Interpretable machine learning for predicting bone metastasis at initial diagnosis in elderly patients with small cell lung cancer: dual-centre development and external validation in two independent cohorts

Abstract Background Small-cell lung cancer (SCLC) is aggressive and often metastatic at diagnosis, and skeletal-related events such as bone pain, pathological fracture and spinal cord compression alter treatment decisions. Prediction tools for bone metastasis in elderly patients with SCLC remain scarce, and external validation and interpretability analyses are limited. We therefore developed and externally validated an interpretable machine learning model for bone metastasis at initial diagnosis in these patients, using routinely available clinical and staging information. Methods We retrospectively enrolled 1,387 consecutive patients aged ≥ 60 years with SCLC treated at two hospitals (2010–2024), randomly divided 7:3 into training and internal validation sets. Two external cohorts were assembled: 128 patients from another hospital (2010–2024) and 19,561 patients from the SEER database (2010–2015). The outcome was bone metastasis at initial diagnosis. Boruta selected features from 16 candidate variables. Six models—logistic regression, decision tree, random forest, XGBoost, LightGBM and artificial neural network—were compared for discrimination and calibration, with decision curve analysis and SHAP interpretation. Results Boruta confirmed four features: N stage, SCLC stage (limited-stage or extensive-stage), liver metastasis and clinical stage. LightGBM was the most stable across cohorts. SHAP identified clinical stage as the largest contributor, with extensive-stage disease, concurrent liver metastasis and higher N stage increasing the predicted risk. Conclusions A LightGBM model based on four features available at initial diagnosis provides stable, interpretable risk stratification for bone metastasis in elderly patients with SCLC and may support clinical decision-making, although recalibration should precede use in other populations.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-16
DOI
https://doi.org/10.21203/rs.3.rs-11570772/v1
Primary Topic
Lung Cancer Research Studies
Type
preprint

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Interpretable machine learning for predicting bone metastasis at initial diagnosis in elderly patients with small cell lung cancer: dual-centre development and external validation in two independent cohorts

Wenhao Hu, Hongyu Chen, Bu Wang, Jiayi Yu et al.
5 links
Research Square
Lung Cancer Research Studies
preprint

Interpretable machine learning for predicting bone metastasis at initial diagnosis in elderly patients with small cell lung cancer: dual-centre development and external validation in two independent cohorts

Wenhao Hu, Hongyu Chen, Bu Wang, Jiayi Yu, Yifei Zheng, Weibo Liu, Dajiang Sun, Zhihao Ma
preprint en

Abstract

Abstract Background Small-cell lung cancer (SCLC) is aggressive and often metastatic at diagnosis, and skeletal-related events such as bone pain, pathological fracture and spinal cord compression alter treatment decisions. Prediction tools for bone metastasis in elderly patients with SCLC remain scarce, and external validation and interpretability analyses are limited. We therefore developed and externally validated an interpretable machine learning model for bone metastasis at initial diagnosis in these patients, using routinely available clinical and staging information. Methods We retrospectively enrolled 1,387 consecutive patients aged ≥ 60 years with SCLC treated at two hospitals (2010–2024), randomly divided 7:3 into training and internal validation sets. Two external cohorts were assembled: 128 patients from another hospital (2010–2024) and 19,561 patients from the SEER database (2010–2015). The outcome was bone metastasis at initial diagnosis. Boruta selected features from 16 candidate variables. Six models—logistic regression, decision tree, random forest, XGBoost, LightGBM and artificial neural network—were compared for discrimination and calibration, with decision curve analysis and SHAP interpretation. Results Boruta confirmed four features: N stage, SCLC stage (limited-stage or extensive-stage), liver metastasis and clinical stage. LightGBM was the most stable across cohorts. SHAP identified clinical stage as the largest contributor, with extensive-stage disease, concurrent liver metastasis and higher N stage increasing the predicted risk. Conclusions A LightGBM model based on four features available at initial diagnosis provides stable, interpretable risk stratification for bone metastasis in elderly patients with SCLC and may support clinical decision-making, although recalibration should precede use in other populations.

Research Square
Hebei North University (CN), Chinese PLA General Hospital (CN), Peking University Cancer Hospital (CN)
National Natural Science Foundation of China, Natural Science Foundation of Hebei Province
Good health and well-being
Lung Cancer Research Studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon