Background: An increased risk of diabetes after COVID-19 exposure has been reported in Caucasians during the early phase of the pandemic, but the effects across viral variants and in non-Caucasians have not been evaluated.
Methods: To address this gap, survival analyses were performed for five outbreak periods. From an anonymized health insurance database REZULT for the employees and their dependents of large companies or government agencies in Japan, 5 matched cohorts were generated based on age, sex, area of residence (47 prefectures), and 7 ranges of medical bills (COVID-19 exposed:unexposed = 1:4). Observation of each matching group began on the same day. Incident diabetes type 1 (T1D) and type 2 (T2D) were defined as the first claim during the target period, including at least 1 year before the start of observation.
Results: T1D accounted for 0.8% of incident diabetes after the first COVID-19 exposure, similar to the non-exposed cohort. Most T2D in the COVID-19 cohort was observed within a few weeks. After further adjustment for the number of days from the start of observation to hospitalization (a time-dependent variable), the hazard ratio for incident T2D ranged from 14.1 to 20.0, with 95% confidence intervals (95%CI) of 8.7 to 32.0, during the 2-month follow-ups from the original strain outbreak to the Delta variant outbreak (by September 2021), and decreased to 2.0, with a 95%CI of 1.6 to 2.5, during the Omicron outbreak (by March 2022). No association was found during the BA.4/5 outbreak (until September 2022). Males had a higher risk, and the trend toward higher risk in older age groups was inconsistent across the periods.
Conclusions: Our large dataset, covering 2019-2023, reports for the first time the impact of COVID-19 on incident diabetes in non-Caucasians. The risk intensity and attributes of post-COVID-19 T2D were inconsistent across outbreak periods, suggesting diverse biological effects of different SARS-CoV-2 variants.
Keywords: COVID-19; Diabetes; Hazard ratio; Health insurance system.