Less Than Half Are Still Taking It

This register has spent a great deal of effort on differences of a few percentage points between drugs. Here is the number that dwarfs all of them. Of 393 people prescribed semaglutide for weight loss in one health system, 44% were still taking it a year later — and those who were lost 13.8% of their weight against 6.4% for those who were not.^[1]

What was measured

Persistence was defined strictly but reasonably: no gap of more than 60 days in twelve months. Adherence used a separate standard — having medication on hand at least 80% of days. Both landed in the same place.

Thirteen point eight percent against six point four. Nothing else on this site produces a gap that size.

The 13.8% figure is worth noticing on its own. It sits close to what the trials report, which means the drug works in ordinary care roughly as advertised — for the minority who keep taking it. The gap between trial results and real-world results is largely a gap in continuation, not in pharmacology.

How you climb the ladder

That is one of the few genuinely actionable findings in this whole literature, and it sits beside another. A randomized trial found that a slower, flexible titration cut withdrawal for gastrointestinal side effects from 19% to 2% while reaching similar final doses and similar results — five clicks instead of a quarter milligram.

Put together: erratic titration produces worse results, and deliberately slower titration keeps people on treatment without costing efficacy. How someone climbs the ladder appears to matter more than which ladder they are on — a conclusion no drug comparison would ever surface.

One caution on direction. People who titrate irregularly may be doing so because of side effects, cost interruptions or supply gaps — so poor titration may be a symptom of the same problems that cause poor results rather than the cause of them. An observational cohort cannot separate those.

The finding that runs against expectation

Women had lower odds of both persistence and adherence than men — odds ratios of 0.32 (95% CI 0.14–0.75) and 0.33 (0.14–0.77). Race, ethnicity and comorbidities showed no significant association with either.

The study does not explain it, and neither will we. ★ It is worth reporting precisely because it is unexpected — and it has since turned up again, in a Scottish national cohort where women were more likely to be classified as intolerant specifically: not taking it and not tolerating it. Different country, different measure, same direction.

What it means for reading everything else

Almost every comparison on this register — 6.5 percentage points for adding cagrilintide, 6.5 for tirzepatide over semaglutide, 3.1 for tripling a dose, 0.75 of an HbA1c point against a competitor — assumes the person is taking the drug. In this cohort, most were not, for most of the year.

• Tolerability outranks efficacy in determining outcome, because a drug someone stops delivers nothing.
• Cost and coverage are clinical variables, not administrative ones, and this cohort spans 2021–2024 — a period of shortages and shifting coverage, which is part of what “persistence” is measuring here.
• Support during the first year is where the leverage is. The authors’ own recommendation is exactly that.
• It is one health system and 393 people. Persistence figures vary widely by setting, insurance and country.

For what happens to weight after people stop, see what happens when you stop; for what those who stopped actually did next, what people actually do after stopping.