Accumulating observational data suggest that metformin use in patients with type 2 diabetes might reduce the risk for death from COVID-19, but the randomized trials needed to prove this are unlikely to be carried out, according to experts.
The latest results, which are not yet peer reviewed, were published online July 31. The study was conducted by Andrew B. Crouse, PhD, of the Hugh Kaul Precision Medicine Institute, University of Alabama at Birmingham, and colleagues.
The researchers found that among more than 600 patients with diabetes and COVID-19, use of metformin was associated with a nearly 70% reduction in mortality after adjustment for multiple confounders.
Data from four previous studies that also show a reduction in mortality among metformin users compared to nonusers were summarized in a “mini review” by André J. Scheen, MD, PhD, published Aug. 1 in Diabetes and Metabolism.
Dr. Scheen, of the division of diabetes, nutrition, and metabolic disorders and the division of clinical pharmacology at Liège (Belgium) University, discussed possible mechanisms behind this observation.
“Because metformin exerts various effects beyond its glucose-lowering action, among which are anti-inflammatory effects, it may be speculated that this biguanide might positively influence the prognosis of patients with [type 2 diabetes] hospitalized for COVID-19,” he said.
“However, given the potential confounders inherently found in observational studies, caution is required before drawing any firm conclusions in the absence of randomized controlled trials,” Dr. Scheen wrote.
Indeed, when asked to comment, endocrinologist Kasia Lipska, MD, of Yale University, New Haven, Conn., said in an interview: “Metformin users tend to do better in many different settings with respect to many different outcomes. To me, it is still unclear whether metformin is truly a miracle drug or whether it is simply used more often among people who are healthier and who do not have contraindications to its use.”
She added, “I don’t think we have enough data to suggest metformin use for COVID-19 mitigation at this point.”
Alabama authors say confounding effects ‘unlikely’
In the retrospective analysis of electronic health records from their institution, Dr. Crouse and colleagues reviewed data from 604 patients who were confirmed to have tested positive for COVID-19 between Feb. 25 and June 22, 2020. Of those individuals, 40% had diabetes.
Death occurred in 11% (n = 67); the odds ratio (OR) for death among those with, vs. without, diabetes was 3.62 (P < .0001).
Individuals with diabetes accounted for >60% of all deaths. In multiple logistic regression, age 50-70 vs. <50, male sex, and diabetes emerged as independent predictors of death.
Of the 42 patients with diabetes who died, 8 (19%) had used metformin, and 34 (81%) had not*, a significant difference (OR, 0.38; P = .0221). Insulin use, on the other hand, had no effect on mortality (P = .5728).
“In fact, with 11% [being] the mortality of metformin users, [this] was comparable to that of the general COVID-19-positive population and dramatically lower than the 23% mortality observed in subjects with diabetes and not on metformin,” the authors said.
The survival benefit observed with metformin remained after exclusion of patients with classic metformin contraindications, such as chronic kidney disease and heart failure (OR, 0.17; P = .0231).
“This makes any potential confounding effects from skewing metformin users toward healthier subjects without these additional comorbidities very unlikely,” Dr. Crouse and colleagues contended.
After further analysis that controlled for other covariates (age, sex, obesity status, and hypertension), age, sex, and metformin use remained independent predictors of mortality.
For metformin, the odds ratio was 0.33 (P = .0210).
But, Dr. Lipska pointed out, “Observational studies can take into account confounders that are measured. However, unmeasured confounders may still affect the conclusions of these studies … Propensity score matching to account for the likelihood of use of metformin could be used to better account for differences between metformin users and nonusers.”
