Immune checkpoint blockade (ICB) provides clinical benefit to a subset of patients with cancer. However, existing biomarkers do not reliably predict treatment response across diverse cancer types. Limited data exist to show how serial circulating tumor DNA (ctDNA) testing may perform as a predictive biomarker in patients receiving ICB. We conducted a prospective phase II clinical trial to assess ctDNA in five distinct cohorts of patients with advanced solid tumors treated with pembrolizumab (NCT02644369). We applied bespoke ctDNA assays to 316 serial plasma samples obtained at baseline and every three cycles from 94 patients. Baseline ctDNA concentration correlated with progression-free survival, overall survival, clinical response and clinical benefit. This association became stronger when considering ctDNA kinetics during treatment. All 12 patients with ctDNA clearance during treatment were alive with median 25 months follow up. This study demonstrates the potential for broad clinical utility of ctDNA-based surveillance in patients treated with ICB.

BACKGROUND:The objective of this study was to investigate the prognostic value of the pretreatment circulating neutrophil count (CNC), circulating monocyte count (CMC), and circulating lymphocyte count (CLC) in human papillomavirus (HPV)-related (HPV1) and HPV-unrelated (HPV-) oropharyngeal cancer (OPC). METHODS: All p16-confirmed HPV1 and HPV-OPC cases treated with chemoradiotherapy from 2000 to 2010 were included. Overall survival (OS) and recurrence-free survival (RFS) were compared for high and low CNCs, CMCs, and CLCs (dichotomized by median values). A multivariate analysis (MVA) confirmed their prognostic value in HPV1 and HPV-tumors, respectively. RESULTS: Five hundred ten HPV1 OPC cases and 192 HPV-OPC cases were included. The HPV1 cohort had lower CNC and CMC values but a CLC similar to that of the HPV-patients (P <.01). The median follow-up was 4.8 years. In the HPV1 cohort, a high CNC or CMC correlated with reduced OS and RFS in comparison with a low CNC or CMC (P <.01 for all), but no difference was evident in OS (P 5.30) or RFS (P 5.10) with the CLC. MVA confirmed that a higher CNC or CMC independently predicted lower OS (hazard ratio [HR] for CNC, 1.14, P <.01; HR for CMC, 2.95, P <.01) and lower RFS (HR for CNC, 1.11, P <.01; HR for CMC, 3.39, P <.01), whereas a higher CLC was associated with higher RFS (HR, 0.66, P 5.03) and marginally higher OS (HR, 0.80, P 5.08). In the HPV-cohort, CNC, CMC, and CLC were not predictive of OS (P 5.16, P 5.86, and P 5.14) or RFS (P 5.61, P 5.59, and P 5.62). CONCLUSIONS: This relatively large cohort study demonstrates that a high CNC and a high CMC independently predict inferior OS and RFS, whereas a high CLC predicts better RFS and marginally better OS in HPV1 OPC patients. This association was not apparent in HPV-patients. Cancer 2015;121:545-55.