Aim
Sex-related differences in atrial fibrillation (AF) are well recognised, yet their impact on ablation outcomes remains incompletely understood. Whether sex-specific differences in cryothermal lesion formation contribute to disparities in post-ablation recurrence is unclear. We aimed to assess the association between procedural cryothermal characteristics and sex-specific outcomes after cryoballoon-based pulmonary vein isolation.
Methods
This retrospective multicentre study analysed patients undergoing fluoroscopy-guided first-time cryoballoon-based pulmonary vein isolation. Temperature-time curves, clinical data, and imaging data were evaluated. The effective cryothermal dose (CTD), based on ex vivo myocardial injury data, was defined as the natural logarithm of the area under the temperature–time curve below −40 °C. Associations with 12-month arrhythmia recurrence were assessed using multivariable logistic regression, adjusting for clinical covariates and centre effects. Non-aggregated temperature–time curves were analysed using a multiple-instance learning (MiLM) approach with patient-level cross-validation, fold-wise calibration, ensemble averaging, and held-out centre validation.
Results
In the cohort (n=2137), men exhibited modestly longer therapy durations (p<0.001) and lower mean temperatures during freeze (p<0.001), resulting in greater cryothermal exposure compared with women (p=0.009). In multivariable analysis, higher cryothermal dose (OR 0.717, p<0.001) was independently associated with lower recurrence risk, indicating a graded dose–response relationship. Left ventricular ejection fraction (p=0.002) and left atrial size were also independent predictors. Although female sex was associated with higher recurrence in unadjusted analyses, it was not independently associated with outcome after adjustment for cryothermal dose and markers of atrial remodelling (p=0.928). The MiLM analysis demonstrated moderate and consistent discrimination (AUC 0.73, 95% CI 0.64-0.82; Brier 0.22), suggesting that conventional temperature-dose metrics captured part of this signal but not all of it. Permutation-based feature importance analysis indicated that minimum-temperature metrics contributed most strongly to model performance.
Conclusions
A higher effective cryothermal dose is strongly associated with reduced AF recurrence. Sex was not an independent predictor after adjustment, indicating that observed outcome differences are largely explained by variation in effective lesion formation. Machine learning-based analysis of cryothermal dynamics supports these findings and underscores the importance of procedural factors in determining outcomes.
| |
All |
Male |
Female |
| n= |
2137 |
1267 |
887 |
| Age |
65.3 ± 11.1 |
62.9 ± 11.4 |
68.9 ± 9.5* |
| BMI |
28.1 ± 4.8 |
28.2 ± 4.3 |
27.9 ± 5.4 |
| proBNP |
1131 ± 2004 |
942 ± 1631 |
1427 ± 2452* |
| LVEF |
56.7 ± 9.6 |
55.6 ± 10.5 |
58.3 ± 8.1* |
| LAD |
46.8 ± 11.6 |
47.7 ± 13.3 |
45.2 ± 7.8 |
| CAD |
23.5% |
28.5% |
16.4%* |
| HYP |
76% |
73.4% |
79.7%* |
| DM |
13.2% |
13.7% |
12.4% |
| 12m Rec |
27.1% |
25.2% |
29.6%* |
Table 1: Cohort Characteristics Age in years; BMI, body mass index in kg/m²; proBNP, pro–B-type natriuretic peptide in ng/L; LVEF, left ventricular ejection fraction in %; LAD, estimated left atrial diameter in mm; CAD, coronary artery disease; HYP, arterial hypertension; DM, diabetes mellitus; and 12m Rec, recurrence within 12 months. *statistically significant difference
Figure 1: A) Sex-specific mean cryothermal temperature curves with standard deviation over normalized application time B) Multiple-instance learning model performance for sex classification.
