A novel radiomic based imaging tool to monitor tumor lymphocyte infiltration and outcome of patients treated by anti-PD-1/PD-L1

Abstract

Background: Tumor infiltrating lymphocytes (TIL) appears necessary to trigger anticancer activity of anti-PD-1/PD-L1. Radiomics consists in the analysis of quantitative data extracted from standard medical imaging to generate imaging biomarkers. We developed a radiomics-based predictor of TIL and investigated whether such signature could predict the outcome of patients treated by anti-PD1/PD-L1. Methods: We first developed a predictive model of tumor infiltrating CD8 T cells with RNA-Seq and raw imaging data (CT -Scan) using random forest in 69 HNSCC patients from the TCGA (The Cancer Genome Atlas)/TCIA (The Cancer Imaging Archive) database. CD8 T cells were estimated by the Microenvironment Cell Populationscounter signature. To validate our tool, this signature was applied to a first independent cohort of 100 patients for which the pathologic TIL was assumed as either high (lymphoma, melanoma, lung, bladder, renal and MSI+ cancers; 30 patients) or low (adenoid cystic carcinoma, low-grade neuroendocrine tumors, uterine leiomyoma; 70 patients). Finally, we applied our signature on a second cohort of 139 patients prospectively enrolled in anti-PD-1/PD-L1 phase 1 trials to infer its relation with patient outcome (Overall Survival). Results: We developed a CD8 radiomics-based signature with six out of the 80 extracted features from CT-scans. As an internal validation, the correlation of this signature with the estimated TCGA CD8 was: spearman's rho=0.81 (P<1e-5). In the first external cohort, this signature was associated with the assumed lymphocyteness (Wilcoxon test, P<0.001). When validating our signature in the second external cohort, the median of the CD8 signature predicted score was used to separate patients into two groups. Patients with high predicted CD8 score had significantly better OS (HR=0.55, 95%CI=0.36-0.86, P=0.009) and the CD8 signature remained significant after multivariate analysis including RMH score and the number of previous lines of treatment (HR=0.48, 95%CI=0.31-0.76, P=0.002). Conclusions: The radiomics-based signature of TIL was validated in two external cohorts. It appears a promising tool to estimate TIL and to infer the outcome of metastatic patients treated with anti-PD1/PD-L1.