How-To Guides

Yes, in this VA target trial emulation study, semaglutide and tirzepatide were associated with significantly lower 12-month all-cause overdose risk than insulin and SGLT2 inhibitors, but not than metformin, sulfonylureas, or DPP4 inhibitors. The authors describe these findings as an observational signal rather than proof of causal benefit.

The study compared semaglutide or tirzepatide initiation with 5 major diabetes medication groups: metformin, insulin, sulfonylureas, dipeptidyl-peptidase 4 inhibitors (DPP4 inhibitors), and sodium-glucose cotransporter-2 inhibitors (SGLT2 inhibitors). The analysis was done as 5 separate 1:1 propensity-matched comparisons.

The primary outcome was all-cause overdose, not opioid overdose alone. The authors used all-cause overdose because there were too few events to support adequately powered opioid-only analyses, and many overdose cases did not specify the substance involved.

Patients were followed for 12 months after medication initiation. Time to overdose was analyzed with Cox proportional hazards models, and participants without an overdose event were censored at 12 months.

Overdose was identified using a combined all-cause overdose measure that included inpatient or emergency department overdose diagnoses documented with ICD codes plus Suicidal Behavior and Overdose Report (SBOR) data from the VA Office of Mental Health and Suicide Prevention. The authors used this combined approach to capture overdoses that might not be fully documented in the formal medical record.

The study included Veterans receiving VA care who had type 2 diabetes mellitus and opioid use disorder and who initiated semaglutide, tirzepatide, or 1 of 5 comparator diabetes medications between January 1, 2020, and December 31, 2024. OUD was defined by ICD coding from at least 1 inpatient or 2 outpatient encounters, and patients had to have a T2DM and OUD diagnosis after January 1, 2014.

Patients were excluded if they had 30 days or less of cumulative prescription coverage or fewer than 2 fills of semaglutide/tirzepatide or the comparator medication starting on the index date. Comparison-group patients were also excluded if they initiated semaglutide/tirzepatide during the 12-month follow-up, and semaglutide/tirzepatide patients were excluded if they started a comparator diabetes medication on the same date.

The researchers used 1:1 nearest-neighbor propensity score matching with a 0.20 standard deviation caliper, performed separately for each comparator group. Matching included demographics, socioeconomic determinants, obesity, chronic pain, psychiatric and substance use diagnoses, prior overdose, diabetes medications, overdose-risk medications such as opioids and sedative-hypnotics, and OUD treatments such as buprenorphine and methadone.

After matching, balance improved substantially. Median standardized mean differences fell from 0.12, 0.11, 0.09, 0.10, and 0.09 before matching to 0.03, 0.05, 0.06, 0.03, and 0.03 for the insulin, SGLT2 inhibitor, DPP4 inhibitor, metformin, and sulfonylurea comparisons, respectively.

For the 2 significant comparisons, the sensitivity analysis suggested relatively strong robustness. The E values indicated that an unmeasured confounder would need hazard ratio associations of 5.70 for the insulin comparison and 7.46 for the SGLT2 inhibitor comparison with both treatment assignment and overdose, beyond the measured confounders, to fully explain away the observed associations.

The article notes that preclinical studies suggest GLP-1 receptor agonists may reduce drug self-administration or drug-seeking behavior, including for opioids, possibly through modulation of dopaminergic neurotransmission involved in reward processing. This study did not test mechanism directly; it evaluated whether an association with overdose was present in clinical observational data.

Semaglutide and tirzepatide were selected because they were available in the VA, are being used in ongoing or planned opioid use disorder trials, and had shown greater efficacy than other GLP-1-based agents in prior observational research cited by the authors. The main analysis evaluated them collectively, although outcomes for semaglutide alone were reported in supplementary material.

• The study was observational, so causality cannot be established and unmeasured confounding may remain despite propensity matching.
• The outcome had to be broadened to all-cause overdose because opioid-only overdose events were too few for adequate power.
• Many overdose cases did not specify the substance involved, limiting opioid versus nonopioid subgroup analysis.
• Some overdoses may have been missed, especially if they occurred outside VA documentation pathways.
• The analysis did not model preinitiation temporal trajectories and did not account for semaglutide or tirzepatide dose because of EHR limitations.
• The authors note that metformin and DPP4 inhibitors may themselves have antiaddiction properties, which could reduce the apparent advantage of semaglutide or tirzepatide in those head-to-head comparisons.

No. The study suggests a potentially important association, but the authors explicitly conclude that randomized controlled trials are needed to determine efficacy in opioid use disorder and to establish causal relationships. These findings may inform clinical discussion in Veterans with type 2 diabetes and OUD, but they do not establish semaglutide or tirzepatide as proven OUD treatment.