Development and validation of risk stratification and shared decision-making tool for catheter ablation for atrial...

The coexistence of atrial fibrillation (AF) and heart failure (HF) presents a significant challenge in risk evaluation and treatment decision-making. This study aimed to develop a shared decision-making tool that aids in risk stratification and guides radiofrequency catheter ablation (RFCA) decisions for patients with AF and HF.

In this multicentre cohort study, we derived a shared decision-making tool by applying unsupervised clustering and supervised learning models to data from the China-AF registry, collected from 31 hospitals between August 1, 2011, and December 31, 2022. External validation was performed using diverse ethnic populations from the international, multicenter, randomized, open-label CABANA trial. The study included patients with AF and HF and excluded the asymptomatic patients. Association of RFCA with prognostic outcomes were assessed and compared across model-identified risk strata, focusing on composite events (cardiovascular death and stroke), all-cause death, cardiovascular hospitalization, major bleeding, and AF recurrence. This study is registered with the Chinese Clinical Trial Registry, ChiCTR-OCH-13003729.

Among 3122 patients in the derivation cohort (1476 females [47.3%] and 1646 males [52.7%]) and the 778 patients in the validation cohort (345 females [44.3%] and 433 males [55.7%]), the tool identified three clusters based on 25 readily accessible clinical features. Incidence rates (per 100 person-years) of composite events were highest in cluster 1 [7.7 (95% CI, 6.9-8.6)], followed by cluster 2 [6.8 (95% CI, 6.1-7.7)], and lowest in cluster 3 [3.8 (95% CI, 3.4-4.4)] (log-rank P < 0.0001). Similar risk stratification was observed for all-cause and cardiovascular mortality. The tool demonstrated consistent risk stratification in the HF with preserved ejection fraction (HFpEF) subgroup and the external validation cohort, with a log-rank P < 0.0001 for composite events. Compared to drug therapy, RFCA was associated with a significantly better prognosis in cluster 1 of the China-AF registry (for composite events: adjusted HR = 0.16; 95% CI, 0.07-0.36, P for interaction = 0.0039), with similar findings observed in the external validation cohort (adjusted HR = 0.19; 95% CI, 0.05-0.73, P = 0.015).

This machine learning-based tool shows promise in facilitating shared decision-making for patients with AF and HF by identifying those most likely to benefit from RFCA following risk stratification. However, as the tool was developed based on observational study data, its effectiveness requires further validation in interventional trials and real-world clinical practice.

The National Key Research and Development Program of China, Beijing Hospitals Authority Yangfan Program, Engineering Research Center of Cardiovascular Diagnostic and Therapeutic Technologies and Devices, Ministry of Education and the National Natural Science Foundation of China.