A model for predicting response to PD-1 inhibitors in NSCLC

点击翻译按钮获取中文摘要

PROJECT SUMMARY Lung cancer is the leading cause of cancer related deaths in the United States (US) and the world, accounting for over 150,000 deaths per year in the US alone. Recently, understanding of the biology of non-small cell lung cancer (NSCLC) has increased. Although many patients are treated with agents targeting specific driver mutations in their tumor, such agents are unavailable for most patients, and resistance eventually emerges. Agents directed against the programmed cell death-1 (PD-1) immune checkpoint have recently shown great promise. Although associated with an objective response rate (ORR) of about 20% in unselected metastatic NSCLC patients, the quality and duration of responses can be profound, particularly in a field accustomed to progression of disease after six months with even the most effective therapies A substantial debate is based on the predictive nature of biomarkers to select patients for therapy. Many were surprised by the results of a study of 495 NSCLC patients I led suggesting an association between ORR and PD-L1 expression. In a training set, we found that staining for PD-L1 in at least half of the tumor cells predicted a greater ORR. When we looked to validate our results in independent patients, the ORR was 45.2% in those with staining in at least half of their tumor cells compared to 16.5% and 10.7% in those with lesser or absent staining respectively. Similar results were seen for progression free and overall survival. Further evidence has been generated looking at other potential biomarkers. We collaborated with others to show that the number of non-synonymous mutations correlated with durable clinical benefit (partial response or stable disease lasting at least 6 months). We also saw correlations with outcome and a history of current or prior cigarette smoking, pre-biopsy CD4+ and CD8+ T cells and expression of certain genes and miRNAs. Yet, no single factor predicts outcome at the level of precision that would be ideal for clinical practice. Further, despite the correlation of each factor with clinical-outcome, the different factors don't correlate with one another particularly strongly. We have banked specimens from well over 100 patients treated with a PD-1 inhibitor to date. In light of recent drug approvals, working with our affiliated satellite offices and a network of community oncologists with whom we collaborate, the TRIO-US network, we will rapidly bank additional high quality specimens that are associated with clinical data. With these specimens, we plan to be able to create models that can effectively predict which patients stand to benefit from PD-1 inhibitors. The specific aims of this project are: 1. Define the clinical characteristics and the properties of the tumor and immune microenvironment that predict response to single agent PD-1 inhibition in a training set 2. Create models to identify the likelihood of benefit from PD-1 inhibition in NSCLC 3. Validate the models generated from the training set samples in an independent set of samples