Comparative validation and local adaptation of gastric cancer risk assessment models in a large multicenter endoscopic screening cohort in China.
4 affiliations
- 1Department of Cancer Prevention, Zhejiang Cancer Hospital, Hangzhou Institute of Medicine (HIM), Chinese Academy of Sciences, Hangzhou, 310022, Zhejiang, China.
- 2Department of Epidemiology, School of Public Health, Nanjing Medical University, Nanjing, 211166, Jiangsu, China.
- 3Department of Cancer Prevention, Zhejiang Cancer Hospital, Hangzhou Institute of Medicine (HIM), Chinese Academy of Sciences, Hangzhou, 310022, Zhejiang, China. [email protected].
- 4Department of Epidemiology, School of Public Health, Nanjing Medical University, Nanjing, 211166, Jiangsu, China. [email protected].
Abstract
Background
Gastric cancer (GC) is a major public health challenge in China, highlighting the need for effective risk stratification to improve early detection. This study aimed to validate existing GC risk assessment models and develop a locally adapted screening approach for GC and precancerous lesions.Methods
From 2022 to 2024, a total of 144,329 participants were recruited through a large multicenter endoscopic screening cohort in Zhejiang Province, China, with baseline data collected via interviews and endoscopic examinations. Participants screened in 2022 were used for model development, whereas those screened during 2023-2024 were used for validation and comparative analyses. Least absolute shrinkage and selection operator regression and logistic regression were used to construct a new model and a weighted point-based risk score. Six previously published models were externally validated for comparison across the validation set. Model performance and clinical utility were assessed and compared using area under the curve (AUC), detection rates, and decision curve analysis for both GC and the composite outcome of GC/high-grade intraepithelial neoplasia (HGIN).Results
A total of 1,419 GC cases were identified in the screening cohort. The newly developed model demonstrated favorable discriminative performance, with an AUC of 0.759 (95% CI: 0.744-0.775) for GC and 0.754 (95% CI: 0.740-0.768) for the composite outcome of GC/HGIN in the validation set. External evaluation of six previously published models showed AUCs ranging from 0.594 (95% CI: 0.558-0.629) to 0.742 (95% CI: 0.725-0.758) for GC, and from 0.602 (95% CI: 0.570-0.635) to 0.744 (95% CI: 0.730-0.759) for GC/HGIN. Risk stratification analyses demonstrated progressively increasing odds of both GC and GC/HGIN across higher-risk quartiles in most evaluated models. In the newly developed model, the highest-risk quartile accounted for 62.25% of GC cases and 58.56% of HGIN cases, indicating substantial enrichment of clinically significant lesions. Decision curve analysis indicated a favorable net benefit of the newly developed model across the evaluated range of threshold probabilities.Conclusions
The newly developed model demonstrated favorable discrimination, stable geographic performance, and effective risk stratification in a large multicenter screening cohort. These findings support its potential utility for risk-adapted GC screening in Chinese populations.
Related
- Cross-cultural adaptation and psychometric evaluation of the Japanese version of the scale for the assessment of non-experts’ AI literacy among medical trainees in a multicenter cross-sectional study
- Validation of an objective visual acuity detector for visual acuity assessment in young children: a multicenter, self-controlled study
- Translation, Cultural Adaptation, and Preliminary Validation of the Persian Version of the Assessment of Medical Education Environment by Teachers (AMEET) Questionnaire
- Cross-cultural adaptation and psychometric validation of the Kannada version of the Extended Effort Assessment Scale (EEAS-K) in adults with hearing loss
- Validation and culture adaptation of the Arabic version of the Rapid Eating Assessment for Participants - Shortened Version (REAP-S): A Confirmatory Factor Analysis and Rasch Analysis Study
BibTeX
@article{li2026comparative,
title = {{Comparative validation and local adaptation of gastric cancer risk assessment models in a large multicenter endoscopic screening cohort in China.}},
author = {Xue Li and Ying Chen and Juan Zhu and Wei-Yan Yu and Huan-Qing Tao and Gang Chen and Hui-Zhang Li and Ling-Bin Du},
journal = {BMC medicine},
year = {2026},
month = jul,
doi = {10.1186/s12916-026-05082-6},
url = {https://europepmc.org/article/MED/42902381},
note = {CC BY},
}
RIS
TY - JOUR
TI - Comparative validation and local adaptation of gastric cancer risk assessment models in a large multicenter endoscopic screening cohort in China.
AU - Xue Li
AU - Ying Chen
AU - Juan Zhu
AU - Wei-Yan Yu
AU - Huan-Qing Tao
AU - Gang Chen
AU - Hui-Zhang Li
AU - Ling-Bin Du
PY - 2026
DA - 2026/07/01
JO - BMC medicine
DO - 10.1186/s12916-026-05082-6
UR - https://europepmc.org/article/MED/42902381
AB - Background Gastric cancer (GC) is a major public health challenge in China, highlighting the need for effective risk stratification to improve early detection. This study aimed to validate existing GC risk assessment models and develop a locally adapted screening approach for GC and precancerous lesions.Methods From 2022 to 2024, a total of 144,329 participants were recruited through a large multicenter endoscopic screening cohort in Zhejiang Province, China, with baseline data collected via interviews and endoscopic examinations. Participants screened in 2022 were used for model development, whereas those screened during 2023-2024 were used for validation and comparative analyses. Least absolute shrinkage and selection operator regression and logistic regression were used to construct a new model and a weighted point-based risk score. Six previously published models were externally validated for comparison across the validation set. Model performance and clinical utility were assessed and compared using area under the curve (AUC), detection rates, and decision curve analysis for both GC and the composite outcome of GC/high-grade intraepithelial neoplasia (HGIN).Results A total of 1,419 GC cases were identified in the screening cohort. The newly developed model demonstrated favorable discriminative performance, with an AUC of 0.759 (95% CI: 0.744-0.775) for GC and 0.754 (95% CI: 0.740-0.768) for the composite outcome of GC/HGIN in the validation set. External evaluation of six previously published models showed AUCs ranging from 0.594 (95% CI: 0.558-0.629) to 0.742 (95% CI: 0.725-0.758) for GC, and from 0.602 (95% CI: 0.570-0.635) to 0.744 (95% CI: 0.730-0.759) for GC/HGIN. Risk stratification analyses demonstrated progressively increasing odds of both GC and GC/HGIN across higher-risk quartiles in most evaluated models. In the newly developed model, the highest-risk quartile accounted for 62.25% of GC cases and 58.56% of HGIN cases, indicating substantial enrichment of clinically significant lesions. Decision curve analysis indicated a favorable net benefit of the newly developed model across the evaluated range of threshold probabilities.Conclusions The newly developed model demonstrated favorable discrimination, stable geographic performance, and effective risk stratification in a large multicenter screening cohort. These findings support its potential utility for risk-adapted GC screening in Chinese populations.
SN - 1741-7015
ER -