Developing a simple clinical risk score for ibrutinib-associated atrial fibrillation.

Abstract

Background: Ibrutinib, a Bruton's tyrosine kinase (BTK) inhibitor, is used in the treatment of B-cell malignancies (e.g., chronic lymphocytic leukemia [CLL]). Initial risk stratification of ibrutinib-associated atrial fibrillation (IAAF) may inform atrial fibrillation (AF) surveillance strategies. The performance of existing AF risk scores to predict incident AF among patients newly treated with ibrutinib is unknown.

Methods: We conducted a single-center retrospective study of all patients without a history of AF who were treated with ibrutinib (2012-2016). Patient demographics were compared between cohorts by the presence of IAAF within 24 months of treatment initiation. First, the predictive ability of established AF risk models was assessed. Secondly, univariate and multivariate analyses were used to create a new IAAF risk model which was compared to established AF risk models by area under the curve (AUC) analysis.

Results: Of 167 patients (66 ± 11 years, 70% male), 24 (14.4%) developed incident IAAF (mean time to IAAF, 7.1 ± 6.3 months). Univariate analysis showed that hypertension (HTN), diabetes (DM), systolic heart failure (HFrEF), and obstructive sleep apnea (OSA) were associated with IAAF. Logistic regression analysis of variables of interest and those with p < 0.1 on univariate analysis demonstrated that left atrial diameter (LAD) > 43 mm and obstructive sleep apnea were independently associated with IAAF. Existing AF risk scores had reasonable performance (AUC, 0.68-0.72). A new simple clinical risk score was developed: OSA 5 points, HFrEF 3 points, DM 2 points, and hyperlipidemia 2 points. This simple IAAF risk score achieved a numerically greater AUC than that of established risk models (AUC = 0.77). There was no statistically significant difference in the AUC performance between risk scores.

Conclusion: Among 167 ibrutinib naïve patients, risk factors for incident AF development resemble those of the general population. However, common AF risk models have moderate predictive ability. Large validation studies are needed to confirm the superior IAAF predictive ability of this simple risk score and investigate the incremental predictive value of echocardiographic variables.