Development of a Hypoxia-Immune Prognostic Classifier for Head-and-Neck Cancer Patients Undergoing Radiotherapy - Results From a Prospective Imaging Trial.

PURPOSE/OBJECTIVE(S) Tumor-infiltrating lymphocytes (TILs) correlate with an improved outcome of head-and-neck squamous cell cancer (HNSCC) patients undergoing radiotherapy and are influenced by the tumor microenvironment and tumor-associated hypoxia. The present analysis is based on a prospective imaging trial and analyzes the value of TILs and tumor-associated hypoxia to stratify HNSCC patients according to their response to radiotherapy. MATERIALS/METHODS A total of 49 patients with locoregionally advanced HNSCCs were prospectively enrolled in this trial and underwent longitudinal hypoxia PET imaging using fluoro-18 misonidazole ([18F]FMISO) in weeks 0, 2 and 5 during chemoradiation. Tumor hypoxia was assumed for a tumor-to-background SUV (contralateral sternocleidomastoid muscle) exceeding 1.4, and an early hypoxia response was defined as a decrease in the normalized maximum intratumoral SUV (FMISO-SUVindex) between weeks 0 and 2. Pre-treatment tumor biopsies were analyzed for TILs and the hypoxia tissue marker carbonic anhydrase IX (CAIX). Trial patients were stratified into 4 subgroups based on their TIL numbers (> 100 vs. < 100 TILs/HPF) and expression of CAIX (above vs. below median H-score), and locoregional control (LRC) and progression-free survival (PFS) rates for all subgroups were assessed using Cox and subsequent concordance analyses (Harrell's C). RESULTS High TIL levels correlated with improved LRC (HR = 0.279, P = 0.011) and PFS (HR = 0.276, P = 0.006). Similarly, a decrease in the FMISO-SUVindex within the first 2 weeks of treatment corresponded with better LRC (HR = 0.321, P = 0.015) and PFS (HR = 0.402, P = 0.043). Harrell's C was 0.68, when TILs and early hypoxia PET response were combined. The hypoxia PET-based hypoxia-immune classifier separated 3 distinct prognostic patient subgroups, a favorable (TILhigh/ early PET response), and intermediate (TILhigh/no early PET response or TILlow/early PET response) and a poor (TILlow/no early PET response) prognostic group with 2-year LRC of 71%, 33% and 0%, respectively. Low pre-treatment tissue levels of CAIX were also prognostic for improved LRC (HR = 0.352, P = 0.050) but not PFS (HR = 0,468, P = 0,087). Harrell's C was 0.66 for CAIX and TILs separately and 0.71 for the combination. The immunohistochemistry-based immune-hypoxia classifier similarly stratified between a favorable (CAIXlow/TILhigh), an intermediate (CAIXlow/TILlow or CAIXhigh/TILhigh) and a poor (CAIXhigh/TILlow) subgroup with 2-year LRC rates of 73%, 62% and 11%, respectively. CONCLUSION We developed a hypoxia PET-based hypoxia-immune classifier that was able to separate HNSCC patients into 3 distinct prognostic subgroups based on their tumor biology. These subgroups could also be stratified in a clinically feasible, immunohistochemistry-based classifier with comparable model accuracy. Therefore, this hypoxia-immune classifier could help to stratify prognoses of HNSCC patients undergoing radiotherapy pending validation in an external cohort.