Objectives Postoperative survival outcomes vary substantially among patients diagnosed with stage I lung adenocarcinoma (LUAD). This study aimed to develop CT-based radiomic subtypes using unsupervised clustering to assess their association with overall survival (OS), systemic nutritional-inflammatory status, and adjuvant chemotherapy benefit. Materials and methods A total of 496 stage I LUAD patients from two independent centers were included. Preoperative CT radiomic features (n = 1218) were extracted, and subtypes were derived using the K-means clustering algorithm. The independent prognostic value of these subtypes, along with their capacity to predict the benefit of adjuvant chemotherapy, was evaluated through multivariable Cox regression and treatment-by-subtype interaction analyses. Results Three radiomic subtypes with significant prognostic differences in OS were identified. The high-risk subtype, Cluster 2, exhibited distinct clinical characteristics and was associated with markedly poorer OS (hazard ratio [HR] = 15.71, p < 0.001, compared to Cluster 0). Cluster 2 also showed an inflammatory imbalance, with elevated systemic immune-inflammation index and neutrophil-to-lymphocyte ratio, and  a decreased lymphocyte-to-monocyte ratio. Notably, a significant interaction was found between subtypes and adjuvant chemotherapy (interaction p < 0.001, Cluster 2 vs Cluster 0). Subgroup analysis indicated that stage IB patients within Cluster 2 derived a significant survival benefit from adjuvant chemotherapy (interaction p = 0.003 vs Cluster 0). Conclusions This study developed a CT-based radiomic subtype system using unsupervised clustering that identifies high-risk stage I LUAD patients with systemic inflammatory imbalance. Notably, these subtypes predict differential survival benefits from adjuvant chemotherapy in high-risk stage IB patients, thereby supporting personalized postoperative treatment strategies. Critical relevance statement This CT-based radiomic subtype system stratifies prognosis and identifies stage I LUAD patients who may benefit from adjuvant chemotherapy, enabling personalized treatment decisions in radiology. Key Points <UnorderedList Mark="Bullet"> <ItemContent> Conventional tumor-node-metastasis (TNM) staging does not adequately capture tumor heterogeneity in stage I LUAD. </ItemContent> <ItemContent> Three CT-based radiomic subtypes were established, with the high-risk subgroup correlating with systemic inflammatory imbalance and poorer OS. </ItemContent> <ItemContent> CT-based radiomic stratification identifies stage IB patients who benefit from adjuvant chemotherapy, supporting personalized postoperative management. </ItemContent> </UnorderedList> Graphical Abstract

Radiomic subtypes predict survival and chemotherapy benefit in stage I lung adenocarcinoma: a multicenter study

期刊论文
出版模式：

开放获取
发表日期：

2026年3月2日

查看全文

Guangyu Tao,
Dongying Wang,
Xin Cheng,
Zhenghai Lu,
Hua Zhong,
Hong Yu,
Wei Nie

摘要

Objectives

Postoperative survival outcomes vary substantially among patients diagnosed with stage I lung adenocarcinoma (LUAD). This study aimed to develop CT-based radiomic subtypes using unsupervised clustering to assess their association with overall survival (OS), systemic nutritional-inflammatory status, and adjuvant chemotherapy benefit.

Materials and methods

A total of 496 stage I LUAD patients from two independent centers were included. Preoperative CT radiomic features (n = 1218) were extracted, and subtypes were derived using the K-means clustering algorithm. The independent prognostic value of these subtypes, along with their capacity to predict the benefit of adjuvant chemotherapy, was evaluated through multivariable Cox regression and treatment-by-subtype interaction analyses.

Results

Three radiomic subtypes with significant prognostic differences in OS were identified. The high-risk subtype, Cluster 2, exhibited distinct clinical characteristics and was associated with markedly poorer OS (hazard ratio [HR] = 15.71, p < 0.001, compared to Cluster 0). Cluster 2 also showed an inflammatory imbalance, with elevated systemic immune-inflammation index and neutrophil-to-lymphocyte ratio, and a decreased lymphocyte-to-monocyte ratio. Notably, a significant interaction was found between subtypes and adjuvant chemotherapy (interaction p < 0.001, Cluster 2 vs Cluster 0). Subgroup analysis indicated that stage IB patients within Cluster 2 derived a significant survival benefit from adjuvant chemotherapy (interaction p = 0.003 vs Cluster 0).

Conclusions

This study developed a CT-based radiomic subtype system using unsupervised clustering that identifies high-risk stage I LUAD patients with systemic inflammatory imbalance. Notably, these subtypes predict differential survival benefits from adjuvant chemotherapy in high-risk stage IB patients, thereby supporting personalized postoperative treatment strategies.

Critical relevance statement

This CT-based radiomic subtype system stratifies prognosis and identifies stage I LUAD patients who may benefit from adjuvant chemotherapy, enabling personalized treatment decisions in radiology.

Key Points

Conventional tumor-node-metastasis (TNM) staging does not adequately capture tumor heterogeneity in stage I LUAD.

Three CT-based radiomic subtypes were established, with the high-risk subgroup correlating with systemic inflammatory imbalance and poorer OS.

CT-based radiomic stratification identifies stage IB patients who benefit from adjuvant chemotherapy, supporting personalized postoperative management.

Graphical Abstract