Introduction

The incidence of brain metastases in patients with renal cell carcinoma (RCC) is hypothesized to have increased in the last two decades. Our objective was to provide an overview of recent incidence trends in patients with primary renal cell carcinoma (RCC) and brain metastases at diagnosis using a nationally representative cancer cohort originating from the United States. Secondary objectives include developing a tool for prediction of brain metastases at diagnosis and to assess their oncological outcomes, and externally validating it in second database._x000D_

Methods

All patients with a primary diagnostic confirmation of RCC within the Surveillance, Epidemiology, and End Results (SEER, 2010-2013) database and the National Cancer Database (NCDB, 2010-2012) were abstracted. The incidence proportions (%IP) and 95% confidence intervals (CI) of brain metastases were calculated overall and according to patient, sociodemographic, and disease characteristics. A 1000-bootstrap corrected multivariable logistic regression models was developed for prediction of brain metastases at diagnosis using the SEER cohort (development). Backward variable selection was conducted to identify the most parsimonious model. Model performance was evaluated via measures of predictive accuracy in the NCDB cohort (validation).

Results

The overall %IP of brain metastases at RCC diagnosis was 1.51% (95% CI: 1.39-1.64) in the SEER and 1.37% (95% CI: 1.29-1.45%) in the NCDB. The odds of harbouring brain metastases at RCC diagnosis varied significantly according to sociodemographic and clinical characteristics. Following backward variable selection within the SEER database, only histology, tumor size, and cN stages were retained in the final model. Predictive accuracy was adequate in the external validation cohort (C-statistic: 0.778). Median time to any death was 6.37 months in patients with brain metastases. After adjusting for confounders, patients with brain metastases were more likely to succumb to any death than those without brain metastases at diagnosis (hazard ratio: 1.87, 95% CI: 1.71-2.05, P<0.001).

Conclusions

The incidence of brain metastases in patients with RCC is increasing. The oncological outcomes of such patients remain poor and their treatment management variable. A clinical risk model including cN-stage, histology and tumor size can predict the presence of brain metastases at diagnosis and may justify baseline imaging in asymptomatic but high-risk patients.

Funding

none