(052) Clinical Responses to Immune Checkpoint Inhibitors in Non-Small Cell Lung Cancer (NSCLC): Patient Characteristics and Predictors

Background: Lung cancer is the leading cause of cancer-related mortality and more than 85% of cases are non-small cell lung cancer (NSCLC). While immune checkpoint inhibitors (ICI) have significantly improved prognosis of NSCLC, not all NSCLC patients respond to ICIs. However, the response rate and predictors of response have not been well described.

Objectives: To define clinical response to ICI, estimate the response rate, and identify clinical predictors of response to ICI in NSCLC patients using electronic medical record (EMR) and cancer registry data from a large cancer center.

Methods: We identified NSCLC patients diagnosed between 01/2011-06/2021 and initiated on ICIs who had next-generation sequencing (NGS) reports available. Demographic, behavioral, and clinical characteristics were abstracted from EMR and EMR-linked cancer registry data. Responses to ICI were defined based on physician’s notes and CT-scan reports as the first documented response after the second cycle of ICI therapy, patients were characterized as responders and non-responders. We recorded reason for discontinuation and tracked the occurrence of adverse events and survival. A multivariate logistic regression (LR) was used to identify predictors of ICI response.

Results: Among 406 NSCLC patients with NGS reports, 53 initiated ICIs and 50 (94%) had documented response to ICI (3 patients were lost to follow up). Among the 50 patients (median age 64, 38% male, median BMI 26 (IQR 23-29), 60% BMI ≥ 25kg/m2, 72% White, 42% never smokers, and 68% metastasis), 30 (60%) responded to ICI. Compared to non-responders, responders were less likely to be never-smokers and symptomatic and to have comorbidities and metastasis to bone and liver. Six patients (12%) discontinued ICIs due to side effects (immune related rash, bleeding, pleural or pericardial effusions, pneumonia, pneumonitis, mucositis). In multivariate LR, patients presenting with weight loss were less likely to respond to ICI (Odds ratio 0.2). Documentation of response in patient charts varied from stable disease, clinical improvement, complete/partial response or continue therapy to document response and no response or progressive disease to document no-response to ICI and needed to be abstracted by manual review.

Conclusions: More than 60% of NSCLC patients responded to ICIs and 12% discontinued ICIs due to side effects. Weight loss is a strong predictor of response to ICI. Information on the ICI response or its predictors were mostly not available in large administrative data, registry, or coded data in EMR. Our results will guide the development of natural language processing (NLP) methods to abstract necessary information in a large multi-center study using institutional EMR data.