Among 52,341 English adults newly diagnosed with type 2 diabetes, the 19,599 who started metformin within six months showed no protection against severe COVID-19. Their hazard ratio for COVID-related hospitalisation or death was 1.17 (95% CI 1.02-1.35) in the intention-to-treat analysis, and 1.11 (0.96-1.29) per-protocol. Long COVID risk was imprecisely estimated (HR 1.39, 0.83-2.33). Bone fracture, a negative control outcome, showed no association, while diabetes-related death, a positive control, was elevated.
Earlier observational studies reported lower COVID mortality among metformin users, fueling hopes of a cheap, repurposed protective drug. The new-user design here directly addresses their main flaw: comparing long-term survivors on metformin with heterogeneous controls. The result lines up with randomised trials, which found little or no mortality benefit. The elevated positive control signals residual confounding, probably because clinicians prescribe metformin to patients with higher baseline glucose or poorer health, so the slight excess risk likely reflects who gets treated, not metformin harming patients. The wide Long COVID interval means that question stays open. For adults with type 2 diabetes, the practical message is to take metformin for glycaemic control and cardiovascular indications, not as COVID protection. This is a preprint that has not been peer reviewed, so estimates may shift. As an assessment, the work is confirmatory but methodologically valuable, helping retire a hypothesis that prevalent-user bias inflated.