Summary
The introduction of immunotherapy (IO) has transformed cancer treatment, particularly for non-small cell lung cancer (NSCLC), yet molecular mechanisms of IO resistance remain elusive due to incomplete clinical and omic data. Notably, while some patients, “long-term responders” (LTR), experience lasting benefits, ∼60% of initial responders develop “acquired resistance” (AR), differing from those with primary resistance (PR) who exhibit no initial response.
Our study leverages real-world data (RWD) for its comprehensive clinical data, capturing therapy lines and responses with extensive follow-up. We examined RWD from ConcertAI PT360® EHRs linked to Caris Life Science molecular data across 2, 176 NSCLC patients, focusing on a filtered cohort (N = 407) to explore IO response and resistance mechanisms. Within this cohort, 43% received IO monotherapy and 57% combination therapy with chemotherapy, predominantly using pembrolizumab. Most patients initiated IO in the first line (90%) or second line (8.2%).
Responders to IO noted improved overall survival (p = 0.0037, median OS 27.6 vs. 12.6 months for non-responders) and longer treatment durations (p = 4.04E-04, median 210 vs. 99 days), compared to non-responders. Biomarker analysis highlighted significantly higher tumor mutational burden (TMB) and immunologic constant of rejection (ICR), along with more PD-L1 positive patients (p = 0.03) among responders, indicating increased immune infiltration. The RNA-seq data revealed 604 significantly differential genes; responders showed higher levels of immune biomarkers (e.g., CD274, CXCL-9, GZMK, TNFRSF9), whereas non-responders exhibited increased immunosuppressive markers such as NNMT and some antigen targets. Gene Set Enrichment Analysis (GSEA) revealed enriched pathways in lymphocyte activation and extracellular structure organization, TGF-b signaling, etc.
Integrative analyses of LTR, AR, and PR groups identified mutations, notably ARID1A, PREX2, and PRKN, enriched in the LTR group, potentially rendering tumor sensitivity to IO and potential new drug targets. Additionally, six distinct molecular patterns were identified, highlighting differences between LTR, AR, and PR groups. Notably, Immunoglobulin G, B cells, and plasma cell activity scores varied significantly between AR and PR, with AR mirroring LTR, implicating these immune cells in initial IO response, potentially linked to immature tertiary lymphoid structures (TLS) before transitioning to AR, manifested with TLS scores differing significantly across groups (p = 1.7E-04).
In summary, our study uses multi-modal RWD to establish an analytical workflow, revealing key clinical features and biomarkers while offer insights into molecular underpinnings of IO long-term response and resistance through multi-omic RWD integration.