# Primary tumor resection: a new hope or an old illusion for patients with metastatic non-small cell lung neuroendocrine tumors?

## TL;DR

This study explores whether removing the primary tumor improves survival for patients with metastatic non-small cell lung neuroendocrine tumors and develops a tool to predict which patients might benefit from surgery.

## Contribution

The study introduces a predictive nomogram based on machine learning to identify patients who may benefit from primary tumor resection in metastatic non-small cell lung neuroendocrine tumors.

## Key findings

- Patients who underwent surgery had significantly longer median overall survival (26 months vs. 11 months) after propensity score matching.
- A logistic regression model with an AUC of 0.760 was used to build a nomogram predicting surgical benefit, showing 30 months vs. 10 months median survival for beneficiary groups.
- The web-based nomogram may aid clinicians in personalized decision-making for patient counseling.

## Abstract

This study aimed to investigate the impact of primary tumor resection (PTR) on survival outcomes for patients with metastatic non-small cell neuroendocrine tumors (mNSCLC-NETs), develop a predictive model to identify which patients may benefit from surgery in terms of survival.

We extracted information on mNSCLC-NET patients from the SEER database. Propensity score matching was used to eliminate bias between surgery and non-surgery groups. The effect of PTR on prognosis was assessed via Kaplan‒Meier analysis with the log-rank test and the Cox proportional hazards model. Feature selection was performed via the Boruta algorithm. Model building utilized fivefold cross-validation and applied five machine learning algorithms. The optimal model was selected and used to construct a visual network nomogram.

Among the 1,776 eligible patients, 12.61% underwent surgery. After PSM, the surgery group showed significantly longer median overall survival (mOS) (26 months vs. 11 months) compared to the non-surgery group. Among the five machine learning models, logistic regression had the highest AUC of 0.760 on the validation set. Therefore, we used a logistic regression model to construct a nomogram. This tool identified beneficiary and non-beneficiary groups, with the former having a longer mOS (30 months vs. 10 months).

Overall, PTR in mNSCLC-NETs could prolong patients survival, and the web-based nomogram can predict patients who may benefit from surgery. This tool may aid clinicians in patient counseling and personalized decision-making.

The online version contains supplementary material available at 10.1186/s12957-025-04063-y.