Development of dose response to Y-90 microsphere treatment of metastatic liver cancer by quantitative analysis of SPECT and PET images

Y-90 microsphere radiotherapy is an option for treating inoperable metastatic liver tumors. This takes advantage of the differing vascular supply of the tumor and normal liver. The radiation dosimetry can be complex due to the non-uniform distribution of the particles. Because of this difficulty, the recorded treatment absorbed dose is often calculated assuming a uniform distribution throughout the entire liver segment. This work represents a retrospective analysis of twelve consecutive patients treated with Y-90 microspheres for colorectal liver metastasis. Absorbed dose to tumor and normal liver tissue was calculated by two methods for comparison. Both were partition methods, one using an average tumor to normal liver vascularity ratio and the other a patient specific vascularity ratio derived from SPECT scans performed pre-treatment. Tumor response was quantitatively evaluated from pre and post treatment PET scans. Site-specific thresholding ROI volumes were used to determine tumor SUV in the image analysis. PET analysis showed a significant response as a whole with an average of 52% +/- 22% decrease in total tumor burden. The range of decrease, representing tumor response in size and metabolism was 17-91%. Dose versus response curves were generated based on the above calculations. The results and statistical analysis indicate that there is a significant difference in the tumor absorbed dose value when calculated by the traditional partition method using an average tumor to normal liver ratio as compared to use of a patient specific tumor to normal liver ratio derived from SPECT images. The paired t-test result demonstrated a significant difference with the t value of 3.06 corresponding to a P of 0.009. A linear regression analysis of each dose response curve allowed a comparison of each dose calculation method as well. There was an increase in the r value for the absorbed dose calculated by the patient specific method in all response parameters. The best fits to the data were achieved using the patient specific method, thus demonstrating that this method better predicts therapeutic outcome than currently practiced methods.