An integrated machine learning-based prognostic model in head and neck cancer using the systemic inflammatory response index and correlations with patient reported financial toxicity.

Objective: To investigate the prognostic utility of systemic inflammatory response index (SIRI) as a biological readout of stress associated immune modulation in head and neck cancer patients who underwent radiation therapy. Methods: Random survival forest machine learning was used to model survival in 568 head and neck cancer patients. SIRI was calculated via pre-treatment bloodwork. Model validation was performed in an external cohort of 345 patients. Baseline financial toxicity (FT) and SIRI were studied in 638 patients. Results: Incorporation of SIRI (with performance status and smoking history) into a machine learning model identified three risk-groups that significantly stratified overall survival (p<0.0001,) and these findings were validated in the external validation cohort (p<0.001.) Increasing levels of FT were significantly associated with increasing SIRI levels. (p=0.001.). Conclusions and Relevance: An integrated machine learning model using clinical features was successfully developed and externally validated. SIRI was significantly associated with increasing FT. Our findings highlight the potential utility of SIRI as a biological marker of FT in head and neck cancer patients.