A records study of patients who stayed on the lowest prescribed dose of semaglutide or tirzepatide for at least six months found modest weight loss on both, more with tirzepatide, published in Biology Methods & Protocols. The abstract lists a string of side effects that were more common on one drug or the other, but in the results section none of those differences survived the authors’ own correction for testing 461 conditions at once. The study used prescription records as a stand-in for so-called microdosing and says plainly that it cannot confirm anyone was microdosing.
What the researchers did
The authors, all employees of the data company nference, searched de-identified health records covering 29 million patients at US academic medical centres. Of 490,072 people ever prescribed semaglutide, 814 had only 0.25 mg prescriptions, at least three of them, spanning six months or more. Of 322,442 people prescribed tirzepatide, 1,016 met the same test at 2.5 mg. Those are the labelled starting doses, which the paper notes are meant for initiation before escalation. After matching on age, sex, body mass index, diabetes status and start year, 534 patients per drug were compared. They averaged about 50 years old, roughly 73% were women, average BMI was 36 to 37, and fewer than one in ten had type 2 diabetes. The median patient had three or four prescriptions over about a year.
Mean weight change at 12 months was −2.2% on low-dose semaglutide and −5.5% on low-dose tirzepatide, and the gap held at every time point. For a 100 kg person that is about 2.2 kg against 5.5 kg. A separate comparison of 753 matched pairs within semaglutide put the low-dose group at −1.4% at 15 months against −12.4% for patients who reached 2.0 mg or more, with more nausea, constipation and hunger in the higher-dose group.
Across 104 diagnosis-code outcomes, nothing differed significantly between the two drugs. In clinical notes, read by software that the authors checked against 100 hand-reviewed documents (93% accuracy), 22 of 461 conditions differed at P < 0.05, including constipation (32.6% vs 22.4%) and acute kidney injury (2.7% vs 0.4%) on tirzepatide and migraine on semaglutide. After false-discovery correction, every one had a q-value above 0.5. When 461 conditions are tested, around 23 would be expected to cross P < 0.05 by chance alone, which is about what was found.
How this fits
Interest in staying on low doses has grown alongside marketing of compounded products, a point the paper itself makes; our guide to compounded semaglutide covers that market. The paper says there is no standardised microdosing protocol, and this study does not supply one. What it adds is the first published look at people who stayed on starting doses in routine care: the weight effect was small, and the side-effect comparison is a list of leads rather than findings. For the side effects established in randomised trials, see our pages on semaglutide and tirzepatide side effects.
In a separate exploratory search of notes mentioning semaglutide and microdosing, 41 of 451 patients had chart-confirmed microdosing. Of those 41, 22 had a documented weight benefit, 16 had no clear documented benefit and 10 had adverse effects or poor tolerability recorded.
What is not yet known
A prescription is not a fill and a fill is not an injection. Three prescriptions in a year can mean deliberate low dosing or simply irregular use, and the authors say they cannot tell those apart. Body mass index stayed modestly unbalanced after matching, and residual confounding is likely in any records study. The authors work for a company that collaborates with drugmakers, though they state none funded or shaped this work. Every adverse-event difference should be read as unconfirmed.
Sources: Murugadoss K, Venkatakrishnan AJ, Soundararajan V. “Comparison of cardiometabolic benefits and tolerability in patients on sustained low doses of semaglutide or tirzepatide.” Biology Methods & Protocols 2026;11(1):bpag048, published 31 August 2026, indexed in PubMed October 2026. doi:10.1093/biomethods/bpag048 (open-access full text, PMC13651699).
