Comparison of imaging-based bone marrow dosimetry methodologies and their dose–effect relationships in [177Lu]Lu-PSMA-617 RLT including a novel method with active marrow localization

Abstract Purpose Establishing accurate methods for red marrow (RM) dosimetry is an important step toward patient-specific treatment guidance. We compared image-based dosimetry methods and investigated their role in predicting changes in blood counts following [ 177 Lu]Lu-PSMA-617 radioligand therapy ( 177 Lu RLT). Methods Four image-based dosimetry methodologies were applied to patients who received 2-bed position serial 177 Lu SPECT/CT after cycle 1 of RLT, with segmentation of all spongiosa within the field-of-view performed on CT using deep learning tools. Cycle 1 RM absorbed doses (ADs) were estimated with: 1) the time-integrated activity (TIA) in segmented spongiosa coupled with MIRD-based S-values (MIRD); 2) the TIA concentration in the segmented aorta (a surrogate for blood-based dosimetry) coupled with MIRD-based S values (MIRD aorta ); 3) the voxel-level TIA map coupled with an in-house Monte Carlo (MC) dosimetry code that incorporated a micro-scale modeling of the spongiosa (MC); and 4) a novel method that utilizes [ 68 Ga]Ga-PSMA-11 PET/CT and [ 99m Tc]Tc-sulfur colloid (SC) SPECT/CT for tumor and marrow localization coupled with the above MC code, modified to allow tumor infiltration of the spongiosa (MC SC+PET ). Spearman rank correlation of AD from the four methods with changes in select blood counts was evaluated. Results Imaging data was available for 20 patients for methods 1–3, while SC images were available for 12 patients for method 4. Cycle 1 AD to the FOV RM was, on average, 1.9 Gy (range: 0.1–8.0 Gy) for MIRD, 0.08 Gy (range: 0.01–0.27 Gy) for MIRD aorta , 2.5 Gy (range: 0.1–10.3 Gy) for MC, and 1.6 Gy (range: 0.1–4.6 Gy) for MC SC+PET . The ADs from MIRD aorta were not concordant with MIRD, MC, or MC SC+PET (|CCC|< 0.01) and were generally underestimates. For 3 patients with high bone tumor burden, MC SC+PET gave lower average AD than MIRD (39%) and MC (53%), potentially due to more accurate localization of marrow and tumor. Cycle 1 RM ADs were correlated with relative change in blood counts at 6-weeks post-cycle 1 with significant correlation observed for neutrophils with MIRD, MC, and MC SC+PET with Spearman rank correlations ranging from r = − 0.61 to r = − 0.88 ( P < 0.01). Correlation with white blood cells at 6-months was also significant with r = − 0.80 ( P < 0.01) for these three methods. MIRD aorta did not correlate with any acute or chronic changes in blood counts. Conclusion The RM AD estimates from the blood-based surrogate were not concordant with the other image-based calculations and did not correlate with changes in blood values. Including patient-specific tumor and marrow distribution information resulted in lower AD for patients with a high bone metastatic burden. These findings have implications for managing hematological toxicities in 177 Lu RLT, especially if dosimetry-guided treatment planning is considered.