Main Session
Sep
30
PQA 09 - Hematologic Malignancies, Health Services Research, Digital Health Innovation and Informatics
3656 - Identifying Genomic Biomarkers of Immunotherapy Response in Advanced NSCLC Using an NLP-Derived Real-World Database
Presenter(s)
Victor Lee, MD, BS - Yale New Haven Hospital, Middletown, CT
V. Lee1, R. Ravella2, D. G. Miller2, E. Yang3, J. McFadden3, D. Billing2, N. Shaverdian2, D. R. Gomez2, and T. L. Chaunzwa2; 1Department of Therapeutic Radiology, Yale School of Medicine, New Haven, CT, 2Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, NY, 3Yale School of Medicine, New Haven, CT
Purpose/Objective(s):
Immune checkpoint inhibitors (ICIs) have transformed non-small cell lung cancer (NSCLC) management. However, response to ICIs remains highly variable, and reliable tumor-specific predictors beyond PD-L1 expression and tumor mutational burden (TMB) are limited. Advances in real-world data (RWD) integration using natural language processing (NLP) have enabled the extraction of meaningful insights from large oncology databases. The MSK-CHORD database, which harmonizes structured and unstructured data from electronic health records, provides an opportunity to explore associations between tumor mutations and overall survival (OS) in NSCLC patients treated with ICIs. By integrating NLP-derived annotations with genomic and clinical data, we aimed to uncover novel biomarkers that could refine patient selection and improve treatment outcomes in the real-world setting.Materials/Methods:
We conducted a retrospective analysis using the MSK CHORD database. Our study included patients with stage IV NSCLC who received at least one ICI (pembrolizumab, atezolizumab, durvalumab, or nivolumab). OS was calculated from the time of first immunotherapy initiation. Genomic data were extracted to identify pathogenic mutations, and Cox proportional hazards models were used to assess the association between individual mutations and OS.Results:
A total of 636 patients were included in the analysis. The median age was 69.1 years. Of these, 518 (81.4%) were current or former smokers, 458 (72.0%) had adenocarcinoma, and 83 (13.1%) had squamous cell carcinoma. Mutations that were associated with worse OS included PTEN (Hazard ratio [HR]: 1.83, 95% Confidence Interval [CI]: 1.16-2.90, p = 0.010); STK11 (HR: 1.64, CI: 1.30-2.05, p < 0.001); KEAP1 (HR: 1.50, CI: 1.21-1.87, p < 0.001); and KMT2D (HR: 1.45, CI: 1.06-1.99, p = 0.022). Mutations that were associated with favorable prognosis included POLE (HR: 0.51, 0.30-0.86, p = 0.013) and NTRK1 (HR: 0.48, CI: 0.26-0.87, p = 0.016). These genomic biomarkers outperformed conventional immunotherapy biomarkers, including PD-L1 expression (HR: 0.69, CI: 0.57-0.83, p < 0.001) and TMB of at least 10 (HR: 0.79, CI: 0.64-0.96, p = 0.019) when predicting OS.Conclusion:
Our findings reveal that specific tumor mutations in NSCLC are strongly associated with survival in ICI-treated patients. PTEN and STK11 mutations, which promote immune evasion via PI3K/AKT activation and STING suppression, were linked to worse outcomes, along with KEAP1 and KMT2D, which regulate oxidative stress and chromatin remodeling. Conversely, POLE, which drives hypermutation, and NTRK1, which enhances tumor immunogenicity, were associated with improved survival. These results highlight the potential of NLP-integrated databases to refine patient selection and optimize immunotherapy.