Immunologic constant of rejection as a predictive biomarker of immune checkpoint inhibitors efficacy in non-small cell lung cancer.

e21162 Background: Immune checkpoint inhibitors (ICI) transformed non-small cell lung cancer (NSCLC) prognosis but response rate remains disappointing. Besides PDL1 expression, tumor mutational burden, quantification of tumor infiltrating lymphocytes or methylation profile failed to confirm their predictive role. The ICR (Immunologic Constant of Rejection) signature is defined as an immune phenotype quantifying the expression of 20 genes all involved in cancer immunity. Our first objective was the assessment of ICR signature predictive value and our secondary objectives were to compare ICR signature with other published immune signatures and to assess the prognostic value of ICR signature. Methods: Our series included 77 paraffin-embedded (FFPE) tumors derived from NSCLC patients treated with ICI as a single agent in at least the second line setting. We also collected clinical and biological data from 4 public data sets to expand and confront our local data. Transcriptomic analysis was performed using nCountertechnology Dx analysis by Nanostring based on microscopic imaging counting relative abundance of hundreds of transcripts using the “nCounter PanCancer Immune Profiling Panel” including the 20 ICR genes. Results: We applied the ICR classification to a series of 44 local tumor samples enhanced with 118 samples from public datasets. Hierarchical clustering on both local cohort and public datasets allowed us to distinguish the four ICR classes, from ICR 1 (“cold” tumors) to ICR 4 (“hot” tumors). Among the 5 datasets, we did not observe any significant difference between ICR groups for age, sex and smoking status. We observed a relative homogeneity in Durable Clinical Benefit (DCB) rate among ICR groups 2 to 4 (median DCB above 20%) unlike ICR group 1 (median DCB under 10%) leading us to consider ICR as a binary signature (ICR 1 vs ICR 2-4). ICR signature indeed showed a significant association with DCB among the 5 pooled datasets (OR = 4.40 [95% Confidence Interval 1.44-13.45] p = 0.950) with uniform distribution between them. Then, we conducted a logistic regression univariate analysis of DCB among various previously published immune signatures and ICR signature was the only one to show a significant association with DCB in this analysis ( p = 0.007). Conclusions: We identified a feasible transcriptomic signature with a promising signal of ICI efficacy prediction in the so far disappointing immune biomarkers field. In a near future, predict ICI efficacy will be a major challenge with three pivotal questions. When, balanced with the new data on ICI use in early-stage lung cancer or widely pretreated patients. How, assessing various techniques to develop composite signatures to better consider every driver of immune response. What, considering the amount of potential new therapeutic targets, starting by the 20 genes of ICR signature.