Measurement of GABA A receptor binding in vivo with [ 11 C]Flumazenil: A test–retest study in healthy subjects ☆ Elina Salmi, a, ⁎ Sargo Aalto, b Jussi Hirvonen, c Jaakko W. Långsjö, c Anu T. Maksimow, c Vesa Oikonen, d Liisa Metsähonkala, e Jussi Virkkala, f Kjell Någren, g and Harry Scheinin h a Investigator, Turku PET Centre, University of Turku and Department of Otorhinolaryngology — Head and Neck Surgery, Turku University Hospital, Finland b Investigator, Department of Psychology, Åbo Akademi University and Turku PET Centre, University of Turku, Finland c Investigator, Turku PET Centre, University of Turku, Finland d Modeller, Turku PET Centre, University of Turku, Finland e Child Neurologist, Department of Child Neurology, Helsinki University Hospital, Finland f Department of Clinical Neurophysiology, Medical Imaging Centre, Pirkanmaa Hospital District, Tampere, Finland g Radiochemist, Turku PET Centre, University of Turku, Finland h Professor, Turku PET Centre and Department of Anesthesiology and Intensive Care, University of Turku, Finland Received 27 June 2007; revised 24 February 2008; accepted 26 February 2008 Available online 4 March 2008 [ 11 C]Flumazenil is widely used in positron emission tomography (PET) studies to measure GABA A receptors in vivo in humans. Although several different methods have been applied for the quantification of [ 11 C]flumazenil binding, the reproducibility of these methods has not been previously examined. The reproducibility of a single bolus [ 11 C]flumazenil measurements was studied by scanning eight healthy volunteers twice during the same day. Grey matter regions were analyzed using both regions-of-interest (ROI) and voxel-based analysis methods. Compartmental kinetic mo- delling using both arterial and reference region input function were applied to derive the total tissue distribution volume (V T ) and the binding potential (BP) (BP P and BP ND ) of [ 11 C]flumazenil. To measure the reproducibility and reliability of each [ 11 C]flumazenil binding parameter, absolute variability values (VAR) and intraclass correla- tion coefficients (ICC) were calculated. Tissue radioactivity concentration over time was best modelled with a 2-tissue compartmental model. V T showed with all methods good to excellent reproducibility and reliability with low VARs (mean of all brain regions) (5.57%–6.26%) and high ICCs (mean of all brain regions) (0.83–0.88) when using conventional ROI analysis. Also voxel-based analysis methods yielded excellent reproducibility (VAR 5.75% and ICC 0.81). In contrast, the BP estimates using pons as the reference tissue yielded higher VARs (8.08%–9.08%) and lower ICCs (0.35–0.80). In conclusion, the reproducibility of [ 11 C]flumazenil measurements is considerably better with outcome measures based on arterial input function than those using pons as the reference tissue. The voxel-based analysis methods are proper alternative as the reliability is preserved and analysis automated. © 2008 Elsevier Inc. All rights reserved. Introduction The benzodiazepine antagonist flumazenil labelled with carbon- 11 ([ 11 C]flumazenil) is widely used in positron emission tomo- graphy (PET) studies to quantify the cerebral GABA A receptors in vivo in human. Because of its high specific to non-specific binding ratios with the regional uptake presenting the known distribution of benzodiazepine receptors in the brain (Möhler and Richards, 1981; Hunkeler et al., 1981), it is highly suitable for this purpose. [ 11 C] Flumazenil has been used in several PET studies to quantify GABA A receptor binding in neurological and psychiatric disorders, such as epilepsy, stroke, congenital syndromes, anxiety disorders and dementia (Savic et al., 1988; Heiss et al., 1998; Lucignani et al., 2004; Malizia et al., 1998; Ihara et al., 2004). Recently, [ 11 C] flumazenil binding has also been used to study the GABAergic effects of different anesthetics (Gyulai et al., 2001; Salmi et al., 2004; Salmi et al., 2005; Salmi et al., 2008). The binding of [ 11 C]flumazenil has earlier been assessed as the total tissue distribution volume (V T ) derived with the methods utilizing arterial input function (Holthoff et al., 1991; Koeppe et al., 1991; Ihara et al., 2004, Hammers et al., 2001; Hammers et al., 2005) or as the binding potential (BP) derived with methods using pons as the reference region (Millet et al., 2002; Lucignani et al., 2004). Although several studies have assessed different aspects of www.elsevier.com/locate/ynimg NeuroImage 41 (2008) 260 – 269 ☆ From the Turku PET Centre and the Department of Anesthesiology and Intensive Care, Turku University Hospital, Turku, Finland. ⁎ Corresponding author. Turku PET Centre, PO Box 52, FIN-20521 Turku, Finland. Fax: +358 2 231 8191. E-mail address: anelsa@utu.fi (E. Salmi). Available online on ScienceDirect (www.sciencedirect.com). 1053-8119/$ - see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2008.02.035