D. H. Hristov1, E. Rahimy2, M. J. Kim1,3, J. Wang3, J. Lewis2, and E. A. Kidd4; 1Department of Radiation Oncology, Stanford University School of Medicine, Palo Alto, CA, 2Department of Radiation Oncology, Stanford University School of Medicine, Stanford, CA, 3Stanford University, Department of Radiation Oncology, Stanford, CA, 4Department of Radiation Oncology, Stanford University, Stanford, CA
Purpose/Objective(s): Dose-volume constraints used for planning bone marrow sparing pelvic radiotherapy such as fractional Active Marrow (AM) volumes under 10 Gy and 20 Gy (V10 and V20) ignore AM heterogeneity and implicitly equate spared AM volume with spared AM load. Using FDG-PET Total Glycolytic Activity (TGA) of the pelvic bone (PB) as an AM load measure, in a single institutional retrospective study of cervical cancer patients, we explore dose-volume and dose-TGA metrics as predictors of grade 3+ hematologic toxicity and nadir WBC. We hypothesize that AM dose-TGA metrics are more predictive than dose-volume ones. Materials/
Methods: Patients on the study underwent definitive chemoradiation with weekly cisplatin, no GCSF administration, had a baseline FDG-PET, and serial CBCs taken. One radiation oncologist contoured retrospectively the PB contours for all patients. Active marrow was segmented at the PB mean SUV. Similarly to AM V10 and V20, pelvic bone TGA10 and TGA20 were calculated but at 1 and 2 weeks prior to WBC nadir (TGA10_1,2W and TGA20_1,2W) to account for known 10-14 day delays between a precursor stem cell division and blood cell release in the circulation. TGAs were normalized to the patients’ PB volumes and aorta SUVs. Differences in predictions were evaluated using logistic regression and GLM (generalized linear effects with variable selection) models. Results: Sixty-six women from 2009 to 2020 were evaluable. The majority had FIGO stage (2018) IIIC1 disease (44%) and underwent IMRT (95%) to the pelvis only (74%), prescribed to a mean dose of 48 Gy; 61% received a nodal boost to a mean dose of 58 Gy. The majority (83%) completed at least 5 cisplatin cycles. The most common hematologic toxicities (HT) were leukopenia, grade 2 (39%) and grade 3 (32%). The PB SUV coefficient of variation was 38% +/- 8%. Baseline PB TGAs were moderately correlated to baseline WBCs (r = 0.44, P = 0.0003). On a multivariate analysis nadir WBC correlated to AM V10, V20, PB TGA10_1,2W, and TGA20_1,2W (P < 0.05). With control for other dosimetric variables, nadir WBC correlated only to TGA10_2W (P = 0.03). Grade 3+ HT predictors were AM V10, V20, PB TGA10_1,2W, and TGA20_1,2W (P < 0.05). The best univariate logistic regression predictor was TGA10_2W. The best GLM model of nadir WBCs included baseline WBCs (P < 0.05), PB TGA10_2W (P < 0.001), and AM V10 (P = 0.05). Conclusion: The expected value of spared TGA below 10 Gy at 2 weeks prior to WBC nadir was predictive for nadir WBC/3+ HT and remained a significant predictor of nadir WBC in GLM with competing, correlated dosimetric predictors. These findings suggest further research is warranted on the informed design of FDG-PET based predictors that consider the heterogeneity of bone marrow and its response to planned spatially and temporarily varying absolute dose distributions.
