Forecasting Time-Dependent Radiation-Induced Lymphocyte Depletion in a Clinical Trial on Early-Stage Lung Cancer

Radiation-induced lymphocyte depletion (RILD) varies substantially among patients, and the extent to which different factors contribute to RILD remains incompletely understood. We present a model to predict time-dependent RILD for patients with early-stage lung cancer treated with stereotactic body radiation therapy (SBRT) by simulating the lymphocyte circulation dynamics in blood and lymph-rich organs.

The Python-based model took radiation therapy data in digital imaging and communications in medicine (DICOM) format from patients as input and simulated the real-time coupled dynamics of lymphocyte circulation between blood and lymph-rich compartments to estimate RILD and predict longitudinal peripheral absolute lymphocyte count (ALC) trajectories up to 6-month post-treatment. The model was optimized using a retrospective cohort of 64 patients and tested on an independent cohort of 51 patients from clinical trial NCT04273893. Spearman's rank correlation and point-biserial correlation were used to analyze the impact of dosimetric parameters and patient characteristics on RILD and the model's ability to account for these parameters. Associations between model-predicted nadir ALC percentage and overall survival/event-free survival were assessed using Kaplan-Meier analysis with log-rank tests and univariate Cox proportional hazards models.

In the independent test cohort, the mean [SD] absolute ALC difference between prediction and measurement was 0.28[0.23]×10⁹ cell/L, with 83% of predictions within 0.5×10⁹ cell/L of observed values. The model successfully captured the contributions of irradiating blood-rich and immune-rich organs to RILD. The model's ability to make longitudinal predictions of lymphocyte numbers in organ compartments beyond blood indicated that irradiation of the lymphatic system was the primary driver of nadir timing and nadir ALC. Predicted nadir ALC percentage was significantly associated with both overall survival and event-free survival.

The model forecasted time-dependent ALC trajectories for patients with lung cancer undergoing SBRT up to 6 months post-treatment. This approach enables estimation of individualized RILD from treatment plans, supporting treatment personalization to mitigate immune toxicity while preserving standard-of-care SBRT.