GLP-1 agonists were associated with lower mortality than SGLT2 inhibitors in serious mental illness

Design

Retrospective target-trial emulation, propensity-matched, new-user active comparator

Population

195,184 matched pairs with bipolar disorder, major depression or schizophrenia

Primary outcome

4-year all-cause mortality

Effect

4.91% vs 6.45%; HR 0.76 (95% CI 0.74 to 0.78)

A target-trial emulation in the TriNetX electronic health record network compared adults starting a GLP-1 receptor agonist with those starting an SGLT2 inhibitor, with separate propensity-matched cohorts for people with and without bipolar disorder, major depression or schizophrenia.

Among 195,184 matched pairs with serious mental illness, four-year mortality was 4.91% with GLP-1 agonists and 6.45% with SGLT2 inhibitors (HR 0.76, 0.74 to 0.78). Absolute reductions were larger than in people without mental illness. In those with type 2 diabetes, semaglutide was associated with lower major adverse cardiovascular events (HR 0.77).

This is observational, and the one-year mortality halving (RR 0.52) is larger than any plausible drug effect in that time — a sign of residual confounding. It supports, but does not prove, preferring a GLP-1 agonist when one is indicated in these patients.

• Check HbA1c, weight and lipids at every antipsychotic review.
• Where a patient with diabetes needs a second agent, a GLP-1 agonist is a reasonable first choice.
• Coordinate with the physician rather than start alone if you do not usually prescribe these.
• In India, semaglutide cost remains a barrier; metformin remains the accessible baseline.

Why it matters

Cardiovascular disease drives most of the excess death in serious mental illness, and this points to a drug class that may narrow it.

Don't overread it

This is an association from health records — a one-year mortality halving is too large to be entirely the drug.

The statistics, in plain English

A hazard ratio of 0.76 means about a quarter lower risk of death at any time point. With nearly 400,000 people, the interval is very narrow, but narrowness reflects sample size, not freedom from bias. People prescribed newer injectables may differ from others in ways records do not capture.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.