TOPLINE:
METHODOLOGY:
- Previous studies have shown that metformin use before and during SARS-CoV-2 infection reduces severe COVID-19 and postacute sequelae of SARS-CoV-2 (PASC), also referred to as long COVID, in adults.
- A retrospective cohort analysis was conducted to evaluate the association between metformin use before and during SARS-CoV-2 infection and the subsequent incidence of PASC.
- Researchers used data from the National COVID Cohort Collaborative (N3C) and National Patient-Centered Clinical Research Network (PCORnet) electronic health record (EHR) databases to identify adults (age, ≥ 21 years) with T2D prescribed a diabetes medication within the past 12 months.
- Participants were categorized into those using metformin (metformin group) and those using other noninsulin diabetes medications such as sulfonylureas, dipeptidyl peptidase-4 inhibitors, or thiazolidinediones (the comparator group); those who used glucagon-like peptide 1 receptor agonists or sodium-glucose cotransporter-2 inhibitors were excluded.
- The primary outcome was the incidence of PASC or death within 180 days after SARS-CoV-2 infection, defined using International Classification of Diseases U09.9 diagnosis code and/or computable phenotype defined by a predicted probability of > 75% for PASC using a machine learning model trained on patients diagnosed using U09.9 (PASC computable phenotype).
TAKEAWAY:
- Researchers identified 51,385 and 37,947 participants from the N3C and PCORnet datasets, respectively.
- Metformin use was associated with a 21% lower risk for death or PASC using the U09.9 diagnosis code (P < .001) and a 15% lower risk using the PASC computable phenotype (P < .001) in the N3C dataset than non-metformin use.
- In the PCORnet dataset, the risk for death or PASC was 13% lower using the U09.9 diagnosis code (P = .08) with metformin use vs non-metformin use, whereas the risk did not differ significantly between the groups when using the PASC computable phenotype (P = .58).
- The incidence of PASC using the U09.9 diagnosis code for the metformin and comparator groups was similar between the two datasets (1.6% and 2.0% in N3C and 2.2 and 2.6% in PCORnet, respectively).
- However, when using the computable phenotype, the incidence rates of PASC for the metformin and comparator groups were 4.8% and 5.2% in N3C and 25.2% and 24.2% in PCORnet, respectively.
IN PRACTICE:
“The incidence of PASC was lower when defined by [International Classification of Diseases] code, compared with a computable phenotype in both databases,” the authors wrote. “This may reflect the challenges of clinical care for adults needing chronic medication management and the likelihood of those adults receiving a formal PASC diagnosis.”
SOURCE:
The study was led by Steven G. Johnson, PhD, Institute for Health Informatics, University of Minnesota, Minneapolis. It was published online in Diabetes Care.