Cumulated clinical experience with checkpoint inhibitors (CPIs) points to a strong need for the identification of predictive biomarkers. Surprisingly, the potential role of the host has not been advocated so far. We developed a custom designed panel of single nucleotide polymorphisms (SNPs) from genes potentially implicated in the response to CPIs.
We studied 94 patients treated in Centre Antoine Lacassagne (Nice, France) with CPI (anti PD-1/PD-L1). High-throughput genotyping of germinal DNA was performed by MassARRAY ImmunoCarta (AGENA Bioscience®) using a custom-panel of 173 SNPs across 90 selected genes (minor allelic frequency ≥5% in the Caucasian population). All tested SNPs were in Hardy-Weinberg equilibrium, and linkage disequilibrium analyses were performed (r2>0.8). A Ridge-penalized logistic regression with 5-fold cross validation was used for the SNP identification to predict objective response (complete or partial response) to treatment.
Median age of patients was 68 (range: 32-85), 67% were male, 51% had non-small cell lung carcinoma (NSCLC), 14% had head and neck squamous cell carcinoma (HNSCC), 15% had renal cell carcinoma (RCC), 13% had melanoma and 7% had another cancer type. Median follow-up was 16.3 months (95%CI: 12.5-18.3). The following SNPs’ decreasing intrinsic weight according to response were selected by the multivariate modeling approach (CCR2: rs1799864, FAS: rs2234767, CD3G: rs3753058, CTLA4: rs5742909, CCL2: rs13900, TNXB: rs12153855, Il1RN: rs419598, PD1: rs11568821, IL17A: rs2275913, IL12B: rs3212227, TLR3: rs7668666, CXCR3: rs2280964, IL10: rs1800871, IL6: rs2069837, TRAF3: rs7145509, VEGFR3: rs307821). In the training set, the accuracy was 0.87 (95%CI: 0.76-0.95; p < 0.001) associated with a sensitivity and a specificity of 0.90 and 0.83, respectively, with a ROC curve AUC at 0.93 (95%CI: 0.87-0.99). In the validation set, the accuracy decreased to 0.71 (95%CI: 0.52-0.85; p < 0.02) associated with a sensitivity and a specificity of 0.82 and 0.60, respectively, and with a ROC curve AUC of 0.85 (95%CI: 0.72-0.98).
These preliminary results point to the feasibility of a signature based on host characteristics for predicting response to CPI.
Centre Antoine Lacassagne, Nice, France.
Has not received any funding.
G. Milano: Paid scientific consultant: Agena Bioscience. S. Shell, R. Everts: Employee of Agena Bioscience. All other authors have declared no conflicts of interest.