Long-term test–retest reliability of striatal and extrastriatal dopamine D_2/3 receptor binding: study with [¹¹C]raclopride and high-resolution PET

Abstract

We measured the long-term test–retest reliability of [¹¹C]raclopride binding in striatal subregions, the thalamus and the cortex using the bolus-plus-infusion method and a high-resolution positron emission scanner. Seven healthy male volunteers underwent two positron emission tomography (PET) [¹¹C]raclopride assessments, with a 5-week retest interval. D_2/3 receptor availability was quantified as binding potential using the simplified reference tissue model. Absolute variability (VAR) and intraclass correlation coefficient (ICC) values indicated very good reproducibility for the striatum and were 4.5%/0.82, 3.9%/0.83, and 3.9%/0.82, for the caudate nucleus, putamen, and ventral striatum, respectively. Thalamic reliability was also very good, with VAR of 3.7% and ICC of 0.92. Test–retest data for cortical areas showed good to moderate reproducibility (6.1% to 13.1%). Our results are in line with previous test–retest studies of [¹¹C]raclopride binding in the striatum. A novel finding is the relatively low variability of [¹¹C]raclopride binding, providing suggestive evidence that extrastriatal D_2/3 binding can be studied in vivo with [¹¹C]raclopride PET to be verified in future studies.

Introduction

[¹¹C]Raclopride is a commonly-used dopamine (DA) D_2/3 receptor antagonist in positron emission tomography (PET) studies assessing D_2/3 receptor availability. It dissociates rapidly from receptors allowing binding equilibrium to be established in vivo within the time span of a PET experiment.^{1, 2} Dopaminergic PET ligands can be used to indirectly measure changes in synaptic DA concentration in vivo in the striatum in response to drugs such as cocaine or amphetamine^{3, 4} and also in response to cognitive challenges.⁵ Dopamine is thought to compete with [¹¹C]raclopride at D_2/3 receptors, and DA release during a challenge is inferred from reduced [¹¹C]raclopride binding.⁶ Obviously, accurate and reproducible measurement of ‘baseline' D_2/3 receptor availability is essential for such studies.

Even though the expression of D_2/3 receptors is highest in the striatum, dopaminergic projections show widespread efferents terminating in limbic, thalamic, and cortical regions.⁷ In extrastriatal areas, [¹¹C]raclopride has been considered to be less adequate for measuring DA activity because [¹¹C]raclopride binding is far lower in extrastriatal than in striatal areas because of the lower density of D_2/3 receptors resulting in low signal-to-noise ratios that hamper the reliability of the measurement.⁸ Despite these concerns, there is evidence that decreases in [¹¹C]raclopride binding potential (BP_ND) can also be observed in extrastriatal regions after drug or behavioral challenges.^{9, 10, 11} Changes in extrastriatal [¹¹C]raclopride binding have also been associated with brain pathologies such as Parkinson's, Huntington's, and Alzheimer's diseases.^{12, 13, 14} The loci of the extrastriatal reductions of [¹¹C]raclopride binding in the abovementioned studies are also consistent with the distribution of extrastriatal D_2/3 receptors reported in postmortem studies, where the thalamus has the highest density of D_2/3 receptors followed by the lateral temporal cortex, the anterior cingulate, and the frontal cortex.¹⁵

The region of interest (ROI) measurement of striatal D_2/3 receptor binding using single-bolus [¹¹C]raclopride injection is highly reliable.^{16, 17, 18} In most test–retest studies, scans have been performed on the same day^{17, 18, 19} and only one study has documented long-term reproducibility.¹⁶ Long-term stability is essential for follow-up studies using pharmacological or behavioral challenges, where the interval between measurements can be relatively long. The reproducibility of [¹¹C]raclopride binding in cortical areas has not yet been fully assessed. The ability to reliably measure DA release in vivo in extrastriatal regions would enable investigation of interactions between striatal and extrastriatal DA systems and provide knowledge of a broader range of functions modulated by DA neurotransmission. The aim of the present study was to measure striatal, thalamic, and cortical test–retest reliability of [¹¹C]raclopride BP using bolus-plus-infusion (B/I) method and high-resolution PET. The novelty of this study lies in the assessment of long-term reproducibility of D_2/3 receptor binding not only in striatal subregions but also in several extrastriatal areas.

Materials and methods

The study protocol was approved by the Ethics Committee of the Hospital District of Southwestern Finland. Subjects were given written information about all relevant issues involved in the study. Written consent was obtained from each subject. This study was performed according to the ethical guidelines given by the Declaration of Helsinki.

Subjects

Seven healthy, right-handed, nonsmoking, and nonmedicated Finnish male university students took part in an ongoing study on the effects of working-memory training on striatal DA release. These subjects belonged to the control group who received no training, and they were chosen for the present test–retest analyses based on optimal motion data. Each subject underwent two [¹¹C]raclopride PET scans, 5 weeks apart. Each scan started with 55 minutes of baseline measurement during which subjects performed a very simple computerized letter-recall task, followed by 25 minutes working memory task. Only the 55 minutes of baseline data was analyzed in this study. During the baseline measurement the subjects were shown 7 to 15 letter sequences, where all the letters were identical (e.g., A – A – A – …). When the sequence ended, the subjects were to report the last four letters by pressing buttons corresponding to A, B, C, and D. To exclude any structural brain abnormalities and to obtain anatomic references, all subjects underwent 3-T magnetic resonance imaging. Age, height, and weight of the subjects were 24±2 years, 176±1 cm, and 69±3 kg, respectively (mean±s.d.).

Preparation of [¹¹C]Raclopride

[¹¹C]Raclopride was prepared as described in our previous papers.^{19, 20}

Positron Emission Tomography Imaging

Positron emission tomography experiments were performed using a brain-dedicated high-resolution PET scanner, the ECAT high-resolution positron emission scanner (HRRT; Siemens Medical Solutions, Knoxville, TN, USA). The HRRT is a dual-layer, crystal-detector scanner allowing for depth-of-interaction measurement for coincident photons. It yields an isotropic 2.5-mm intrinsic spatial resolution. Spatial resolution in reconstructed images varies in radial and tangential directions from ~2.5 to 3 mm and in axial directions from 2.5 to 3.5 mm in the 10-cm field of view covering most of the brain.²¹ The left antecubital vein was cannulated and [¹¹C]raclopride bolus injections were administered intravenously, followed immediately by continuous infusion of [¹¹C]raclopride for 80 minutes (data from the first 55 minutes of acquisition was used for test–retest analyses). A bolus-to-infusion rate ratio (K_bol) of 105 minutes and a bolus component of 50% (of the total volume of the tracer) were chosen according to an optimization study by Watabe et al,²² as well as earlier studies by Mawlawi et al¹⁸ and Martinez et al.²³ The minimum specific activity in any scan was 205 MBq/nmol and there were no statistically significant differences in the injected doses, P=0.97 (bolus)/P=0.86 (infusion); the specific radioactivities, P=0.87 (bolus)/P=0.88 (infusion); or the injected masses, P=0.78 (bolus)/P=0.86 (infusion) between the test and retest studies. Radiochemical purity was >99% in all cases. Before each emission scan, a transmission scan was performed using a ¹³⁷Cs point source. Emission scans started at the injection time, and data were collected in list-mode format.

Image Processing

List-mode format PET measurements were histogrammed into series of three-dimensional sinograms using span 9 and a maximum ring difference of 67, resulting in 2,209 sinograms in 16 segments. Image reconstructions were made using an ordinary Poisson ordered subset expectation maximization algorithm with 16 subsets and 8 iterations and a voxel size of 1.22 mm × 1.22 mm × 1.22 mm. The time-frame sequence was 8 × 2, 4 × 3, 2 × 4, 1 × 5, 1 × 6, and 1 × 8 minutes. Dynamic image data were first corrected for frame-by-frame motion and between-scan misalignment using the realign function in SPM8 (Wellcome Trust Centre for Neuroimaging, London, UK). Second, anatomic T1-MR images were coregistered to a PET sum image, which was followed by MR tissue segmentation and spatial normalization using the unified segmentation algorithm in SPM8 (Wellcome Trust Centre for Neuroimaging).²⁴ Unified segmentation yielded tissue probability maps for gray and white matter and deformation fields for mapping individual images into Montreal Neurological Institute space and, conversely, for mapping images in Montreal Neurological Institute space into individual space.

Motion Data

An individually fitted thermoplastic mask was used for each subject to minimize head movements. To monitor head movements, an external position detector (Polaris Vicra, Northern Digital, Waterloo, Ontario, Canada) was used. The Polaris Vicra system (Northern Digital) tracks the position of passive infrared reflectors placed on top of the mask. The position of the reflectors was averaged over the transmission scans, which were used as reference positions for later tracking. External-motion estimates were obtained by calculating the position difference between the reference and a given time point. Points tracked by the Polaris system (Northern Digital) were located at the forehead close to the central line. Frame-to-frame motion estimates were obtained using the realign function of Statistical Parametric Mapping, version 8 (SPM8, revision 4290, Wellcome Trust Centre for Neuroimaging), which applies a mutual information maximization algorithm for rigid image registration. The frame at 4 to 6 minutes was used as a reference to obtain low within-frame motion and high-count statistics. Each frame was registered to the reference frame and a transformation matrix was used to generate motion-corrected image sequences, and to calculate displacement of a reference point at the forehead close to the central line. Frame-to-frame displacement of the reference point within SPM8 (Wellcome Trust Centre for Neuroimaging) was compared with average displacement of the Polaris (Northern Digital) reference point with very good agreement. Thus, frame-to-frame motion estimates from SPM8 (Wellcome Trust Centre for Neuroimaging) were used to exclude suspect data from final analysis. We considered displacements above 2.0 mm from the reference position to be critical, and if this threshold was exceeded in more than 10% of all frames in either of the scans, the subject was excluded from test–retest analysis.

Regions of Interest

Subcortical ROIs and reference region were drawn using in-house software (Carimas 2.8, Turku PET Centre, Turku, Finland) on coregistered magnetic resonance imagings. Thus, an identical set of ROIs was applied for both [¹¹C]raclopride scans of the same subject. Regions of interest were defined in coronal plane for the caudate nucleus, putamen, and ventral striatum (VST). The caudate nucleus and putamen were divided anatomically into precommissural and postcommissural sections. The VST was defined according to Mawlawi et al.¹⁸ The thalamus and cerebellum were drawn on transaxial slices. Neocortical ROIs were created using FreeSurfer software (version 5.3; http://surfer.nmr.mgh.harvard.edu/), which can be used to produce ROIs for PET data analysis with high anatomic accuracy.²⁵ The automatic processing of individual T1-weighted magnetic resonance images with FreeSurfer includes removal of nonbrain tissue, segmentation of subcortical structures, automated Talairach transformation, intensity normalization, tessellation of the gray–white matter boundary, automated topology correction, and surface deformation after intensity gradients to optimally place the tissue boundaries at the location where the greatest shift in intensity defines the transition to the other tissue class (more details in http://surfer.nmr.mgh.harvard.edu/). The results of the automatic processing were visually inspected and manually corrected if needed. Cortical gray matter was parcellated with a probabilistic labeling algorithm by inflating the gray–white matter boundary with overlaying curvature information of the inflated surface.²⁶ Neocortical ROIs used in the present study included anterior cingulate, superior frontal gyrus, inferior frontal gyrus, dorsolateral prefrontal cortex, orbitofrontal cortex, and temporal cortex.

Quantification of [¹¹C]Raclopride Binding

Specific [¹¹C]raclopride binding to D_2/3 receptors was estimated using the simplified reference tissue compartmental modeling (SRTM)²⁷ at both ROI and voxel levels. As reference tissue, we selected the cerebellum that is devoid of D_2/3 receptors.² Simplified reference tissue compartmental modeling was used to obtain BP_ND values to estimate specific binding.²⁸ For ROI-level model fitting, we used nonlinear, weighted fitting based on actual count statistics implemented in an in-house software (http://www.turkupetcentre.net/software/doc/fit_srtm_3_1_0.xml). For voxel-level model fitting, we used a linearized model using a basis-function approach²⁹ implemented also in in-house software (http://www.turkupetcentre.net/programs/doc/imgbfbp.html). As a comparison, we also used ratios of area under the curve (from this point onward denoted as RM=ratio method) from 36 to 55 minutes to calculate BP_ND values representing specific binding of [¹¹C]raclopride during equilibrium condition. In this method, radioactivity concentration ratios of the target region and cerebellar reference region are calculated using the formula BP_ND=target−reference/reference.³⁰

Statistical Analyses

Test–retest variability (VAR%) was calculated as follows:

where T is the first (test) estimate and RT is the second (retest) estimate. Test–retest consistency was measured with an intraclass correlation coefficient (ICC). The ICC can have a value between −1 and 1. Values closer to 1 indicate that most of the variance is because of between-subject rather than within-subject variation, which indicates good reliability. Values below 0 indicate greater within-subject than between-subject variation indicating poor reliability. Paired two-tailed t-tests were applied to the injected doses, specific radioactivities, injected masses, and BP_ND estimates between the test and the retest scans. Statistical analyses were applied to BP_ND estimates at both ROI and voxel levels. Voxel-level analyses were performed by spatially normalizing individual parametric images into Montreal Neurological Institute space, using the unified segmentation algorithm in SPM8 (Wellcome Trust Centre for Neuroimaging).

Results

Regional [¹¹C]raclopride BP_ND values and their test–retest characteristics are presented in Table 1. BP_ND values are expressed as the mean±s.d. In the striatum, [¹¹C]raclopride BP_ND values were high in all subregions, ranging from 3.48±0.27 (VST) to 4.55±0.37 (posterior putamen). In the thalamus, the mean BP_ND was 0.70±0.06 and in the cortical areas BP_ND values ranged from 0.30±0.06 (superior frontal gyrus) to 0.38±0.07 (dorsolateral prefrontal cortex). There were no reliable differences in [¹¹C]raclopride binding between scans in any of the regions measured (P=0.17 to 0.97). No noticeable lateralization was observed in the striatal regions (P=0.17 to 0.98 for test and P=0.44 to 0.94 for retest). In the cortex, there was statistically lower BP_ND in the left side of superior frontal gyrus and caudal frontal middle cortex (data not shown). Absolute variability and ICC values indicated very good reproducibility throughout the striatum (3.5%, 0.88 to 9.5%, 0.54). Thalamic VAR and ICC was also very good, 3.7% and 0.92, respectively. Cortical areas showed good to moderate reproducibility, with VAR and ICC ranging from 6.1%, 0.79 (temporal cortex) to 13.1%, 0.67 (superior frontal gyrus). Maps of voxel-level VAR and ICC are shown in Figures 1A and 1B, respectively. Table 2 shows regional [¹¹C]raclopride BP_ND values and their test–retest characteristics using RM-based modeling. There were no statistically significant (P=0.28 to 0.98) differences between the VAR values obtained using SRTM and RM in any studied regions.

Table 1. SRTM-based regional [¹¹C]raclopride BP_ND values and their test–retest characteristics (N=7, SRTM 0 to 55 minutes).

Region	Scan 1 (BPND)		Scan 2 (BPND)		Both scans (BPND)		VAR (%)	Diff (%)	Between scans
	Mean±s.d.	COV (%)	Mean±s.d.	COV (%)	Mean±s.d.	COV (%)	Mean±s.d.	Range	T-test	ICC
Putamen	4.47±0.40	8.86	4.41±0.29	6.58	4.44±0.34	7.55	3.93±2.41	−6.71 to 4.72	0.44	0.83
Putamen anterior	4.32±0.36	8.31	4.30±0.26	6.05	4.31±0.30	6.99	4.34±3.36	−8.19 to 6.72	0.84	0.72
Putamen posterior	4.60±0.42	9.21	4.50±0.33	7.32	4.55±0.37	8.09	3.53±2.16	−5.67 to 2.96	0.19	0.88
Caudate nucleus	4.07±0.38	9.30	4.07±0.29	7.10	4.07±0.32	7.95	4.52±2.64	−6.39 to 7.7	0.97	0.82
Caudate nucleus anterior	4.10±0.39	9.48	4.10±0.30	7.29	4.10±0.33	8.12	4.15±3.11	−6.61 to 7.71	1.00	0.84
Caudate nucleus posterior	3.70±0.53	14.22	3.59±0.39	11.00	3.64±0.45	12.38	9.48±7.11	−22.14 to 10.94	0.53	0.54
Ventral striatum	3.52±0.28	7.87	3.45±0.29	8.29	3.48±0.27	7.84	3.92±3.26	−9.58 to 6.57	0.29	0.82
Thalamus	0.69±0.07	9.81	0.70±0.06	8.96	0.70±0.06	9.03	3.67±1.25	−4.25 to 5.47	0.59	0.92
DLPFC	0.38±0.07	19.82	0.38±0.07	18.74	0.38±0.07	18.54	8.76±4.87	−12.62 to 16.97	0.76	0.86
Anterior cingulate	0.36±0.06	15.12	0.37±0.06	16.84	0.37±0.06	15.46	7.36±7.91	−16.37 to 19.4	0.48	0.84
Inferior frontal gyrus	0.32±0.04	13.79	0.35±0.05	12.92	0.34±0.05	13.46	9.68±7.75	−4.3 to 20.12	0.07	0.64
Superior frontal gyrus	0.30±0.05	17.91	0.31±0.06	20.94	0.30±0.06	18.77	13.06±7.39	−15.46 to 30.25	0.78	0.67
OFC	0.31±0.04	13.49	0.33±0.05	15.62	0.32±0.05	14.22	7.48±4.25	−12.53 to 13.23	0.21	0.86
Temporal cortex	0.32±0.03	9.23	0.34±0.04	10.80	0.33±0.03	10.01	6.09±3.13	−4.81 to 10.65	0.05	0.79

BP_ND, binding potential according to the simplified reference tissue model; COV, coefficient of variation; Diff, relative difference of means; DLPFC, dorsolateral prefrontal cortex; ICC, intraclass correlation coefficient; OFC, orbitofrontal cortex; SRTM, simplified reference tissue compartmental modeling; VAR, absolute variability.

Figure 1.

(A) Map of voxel-level VAR (absolute variability %) where the (red to yellow) color scale denote to variability of 0% to 30%. Values closer to 0% (red) indicate small variability and closer to 30% (yellow) indicate greater variability between test and retest scans. (B) Map of voxel-level ICC (intraclass correlation coefficient) where the (red to yellow) color scale represent ICC value of 0 to 1. Values closer to 1 (yellow) indicate that most of the variance is because of between-subject rather than within-subject variation, which indicates good reliability. Negative values indicate vice versa (not shown).

Table 2. RM-based regional [¹¹C]raclopride BP_ND values and their test–retest characteristics (N=7, RM 36 to 55 minutes).

Region	Scan 1 (BPND)		Scan 2 (BPND)		Both scans (BPND)		VAR (%)	Diff (%)	Between scans
	Mean±s.d.	COV (%)	Mean±s.d.	COV (%)	Mean±s.d.	COV (%)	Mean	Range	T-test	ICC
Putamen	4.36±0.42	9.64	4.22±0.28	6.59	4.29±0.35	8.18	5.06±4.82	−11.9 to 3.47	0.22	0.65
Putamen anterior	4.24±0.38	9.05	4.13±0.25	5.98	4.18±0.32	7.54	5.2±5.28	−12.0 to 4.91	0.38	0.54
Putamen posterior	4.47±0.45	10.02	4.30±0.31	7.23	4.39±0.38	8.69	4.84±4.48	−11.64 to 1.95	0.12	0.72
Caudate nucleus	3.99±0.40	10.03	3.90±0.26	6.78	3.95±0.33	8.33	3.94±3.37	−8.34 to 4.58	0.30	0.82
Caudate nucleus anterior	4.01±0.41	10.12	3.93±0.28	7.08	3.97±0.34	8.49	3.84±3.13	−7.84 to 4.58	0.30	0.84
Caudate nucleus posterior	3.62±0.58	15.91	3.45±0.40	11.49	3.54±0.48	13.68	7.3±7.61	−20.61 to 4.22	0.27	0.70
Ventral striatum	3.43±0.27	7.92	3.31±0.26	7.76	3.37±0.26	7.77	4.88±2.52	−9.24 to 4.51	0.06	0.78
Thalamus	0.69±0.07	9.83	0.70±0.07	9.59	0.70±0.06	9.34	8.11±6.7	−9.12 to 24.15	0.81	0.37
DLPFC	0.39±0.08	19.88	0.38±0.07	18.32	0.38±0.07	18.37	10.60±7.02	−14.03 to 26.4	0.97	0.75
Anterior cingulate	0.38±0.05	11.97	0.37±0.07	18.36	0.37±0.06	14.89	8.69±8.16	−20.05 to 7.36	0.46	0.76
Inferior frontal gyrus	0.34±0.05	14.39	0.35±0.04	10.13	0.35±0.04	11.96	9.03±4.59	−11.37 to 15.52	0.62	0.66
Superior frontal gyrus	0.31±0.05	17.62	0.30±0.06	20.94	0.31±0.06	18.59	13.32±12.05	−15.8 to 44.02	0.98	0.50
OFC	0.33±0.05	14.55	0.33±0.05	15.91	0.33±0.05	14.65	5.92±7.05	−18.2 to 8.83	0.98	0.88
Temporal cortex	0.33±0.03	10.36	0.34±0.03	9.63	0.34±0.03	9.65	5.28±3.22	−5.69 to 12.45	0.43	0.83

BP_ND, binding potential according to the simplified reference tissue model; COV, coefficient of variation; Diff, relative difference of means; DLPFC, dorsolateral prefrontal cortex; ICC, intraclass correlation coefficient; OFC, orbitofrontal cortex; VAR, absolute variability.

Discussion

We explored the long-term test–retest reliability of striatal, thalamic, and neocortical DA D_2/3 receptor binding using the B/I method with [¹¹C]raclopride and the HRRT scanner. A cohort of seven male subjects took part in an ongoing study on the effects of working-memory training on striatal DA release and they were scanned twice with a 5-week interval while they performed same, very simple computerized letter-recall task during both PET measurements. They were selected for the present test–retest analyses based on optimal motion data. The analyses were performed using SRTM modeling and by calculating ratios of areas under the curve (RM; see Materials and Methods section). In a previous test–retest study during rest, we showed that the use of the HRRT scanner increased the signal from the striatum and thalamus, resulting in higher BP_ND values compared with previous studies using scanners with lower resolution.¹⁸ In the current study, we found that also in the regions characterized with low densities of D_2/3 receptors such as the neocortex, the measurement of [¹¹C]raclopride binding turned out to be fairly robust. Protocol optimization by homogenizing the scanning conditions (simple cognitive task) and by eliminating the effect of movement may have contributed to this positive outcome. The rank order of observed BP_ND values was in line with the known density of both striatal and extrastriatal D_2/3 receptors.¹⁵ Moreover, the rostrocaudal trend of striatal [¹¹C]raclopride binding detected in our previous study was replicated, as the highest BP_ND values were observed in the posterior putamen and in the anterior caudate nucleus.²⁰

The reproducibility of a single bolus of [¹¹C]raclopride binding in human striatum has been evaluated in various studies using different time intervals.^{16, 17, 18} In the earlier studies, the striatum was evaluated as a whole structure¹⁶ and later subdivided into the putamen and caudate nucleus.¹⁹ Mawlawi et al¹⁸ were the first group to use B/I method and measure VST separately. Moreover, most experiments have examined short-term test–retest reliability with scans performed within 24 hours. There is one study on long-term stability of [¹¹C]raclopride binding, where two sets of scan pairs (intervals of 3 to 7 weeks and 6 to 11 months) were used.¹⁶ They reported striatal VAR of 5.5% with a 3- to 7-week interval, and when the interval was prolonged to 6 to 11 months, VAR increased to 10.4%. In the present study, the VAR and ICC measured for the whole striatum were excellent: 3.9% and 0.86, respectively. Substriatal outcome was also very good, ranging from 3.5%, 0.88 (posterior putamen) to 9.5%, 0.54 (posterior caudate nucleus). It should be noted that the volume of the postcommissural caudate is rather small, causing an increased partial volume effect and noise, both of which hamper reproducibility. Thalamic VAR and ICC values were 3.7%, 0.92, respectively, and thus, slightly better than in our previous test–retest study¹⁸ (4.59%, 0.66 medial part and 6.69%, 0.44 lateral part) and clearly superior to the test–retest results obtained by Mawlawi et al¹⁸ using B/I method (17.4%, 0.39). Altogether, our results are in good agreement with previous test–retest reports and indicate good long-term reliability of [¹¹C]raclopride binding in human striatum, further validating the use of [¹¹C]raclopride in imaging D_2/3 receptor functions.

The density of D_2/3 receptors in extrastriatal regions is only 2% to 8% of that in the striatum.³¹ Because [¹¹C]raclopride has moderate in vivo affinity and a relatively low signal-to-noise ratio (affinity compared with nonspecific binding) in low D_2/3 receptor density areas such as the cortex, [¹¹C]raclopride is not ideal to quantify D_2/3 receptor availability in those areas.⁸ With the introduction of high-affinity radioligands such as [¹¹C]FLB 457 and [¹⁸F]fallypride, it has become possible to visualize and quantify extrastriatal D_2/3 receptors.^{32, 33} However, these ligands have some disadvantages. [¹¹C]FLB 457 clears from the striatum much more slowly than [¹¹C]raclopride so that washout of the ligand from the striatum is too slow for quantitative imaging when coupled with the rapid decay rate of C-11.³⁴ Thus, [¹¹C]FLB 457 can only be used for imaging extrastriatal regions. With [¹⁸F]fallypride, it is possible to provide simultaneous quantitative measures of D_2/3 receptor binding in the striatum and in extrastriatal regions by prolonging the scanning session, but at the same time it is impossible to perform two scans within the same day because of the long decay of F-18. This fact may increase within-subject variability and decrease the chances to detect subtle changes in BP_ND.

Although [¹¹C]raclopride is suboptimal for measuring extrastriatal DA transmission, there is evidence that decreases in [¹¹C]raclopride BP_ND can be observed in extrastriatal regions after drug or behavioral challenges.^{9, 10, 11} Changes in extrastriatal [¹¹C]raclopride binding have also been associated with brain pathologic assesment, such as Parkinson's, Huntington's, and Alzheimer's diseases.^{12, 13, 14} Moreover, a recent study using diffusion-weighted magnetic resonance imaging and PET found that connectivity-based subdivision of human striatum improved the evaluation of regional differences in DA transmission, supporting an association between striatal DA release and cortical connectivity.³⁵ Thus, as major neurodegenerative disorders have been associated with alterations in corticostriatal circuits, imaging DA transmission simultaneously in striatal and extrastriatal regions is a major advantage. Stokes et al¹¹ estimated reproducibility of cortical [¹¹C]raclopride binding from a separate cohort of subjects (controls) where Δ9-tetrahydrocannabinol was administered to quantify possible alterations of DA transmission in the human cortex. Region of interest analyses including right middle frontal gyrus, left superior frontal gyrus, and left superior temporal gyrus yielded relatively poor VAR values ranging from 18% to 24%. With high-affinity ligands such as [¹⁸F]fallypride and [¹¹C]FLB 457 cortical test–retest reliability has been reported to be ~10% and 5.3% to 10.4%, respectively.^{36, 37} In the current study, cortical VAR and ICC values were good to moderate, ranging from 6.1%, 0.79 (temporal cortex) to 13.1%, 0.67 (superior frontal gyrus). This is clearly superior to previous results by Stokes et al¹¹ and comparable to test–retest parameters obtained with high-affinity ligands.^{36, 37}

In addition to SRTM²⁷ estimation of the BP_ND values, we also calculated ratios (=RM, as described earlier in the Materials and Methods section) of area under the curve from 36 to 55 minutes. RM is considered to be a gold standard in activation-type studies when using a B/I protocol.³⁰ The validity of both SRTM and RM modeling using B/I of [¹¹C]raclopride has previously been evaluated by Mawlawi et al,¹⁸ and they found SRTM and RM (denoted in their paper as ‘equilibrium V3') to yield very similar values. In their test–retest analyses SRTM provided slightly but significantly better reproducibility in the caudate nucleus and putamen and equivalent reproducibility in VST and the thalamus. In our study, SRTM- and RM-based BP_ND showed rather equal reproducibility of [¹¹C]raclopride binding (see Tables 1 and 2, respectively) and no statistically significant differences in VAR values were observed in any regions measured (P=0.28 to 0.98).

The PET method is very sensitive to motion during image acquisition and greatly biases BP estimates.³⁸ Thus, the optimization of a study starts with careful planning of the setup, including optimal head support during data acquisition, which is especially critical when studying small structures or areas with a low signal-to-noise ratio. At the stage of analysis, if data are found to be affected by movement, all necessary steps to correct the bias should be taken. There is no unambiguous threshold for the amount of motion that could be tolerable during a PET scan. However, it is conceivable that an amount of motion that is less than the scanner resolution may be regarded as tolerable. The high-resolution scanner used in the present study is characterized by an isotropic 2.5-mm intrinsic spatial resolution. Thus, we considered displacements of >2.0 mm from the reference position to be critical, and if this threshold was exceeded in more than 10% of all frames in either scan, the subject was excluded from the test–retest analysis. By using this rather rigorous criterion for exclusion, we sought to ensure that head movements were not hampering the analysis of cortical D_2/3 receptor binding, which is known to be challenging using [¹¹C]raclopride. In light of our results, this optimization appears to have been beneficial.

Subjects who participated in this study were initially recruited with the goal of studying the impact of working-memory training on DA transmission. The seven male subjects who were chosen for test–retest analyses were from the control group, so during the 5-week retest interval they did not practice any cognitive task. Thus, the scan pairs were identical as the same extremely simple letter-recall task was administered, and the same scan protocol was used during both measurements. We considered this to be a better alternative than rest, because functional magnetic resonance imaging studies have showed that the rest condition presumably might give subjects an opportunity for day dreaming, self-reflection, problem solving, and other uncontrolled mental activities that may affect the measurement.³⁹ Conceivably, the fact that our subjects focused their attention on the same simple cognitive task served to homogenize their DA transmission, increasing the detectability of radioligand binding. It may also have helped us to achieve good reproducibility of inherently weaker cortical [¹¹C]raclopride D_2/3 receptor binding. In our previous test–retest study, we showed that the use of the HRRT scanner increased the signal from the striatum and thalamus, resulting in higher BP_ND values compared with previous studies using scanners with lower resolution.¹⁹ Stokes et al¹¹ used a lower resolution (~5 mm) whole-body PET scanner (ECAT HR+ 962 scanner (CTI/Siemens Medical Solutions, Knoxville, TN, USA)) and reported cortical BP_ND values of 0.18 (right middle frontal gyrus) to 0.23 (left superior temporal gyrus), whereas the present study using the HRRT scanner yielded superior values in the range of 0.30 (superior frontal gyrus) to 0.38 (dorsolateral prefrontal cortex). Thus, the higher BP_ND estimates enabled by the superior resolution of the HRRT scanner might also add to the good cortical reproducibility obtained in this study.

Some potential limitations of the study should be noted. First, the sample size was small (N=7). The subjects were selected for the test–retest analyses based on optimal motion data to ensure that head movements were not hampering the analysis of cortical D_2/3 receptor binding, which is known to be challenging using [¹¹C]raclopride. Despite the limited number of subjects, our results indisputably show that with protocol optimization it is possible to improve the reliability of cortical measurement of [¹¹C]raclopride binding, as discussed above. Second, [¹¹C]raclopride has moderate in vivo affinity and a relatively low signal-to-noise ratio in low D_2/3 receptor density areas such as the cortex.¹¹ The fact that [¹¹C]raclopride binding to extrastriatal D_2/3 receptors along with acceptable test–retest reliability observed in this study might appear puzzling given previous failures to convincingly show [¹¹C]raclopride binding outside the striatum. Especially, it was found that in the temporal and frontal cortices the distribution ratios were, however, only a few percent higher for [¹¹C]raclopride than for [¹¹C]FLB 472 (the inactive enantiomer of raclopride), indicating that if D_2/3 DA receptors are present in the human neocortex, then their density is quite low.⁸ Having said this, one has to remember that during the last almost 30 years the sensitivity and resolution of the scanners have improved considerably. The ability of [¹¹C]raclopride to reflect the true levels of specific binding to D_2/3 receptors in the cortical areas is not possible to assess based on the present results. The amount of nonspecific binding of [¹¹C]raclopride in the extrastriatal areas has not yet been fully assessed and we suggest that to further investigate if [¹¹C]raclopride truly could be used in exploring extrastriatal areas, displacement studies using, e.g., haloperidol, would be highly beneficial.

Conclusions

We examined long-term reliability of [¹¹C]raclopride as a DA D_2/3 receptor marker in the human striatum, thalamus, and neocortex. Our results are in good agreement with past research and show the feasibility of the B/I method and good long-term stability of [¹¹C]raclopride BP_ND estimates. To our knowledge, this is the first test–retest study on cortical D_2/3 receptor binding using [¹¹C]raclopride and suggests that extrastriatal D_2/3 binding might be possible to study with this ligand although further validation is first needed. Careful planning of the study protocol and well-optimized conditions during scanning, e.g., reducing subject motion, could improve the possibilities to evaluate interactions between striatal and extrastriatal DA transmission using [¹¹C]raclopride.

The authors declare no conflict of interest.