Confounding by indication in retrospective studies of intracerebral hemorrhage: antiepileptic treatment and mortality.

INTRODUCTION: Intracerebral hemorrhage (ICH) is a highly fatal disease with few proven treatments. Data to guide clinician decisions for therapies, including antiepileptic drugs (AED), are limited. Published studies on AED treatment in ICH have provided conflicting results. We investigated the effect of AED treatment on 90-day mortality after ICH in a large prospectively ascertained cohort. METHODS: We conducted a retrospective analysis of a prospectively assembled cohort of patients with ICH in the supratentorial regions, comparing 90-day mortality and modified Rankin Score among 543 patients treated with AED during hospitalization and 639 AED-free ICH. Supratentorial ICH location was categorized as lobar or deep hemispheric. RESULTS: Multivariate analysis demonstrated an association between AED treatment and reduced 90-day mortality in supratentorial ICH (OR = 0.62, 95 % CI 0.42-0.90, p = 0.01) and the subset of lobar ICH (OR = 0.49, 95 % CI 0.25-0.96, p = 0.04). When analyses were restricted to subjects surviving longer than 5 days from ICH, however, no association between AED treatment and a 90-day outcome, regardless of hemorrhage location (all p > 0.15), was detected, despite more than adequate power to detect the originally observed association. CONCLUSION: These results suggest that AED treatment in acute ICH is not associated with 90-day mortality or outcome and that any detected association could arise by confounding by indication, in which the most severely affected patients are those in whom AEDs are prescribed. They provide a cautionary example of the limitations of drawing conclusions about treatment effects from observational data.