Extended Data Fig. 1: Characterization of cohorts and association of TMB, PD-L1, and CD8 infiltrate with outcomes.

a. Comparable PFS outcomes are seen in the biomarker analysis cohorts relative to the full dataset. Cox proportional hazards model used to estimate hazard ratio (and 95% CI) between placebo and avelumab in subgroups with genomic data. b. Tumor mutation burden (TMB) is higher in HPV negative (median=5.37 Mut/Mb, IQR=3.95-7.23, N=272) vs. HPV positive (median=4.26 Mut/Mb, IQR=2.53-6.53, N=149) (p < 0.001; Wilcoxon-rank sum test). c. Associations between PFS, treatment, and classical biomarkers were comparable in HPV-positive and HPV-negative patients (Cox proportional hazards model and 95% CI). d. Association between PFS in placebo/CRT arm and expression of PD-L1 by tumor cells (left) vs immune cells (right). PD-L1 high vs low populations (log-rank test).