3113 - Monte Carlo-Based Per-Transit Cardiac Lymphocyte Dose Modeling and Association with Radiation-Induced Lymphopenia in Locally Advanced NSCLC

12:30pm - 01:45pm ET

Poster Hall - Exhibit Hall A

Screen: 21

POSTER

Presenter(s)

Verny Rodriguez, MD - University of Kentucky, Lexington, KY

V. Rodriguez¹, R. C. McGarry¹, E. L. Johnson¹, J. Wang², Y. Wu², K. Shang³, M. Ge², S. Zou², Y. Zhang², J. Gorman¹, D. Lockhart¹, and W. Yan¹; ¹Department of Radiation Oncology, University of Kentucky, Lexington, KY, ²Department of Radiation Oncology, the Fourth Hospital of Hebei Medical University, Shijiazhuang, Hebei, China, ³Department of Radiotherapy, the Fourth Hospital of Hebei Medical University, Shijiazhuang, China

Purpose/Objective(s):

Radiation-induced lymphopenia (RIL) is associated with inferior outcomes following definitive chemoradiation (chemoRT) for locally advanced NSCLC. Existing models, including Estimated Dose to Immune Cells (EDIC), estimate cumulative circulating blood dose but do not model dose delivered during individual systemic circulation transits. We developed ICE3 (Immune Cell Circulatory Exposure Engine), a Monte Carlo–based per-transit cardiac dose model and tested whether accumulated dose across circulation is associated with severe RIL.

Materials/Methods:

Forty-five patients with locally advanced NSCLC treated with definitive concurrent chemoRT (UKY: 10–30 fx; China: 56 Gy/28 fx) were retrospectively analyzed. RT datasets were integrated with longitudinal absolute lymphocyte count (ALC) data. A Monte Carlo circulation model simulated 20,000 lymphocyte trajectories (mean circulation time 60 s) through a 3-compartment serial model (Body ? Heart ? Lung ? Body). UKY data (N=25) utilized full DICOM RT datasets; whereas China data (N=20) were DVH-derived. ICE3 modeling for China cohort was DVH-based without voxel-level spatial reconstruction. ICE3 Dmax was defined as the maximum accumulated dose to any simulated lymphocyte across treatment. ICE3 D90, D50, D100, and Mean were defined per DVH convention (D100 = minimum dose; Dmax = maximum dose). Outcomes included ALC nadir and Grade =3 RIL (ALC <0.5 K/µL). Associations were evaluated using Spearman correlation and ROC analysis. Analyses were done in Python/R; AI assisted coding only. Analyses were exploratory.

Results:

Complete ALC data was available for 35/45 patients. In these patients, Grade =3 lymphopenia was present in 66% (23/35). In the pooled cohort, EDIC V1.1 correlated significantly with ALC nadir (?=-0.465, p=0.005) and ALC % decline (?=+0.465, p=0.004). ICE3 Dmax showed a stronger correlation with ALC nadir in the UKY DICOM cohort (?=-0.590, p=0.010) with AUC=0.733 for Grade =3 classification, whereas EDIC did not (? = -0.271, p = 0.278). ICE3 Dmax demonstrated moderate discrimination for Grade =3 RIL (AUC 0.733; 95% CI 0.50–0.95), compared with EDIC (AUC 0.611). ICE3 Dmax was correlated with mean heart dose (? = 0.72, p < 0.001), limiting assessment of statistical independence. No multivariable modeling was performed due to sample size constraints, and independence was not established.

Conclusion:

In this hypothesis-generating pilot study, peak per-transit lymphocyte dose demonstrated an association with severe RIL and ALC nadir. Effect estimates were limited by small sample size and collinearity with mean heart dose. These findings support the biologic plausibility of per-transit immune dose modeling. Inclusion of two distinct treatment cohorts supports preliminary external generalizability. ICE3 represents a complementary mechanistic framework to EDIC. Prospective multi-site validation is ongoing to determine reproducibility and independence from conventional cardiac dose metrics.