Predictors of early immunotherapy response in head and neck cancer: Per lesion analysis of a prospective randomized trial with nivolumab.

Abstract

6041 Background: The capacity for radiation and checkpoint inhibitors to elicit clinical responses is impacted by tumor immunogenicity and the immune microenvironment. We sought to determine whether head and neck primary site and metastatic tumor location was associated with initial response in non-irradiated lesions. Methods: We evaluated response in 144 non-irradiated lesions from 59 patients with metastatic head and neck cancer enrolled on a phase II randomized controlled trial of nivolumab with stereotactic body radiotherapy (n=30) vs. nivolumab alone (n=29). Nivolumab was administered 3 mg/kg intravenously every 2 weeks. Radiated lesions were treated with 27 Gy / 3 fractions to a single lesion within 14 days of the first dose of nivolumab. Non-target lesion progression was defined ≥30% increase in the greatest axial diameter 8 weeks after enrollment. Fisher’s exact test with nested bootstrap resampling was used for univariate analysis. Logistic regression with a mixed random effects term was used for multivariate analysis. Results: Primary tumor site, metastatic tumor organ sites, and the unadjusted likelihood of progressive disease by site are listed in the table. On multivariate logistic regression controlling for PD-L1 status (p=0.66) and viral status (p=0.29), lymph node metastases (OR 0.79, p=0.0064) were associated with decreased risk of progression, while liver metastases (OR 1.39, p=0.014) and oral cavity primaries (OR 1.56, p=0.018) were associated with increased risk of progression at 8 weeks, using lung metastases and larynx/hypopharynx primaries as reference. Conclusions: Primary tumor subsite and metastasis location were predictors of response or stable disease following treatment with nivolumab. Metastases from oral cavity primaries and metastases to the liver were at increased risk of early progression. Clinical trial information: NCT02684253. [Table: see text]