Prediabetes medication: what the prevention trials showed

Read the populations before comparing rows. The DPP, STOP-NIDDM and ACT NOW enrolled people selected for high glucose. The semaglutide and tirzepatide results come from weight-management trials in people with obesity or overweight ^[6][8], and SELECT enrolled people who already had heart disease ^[7]. No trial has compared these drugs head to head for diabetes prevention, so the table cannot rank them.

The DPP is the trial behind almost every metformin-for-prediabetes discussion. Over an average of 2.8 years, new diabetes occurred at 11.0 cases per 100 person-years on placebo, 7.8 on metformin and 4.8 with the intensive lifestyle programme ^[2]. The lifestyle programme was significantly more effective than metformin. Put another way, about 6.9 people had to follow the lifestyle programme for three years to prevent one case, against 13.9 on metformin ^[2].

The long follow-up tells a more modest story. After 15 years, the reduction was 27% for the lifestyle group and 18% for metformin, and cumulative diabetes incidence was 55%, 56% and 62% for lifestyle, metformin and placebo ^[3]. Combined small-vessel complications (kidney, eye and nerve disease) did not differ between the groups overall ^[3]. Metformin delayed diabetes; it did not stop the condition progressing in most people.

The ADA Standards of Care 2026 put a structured lifestyle programme modelled on the DPP first ^[1]. For medicine, they say metformin should be considered in adults at high risk of type 2 diabetes, especially:

• people aged 25 to 59 with a BMI of 35 kg/m² or more;
• people with higher fasting glucose (about 110 mg/dL or more) or higher A1C (about 6.0% or more);
• people with a history of gestational diabetes ^[1].

The same section suggests periodic vitamin B12 checks in people on metformin long term, since metformin can lower B12 levels ^[1]. It also notes that this use of metformin is not FDA-approved ^[1].

Not every diabetologist agrees that metformin belongs in prediabetes at all. A 2020 Diabetes Care commentary argued against it: about two-thirds of people with prediabetes never develop diabetes, about one-third return to normal glucose, and prediabetes-range glucose does not cause the small-vessel complications of diabetes ^[9]. Its author proposed close follow-up of the highest-risk groups, with metformin started at diagnosis of diabetes rather than before ^[9].

No. Metformin's FDA label covers glycemic control in type 2 diabetes ^[10], and the ADA notes that its use for prevention is not an approved indication ^[1]. Pioglitazone's label is likewise for type 2 diabetes ^[11]. Semaglutide and tirzepatide are approved for type 2 diabetes ^[12][13]; their prevention results came from trials of weight management in people with obesity or overweight ^[6][8], and none of these labels lists prediabetes.

That matters for two practical reasons that searchers ask about: prescribing in prediabetes is a clinician's off-label judgment, and US insurance coverage often follows the label. The site does not track coverage. The semaglutide and tirzepatide records list what each drug is approved for.

Acarbose slows carbohydrate digestion. In STOP-NIDDM it lowered the relative hazard of diabetes to 0.75, but 31% of participants on acarbose stopped treatment early, against 19% on placebo ^[4].

Pioglitazone produced the largest relative reduction of any tablet, 72% in ACT NOW, at the cost of weight gain and edema ^[5]. Its label carries a boxed warning that it can cause or worsen heart failure ^[11].

Semaglutide and tirzepatide produced the largest absolute effects, in people with obesity. In SELECT, semaglutide held A1C 0.32 points below placebo and more people returned to normal glucose, 69.5% vs 35.8% at week 156; weight loss explained about a quarter of the effect ^[7]. The trial authors concluded that semaglutide did not slow the gradual rise in glucose over time ^[7]. In SURMOUNT-1, 1.3% of the tirzepatide group had diabetes at the end of 176 weeks of treatment; after 17 more weeks off the drug the figure was 2.4%, against 13.7% on placebo ^[8]. The effect lasts while the drug is taken; the trials do not show that it persists after stopping.

The amylin drugs covered elsewhere in this guide have no prevention trial result here. CagriSema is not approved; Novo Nordisk filed it with the FDA for weight management in December 2025 ^[14].

The trials answer a narrower question: which medicine delayed diabetes, in whom, and at what cost. On that evidence:

• Longest record: metformin, with 15 years of follow-up and an inexpensive generic ^[3].
• Largest effect in people with obesity: tirzepatide and semaglutide, in trials that were testing weight loss ^[8][7].
• Better than metformin in the one direct comparison: the intensive lifestyle programme ^[2].

There is no single best prediabetes medication, because no trial has compared the drugs with each other and the people studied differ. How clinicians weigh weight, age, glucose level, gestational-diabetes history and heart risk is set out in the ADA recommendations above ^[1]. Where prediabetes has already become diabetes, the type 2 diabetes remission guide covers what the trials show about reversing it, and glucose pills covers the tablets.