Within twelve weeks of its approval for weight management, the semaglutide pill was being started almost as often as the injection. A research letter in JAMA Network Open, drawn from the medical records of nearly 48 million US adults who qualified for an obesity drug, counted 148,792 people who began oral semaglutide between 22 December 2025 and 15 March 2026. Just over half had never taken a GLP-1 drug before. Tirzepatide injections were still the most commonly started treatment throughout.
What the records showed
The University of Pennsylvania team used Epic Cosmos, a pooled database of electronic health records from more than 2,000 hospitals. They counted adults with a body mass index of 30 or more, or 27 or more with a weight-related condition such as high blood pressure or sleep apnea, and excluded anyone with type 2 diabetes so the count would reflect weight management. That left 47,894,786 eligible adults.
Over the twelve weeks, 148,792 of them started oral semaglutide, against 227,040 who started injectable semaglutide and 355,513 who started injectable tirzepatide. Of the oral starts, 87.6% were the Wegovy pill and 12.4% Rybelsus, the diabetes version of the same drug. The researchers measured new starts per 10,000 eligible adults per week. Oral semaglutide rose from almost zero to a peak of 4.56 per 10,000 (95% CI 4.50 to 4.62) in the week of 2 March, which briefly put it level with, and then slightly above, the injectable form. Tirzepatide ran at roughly 5 to 7 per 10,000 a week across the whole period.
Who took the pill differed a little. Oral starters averaged 53.3 years old, against 49.7 for injectable semaglutide and 50.5 for tirzepatide; 72.7% were women; their average BMI was 36.6, a point or so lower than the injection groups; and 13.4% were recorded as on Medicare, against 9.3% and 10.2%. Of the 148,792, 78,861 (53.0%) had no GLP-1 prescription in the previous year, which means close to half were switching from something else, most often an injection.
Expansion, switching, and what the records miss
The finding the authors highlight is that the pill did both jobs at once: it brought in people who had not used these drugs, and it moved existing users off injections. It does not say how either group fared. The pill and the injection are the same molecule but are not interchangeable milligram for milligram, and the approved doses differ on each label. The letter also notes that orforglipron, a second oral option, was approved in April 2026, after this study window closed.
The bigger caution is about where people actually get these drugs. The letter counts prescriptions written inside health systems that use Epic. Only 0.8% of oral starters were recorded as self-pay, and that column reflects insurance on file, not who paid at the pharmacy. Prescriptions from online and direct-to-consumer telehealth services, a route many cash-paying buyers use, were mostly not captured, as the authors acknowledge. If the pill has been especially popular through those channels, this study undercounts it. Our guide to what these drugs cost without insurance covers the self-pay routes the records leave out.
What is not yet known
This is a short, descriptive research letter. It counts prescriptions written, not prescriptions filled or doses taken, and the records do not state why a drug was prescribed. No statistical testing was done, and nothing was adjusted for, so the differences between groups describe who got which drug rather than why. Twelve weeks says nothing about how many people stay on the pill, which is the long-standing weakness of GLP-1 treatment in general, or whether people who switched from injections kept their weight off. The authors call for longer studies of persistence and outcomes; until they report, early popularity is the only thing this establishes. The study was funded by the National Institutes of Health; one senior author reported unrelated consulting fees.
Sources: Lei Y, Tang H, Jian X, Asch DA, Chen Y. Early Adoption and Diffusion of Oral Semaglutide in US Adults With Obesity. JAMA Network Open, research letter, published 7 October 2026 (open-access full text read via PubMed Central PMC13647145).
