A scoring tool to predict mortality and dependency after cerebral venous thrombosis

Abstract

Background and purpose: A prognostic score was developed to predict dependency and death after cerebral venous thrombosis (CVT) to identify patients for targeted therapy in future clinical trials. Methods: Data from the International CVT Consortium were used. Patients with pre-existent functional dependency were excluded. Logistic regression was used to predict poor outcome (modified Rankin Scale score 3–6) at 6 months and Cox regression to predict 30-day and 1-year all-cause mortality. Potential predictors derived from previous studies were selected with backward stepwise selection. Coefficients were shrunk using ridge regression to adjust for optimism in internal validation. Results: Of 1454 patients with CVT, the cumulative number of deaths was 44 (3%) and 70 (5%) for 30 days and 1 year, respectively. Of 1126 patients evaluated regarding functional outcome, 137 (12%) were dependent or dead at 6 months. From the retained predictors for both models, the SI2NCAL2C score was derived utilizing the following components: absence of female-sex-specific risk factor, intracerebral hemorrhage, infection of the central nervous system, neurological focal deficits, coma, age, lower level of hemoglobin (g/l), higher level of glucose (mmol/l) at admission, and cancer. C-statistics were 0.80 (95% confidence interval [CI] 0.75–0.84), 0.84 (95% CI 0.80–0.88) and 0.84 (95% CI 0.80–0.88) for the poor outcome, 30-day and 1-year mortality model, respectively. Calibration plots indicated a good model fit between predicted and observed values. The SI2NCAL2C score calculator is freely available at www.cerebralvenousthrombosis.com. Conclusions: The SI2NCAL2C score shows adequate performance for estimating individual risk of mortality and dependency after CVT but external validation of the score is warranted.