ABSTRACT
The application of an individualised dosimetric procedure for radioiodine therapy requires the intensive use of resources in nuclear medicine facilities. In practice, the amount of data taken per patient is too limited to obtain an accurate estimate of the absorbed dose in the thyroid. The individualised absorbed dose estimates can be enhanced using statistical tools for population-based approaches. The aim of this work was to build a population biokinetic model of thyroid uptake and elimination of radioiodine using a nonlinear mixed-effects approach in patients with Graves' disease. Input data for the model development were taken from a dosimetric method based on 123I imaging data. 123I decay-corrected uptake values were estimated at 4, 24, and 96â¯h post-administration and for 58 patients. The root mean squared error (RMSE) for predicted 123I uptake values by the fitted model was 4%. The root mean squared error of prediction (RMSEP) for out-of-sample 123I uptake values, computed by a leave-one-out cross-validation, was 12%. We calculated 131I activity to administer from out-of-sample predicted 123I uptake values and compared the result with that calculated from observed 123I uptake values. RMSEP values for therapeutic activity revealed that there were measuring points with higher weight than others in the model. The mixed-effects approach can be used to enhance the accuracy of dosimetric calculations in therapies using 131I. Assessing the accuracy of the predictive model enables choosing among different time-sampling schedules of the radioiodine thyroid uptake curve. This methodology can also be applied in other areas of radiation dosimetry.
