3299 - High-Throughput Dosimetric Profiling of Pelvic Organs at Risk and Lymphopenia in Prostate Radiotherapy
Presenter(s)
L. Kuncman1,2, K. Stawiski2,3, M. Bilski4,5, C. Franzese6,7, M. Orzechowska1, K. Bulacz2, M. Lisik-Habib2, M. Jobczyk8, J. Fijuth1,2, and B. A. Jereczek-Fossa9,10; 1Medical University of Lodz, Lodz, Poland, 2Copernicus Memorial Hospital in Lodz Comprehensive Cancer Center and Traumatology, Lodz, Poland, 3Medical Univeristy of Lodz, Lodz, Poland, 4Affidea Nu-med Center of Oncology Diagnostics and Therapy, Zamosc, Poland, 5Department of Radiotherapy, Medical University of Lublin, Lublin, Poland, 6Humanitas University, Milan, Italy, 7IRCCS Humanitas Research Hospital, Milan, Italy, 8John Paul II Provincial Hospital, Belchatów, Poland, 9IEO European Institute of Oncology IRCCS, Milan, Italy, 10University of Milan, Milan, Italy
Purpose/Objective(s):
Determinants of radiation-induced lymphopenia (RIL) in prostate radiotherapy, including organ-at-risk (OAR) and dose-range effects, remain unclear, limiting mitigation strategies and informed pelvic field design, including elective nodal irradiation. We hypothesized that high-dimensional pelvic dosimetry screening could identify dominant predictors of clinically significant RIL and clarify its association with overall survival (OS).Materials/Methods:
We retrospectively analyzed consecutive patients treated with external-beam radiotherapy in 2021 for localized prostate adenocarcinoma, including both radical and post-prostatectomy radiotherapy. All complete blood counts performed for any indication from radiotherapy start through follow-up were included. Delivered-plan dose–volume data were screened across 13 pelvic candidate structures potentially contributing to RIL and 62 dose–volume histogram descriptors per structure (806 features). Random-forest screening (permutation importance) ranked dosimetric predictors; ROC/Youden analysis defined an interpretable cutoff with bootstrap stability assessment. OS was analyzed in a multivariable Cox model (age, EAU risk group, RIL, and the top dosimetric predictor dichotomized at the cutoff), with bootstrap internal validation.Results:
The cohort included 263 patients with 1901 blood counts. CTCAE grade =2 lymphopenia occurred in 124 (47.1%) and grade =3 in 50 (19.0%). In high-dimensional feature ranking, pelvic_bones_3mm_V10Gy(%) was the strongest predictor of grade 2–4 lymphopenia (AUC 0.925), with an optimal cutoff of 34.26%. Bootstrap validation demonstrated a stable cutoff distribution centered around 34–35%. Unadjusted Kaplan–Meier analysis showed significantly worse OS in patients with CTCAE grade =2 lymphopenia than in those with CTCAE grade 0–1 lymphopenia. To identify confounding, a Cox model including age and EAU risk group indicated age as a confounder (HR 1.05; 95% CI 1.00–1.11; p=0.0435), whereas EAU risk group terms were not significant. Age-adjusted Kaplan–Meier estimation remained significant (p=0.008); median OS was not reached in either group, and restricted mean survival time was higher for CTCAE grade 0–1 versus grade =2 lymphopenia (58.8±0.8 vs 52.6±1.6 months). In multivariable Cox analysis, RIL was associated with worse OS (HR 2.44; 95% CI 1.11–5.35; p=0.026) after adjustment for age and pelvic_bones_3mm_V10Gy(%) group (C-index 0.675; optimism-corrected C-index 0.642).Conclusion:
High-throughput pelvic OAR dosimetry identified V10Gy as a dominant, stable predictor of clinically significant RIL. RIL was independently associated with worse OS, supporting RIL-aware pelvic field decisions. To our knowledge, this is the first study linking RIL, OS, and high-throughput OAR dosimetric predictors in prostate cancer radiotherapy.