Prediction of the individual risk for the development of brain metastases in patients with non-oncogene-addicted non-small cell lung cancer: Real-world data from the German prospective CRISP registry (AIO-TRK-0315)

Abstract

Brain metastases (BM) from non-small cell lung cancer (NSCLC) are a relevant cause of morbidity and mortality. This study aims to identify a population at high risk for BM in order to inform personalized follow-up strategies in the future. We have therefore developed and validated a point-based risk classifier for the risk of BM1 in patients with metastatic non-oncogene-addicted NSCLC without BM at disease onset. A total of 3139 patients were included from the German CRISP lung cancer registry and randomly assigned to a derivation and validation set (2:1). In the derivation set, a Fine and Gray model was developed considering BM and death competing risks. A point-based risk classifier was derived and evaluated in the validation set. Of the baseline variables included in the risk model, ≥4 versus 0 extrathoracic metastatic sites (hazard ratio [HR], 2.69; 95% CI, 1.45-5.00), N3 vs. N0 disease (2.29; 1.41-3.72), and age < 65 vs. ≥75 years (2.03; 1.39-2.98) were the strongest predictors of BM development. Using the risk classifier, high-risk patients in the derivation set had a cumulative incidence of BM of 23.7% (95% CI, 16.7-30.1) at 18 months compared to 10.3% (8.9-11.7) in low-risk patients (HR, 2.44; 95% CI, 1.87-3.19). In the validation set, the cumulative incidence of BM at 18 months was 14.7% (11.5-17.8) and 8.9% (6.2-11.6) in the high-risk and low-risk groups, respectively (HR, 1.54; 95% CI, 1.08-2.20). In conclusion, our point-based risk classifier is able to identify NSCLC patients at high risk for BM development that may benefit from intensified brain surveillance.

Keywords: NSCLC; brain metastases; follow‐up; risk factors; risk model.