Associations between interaction effects of mutational events and efficient immunotherapy in squamous NSCLC: Results from 373 patients.

Abstract

e21136 Background: Predictive biomarkers might not be similarly predictive in every single subpopulation, and they might be merely valid under certain circumstances, which is termed by “interaction effect”. Here, we proposed to investigate the interaction effect of mutational events on predicting efficacy of immune checkpoint inhibitors (ICIs), and to explore whether adding these interaction terms could optimize the prediction model based on genomic alterations. Methods: In total, 373 patients with squamous NSCLC were included. 191 patients with both PFS and OS data receiving monotherapy from the MSKCC, DFCI, SNCC, and POPLAR/OAK cohorts were included in the training set. 130 patients from the POPLAR/OAK cohort treated with docetaxel were included in the control set. Validation set-1 consists of 34 patients receiving monotherapy from the NCC and MSKCC cohorts, and validation set-2 comprises 18 patients treated with combination therapy from the MSKCC cohort. Risk models using single mutations (uni-model) and both single mutations and interactions (inter-model) were developed via multivariable Cox proportional hazards regression model. Results: Four single mutational events, including mutations of NOTCH1/2/3, LRP1B, RB1, and members in PI3K pathway, and two interactions ( TP53* NFE2L2 and TP53*HRR pathway) were identified as predictive biomarkers. The score of inter-model exhibited higher correlation with immunotherapeutic PFS in the training sets, compared to the uni-model ( Dxy: 0.196 vs. 0.175). Uni-score low was barely associated with the PFS (HR=0.32, 95% CI 0.07-1.50, P=0.15) and OS (HR=1.20, 95% CI 0.45-3.17, P=0.71) in the validation set-1, and the PFS in the validation set-2 (HR=0.96, 95% CI 0.31-2.95, P=0.94). Inter-score low showed better association with longer PFS (HR=0.17, 95% CI 0.03-0.84, P=0.029) and OS (HR=0.33, 95% CI 0.12-0.92, P=0.033) in the validation set-1, and longer PFS in the validation set-2 (HR=0.50, 95% CI 0.13-1.85, P=0.30). In addition, no association was observed between inter-score low and the PFS and OS in the control set, indicating that inter-score low might not be a prognostic biomarker, but a positive predictive biomarker to efficacious immunotherapy. Conclusions: Two interaction effects of mutational events on predicting efficacious immunotherapy in squamous NSCLC ( TP53* NFE2L2 and TP53*HRR pathway) were identified. Adding these interaction terms to the prediction model could optimize the predictive utility. Deeper understanding of these inter-relations might provide insights for basic research on anti-tumor immunity.