Impact of Follow-Up Record Grouping on Radiation Epidemiology Studies

In the age of chronic, low-dose radiation exposure studies, it is imperative that cohorts are large enough to detect radiation-associated health outcomes with precision. To accommodate increased subject numbers, statistical software capabilities have recently expanded to support datasets with over 50 million person-years (rows) of data. Previously, to perform Cox proportional hazards regression on large datasets, analysts grouped annual dose records into uniform intervals. This method enabled analyses of pooled cohorts larger than possible with traditional radiation epidemiology software. However, combining records within a dataset may mask important time dynamics, especially for individuals with a limited number of annual records. In this work, a prominent cohort from the Million Person Study is analyzed with and without person-year grouping. Changes in risk estimates are reported for a variety of person-year group sizes, grouping methods, and health outcomes. These comparisons inform the efficacy of the previously used dataset size-reduction method while highlighting the benefits of recent advancements in epidemiology software.