Background
Flumazenil (FMZ) is a specific, reversibly bound high-affinity neutral antagonist of the benzodiazepine site at the β-aminobutyric acid (GABA)-A receptor. In general, a single tissue compartment model with plasma input is used for analysis. For clinical studies, however, reference tissue models have also been used.
Objective
To determine the relation between different plasma input and reference tissue models and to define the optimal model for analysis of clinical FMZ studies. Methods: Eleven drug free patients (4 male; age range 22–54, average 37.5±10.8 years) with a major depressive episode (DSM IV) and nine healthy controls (6 male; age range 22–43, average 32.4±7.4 years) were included. A dynamic 3D scan with a duration of 60 minutes was acquired following bolus injection of ∼370 MBq FMZ. During the scan, arterial whole blood was monitored continuously using an on-line detection system and, at set times, discrete samples were taken in order to derive the metabolite corrected plasma curve. In addition, for each subject a T1-weighted structural MRI scan was acquired, using a 1.5 Tesla MRI scanner. The MRI scans were co-registered with summed FMZ images. Regions of interest (ROI) were defined on these co-registered MRI scans for the following structures: anterior, ventrolateral, dorsolateral and orbitomedial prefrontal cortex, anterior and posterior cingulate, medial and lateral temporal lobe and insular area, parietal and occipital area, cerebellum, hippocampus, putamen, thalamus, pons and white matter. ROIs were projected onto the dynamic FMZ scans, generating time activity curves for each region. Data were analysed using both single tissue (1 T) and reversible two tissue (2 T) compartment models with plasma input, providing values of the volume of distribution (Vd) of FMZ. In addition, both full (FRTM; 4 parameters) and simplified (SRTM; 3 parameters) reference tissue models were investigated, resulting in measures of binding potential (BP). In the latter analysis, both pons and white matter were investigated as reference tissue.
Results
Due to technical problems, 1 control and 2 patient scans could not be analysed. According to Akaike and Schwarz criteria, for most patients the 1 T model was selected for pons, white matter and the basal structures putamen and thalamus and the 2 T model for the other structures. Across all structures and patients, the 2 T model was selected in 61% of the fits. For these cases, Vd values obtained with the 2 T model were, on average, 6±4% higher than those obtained with the 1 T model. Across all structures and patients, the SRTM model was selected in 82 and 73% of the fits when pons and white matter were used as reference tissue, respectively. Pearson Correlation between BP from SRTM and Vd from the 2 T model was 0.997 and 0.988 using pons and white matter as reference tissue, respectively.
Conclusion
In contrast to previous studies, for cortical structures, a 2 T model provides slightly better fits in the majority of cases. Use of a 1 T model may result in a slight (∼6%) underestimation of Vd. Use of SRTM with pons as reference tissue is a good alternative for clinical studies.