Commentary: Concerns regarding the predictive performance of the composite index for mitral regurgitation improvement after TAVI

Résumé

First, the development of a multivariable predictive model using only 53 patients creates a substantial risk of overfitting. According to TRIPOD recommendations, prognostic models should meet adequate sample-size requirements, with at least 10-20 events per variable to ensure model stability¹. In this study, the improvement event occurred in about 36 patients, which is insufficient for a multi-parameter model and strongly predisposes to optimistic estimates.Second, the authors report an AUC of 0.909 and characterize it as "outstanding discriminative ability." However, this value was calculated on the same dataset used for model derivation. Without internal validation techniques-such as bootstrap resampling or k-fold cross-validation-the performance is likely optimistic². The lack of external validation further limits confidence in the generalizability of the proposed index.Third, the article implies potential clinical utility of the index. Yet predictive models intended for clinical decision-making must undergo rigorous validation, include calibration assessment, and be tested in independent populations³. Without these steps, the index cannot be considered ready for clinical application and should instead be viewed as preliminary and hypothesis-generating.Fourth, qualitative echocardiographic variables such as "eccentric MR" require strict standardization. Echocardiographic assessment of regurgitation is highly dependent on hemodynamic conditions and imaging windows, and inconsistent definitions may introduce classification bias⁴.Finally, the manuscript would benefit from stronger adherence to TRIPOD reporting standards, including transparent reporting of model development, handling of missing data, and justification of sample size¹.In conclusion, while Xiao et al. address an important question, the composite index should be interpreted with caution due to methodological limitations intrinsic to small-sample predictive modeling without validation. Larger multicenter studies are needed before clinical implementation can be considered.

Citer ce document

Accès au document

Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter

Voir l'article sur le site de la revue

Auteur(s)

Statistiques

Consultations : 1

Téléchargements : 0