Ability of the Khorana score to predict recurrent thromboembolism in cancer patients with ischemic stroke.

Cancer patients with acute ischemic stroke (AIS) are high-risk for recurrent thromboembolic events (RTE). Currently, no risk stratification model exists to predict RTE in this population. We tested the hypothesis that the Khorana score, a validated risk model for predicting venous thromboembolism in cancer patients, can effectively classify RTE risk in cancer patients with AIS. We retrospectively identified adults with active solid or hematological cancer diagnosed with AIS at a tertiary-care cancer center from 2005 to 2010. The Khorana score at the time of index stroke was calculated. The primary outcome was arterial or venous RTE. Cox regression was used to evaluate the association of the Khorana score and its individual components with RTE. Harrell's c-statistic was used to calculate the score's discriminatory ability. Among 263 AIS patients, median survival was 84 days (IQR 24-149 days) and 90 (34%) had RTE. The median Khorana score was 2 (IQR 1-2, range 0-5). Cumulative rate of RTE was 28% among patients who scored 0, 36% with scores of 1-2, and 32% with scores of 3-6. The overall Khorana score was marginally associated with RTE (HR, 1.14; 95% CI, 1.02-1.28). Of its individual components, only leukocytosis was associated with RTE (HR adjusted for other components, 1.45; 95% CI 1.11-1.90). The score's c-statistic for predicting RTE was 0.57. In this study, the Khorana score had poor discriminatory ability for predicting RTE in cancer patients with AIS. Future research is needed to identify better methods for predicting RTE in this high-risk population.