Nonlinear Mixed-Effects Modeling Approach for Simplified Reference Tissue Model

Objective: The conventional approach to the analysis of dynamic PET data can be described as a two-stage approach. In Stage 1, each individual's kinetic parameter estimates are obtained by modeling their PET data. Then in Stage 2, those parameter estimates are treated as though they are the observed data and compared across subjects and groups using standard statistical analyses. In this context, we explore the application of nonlinear mixed-effects (NLME) model under the assumptions of simplified reference tissue model. Methods: In the NLME framework, all subject's PET data are modeled simultaneously and the estimation of kinetic parameters and statistical inference across subjects are performed jointly. Results: In simulated [C-11]WAY100635 data, this NLME approach shows improved power (6-27% increase) for detecting group differences and greater consistency of population (1.13-1.44 times greater) and individual-level parameter estimation compared to the two-stage approach applying simplified reference tissue model for pharmacokinetic modeling of PET data. We applied our NLME approach to clinical PET data and observed shrinkage of individual-level parameters that is inherent in this modeling structure. Conclusion: The proposed approach is more powerful and accurate than the two-stage approach under the assumptions of simplified reference tissue model in PET data. Significance: The stability of the NLME approach not only improves the efficiency of collected data, but also comes with no additional financial cost and negligible computation cost.