Validation of Embolic and Bleeding Risk Scores in Patients With Atrial Fibrillation and Cancer

Abstract

Little is known about the prediction of atrial fibrillation (AF) risk scores in patients with cancer. The aim of this study was to assess the predictive ability of the CHA₂DS₂-VASc and HAS-BLED scores in patients with AF and cancer. Overall, 16,056 patients with AF diagnosed between 2014 and 2018 from a Spanish health area, including 1,137 patients with cancer, were observed during a median follow-up of 4.9 years. Although discrimination was similar between patients with cancer and patients without cancer who were treated with anticoagulation therapy (0.56 and 0.58), in patients with cancer who were not treated with anticoagulation therapy, c-statistic of CHA₂DS₂-VASc was poor and significantly lower than in the patients without cancer (0.42 vs 0.65). The overall precision of the CHA₂DS₂-VASc score was good throughout the follow-up (Brier score < 0.1), in patients with and without cancer. Regarding the HAS-BLED score, calibration and discrimination were poor in patients with cancer (c-statistic 0.51), similar to those in patients without cancer (c-statistic 0.53). In patients with cancer who were not treated with anticoagulation therapy, the embolic risk CHA₂DS₂-VASc score = 1 was similar to CHA₂DS₂-VASc score ≥ 2. Only patients with AF and cancer and CHA₂DS₂-VASc score = 0 presented a low risk of embolic events (negative predictive value 100%). A HAS-BLED score > 3 was not associated with higher bleeding risk in patients with cancer (p > 0.05). In summary, in patients with cancer and with AF, neither the CHA₂DS₂-VASc score nor the HAS-BLED score was useful for predicting embolic and hemorrhagic events, respectively. However, a CHA₂DS₂-VASc score 0 is useful to identify patients with AF and cancer who are at low embolic risk.