Bordeaux Segalen University

Causal machine learning (ML) provides an efficient way of identifying heterogeneous treatment effect groups from hundreds of possible combinations, especially for randomized trial data.

The aim of this paper is to illustrate the potential of applying causal ML on the DECAAF II trial data. We proposed a causal ML model to predict the treatment response heterogeneity.

We applied causal tree learning to the DECAAF II trial data as an example of real applications, identifying subgroups that may be superior when subject to one of the treatments over the other through an easily interpretable process. For each subgroup identified, the characteristics were summarized, and the relationship between treatment arms and risk for recurrence of atrial tachyarrhythmia (aTA) among subjects was assessed.

Causal tree learning demonstrated that, among all the preablation predictors, dividing subgroups according to age, with a cutoff of 58 years, provides the most heterogeneous subgroups in response to fibrosis-guided ablation in addition to pulmonary vein isolation (PVI) compared with PVI alone. The difference in the risk of recurrence of aTA between 2 treatments was nonsignificant in older patients (hazard ratio [HR] 1.06; 95% confidence interval [CI] 0.77-1.47; P = .72). However, among the younger patients, the risk of aTA recurrence was significantly lower in the fibrosis-guided ablation group compared with PVI-only (HR 0.50; 95% CI 0.28-0.90); P = .02).

Applying causal ML on random controlled trial datasets helped us identify groups of patients that profited from the treatment of interest in an efficient and unbiased manner.

The Efficacy of Delayed Enhancement MRI-Guided Fibrosis Ablation vs Conventional Catheter Ablation of Atrial Fibrillation randomized trial showed no difference in atrial fibrillation (AF) recurrence with additional delayed enhancement magnetic resonance imaging (DE-MRI) fibrosis-targeted ablation to pulmonary vein isolation (PVI) in persistent AF.

We evaluated the effect of lesion delivery on ablation-induced scarring and AF recurrence.

Lesions delivered, targeting fibrotic and nonfibrotic areas identified from preablation DE-MRI, were studied in relation to ablation-induced scarring on 3-month DE-MRI, including their association with arrhythmia recurrence.

A total of 593 patients treated with radiofrequency were analyzed: 293 (49.4%) underwent PVI and 300 (50.6%) underwent additional fibrosis-guided ablation. Lesion analysis showed that 80.9% in the MRI fibrosis-guided group vs 16.5% in the PVI group (P < .001) had ≥40% of baseline fibrosis targeted. MRI assessment of ablation-induced scar showed that 44.8% of fibrosis-guided ablation and 15.5% of PVI had ≥40% of their fibrosis covered by scar (P < .001), demonstrating significant attenuation from lesions delivered to scar formed. In the overall population, fibrosis coverage with scar was not associated with recurrence (hazard ratio [HR] 0.90; 95% confidence interval [CI] 0.80-1.01; P = .08 per 20% increase). In patients with baseline fibrosis < 20%, fibrosis coverage with scar was associated with lower recurrence than PVI (HR 0.85; 95% CI 0.73-0.97; P = .03), whereas the association was not significant when baseline fibrosis ≥ 20% (HR 0.97; 95% CI 0.80-1.17; P = .77). Significant center variation was observed in fibrosis targeting and coverage with scarring.

Radiofrequency ablation lesions do not uniformly result in scar formation. A post hoc analysis suggests reduced arrhythmia recurrence when ablation-induced scarring covers fibrotic regions in patients with low baseline fibrosis.