Most of what gets written about GLP-1 medications covers the first several months — the period where the weight loss curve is steepest and the before/after photos come from. Year two gets a lot less coverage, mostly because the data has been thin. On August 12 and 13, Epic Research published two studies that fill part of that gap, drawn from tens of thousands of real patient records. The headline is reassuring. The fine print is the part worth reading.
What the studies looked at
Epic Research pulled from Cosmos, a dataset of more than 310 million patient records across 2,100+ hospitals and 49,000+ clinics. Two cohorts were built:
- 25,213 adults without diabetes who started a GLP-1, stayed on it continuously for at least a year, and lost at least 5 pounds in that first year
- 38,621 adults with diabetes who met the same criteria
Researchers then tracked what happened to their weight from month 13 through month 24, measuring it against each patient’s own weight at the one-year mark.
What they found
For people without diabetes, the second year looked good for most who stayed on treatment. Among semaglutide patients, 53.1% held their weight within a narrow band of where they were at year one, and another 31.9% kept losing. Only 14.9% regained at least a quarter of their first-year loss. Tirzepatide patients did somewhat better on the regain measure — 52.9% held steady, 39.3% kept losing, and roughly 8% regained a quarter or more.
The diabetes cohort showed a similar shape but different numbers: 37.7% of semaglutide patients and 47.3% of tirzepatide patients maintained their year-one loss, while 26.9% and 14.2% respectively regained at least a quarter of it. A full return to starting weight by year two was uncommon in both drugs — 4.6% for semaglutide, under 2% for tirzepatide.
Put simply: among people who stuck with treatment, weight regain in year two was the exception, not the rule, and tirzepatide outperformed semaglutide on that specific measure in both groups.
What this does not establish
This is a study of people who already succeeded and stayed. Every patient in both cohorts had already lost at least 5 pounds in year one and stayed on the medication continuously for a full year with no gap longer than 60 days. That’s not a random sample of everyone who starts a GLP-1 — it’s the subset who responded early and didn’t stop. People who quit in month three because of nausea, people whose insurance dropped coverage, people who never lost meaningful weight in the first place: none of them are in this dataset. The maintenance rates above describe what happens after you clear that first bar, not the odds of clearing it. Epic Research’s own authors note this limitation directly.
EHR weight data is noisy in ways a trial isn’t. These are weights pulled from routine clinical visits, not standardized study measurements taken under the same conditions each time. Different scales, different clothing, different time of day, different visit frequency — all of that adds noise that a randomized trial controls for and an insurance claims database doesn’t.
Association, not a controlled comparison. Tirzepatide’s lower regain rate is consistent with what randomized trials have also shown — tirzepatide tends to produce more weight loss than semaglutide in head-to-head comparisons — but this particular analysis wasn’t designed to isolate why the numbers differ. Patients prescribed one drug over the other aren’t interchangeable groups; prescribing patterns, cost, and insurance formularies all shape who ends up on which medication.
It says nothing about compounded preparations specifically. Cosmos draws from prescription records tied to branded and generic dispensing in the U.S. health system, which means this dataset is dominated by the FDA-approved manufactured products — Ozempic, Wegovy, Mounjaro, Zepbound. Compounded semaglutide and tirzepatide are prepared individually by licensed 503A pharmacies and are not FDA-approved products in their own right. There’s no reason to expect the biology to behave differently, but nobody has run this exact analysis on a compounded-only population, so treat that as an open question rather than a settled one.
Why maintenance data matters more than the first-month number
A lot of GLP-1 marketing leans on the steepest part of the curve — the number at month three or month six, when the rate of loss is fastest. That number is real, but it’s also the least useful one for deciding whether this is a sustainable path for you, because it says nothing about what happens once the initial loss levels off. This data is a better — if still partial — answer to the question that actually matters: if it works for you and you stay on it, does it hold. For most people who got that far in this dataset, it did.
What it can’t tell you is whether you’ll be one of the people who gets that far, what dose you’ll need to stay there, or what your own regain risk looks like if your circumstances change. Those are the questions a licensed provider reviewing your history and lab work is positioned to answer — a national average from an insurance database isn’t built to predict one person’s trajectory, including yours.
This post is general information, not medical advice. It does not describe a guaranteed outcome of any NoBsRx program. Compounded medications are not FDA-approved. Treatment is never guaranteed — eligibility, dosing, and every treatment decision are made by independent licensed providers based on an individual clinical assessment, and availability varies by state. If you are considering starting, stopping, or changing a GLP-1 medication, talk to a licensed clinician about your own history and goals.