Encrypted login | home

Program Information

Molecular-Imaging Based Assessment of Liver Complications for Yttrium-90 Microsphere Treatments: Can Existing NTCP Models Explain Clinical Outcomes?

M Lin

M Lin*, E Choi , M Chuong , B Saboury , F Moeslein , W D'Souza , M Guerrero , University of Maryland School of Medicine, Baltimore, MD


WE-AB-204-2 (Wednesday, July 15, 2015) 7:30 AM - 9:30 AM Room: 204

Purpose: To evaluate weather the current radiobiological models can predict the normal liver complications of radioactive Yttrium-90 (⁹⁰Y) selective-internal-radiation-treatment (SIRT) for metastatic liver lesions based on the post-infusion ⁹⁰Y PET images.

Methods: A total of 20 patients with metastatic liver tumors treated with SIRT that received a post-infusion ⁹⁰Y-PET/CT scan were analyzed in this work. The 3D activity distribution of the PET images was converted into a 3D dose distribution via a kernel convolution process. The physical dose distribution was converted into the equivalent dose (EQ2) delivered at 2 Gy based on the linear-quadratic (LQ) model considering the dose rate effect. The biological endpoint of this work was radiation-induce liver disease (RILD). The NTCPs were calculated with four different repair-times (T1/2_Liver_Repair= 0,0.5,1.0,2.0 hr) and three published NTCP models (Lyman-external-RT, Lyman 90Y-HCC-SIRT, parallel model) were compared to the incidence of RILD of the recruited patients to evaluate their ability of outcome prediction.

Results: The mean normal liver physical dose (avg. 51.9 Gy, range 31.9-69.8 Gy) is higher than the suggested liver dose constraint for external beam treatment (~30 Gy). However, none of the patients in our study developed RILD after the SIRT. The estimated probability of 'no patient developing RILD' obtained from the two Lyman models are 46.3% to 48.3% (T1/2_Liver_Repair= 0hr) and <1% for all other repair times. For the parallel model, the estimated probability is 97.3% (0hr), 51.7% (0.5hr), 2.0% (1.0hr) and <1% (2.0hr).

Conclusion: Molecular-images providing the distribution of ⁹⁰Y enable the dose-volume based dose/outcome analysis for SIRT. Current NTCP models fail to predict RILD complications in our patient population, unless a very short repair-time for the liver is assumed. The discrepancy between the Lyman ⁹⁰Y-HCC-SIRT model predicted and the clinically observed outcomes further demonstrates the need of an NTCP model specific to the metastatic liver SIRT.

Contact Email: