Scholarlink

Objectives <p>In colorectal cancer (CRC) patients, body composition (BC) has been recognized as a patient-specific imaging biomarker associated with prognosis. This study aims to construct a joint prediction model based on computed tomography (CT) quantitative body composition and clinical parameters using the Least Absolute Shrinkage and Selection Operator (LASSO) regression model and evaluate its prognostic value in patients undergoing curative surgery for CRC.</p> Materials and methods <p>This multicenter retrospective study analyzed 377 CRC patients who underwent surgical resection from 2017 to 2022. The cohort was split into a training set and a validation set. CT images at the L3 and umbilical levels within 3 months pre-surgery were used to quantify body composition. LASSO regression and COX proportional hazard regression identified independent prognostic factors and developed nomogram prediction models. Model performance was assessed using C-index, receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis (DCA), and external validation.</p> Results <p>LASSO regression identified 7 clinical indicators strongly associated with OS and DFS, respectively. Nomogram prediction models were then created incorporating body composition indicators. ROC curves showed the area under the curve (AUC) of 0.915 and 0.882 for 5 years OS and DFS, respectively. The combined model significantly outperformed TNM staging alone, with AUCs of 0.915 vs. 0.788 for 5 years OS in the training set, and 0.910 vs. 0.735 in the validation set. Calibration and DCA curves indicated strong predictive ability and clinical effectiveness.</p> Conclusion <p>The combined prediction model based on CT-quantified body composition and clinical indicators effectively predicts CRC prognosis, outperforming the single TNM stage model.</p> Clinical trial number <p>Retrospective study, not applicable.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Construction of a prognostic model for colorectal cancer based on clinical parameters and quantitative body composition using CT imaging

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月7日

查看全文
  • Zhihao Liu,
  • Mingming Song,
  • Yixin Heng,
  • Tong Nie,
  • Jiaxin Xu,
  • Xiaoyu Wu,
  • Liming Shen,
  • Yinghao Cao,
  • Feihong Wu,
  • Le Qin

摘要

Objectives

In colorectal cancer (CRC) patients, body composition (BC) has been recognized as a patient-specific imaging biomarker associated with prognosis. This study aims to construct a joint prediction model based on computed tomography (CT) quantitative body composition and clinical parameters using the Least Absolute Shrinkage and Selection Operator (LASSO) regression model and evaluate its prognostic value in patients undergoing curative surgery for CRC.

Materials and methods

This multicenter retrospective study analyzed 377 CRC patients who underwent surgical resection from 2017 to 2022. The cohort was split into a training set and a validation set. CT images at the L3 and umbilical levels within 3 months pre-surgery were used to quantify body composition. LASSO regression and COX proportional hazard regression identified independent prognostic factors and developed nomogram prediction models. Model performance was assessed using C-index, receiver operating characteristic (ROC) curves, calibration curves, decision curve analysis (DCA), and external validation.

Results

LASSO regression identified 7 clinical indicators strongly associated with OS and DFS, respectively. Nomogram prediction models were then created incorporating body composition indicators. ROC curves showed the area under the curve (AUC) of 0.915 and 0.882 for 5 years OS and DFS, respectively. The combined model significantly outperformed TNM staging alone, with AUCs of 0.915 vs. 0.788 for 5 years OS in the training set, and 0.910 vs. 0.735 in the validation set. Calibration and DCA curves indicated strong predictive ability and clinical effectiveness.

Conclusion

The combined prediction model based on CT-quantified body composition and clinical indicators effectively predicts CRC prognosis, outperforming the single TNM stage model.

Clinical trial number

Retrospective study, not applicable.