Novel AI-Based Score Could Pinpoint AF Risk in Patients With CLL on BTKis

A new scoring system could help more accurately characterize a patient’s risk of developing atrial fibrillation (AF) while taking Bruton tyrosine kinase inhibitors (BTKis) for chronic lymphocytic leukemia (CLL), according to a study published in the European Journal of Haematology.¹

Other AF prediction models have been proposed, including one with a machine-learning approach, but earlier models may not apply to patients taking second-generation BTKis. | Image Credit: sovova - stock.adobe.com

BTKis have had a major positive impact on the treatment of CLL, but they also come with a risk of significant adverse effects, including AF. Corresponding author Tamar Tadmor, MD, of the Bnai Zion Medical Center, in Israel, and colleagues noted that the first BTKi, ibrutinib, carries the highest reported incidence of AF. However, second-generation BTKi’s, including acalabrutinib (Calquence; AstraZeneca) and zanubrutinib (Brukinsa; BeiGene), are still associated with a significant risk of AF.² In fact, they noted that CLL itself had been linked with AF risk even before the development of BTKis.

Though a number of AF prediction models have been proposed, including one with a machine-learning approach, Tadmor and colleagues explained that earlier models are centered on AF risk in patients taking ibrutinib and therefore may not apply to patients taking second-generation BTKis.¹

“Additionally, they heavily emphasize traditional cardiovascular risk factors—such as age and hypertension—while underutilizing unbiased variables routinely collected during CLL follow-up, such as infection history or baseline blood cultures,” Tadmor and colleagues wrote.

The investigators turned to electronic health records from patients at Maccabi Healthcare Services, the second-largest health care organization in Israel. They identified 3964 patients with a CLL diagnosis, including 208 patients who initiated BTKi therapy during the study period. Of those, 125 received ibrutinib and the other 83 received acalabrutinib. A total of 16 patients developed AF during the study’s follow-up period; 13 of those patients were receiving ibrutinib.

The investigators used the Risk-Calibrated Supersparse Linear Integer Model algorithm to analyze patient characteristics and create an optimal scoring model to assess AF risk. They found that factors such as age, hypertension, and sex were important. However, the algorithm also identified low estimated glomerular filtration rate (less than 30 mL per minute), elevated absolute monocytes (higher than 1100/μL), elevated C-reactive protein (0.5-1 mg/L), elevated creatine kinase (above upper normal limit), and elevated beta-2-microglobulin (higher than 2.5 mg/L) as factors associated with AF risk. Lastly, they confirmed that the type of BTKi used was important, as ibrutinib was associated with a higher risk of AF.

The authors noted previous research has also identified some of the same risk factors for AF, including high monocyte counts.

“Overall, these findings underscore the multifaceted role of monocytes in AF, involving inflammatory pathways, structural remodeling, and fibrotic processes, which collectively contribute to the risk and progression of AF,” they wrote.

The investigators then used their findings to identify 3 risk groups: low, moderate, and high risk. In the latter 2 groups, the median AF-free survival was 56 months and 28 months, respectively. The median AF-free survival was not reached in the low-risk group.

Taken together, the authors said their scoring system—which they call the CLLAF Score—has several advantages over existing scoring methods, including the inclusion of patients taking a second-generation BTKi and the reliance on easily obtainable parameters. They noted that it outperforms existing scoring systems.

Tadmor and colleagues said there are a number of important limitations to their findings. Among them, the study was retrospective and did not include external validation. In addition, they said their cohort had a low incidence of severe arrhythmias, which may limit the generalizability of the results.

Still, they said their findings offer the potential for a novel, unbiased model to assess AF risk in treatment-naive patients and that CLLAF scores can be used to discuss the risks and benefits of particular therapies with patients.

References

1. Tadmor T, Melamed G, Alapi H, Rokach L. CLLAF score-a new risk score for predicting atrial fibrillation in treatment-naive CLL patients initiating first- and second-generation BTK inhibitor therapy. Eur J Haematol. 2025;114(6):1011-1019. doi:10.1111/ejh.14400
2. Ahn IE, Brown JR. Targeting Bruton's tyrosine kinase in CLL. Front Immunol. 2021;12:687458. doi:10.3389/fimmu.2021.687458