26 Validation of the Primary Immune Response (PIR) test in advanced non-small cell lung cancer (NSCLC): blinded retrospective analyses from the POPLAR and OAK trials

Abstract

Background Biomarkers of immune checkpoint inhibitor (ICI) efficacy can be used for patient selection. PD-L1 expression in tumor tissue is used to determine eligibility for combination or monotherapy in 1st line NSCLC.1, 2 The liquid-biopsy mass spectrometry-based PIR test was developed to capture the role of patient biology on ICI outcomes.3 The test, stratifying patients into Resistant, Intermediate, and Sensitive groups, was associated with outcome on nivolumab treatment in 2nd line NSCLC patients.3 In this study, we blind validated PIR classifications in two large clinical studies (POPLAR4 and OAK5) of advanced NSCLC patients treated in the second or third line with atezolizumab. Methods Pretreatment serum samples from patients assigned to receive atezolizumab in the two studies (POPLAR (NCT01903993) and OAK (NCT02008227)) underwent PIR testing blinded to all clinical data. Association of test classification, as Sensitive vs Not Sensitive (Resistant+Intermediate) and Resistant vs Not Resistant (Sensitive+Intermediate), with overall survival (OS) and progression-free survival (PFS) was investigated using Cox proportion hazards models in univariate and multivariate analysis. Results PIR classifications were generated for 133 (POPLAR) and 403 (OAK) samples; the remaining available samples failed test QC, mainly due to hemolysis. PIR classified the POPLAR samples as 53 (40%) Resistant, 25 (19%) Intermediate, 55 (41%) Sensitive and the OAK samples as 154 (38%) Resistant, 89 (22%) Intermediate, and 160 (40%) Sensitive. In both cohorts, OS and PFS were better in the Not Resistant vs Resistant group (figure 1). OS and PFS were superior in the Sensitive vs Not Sensitive group in the POPLAR cohort, while OS was better and PFS showed indications of superiority in the OAK cohort (figure 2). Multivariate analysis within the OAK cohort showed that test classification predicted OS when adjusted for baseline factors, including PD-L1 negative vs positive, with hazard ratio 0.51 (95% confidence interval (CI) 0.40-0.65) for Resistant vs Not Resistant and 0.65 (CI: 0.50- 0.83) for Sensitive vs Not Sensitive. Conclusions The PIR test stratified outcomes for patients treated with atezolizumab in second and third line NSCLC even when adjusted for PD-L1 expression. The combination of both tumor and host biomarkers appears to provide a more specific prognosis of NSCLC treated with ICIs. (Figure Presented).