Integrating sarcopenia and non-contrast CT radiomics for preoperative prediction of survival in sarcomatoid renal cell carcinoma

TL;DR

This study combines muscle mass measurements and CT scan features to predict survival in a rare and aggressive kidney cancer before surgery.

Contribution

The novel integration of sarcopenia and non-contrast CT radiomics for preoperative survival prediction in sarcomatoid renal cell carcinoma.

Findings

A combined model of sarcopenia and radiomic features outperformed radiomics-only models in predicting survival.

The model showed strong performance with AUCs of 0.849, 0.804, and 0.819 for 1-, 2-, and 3-year survival.

Sarcopenia and clinical factors were identified as independent prognostic indicators.

Abstract

Sarcomatoid renal cell carcinoma (sRCC) is an aggressive subtype with a poor prognosis. Preoperative prognostic tools are lacking, and the predictive value of sarcopenia combined with radiomic features from non-contrast CT remains unexplored.

In this retrospective study, 121 pathologically confirmed sRCC patients were enrolled. Sarcopenia was assessed using muscle mass measurements at the L3 level on preoperative non-contrast CT. Radiomic features were extracted from tumor regions of interest. Least absolute shrinkage and selection operator (LASSO) and Cox regression were used to select features and construct prognostic models for overall survival (OS). A combined model integrating sarcopenia status and radiomic signature (Rad-score) was developed and evaluated regarding its discrimination, calibration, and clinical utility.

Multivariable analysis identified paravertebral muscle-defined sarcopenia (HR = 3.046, p = 0.029), platelet-to-neutrophil ratio, hemoglobin-albumin-lymphocyte-platelet score, tumor size, and N stage as independent prognostic factors. The combined model (clinical + Rad-score) demonstrated superior predictive performance for 1-, 2-, and 3-year OS, with AUCs of 0.849, 0.804, and 0.819, respectively, and significantly outperformed the radiomics-only model (p = 0.002). Calibration curves and decision curve analysis confirmed its clinical applicability.

The integration of sarcopenia and non-contrast CT radiomics provides a valuable preoperative tool for predicting survival in sRCC patients, facilitating individualized risk stratification and clinical decision-making.

Figures