# Data-driven prognostic factors analysis and personalized follow-up strategies for post-progression survival in locally advanced esophageal squamous cell carcinoma after definitive chemoradiotherapy

## TL;DR

This study identifies factors affecting survival after cancer recurrence in esophageal cancer patients and proposes personalized follow-up strategies based on data analysis.

## Contribution

The study introduces a data-driven model for predicting post-progression survival and tailoring follow-up care in esophageal squamous cell carcinoma patients.

## Key findings

- Prognostic factors like N stage, tumor length, and blood markers were linked to post-progression survival.
- Conditional survival analysis showed improved outcomes with longer survival times in risk groups.
- Personalized follow-up strategies were proposed based on risk scores and recurrence risks.

## Abstract

This study investigates clinical characteristics influencing post-progression survival (PPS) in locally advanced esophageal squamous cell carcinoma (ESCC) after definitive chemoradiotherapy (dCRT), aiming to develop individualized follow-up strategies using conditional PPS.

The correlation between PPS and overall survival (OS) using Spearman correlation analysis. LASSO regression, Cox regression, and machine-learning methods were employed to identify prognostic factors, and a prediction model was constructed. The Shapley additive explanations (SHAP) method was used to interpret the model. Conditional PPS survival rates and recurrence risks were analyzed.

This study enrolled 741 patients, with a median follow-up of 27.2 months. PPS was positively correlated with OS. Prognostic factors included: N stage, tumor length, chemotherapy cycles, platelet-to-albumin ratio, lymphocyte-to-monocyte ratio, age, body mass index, radiotherapy dose, and neutrophil to monocyte to lymphocyte ratio. Calibration curves, decision curves, and ROC curves demonstrated the model’s stability and predictive performance. Subgroup analyses suggested shorter PPS in high-risk patients. After adjusting for other confounders, multi-model analyses continued to show a positive association between the risk score and unfavorable PPS. Conditional PPS analyses across different risk groups revealed that, with increasing survival time, conditional PPS extended correspondingly, and the relapse risk gradually decreased. Finally, individualized follow-up strategies were proposed, indicating intensified monitoring for high-risk patients.

This study fills the research gap in the influencing factors of PPS and personalized follow-up strategies for patients with locally advanced ESCC after dCRT, and provides important clinical evidence for promoting the transformation of post-recurrence management from ‘experience-driven’ to ‘data-driven’.