139P - Multimodal prognosis modeling of advanced NSCLC treated with first-line immunochemotherapy: Integrating genomic and microenvironmental data

Background

Immunochemotherapy has emerged as the standard treatment for advanced non-small cell lung cancer (NSCLC). However, this combination also introduces certain challenges, including increased toxicity and the necessity for predictive biomarkers to identify patients most likely to benefit from immunochemotherapy.

Methods

In this prospective study involving 162 untreated advanced NSCLC patients at the Chinese PLA General Hospital, next-generation sequencing testing was conducted using an 1123-gene panel before initiating first-line immunochemotherapy. Lasso regression was used to identify core tumor mutation characteristics and a deep learning model based on Whole Slide Images (WSI) was employed to recognize tumor microenvironment (TME) features. Based on the information above, a prognosis model was established and optimized.

Results

Firstly, a Risk Score (RS) model was established based on tumor mutation patterns, including the RTK-RAS pathways, as well as ARID1B, ABCC2, NIPBL, DYNC2H1, SETD2, and FAF1. The high-risk score (HRS) patients exhibited significantly shorter progression-free survival (PFS) (HR=3.44; 95% CI 2.04–5.82; p<0.0001) and overall survival (OS) (HR=2.05; 95% CI 1.05–4.18; p=0.032) compared to the low-risk score (LRS) group. Secondly, the proportions of five cell types in TME were recognized by WSI model. Patients characterized by a higher proportion of epithelial cells demonstrated longer PFS (HR=0.012; 95% CI 7.49e-05–2.076; p=0.093). Finally, the comprehensive findings indicated that patients with LRS and a higher proportion of epithelial cells may significantly benefit from immunochemotherapy with longer PFS (HR=0.238; 95% CI 0.13-0.46; p<0.0001). The time-dependent ROC curve area under the curve for predicting PFS reached 0.807.

Conclusions

By integrating the RS model with genomic alterations and a deep learning-based microenvironment cell model, it is feasible to generate multimodal prognostic biomarkers that serve as guides for first-line immunochemotherapy of advanced NSCLC patients. These comprehensive biomarkers offer novel insights into personalized treatment strategies.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

The Chinese PLA General Hospital.

Funding

Beijing ChosenMed Clinical Laboratory Co, Ltd.

Disclosure

S. Liu: Other, Personal, Full or part-time Employment: Beijing ChosenMed Clinical Laboratory Co, Ltd.. All other authors have declared no conflicts of interest.