This is an observational follow-up of a randomised trial, not a fresh trial. Between 1996 and 1999, 3,234 adults with prediabetes were randomly assigned in the Diabetes Prevention Program to intensive lifestyle intervention, metformin or placebo. Medicare claims data through 2021 were available for 1,173 of them who consented, by which point the median age was 74 years and 68% were women. The outcome was multimorbidity, defined as at least two of 15 chronic conditions.
By the end of follow-up, 997 people (85%) had accumulated at least two conditions, with a median of five. Those randomised to lifestyle were less likely to reach that point than those on placebo (hazard ratio 0.79, 95% CI 0.68-0.93). Metformin showed no such difference (HR 0.91, 95% CI 0.78-1.07). When the analysis was restricted to pairs of the costliest conditions, the lifestyle effect was larger (HR 0.57, 95% CI 0.38-0.85). Crucially, the finding held when diabetes itself was removed from the definition, so this is not simply delayed diabetes being counted twice.
This gives a concrete answer to the patient who asks what a prevention programme actually buys them. It is a longer and broader payoff than the original trial measured, and it is the sort of number worth quoting when someone is deciding whether the effort is worth it. The metformin result is not a reason to stop metformin: the original trial showed it prevented diabetes. It simply did not spread to the wider burden of chronic illness the way structured lifestyle change did.
- When counselling someone with prediabetes, describe the benefit in terms of chronic illness accumulated over decades, not only progression to diabetes
- Record whether a patient has ever completed a structured lifestyle programme, not just whether they were advised to lose weight
- Do not read this as a reason to withhold or withdraw metformin in prediabetes; its diabetes-prevention effect is separate and established
- Offer periodic re-entry into lifestyle support rather than a single course; this cohort had booster sessions for years
- For Indian patients, where diabetes onset is roughly a decade earlier, the window in which prevention can act is longer, so start these conversations at younger ages
The statistics, in plain English
A hazard ratio of 0.79 means the lifestyle group reached multimorbidity at about four-fifths the rate of placebo over follow-up. Its confidence interval, 0.68 to 0.93, sits entirely below 1.0, so the reduction is unlikely to be chance. Metformin's interval, 0.78 to 1.07, crosses 1.0: the data are compatible with a modest benefit, no effect, or a slight harm, so the honest reading is that no effect was shown, not that an effect was ruled out. Note the design limit as well. Only 1,173 of the original 3,234 participants had linked claims data and consented, and treatment was unmasked after the trial ended, so this is an association within a randomised cohort rather than a randomised result. The larger effect on the costliest condition pairs (0.57) rests on fewer events, which is why its interval is much wider.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for diabetes & endocrinology, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free