No abstract

No abstract

Different types of therapy are currently being used to treat non-small cell lung cancer (NSCLC) depending on the stage of tumor and the presence of potentially druggable mutations. However, few biomarkers are available to guide clinicians in selecting the most effective therapy for all patients with various genetic backgrounds. To examine whether patients’ mutation profiles are associated with the response to a specific treatment, we collected comprehensive clinical characteristics and sequencing data from 524 patients with stage III and IV NSCLC treated at Atrium Health Wake Forest Baptist. Overall survival based Cox-proportional hazard regression models were applied to identify mutations that were “beneficial” (HR < 1) or “detrimental” (HR > 1) for patients treated with chemotherapy (chemo), immune checkpoint inhibitor (ICI) and chemo+ICI combination therapy (Chemo+ICI) followed by the generation of mutation composite scores (MCS) for each treatment. We also found that MCS is highly treatment specific that MCS derived from one treatment group failed to predict the response in others. Receiver operating characteristics (ROC) analyses showed a superior predictive power of MCS compared to TMB and PD-L1 status for immune therapy-treated patients. Mutation interaction analysis also identified novel co-occurring and mutually exclusive mutations in each treatment group. Our work highlights how patients’ sequencing data facilitates the clinical selection of optimized treatment strategies.

Introduction: Currently, several types of treatment can be used to treat non-small cell lung cancer (NSCLC) depending on a potential druggable mutation or stage of cancer. However, a limited number of biomarkers are available to guide clinicians in selecting the most effective therapy for all patients. Methods: The clinical characteristics of 642 NSCLC patients and tumor sequencing data were collected retrospectively at Atrium Health Wake Forest Baptist. Cox-proportional hazard regression models were fit to identify mutations that were “beneficial” (hazard ratio &lt; 1) or “detrimental” (hazard ratio &gt; 1) for patients on different treatment regimens, followed by the generation of mutation composite scores (MCS) for each treatment. The overall survival (OS) of patients receiving each treatment was plotted based on the patients’ MCS, and receiver operating characteristics (ROC) curves tested the predictive power of each MCS for each treatment group. We also identified novel co-occurring and mutually exclusive mutations in each treatment group by mutation interaction analysis. Results: We identified treatment-specific mutations associated with either a better or worse OS. The MCS generated for each treatment group significantly enhanced the prediction power compared to a single mutation with limited application in patients with rare mutations. Mutation signatures to chemotherapy (NTRK1, FBXW7, BRAF, MPL, KRAS, and GATA3) and immunotherapy (MAP2K1, EGFR, CDK4, NTRK1, and NOTCH1) have a comparable prediction power with actual clinical response. Conclusions: NSCLC patients’ responses to specific treatments are diverse because of tumor heterogeneity. Our work demonstrates how analyzing patients’ sequencing data facilitates the clinical selection of optimized treatment strategies. Citation Format: Margaret R. Smith, Yuezhu Wang, Ralph D' Agostino, Yin Liu, Jimmy Ruiz, Thomas Lycan, George Oliver, Umit Topaloglu, Jireh Pinkney, Mohammed N. Abdulhaleem, Michael D. Chan, Michael Farris, Jing Su, Fei Xing. Treatment prognostic signature of patients with non-small cell lung cancer: a retrospective single-institutional study [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 975.

Background Leptomeningeal failure (LMF) represents a devastating progression of disease following resection of brain metastases (BrM). We sought to identify a biomarker at time of BrM resection that predicts for LMF using mass spectrometry-based proteomic analysis of resected BrM and to translate this finding with histochemical assays. Methods We retrospectively reviewed 39 patients with proteomic data available from resected BrM. We performed an unsupervised analysis with false discovery rate adjustment (FDR) to compare proteomic signature of BrM from patients that developed LMF versus those that did not. Based on proteomic analysis, we applied trichrome stain to a total of 55 patients who specifically underwent resection and adjuvant radiosurgery. We used competing risks regression to assess predictors of LMF. Results Of 39 patients with proteomic data, FDR revealed type I collagen-alpha-1 (COL1A1, p=0.045) was associated with LMF. The degree of trichrome stain in each block correlated with COL1A1 expression (β=1.849, p=0.001). In a cohort of 55 patients, a higher degree of trichrome staining was associated with an increased hazard of LMF in resected BrM (Hazard Ratio 1.58, 95% CI 1.11-2.26, P=0.01). Conclusion The degree of trichrome staining correlated with COL1A1 and portended a higher risk of LMF in patients with resected brain metastases treated with adjuvant radiosurgery. Collagen deposition and degree of fibrosis may be able to serve as a biomarker for LMF.