Metformin within 6 days of COVID-19 linked to 53% lower risk of long COVID or death in health records

An observational analysis of electronic health records, published in the Journal of Medical Virology, found that metformin prescribed within 6 days of documented SARS-CoV-2 infection was associated with a 53% lower risk of long COVID or death over 6 months than comparator medications wiley.comJournal of Medical Virologywiley.comOpen source ↗. After weighting, each group comprised 248 patients: 10 (4.0%) in the metformin group and 21 (8.5%) in the control group developed long COVID or died. The data came from a nationwide US database of more than 21 million people with documented infection.

Study design and patient groups

The authors used a new-user design framed as a target trial emulation. Metformin was compared with a prescription for fluvoxamine, fluticasone, ivermectin or montelukast. Patients had to have a clinic visit in both the 0–6 and 6–12 months before infection. Those who had received metformin or a comparator in the previous 12 months were excluded, as were patients with documented indications for chronic metformin use, including diabetes and prediabetes. Follow-up was 180 days.

Long COVID was defined by a diagnosis code or a computable phenotype, and death was included in the outcome as a competing risk. The average treatment effect was estimated with entropy balancing and a weighted log-linear model. After weighting, the average age was 53 (16); 16% of patients were Black and 16% were Hispanic.

Results and limitations

The adjusted risk ratio (aRR) was 0.47 (95% CI 0.25–0.89, p = 0.02). According to the authors, any risk reduction between 75% and 11% is highly compatible with the data wiley.comJournal of Medical Virologywiley.comOpen source ↗. For prescriptions on Days 0–1 after infection the aRR was 0.39 (95% CI 0.12–1.24), and for Days 0–14 it was 0.75 (95% CI 0.52–1.08). Both of these intervals include 1.

The authors note the limits of an observational design: missing data and misclassification of exposures, outcomes and confounders. Vaccination data were not available. Nirmatrelvir–ritonavir could be obtained at pharmacies without a visit and may be missing from the records. Home testing may have left infection counts and dates incomplete. Higher glycated hemoglobin in the metformin group suggests it may have included people with undiagnosed diabetes or prediabetes.

An earlier randomized trial of metformin restricted enrollment by age and body mass index, and only 22% of its participants were enrolled in the Omicron era. The authors say that this analysis, based on diagnoses in health records, matters for the reproducibility of trial results in which the same outcome was ascertained by participant report. They state that further clinical trials are needed.