Brain metastases (BMs) account for a relatively large proportion of patients with metastatic non-small cell lung cancer (NSCLC) both at the time of diagnosis and during the course of disease, which is a major cause leading to a significant increase in cancer-related mortality for NSCLC patients. We aimed to build a novel risk classification system to predict overall survival (OS) of patients presenting with BMs at the first diagnosis of NSCLC.
Data on NSCLC patients with BMs was extracted from the SEER database between 2010 and 2013. All cases were randomly divided into the training cohort and validation cohort (7:3). Cox regression was performed to select the independent predictors of OS. A nomogram was established for predicting 6-, 12- and 18-month OS. The prognostic performance of nomogram was evaluated using concordance indexes (C-indexes), calibration curves, and decision curve analyses (DCAs). All statistical analysis was performed with R software.
According to the multivariate Cox regression of the training set, the following variables were contained in the nomogram prognostic model: age, marital status, sex, race, primary site, histological type, T stage, N stage, metastatic pattern, whether received radiotherapy and chemotherapy. The nomogram displayed good accuracy in predicting 6, 12 and 18 months OS. The C-indexes of training set and validation set were 0.729 and 0.723, respectively. The calibration curves showed excellent consistency between the actual results and the nomogram prediction, and the DCAs exhibited great clinical utility of the model. A new risk classification system based on nomogram scores was introduced to classify NSCLC patients with BMs into three subgroups. In the training cohort, the median OS of patients in the low-, intermediate- and high-risk groups was 11.0 months (95% confidence interval [CI] 10.3–11.7), 5.0 months (95% CI 4.6–5.4), and 2.0 months (95% CI 1.9–2.1), respectively.
We established a risk stratification system based on nomogram scores to predict the OS of patients presenting with BMs at initial diagnosis of NSCLC. These predictive models can act as auxiliary tools to guide treatment strategies and prognostic evaluation.
Clinical trial identification
Legal entity responsible for the study
Has not received any funding.
All authors have declared no conflicts of interest.