The Validation of a Nomogram for Predicting Recurrence in Patients with Non-Valvular Atrial Fibrillation...

To utilize the developed nomogram for evaluating the risk of recurrence in non-valvular atrial fibrillation (NVAF) patients after radiofrequency catheter ablation (RFCA) and compare the model's performance with the APPLE, ATLAS, and Antwerp scores.

242 patients with NVAF requiring RFCA were enrolled. These patients were randomly divided into a training cohort (n=169) and a validation cohort (n=73) according to 7:3. A nomogram was developed based on LAVI, RAVI, SII, NYHA classification, CHA₂DS₂-VASc score to estimate the risk of AF recurrence after RFCA. The APPLE, ATLAS, and Antwerp scores were calculated using the "pROC" package in R software. The AUC value of the nomogram compared with each of the three scores was evaluated using the DeLong test. The integrated discrimination improvement and net reclassification index were calculated to compare the predictive performance of the nomogram against the scores in R software.

The nomogram achieved significantly higher values with an AUC of 0.837 (95% CI: 0.774-0.899) in the training cohort and 0.895 (95% CI: 0.823-0.968) in the validation cohort (all P < 0.05) than the three scores. It also achieved better positive and negative predictive values, indicating enhanced discriminatory power. By integrating multidimensional parameters and optimizing risk stratification, it significantly reduced misjudgment rates. Furthermore, the model demonstrated a more balanced sensitivity-specificity profile and greater predictive stability than single-dimensional scores. It also provides more robust clinical decision support for predicting post-RFCA recurrence across diverse datasets.

The APPLE, ATLAS, and Antwerp scores all demonstrated effectiveness in predicting AF recurrence after RFCA in patients with NVAF. Among these established scoring systems, the APPLE score showed better performance compared to the other two. More importantly, our newly developed nomogram exhibited superior performance compared to all three existing scores, demonstrating a marked improvement in predicting the risk of AF recurrence. While our model represents a promising tool, it is still in the preliminary stage and requires further validation in larger, multi-center, prospective cohorts to confirm its generalizability.