Metformin is a cheap, old diabetes pill that people search for as a "poor man's Ozempic". The trials say it moves weight a little — about 2% of body weight on average — and that it is not a GLP-1 drug. It is approved in the US only to control blood sugar in type 2 diabetes. Weight is not on its label, so taking it for weight alone is off-label use.
This page reports what the label and the long-running trials measured, read on 27 October 2026: the current metformin label via openFDA, the Diabetes Prevention Program (DPP) and its 15-year follow-up, a 2024 meta-analysis of 49 trials, and the 2026 mechanism research. The GLP-1 figures sit beside it with each trial's population named. They are never merged into a ranking. The wider set of non-GLP-1 options is on our page on GLP-1 alternatives.
What the label says, and what it does not
The current US label (read via openFDA; a metformin extended-release label effective 9 September 2026) indicates metformin "as an adjunct to diet and exercise to improve glycemic control in adults with type 2 diabetes mellitus." That is the whole indication.
The label does record weight, almost in passing, and the numbers are worth reading. In a 16-week placebo-controlled study of the extended-release tablets in type 2 diabetes, mean body weight changed by −1.3, −1.3, −0.7, −1.5 and −2.2 lbs across five metformin dose groups and by −1.8 lbs on placebo. In that study, metformin did not separate from placebo on weight at all. It did lower fasting glucose sharply, which is what it is approved for.
The label carries a boxed warning for lactic acidosis, a rare but serious build-up of acid in the blood, with risk factors including kidney impairment, heavy alcohol use and contrast-dye procedures. Its common side effects are gastrointestinal: in the label's placebo-controlled trials, diarrhea occurred in 10% against 3% on placebo and nausea or vomiting in 7% against 2%. It also states that in 29-week trials about 7% of patients saw previously normal vitamin B12 levels fall below normal, and asks for B12 to be checked every two to three years.
The Diabetes Prevention Program: the best weight data metformin has
The DPP (Knowler et al., NEJM 2002) randomised 3,234 adults with prediabetes — average age 51, average BMI 34.0, 68% women, 45% from minority groups — to placebo, metformin 850 mg twice daily, or an intensive lifestyle programme aiming for at least 7% weight loss. Its main question was diabetes prevention: metformin cut new cases by 31% and lifestyle by 58% against placebo. Weight was a secondary measure, and the long follow-up is what makes it useful here.
| Period | Metformin | Placebo | Source |
|---|---|---|---|
| Double-blind phase (about 2 years) | −2.06% | −0.02% | DPP Research Group, Diabetes Care 2012 |
| Open-label follow-up (to about 10 years) | −2.0% | −0.2% | same paper |
| Waist, double-blind phase | −2.13 cm | −0.79 cm | same paper |
Two findings from that paper matter for anyone reading the headline. Weight loss "was directly related to adherence" — people who kept taking the tablets lost more. And the small loss was durable: the gap with placebo held for at least ten years, which almost no weight intervention can show.
A later analysis (Apolzan et al., Annals of Internal Medicine 2019) followed the people who lost at least 5% in the first year — 28.5% of the metformin group, 62.6% of the lifestyle group and 13.4% of the placebo group. Between years 6 and 15, those originally on metformin maintained a mean loss of 6.2%, against 3.7% for lifestyle and 2.8% for placebo. The authors label it post hoc and note it compares non-randomised subsets, so it describes who kept weight off. It does not show that metformin made them keep it off.
Trials in people without diabetes
The largest pooled look is a 2024 meta-analysis of 49 randomised trials in adults with overweight or obesity and no diabetes (Haber et al., Diabetes, Obesity and Metabolism 2024). Against placebo, metformin lowered BMI by 0.56 kg/m² on average, at doses from 500 to 2,550 mg a day, and by 2.53% at 1,700 mg a day. It did no better than lifestyle programmes, and orlistat — an approved weight-loss drug — beat it. The authors describe the trials as of "low quality" with large variation in how outcomes were measured.
Put in everyday terms (our arithmetic): for a person 1.70 m tall, 0.56 kg/m² of BMI is about 1.6 kg.
Is metformin a GLP-1?
No. Metformin belongs to a class called biguanides and does not act on the GLP-1 receptor. The question is common because both lower blood sugar, both are linked to eating less, and metformin is sometimes prescribed alongside a GLP-1 drug in type 2 diabetes.
Researchers have proposed two signals for metformin's modest effect on appetite and weight, both still mechanism research, not settled explanation:
- GDF15, an appetite-suppressing hormone. Metformin raises its blood level, and in mice the gut lining, not the liver, supplied most of that rise (Kincaid et al., Sci Rep 2024; the state of the evidence is reviewed in Nature Reviews Endocrinology 2023).
- Lac-Phe (N-lactoyl-phenylalanine), a metabolite that also rises after hard exercise. A paper in Cell on 22 October 2026 identified a gut transporter, MCT6, that exports it into the blood; mice lacking MCT6 were resistant to metformin-associated weight loss (Cell 2026). That is a mouse finding.
How the GLP-1 drugs themselves work is on our page on how GLP-1 drugs work.
Beside the GLP-1 trial figures
These numbers come from separate trials in different populations. They are set side by side so the scale is visible, not ranked.
| Medicine | Trial and population | Length | Weight change vs placebo |
|---|---|---|---|
| Metformin 850 mg twice daily | DPP, adults with prediabetes, BMI 34 | ~2 years | −2.06% vs −0.02% |
| Metformin (pooled doses) | 49 trials, adults without diabetes | ≥3 months | BMI −0.56 kg/m² vs placebo |
| Semaglutide 2.4 mg weekly (Wegovy) | STEP 1, adults with obesity, no diabetes (NEJM 2021) | 68 weeks | −14.9% vs −2.4% |
| Tirzepatide 15 mg weekly (Zepbound) | SURMOUNT-1, adults with obesity, no diabetes (NEJM 2022) | 72 weeks | −20.9% vs −3.1% |
The difference is large enough that the different trial designs cannot close it. It still remains a comparison across trials, not a measurement from one. The question people actually face is usually cost and access, set out on our pages on the cheapest GLP-1 without insurance and how to get a GLP-1 prescription.
Where metformin's weight story fits
- Prediabetes. The DPP is the reason metformin is discussed for prediabetes at all; its diabetes-prevention results and the GLP-1 trial data in people with prediabetes are on GLP-1 drugs for prediabetes.
- PCOS. Metformin is prescribed off-label in polycystic ovary syndrome; the GLP-1 trials there are on GLP-1 drugs for PCOS.
- Type 2 diabetes. Metformin is usually the first diabetes medicine, and GLP-1 drugs are often added to it — the SUSTAIN trials on our Ozempic for type 2 diabetes page were run largely on top of metformin.
- "Natural" GLP-1 claims. Supplements sold as metformin-like or GLP-1-like, such as berberine, are covered on natural GLP-1.
What the record does not show
- No trial has tested metformin head to head against a GLP-1 drug for weight loss in people without diabetes.
- The 49-trial meta-analysis is built on trials its own authors rate as low quality.
- The 15-year maintenance figure comes from a self-selected subgroup.
- The mechanism work on GDF15 and Lac-Phe does not yet predict who will lose weight on metformin.
Whether metformin, a GLP-1 drug, another medicine or none suits a person is a decision for their doctor, who can weigh kidney function, other medicines and B12 status.
Sources
Read 27 October 2026: metformin hydrochloride extended-release label via openFDA (effective 2026-09-09; indication, boxed warning, sections 5.2, 6.1 and 14); Knowler et al., NEJM 2002 (PMID 11832527); Diabetes Prevention Program Research Group, Diabetes Care 2012 (PMID 22442396); Apolzan et al., Ann Intern Med 2019 (PMID 31009939); Haber et al., Diabetes Obes Metab 2024 (PMID 38468148); Kincaid et al., Sci Rep 2024 (PMID 38253650); Nat Rev Endocrinol 2023 (PMID 36271034, cited as a review); Cell 2026 (PMID 42913628); Wilding et al., NEJM 2021 (PMID 33567185); Jastreboff et al., NEJM 2022 (PMID 35658024). Corrections go to the contact page.