At one urban clinic, patients on semaglutide lost about 6.5 percent of their body weight over a year. Patients on dulaglutide, an older drug in the same family, lost barely 1 percent.
That gap comes from a retrospective study published in the American Journal of Managed Care ↗, which compared the two weekly injections in 120 adults with type 2 diabetes at an academic internal medicine clinic. The patient group was urban and predominantly African American, a population the head-to-head trials that defined these drugs barely included.
Both drugs switch on the GLP-1 receptor ↗, the gut-hormone signal that curbs appetite and improves blood sugar. Dulaglutide ↗, sold as Trulicity, reached the market first and is the weaker of the two on weight. Semaglutide ↗, sold as Ozempic, came later and hits harder. A randomized head-to-head trial, SUSTAIN 7, already showed semaglutide beating dulaglutide on both weight and blood sugar back in 2018. What this study tested is whether that edge survives outside a controlled trial, in patients who look different from the ones SUSTAIN 7 enrolled.
It did, and by a wide margin. Over 52 weeks, semaglutide patients lost 6.51 percent of their body weight against 1.14 percent for dulaglutide (p = 0.001). The responder numbers are starker. Just over half the semaglutide group (51.7 percent) lost at least 5 percent of their weight, the threshold doctors treat as clinically meaningful, versus a quarter of the dulaglutide group (25 percent). For a 10 percent loss the split was 31.7 percent against 8.3 percent, close to a fourfold difference. Blood sugar, measured by HbA1c, fell more on semaglutide too (a 2.15 versus 1.36 percentage-point drop, p = 0.033), though the share of patients reaching target HbA1c was not statistically different between the groups.
The reason the population matters is not politeness. Drug response can differ across groups for reasons of genetics, diet, body composition, and access to care, and a result measured almost entirely in white cohorts, as SUSTAIN 7 and most pivotal GLP-1 trials were, is a weaker guide for a patient who was not in those cohorts. A study run in a group the pivotal trials underrepresented is doing real work even when it confirms the expected direction.
The limits are obvious and the authors do not hide them. This is 60 patients per drug at a single clinic, looked at after the fact rather than randomized, which is how confounding sneaks in: the sicker or heavier patient may have been steered toward the stronger drug in the first place. It is a signal, not a verdict. But it is a signal pointing the same way the trials did, in the patients the trials left out, which is exactly the kind of confirmation real-world data is supposed to provide. ↯