Which of the two is stronger is the most-asked question in this category and the one with the least direct evidence behind it. The two drugs were tested against placebo, not against each other, so the comparison usually offered is an eyeball across separate trials with different populations. The study below is a better attempt than that, and still not the thing people take it for. It also sits upstream of a price question, which we treat separately in who publishes a dose ladder.
What the study did
Researchers used electronic health records linked to dispensing data across a group of US health systems to find adults with overweight or obesity who started either drug between May 2022 and September 2023 [1]. Everyone included had a year of regular care beforehand, no prior GLP-1 use, a prescription within 60 days of starting, and a recorded baseline weight.
Of 41,222 adults meeting the criteria — 32,029 on semaglutide and 9,193 on tirzepatide — 18,386 remained after propensity score matching. Mean age was 52.0, 70.5% were female, 52.0% had type 2 diabetes, and mean baseline weight was 110 kg. Both drugs were in formulations labeled for type 2 diabetes, prescribed on or off label.
What it found
On treatment, people receiving tirzepatide were significantly likelier to reach every weight threshold measured: a hazard ratio of 1.76 (95% CI 1.68 to 1.84) for losing 5% or more, 2.54 (95% CI 2.37 to 2.73) for 10% or more, and 3.24 (95% CI 2.91 to 3.61) for 15% or more.
Weight change itself diverged steadily: a difference of −2.4% at three months (95% CI −2.5 to −2.2), −4.3% at six (95% CI −4.7 to −4.0) and −6.9% at twelve (95% CI −7.9 to −5.8). Rates of gastrointestinal adverse events were similar between the groups, which is the finding most likely to surprise anyone who has read the two drugs’ separate trial write-ups.
What a cohort study cannot do
Nobody was randomized. Prescribers chose which drug to write, and the reasons they chose — insurance, supply, comorbidity, how motivated the patient seemed — are not fully in the record and not fully removable by matching. During the window studied, availability of both drugs was erratic, which is exactly the kind of factor that correlates with who gets what.
The direction is nonetheless consistent with what the placebo-controlled trials imply separately, and the gradient across thresholds is hard to produce by confounding alone. The honest summary is that tirzepatide probably does more, by an amount this study estimates rather than establishes.
What this changes about choosing
Less than it looks. Whether a seller will tell you the dose, what the price does as that dose climbs, and whether you can reach anyone are questions that apply identically to both molecules, and they decide more outcomes than the choice between them — which is why the questions they leave open is molecule-agnostic. If you are choosing on format rather than molecule, the oral evidence is in what the pill actually does, and the question of what happens when either drug stops is in stopping and weight regain.