Abstract
Background
Systemic inflammatory response plays a critical role in tumor progression and immune evasion. Elderly patients with locally advanced cervical squamous cell carcinoma (CSCC) receiving definitive chemoradiotherapy (dCRT) exhibit high prognostic heterogeneity, and there is a lack of simple and effective risk stratification tools. This study aimed to evaluate the prognostic value of pretreatment peripheral blood inflammatory biomarkers and to construct a recursive partitioning analysis (RPA) risk stratification mode.
Methods
A total of 441 elderly patients with locally advanced CSCC who received dCRT were included. Variable selection was performed using LASSO regression, independent prognostic factors were identified via multivariate Cox regression, and an RPA model was constructed. The performance of the final RPA model was internally validated using bootstrap resampling (1,000 resamples) and 10-fold cross-validation, with the concordance index (C-index), area under the curve (AUC), calibration curves, and decision curve analysis (DCA) used as evaluation metrics.
Results
LASSO regression identified seven potential predictive factors from 18 candidate variables. Multivariable Cox regression further confirmed that NLR and SIRI were independent prognostic factors for overall survival and progression-free survival, and SHAP visualization analysis further validated their core contributions. Elevated NLR and SIRI levels were significantly associated with larger tumor diameter, higher FIGO stage, and poorer prognosis. Restricted cubic spline analysis demonstrated a linear association between NLR/SIRI and OS as well as PFS. A RPA risk stratification model based on NLR, SIRI, and FIGO stage classified patients into low-risk (5-year OS rate: 82.1%), intermediate-risk (70.0%), and high-risk (28.6%) groups. This model showed superior predictive performance compared to traditional FIGO stage and individual inflammatory biomarkers and remained robust across all subgroups.
Conclusion
Pretreatment NLR and SIRI are independent prognostic biomarkers for survival in elderly patients with locally advanced CSCC undergoing dCRT. The RPA model, which integrates NLR, SIRI and FIGO stage, can be used as an auxiliary prognostic stratification tool, but further external validation is required.