Abstract
1633
Objectives: 90Y-microspheres Radioembolization (90Y-RE) with is an emerging form of radiotherapy in the management of unresectable hepatocellular carcinoma (HCC) that requires precise delivery of radiation to planned treatment volumes to ensure efficacy. We characterized the differences in uptake and absorbed doses to tumor and normal liver within the treated liver (NL) between planned (99mTc-MAA) and delivered (90Y-glass microspheres) treatments for HCC patients.
Methods: We performed retrospective dosimetry analysis of a single arm, single-institution phase II study of sorafenib followed by 90Y-RE in advanced HCC patients (n=34). 90Y-RE treatment planning was done per the package insert - prescribing a mean nominal absorbed dose (120 Gy) to treated liver volume (containing tumors and NL) using 1-compartment MIRD model. Quantitative 99mTc-MAA and 90Y bremsstrahlung SPECT/CT’s, acquired as standard-of-care, were input into MIM 90Y-SurePlan software to compute planned and delivered voxel-level dose distributions, respectively. Treated liver volumes and tumors (D>2.5cm; ≤3 per patient; n=53) were segmented by an interventional radiologist using diagnostic CT/MRI of the liver registered to 99mTc-MAA SPECT/CT. All tumors with 1 cm<D<2.5 cm were segmented and included as non-target tumors. 99mTc-MAA SPECT/CT was registered to the corresponding 90Y SPECT/CT; the tumor/liver contours were auto-transferred and manually adjusted. Planned and delivered mean doses to tumors and NL were computed. The tumor to NL uptake ratio (TNR) was calculated for all patients. Summary statistics and linear correlation (r) analysis were performed. Bland-Altman analysis was performed to assess the mean difference and 95% limits of agreement (95%CI) between planned and delivered dose estimates.
Results: While population-median (range) of treated liver dose was similar between planned 120 (85-145) Gy and delivered 116.5 (84-140) Gy doses; patient-specific tumor and NL compartments showed large differences. The TNR were correlated (r=0.55, p<0.001) with similar median (range) of 2.5 (0.3-8.4) and 2.1 (0.7-6.9) in planned and delivered images. However, paired differences between planned and delivered TNR showed a wide 95%CI range of -3.0 to 4.0. In 14/34 (41%) cases the delivered TNR was within 0.5 of planned (or equivalent); but in 15/34 (44%) cases the TNR differences were >1.0 (or discordant). Mean tumor doses were well correlated (r=0.76, p<0.001) but population-median (range) delivered dose 165 (40-764) Gy was ~40% lower than planned dose 231 (9-627) Gy. Furthermore, paired difference between planned and delivered tumors doses were large with mean absolute and fractional differences (95%CI) of 91 (-145 to 226) Gy and 48% (-142% to 238%). Only in 18/53 (34%) cases were tumor doses delivered within 20% of planned doses (or equivalent); while in 10/53 (19%) cases tumor doses delivered were <20% of planned doses (or under-treated). The mean dose to NL were well correlated (r=0.69, p<0.001) with similar population-median (range) planned 87 (20-117) Gy and delivered 89 (21-136) Gy doses. However, paired differences between planned and delivered NL doses were found to have wide 95%CI range for absolute and fractional 95%CI differences of -41 to 35 Gy and -55% to 49%. Only in 22/34 (65%) cases were the absolute difference between planned and delivered NL doses found to be <20% (or equivalent); while in 12/34 (35%) cases the differences were found to be >20% (or discordant).
Conclusions: Patient-specific tumor and NL compartments showed large differences due to alterations in distribution and variations in TNR between MAA and Y90; therefore MAA-based planning doses are not reliable to estimate tumor response or NL toxicities. Post therapy imaging with PET or SPECT is essential for dosimetry and response assessment. Analysis of planned and delivered dose correlations with tumor response and toxicities are underway.