Radioembolization of Hepatocellular Carcinoma with (90)Y Glass Microspheres: No Advantage of Voxel Dosimetry with Respect to Mean Dose in Dose–Response Analysis with Two Radiological Methods

SIMPLE SUMMARY: We confirmed that the non-uniformity of an intra-lesion dose distribution, which was introduced in calculations as voxel dosimetry, did not significantly improve the AUC values of the dose–response relationship with respect to the mean dose. This was probably derived from the strong correlations (all p < 0.0001) among all voxel-based dosimetric variables (minimum Spearman correlation coefficient: 0.67) caused by the limited spatial resolution of nuclear medicine images. Responses were assessed with mRECIST and with an experimental densitometric method with a response threshold optimized at 20% HU variation. Significant dose–response agreement was obtained only with the densitometric method and only with post-therapy 90Y-PET data. More unexpectedly, the injection of Theraspheres™ on day 8 from the reference date rather than on day 4 worsened the dose–response correlation and reduced the efficacy at high doses. This may be explained by the increased non-uniformity following the non-linear mega-clustering effect triggered by the higher number of microspheres/GBq injected on day 8. ABSTRACT: In this confirmatory study, we tested if a calculation that included the non-uniformity of dose deposition through a voxel-based dosimetric variable Ψ was able to improve the dose–response agreement with respect to the mean absorbed dose D. We performed dosimetry with (99m)Tc-MAA SPECT/CT and (90)Y-PET/CT in 86 patients treated 8 instead of 4 days after the reference date with 2.8 times more (90)Y glass microspheres/GBq than in our previous study. The lesion-by-lesion response was assessed with the mRECIST method and with an experimental densitometric criterion. A total of 106 lesions were studied. Considering Ψ as a prognostic response marker, having no Ψ provided a significantly higher AUC than D. The correlation, t-test, and AUC values were statistically significant only with the densitometric method and only with post-therapy dosimetry. In comparison with our previous study, the dose–response correlation and AUC values were poorer (maximum r = 0.43, R(2) = 0.14, maximal AUC = 0.71), and the efficacy at a high dose did not reach 100%. The expected advantages of voxel dosimetry were nullified by the correlation between any Ψ and D due to the limited image spatial resolution. The lower AUC and efficacy may be explained by the mega-clustering effect triggered by the higher number of microspheres/GBq injected on day 8.