Objective:

Despite growing interest in neoadjuvant therapies, there are no methods to predict radio- (RT) or chemoradiotherapy (CRT) response in head and neck squamous cell carcinoma (HNSCC). The aim of this research was to study the effect of neoadjuvant RT or CRT on the tumor immune landscape and patient survival in HNSCC.

Methods:

All HNSCC patients treated with neoadjuvant RT or CRT (n = 53) were identified from a retrospective cohort of 1033 patients. Pre- and post-neoadjuvant cancer samples from the same patient were analyzed with biomarkers related to cancer immunology: tumor-infiltrating lymphocytes (CD8), tumor-associated macrophages (CD68, CD206, Clever-1), immune response regulator (PD-L1) and histologic tumor necrosis. Outcomes of interest were individual immune landscape profiling and its impact on 5-year overall survival (OS) in HNSCC patients treated with neoadjuvant RT/CRT.

Results:

Results from 588 whole-section stainings revealed multiple statistically significant alterations in immune landscape in response to RT/CRT. Pretreatment tumor necrosis was the most useful biomarker in predicting poor outcome, as the OS was 14.3% with necrosis and 48.5% without necrosis (HR 2.87; 95% CI: 1.23 to 6.66, p=0.014). In addition, an artificial intelligence-based (AI) deep learning method for identifying tumor necrosis from histopathological specimens was successfully developed. The predictive role of histological necrosis in neoadjuvant RT/CRT was validated in additional samples from 171 HNSCC patients untreated with neoadjuvant therapy.

Conclusions:

Detection of tumor necrosis and AI-driven deep learning effectively predict neoadjuvant RT/CRT responses in HNSCC.