A large cohort study using nationwide data from Sweden and Denmark found no statistically significant increased risk of suicide death among users of glucagon-like peptide-1 receptor agonists compared with users of sodium-glucose cotransporter-2 inhibitors.
The study, published in JAMA Internal Medicine, examined 298,553 adults who initiated either medication between 2013 and 2021. During a mean follow-up of 2.5 years, 77 suicide deaths occurred among glucagon-like peptide-1 (GLP-1) receptor agonist users compared to 71 among sodium-glucose cotransporter-2 (SGLT-2) inhibitor users. The weighted incidence rates were 0.23 versus 0.18 events per 1000 person-years, respectively. The absolute risk difference was 0.05 events per 1000 person-years.
The study also found no increased risk for secondary outcomes. The hazard ratio (HR) for the composite of suicide death and nonfatal self-harm was 0.83, and for incident depression and anxiety-related disorders it was 1.01.
Methods and Results
Researchers used an active-comparator new-user design, analyzing data from national health and administrative registers. The study included 124,517 new users of GLP-1 receptor agonists (77,495 in Sweden, 47,022 in Denmark) and 174,036 new users of SGLT-2 inhibitors (108,881 in Sweden, 65,155 in Denmark), aged 18 to 84 years.
Among GLP-1 receptor agonist users, the mean age was 60 years and 45% were women. The most commonly used GLP-1 receptor agonists were liraglutide (50%) and semaglutide (41%).
Median follow-up times varied by country:
- Sweden: 2.8 years for GLP-1 users; 2.1 years for SGLT-2 users
- Denmark: 2.1 years for GLP-1 users; 2.2 years for SGLT-2 users
Country-specific results for the primary outcome were:
- Sweden: HR 1.44
- Denmark: HR 0.94
Subgroup analyses showed:
- History of psychiatric disorders: HR 1.25
- No history of psychiatric disorders: HR 1.44
- Liraglutide users: HR 1.35
- Semaglutide users: HR 0.74
Additional analyses using an as-treated exposure definition and restricting follow-up to the first year yielded similar results.
Secondary Outcomes and Sensitivity Analyses
For the self-harm analysis, after excluding those with nonfatal self-harm within 3 months prior to cohort entry, 124,459 GLP-1 receptor agonist users and 173,985 SGLT-2 inhibitor users were included. The HR for self-harm was 0.77.
The incident depression and anxiety-related disorders analysis, after excluding those with previous psychiatric disorders, included 72,420 GLP-1 receptor agonist users and 111,083 SGLT-2 inhibitor users.
A sensitivity analysis defining incident depression and anxiety-related disorders using only diagnoses registered during health care visits yielded an HR of 1.17.
Baseline Characteristics
Before weighting, baseline characteristics showed some differences between groups:
Current use of antidepressants (past 6 months):
- Sweden: 19.0% GLP-1 users versus 14.7% SGLT-2 users
- Denmark: 16.3% GLP-1 users versus 12.6% SGLT-2 users
Previous depression or anxiety-related disorder:
- Sweden: 12.8% GLP-1 users versus 9.1% SGLT-2 users
- Denmark: 4.8% GLP-1 users versus 3.4% SGLT-2 users
Cardiovascular disease:
- Sweden: 28.7% GLP-1 users versus 41.0% SGLT-2 users
- Denmark: 28.7% GLP-1 users versus 36.6% SGLT-2 users
Obesity diagnosis:
- Sweden: 18.8% GLP-1 users versus 9.9% SGLT-2 users
- Denmark: 23.0% GLP-1 users versus 12.5% SGLT-2 users
Prescription drug use in the past year also differed:
Metformin:
- Sweden: 69.5% GLP-1 users versus 74.8% SGLT-2 users
- Denmark: 70.3% GLP-1 users versus 81.1% SGLT-2 users
Insulin:
- Sweden: 39.3% GLP-1 users versus 21.6% SGLT-2 users
- Denmark: 20.1% GLP-1 users versus 10.1% SGLT-2 users
Limitations
The study primarily included patients with type 2 diabetes, limiting generalizability to those using GLP-1 receptor agonists for obesity without diabetes. While 25% of users were followed for at least 4.7 years, longer-term risks could not be assessed. The upper limit of the confidence interval was compatible with up to an 88% relative risk increase of suicide death, corresponding to an absolute risk increase of no more than 0.16 per 1000 person-years.
Conflict of interest disclosures can be found in the published study.