Main Session
Sep 29
PQA 06 - Genitourinary Cancer, Gynecological Cancer, and Health Care Access and Engagement

3383 - Association of Pre-Treatment Sacral Bone Marrow Morphology with Radiation-Induced Lymphopenia in Prostate Cancer

02:15pm - 03:30pm ET
Poster Hall - Exhibit Hall A
Screen: 19
POSTER

Presenter(s)

Konrad Stawiski, MD, PhD - Medical University of Lodz, Lodz, Lodzkie

K. Stawiski1,2, L. Kuncman3, M. Bilski4, C. Franzese5, M. Orzechowska6, K. Bulacz3, M. Lisik-Habib7, M. Jobczyk8, J. Fijuth6, and B. A. Jereczek-Fossa9; 1Department of Biostatistics and Translational Medicine, Medical University of Lódz, Lodz, Poland, 2Dana-Farber Cancer Institute, Boston, MA, 3Copernicus Memorial Hospital in Lodz Comprehensive Cancer Center and Traumatology, Lodz, Poland, 4Department of Radiotherapy, Medical University of Lublin, Lublin, Poland, 5IRCCS Humanitas Research Hospital, Milan, Italy, 6Medical University of Lodz, Lodz, Poland, 7Copernicus Memorial Hospital in Lodz, Lodz, Poland, 8John Paul II Provincial Hospital, Belchatów, Poland, 9IEO European Institute of Oncology IRCCS, Milan, Italy

Purpose/Objective(s): Pelvic nodal irradiation (PNI) causes radiation-induced lymphopenia (RIL) in prostate cancer, yet PNI decisions rely on staging without assessing bone marrow reserve. The pelvis and lower spine harbor most adult active hematopoietic marrow. We hypothesized that sacral bone morphology on planning CT identifies patients at higher risk for grade =2 RIL (nadir ALC <0.8×10?/L), and that this risk differs by PNI status. If confirmed, CT-derived marrow assessment could inform PNI decisions.

Materials/Methods: We studied 260 consecutive prostate cancer patients treated with VMAT (median primary PTV dose 70 Gy, nodal 53 Gy) in 2021-2022. TotalSegmentator v2 segmented pelvic bony structures on non-contrast planning CTs. PyRadiomics v3.0.1 extracted shape and first-order features; 14 sacral shape features from the pre-specified sacral ROI were tested against grade =2 lymphopenia with Bonferroni correction. Logistic regression compared cross-validated (5-fold × 20 repeats) AUC for age, age+PNI, and age+PNI+sacral morphology. Calibration was assessed by slope and Brier score. Dose-morphology correlations tested whether morphology acts as a dose proxy. Landmark Cox regression (from end of RT) assessed OS adjusted for age, PNI, ISUP grade, and intent.

Results: Grade =2 lymphopenia occurred in 120/260 (46%), with PNI as the dominant driver (73/84 [87%] vs 47/176 [27%], p<0.001). Sacral flatness (least-to-greatest principal axis ratio) predicted lymphopenia after Bonferroni correction (r=0.207, adjusted p=0.011; multivariate OR 1.58/SD, p=0.007). Adding sacral flatness to age+PNI improved cross-validated AUC from 0.756 to 0.823 (nested CV: 0.807; flatness selected 95/100 folds). The model was well calibrated (slope 0.98, Brier 0.164). Sacral flatness did not correlate with body or femoral head DVH (all |r|<0.05, p>0.4, n=256); mediation analysis using available dose surrogates showed ~0% mediated by dose. Stratification by flatness tertile and PNI yielded lymphopenia rates from 23% to 94%. In landmark analysis (40 deaths, 3.2y follow-up), grade =2 lymphopenia was associated with worse OS (HR 3.56, 95% CI 1.70–7.42, p<0.001).

Conclusion: Sacral bone morphology on planning CT predicts RIL beyond PNI and age, with good calibration. This effect was not explained by dose geometry. High sacral flatness combined with PNI identifies a subgroup with 94% lymphopenia risk. If validated, quantitative marrow assessment from routine planning CTs could inform PNI decisions.