A B S T R A C TPurpose: To describe the retention rates of first antiepileptic drugs (AEDs) in patients with poststroke epilepsy on a nationwide scale. Methods: The Swedish Stroke Register, which has 94% coverage and high-resolution data on stroke, comorbidities, and disability, was cross-referenced to the National Patient Register, Drug Register, and Cause-of-Death Register. Patients with onset of AED-treated epilepsy after stroke in 2005-2010 were included. An algorithm based on prescription renewal intervals was used to analyze treatment data until the end of 2014. Results: A total of 4991 patients were included. First AEDs analyzed were carbamazepine (n = 2373), valproic acid (n = 943), levetiracetam (n = 555), lamotrigine (n = 519), phenytoin (n = 176), and oxcarbazepine (n = 89). The five-year retention rate was highest for lamotrigine (75%, 95%CI:70.4-79.4), followed by levetiracetam (69%, 95%CI:62.9-74.3), oxcarbazepine (68%, 95%CI:55.2-79.8), valproic acid (62%, 95%CI:57.8-66.4), carbamazepine (60%, 95%CI:57.6-62.4), and phenytoin (55%, 95%CI:45.2-64.0). There were minor differences in baseline characteristics with low levels of disability being slightly more common in patients treated with lamotrigine and levetiracetam. Atrial fibrillation and hypertension were more common in patients treated with levetiracetam, and atrial fibrillation was less common in patients treated with carbamazepine. In a Cox model adjusted for baseline characteristics, the risk of discontinuation was lower for lamotrigine (HR 0.53, 95%CI:0.43-0.67) and levetiracetam (HR 0.75, 95%CI:0.60-0.94) when compared to carbamazepine. Conclusions: Lamotrigine and levetiracetam have higher retention rates than carbamazepine in poststroke epilepsy. This is in agreement with existing small RCTs in this patient group.
Material and methods
Study design and study cohortThis was a retrospective cohort study based on nation-wide registers. The cohort was selected from patients registered in the Swedish Stroke Register (Riksstroke) between 2005-2010, based on information https://doi.