UroPredict: Machine learning model on real-world data for prediction of kidney cancer recurrence (UroCCR-120)

Gaëlle Margue,Loïc Ferrer,Guillaume Etchepare,Pierre Bigot,Karim Bensalah,Arnaud Mejean,Morgan Roupret,Nicolas Doumerc,Alexandre Ingels,Romain Boissier,Géraldine Pignot,Bastien Parier,Philippe Paparel,Thibaut Waeckel,Thierry Colin,Jean-Christophe Bernhard

DOI: https://doi.org/10.1038/s41698-024-00532-x

2024-02-23

npj Precision Oncology

Abstract:Abstract Renal cell carcinoma (RCC) is most often diagnosed at a localized stage, where surgery is the standard of care. Existing prognostic scores provide moderate predictive performance, leading to challenges in establishing follow-up recommendations after surgery and in selecting patients who could benefit from adjuvant therapy. In this study, we developed a model for individual postoperative disease-free survival (DFS) prediction using machine learning (ML) on real-world prospective data. Using the French kidney cancer research network database, UroCCR, we analyzed a cohort of surgically treated RCC patients. Participating sites were randomly assigned to either the training or testing cohort, and several ML models were trained on the training dataset. The predictive performance of the best ML model was then evaluated on the test dataset and compared with the usual risk scores. In total, 3372 patients were included, with a median follow-up of 30 months. The best results in predicting DFS were achieved using Cox PH models that included 24 variables, resulting in an iAUC of 0.81 [IC95% 0.77–0.85]. The ML model surpassed the predictive performance of the most commonly used risk scores while handling incomplete data in predictors. Lastly, patients were stratified into four prognostic groups with good discrimination (iAUC = 0.79 [IC95% 0.74–0.83]). Our study suggests that applying ML to real-world prospective data from patients undergoing surgery for localized or locally advanced RCC can provide accurate individual DFS prediction, outperforming traditional prognostic scores.

oncology

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to more accurately predict the individual disease - free survival (DFS) in patients after renal cell carcinoma (RCC) surgery. The existing prognostic scoring systems offer relatively limited predictive performance, leading to challenges in postoperative follow - up recommendations and in selecting patients who may benefit from adjuvant therapy. By applying machine learning (ML) to analyze real - world data, the study aims to develop a model that can provide more precise individual DFS predictions, thus surpassing traditional prognostic scoring systems.

Specifically, the research objectives include:
1. **Predicting individual DFS based on baseline multimodal data**: Using patients' clinical, pathological and biological data, predict patients' disease - free survival after surgery through a machine - learning model.
2. **Stratifying patients into different risk groups**: Identify patient groups with extremely low and high recurrence risks, so as to develop more personalized management plans for these patients, such as reducing the follow - up intensity for low - risk patients or considering providing adjuvant therapy for high - risk patients.

By solving these problems, the study hopes to improve the accuracy of predicting the recurrence risk of patients after kidney cancer surgery, thereby optimizing patients' treatment and follow - up strategies.