Metformin associated with less incident seizures and fewer seizures: understanding why by deconstructing correlation in electronic health records.

The paper by Mbizvo and colleagues demonstrated a beneficial association between metformin in people without evidence for pre-diabetes on the rate of new onset seizures after cerebral injury (Cohort 1) and markers of seizure control in people with epilepsy (Cohort 2). This is both exciting and an opportunity to deconstruct correlation in retrospective electronic health record (EHR) studies to guide clinical care and future evaluations of causation.1

The first step of evaluating EHR studies is understanding the inclusion and exclusion criteria. To address the confound of pre-diabetes, this study focused on a likely less than 10% of metformin users by excluding patients with any known evidence of pre-diabetes or diabetes. Evidence included the International Classification of Diseases codes, medication records and A1c ≥5.7% or serum glucose ≥100 mg/dL. Other measurable reasons for…