Developing a Predictive Model for Clinical Outcomes of Advanced Non-Small Cell Lung Cancer Patients Treated With Nivolumab

A single biomarker cannot account for the heterogeneous tumor biology and immune interplay in patients with advanced non–small cell lung cancer who receive programmed death inhibitor. This article reports our initial model development that incorporates differential weightings of clinical and hematologic variables for a future algorithm. The immunotherapy, Sex, Eastern Cooperative Oncology Group performance status, Neutrophil-to-lymphocyte ratio (NLR), and Delta NLR are incorporated into the model that categorizes patients into different risk groups and significantly discriminates each group's clinical outcome. Introduction: Despite significant improvement of clinical outcomes of advanced non–small-cell lung cancer (NSCLC) patients treated with immunotherapy, our knowledge of optimal biomarkers is still limited. Patients and Methods: We retrospectively evaluated 159 advanced NSCLC patients in our institution treated with nivolumab after disease progression during platinum-based chemotherapy. We correlated several variables with progression-free survival (PFS) to develop the immunotherapy, Sex, Eastern Cooperative Oncology Group performance status, Neutrophil-to-lymphocyte ratio (NLR), and Delta NLR (iSEND) model. We categorized patients into iSEND good, intermediate, and poor risk groups and evaluated their clinical outcomes. Performance of iSEND at 3, 6, 9, and 12 months was evaluated according to receiver operating characteristic (ROC) curves and internally validated using bootstrapping. The association of iSEND risk groups with clinical benefit was evaluated using logistic regression. Results: Median follow-up was 11.5 months (95% confidence interval [CI], 9.4-13.1). There were 50 deaths and 43 with disease progression without death. PFS rates at 3, 6, 9, and 12 months were 78.4%, 63.7%, 55.3%, and 52.2% in iSEND good; 79.4%, 44.3%, 25.9%, and 19.2% in iSEND intermediate; and 65%, 25.9%, 22.8%, and 17.8% in iSEND poor. Time-dependent area under ROC curves of iSEND for PFS at 3, 6, 9, and 12 months were 0.718, 0.74, 0.746, and 0.774. The iSEND poor group was significantly associated with progressive disease at 12 ± 2 weeks (odds ratio, 9.59; 95% CI, 3.8-26.9; P <.0001). Conclusion: The iSEND model is an algorithmic model that can characterize clinical outcomes of advanced NSCLC patients receiving nivolumab into good, intermediate, or poor risk groups and might be useful as a predictive model if validated independently.