The Benefit That Stops

The study appeared in BMJ Medicine on March 18, 2026, alongside a Washington University in St. Louis release; ScienceDaily carried its account on September 21. Researchers followed 333,687 U.S. veterans with type 2 diabetes for up to three years. Continuous GLP-1 treatment was associated with lower cardiovascular risk; interruptions and discontinuation were associated with less protection.

The study, by Yan Xie, Taeyoung Choi and Ziyad Al-Aly in BMJ Medicine, is described as a target-trial emulation. That means an observational analysis designed to approximate a trial, not an experiment randomly assigning people to stop their medication. The distinction belongs near the beginning, before a percentage starts impersonating destiny.

According to the release, 132,551 participants had been prescribed GLP-1 drugs and 201,136 had been prescribed sulfonylureas, another class of diabetes medicines. Researchers reassessed GLP-1 treatment status every six months. Twenty-six percent of GLP-1 users stopped altogether; roughly another 23 percent had a gap of at least six months before restarting.

Those who continued GLP-1 treatment throughout the three years had an 18 percent lower risk of major cardiovascular events than those taking sulfonylureas, the release reports—about four fewer events per 100 people over that period. The account describes the outcome as including heart attack, stroke and death.

The discontinuation comparison used a different reference group. After two years off GLP-1 treatment, risk was reported as 22 percent higher than among continuous users. That is a relative difference, not a claim that 22 percent of people who stop will have an event, nor an increase of 22 percentage points. Keeping the denominator attached is less exciting than a frightening headline. It is also how numbers remain information.

These findings concern veterans with type 2 diabetes. They cannot simply be extended to everyone using these medicines for weight management or other indications. Differences between people who continue and people who stop can complicate observational comparisons. The full publisher article was blocked during checking, so this account relies on the accessible university release rather than an independent assessment of its detailed methods.

Still, the question the report raises is larger than whether patients follow instructions.

Al-Aly identifies cost, side effects and shortages as reasons people stop, and calls for health systems to support continued treatment. The release does not establish how much each barrier explains the observed outcomes. But those barriers change what “continuity” asks of someone. A prescription is not the same thing as an affordable supply, manageable side effects and a clinician available when something goes wrong.

Here the coherenceist question is concrete: who must hold everything together for a treatment to remain possible? If continuity matters, care cannot end at the moment a medicine is prescribed. Nor should a population-level association become pressure to persist with a treatment regardless of an individual's circumstances. Decisions about continuing, stopping or restarting belong with the patient and their treating clinician.

The useful response is neither panic nor a lecture about willpower. It is to make the ongoing work visible—and share it. Biology has declined to provide a finish line. Health care can at least avoid leaving people to run the rest of the course alone.