# Establishing and validating models integrated with hematological biomarkers and clinical characteristics for the prognosis of non-esophageal squamous cell carcinoma patients

## TL;DR

This study developed and validated two models to predict prognosis in non-esophageal squamous cell carcinoma patients using blood markers and clinical data.

## Contribution

The RSF model outperformed existing methods in predicting survival outcomes for non-esophageal squamous cell carcinoma patients.

## Key findings

- The RSF model achieved a higher concordance index (C-index) than the LASSO Cox model and TNM staging.
- Low-risk subgroups had better survival outcomes than high-risk subgroups in both models.
- The RSF model showed higher AUC values and better predictive power for individualized treatment guidance.

## Abstract

This study aimed to construct a novel model and validate its predictive power in non-esophageal squamous cell carcinoma (NESCC) patients.

This retrospective study included 151 patients between October 2006 and September 2016. The LASSO Cox and Random Survival Forest (RSF) models were developed with the help of hematological biomarkers and clinical characteristics. The concordance index (C-index) was used to assess the prognostic power of the LASSO Cox model, RSF model, and TNM staging. Based on the risk scores of the LASSO Cox and RSF models, we divided patients into low-risk and high-risk subgroups.

We constructed two models in NESCC patients according to LASSO Cox regression and RSF models. The RSF model reached a C-index of 0.841 (95% CI: 0.792–0.889) in the primary cohort and 0.880 (95% CI: 0.830–0.930) in the validation cohort, which was higher than the C-index of the LASSO Cox model 0.656 (95% CI: 0.580–0.732) and 0.632 (95% CI: 0.542–0.720) in the two cohorts. The integrated C/D area under the ROC curve (AUC) values for the LASSO Cox and RSF models were 0.701 and 0.861, respectively. In both two models, Kaplan-Meier survival analysis and the estimated restricted mean survival time (RMST) values indicated that the low-risk subgroup had a better prognostic outcome than the high-risk subgroup (p < 0.05).

The RSF model has better prediction power than the LASSO Cox and the TNM staging models. It has a guiding value for the choice of individualized treatment in patients with NESCC.