Un nuevo modelo machine learning para predecir los resultados de la Novel Trifecta y la validación de su pronóstico en una gran serie unicéntrica de nefrectomías parciales mínimamente invasivas

Abstract

Introduction

Novel trifecta has been proposed to define surgical success in patients undergoing partial nephrectomy (PN). This study aimed to evaluate its role in predicting survival outcomes and to identify the predictors of trifecta achievement via a machine-learning based model.

Methods

Data from consecutive patients who underwent minimally invasive PN between 2003 and 2022 were retrospectively analysed. Surgical success was defined through the achievement of novel trifecta (No Clavien-Dindo ≥ 3 postoperative complications, negative surgical margins and < 30% reduction in postoperative eGFR). Kaplan-Meier curves were used to assess the impact of trifecta achievement on survival. Automated chi-squared interaction detection (CHAID), a recursive machine-learning algorithm, was applied to predict trifecta achievement.

Results

A total of 465 patients were included, with a median follow-up of 72 (IQR: 38-108) months. The trifecta was achieved in 73% of cases and was associated with significantly improved overall survival (85% vs. 69.8%; P = .04). Solitary kidney status, patient age, and surgical experience were the most relevant predictors, allowing classification into three clusters. Trifecta rates were 43.6% in cluster #1 (solitary kidney), 71.2% in cluster #2 (non-solitary kidney with surgical experience ≤ 80 cases or > 80 cases and patients-age ≥ 65), and 88.5% in cluster #3 (non-solitary kidney, patients-age < 65, surgicall experience > 80 cases).

Conclusions

Novel trifecta achievement after PN correlates with overall survival but not with other oncological outcomes. A predictive model was developed, highlighting surgical experience as the only modifiable factor. Tailored perioperative strategies and advanced surgical training may improve trifecta achievement.