Predictive Nomograms for Overall Survival and Cancer-specific Survival of Patients with Non-small Cell Lung Cancer and Concurrent Brain Metastasis: A SEER Database Analysis

Introduction: Brain metastasis in non-small cell lung carcinoma (NSCLC) forebodes poor prognosis. Identifying prognostic factors and constructing a nomogram to predict overall survival (OS) and cancer-specific survival (CSS) is essential to guide treatment in such patients.

Methods: SEER database was surveyed between 1975 to 2019 to include data on patients diagnosed with synchronous NSCLC and brain metastasis. We identified and randomized 1743 patients into a training (n=1220) and validation group (n=523) using a 7:3 ratio. Cox proportional-hazards regression models were used to determine significant prognostic factors used to build the OS and CSS nomograms. The predictive value of both nomograms at 6-, 12-, 18-, and 24-months was assessed using c-indices, AUC values, and calibration curves.

Results: Overall, patients were predominantly in the ≥60 years group (n=1325/1743; 76%), and 53.24% were females. The median OS and CSS were 9 months (95% CI: 8.13-9.87) and 11 months (95% CI: 9.84-12.15), respectively. Multivariate Cox regression identified age, race, sex, histology, T stage, metastasis to bone, metastasis to liver, surgery, radiotherapy and chemotherapy status as independent prognostic factors affecting OS and CSS. In the training cohort, the C-indices for the OS nomogram and the CSS nomogram were 0.743 (95% CI: 0.723-0.763) and 0.758 (95% CI: 0.710-0.754), respectively. In the validation cohort, the C-indices for the OS nomogram and CSS nomogram were 0.759 (95% CI: 0.699-0.819) and 0.746 (95% CI: 0.682-0.810), respectively. The AUC values at 6-, 12-, 18-, and 24-months were 0.777, 0.758, 0.740, and 0.730 for OS and 0.775, 0.744, 0.731, and 0.744 for CSS, respectively. The calibration curves showed excellent consistency between the predicted and actual survival probabilities for both OS and CSS.

Conclusion : We established a nomogram that demonstrated a great capability at predicting OS and CSS up to 2-years using multiple relevant demographic and clinical features in NSCLC patients with concurrent brain metastasis.