The association between exposure to GLP-1 receptor agonists and risk of suicide-related and self-harm-related events in diabetic patients from the UK general practice IQVIA Medical Research Data (IMRD) database

Background: The study was triggered by a safety signal of suicidal and self-injurious ideation after exposure to liraglutide and semaglutide, glucagon-like peptide-1 receptor agonists (GLP-1a), authorized for weight management and type-2 diabetes mellitus (T2DM).

Objectives: To evaluate whether initiating GLP-1a is associated with an increased risk of suicide- and self-harm related events compared to initiating sodium-glucose co-transporter 2 inhibitors (SGLT-2i) in T2DM.

Methods: This was a new user, active comparator cohort study using the IMRD database from March 2013 to June 2023.
Patients who first initiated a GLP-1a or a SGLT-2i within the study period were eligible. Treatment initiation was the index-date. Additional eligibility criteria evaluated at baseline were a confirmed T2DM diagnosis, use of biguanides and no history of outcome.
The outcome was a composite of suicide-related (suicidal ideation, attempted suicide, completed suicide) and self-harm-related (deliberate self-harm/self-injury, self-inflicted harm, self-injurious behavior) diagnoses.
Patients were followed from index to first outcome event. In an intention-to-treat (ITT) analysis, follow-up was censored at outcome, death, data-exit, or study-end. In an on-treatment (OT) analysis, patients were additionally censored at treatment discontinuation or switch to the other treatment class.
Inverse probability of treatment weighting was used to adjust for baseline confounding and inverse probability of censoring weighting was used in the OT to also adjust for potential selection bias due to treatment discontinuation.
Incidence rates (IR) were calculated as the number of events divided by person time and hazard (HR) ratios were estimated using proportional hazards Cox regression.

Results: 6,207 GLP-1a and 20,855 SGLT-2i initiators were eligible for the study.
In the ITT analysis, the crude IRs were 0.50 (95% CI: 0.42-0.60) and 0.30 (0.26-0.35) events per 100 person-years for GLP-1a and SGLT-2i respectively, corresponding to a crude HR of 1.71 (1.35-2.16). After confounding adjustment, the HR dropped to 1.10 (0.86-1.41). In the OT analysis, the crude IRs were 0.48 (0.37-0.62) for GLP-1a and 0.29 (0.24-0.35) for SGLT-2i, with a crude HR of 1.65 (1.20-2.26) and an adjusted HR of 1.16 (0.83-1.61). In an ITT analysis restricted to obese patients, the crude IRs were 0.55 (0.46 - 0.66) for GLP-1a and 0.35 (0.29 - 0.41) for SGLT2i, corresponding to a HR of 1.66 (1.29-2.13) which dropped to 1.33 (1.02-1.72) after confounding adjustment.

Conclusions: Our study does not support an increased risk of suicide-/self-harm- related events associated with the use of GLP1a versus SGLT2i in the general T2DM population but the contrast might be higher in the obese subpopulation, which warrants further investigation.