Personalized survival models for choosing irradiated tumor burden on combined radiotherapy and immune checkpoint inhibitors in stage IV NSCLC

Abstract

Purpose: To assess the effects of irradiated tumor burden (ITB) and key prognostic factors for patients with advanced non-small cell lung cancer (NSCLC) who received a combination treatment of radiotherapy (RT) and immune checkpoint inhibitors (ICI). Methods: We retrospectively reviewed 79 patients with stage 4 NSCLC treated with ICI after RT. The linear and non-linear prediction models were trained to predict individual risk-based on the optimal parameter set selected from 14 clinical features including patient, tumor, treatment, and immunologic factors. We introduced a new prognostic factor, ITB, as the volume fraction of tumor that irradiated to investigate the impact of radiotherapy for the combination treatment. ITB's correlation with clinical variables was examined using log-rank survival tests and Pearson correlation maps. Results: Overall survival (OS) prediction indicates the non-linear random survival forest (RSF) model surpassed both the Cox linear model and gradient boosting model (RSF c-index = 0.78 90 %CI [0.77-0.80], Cox c-index = 0.75 90 %CI [0.74-0.76], GBM c-index = 0.77 90 %CI [0.76-0.78]; t-test P < 0.001 between RSF and Cox). Major prognostic factors for the final OS models are ITB, metastasis extent, and RT-induced lymphopenia. Patients with an ITB over 50 % exhibited improved outcomes (Median OS: 15.8 vs 7.9 months, p < 0.001; median progression-free survival (PFS): 12.5 vs 7.4 months, p < 0.001). ITB demonstrated independent predictive power, as no linear correlation with other variables was evident. Conclusion: Our findings endorse RT followed by ICI as a favorable protocol for advanced NSCLC, advocating for ITB's integration into future RT-ICI trial designs due to its significant prognostic implication.