266 (1.0) | 0.014 |
| I50 (Heart failure) | 1041 (3.4) | 8668 (17.0) | 0.464 | 1022 (4.0) | 1094 (4.3) | 0.014 |
| I48 (Atrial fibrillation and flutter) | 875 (2.8) | 4518 (8.9) | 0.261 | 823 (3.2) | 848 (3.3) | 0.005 |
| I12 (Hypertensive chronic kidney disease) | 598 (1.9) | 1268 (2.5) | 0.039 | 494 (1.9) | 489 (1.9) | 0.001 |
| I13 (Hypertensive heart and chronic kidney disease) | 253 (0.8) | 2154 (4.2) | 0.219 | 242 (0.9) | 272 (1.1) | 0.012 |
| N18.3 (Chronic kidney disease, Stage 3 (moderate)) | 648 (2.1) | 2240 (4.4) | 0.131 | 579 (2.3) | 583 (2.3) | 0.001 |
| N18.4 (Chronic kidney disease, stage 4 (severe)) | 179 (0.6) | 444 (0.9) | 0.035 | 145 (0.6) | 139 (0.5) | 0.003 |
| N18.9 (Chronic kidney disease, unspecified) | 526 (1.7) | 1924 (3.8) | 0.128 | 463 (1.8) | 463 (1.8) | <0.001 |
| N18.6 (End stage renal disease) | 254 (0.8) | 164 (0.3) | 0.066 | 126 (0.5) | 114 (0.4) | 0.007 |
| N18.2 (Chronic kidney disease, stage 2 (mild)) | 133 (0.4) | 446 (0.9) | 0.056 | 113 (0.4) | 105 (0.4) | 0.005 |
| N18.5 (Chronic kidney disease, stage 5) | 55 (0.2) | 57 (0.1) | 0.017 | 31 (0.1) | 31 (0.1) | <0.001 |
| E11.5 (Type 2 diabetes mellitus with circulatory complications) | 617 (2.0) | 1802 (3.5) | 0.095 | 536 (2.1) | 551 (2.1) | 0.004 |
| E11.2 (Type 2 diabetes mellitus with kidney complications) | 1285 (4.1) | 3827 (7.5) | 0.145 | 1078 (4.2) | 1105 (4.3) | 0.005 |
| E11.6 (Type 2 diabetes mellitus with other specified complications) | 5271 (17.0) | 9713 (19.1) | 0.055 | 4131 (16.1) | 4178 (16.3) | 0.005 |
| E11.4 (Type 2 diabetes mellitus with neurological complications) | 1455 (4.7) | 2687 (5.3) | 0.027 | 1182 (4.6) | 1176 (4.6) | 0.001 |
| E11.9 (Type 2 diabetes mellitus without complications) | 9624 (31.0) | 17 751 (34.9) | 0.083 | 7511 (29.2) | 7451 (29.0) | 0.005 |
| Service | ||||||
| Office or Other Outpatient Services | 8442 (27.2) | 10 193 (20.0) | 0.169 | 5875 (22.9) | 5897 (22.9) | 0.002 |
| Emergency Department Services | 2808 (9.0) | 7389 (14.5) | 0.171 | 2358 (9.2) | 2339 (9.1) | 0.003 |
| Hospital Inpatient and Observation Care Services | 2021 (6.5) | 7990 (15.7) | 0.296 | 1831 (7.1) | 1856 (7.2) | 0.004 |
| Preventive Medicine Services | 1642 (5.3) | 1244 (2.4) | 0.148 | 945 (3.7) | 932 (3.6) | 0.003 |
| Medication | ||||||
| Antianginals | 853 (2.7) | 5428 (10.7) | 0.321 | 836 (3.3) | 847 (3.3) | 0.002 |
| Angiotensin II Inhibitor | 2584 (8.3) | 7203 (14.2) | 0.186 | 2105 (8.2) | 2058 (8.0) | 0.007 |
| Beta Blockers/Related | 3447 (11.1) | 11 923 (23.4) | 0.331 | 3021 (11.8) | 3043 (11.8) | 0.003 |
| Antiarrhythmics | 2687 (8.7) | 8565 (16.8) | 0.247 | 2297 (8.9) | 2313 (9.0) | 0.002 |
| Calcium Channel Blockers | 2607 (8.4) | 7214 (14.2) | 0.184 | 2187 (8.5) | 2153 (8.4) | 0.005 |
| Antihypertensives, Other | 1322 (4.3) | 4284 (8.4) | 0.172 | 1170 (4.6) | 1185 (4.6) | 0.003 |
| ACE Inhibitors | 2868 (9.2) | 5960 (11.7) | 0.081 | 2347 (9.1) | 2361 (9.2) | 0.002 |
| Sedatives/Hypnotics | 2676 (8.6) | 8156 (16.0) | 0.227 | 2296 (8.9) | 2307 (9.0) | 0.001 |
| CNS Stimulants | 383 (1.2) | 379 (0.7) | 0.049 | 233 (0.9) | 230 (0.9) | 0.001 |
| CNS Medications, Other | 1907 (6.1) | 3776 (7.4) | 0.051 | 1527 (5.9) | 1511 (5.9) | 0.003 |
| Antilipemic Agents | 5736 (18.5) | 14 442 (28.4) | 0.236 | 4814 (18.7) | 4787 (18.6) | 0.003 |
| Adrenal Corticosteroids | 3198 (10.3) | 6610 (13.0) | 0.084 | 2507 (9.8) | 2528 (9.8) | 0.003 |
| Analgesics | 6294 (20.3) | 17 529 (34.5) | 0.322 | 5405 (21.0) | 5358 (20.8) | 0.004 |
| Anticonvulsants | 1980 (6.4) | 3951 (7.8) | 0.054 | 1567 (6.1) | 1540 (6.0) | 0.004 |
| Antipsychotics | 540 (1.7) | 1519 (3.0) | 0.082 | 441 (1.7) | 431 (1.7) | 0.003 |
| Laboratory | ||||||
| Haemoglobin A1c/Haemoglobin. total in Blood | 8.57 ± 2.44 | 8.46 ± 2.22 | 0.047 | 8.74 ± 2.39 | 8.76 ± 2.39 | 0.008 |
| 0%–6.50% | 2487 (8.0) | 3415 (6.7) | 0.05 | 1433 (5.6) | 1383 (5.4) | 0.009 |
| 6.50%–8% | 3420 (11.0) | 6836 (13.4) | 0.074 | 2559 (10.0) | 2470 (9.6) | 0.012 |
| 8–9.50% | 1949 (6.3) | 4136 (8.1) | 0.072 | 1619 (6.3) | 1671 (6.5) | 0.008 |
| 9.50–% | 3632 (11.7) | 5408 (10.6) | 0.034 | 2788 (10.8) | 2806 (10.9) | 0.002 |
| Cholesterol in LDL [Mass/volume] in Serum or Plasma | 104 ± 41.6 | 95.8 ± 43 | 0.198 | 102 ± 42.4 | 101 ± 43.1 | 0.018 |
| 0–130 mg/dL | 5835 (18.8) | 10 583 (20.8) | 0.051 | 4298 (16.7) | 4217 (16.4) | 0.008 |
| 130–200 mg/dL | 1774 (5.7) | 2284 (4.5) | 0.056 | 1157 (4.5) | 1178 (4.6) | 0.004 |
| 200–270 mg/dL | 321 (1.0) | 446 (0.9) | 0.016 | 243 (0.9) | 232 (0.9) | 0.004 |
| 270–mg/dL | 68 (0.2) | 137 (0.3) | 0.01 | 55 (0.2) | 54 (0.2) | 0.001 |
| Triglyceride [Mass/volume] in Serum, Plasma or Blood | 204 ± 205 | 200 ± 248 | 0.018 | 211 ± 205 | 222 ± 276 | 0.047 |
| 0–130 mg/dL | 3064 (9.9) | 5765 (11.3) | 0.048 | 2108 (8.2) | 2124 (8.3) | 0.002 |
| 130–200 mg/dL | 2552 (8.2) | 3703 (7.3) | 0.035 | 1802 (7.0) | 1739 (6.8) | 0.01 |
| 200–270 mg/dL | 1292 (4.2) | 1868 (3.7) | 0.025 | 969 (3.8) | 964 (3.8) | 0.001 |
| 270–mg/dL | 1594 (5.1) | 2358 (4.6) | 0.023 | 1252 (4.9) | 1248 (4.9) | 0.001 |
| BMI | 37.1 ± 8.12 | 32.7 ± 6.71 | 0.586 | 35.5 ± 7.4 | 35.2 ± 7.3 | 0.04 |
| 0–30 kg/m2 | 2588 (8.3) | 10 585 (20.8) | 0.359 | 2506 (9.7) | 2482 (9.7) | 0.003 |
| 30–35 kg/m2 | 3614 (11.6) | 8468 (16.6) | 0.144 | 3239 (12.6) | 3208 (12.5) | 0.004 |
| 35–40 kg/m2 | 3227 (10.4) | 4614 (9.1) | 0.045 | 2454 (9.5) | 2459 (9.6) | 0.001 |
| 40–45 kg/m2 | 2074 (6.7) | 2212 (4.3) | 0.102 | 1352 (5.3) | 1343 (5.2) | 0.002 |
| 45–kg/m2 | 2108 (6.8) | 1638 (3.2) | 0.164 | 1104 (4.3) | 1112 (4.3) | 0.002 |
Note: SI conversion factors: To convert HbA1c to proportion of total HbA1c, multiply by 0.01; to convert LDL and total cholesterol to mmol/L, multiply by 0.0259; to convert Triglyceride to mmol/L, multiply by 0.0113.
Abbreviations: aSMD: absolute standardized mean difference; GLP1 RA: Glucagon‐like peptide‐1 Receptor Agonist; SGLT2i: Sodium‐glucose cotransporter‐2 inhibitor; CNS: central nervous system; LDL: low‐density lipoprotein; BMI: body mass index.
3.2. Primary and secondary outcomes
Over a one‐year follow‐up period, the use of GLP‐1 RAs was associated with a significantly higher risk of the primary composite outcome of incident depression compared to the use of SGLT2is (Table 2). The incidence was 17.0% (4359 events) in the GLP‐1 RA group versus 14.8% (3802 events) in the SGLT2i group, yielding a hazard ratio (HR) of 1.09 (95% Confidence Interval [CI], 1.04 to 1.14; p = 0.0001). The absolute risk difference at 1 year was 2.2% (95% CI, 1.5% to 2.8%). E‐values were 1.40 for the point estimate and 1.24 for the lower limit of the confidence interval. The cumulative incidence curves for depression began to diverge after approximately 2 months of follow‐up (Figure 2), with the GLP‐1 RA group showing a consistently higher cumulative incidence throughout the remainder of the year. Quantified by the restricted mean survival time (RMST) analysis, over the one‐year follow‐up, the GLP‐1 RA group experienced 3.3 days fewer on average free from depression compared to the SGLT2i group (95% CI for the difference, 1.60 to 5.02 days; p < 0.001). Flexible spline models showed that the time‐varying HR rose during early follow‐up, continued to increase until approximately 6 months (day 181), after which it stabilized at approximately 1.12 through the end of follow‐up (Figure 3).
TABLE 2.
Primary, secondary, sensitivity, and subgroup analyses for the risk of incident depression and other outcomes associated with GLP‐1 RA use compared to SGLT2i use.
| Patients and events | HR | Log‐rank P | PHA | |
|---|---|---|---|---|
| Primary Outcome (Depression and antidepressants) | 25 704 (4359 vs. 3802) | 1.09 (1.04, 1.14) | 0.0001 | 0.40 |
| Secondary Outcome | ||||
| Depression | 25 704 (1753 vs. 1475) | 1.12 (1.05, 1.20) | 0.001 | 0.70 |
| Antidepressants | 25 704 (3778 vs. 3348) | 1.07 (1.02, 1.12) | 0.005 | 0.88 |
| Deceased | 25 704 (239 vs. 303) | 0.74 (0.63, 0.88) | 0.001 | 0.003 |
| Primary Outcome and Deceased | 25 704 (4528 vs. 4016) | 1.07 (1.03, 1.12) | 0.002 | 0.69 |
| Sensitivity Analysis | ||||
| With TTE compatible protocol | 34 660 (5801 vs. 5256) | 1.06 (1.02, 1.10) | 0.002 | 0.20 |
| With prior EHR record | 12 195 (2059 vs. 1944) | 1.03 (0.97, 1.09) | 0.41 | 0.81 |
| Elderly with prior EHR record | 5156 (865 vs. 759) | 1.11 (1.01, 1.22) | 0.038 | 0.95 |
| Only DM (No overweight or obese) | 35 568 (5087 vs. 4353) | 1.11 (1.07, 1.16) | <0.0001 | 0.05 |
| Cohort since 2014 | 26 709 (4583 vs. 3995) | 1.09 (1.05, 1.14) | <0.0001 | 0.93 |
| 2 months of grace period | 23 640 (3952 vs. 3456) | 1.09 (1.04, 1.14) | <0.0001 | 0.48 |
| Primary outcome analysed on 12th Sep | 28 445 (5002 vs. 4035) | 1.11 (1.07, 1.16) | <0.0001 | 0.41 |
| Exclude ESRD and pregnancy (12th Sep) | 28 380 (5024 vs. 4246) | 1.13 (1.08, 1,18) | <0.0001 | 0.23 |
| Compared with DPP‐4i (12th Sep) | 33 861 (5972 vs. 5194) | 1.12 (1.08, 1.16) | <0.0001 | 0.91 |
| Landmark Analysis | ||||
| 1 month–3 years | 25 311 (5797 vs. 5128) | 1.10 (1.06, 1.15) | <0.0001 | 0.32 |
| 1 month–6 months | 25 311 (3018 vs. 2640) | 1.08 (1.03, 1.14) | 0.003 | 0.34 |
| 2 weeks–6 months | 25 311 (3349 vs. 3007) | 1.06 (1.01, 1.11) | 0.029 | 0.06 |
| 2 weeks–1 year | 25 311 (4534 vs. 4040) | 1.07 (1.02, 1.11) | 0.004 | 0.31 |
| 2 weeks–3 years | 25 311 (5989 vs. 5379) | 1.09 (1.05, 1.13) | <0.0001 | 0.10 |
| 2 weeks–2 months | 25 309 (1692 vs. 1540) | 1.05 (0.98, 1.13) | 0.15 | 0.008 |
| Subgroup analysis | ||||
| Age | Interaction P | 0.028 | ||
| Elder | 10 350 (1784 vs. 1467) | 1.15 (1.08, 1.24) | <0.0001 | 0.57 |
| Younger | 14 103 (2330 vs. 2111) | 1.04 (0.98, 1.10) | 0.19 | 0.21 |
| Sex | Interaction P | 0.953 | ||
| Male | 12 237 (1616 vs. 1413) | 1.09 (1.01, 1.17) | 0.025 | 0.55 |
| Female | 11 973 (2524 vs. 2204) | 1.09 (1.03, 1.15) | 0.004 | 0.55 |
| Previous Psychiatric History | Interaction P | 0.735 | ||
| With Psychiatric History | 4816 (1670 vs. 1476) | 1.08 (1.01, 1.16) | 0.035 | 0.17 |
| Without History | 20 011 (2454 vs. 2130) | 1.10 (1.03, 1.16) | 0.002 | 0.64 |
| GLP‐1 RA Agent | Interaction P | 0.092 | ||
| Semaglutide | 16 153 (2754 vs. 2475) | 1.07 (1.01, 1.13) | 0.022 | 0.24 |
| Tirzepatide | 3731 (705 vs. 629) | 1.07 (0.96, 1.12) | 0.196 | 0.99 |
| Liraglutide | 10 025 (1912 vs. 1586) | 1.17 (1.09, 1.25) | <0.0001 | 0.17 |
| Unmatched and NCO | ||||
| Unmatched Cohort | 31 056 (5567) vs. 50 847 (6630) | 1.26 (1.22, 1.31) | <0.0001 | 0.02 |
| (NCO) Acute appendicitis | 25 704 (14 vs. 12) | 1.10 (0.51, 2.37) | 0.81 | 0.40 |
| (NCO) Contusion of knee | 25 704 (34 vs. 31) | 1.03 (0.63, 1.67) | 0.91 | 0.66 |
Abbreviations: DM, diabetes mellitus; DPP‐4i, dipeptidyl peptidase‐4 inhibitor; EHR, electronic health record; ESRD, end‐stage renal disease; GLP‐1 RA, glucagon‐like peptide‐1 receptor agonist; HR, hazard ratio; Hx, history; Log‐rank p, p‐value from the log‐rank test; NCO, negative control outcome; PHA, p‐value for the proportional hazard assumption test; TTE, target trial emulation.
FIGURE 2.
Cumulative incidence of depression. The shaded region represents the 95% confidence interval (CI). GLP1 RA, glucagon‐like peptide‐1 receptor agonist; SGLT2i, sodium‐glucose cotransporter‐2 inhibitor.
FIGURE 3.
Time‐varying hazard ratio for incident depression with GLP1 RA versus SGLT2i. The solid line represents the adjusted hazard ratio (HR) and the shaded region represents the 95% confidence interval (CI) for the risk of depression in patients initiating GLP1 RA compared to SGLT2i (reference). A value above the dashed line (HR >1.0) indicates a higher risk for the GLP1RA group. GLP1 RA, glucagon‐like peptide‐1 receptor agonist; SGLT2i, sodium‐glucose cotransporter‐2 inhibitor.
When the components of the primary outcome were examined separately, GLP‐1 RA use was associated with an increased risk of both a new depression diagnosis (HR 1.12; 95% CI, 1.05 to 1.20) and the initiation of a new antidepressant (HR 1.07; 95% CI, 1.02 to 1.12). As a secondary outcome, GLP‐1 RA use was associated with a significantly lower risk of all‐cause mortality compared to SGLT2i use (HR 0.74; 95% CI, 0.63 to 0.88; p = 0.001).
3.3. Sensitivity and subgroup analyses
The main findings remained largely consistent across a series of prespecified sensitivity analyses (Table S3). Methodological robustness analyses showed that in the target trial emulation (which relaxed mutual exclusivity), the increased depression risk remained significant but slightly attenuated (HR 1.06; 95% CI, 1.02 to 1.10). The negative‐control outcome analyses were not significant. Extended follow‐up and population analyses confirmed the robustness of our results. The increased risk persisted when extending follow‐up to 3 years (HR 1.10; 95% CI, 1.06 to 1.15) and when restricting to patients with T2D only, regardless of weight status (HR 1.11; 95% CI, 1.07 to 1.16). However, when restricting to patients with prior electronic health records, the association was attenuated and non‐significant (HR 1.03; 95% CI, 0.97 to 1.09). Similarly, landmark analysis of the early period (2 weeks to 2 months) showed no significant association (HR 1.05; 95% CI, 0.98 to 1.13).
In subgroup analyses, a significant interaction was observed with age (p for interaction = 0.028), with a more pronounced risk increase among older adults (≥65 years) (HR 1.15; 95% CI, 1.08 to 1.24) compared to younger adults (<65 years). No significant interaction was found for sex or prior psychiatric history. Among individual GLP‐1 RA agents, the association was statistically significant for liraglutide (HR 1.17; 95% CI, 1.09 to 1.25) and semaglutide (HR 1.07; 95% CI, 1.01 to 1.13), but did not reach statistical significance for tirzepatide (HR 1.07; 95% CI, 0.96 to 1.12).
4. DISCUSSION
In this large, new‐user, active‐comparator cohort study of adults with type 2 diabetes and overweight or obesity, we found that initiating a GLP‐1 receptor agonist was associated with a 9% higher relative rate of incident depression compared to initiating an SGLT2i. It is important to contextualize this finding, as emerging evidence suggests SGLT2 inhibitors and DPP‐4 inhibitors may have potentially beneficial effects on mood. 28 , 29 Thus, our findings should not be interpreted as GLP‐1 RAs having an absolute risk of causing depression. Given the high baseline depression incidence (14.8% annually), this translated to an absolute risk difference of 2.2% and a number needed to harm of 45 over 1 year. As the use of GLP‐1 RAs continues to expand rapidly, this risk can impact a substantial number of patients at the population level. While many large‐scale randomized controlled trials (RCTs) have not flagged depression as a significant adverse event, 9 , 30 our real‐world data suggest this might be due to limitations inherent in trial design. RCTs are typically not powered for psychiatric endpoints and often enroll highly selected, lower‐risk populations. 31 Moreover, by using an active‐comparator design and integrating laboratory‐based covariates, our analysis more rigorously mitigates confounding than prior observational studies that relied on non‐user comparators or claims data without lab information. 7 , 10 , 12
While our primary finding was statistically significant, its robustness warrants careful consideration. The E‐value was 1.4, suggesting that an unmeasured confounder associated with both the exposure and outcome by a hazard ratio of 1.4 could potentially explain away the observed association. For example, unmeasured patient preferences could create selection bias. Individuals with greater concern about body weight may be predisposed to depression and also more likely to opt for an injectable GLP‐1 RA over an oral alternative. The observed risk may be partially attributable to these unmeasured patient factors, warranting a cautious interpretation. Sensitivity analyses revealed that the association, while directionally consistent, was attenuated in certain scenarios. For instance, requiring a prior healthcare visit for inclusion rendered the result non‐significant, suggesting that unmeasured confounding related to healthcare‐seeking behaviour or underlying health status may still persist despite comprehensive propensity score matching. Furthermore, the temporal pattern of the risk provides important clues. The landmark analysis from 2 weeks to 2 months was not statistically significant, indicating the risk difference does not emerge immediately after treatment initiation.
Our analysis of the time‐varying hazard ratio shows that the risk for depression is not constant. The HR increased until approximately 6 months and then levelled off around 1.12. This dynamic suggested a period of escalating risk during the first 6 months. Such time‐varying patterns, where the relative effect of GLP‐1 RAs changes over time and can lead to crossing survival curves, are consistent with findings from several large‐scale studies on similar endpoints. 10 , 12 The delayed onset implies a gradual process of complex biopsychosocial adjustment rather than an acute drug reaction. Although we cannot disentangle these pathways, our findings suggest the association extends beyond a direct pharmacological effect on central mood circuits. 32 , 33
The subgroup analyses provided further nuance, revealing a significant interaction with age. The increased risk of depression was more pronounced in older adults (≥65 years), a vulnerable population in whom the clinical and functional consequences of depression can be particularly severe. 34 This finding highlights a specific subgroup that may benefit from heightened clinical vigilance. The association was statistically significant for the individual agents liraglutide and semaglutide but did not reach significance for tirzepatide. A recent pharmacovigilance study also reported drug‐specific effects on depression. 35 It remains to be determined whether tirzepatide's dual incretin mechanism, involving glucose‐dependent insulinotropic polypeptide (GIP), confers distinct neuropsychiatric effects compared with pure GLP‐1 RAs. 36 , 37 However, this observation should be interpreted with caution due to the potential for reduced statistical power in subgroup analyses. Given its increasing clinical adoption, future head‐to‐head studies focusing on tirzepatide are warranted to clarify its potential differential effects. While our primary analysis suggests an association with a higher incidence of depression, a secondary analysis also found an association between GLP‐1 RA use and a 26% lower rate of all‐cause mortality compared to SGLT2i use. This observation is consistent with findings from cardiovascular outcome trials. 38 , 39 Therefore, clinicians should consider the potential for mood‐related changes alongside the established cardiometabolic benefits of GLP‐1 RAs.
These findings have direct implications for clinical practice. Guided by the principle of do no harm, clinicians should be cognizant of this potential, albeit small, risk of mood changes when prescribing GLP‐1 RAs. This does not suggest withholding these highly effective therapies but rather calls for a proactive approach to patient care. This includes engaging in shared decision‐making to discuss the full spectrum of potential benefits and risks, as well as implementing routine monitoring for mood symptoms, particularly beyond the first few months of treatment and especially in older patients. Timely referral to mental health specialists should be considered for patients who develop concerning symptoms.
4.1. Strengths and limitations
The strengths of our study include its large, real‐world population and the robust new‐user, active‐comparator design, which minimizes critical biases. Additionally, our analysis of time‐varying hazard ratios captured subacute effects that might be missed by conventional approaches. Nevertheless, our study has limitations inherent to retrospective observational research. First, despite extensive covariate adjustment, residual confounding from unmeasured variables is always possible. Second, our reliance on administrative and diagnostic codes for outcome ascertainment is susceptible to misclassification bias. Third, in the analysis limited by the TriNetX platform, we were unable to conduct more granular per‐protocol or time‐varying analyses, which could be influenced by treatment discontinuation or switching. Moreover, as laboratory data were not prospectively collected, stratified analyses based on such measures could not be performed. Fourth, the cases in TriNetX are derived mostly from medical centres in the US, and we focused on overweight or obese patients with comorbid T2D. Therefore, the generalizability to other populations or to patients with obesity alone is limited. In addition, because we excluded co‐initiation of other glucose‐lowering agents (e.g., metformin) at the index date to enhance internal validity, the applicability of our findings to typical combination therapy in real‐world practice may also be limited.
5. CONCLUSION
In a real‐world setting, initiating a GLP‐1 RA was associated with a small but statistically significant increase in the risk of incident depression compared to an SGLT2i, an effect most pronounced during the subacute period and among older adults. This risk must be carefully balanced against the substantial mortality benefit afforded by GLP‐1 RAs. Future research should aim to identify other potentially susceptible subgroups and explore the biopsychosocial mechanisms linking GLP‐1 RA therapy and mood, such as the rate and magnitude of weight loss, through prospective cohort studies and pragmatic randomized controlled trials. Ongoing synthesis of real‐world evidence, including independent meta‐analyses, will be essential to refine clinical practice guidelines.
AUTHOR CONTRIBUTIONS
Concept and design: Yu Chang and Ming‐Hong Hsieh. Acquisition, analysis, or interpretation of data: Yu Chang and Cheng‐Chen Chang. Drafting of the manuscript: Yu Chang. Critical review of the manuscript for important intellectual content: Yu Chang, Ming‐Hong Hsieh, Po‐Chung Ju, and Cheng‐Chen Chang. Statistical analysis: Yu Chang. Supervision: Cheng‐Chen Chang. All authors read and approved the final manuscript.
FUNDING INFORMATION
Not applicable.
CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.
ETHICS STATEMENT
The study protocol received approval from the Institutional Review Board of Chung Shan Medical University (Case No. CS2‐25111).
Supporting information
Data S1: Supporting Information.
ACKNOWLEDGEMENTS
The authors have nothing to report.
Chang Y, Hsieh M‐H, Ju P‐C, Chang C‐C. Risk of depression with GLP‐1 receptor agonists use in overweight or obese adults with type 2 diabetes: A new‐user, active‐comparator cohort study. Diabetes Obes Metab. 2026;28(1):197‐209. doi: 10.1111/dom.70175
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the TriNetX research network but restrictions apply to the availability of these data. Data may be available from the authors upon reasonable request and with permission of TriNetX.
REFERENCES
- 1. Gilbert O, Gulati M, Gluckman TJ, et al. Concise clinical guidance: an ACC expert consensus statement on medical weight management for optimization of cardiovascular health. JACC. 2025;86(7):536–555. doi: 10.1016/j.jacc.2025.05.024 [DOI] [PubMed] [Google Scholar]
- 2. Yang R, Zhang L, Guo J, et al. Glucagon‐like Peptide‐1 receptor agonists for obstructive sleep apnea in patients with obesity and type 2 diabetes mellitus: a systematic review and meta‐analysis. J Transl Med. 2025;23(1):389. doi: 10.1186/s12967-025-06302-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Sánchez‐Garrido MA, Serrano‐López V, Ruiz‐Pino F, et al. Superior metabolic improvement of polycystic ovary syndrome traits after GLP1‐based multi‐agonist therapy. Nat Commun. 2024;15(1):8498. doi: 10.1038/s41467-024-52898-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Nevola R, Epifani R, Imbriani S, et al. GLP‐1 receptor agonists in non‐alcoholic fatty liver disease: current evidence and future perspectives. Int J Mol Sci. 2023;24(2):1703. doi: 10.3390/ijms24021703 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Perkovic V, Tuttle KR, Rossing P, et al. Effects of Semaglutide on chronic kidney disease in patients with type 2 diabetes. N Engl J Med. 2024;391(2):109‐121. doi: 10.1056/NEJMoa2403347 [DOI] [PubMed] [Google Scholar]
- 6. Lu W, Wang S, Tang H, Yuan T, Zuo W, Liu Y. Neuropsychiatric adverse events associated with glucagon‐like peptide‐1 receptor agonists: a pharmacovigilance analysis of the FDA adverse event reporting system database. Eur Psychiatry. 2025;68(1):e20. doi: 10.1192/j.eurpsy.2024.1803 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Kornelius E, Huang JY, Lo SC, Huang CN, Yang YS. The risk of depression, anxiety, and suicidal behavior in patients with obesity on glucagon like peptide‐1 receptor agonist therapy. Sci Rep. 2024;14(1):24433. doi: 10.1038/s41598-024-75965-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Wang W, Volkow ND, Berger NA, Davis PB, Kaelber DC, Xu R. Association of semaglutide with risk of suicidal ideation in a real‐world cohort. Nat Med. 2024;30(1):168‐176. doi: 10.1038/s41591-023-02672-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Chen X, Zhao P, Wang W, Guo L, Pan Q. The antidepressant effects of GLP‐1 receptor agonists: a systematic review and meta‐analysis. Am J Geriatr Psychiatry. 2024;32(1):117‐127. doi: 10.1016/j.jagp.2023.08.010 [DOI] [PubMed] [Google Scholar]
- 10. Shapiro SB, Yin H, Yu OHY, Rej S, Suissa S, Azoulay L. Glucagon‐like peptide‐1 receptor agonists and risk of suicidality among patients with type 2 diabetes: active comparator, new user cohort study. BMJ. 2025;388:e080679. doi: 10.1136/bmj-2024-080679 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. De Giorgi R, Ghenciulescu A, Dziwisz O, et al. An analysis on the role of glucagon‐like peptide‐1 receptor agonists in cognitive and mental health disorders. Nat Mental Health. 2025;3(3):354‐373. doi: 10.1038/s44220-025-00390-x [DOI] [Google Scholar]
- 12. Tang H, Lu Y, Donahoo WT, et al. Glucagon‐like Peptide‐1 receptor agonists and risk for depression in older adults with type 2 diabetes : a target trial emulation study. Ann Intern Med. 2025;178(3):315‐326. doi: 10.7326/ANNALS-24-01347 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Ueda P, Söderling J, Wintzell V, et al. GLP‐1 receptor agonist use and risk of suicide death. JAMA Intern Med. 2024;184(11):1301‐1312. doi: 10.1001/jamainternmed.2024.4369 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Chang Y, Huang SS, Hsu WY, Liu YC. Mapping psychiatric comorbidity network: a pilot multi‐method weighted network analysis with a focus on key disorders. Compr Psychiatry. 2025;141:152608. doi: 10.1016/j.comppsych.2025.152608 [DOI] [PubMed] [Google Scholar]
- 15. U.S. Food & Drug Administration . Update on FDA's ongoing evaluation of reports of suicidal thoughts or actions in patients taking a certain type of medicines approved for type 2 diabetes and obesity. Published online February 19, 2025. Accessed August 5, 2025. https://www.fda.gov/drugs/drug-safety-and-availability/update-fdas-ongoing-evaluation-reports-suicidal-thoughts-or-actions-patients-taking-certain-type
- 16. European Medicines Agency (EMA) . Meeting highlights from the Pharmacovigilance Risk Assessment Committee (PRAC) 8–11 April 2024 | European Medicines Agency (EMA). 2024. April 12, 2024. Accessed August 5, 2025. https://www.ema.europa.eu/en/news/meeting-highlights-pharmacovigilance-risk-assessment-committee-prac-8-11-april-2024
- 17. Yapici‐Eser H, Appadurai V, Eren CY, et al. Association between GLP‐1 receptor gene polymorphisms with reward learning, anhedonia and depression diagnosis. Acta Neuropsychiatr. 2020;32(4):218–225. doi: 10.1017/neu.2020.14 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Cork SC, Richards JE, Holt MK, Gribble FM, Reimann F, Trapp S. Distribution and characterisation of glucagon‐like peptide‐1 receptor expressing cells in the mouse brain. Molecular Metabolism. 2015;4(10):718‐731. doi: 10.1016/j.molmet.2015.07.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Batterham PJ, Christensen H, Calear AL. Anxiety symptoms as precursors of major depression and suicidal ideation. Depress Anxiety. 2013;30(10):908‐916. doi: 10.1002/da.22066 [DOI] [PubMed] [Google Scholar]
- 20. Sheu WHH, Chan SP, Matawaran BJ, et al. Use of SGLT‐2 inhibitors in patients with type 2 diabetes mellitus and abdominal obesity: an Asian perspective and expert recommendations. Diabetes Metab J. 2020;44(1):11‐32. doi: 10.4093/dmj.2019.0208 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Tsushima Y, Lansang MC, Makin V. The role of SGLT‐2 inhibitors in managing type 2 diabetes. CCJM. 2021;88(1):47‐58. doi: 10.3949/ccjm.88a.20088 [DOI] [PubMed] [Google Scholar]
- 22. Fanelli G, Raschi E, Hafez G, et al. The interface of depression and diabetes: treatment considerations. Transl Psychiatry. 2025;15(1):22. doi: 10.1038/s41398-025-03234-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Dong D, Liu X, Ma L, et al. Dapagliflozin inhibits the activity of lateral habenula to alleviate diabetes mellitus‐induced depressive‐like behavior. Exp Neurol. 2023;366:114448. doi: 10.1016/j.expneurol.2023.114448 [DOI] [PubMed] [Google Scholar]
- 24. Haneuse S, VanderWeele TJ, Arterburn D. Using the E‐value to assess the potential effect of unmeasured confounding in observational studies. JAMA. 2019;321(6):602‐603. doi: 10.1001/jama.2018.21554 [DOI] [PubMed] [Google Scholar]
- 25. Liu N, Zhou Y, Lee JJ. IPDfromKM: reconstruct individual patient data from published Kaplan‐Meier survival curves. BMC Med Res Methodol. 2021;21(1):111. doi: 10.1186/s12874-021-01308-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Uno H, Claggett B, Tian L, et al. Moving beyond the Hazard ratio in quantifying the between‐group difference in survival analysis. J Clin Oncol. 2014;32(22):2380‐2385. doi: 10.1200/JCO.2014.55.2208 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Royston P, Parmar MKB. Flexible parametric proportional‐hazards and proportional‐odds models for censored survival data, with application to prognostic modelling and estimation of treatment effects. Stat Med. 2002;21(15):2175‐2197. doi: 10.1002/sim.1203 [DOI] [PubMed] [Google Scholar]
- 28. Liebers DT, Ebina W, Iosifescu DV. Sodium‐glucose Cotransporter‐2 inhibitors in depression. Harv Rev Psychiatry. 2023;31(4):214‐221. doi: 10.1097/HRP.0000000000000374 [DOI] [PubMed] [Google Scholar]
- 29. Zoicas I, Mühle C, von Hörsten S, Plank AC, Kornhuber J. Sitagliptin as a therapeutic approach for social anxiety disorder: the role of DPP4 and NPY in modulating social fear and comorbid depressive‐like behavior in mice. Neuropsychopharmacology. 2025;50:1724‐1731. doi: 10.1038/s41386-025-02146-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Pierret ACS, Mizuno Y, Saunders P, et al. Glucagon‐like peptide 1 receptor agonists and mental health: a systematic review and meta‐analysis. JAMA Psychiatry. 2025;82(7):643‐653. doi: 10.1001/jamapsychiatry.2025.0679 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Singh S, Loke YK. Drug safety assessment in clinical trials: methodological challenges and opportunities. Trials. 2012;13(1):138. doi: 10.1186/1745-6215-13-138 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Arillotta D, Floresta G, Guirguis A, et al. GLP‐1 receptor agonists and related mental health issues; insights from a range of social media platforms using a mixed‐methods approach. Brain Sci. 2023;13(11):1503. doi: 10.3390/brainsci13111503 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Moszak M, Marcickiewicz J, Pelczyńska M, Bogdański P. The interplay between psychological and neurobiological predictors of weight regain: a narrative review. Nutrients. 2025;17(10):1662. doi: 10.3390/nu17101662 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Szymkowicz SM, Gerlach AR, Homiack D, Taylor WD. Biological factors influencing depression in later life: role of aging processes and treatment implications. Transl Psychiatry. 2023;13(1):160. doi: 10.1038/s41398-023-02464-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Wang M, Chen X, Liu Z, et al. Exploring potential associations between GLP‐1RAs and depressive disorders: a pharmacovigilance study based on FAERS and VigiBase data. eClin Med. 2025;86:103385. doi: 10.1016/j.eclinm.2025.103385 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Liu QK. Mechanisms of action and therapeutic applications of GLP‐1 and dual GIP/GLP‐1 receptor agonists. Front Endocrinol. 2024;15:15. doi: 10.3389/fendo.2024.1431292 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Zhou Q, Lei X, Fu S, et al. Efficacy and safety of tirzepatide, dual GLP‐1/GIP receptor agonists, in the management of type 2 diabetes: a systematic review and meta‐analysis of randomized controlled trials. Diabetol Metab Syndr. 2023;15(1):222. doi: 10.1186/s13098-023-01198-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Lee MMY, Sattar N, Pop‐Busui R, et al. Cardiovascular and kidney outcomes and mortality with Long‐acting injectable and Oral glucagon‐like peptide 1 receptor agonists in individuals with type 2 diabetes: a systematic review and meta‐analysis of randomized trials. Diabetes Care. 2025;48(5):846‐859. doi: 10.2337/dc25-0241 [DOI] [PubMed] [Google Scholar]
- 39. Sattar N, Lee MMY, Kristensen SL, et al. Cardiovascular, mortality, and kidney outcomes with GLP‐1 receptor agonists in patients with type 2 diabetes: a systematic review and meta‐analysis of randomised trials. Lancet Diabetes Endocrinol. 2021;9(10):653‐662. doi: 10.1016/S2213-8587(21)00203-5 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1: Supporting Information.
Data Availability Statement
The data that support the findings of this study are available from the TriNetX research network but restrictions apply to the availability of these data. Data may be available from the authors upon reasonable request and with permission of TriNetX.