Machine learning algorithms in predicting survivability of patients with renal cell carcinoma after nephrectomy: a...

Poor prognosis brings great physical suffering and financial burden to patients with renal cell carcinoma (RCC) after nephrectomy. This study aims to explore the application of machine learning for feature selection in predicting survivability and construct a well-performed prognostic model for identifying and managing the high-risk patients.

We retrospectively analyzed 737 patients with RCC after nephrectomy. Important features were respectively selected by least absolute shrinkage and selection operator (LASSO) regression and random survival forest (RSF), and the LASSO-Cox model and RSF-Cox model were constructed in conjunction with Cox regression. And their predictive performance were evaluated and compared by the C-index, calibration curve, decision curve analysis (DCA), area under the curve (AUC) of the receiver operating characteristic (ROC), and Kaplan-Meier curve. Besides, a Cox model was constructed using all clinical variables and compared with the C-index and AUC of the two models described above to demonstrate the necessity of feature selection.

A total of 725 cases fitted this study ultimately, of which 48 died during the period of follow-up. The shared variables for the two models were tumor size, preoperative plasma fibrinogen content, N stage, and Fuhrman grade. In the training set, the C-index of the Cox, LASSO-Cox and RSF-Cox was 0.863, 0.893 and 0.874, and the 5-year AUC was 0.816, 0.880 and 0.837. And in the validation set, the C-index was 0.837, 0.856 and 0.821, and the 5-year AUC was 0.790, 0.855 and 0.852. The calibration and DCA curves suggested that the LASSO-Cox model outperformed the RSF-Cox model in survival prediction and net benefit. Significant survival differences were observed between the low and high-risk groups.

The LASSO-Cox model we constructed has been simplified and obtained higher efficiency, which can help to inform early intervention and clinical decision-making.