Validation of a prediction model for early reconnection after cryoballoon ablation.

Background: We previously developed an early reconnection/dormant conduction (ERC) prediction model for cryoballoon ablation to avoid a 30-min waiting period with adenosine infusion. We now aimed to validate this model based on time to isolation, number of unsuccessful cryo-applications, and nadir balloon temperature.

Methods: Consecutive atrial fibrillation patients who underwent their first cryoballoon ablation in 2018-2019 at the Leiden University Medical Center were included. Model performance at the previous and at a new optimal cutoff value was determined.

Results: A total of 201 patients were included (85.57% paroxysmal AF, 139 male, median age 61 years (IQR 53-69)). ERC was found in 35 of 201 included patients (17.41%) and in 41 of 774 veins (5.30%). In the present study population, the previous cutoff value of - 6.7 provided a sensitivity of 37.84% (previously 70%) and a specificity of 89.07% (previously 86%). Shifting the cutoff value to - 7.2 in both study populations resulted in a sensitivity of 72.50% and 72.97% and a specificity of 78.22% and 78.63% in data from the previous and present study respectively. Negative predictive values were 96.55% and 98.11%. Applying the model on the 101 patients of the present study with all necessary data for all veins resulted in 43 out of 101 patients (43%) not requiring a 30-min waiting period with adenosine testing. Two patients (2%) with ERC would have been missed when applying the model.

Conclusions: The previously established ERC prediction model performs well, recommending its use for centers routinely using adenosine testing following PVI.