Predictive Performance of HAS-BLED Score in Patients with Atrial Fibrillation and Cancer: A Meta-Analysis

Patients with atrial fibrillation (AF) and history of cancer face unique bleeding risks, complicating the applicability of standard bleeding risk scores like HAS-BLED. This meta-analysis aimed to evaluate the performance of HAS-BLED in predicting bleeding events in this high-risk population.

The MEDLINE, PubMed, and EMBASE databases were searched from 1st of January 2010 to 30th of November 2024 for relevant studies using keywords, such as "AF" "cancer" "bleeding," and "HAS-BLED." Data on C-statistics were extracted to assess the predictive performance of HAS-BLED score.

Our analysis included seven retrospective cohort studies, recruiting a total of 436,102 patients. The quality of the included studies was deemed acceptable for analysis. The reported C-statistics for HAS-BLED score varied widely across studies, ranging from 0.45 to 0.77. Subgroup analyses demonstrated moderate discrimination in patients with breast cancer (0.56-0.80), prostate cancer (0.58-0.72), and lung cancer (0.59-0.80), while poorer performance was observed in hematological malignancies (0.45-0.70) and in anticoagulated patients (pooled C-statistic = 0.55; 95% confidence interval: 0.54-0.56). Significant heterogeneity was observed in the overall analysis and most subgroups (I2 > 90%), except for the anticoagulated subgroup. A sensitivity analysis excluding the largest study reduced heterogeneity and improved funnel plot symmetry, indicating that study size contributed to variability in HAS-BLED performance.

The HAS-BLED score has shown variable predictive abilities in AF patients with cancer ranging from poor to good, with notable heterogeneity across studies secondary to various contributing factors. This emphasizes the need for individualized risk assessment tailored to the unique characteristics of cancer patients to effectively guide clinical decision-making.