Stevenson-Hoare · BMC public health 2023 · matched retrospective cohort study · n=197,703 diabetic patients plus matched controls

Comparison of long-term effects of metformin on longevity between people with type 2 diabetes and matched non-diabetic controls.

Cited 24 times in the scientific literature.

Level 3 - non-randomized controlled study

Retrospective non-randomized matched cohort study using electronic health records.

PubMed 37131166 · doi:10.1186/s12889-023-15764-y · record verified 2026-08-30

What was done

Researchers analyzed electronic health records from the Secure Anonymised Information Linkage (SAIL) dataset in Wales, UK, tracking patients with type 2 diabetes treated with metformin (N = 129,140) or sulphonylurea (N = 68,563) alongside matched non-diabetic controls. Controls were matched on sex, age, smoking status, and history of cardiovascular disease and cancer. Survival analysis evaluated post-treatment survival across simulated follow-up windows extending up to 20 years.

What was found

Over the full 20-year observation period, type 2 diabetes patients treated with metformin had shorter survival times than matched non-diabetic controls, as did patients treated with sulphonylurea. Controlling for age, metformin-treated patients had better survival than sulphonylurea-treated patients. Metformin therapy showed a survival benefit over matched non-diabetic controls within the first 3 years of treatment, but this reversed after 5 years. Specific numerical values (such as hazard ratios, survival times, or confidence intervals) were not reported in the abstract.

Why it matters

Prior studies suggesting that metformin extends lifespan beyond that of healthy non-diabetic individuals were limited by shorter follow-up periods. These findings indicate that the apparent early survival benefit of metformin does not persist and fails to offset the long-term mortality risks of type 2 diabetes over 20 years.

Limits

The abstract reports no numerical effect estimates, confidence intervals, or p-values. As an observational study based on electronic health records, it is subject to confounding by indication, unmeasured lifestyle factors, and treatment changes over time. The exact number of matched non-diabetic controls is not reported in the abstract, and data are drawn from a single geographic region (Wales, UK).

Cited by