Abstract TMP105: Development and Validation of a Risk Prediction Model for Ischemic Stroke Among Individuals Newly Diagnosed With Cancer

Abstract

Introduction: Cancer is an emerging risk factor for ischemic stroke. The risk of stroke is highest during the first year after a new diagnosis of cancer, but no tools exist to identify which patients are at the highest risk.. Methods: Using linked clinical and administrative health databases, we conducted a population-based retrospective cohort study of adults in Ontario, Canada with newly diagnosed cancer from 2010 - 2021 (excluding non-melanoma skin cancer & central nervous system malignancies). Patients were randomly selected for model derivation (60%) or validation (40%). The final model predicting stroke within 1 year following cancer diagnosis was derived using multivariable Fine-Gray regression with candidate predictors selected via backward elimination. Sub-distribution adjusted hazard ratios (aHR) and 95% confidence intervals (CI) were calculated, where all-cause mortality was treated as a competing event. Model performance of the validation cohort was assessed using the C-statistic & calibration plots for discrimination and calibration, respectively. Results: Of the 698,566 eligible patients, 418,911 were randomly allocated to derivation, and 279,576 to validation. The overall rate of stroke per 1000 person-years was 6.7 (6.4 - 6.9) for the derivation cohort. The final model included 22 predictors: age, sex, long-term care residency, history of heart failure, hypertension, dementia, asthma, atrial fibrillation, dyslipidemia, liver disease, ischemic stroke, transient ischemic attack, valvular disease, venous thromboembolism, hospitalization within the last 3 months, cancer type, cancer stage, cancer surgery or chemotherapy 3 months following diagnosis, and several 2-way interactions with age & cancer type. Discrimination was good, with a c-statistic of 0.73 in the validation cohort. The model was well calibrated, with points following the 45-degree line (Fig 1). Conclusion: We derived and validated a risk prediction model for ischemic stroke in patients with a new cancer diagnosis with good discrimination. Although our results require external validation, it has potential to identify individuals at highest risk for future randomized trials.