Reference region modeling approaches for amphetamine challenge studies with [¹¹C]FLB 457 and PET

Abstract

Detecting fluctuations in synaptic dopamine levels in extrastriatal brain regions with [¹¹C]FLB 457 and positron emission tomography (PET) is a valuable tool for studying dopaminergic dysfunction in psychiatric disorders. The evaluation of reference region modeling approaches would eliminate the need to obtain arterial input function data. Our goal was to explore the use of reference region models to estimate amphetamine-induced changes in [¹¹C]FLB 457 dopamine D2/D3 binding. Six healthy tobacco smokers were imaged with [¹¹C]FLB 457 at baseline and at 3 hours after amphetamine (0.4  to 0.5 mg/kg, per os) administration. Simplified reference tissue models, SRTM and SRTM2, were evaluated against the 2-tissue compartmental model (2TC) to estimate [¹¹C]FLB 457 binding in extrastriatal regions of interest (ROIs), using the cerebellum as a reference region. No changes in distribution volume were observed in the cerebellum between scan conditions. SRTM and SRTM2 underestimated binding, compared with 2TC, in ROIs by 26% and 9%, respectively, with consistent bias between the baseline and postamphetamine scans. Postamphetamine, [¹¹C]FLB 457 binding significantly decreased across several brain regions as measured with SRTM and SRTM2; no significant change was detected with 2TC. These data support the sensitivity of [¹¹C]FLB 457 for measuring amphetamine-induced dopamine release in extrastriatal regions with SRTM and SRTM2.

Introduction

Nicotine and other drugs of abuse exert their rewarding, reinforcing, and motivational effects through the mesolimbic and mesocortical dopamine systems.¹ Striatal dopamine changes have been reliably detected using positron emission tomography (PET) with dopamine D2/D3 radiotracers, such as [¹¹C]PHNO and [¹¹C]raclopride, after an amphetamine challenge.^{2, 3, 4} The ability to measure changes in extrastriatal dopamine levels, including thalamus, amygdala, prefrontal cortex, as well as other cortical areas, is useful for further advancing our understanding of the neuronal mechanisms that underlie addictive disorders.⁵

The high-affinity radiotracer [¹¹C]FLB 457, 5-bromo-N-[[(2S)-1-ethyl-2-pyrrolidinyl]methyl]-3-methoxy-2-(methoxy-¹¹C) benzamide has shown sensitivity and reliability for measuring amphetamine-induced dopamine release at extrastriatal dopamine D2/D3 receptors when adhering to scan constraints, such as mass limits < ~0.6 μg and collection of emission data for at least 90 minutes after radiotracer injection.^{6, 7} Another proviso is that the quantification of amphetamine-induced changes using reference tissue-input based models may lead to an underestimation in non-displaceable binding potential (BP_ND),⁸ as compared with arterial-input based models. The attractive feature of reference tissue modeling approaches is the circumvention of collecting arterial blood samples, a process that is invasive, may cause the subject discomfort, and is not always attainable. However, for reference region modeling the reference region should be validated for a given radiotracer with a displacement study to ensure the lack of specific binding. For radiotracers that target dopamine D2/D3 receptors using radioligands with moderate affinity, the cerebellum is typically used as the reference region due to a negligible amount of D2/D3 receptors that can be displaced (i.e., non-displaceable=free and nonspecifically bound).⁹

Some studies with [¹¹C]FLB 457 have reported a reduction in cerebellum distribution volume, V_T(CER), postD2 antagonist or postpsychostimulant challenge, suggesting specific binding in the cerebellum.^{10, 11, 12} A more recent [¹¹C]FLB 457 blocking study with aripiprazole (D2/D3 partial agonist) was performed to evaluate the fractional contribution to specific D2/D3 binding in the cerebellum, in addition to the pons and centrum semiovale (CESVL), to assess potential reference regions.¹³ The change in V_T before and after aripiprazole was lower in the CESVL (−3%) and the pons (−10%), compared with the cerebellum (−17%). However, a reevaluation of previous data yielded lower variability with test-retest and amphetamine-induced changes in BP_ND and greater sensitivity to detect amphetamine-induced DA release across regions when V_T(CER) was used to estimate the nondisplaceable distribution volume, V_ND, compared with V_T(PONS) and V_T(CESVL). Other studies have supported reference region modeling approaches for the analysis of [¹¹C]FLB 457 BP_ND for low D2/D3 density regions,^{14, 15, 16, 17} but the degree of underestimation with amphetamine challenge studies and the sensitivity for detecting amphetamine-induced dopamine release, as compared with arterial input model methods, has not been examined.

The aim of this study was to validate the use of reference region models for detecting changes in extrastriatal dopamine release because obtaining an arterial input function in human subjects is not always feasible. Our approach was to evaluate the simplified reference tissue models (SRTM), SRTM and SRTM2, with the 2-tissue compartmental model (2TC), for the analysis of [¹¹C]FLB 457 BP_ND before and after an amphetamine challenge in tobacco smokers. The change in cerebellum V_T was assessed, and the cerebellum was used to estimate BP_ND for each of the models across extrastriatal regions.

Materials and methods

Human Subjects

Positron emission tomography scans were performed in six tobacco smokers (4 male and 2 female, 39.5±6.8 years old, 83.8 kg±13.7 kg). Written informed consent was obtained from all subjects. Subjects were medically and mentally healthy and were screened for the following: no current or history of medical illnesses, no prescription or illicit drug use, and based on psychiatric assessments (e.g., SCID and DSMIV Axis I) and mood measures (e.g., depression, anxiety, and impulsivity). Tobacco smoking inclusion criteria included: Fagerstrom Test for Nicotine Dependence of at least 3, smoking cigarettes daily for at least 1 year, and during intake evaluation, carbon monoxide levels greater than 8 p.p.m., plasma nicotine and cotinine levels greater than 10 and 50 ng/mL, respectively. Across subjects, cigarettes smoked per day and number of years smoked were 13.3±5.2 and 20.8±4.2, respectively. Females had negative pregnancy tests at intake and on the day of the scan. Approval for the study protocol was obtained from The Human Investigation Committee, Yale University School of Medicine, and Yale-New Haven Hospital Radiation Safety. The conducted study adhered to the Protection of Human Subjects of Research and Ethical Principle and Guidelines.

Each subject participated in two [¹¹C]FLB 457 PET scans on the same day, one baseline scan and one scan 3 hours after amphetamine administration. On a separate day, each subject had one MR scan, required to delineate anatomic information from the PET data.

Radiosynthesis of [¹¹C]FLB 457

[¹¹C]CO₂ was produced by a 16.5-MeV GE PETtrace cyclotron with 60 μA irradiation of a nitrogen target for 40 minutes. The [¹¹C]CO₂ was first converted to [¹¹C]methyl iodide then converted to [¹¹C]methyl triflate with either GE FXC Pro or Upgrade synthesis modules. The [¹¹C]methyl triflate, under helium stream, was bubbled into a solution of FLB 604 (0.3  to 0.6 mg), 5N NaOH (8 μL), and acetone (400 μL). The reaction proceeded for 5 minutes at room temperature, then the solution was diluted with 1 mL deionized water and injected onto a reverse-phase HPLC column (Phenomenex Prodigy ODS, 250 × 10 mm, 10 μm particle size). Using a mobile phase of 25% acetonitrile, 75% 0.1 mol/L ammonium formate, containing 0.03% ascorbic acid (pH 4.2) at a flow rate of 5 mL/min, the radioligand eluted and was collected after 16  to 17 minutes. The collected fraction was diluted with 50 mL deionized water containing 400 mg ascorbic acid. The diluted product was trapped on a Waters C18 Sep-Pak (Milford, MA, USA) and washed with 10 mL deionized water containing 10 mg ascorbic acid. The final product was eluted from the Sep-Pak using 1.0 mL of absolute ethanol (USP) and diluted with 10 mL of 0.9% sodium chloride (for injection, USP). The average specific activity was 978.0 ±473.9 MBq/nmol (26.4 ±12.8 mCi/nmol, n=12) at end-of-synthesis, with chemical and radiochemical purities of ⩾93% and ⩾97%, respectively.

Input Function and Free Fraction Measurements

The arterial input functions were collected for all scans and were corrected for the presence of radiometabolites. For the first 7 minutes after injection, an automated blood counter (PBS-101; Veenstra Instruments, Joure, The Netherlands) with a peristaltic pump at a rate of 4 mL/min was used to measure the radioactivity continuously in whole blood. Manual sequential blood samples were also drawn (2  to 10 mL) at 3, 5, 7 , 10, 15, 20, 30, 50, 50, 60, 70, and 90 minutes after injection. The radioactivities of manual whole blood and plasma obtained from each corresponding sample via centrifugation (2,930 g at 4°C for 5 minutes) samples were counted in a cross-calibrated gamma counter (1480 WIZARD; Perkin-Elmer, Waltham, MA, USA). The plasma time-activity curve (TAC) was merged and extrapolated from these two sets of data. The end of the total plasma curve was fitted to a sum of exponentials to reduce noise in the input function.

Radiometabolites were measured in the plasma from arterial blood samples collected at 5, 15, 30, 60, and 90 minutes after injection using the automatic column-switching HPLC method¹⁸ to determine the parent fraction. Plasma samples were initially treated with urea (8 mol/L), and then loaded onto a capture column (19 mm × 4.6 mm) packed with Phenomenex SPE C18 Strata-X sorbent and eluted with 1% acetonitrile in water at 2 mL/min. At 4 minutes, the activity trapped on the capture column was back-flushed onto an analytical HPLC column (Phenomenex Luna Phenyl hexyl, Torrance, CA, USA; 5 μm, 250 mm × 4.6 mm) eluted with 34% acetonitrile in 0.1 mol/L ammonium formate at 1.70 mL/min. An automated fraction collected the HPLC eluent and was counted in the gamma well counter. The parent fraction (retention time of ~10.5 minutes) was determined as the ratio of the sum of radioactivity containing the parent to the total amount of radioactivity. Parent fraction data were fitted with an inverted gamma function for five subjects and with a bounded sum of exponentials for one subject, as determined by the quality of fit. The arterial input function was calculated as the product of the fitted total plasma curve and the fitted parent fraction curve.

To measure the unbound portion, or free fraction (f_P), of [¹¹C]FLB 457, an ultrafiltration-based method was used. [¹¹C]FLB 457 (~7.4 MBq) was mixed with arterial blood (6 mL) drawn immediately before tracer injection. After 10 minutes at room temperature, the spiked blood sample was centrifuged at 2,930 g for 5 minutes to separate the plasma. Plasma aliquots (0.3 mL) were loaded onto the reservoir of the EMD Millipore Centrifree ultrafiltration device (Billerica, MA, USA) in triplicate and centrifuged at 1,228 g for 20 minutes. Free fraction was determined by calculating the ratio of the radioactivity concentration in the ultrafiltrate to the total activity in plasma. The amount of nonspecific binding of [¹¹C]FLB 457 to the filter was determined, as described above, by spiking the sample of saline with [¹¹C]FLB 457. The ultrafiltrate to spiked saline ratio was 97.51±0.86 (n=12), indicating negligible filter retention.

Amphetamine Administration and Plasma Levels

Amphetamine (0.4 to 0.5 mg/kg, PO) was administered 3 hours before the second [¹¹C]FLB 457 injection. Blood samples were collected to measure plasma amphetamine levels before amphetamine administration, t=−180 minutes, and at −120, −60, 0, 45, and 90 minutes, relative to the start time of the second [¹¹C]FLB 457 injection (t=0 minute).

Positron Emission Tomography Scans and Image Reconstruction

Positron emission tomography scans were performed on the ECAT EXACT HR+ (Siemens/CTI, Knoxville, TN, USA). Before each radiotracer administration, a 6-minute transmission scan was acquired, necessary for attenuation correction of the PET emission data. [¹¹C]FLB 457 was injected intravenously as a bolus over 1 minute by a computer-controlled pump (Harvard Apparatus, Holliston, MA, USA), and emission data were collected for 90 minutes.

Emission data were collected and sinograms were reconstructed with all corrections (attenuation, normalization, scatter, randoms, and deadtime) into a sequence of 27 frames: 6 × 30 seconds; 3 × 1 minutes; 2 × 2 minutes; 16 × 5 minutes. Final image dimension and voxel size were 128 mm³ × 128 mm³ × 63 mm³ and 2.06 mm³ × 2.06 mm³ × 2.43 mm³, respectively. Motion correction on the dynamic data was performed by registering each frame to an early frame (i.e., the first 10 minutes after injection) using a 6-parameter mutual information algorithm (FMRIB's Linear Image Registration Tool, FMRIB Software library, version 3.2).¹⁹

Magnetic Resonance Scanning and Processing

T1-weighted magnetic resonance (MR) images were acquired on a 3T Trio whole-body scanner (Siemens Medical Systems, Erlangen, Germany) with a circularly polarized head coil. The final MR image dimension and pixel size were 256 mm³ × 256 mm³ × 176 mm³ and 0.98 mm³ × 0.98 mm³ × 1.0 mm³, respectively. Postprocessing of the MR images included a skull- and muscle stripping procedure so that only the brain remained in the image field-of-view (FMRIB's Brain Extraction Tool, http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/BET), before coregistration with the PET images.

Positron Emission Tomography Image Processing

Positron emission tomography images were aligned to the MR via a rigid registration with mutual information. Each MR image was normalized to Montreal Neurological Institute space²⁰ using an affine linear plus nonlinear registration (Bioimage Suite 2.5, http://www.bioimagesuite.org/index.html), to extract regions-of-interest (ROIs) from the automated anatomic labeling (AAL) template.²¹ The ROIs were then mapped from the AAL template to PET space via the two transformations (e.g., PET-MR and MR-AAL template) to compute TACs in the following regions: amygdala (0.71 cm³), anterior cingulate cortex (4.14 cm³), dorsolateral prefrontal cortex (11.83 cm³), hippocampus (2.86 cm³), occipital cortex (15.36 cm³), orbitofrontal cortex (11.07 cm³), parietal cortex (12.83 cm³), temporal cortex (32.85 cm³), thalamus (3.29 cm³), and ventromedial prefrontal cortex (10.21 cm³), and cerebellum (15.98 cm³). The AAL template was used to divide the prefrontal cortex into dorsolateral, orbitofrontal, and ventromedial. The dorsolateral prefrontal cortex was defined by combining the frontal superior, frontal mid, and frontal inferior triangularis corresponding to Brodmann's areas 9 and 46.²² The orbitofrontal cortex was delineated by combining the frontal superior medial orbital and frontal medial orbital regions, and the ventromedial prefrontal cortex was defined by combining the frontal superior orbital and frontal superior medial regions.

Tracer Kinetic Modeling

The 2TC,²³ SRTM,⁹ and SRTM2²⁴ were used for the kinetic analysis of regional BP_ND⁸ [¹¹C]FLB 457 TACs. For 2TC, the volume of distribution (V_T) in each ROI was estimated using the tissue TAC and the metabolite-corrected arterial input function. With the cerebellum as the reference region BP_ND(2TC) was computed as

where V_T(ROI) and V_T(CER) are the volumes of distribution in the ROI and the cerebellum region, respectively.

The cerebellum was the reference input function for SRTM and SRTM2, where BP_ND was computed as

Parameters R₁, k₂, and k₂' are estimated directly from SRTM, where k₂ (1/minute) and k₂' (1/minute) are the rate constants of tracer efflux to the blood from the tissue in the ROI and reference tissue, respectively. R₁ is the ratio of tracer influx from the blood to tissue in the ROI and reference tissue, K₁ and K₁' (mL/cm³/min), respectively. For SRTM2 only R₁ and k₂ are estimated where k₂' was shared across ROIs for each scan using coupled fits, where the TACs are fit simultaneously.²⁵

Using an approximation of noise-equivalent counts for each frame, the data were weighted in the fits.²⁶ The models were evaluated in terms of quality of fit to the data, and the F-test (P<0.05) was used to determine whether the ROI TAC fits across scans were statistically different between SRTM (3-parameter per ROI) and SRTM2 (2-parameter per ROI and 1 global parameter) models. BP_ND(SRTM) and BP_ND(SRTM2) were compared with BP_ND(2TC), the standard for comparison. This analysis was limited to data points for which the standard error (SE) for 2TC BP_ND(2TC) was less than 20% (%SE=SE/BP_ND × 100), to include only reliable 2TC estimates of BP_ND.

Percent change in BP_ND (%ΔBP_ND) from baseline to amphetamine challenge was computed as

Data are reported as mean±s.d. Statistical analysis within each group was performed with two-tailed, paired t-tests with P<0.05, without correction for multiple comparisons.

Results

Scan Parameters

There were no differences between baseline and postamphetamine scans in injected dose, injected mass, or specific activity. Plasma-free fraction (f_P) was significantly different between scan conditions (P<0.05) (Table 1). In the cerebellum reference region, no significant change was observed in [¹¹C]FLB 457 V_T (−8±10%) nor V_T/f_p (−1±13%) between scan conditions (Table 1). Cerebellum V_T and V_T/f_p for each subject at baseline and postamphetamine challenge appear in Supplementary Figure 1.

Table 1. Scan parameters for [¹¹C]FLB 457 (n=6 subjects).

	Baseline	Postamphetamine
Injected dose (MBq)	334±49	338±58
Specific activity (MBq/nmol)	376±178	407±172
Injected mass (μg)	0.40±0.15	0.37±0.14
Plasma-free fraction (fp)	0.33±0.02	0.30±0.02*
Cerebellum VT /fP (mL/cm3)	10.8±2.0	10.6±1.4
Cerebellum VT (mL/cm3)	3.5±0.7	3.2±0.4

Specific activity was at time of injection. There was no significant difference between scan conditions except for plasma free (*P<0.05).

Plasma Amphetamine Levels

Amphetamine levels peaked at 2 hours at 64.5 ±14.0 ng/mL, remained elevated at the start of the second scan at 62.7±12.5 ng/mL (3 hours after amphetamine), and were steady throughout the duration of the scan at 58.7±8.5 ng/mL and 56.4 ng/mL±8.6 ng/mL (3 hours 45 minutes and 4 hours 30 minutes after amphetamine, respectively) (Figure 1). In one subject amphetamine levels peaked at 1 hour at 107.8 ng/mL, while the other subjects ranged from 30  to 45 ng/mL. In the same subject, the plasma level was 82.6 ng/mL at the start of scan 2 and remained ~10  to 20 ng/mL higher than the other 5 subjects for the duration of the scan.

Figure 1.

Amphetamine levels in the plasma (pk). Times are relative to the second [¹¹C]FLB457 injection at t=0 minute. Data points are the mean across subjects (n=6), and bars represent standard deviation.

Model Comparison

Model fits

All the models (2TC, SRTM, and SRTM2) produced good fits for low binding regions (Figure 2). Fit quality was slightly poorer for SRTM2, noticeably in the amygdala. On the basis of the F-test, [F(9,240)=1.92, P<0.05], SRTM had better fits across regions than SRTM2 for all scans.

Figure 2.

Typical regional time-activity curves from one representative subject and model fits for (A) 2-tissue compartmental model (2TC), (B) simplified reference tissue model (SRTM), (C) and SRTM2.

Comparison of [ ¹¹ C]FLB 457 BP_NDwith reference region input models and 2-tissue compartmental model

SRTM and SRTM2 underestimated 2TC BP_ND for ROIs by 26% and 9%, respectively (Figure 3). The bias observed was driven mainly by higher-binding ROIs, the thalamus and amygdala. Bias across scan conditions was consistent, where baseline BP_ND(2TC)=0.74BP_ND(SRTM)+0.20, R²=0.98, and amphetamine challenge BP_ND(2TC)=0.75BP_ND(SRTM)+0.16, R²=0.98. Similarly, bias was consistent for BP_ND(SRTM2) across conditions, where baseline BP_ND(2T)=0.92BP_ND(SRTM2)+0.04, R²=0.96, and amphetamine challenge BP_ND(2T)=0.90BP_ND(SRTM2)+0.02, R²=0.98. Due to the low intercept in the aforementioned regression equations, the bias induced by SRTM2 on the estimation of %ΔBP_ND, computed as , is expected to be lower than 5% in regions where the 2TC BP_ND is larger than 0.5.

Figure 3.

Model comparison of simplified reference tissue model (SRTM) and SRTM2 versus 2-tissue compartmental model (2TC) for [¹¹C]FLB 457 BP_ND for baseline and postamphetamine conditions in all ROIs evaluated. Both (A) SRTM and (B) SRTM2 underestimated 2TC BP_ND indicated with the regression fit (solid line). The identity plot (dashed line) was added for reference. Only 2TC BP_ND estimates with standard error <20% were used in these comparisons.

Analysis of [¹¹C]FLB 457 %ΔBP_ND with 2-tissue Compartmental Model, Simplified Reference Tissue Model, and Simplified Reference Tissue Model 2

Model estimates of BP_ND and %ΔBP_ND for baseline and postamphetamine challenge scans are shown in Table 2. With the 2TC model, mean %ΔBP_ND decreased nonsignificantly from the baseline to amphetamine challenge scans in the hippocampus, temporal cortex, and thalamus. Reductions in %ΔBP_ND were observed with SRTM and SRTM2 across all regions. With SRTM and SRTM2, BP_ND was reduced significantly (P⩽0.05) postamphetamine in the amygdala, dorsolateral prefrontal cortex, hippocampus, and temporal cortex, and anterior cingulate cortex (SRTM2 only). Variability in %ΔBP_ND was smallest with SRTM2 in 6 of the 10 regions examined. For each subject, baseline and postamphetamine BP_ND values were plotted and shown in Supplementary Figure 2.

Table 2. [¹¹C]FLB457 BP _ND at baseline and postamphetamine challenge for 2TC, SRTM, and SRTM2 (mean±s.d.).

	2TC				SRTM				SRTM2
	Baseline	Challenge	ΔBPND(%)	P	Baseline	Challenge	ΔBPND(%)	P	Baseline	Challenge	ΔBPND(%)	P
Amyg	3.24±0.49	3.22±0.54	2.5±29.5	0.957	2.52±0.27	2.33±0.18	−8.9±6.4	0.021*	3.29±0.33	2.84±0.23	−13.4±7.5	0.008*
Ant. cing.	1.01±0.19	1.02±0.19	0.8±7.7	0.852	1.05±0.17	1.03±0.06	−8.5±9.1	0.060	1.00±0.18	0.94±0.21	−6.7±7.5	0.047*
Dlpfc	0.60±0.15	0.62±0.14	3.8±12.5	0.603	0.68±0.14	0.64±0.11	−11.5±8.7	0.008*	0.60±0.15	0.56±0.16	−7.3±5.2	0.003*
Hipp	1.40±0.17	1.25±0.31	−11.6±11.6	0.068	1.29±0.17	1.23±0.26	−12.9±9.7	0.016*	1.14±0.18	1.02±0.25	−11.5±9.5	0.023*
Occ	0.45±0.14	0.48±0.16	5.1±9.7	0.327	0.52±0.16	0.47±0.11	−8.0±14.7	0.223	0.45±0.16	0.42±0.18	−7.2±12.0	0.213
Ofc	0.51±0.14	0.54±0.15	10.4±39.0	0.722	0.56±0.13	0.51±0.16	−9.2±22.5	0.389	0.49±0.14	0.46±0.16	−7.1±13.7	0.375
Par	0.45±0.14	0.50±0.20	11.2±19.7	0.313	0.53±0.18	0.49±0.12	−10.1±16.3	0.282	0.46±0.19	0.46±0.22	−3.0±12.6	0.892
Temp	1.14±0.21	1.13±0.25	−1.3±7.1	0.781	1.12±0.20	1.08±0.21	−8.4±8.1	0.036*	1.10±0.21	1.02±0.26	−7.7±7.4	0.046*
Thal	3.61±0.73	3.36±0.77	−7.1±6.9	0.057	2.99±0.39	2.73±0.44	−7.5±8.5	0.084	3.26±0.39	2.97±0.65	−9.4±10.7	0.097
Vmpfc	0.51±0.17	0.54±0.14	13.6±22.3	0.211	0.58±0.15	0.58±0.06	−9.4±12.9	0.084	0.51±0.15	0.47±0.17	−9.6±18.7	0.186

Abbreviations: Amyg, amygdala; Ant. cing, Anterior cingulate cortex; Dlpfc, Dorsolateral prefrontal cortex; Hipp, Hippocampus; Occ, Occipital cortex; Ofc, Orbitofrontal cortex; Par, Parietal cortex; SRTM, simplified reference tissue model; Temp, Temporal cortex; Thal, Thalamus; Vmpfc, Ventromedial prefrontal cortex; 2TC, 2-tissue compartmental model.

Significance level was *P<0.05 (two-tailed, paired t-test, uncorrected for multiple comparisons) of the difference between baseline and postamphetamine challenge scans.

Discussion

Positron emission tomography imaging studies can be complicated by the need to obtain arterial input functions. In human subjects, placement of arterial lines is invasive and not always attainable. In the current study, we compared reference region input modeling methods, SRTM and SRTM2, with the arterial input modeling method, 2TC, for the analysis of [¹¹C]FLB 457 BP_ND to determine the feasibility of analyzing the data using a reference region approach. Other reference region approaches, MRTM, MRTM2, and Logan (with different t* values), were compared with 2TC for the analysis of BP_ND, evaluated preliminarily in the first four subjects' pre- and postamphetamine data sets. All approaches were negatively biased compared with BP_ND(2TC) and showed similar effects as SRTM and SRTM2. SRTM and SRTM2 estimates can be the least noisy due to lower number of parameters, compared with 2TC, or since all frame data points are included in fit, compared with MRTM, MRTM2, and Logan reference approaches. Thus, subsequent analyses of BP_ND were performed with SRTM and SRTM2.

The negative bias in BP_ND for SRTM and SRTM2, observed in Figure 3, is expected since SRTM works optimally for tracers that are well-fitted with the one-tissue compartment model, which is not the case for [¹¹C]FLB 457. When the gold standard is the 2TC model, it is expected that SRTM would introduce a bias, as seen for other tracers.^{23, 27, 28} The amplitude of this bias would be different depending on the ROI properties: e.g., simulation studies with 5HT_1A radioligand [¹¹C]WAY-100635 showed that the negative bias with SRTM, using the cerebellum as reference input, is increased in regions with higher receptor density and with decreasing estimates of R₁ (ratio of tracer delivery rate to the target and reference ROI).²⁸ In this study, the bias was indeed larger in the higher binding thalamus and amygdala regions (BP_ND >2), while the other ROIs with (BP_ND <2) lay on the line of identity.

More importantly, the negative bias in BP_ND with SRTM (26%) and SRTM2 (9%) was consistent between baseline and postamphetamine conditions such that the bias cancels out when computing the primary outcome measure, %ΔBP_ND. Postamphetamine reductions in BP_ND were observed with SRTM and SRTM2 across all regions examined, but not in all ROIs with 2TC. Additionally, SRTM and SRTM2 were more reliable models for detecting significant differences in BP_ND after the amphetamine challenge across several extrastriatal regions that were not detected with 2TC.

Plasma Amphetamine Levels

After amphetamine administration, plasma levels peaked at 2  to 3 hours for five subjects, however, one subject's plasma levels peaked early at 1 hour. This subject did not strongly influence the results when excluded from the analysis; ROIs with a significant decrease in BP_ND after amphetamine were retained. Importantly, at a dose of 0.4 or 0.5 mg/kg of amphetamine, plasma levels were consistent across subjects at the start of the second [¹¹C]FLB 457 scan.

Modeling Methods

All models visually produced good fits for low binding regions but were slightly poorer for SRTM2, more noticeable in the relatively higher binding regions. Statistically, SRTM produced better fits than SRTM2 across regions for all scans based on the F-test. However, compared with 2TC, SRTM2 was 17% less biased than SRTM for BP_ND. When comparing 2TC with SRTM, the frequency of cases where the relative standard error on BP_ND estimates exceeded 20% was 11% and 3%, respectively. Larger standard errors in BP_ND estimates are due to an increased number of parameters estimated with 2TC (4 parameters) versus SRTM (3 parameters). Another source of error in BP_ND(2TC) may be due to random fluctuations in arterial input function data. These factors that yield noise in BP_ND(2TC) may explain the lack of sensitivity to detect significant differences post-amphetamine, only 3 of the 10 ROIs examined had a mean reduction in BP_ND, and the larger variability in %ΔBP_ND across subjects. On the basis of these data, reference region input modeling approaches decrease the noise in BP_ND estimates, increase the sensitivity to detect significant changes postamphetamine, and reduce the variability in %ΔBP_ND across subjects.

Cerebellum as a Reference Region

Previous studies have reported specific binding of [¹¹C]FLB 457 in the cerebellum that would lead to an underestimation in BP_ND.^{10, 11, 12} This underestimation applies to the 2TC model as well as to the SRTM and SRTM2 models, since 2TC BP_ND values were computed using the total V_T estimate for the cerebellum. Using the cerebellum as a reference in 2TC to compute BP_ND in amphetamine challenge studies induces a bias in the estimation of %ΔBP_ND. For example, if we assume that 17% of the cerebellum V_T correspond to specific binding,¹³ then %ΔBP_ND would be overestimated by ~25% (i.e., a true 16% decrease in the true BP_ND would lead to a %ΔBP_ND estimate of −20%) in a region which true BP_ND is 1. This bias decreases in regions with larger true BP_ND (it would be ~11% in a region which true BP_ND is 2). Thus, this underestimation of BP_ND in 2TC leads to an overestimation of %ΔBP_ND.

Other studies have compared other approaches for the quantification of [¹¹C]FLB 457 binding, such as transient equilibrium and linear graphical analysis, and have shown that SRTM would produce the least biased estimates of BP_ND.^{14, 15, 16} The aim of this study was to examine the degree of underestimation of BP_ND in amphetamine challenge studies, and the sensitivity for detecting amphetamine-induced dopamine release, as compared with an arterial input model method. The lack of change in cerebellum V_T and V_T/f_P pre- to postamphetamine helps validate the use of the cerebellum in amphetamine challenge studies, since the lack of amphetamine effect in the cerebellum avoids adding an extra bias and variability to %ΔBP_ND. Moreover, the low and consistent bias between 2TC and SRTM2 BP_ND estimates in the two conditions, and the lower variability of SRTM(2) BP_ND estimates compared with 2TC, further validates the use of reference region modeling approaches for future amphetamine challenge studies with [¹¹C]FLB 457 and PET.

Limitations

Data from six tobacco-smoking subjects were compared for the analysis of reference region input modeling approaches for amphetamine challenge studies with [¹¹C]FLB 457. One limitation of the study is the small sample size, but with SRTM and SRTM2 we were able to detect significant reductions in BP_ND postamphetamine across regions that were consistent with another study.⁶ Another possible drawback of this study is the variability in %ΔBP_ND between tobacco-smoking subjects. Variability among subjects may have been attributed to smoking habit. To examine this, number of pack years (number of cigarettes smoked per day × number of years smoked/20) was computed for all subjects with a mean±s.d. of 14±7. Regression analysis was performed, and there were no significant correlations between pack years and reduction in BP_ND(SRTM) nor BP_ND(SRTM2), averaged across ROIs. In addition, the variability is comparable to that previously reported in a larger group of 12 healthy control subjects.^{6, 7}

Conclusion

These data support the sensitivity of [¹¹C]FLB 457 and PET with SRTM and SRTM2 to detect amphetamine-induced dopamine release in extrastriatal regions in tobacco smokers. No changes in distribution volume were observed in the cerebellum. Both SRTM and SRTM2 can be used to estimate [¹¹C]FLB 457 BP_ND across brain regions with lower variability and consistent bias across baseline and amphetamine conditions, with lower bias in BP_ND with SRTM2, compared with 2TC. Thus, the need for arterial input data for future amphetamine challenge studies is not strictly required.

The authors declare no conflict of interest.