Voxel-Based Dosimetry Predicts Hepatotoxicity in Hepatocellular Carcinoma Patients Undergoing Radioembolization with 90 Y Glass Microspheres

Personalized dosimetry holds promise to improve radioembolization treatment outcomes in hepatocellular carcinoma (HCC) patients. To this end, tolerance absorbed doses for nontumor liver tissue are assessed by calculating the mean absorbed dose to the whole nontumor liver tissue (AD-WNTLT), which may be limited by its neglect of nonuniform dose distribution. Thus, we analyzed whether voxel-based dosimetry could be more accurate in predicting hepatotoxicity in HCC patients undergoing radioembolization. In total, 176 HCC patients were available for this retrospective analysis; of these, 78 underwent partial- and 98 whole-liver treatment. Posttherapeutic changes in bilirubin were graded using the Common Terminology Criteria for Adverse Events. We performed voxel-based and multicompartment dosimetry using pretherapeutic Tc-labeled human serum albumin SPECT and contrast-enhanced CT/MRI and defined the following dosimetry parameters: AD-WNTLT; the nontumor liver tissue volume exposed to at least 20 Gy (V20), at least 30 Gy (V30), and at least 40 Gy (V40); and the threshold absorbed dose to the 20% (AD-20) and 30% (AD-30) of nontumor liver tissue with the lowest absorbed dose. Their impact on hepatotoxicity after 6 mo was analyzed using the area under the receiver-operating-characteristic curve; thresholds were identified using the Youden index. The area under the curve for prediction of posttherapeutic grade 3+ increases in bilirubin was acceptable for V20 (0.77), V30 (0.78), and V40 (0.79), whereas it was low for AD-WNTLT (0.67). The predictive value could further be increased in the subanalysis of patients with whole-liver treatment, where a good discriminatory power was found for V20 (0.80), V30 (0.82), V40 (0.84), AD-20 (0.80), and AD-30 (0.82) and an acceptable discriminatory power was found for AD-WNTLT (0.63). The accuracies of V20 ( = 0.03), V30 ( = 0.009), V40 ( = 0.004), AD-20 ( = 0.04), and AD-30 ( = 0.02) were superior to that of AD-WNTLT but did not differ significantly from each other. The respective thresholds were 78% (V30), 72% (V40), and 43 Gy (AD-30). Statistical significance was not reached for partial-liver treatment. Voxel-based dosimetry may more accurately predict hepatotoxicity than multicompartment dosimetry in HCC patients undergoing radioembolization, which could enable dose escalation or deescalation with the intent to optimize treatment response. Our results indicate that a V40 of 72% may be particularly useful in whole-liver treatme