On-Treatment Hematological Markers are Associated with Increased Hospitalization and Emergency Department (ED) Visits in Head and Neck Squamous Cell Carcinomas (HNSCC)

Abstract

<h3>Purpose/Objective(s)</h3> As healthcare costs continue to rise, there are growing needs to accelerate the development of value-based care in HNSCC patients treated with chemoradiation (CRT). Literature has reported that up to 35% HNSCC patients are hospitalized secondary to treatment-related toxicity during or shortly after CRT. We hypothesize that on-treatment immune and inflammatory markers (e.g., absolute lymphocyte count ALC, neutrophil-to-lymphocyte ratio NLR & C-reactive protein CRP) reflect treatment injury and can identify patients at higher risks for hospitalization/ED visits and increased care costs within the first 3 months of treatment start. <h3>Materials/Methods</h3> Eligibility criteria included newly diagnosed M0 HNSCC patients treated with neck radiation ± systemic therapy in 2015-19 with baseline ALC and ≥ 1 on-treatment ALC data point. We collected weekly laboratory data at baseline and during the first 8 weeks of treatment. Data was computed for 6 biomarkers: nadir ALC, max NLR, max CRP, max ALC decrease, max NLR increase & max CRP increase. We randomly allocated the sample into training & validation datasets (70/30% split with 222/94 patients). Missing values were omitted in our analysis (0% missing for ALC, <1% for NLR & 78% for CRP). Categorical data was collected for any documented hospitalization/ED visits within 3 months since treatment start ("yes" or "no"). In the training dataset, we computed the optimal threshold – defined as maximum sensitivity (Se) with ≥80% specificity (Sp) - for each of the 6 biomarkers. In the independent testing dataset, we used established thresholds to categorize patients into low and high-risk groups (LR, HR) and conducted Se/Sp analyses. Unpaired t-test was performed for ALC & NLR metrics with normal approximation of a binomial distribution assuming unequal variance. The sample size for CRP metrics was too small to assume normal distribution for t-test analysis. <h3>Results</h3> 316 patients met our eligibility criteria. In the testing dataset, we observed a consistent 8-18% higher rate of hospitalization or ED visits in high-risk patients (all p-values <0.0001*) (Table). Detection Se was 28-63% and Sp 80-100%. Of note, PSA has a similar prediction performance in diagnosing prostate cancer at ≥ 4 ng/mL with Se 21% & Sp 91%. Interestingly, CRP metrics have the best performance with Se 42-63% & Sp 94-100%, but we need to interpret this finding with caution given the small sample size. <h3>Conclusion</h3> On-treatment hematological & immune markers identify patients at risk for hospitalization & ED visit during and short after CRT. In future directions, strategies directed towards limiting the hematologic consequences of CRT may offer higher value care for HNSCC patients.