P081] A population-based and mixed model approach for enhancing iodine biokinetics in radioiodine therapy for hyperthyroidism

Abstract

Purpose Individualized calculation of radioiodine 131 I activity for the treatment of benign thyroid disease is imperative according to Euratom Directive 59/13. In general, measurements taken per patient are very limited to obtain an accurate estimate of the absorbed dose in the thyroid. The application of statistical tools can improve the accuracy of the biokinetic parameters used to calculate personalized activity. The aim of this work was to model the population biokinetics of thyroid for hyperthyroid patients using a nonlinear mixed-effects approach. Methods The model data were obtained from planar scintigraphic images obtained 4, 24 and 96 hours after the pretherapeutic administration of 111 MBq of 123 I-NaI . A total of 174 uptake measurements corrected for physical decay in patients who received 131 I - NaI for Graves’ disease were included in the study. The root mean squared error (RMSE), coefficient of determination (R2) and leave-one-out cross-validation (LOOCV) method were used to evaluate the goodness of the fit and performance of the model. The predicted and calculated uptake values were used to calculate the activity of 131 I for giving 120 Gy in thyroid from the MIRD (Medical Internal Radiation Dose) formula and the differences were evaluated using RMSE. Results The model fit to the data was very good (R2=0.9977 and RMSE=4%) and the standard deviation of the residual error was of 3%. The fitted model predicts with high accuracy the left-out observations and the calculated activities of I-131(RMSE=12%) and especially, when the 24-h observation was left out (RMSE=7% and 8.5%,respectively). The highest RMSE corresponded to the 96-hours measurement. Conclusions A population-based and mixed model approach can enhance the accuracy of biokinetic parameters included in the MIRD formula for the calculation of the optimum therapeutic activity of 131I. A high ability for predicting new observations and observed uptake values was found. The quantification of the accuracy of prediction of the model enables choose among different time sampling schedules of the radioiodine thyroid uptake curve. In our case, we could use a 2-point schedule with a low loss of accuracy against a 3-points schedule. This methodology could be extended to other areas where repeated measurements are made.