The Diabetes Prevention Program randomised 3234 adults with prediabetes to intensive lifestyle intervention, metformin or placebo starting in 1996. This is an observational follow-up of 1173 of them who consented to linkage of US Medicare claims through 2021, at a median age of 74. The outcome was multimorbidity, defined as two or more of 15 chronic conditions.
By the end of follow-up 997 participants (85%) had at least two conditions, with a median of five. Compared with placebo, the lifestyle group had a lower risk of multimorbidity (hazard ratio 0.79, 95% CI 0.68 to 0.93) after adjustment. Metformin showed no difference (HR 0.91, 95% CI 0.78 to 1.07). The lifestyle effect held when diabetes itself was excluded from the definition, and was larger when restricted to pairs of the costliest conditions (HR 0.57, 95% CI 0.38 to 0.85).
The finding worth carrying is that a time-limited behavioural programme run decades ago was still associated with less accumulated chronic disease in the eighth decade of life, and that this was not simply diabetes prevention by another name. Metformin, which does prevent diabetes in this cohort, did not deliver the same breadth of benefit - a reminder that a drug which fixes one pathway does not do the work of changing how someone lives.
In clinic this is an argument for treating the lifestyle referral as a real intervention with a real dose, not as a line in the discharge letter. It is also a caution about how these results were obtained: this is an observational follow-up of consenting survivors, so the people analysed are healthier and more engaged than the original cohort, and the randomisation no longer fully protects the comparison. The direction is convincing; the exact size is not.
- Name the lifestyle programme, the contact frequency and who runs it when you refer - a vague instruction to lose weight is not the intervention studied
- Set and record a weight target and a physical activity target, and review both at the next visit
- Do not treat metformin as a substitute for a structured programme in prediabetes
- Screen for the other conditions that make up multimorbidity - hypertension, dyslipidaemia, chronic kidney disease, depression - rather than tracking glucose alone
- In Indian practice, screen for dysglycaemia from a younger age and at a lower BMI than Western thresholds suggest
The statistics, in plain English
A hazard ratio of 0.79 means the lifestyle group accumulated multimorbidity at about four-fifths the rate of placebo over the follow-up; the interval from 0.68 to 0.93 sits entirely below 1.0, so a real effect is likely. Metformin's interval, 0.78 to 1.07, crosses 1.0 - the data are compatible with a modest benefit, no effect, or slight harm, so no conclusion can be drawn either way. The striking HR of 0.57 for the costliest condition pairs comes from a smaller subgroup with a much wider interval (0.38 to 0.85) and should be read as a signal, not a measurement. Most importantly, this stage of the study is observational: only 1173 of the original 3234 contributed data, and consent to record linkage is not random, so unmeasured differences between the groups can no longer be ruled out.
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