A multicenter prospective cohort study developing and validating a SIRI-based machine learning model and simplified...

Area of Science:

• Medical research
• Biomarker discovery
• Inflammation and disease

Background:

• Postherpetic neuralgia (PHN) is a severe complication of herpes zoster (HZ).
• Early identification of patients at high risk for PHN remains a clinical challenge.
• The systemic inflammatory response index (SIRI) is a novel inflammatory marker with prognostic value in various diseases, but its role in PHN risk has not been established.

Purpose of the Study:

• To investigate the association between SIRI and the risk of developing PHN.
• To develop and validate a predictive model for PHN using machine learning algorithms.
• To create a clinically practical risk stratification tool for PHN.

Main Methods:

• A cohort of 1361 patients with herpes zoster was analyzed.
• Candidate predictive factors were screened using logistic regression and LASSO regression.
• The predictive performance of 8 machine learning algorithms was compared, with XGBoost selected for model development.
• A simplified risk scoring table was constructed using SHapley Additive exPlanations (SHAP) values.

Main Results:

• SIRI was identified as an independent predictive factor for PHN (OR = 1.448, P = 0.005).
• Six core predictive variables were identified: age, SIRI, numerical rating scale (NRS) score, time to treatment, rash location, and neutrophil-to-albumin ratio (NAR).
• The XGBoost model achieved high predictive performance with AUC values ranging from 0.857 to 0.900 across training, internal, and external validation sets.
• The SHAP-based risk scoring table showed an AUC of 0.904 in external validation.

Conclusions:

• SIRI is a novel and independent biomarker for predicting PHN development.
• The SIRI-based XGBoost model demonstrates excellent predictive accuracy for PHN.
• The developed simplified risk scoring table is clinically practical for early risk stratification and intervention in patients at risk of PHN.