The EASY-WPW algorithm in practice: real-world accuracy in predicting accessory pathway locations

Abstract Introduction Accurate localization of accessory pathways (APs) is essential for effective ablation therapy. The EASY-WPW algorithm, published in 2023, is designed to help clinicians predict AP locations based on baseline ECGs, potentially enhancing procedural efficiency. However, external validation of the algorithm's accuracy and generalizability remains limited. Aim This study aimed to assess the sensitivity and accuracy of the EASY-WPW algorithm in a cohort of patients who underwent successful AP ablation at a tertiary center in Portugal. Methods We conducted a retrospective analysis of patients who underwent successful AP ablation from 2021 to 2024. Baseline characteristics and procedural data were collected. Two electrophysiology department members, including a cardiology resident and an electrophysiologist (EP), independently assessed baseline ECGs using the EASY-WPW algorithm. Additionally, an accessible artificial intelligence (AI)-based tool (ChatGPT) was utilized to apply the algorithm to ECGs. The algorithm's sensitivity and positive predictive value (PPV) were calculated based on results from electrophysiological (EP) studies. Results A total of 154 patients (72.5% male, n=111) with a mean age of 33.9 ± 18.5 years were included. The mean body mass index (BMI) was 24.3 ± 5.4 kg/m², and the mean heart rate was 75.0 ± 17.3 bpm. Among the cohort, 27.1% were under 18 years old. The algorithm correctly identified AP location in 59% of cases, yielding a PPV of 57.8% and a sensitivity of 55.7%. PPVs for right-sided, left-sided, septal, and lateral pathways were 55.6%, 62.5%, 58.3%, and 56.8%, respectively, with no significant differences between them. No significant differences in algorithm accuracy were found based on patient age (t=0.73, p=0.23), BMI (t=-10.16, p=0.311), or heart rate (t=-1.38, p=0.16). Similarly, clinician experience (resident vs. electrophysiologist) did not significantly affect accuracy, with good agreement between resident and EP (55.8% vs. 59%; kappa=0.76; p=0.001). ChatGPT’s performance in predicting AP locations was significantly lower compared to clinicians (26.5% vs. 59%). Conclusion In this cohort, the EASY-WPW algorithm showed moderate accuracy, correctly localizing APs in 59% of cases. Patient age, BMI, heart rate, and clinician experience did not significantly impact algorithm performance. Our findings suggest that clinicians, especially those new to electrophysiology, should exercise caution when applying simplified algorithms across diverse populations. Furthermore, while the use of ChatGPT may be tempting, it is currently not advisable for this purpose given its lower accuracy in predicting AP locations.