An international cohort study of valproate withdrawal risks in young men and women

<h3>Background</h3> Whether there are any potential risks of personal harm or morbidity (including hospital admissions, injuries, or death) associated with valproate withdrawal has not yet been quantified for either men or women on valproate. We aimed to investigate this. <h3>Methods</h3> We performed a retrospective cohort study of internationally-derived electronic health data within the TriNetX research platform. Included were men and women aged 16–54 years with ≥1 epilepsy disease or symptom code between 01/12/2017–01/12/2018 and ≥2 valproate prescriptions over the preceding two years (01/01/2015–30/11/2017). 5-year propensity-matched risks of mortality and a range of morbidity outcomes were compared between those remaining on vs. withdrawn from valproate during the 01/12/2017–01/12/2018 recruitment period, regardless of whether switched to another antiseizure medication. Survival analysis was undertaken using Cox-proportional hazard models. <h3>Findings</h3> 8,834 men and 5,156 women taking valproate were recruited. 27% of men and 36% of women were withdrawn from treatment. Withdrawal was associated with significantly increased risks of A&amp;E attendance (HRs overall: 1.270, CI 1.186–1.360, men: 1.219, CI 1.115–1.332, women: 1.185, CI 11.073–1.309), hospital admission (HRs overall: 1.146, CI 1.065–1.233, men: 1.140, CI 1.032–1.260, women: 1.171, CI 1.053–1.303), falls (HRs overall: 1.247, CI 1.096–1.420, men: 1.286, CI 1.075–1.537), injuries (HRs overall: 1.104, CI 1.026–1.188, men: 1.113, CI 1.013–1.222), burns (HRs overall: 1.641 1.093–2.463), and new-onset depression (HRs overall 1.273, CI 1.077–1.505, women: 1.305, CI 11.03–1.639). <h3>Interpretation</h3> Results help optimise informed decision-making about valproate withdrawal.