Risk of suicidal attempt and GLP-1RAs in type 2 diabetes population in Taiwan: a case-crossover and case-case-time-control study

Background: Glucagon-like peptide-1 receptor agonist (GLP-1RA) is a widely used medication for managing type 2 diabetes and weight control. Nevertheless, following the European Medicines Agency's (EMA) July 2023 announcement about the possible link between GLP-1RAs and increased suicidal attempts, the need for additional evidence to shape future policies has become pressing. Moreover, given the significant global prevalence of type 2 diabetes and the complexity of healthcare in these cases, it’s critical to determine whether such an association indeed exists in the type 2 diabetes population. Here, we initiate a study employing a case-crossover design trying to address the sophisticated characteristics within the population. Additionally, we're implementing a case-case-time-control (CCTC) design to mitigate any potential biases in exposure time trends that could bias the results.

Objectives: To evaluate the association between suicidal attempt and the use of GLP-1RAs in type 2 diabetes population in Taiwan.

Methods: The study applied both case-crossover and CCTC designs. The individuals enrolled in Taiwan's National Health Insurance Research Database (NHID) from 2011 to 2021 were included. We identified the patients with incident suicidal attempt event. The index date (day0) was defined as the date when the suicidal attempt was coded, and the hazard period, washout period, and referent period were defined retrospectively (90 days for each period). For CCTC design, we matched each case with 1 “future case” on age and sex to adjust for exposure time trends. The baseline period was within 1 year period before the index date. The exposure was defined as all the GLP-1RAs approved for type 2 diabetes treatment in Taiwan, including exenatide, lixisenatide, dulaglutide, liraglutide, and semaglutide. Odds ratios (ORs) and 95% confidence intervals (CIs) between hazard period and referent period was obtained by conditional logistic regression.

Results: We identified 1,243,120 cases in case-crossover analysis and 971,789 eligible case-control pairs in CCTC analysis. The odds ratio of case-crossover analysis is 1.99 (1.66, 2.40). The odds ratios of CCTC analysis are 2.21 (1.80, 2.72) and 1.82 (1.46, 2.28) for case group and control group respectively, and the adjustment OR from CCTC is 1.21 (0.90, 1.64).

Conclusions: According to our study, although the significant result was observed in the case-crossover design, the association between GLP-1RAs and suicidal attempt did not reach statistical significance after applying CCTC design, explained the possible bias of exposure time-trend in the case-crossover manners.