Phase 3 obesity trials barely enroll anyone in their eighties, so for this age group the randomized evidence essentially does not exist and never will. That makes a matched cohort with a mean age of 81.6 the best available answer rather than a second-rate one [1]. Who reaches these drugs at all, and at what age, is the subject of who the price filtered out.
Across up to five years of follow-up, GLP-1 users had lower all-cause mortality (HR 0.82, 95% CI 0.77–0.88), fewer major cardiovascular events (HR 0.86), fewer major kidney events (HR 0.86) and fewer hospitalizations of any kind (HR 0.91). Taken at face value that is a substantial benefit in the oldest patients anyone prescribes to.
The selection problem in this age band is severe and specific. Starting an 82-year-old on a weekly injection that must be stored, drawn and self-administered, and that causes nausea and weight loss, is a decision clinicians make for patients who have years ahead of them, manage their own medication and are not visibly declining. Choosing a DPP-4 tablet instead is often a decision about someone frailer. An 18% mortality difference is roughly what that sorting would produce with no drug effect at all — the same structure as the falls and fractures cohort, which used the same comparator in an overlapping population.
That does not make the finding worthless. Propensity matching on recorded characteristics removes the crude differences, the direction is consistent with everything else known about the class, and a 9% reduction in all-cause hospitalization is the kind of outcome that is hard to manufacture out of selection alone. It does mean the honest reading is that these drugs appear not to harm very old patients and may help them, which is a weaker claim than the hazard ratios look.
For anyone buying on behalf of an older parent, the practical questions are the ones the study cannot answer: whether they can manage the injection, whether weight loss at that age costs muscle they cannot spare, and whether anyone is monitoring. Those are covered from the cost side in covered and still not treated and from the clinical-support side in what patients said when asked.