Restricted maximum likelihood estimation of PET neuroreceptor occupancy in the absence of a reference region

Brain positron emission tomography (PET) is a useful technique for estimating the neuroreceptor occupancy of a drug in vivo. In the absence of a reference region, occupancy can be obtained from an "occupancy plot" with ordinary least squares (OLS) regression. However, OLS has been found to return inefficient occupancy estimations. The aim of this study was to improve the accuracy and precision of occupancy estimations.

Within a simulation framework, the efficiency of several model II regression approaches (accounting for error in the independent variable) and restricted maximum likelihood estimator (REML, specifically modeling the drug occupancy) was compared to the efficiency of OLS.

Efficiency of REML was 171%-210% the efficiency of OLS, while model II regressions were found to be substantially less efficient.

In the absence of a reference region, it is recommended to use occupancy REML instead of OLS in order to increase the validity of occupancy estimations and thus decrease the costs of PET research.