EPI‐VT Score: A Step Forward or a Misstep?

I read with curiosity the article authored by Kalaph et al. proposing a clinical score to determine the need for epicardial ablation in patients with ventricular tachycardia (VT) [1]. The authors claim that their score is 99% accurate in predicting the need for combined epicardial approach in patients undergoing endocardial VT ablation. While the simplicity of the score is commendable and the decision to exclude MRI findings is well justified, several methodological concerns arise upon closer examination of the article. First and foremost, the authors used data from a retrospective cohort; however, the sample size was not pre-calculated, which raises questions about the reliability of the findings. With less than 100 events of interest—epicardial ablations—the risk of underpowering the study is considerable [2]. Secondly, although the authors pre-specified some recognized predictors for their model, they fail to report a regression analysis or the process of converting coefficients into a numerical score [3]. This lack of defined scoring suggests that the assignment of scores to predictors may have been arbitrary. Third, the area under the curve (AUC) is only one of three parameters used to assess the performance of a prediction model: the authors inexplicably omitted the calibration slope and calibration in the large [2, 4]. Fourth, in the absence of internal validation, attributing an astonishing 99% discrimination ability to the model raises concerns about data overfitting [5]. Further questions emerge as one carefully reviews the article. For instance, why were well-established ECG predictors, such as pseudo-delta wave, negative QRS in limb leads, and the maximum deflection index (MDI), not selected as predictors [6, 7]? Additionally, a much larger multicenter retrospective analysis reported a need for epicardial mapping in almost 45% of their patients with nonischemic cardiomyopathy. In contrast, the proportion of epicardial ablations for this subgroup of patients was nearly 90% in the series by Kalaph (see Table 3 in the original article): is this an unacknowledged selection bias [1, 8]? Lastly, what is the clinical benefit of adjunctive epicardial ablation? A fully developed clinical score is useful when it accurately predicts an event with a reasonable net benefit [2]. The authors only report an unadjusted VT recurrence rate of 25% in the epicardial ablation group compared to 31% in the endocardial-only group after 1 year, and they regrettably do not mention acute procedural outcomes such as the incidence of complications, which in other series ranges from less than 1 to more than 20% [9]. Although the authors of the score occasionally disclose that their study is an exploratory retrospective cohort analysis, they disproportionately highlight results of an underpowered and incomplete analysis that is rather hypothesis-generating. In 1984, three Italian students deceived the art world by sculpting a fake Modigliani head, which was later found in a riverbed where the artist had allegedly discarded some of his sculptures 75 years prior. The clinical application of this tool should be delayed until proper analysis is conducted. The author has nothing to report.