Scholarlink

Background <p>Post-progression survival (PPS) is a critical endpoint in oncology, yet predictors of PPS and data-driven strategies for post-progression management in locally advanced rectal cancer (LARC) remain undefined. This study aimed to identify prognostic factors for PPS, develop a predictive tool, and formulate risk-adapted surveillance strategies based on conditional survival to optimize long-term patient management.</p> Methods <p>The relationship between PPS and overall survival (OS) using spearman analysis. Independent predictors of PPS were identified through a combination of LASSO regression, machine learning approaches, and Cox regression, and a PPS prediction model was constructed. Model interpretability was elucidated using SHAP. Conditional PPS survival rates and annual recurrence risks were evaluated using Kaplan-Meier curve. Restricted cubic spline (RCS) analyses were employed to explore the association between a risk score and PPS.</p> Results <p>Among the cohort, 161 patients experienced disease progression. PPS was positively correlated with OS. Vascular tumor thrombus, perineural invasion, extramural venous invasion, body mass index, tumor size, and lymphocyte-to-monocyte ratio were identified as independent prognostic factors. The model achieved AUCs of 0.767, 0.801, and 0.802 at 1, 3, and 5 years, respectively, outperforming individual variables and TNM staging. The C-index of the model is 0.745. RCS analysis indicated a nonlinear relationship between higher risk score and worse PPS. After adjusting for other confounders, multi-model analysis demonstrated further supporting the positive association between risk score and adverse outcomes. Conditional survival analyses revealed that, across risk groups, longer PPS corresponded to higher survival probabilities and lower recurrence risks. Based on these findings, follow-up strategies should implement intensive surveillance for high-risk patients while adopting relatively streamlined monitoring for low-risk patients.</p> Conclusion <p>This study provides the first integrated, data-driven framework for predicting PPS and personalizing surveillance in LARC. The validated nomogram and conditional survival-based strategies enable risk-adapted follow-up, supporting a shift from empirical to evidence-based management after disease progression.</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

Leveraging data-driven risk-adapted surveillance strategies after disease progression in locally advanced rectal cancer

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月3日

查看全文
  • Jianjian Qiu,
  • Yuling Ye,
  • Yilin Yu,
  • Zhiping Wang,
  • Liang Hong,
  • Lingdong Shao,
  • Baihua Yang,
  • Junxin Wu

摘要

Background

Post-progression survival (PPS) is a critical endpoint in oncology, yet predictors of PPS and data-driven strategies for post-progression management in locally advanced rectal cancer (LARC) remain undefined. This study aimed to identify prognostic factors for PPS, develop a predictive tool, and formulate risk-adapted surveillance strategies based on conditional survival to optimize long-term patient management.

Methods

The relationship between PPS and overall survival (OS) using spearman analysis. Independent predictors of PPS were identified through a combination of LASSO regression, machine learning approaches, and Cox regression, and a PPS prediction model was constructed. Model interpretability was elucidated using SHAP. Conditional PPS survival rates and annual recurrence risks were evaluated using Kaplan-Meier curve. Restricted cubic spline (RCS) analyses were employed to explore the association between a risk score and PPS.

Results

Among the cohort, 161 patients experienced disease progression. PPS was positively correlated with OS. Vascular tumor thrombus, perineural invasion, extramural venous invasion, body mass index, tumor size, and lymphocyte-to-monocyte ratio were identified as independent prognostic factors. The model achieved AUCs of 0.767, 0.801, and 0.802 at 1, 3, and 5 years, respectively, outperforming individual variables and TNM staging. The C-index of the model is 0.745. RCS analysis indicated a nonlinear relationship between higher risk score and worse PPS. After adjusting for other confounders, multi-model analysis demonstrated further supporting the positive association between risk score and adverse outcomes. Conditional survival analyses revealed that, across risk groups, longer PPS corresponded to higher survival probabilities and lower recurrence risks. Based on these findings, follow-up strategies should implement intensive surveillance for high-risk patients while adopting relatively streamlined monitoring for low-risk patients.

Conclusion

This study provides the first integrated, data-driven framework for predicting PPS and personalizing surveillance in LARC. The validated nomogram and conditional survival-based strategies enable risk-adapted follow-up, supporting a shift from empirical to evidence-based management after disease progression.