Sex-associated differences in lesion delivery and atrial fibrillation recurrence after cryoballoon ablation

Introduction

Atrial fibrillation (AF) is more prevalent in men, but women have greater symptom burden, more advanced remodelling, and more non-pulmonary-vein triggers. After catheter ablation, women show higher rates of durable pulmonary vein isolation (PVI), recurrence, and adverse events.^1–3 While anatomical differences may contribute, it remains unclear whether ablation-dosing is comparable between sexes.^4,5

Methods

Under ethics approval (EA2/102/24), 2137 consecutive patients undergoing fluoroscopy-guided first-time PVI with a fixed-size 28-mm cryoballoon at six German hospitals were included. Phrenic nerve monitoring was compound motor action potential (CMAP)-guided during right-sided applications. Oesophageal temperature monitoring was operator-dependent and non-standardized. Except for CMAP-related safety termination, freezes were ended time-based. Recurrence was defined as arrhythmia ≥30 s or physician-documented recurrence. Patients without 12-month follow-up were excluded. Clinical and imaging data were extracted, with left atrial size standardized to diameter (LAD) using validated conversions.⁶ Temperature-time curves were analysed to derive cumulative cryothermal exposure. Effective cryothermal dose (CTD) was defined as the area under the temperature-time curve below −40°C, a pragmatic threshold based on ex vivo myocardial injury data and expected probe-to-surface temperature differences.⁷ Natural logarithmic transformation (ln[CTD]) accounted for diminishing effects at lower temperatures. Curves not meeting predefined criteria (typical profile, duration >90 s, minimum temperature ≤−10°C) were excluded. Continuous and categorical variables were analysed using standard tests.

Recurrence predictors were assessed using cumulative CTD-focused multivariable logistic regressions adjusted for clinical covariates and centre effects, evaluated by calibration and bootstrap resampling. To assess whether sex-associated procedural information was present in non-aggregated temperature-time curves, we trained multiple-instance learning (MiLM) models. Patients were represented as padded and masked bags of resampled application curves. Models were trained on five centres (n = 11 707 curves) and validated in one held-out centre (n = 724 curves). Patient-level cross-validation, fold-wise calibration, and ensemble averaging were used for internal evaluation. A conventional logistic-regression comparator using prespecified cryothermal summary variables benchmarked the full-curve signal. Feature importance was assessed by permutation-based change in Brier score.

Results

Patients showed largely comparable baseline characteristics (Figure 1A) and had persistent AF in 55% (♂ 56% vs. ♀ 52%; P = 0.058). Within 12 months, arrhythmia recurrence occurred in 27% of patients with a higher frequency in women (♂ 25.2% vs. ♀ 29.6%, P = 0.025).

Figure 1.

(A) Baseline characteristics of the overall cohort and stratified by sex. Age in years; BMI, body mass index in kg/m²; CAD, coronary artery disease; DM, diabetes mellitus; GFR, glomerular filtration rate in mL/min/1.73 m²; HYP, arterial hypertension; LAD, estimated left atrial diameter in mm; LVEF, left ventricular ejection fraction in %; proBNP, pro-B-type natriuretic peptide in ng/L; QRS, QRS duration in ms; and Rec12 m, recurrence within 12 months. Asterisks indicate significant differences between sexes. (B) Sex-specific mean cryothermal temperature curves with standard deviation over normalized application time; individual outliers were reviewed. (C) Cryothermal dose by sex and recurrence status, and recurrence rates across dose quartiles. CTD, cryothermal dose in °C·s; QRT, quartile. (D) Multiple-instance learning model performance for sex classification, including internal and external validation and permutation-based feature importance. Temp, temperature; min., minimum.

On average, 6 ± 2 temperature recordings passed criteria for clinically meaningful cryothermal exposure. Men demonstrated modestly longer therapy (♂ 1582 ± 712 vs. ♀ 1478 ± 652 s, P < 0.001) and freeze-duration (P = 0.030) as well as lower minimal (♂ −55 ± 6 vs. ♀ −53 ± 6°C, P < 0.001) and mean temperatures during freeze (♂ −43 ± 5 vs. ♀ −42 ± 5°C, P < 0.001). Consequently, men exhibited greater freeze exposure, with longer time below −40°C (♂ 612 ± 390 vs. ♀ 543 ± 334 s, P < 0.001) and higher CTD (♂ 4512 ± 3956 vs. ♀ 3487 ± 3093, P < 0.001) and ln(CTD) (♂ 6.21 ± 0.77 vs. ♀ 6.13 ± 0.70, P = 0.009). Maximal (P = 0.042) and mean temperature drop rate (♂ −1.56 ± 0.28 vs. ♀ −1.64 ± 0.30°C/s, P < 0.001) were steeper in women (Figure 1B, 1C).

The multivariable logistic-regression model (n = 1509; 26.7% 12-month recurrence), including sex, age, LVEF, LAD, ln(CTD), treatment centre, and comorbidities, was statistically significant (P < 0.001) and demonstrated good calibration. Bootstrap validation confirmed robustness [95% bias-corrected and accelerated (BCa) confidence interval (CI) −0.495 to −0.182, P < 0.001].

Higher ln(CTD) (OR 0.717, 95% CI 0.612–0.838, P < 0.001) and higher LVEF (OR 0.979 per %, 95% CI 0.968–0.991, P < 0.001) were independently associated with a lower recurrence risk. A graded dose–response relationship across CTD quartiles was observed (P = 0.002). Larger LAD (OR 1.013 per mm, 95% CI 0.999–1.028, P = 0.067) and greater age showed a trend towards higher recurrence risk (OR 1.011 per year, 95% CI 0.999–1.023, P = 0.062). Other covariates, including sex (OR 1.012, 95% CI 0.784–1.307, P = 0.928), were not independently contributing to recurrence. No sex-by-thermal-dose interaction was observed (P = 0.699). Sensitivity analyses using alternative thresholds showed consistent results for CTD below −30°C (OR 0.729, P = 0.003), whereas a more restrictive threshold of −50°C, which is below the mean freeze temperature of −42.2°C, was not significant (OR 0.948, P = 0.157).

In exploratory full-curve analysis, MiLM identified sex-associated information in cryothermal curves with modest internal discrimination [ensemble area-under-the-curve (AUC) 0.66; Brier 0.21] and similar external performance (AUC 0.73, 95% CI 0.64–0.82; Brier 0.22). At the optimized threshold, performance was balanced internally (accuracy 0.68, sensitivity 0.73, specificity 0.69, F1 0.69) and externally (accuracy 0.64, sensitivity 0.66, specificity 0.61, F1 0.64). A conventional logistic-regression comparator based on prespecified cryothermal summary variables achieved lower discrimination (AUC 0.607, 95% CI 0.581–0.633), suggesting that conventional temperature-dose metrics captured part, but not all, of this signal. Importance analysis indicated that minimum-temperature metrics contributed most strongly to model performance (Figure 1D). These findings support the presence of sex-associated procedural patterns.

Discussion

Higher CTD showed a consistent, graded association with lower recurrence, supporting a dose–response relationship.⁸ Lower CTD was associated with higher recurrence and attenuated the unadjusted association between female sex and recurrence, suggesting that procedural exposure may have partly contributed to observed sex-related differences in outcome. Exploratory full-curve MiLM supported reproducible sex-associated procedural patterns.

Women received lower effective cryothermal exposure despite comparable procedural endpoints. However, although CTD may be modifiable, increasing CTD may pose risks. The observed sex-related difference may reflect variation in balloon–tissue coupling due to anatomical or procedural factors.^9,10 Men have been reported to exhibit greater anterior and septal atrial wall thickness, larger pulmonary vein diameters and cross-sectional areas, and longer pulmonary vein sleeves. Pulmonary vein wall thickness has not been directly studied in this context. Comparable temperature-decline dynamics argue against a dominant isolated role of intrinsic tissue characteristics, but do not exclude subtler anatomical or procedural mechanisms.

Limitations

Retrospective design, centre variability, and incomplete procedural-strategy capture introduce potential bias and residual confounding. CTD should be interpreted as a cryothermal exposure marker, not as a validated tissue-level lesion metric or directly titratable treatment target. Vein-specific applications, operator strategy, occlusion quality, and direct tissue characteristics were unavailable, limiting mechanistic interpretation. Follow-up reflected routine, Holter-ECG-based care. Therefore, asymptomatic recurrence may have been underestimated, and AF burden could not be assessed.

Conclusion

Women exhibited higher recurrence rates despite similar procedural endpoints. This difference was not independently attributable to sex after adjustment and was associated with lower cryothermal exposure. Exploratory full-curve modelling supported sex-associated procedural patterns. Strategies assessing effective lesion formation may improve outcomes.

Funding

F.H. was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through CRC 1470, the Berlin Institute of Health (BIH), and the German Federal Ministry of Education and Research (BMBF).

Data availability

The data underlying this article will be shared at a reasonable request by the corresponding author.

Artificial intelligence disclosure

Artificial intelligence-assisted tools were used to refine language, improve clarity, and enhance conciseness. They did not contribute to study design, data analysis, content generation, or interpretation of results.