Development and validation of a nomogram to predict postoperative cancer-specific survival of patients with nonmetastatic T3a renal cell carcinoma.

PURPOSE

To establish a nomogram for the prediction of postoperative cancer-specific survival (CSS) in patients with nonmetastatic T3a renal cell carcinoma (RCC).

METHODS

The Surveillance, Epidemiology, and End Results database were searched for patients with pT3aN0-1M0 RCC between 2010 and 2018. The patients were randomly stratified into the training and verification group (7:3 ratio). Using Cox regression analysis, the predictors for the CSS in the training group were integrated to establish the nomogram for predicting the 3-year and 5-year CSS. Harrell's concordance index (C-index), time-dependent receiver operating characteristic curve, decision curve analysis, and Kaplan-Meier survival analysis were used to evaluate the nomogram performance.

RESULTS

A total of 5,791 pT3aN0-1M0 RCC cases with eligible data were selected from the Surveillance, Epidemiology, and End Results database. Age, tumor size, surgery type, Fuhrman grade, histological type, sarcomatoid, N stage, and invasion patterns were identified as the significant predictors for CSS to establish the nomogram. The C-indices of the nomogram were 0.774 (95% CI: 0.753-0.795) and 0.777 (95% CI: 0.745-0.809) for the training and verification group, respectively. The calibration of the nomogram revealed consistency between the predicted and observed survival. The area under the 3-year and 5-year CSS receiver operating characteristic curves were 0.773 and 0.786 in the training group, respectively. Decision curve analysis showed the optimal application of the model in clinical decision-making. According to the cutoff values of prognostic indices, patients with low-risk showed better CSS than those with high-risk in both training and verification groups (both P< 0.0001).

CONCLUSION

The current nomogram could effectively predict the CSS of patients with nonmetastatic T3a RCC, and could be used to identify patients who might need a compact interval of follow-up and postoperative adjuvant systemic treatment. The limitations included the retrospective nature, absence of external validation, and several unmeasured variables related to the selection bias of surgery type. The results should be interpreted with caution.