Optimizing primary tumor resection decisions for non-small cell lung cancer patients with brain metastases: analysis based on the SEER database and development and external validation of predictive models

Abstract

Background: Brain metastases (BM) in non-small cell lung cancer (NSCLC) are associated with poor prognosis, and the role of primary tumor resection (PTR) remains controversial. This study aims to evaluate the role of PTR in NSCLC patients with BM and develop a predictive tool to identify optimal surgical candidates.

Methods: We analyzed data from 608 patients who underwent PTR and 18,529 patients who did not, sourced from the Surveillance, Epidemiology, and End Results (SEER) database. Propensity score matching (PSM), a statistical method to balance baseline characteristics between treatment groups, was performed to minimize selection bias. We utilized Kaplan-Meier (KM) survival analysis, Cox regression analysis, and PSM to assess the impact of PTR on the prognosis of BM patients. Logistic regression analysis was performed to identify and quantify the clinical factors that influence the benefits of PTR in BM patients, which informed the development of a predictive model. The model's predictive accuracy and clinical applicability were evaluated using the Concordance index (C-index), F1 score, receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA), with 42 patients from Fujian Provincial Hospital serving as the external validation cohort.

Results: KM survival analysis revealed significant survival differences between the PTR and non-PTR groups, both before and after PSM, with median survival times of 15 and 7 months, respectively. PTR was identified as an independent clinical factor affecting the prognosis of BM patients. In the training set, the C-index for the model was 0.7857 [95% confidence interval (CI): 0.7387-0.8327]. In the validation set, the C-index was 0.7441 (95% CI: 0.6674-0.8208), and in the external validation set, it was 0.7759 (95% CI: 0.5855-0.9662). The F1 score of the predictive model was 0.7667. Analyses of the ROC curve, calibration curve, and DCA indicated that the model exhibits strong accuracy and clinical applicability.

Conclusions: This study indicated that PTR can enhance the prognosis of NSCLC patients with BM and also establishes an effective screening tool for quantifying the likelihood of patient benefiting from PTR treatment.

Keywords: Non-small cell lung cancer (NSCLC); Surveillance, Epidemiology, and End Results (SEER); brain metastases (BM); primary tumor resection.

Figure 1

This research outlines the patient selection process and study design workflow. NSCLC, non-small cell lung cancer; PSM, propensity score matching; PTR, primary tumor resection; SEER, Surveillance, Epidemiology, and End Results.

Figure 2

Kaplan-Meier survival curves before (A) and after (B) propensity score matching. PTR, primary tumor resection.

Figure 3

Nomogram for predicting the probability of PTR benefit in patients with NSCLC and brain metastasis. N, node; NSCLC, non-small cell lung cancer; PTR, primary tumor resection; T, tumor.

Figure 4

A web-based nomogram for predicting the probability of PTR benefit in NSCLC patients with brain metastases. NSCLC, non-small cell lung cancer; PTR, primary tumor resection.

Figure 5

ROC curves for the training set, validation set, and external validation set. AUC, area under the curve; ROC, receiver operating characteristic.

Figure 6

Calibration curves for the training set (A), validation set (B), and external validation set (C).

Figure 7

DCA curves for the training set (A), validation set (B), and external validation set (C). DCA, decision curve analyses.

Figure 8

The Kaplan-Meier survival curves for the high and low probability survival groups curves in the SEER cohort (A) and the external cohort (B) are shown. SEER, Surveillance, Epidemiology, and End Results.