Dosimetric analysis with post-treatment Y-90 imaging after transarterial radioembolization (TARE) treatment of liver tumours and its role in prediction of response to treatment

Aim: Liver tumours, both primary and secondary, are significant contributors to cancer-related deaths worldwide. Radioembolization is a local/regional treatment method applicable to these patient groups. In this study, we aimed to investigate the role of voxel-based dosimetry using 90Y PET/MR images after radioembolization treatment with 90Y glass and resin microspheres in predicting of response to treatment and its correlations with pre-treatment 99mTc-MAA SPECT/CT. As a secondary objective, we aimed to identify the reasons for toxicity and complications observed after treatment. Materials and Methods: A total of 97 treatments for 84 patients, with 60 using glass microspheres and 37 using resin microspheres were included in this study both retrospectively and prospectively. The majority of patients had colorectal cancer (44%) and hepatocellular carcinoma (28%) and the remaining patients had cholangiocellular, breast, pancreas, adrenal gland, thyroid, neuroendocrine carcinomas and uveal melanoma. Tumour and normal liver parenchyma in post-treatment 90Y PET/MR images were segmented, average dose values and dose-volume histograms were calculated for each compartment using 90Y PET images. Regression analyses were performed to investigate the distribution of dosimetric parameters between treatment response categories (complete response/partial response/stable/progression) which are determined in pre- and post-treatment PET images. ROC analyses were applied to determine the discriminatory power of these parameters in predicting treatment response and complete treatment response. Correlation between parameters obtained from 90Y PET/MRI and 99mTc-MAA SPECT/CT was also investigated. The success of predicting treatment response based on average tumor doses obtained from 90Y PET/MRI and 99mTc-MAA SPECT/CT were compared using ROC analyses. For toxicity assessment, clinical and dosimetric parameters associated with an increase in Child-Pugh score after treatment were evaluated. Results: Regression and ROC analyses showed that average tumour dose and almost all D values which are calculated from PET/MRI were statistically significant in predicting treatment response and complete treatment response. Also, in treatments with glass microspheres D60 and D70 values showed statistically significantly higher discriminatory power for predicting complete treatment response compared to average dose delivered to tumour. Threshold values with high specificity for predicting response and complete response to treatment were calculated. In comparison of PET/MRI and SPECT/CT, pre-treatment SPECT/CT accurately predicted normal liver parenchyma dose values, but prediction of T/N and tumour doses were less accurate. PET/MRI demonstrated superior discriminatory power in predicting response to treatment compared to SPECT/CT, based on average tumor dose values. ROC analysis for predicting of complete response to treatment returned an AUC value which is not statistically significantly different from 0.5 for SPECT/CT. Distribution of pre-treatment albumin level, ALBI score, and tumor volume percentage in the liver were significantly differed between groups with and without post-treatment Child-Pugh score increase. However, no statistically significant difference was found in the analysis of dosimetric data for predicting liver function impairment between the two groups. Conclusions: Voxel-based dosimetric analysis after TARE treatment with post-treatment 90Y PET/MRI can accurately predict response and complete response to the radioembolization treatment. 99mTc-MAA SPECT/CT can predict normal tissue dose and, to a lesser extent, T/N ratio and tumor doses. It still maintains its relevance in treatment planning, despite having lower discriminatory power for predicting treatment response compared to PET/MRI.