Pretreatment Risk Model for Radiation-Induced Lymphopenia Is Associated With Adjuvant Durvalumab Efficacy in Patients With Unresectable Stage III NSCLC

PURPOSE: Severe radiation-induced lymphopenia (RIL) during concurrent chemoradiation therapy (CCRT) for NSCLC has been associated with poorer outcomes and reduced immunotherapy efficacy. Because RIL often develops late during CCRT, identifying patients at risk before treatment may be clinically relevant. This study aimed to develop and validate a nomogram based on pretreatment predictors for severe RIL, and secondarily to explore associations between predicted RIL risk and adjuvant durvalumab-associated survival. METHODS AND MATERIALS: A retrospective development cohort of consecutive patients with NSCLC treated with CCRT (2010-2019) was established, along with an independent external validation cohort from other institutions. A multivariable logistic regression model was developed to predict severe RIL, with internal and external validation. Survival analyses (progression-free and overall survival) were performed as exploratory, hypothesis-generating analyses stratified by predicted RIL risk and durvalumab use. RESULTS: Severe RIL was defined as an absolute lymphocyte count nadir of 30 radiation therapy fractions, higher mean lung dose, and lower baseline absolute lymphocyte count (c-statistic: 0.70). External validation (130 patients, 41 [32%] with severe RIL) yielded similar discrimination (c-statistic: 0.69). In exploratory analyses, durvalumab use was associated with improved survival in patients with a low predicted risk of severe RIL, whereas no statistically significant association was observed in those with a high predicted risk, in both cohorts. CONCLUSIONS: We developed and externally validated a pretreatment prediction model for severe RIL during CCRT for NSCLC with consistent performance. In exploratory analyses, an association between durvalumab use and improved survival was observed in patients with a low predicted risk of severe RIL, but not in those with a high predicted risk. This model may help identify patients for lymphopenia-mitigating strategies and inform more personalized immunotherapy approaches, pending prospective validation.