Towards Personalized Dosimetry In Patients With Intrahepatic Cholangiocarcinoma Treated With Y-90-Loaded Glass Microsphere Selective Internal Radiation Therapy.

121 Purpose: Selective internal radiation therapy (SIRT) with yttrium-90 (90Y) is a valuable treatment in unresectable intrahepatic cholangiocarcinoma (ICC). Despite being a fundamental tool for this technique, a very sparse number of studies reported the value of dosimetry in this pathology. This work aimed to evaluate the agreement between technetium-99 (99mTc)-labeled macroaggregated albumin (MAA) SPECT/CT predictive dosimetry and 90Y PET/CT post-treatment dosimetry, and to assess the added value to predict response in ICC patients. Methods: We retrospectively compared the 99mTc-MAA SPECT/CT results of 19 treatment sessions (16 patients, 65 tumors) with 90Y-SIRT using glass microspheres and the post-treatment 90Y PET/CT data. Dose Volume Histograms (DVHs) derived from both modalities were calculated using Kernel based convolution (KBC) and Local Deposition Method (LDM). Absorbed dose (AD) using SPECT/CT data were computed using both relative (SPECTrel) and absolute (SPECTabs) normalization. A rigid registration between SPECT/CT and PET/CT was performed. The correlation of AD to tumors and whole liver normal tissue before and after treatment was assessed using the Mann-Whitney U test associated with the Lin’s concordance index. The significance of prognostic factors was tested using a univariate binary logistic regression at both lesion and patient levels. A sub-cohort of 35 lesions which had a volume > 10cm3 was also studied to minimize registration error and partial volume effect. Survival analysis was conducted through both univariate and multivariate Cox proportional-hazards ratio analyses with a variable selection performed with the elastic-net Cox regression model. Results: Median injected activity was 2.37 GBq (range 0.43-4.19). Mean tumor AD was 242+/- 121 Gy. Unlike SPECTabs-based computation, all AD metrics calculated using SPECTrel, regardless of the computation method, showed a good correlation when comparing with PET-based computation (p > 0.05 and ρc > 0.8). Computation using KBC and LDM exhibited a very high concordance (ρc > 0.99). Median overall-survival was 13.7 months (range: 1.39-39.06) and median progression-free-survival (PFS) was 12.1 months (range: 1.39 - 39.06). No permanent liver toxicities were observed. Ten lesions presented complete response (15%), 42 partial response (65%) and 13 stable disease (20%) according to mRECIST criteria. No progressive lesion was recorded. At a patient level analysis, only the tumor volume and the bi-lobar involvement were significantly associated with response at univariate analysis. No threshold was found between tumor AD and response. However, several DVH-based metrics were significantly higher in the responding lesions than in the non-responding ones using SPECT-based or PET-based data (namely: D95, D70, V50, V30 and dose coefficient of variation). This holds true for only SPECT-based data when considering lesions with volume greater than 10 cm3. Only ECOG was significantly associated with PFS at a multivariate analysis while no variables were retained for the OS analysis. No dosimetric variable was significantly associated with survival. Conclusions: 99mTc-MAA SPECTrel data showed a high concordance with 90Y PET/CT data in ICC tumors and particularly for large lesions (volume > 10 cm3). DVH calculated from SPECT proved to be valuable to estimate tumor response. These preliminary results paved the way to a new workflow for the computation of the required injected activity. Supplementary data including more patients will be presented at the meeting.