Developing Genomic Predictive Biomarkers for Survival Benefit from Adjuvant Chemotherapy in Early-Stage Lung Cancer Patients

Abstract

Surgical resection only remains the standard choice for the treatment of early-stage non-small cell lung cancer (NSCLC) patients. Preliminary studies suggest that the application of adjuvant chemotherapy with surgery (ACT) is associated with a better prognosis for more severe NSCLC patients compared to those who only underwent surgical resection. However, at an individual level, not all patients may benefit from ACT. Given the well-known adverse effects and toxicity of ACT, finding the patients that are most likely to benefit from ACT is paramount. Thus, the purpose of this research is to utilize gene expression and clinical data from lung cancer patients to develop a statistical decision support algorithm to find predictive genomic biomarkers and identify subgroups of patients who benefit from ACT. Cox regression models are trained using a randomized controlled trial gene expression data from the National Cancer Institute utilizing explicit treatment interaction terms. To handle high dimensions inherent in gene expression data, a modified-covariate regularized Cox regression model with lasso penalty is applied to find the most significant interacting covariates. Then risk scores are constructed from these models and are used to stratify patients into a high risk or low risk group respective to ACT treatment. After applying the model to an independent validation genomic data set, our methods show that patients who underwent the treatment according to their risk group exhibit a slightly higher survival rate than those who do not.