PhD Student UNC Chapel Hill Gillings School of Global Public Health, United States
Background: Post-acute sequelae of COVID-19 (PASC) is an emerging epidemic that affects up to 20% of individuals following infection. The pathophysiology of the condition suggests immune altering effects causing prolonged symptoms. Metformin may have immune-modulatory effects that could reduce the risk of PASC in persons on metformin at the time of SARS-CoV-2 infection.
Objectives: To estimate the risk of PASC in adults with type 2 diabetes mellitus (T2DM) treated with metformin compared with sulfonylureas (SU), or dipeptidyl peptidase 4 inhibitors (DPP4is).
Methods: This retrospective study used patient level data from the National Covid Cohort Collaborative (N3C) between October 2021 and October 2022. Our study population consisted of adults ≥ 18 years with T2DM who had at least one outpatient healthcare encounter in each of the two 6-months periods before COVID-19 diagnosis, and who had metformin, SU, or DPP4i, and no other diabetes medications, reported on their active medication list in the 90 days prior to diagnosis. Patients with combination therapy of the study comparator drugs were excluded. Follow-up began after a positive test for SARS-CoV-2 infection until the outcome of interest, death, loss to follow-up, or administrative censoring. PASC was defined based on presence of International Classification of Disease tenth edition diagnoses codes (U09.9). We calculated the 1-year incidence of PASC using the Kaplan-Meier estimator and treating death as a censoring event. We estimated crude 1-year risk ratios (RR) and risk differences (RD).
Results: During the study period, we identified 2,667 SARS-CoV-2 infections of which 80% (n=2,128), 15% (n=400), and 5% (n=139) occurred in users of metformin, SU, and DPP4i respectively. Compared to the users of SU or DPP4i, patients treated with metformin were younger and a higher body mass (BMI). Users of metformin also had a lower prevalence of hypertension and other cardiovascular diseases, including heart failure and coronary artery disease. The 1-year risk of getting a code for PASC was 2.20%, 1.53%, and 2.39% for metformin, SU, and DPP4i respectively (RR metformin vs SU: 1.60 [95% CI: 0.64, 4.01]; RD metformin vs SU: 8.0 per 1,000 [95% CI: -5.2, 21.3], RR metformin vs DPP4i: 0.91[95% CI: 0.29, 2.91]; RD metformin vs DPP4i: -2.0 per 1,000 [95% CI: -29.0, 25.0] ).
Conclusions: In this preliminary study, we found a low incidence of PASC in our study population leading to imprecise estimates of the effect of these medications, likely due to reduced sensitivity of the diagnosis code in capturing PASC. Future work will use a machine learning algorithm developed to predict PASC in this population to estimate risk and account for measured confounding and competing events, providing more evidence on the effect of these medications on PASC.
