[11C]flumazenil metabolite measurement in plasma is not necessary for accurate brain benzodiazepine receptor quantification

In this work, a mathematical correction for metabolites has been validated which estimates the relative amount of [C-11]flumazenil ([C-11]FMZ) in the total plasma curve from the tissue kinetic data without the need for direct metabolite measurement in blood plasma samples. Kinetic data were obtained using a 90-min three-injection protocol on five normal volunteers. First, the relative amount of [C-11]FMZ in plasma was modelled by a two-parameter exponential function. The parameters were estimated either directly by fitting this model to the blood plasma metabolite measurements, or indirectly from the simultaneous fitting of tissue time activity curves from several brain regions with a non-linear FMZ kinetic model. Second, the direct and indirect metabolite corrections were fixed and the FMZ compartmental parameters were determined on a regional basis in the brain. The validation was performed by comparing the regional values of benzodiazepine receptor density B-max and equilibrium dissociation constant K-d Obtained with the direct metabolite correction with those values obtained with the indirect correction. For B-max the correlation coefficient r(2) was above 0.97 for all subjects and the slope values of the linear regression were within the interval [0.97, 1.2]. For K-d, r(2) was above 0.96, and the slope values of the linear regression were within the interval [0.99, 1.1]. Simulation studies were performed in order to evaluate whether this metabolite correction method could be used in a clinical protocol where only a single [C-11]FMZ injection and a linear compartmental model are used. The resulting [C-11]FMZ distribution volume estimates were found to be linearly correlated with the true values, with r(2) = 1.0 and a slope value of 1.1. The mathematical metabolite correction proved to be a feasible and reliable method to estimate the relative amount of [C-11]FMZ in plasma and the compartmental model parameters for time-injection protocols. Although validation with real data is necessary, simulation results suggest that our analysis method may also be applied to single-injection protocols.