Abstract
INTRODUCTION: The incidence of pneumonitis, a treatment-related adverse event (AE) in non-small cell lung cancer (NSCLC) patients, has been studied in the United States mostly through clinical trials and retrospective chart reviews. Few analyses of real-world data have been published. This study of a large nationally representative health records database estimated the incidence and predictors of pneumonitis among treated NSCLC patients between 2008 and 2018. METHODS: The Optum(®) electronic health records (EHR) database includes data on over 80 million patients from more than 50 healthcare plans. The cohort of primary NSCLC patients was identified using ICD-9/10 codes. Natural language processing of unstructured data from physicians' notes facilitated extraction of biomarker (epidermal growth factor receptor [EGFR] and programmed death ligand-1 [PD-L1]) status. Cumulative incidence was estimated as the proportion with pneumonitis overall, by clinical characteristics, and line of therapy (LOT) after diagnosis and treatment. Univariate analysis of incidence rates (cases/1000 person-years) enabled the identification of significant predictors of risk. Competing risk regression identified predictors of pneumonitis. RESULTS: The cohort included 81,628 patients. Overall, 19.0% developed pneumonitis during any LOT, with a cumulative incidence of 33.7% and 17.0% for patients with a prior history of pneumonitis and those without, respectively. Univariate analyses revealed several factors associated with pneumonitis (p < 0.05). While factors varied between LOTs, common factors included male gender, squamous histology, history of diabetes or pneumonitis, EGFR-negative status, monotherapy immunomodulatory drugs, or history of radiation therapy. Multivariable competing risk regression showed that male gender, history of pneumonitis, EGFR-negative status, use of other targeted therapies, use of immunomodulatory drugs, and history of radiation therapy predicted pneumonitis. CONCLUSION: Pneumonitis is significantly associated with NSCLC treatment. Knowledge of its predictors identified in this study may help devise strategies to mitigate its impact, enhancing treatment adherence and improving outcomes.