Machine Learning Incorporating Host Factors for Predicting Survival in Head and Neck Squamous Cell Carcinoma Patients

Abstract

Prognostication for cancer patients is integral for patient counseling and treatment planning, yet providing accurate prediction can be challenging using existing patient-specific clinical indicators and host factors. In this work, we evaluated common machine learning models in predicting head and neck squamous cell carcinoma (HNSCC) patients’ overall survival based on demographic, clinical features and host factors. We found random survival forest had best performance among the models evaluated, which achieved a C-index of 0.729 and AUROC of 0.792 in predicting two-year overall survival. In addition, we verified that host factors are independently predictive of HNSCC overall survival, which improved the C-index by a margin of 0.026 and the AUROC by 0.034. Due to the strong correlation among host factors, we showed that proper dimension reduction is an important step before their incorporation into the machine learning models, which provides a host factor score reflecting the patients’ nutrition and inflammation status. The score by itself showed excellent discriminating capacity with the high-risk group having a hazard ratio of 3.76 (1.93–7.32, p < 0.0001) over the low-risk group. The hazard ratios were further improved to 7.41 (3.66–14.98, p < 0.0001) by the random survival forest model after including demographic and clinical features.

Association of Body Mass Index With Outcomes Among Patients With Head and Neck Cancer Treated With Chemoradiotherapy

“…[22][23][24][25]28,29 Such discrepancies may be due to a nonlinear association between BMI and survival, with the highest survival seen in the overweight BMI range. 33,34 To our knowledge, this is the first report for head and neck cancer to show that overweight BMI and obese BMI are associated with complete metabolic response on follow-up PET-CT. Our finding is consistent with another study suggesting a higher likelihood of pathologic complete response in…”

“…32,51 Although a few other studies have suggested that a higher BMI is associated with improved survival, 41 they also included patients with an underweight BMI as a reference group, which was previously shown to be associated with worse survival outcomes. 33,34 Although a lack of association between BMI, HPV, and survival in our study may be due to smaller subgroup sample sizes, interaction among these variables warrants further investigation. For example, despite adipose tissue-promoting pathways, including PI3K-PTEN-Akt-mTOR and Ras-Raf-MAPK associated with HPV-associated head and neck cancers, 52 patients with a high BMI were more…”

Association of Pack-Years of Cigarette Smoking With Survival and Tumor Progression Among Patients Treated With Chemoradiation for Head and Neck Cancer

“… 40 , 41 Another method to improve patient selection for survival outcomes is to incorporate machine learning to account for complex interactions with other host factors. 42 Although a phase 3 trial failed to show improved locoregional control with radiation dose escalation, 43 a recent study also investigated the feasibility of adaptive dose escalation among those with poor survival outcomes. 44 Further studies would be warranted to improve risk stratification and tailor effective treatment options.…”

Exercise and Nutrition Interventions in Head and Neck Cancer

“…In a recent work, Yu et al [1] evaluated almost 600 primary HNSCC patients treated with definitive or post-operative RT. The authors showed through a machinelearning model that the main predictors of patients' overall survival were performance status, body-mass index (BMI) and the host factors reflecting the patients' nutrition and inflammation status.…”