Test–retest reliability of the novel 5-HT1B receptor PET radioligand [11C]P943

  • Aybala Saricicek
  • Jason Chen
  • Beata Planeta
  • Barbara Ruf
  • Kalyani Subramanyam
  • Kathleen Maloney
  • David Matuskey
  • David Labaree
  • Lorenz Deserno
  • Alexander Neumeister
  • John H. Krystal
  • Jean-Dominique Gallezot
  • Yiyun Huang
  • Richard E. Carson
  • Zubin Bhagwagar
Original Article

DOI: 10.1007/s00259-014-2958-5

Cite this article as:
Saricicek, A., Chen, J., Planeta, B. et al. Eur J Nucl Med Mol Imaging (2015) 42: 468. doi:10.1007/s00259-014-2958-5

Abstract

Purpose

[11C]P943 is a novel, highly selective 5-HT1B PET radioligand. The aim of this study was to determine the test–retest reliability of [11C]P943 using two different modeling methods and to perform a power analysis with each quantification technique.

Methods

Seven healthy volunteers underwent two PET scans on the same day. Regions of interest (ROIs) were the amygdala, hippocampus, pallidum, putamen, insula, frontal, anterior cingulate, parietal, temporal and occipital cortices, and cerebellum. Two multilinear radioligand quantification techniques were used to estimate binding potential: MA1, using arterial input function data, and the second version of the multilinear reference tissue model analysis (MRTM2), using the cerebellum as the reference region. Between-scan percent variability and intraclass correlation coefficients (ICC) were used to assess test–retest reliability. We also performed power analyses to determine the method that would allow the least number of subjects using within-subject or between-subject study designs. A voxel-wise ICC analysis for MRTM2 BPND was performed for the whole brain and all the ROIs studied.

Results

Mean percent variability between two scans across regions ranged between 0.4 % and 12.4 % for MA1 BPND, 0.5 % and 11.5 % for MA1 BPP, 16.7 % and 28.3 % for MA1 BPF, and between 0.2 % and 5.4 % for MRTM2 BPND. The power analyses showed a greater number of subjects were required using MA1 BPF compared with other outcome measures for both within-subject and between-subject study designs. ICC values were the highest using MRTM2 BPND and the lowest with MA1 BPF in ten ROIs. Small regions and regions with low binding had lower ICC values than large regions and regions with high binding.

Conclusion

Reliable measures of 5-HT1B receptor binding can be obtained using the novel PET radioligand [11C]P943. Quantification of 5-HT1B receptor binding with MRTM2 BPND and with MA1 BPP provided the least variability and optimal power for within-subject and between-subject designs.

Keywords

Serotonin 5-HT1B Positron emission tomography Graphical analysis 

Introduction

The serotonergic (5-HT) system is implicated in the pathophysiology of numerous neuropsychiatric disorders, perhaps most notably major depressive disorder (MDD) [1, 2]. With the exception of the ionotropic 5-HT3 receptor, the remaining six 5HT receptors are metabotropic [3]. The 5-hydroxytryptamine 1 (5-HT1) receptor family includes three subtypes of receptors that inhibit the formation of cAMP (cyclic adenosine monophosphate) that have been shown to be functional in vivo: the 5-HT1A, 5-HT1B, and 5-HT1Dreceptors [3].

The 5-HT1B receptor is a G-protein-coupled metabotropic receptor [4, 5], the majority of which are postsynaptic (reviewed by Clark and Neumaier [6]) and a minority serve as terminal autoreceptors controlling 5-HT release. The development of two novel 5-HT1B PET radioligands, [11C]P943 and [11C] AZ10419369, has allowed novel insights into the system [7, 8]. P943 (R-1-[4-(2-methoxy-isopropyl)-phenyl]-3-[2-(4-methyl-piperazin-1-yl) benzyl]-pyrrolidin-2-one) is a 5-HT1B antagonist that binds with high affinity to the human 5-HT1B receptors (Ki = 1.2 nmol/L) [9]. It has tenfold greater affinity for the 5-HT1B receptor relative to the 5-HT1D receptor and at least two orders of magnitude greater affinity for the 5-HT1B receptor than for all other targets [9]. However, there is a relatively low density of 5-HT1D receptors in the human brain [10]; therefore, it is expected that only a negligible fraction of [11C]P943 in vivo binding reflects binding to 5-HT1D receptors. Using [11C]P943 PET, a number of studies have investigated the role of the 5-HT1B receptor in the pathophysiology of MDD, pathological gambling, posttraumatic stress disorder, and alcohol dependence [7, 11, 12, 13].

The aim of this study was to determine test–retest reliability of [11C]P943 binding potential (BP) in healthy volunteers using two different radioligand quantification techniques, and to determine the number of subjects required for within-subject and between-subject study designs using these techniques. The two modeling techniques were both multilinear analyses: MA1 [14], using arterial input function data, and the second version of the multilinear reference tissue model (MRTM2) analysis [15], using the cerebellum as the reference region. These two methods have several advantages over conventional compartment modeling using one or two tissue compartments. Often the one-compartment model does not adequately fit the time–activity curves (TACs), whereas the two-compartment model can produce parameter estimates, specifically the volume of distribution (VT), with high statistical uncertainty. In contrast, the multilinear methods tend to produce estimates with little or no bias and have reasonable precision. Both were chosen on the basis of previous experience with quantification of [11C]P943 in healthy volunteers [16].

Materials and methods

Subjects

A total of seven healthy volunteers (five men, two women) with no current or life-time personal or family history of psychiatric disorders, between the ages of 19 and 24 years (mean ± SD 22 ± 2.2 years) participated in the study. All subjects underwent a comprehensive screening assessment including a clinician interview for a complete psychiatric and medical history, Structured Clinical Interview for DSM-IV Disorders [17] and other standardized psychiatric assessments, physical examination, routine blood tests, pregnancy test, urine toxicology, and electrocardiogram.

Exclusion criteria included: the diagnosis of life-time or current Axis I or II disorders as judged both by the SCID and a semistructured clinical interview with an experienced clinician; current smoker; current or past serious medical or neurological illness which would preclude participation; the diagnosis of life-time or current Axis I disorders in first degree relatives; current pregnancy (as documented by pregnancy testing at screening and on the day of the PET imaging study) or breast feeding; general MRI exclusion criteria; and significant alcohol or illicit substance abuse or dependence in the past 3 months. All subjects were medication-free for a minimum of 6 weeks at the time of the scan.

The study was performed under protocols approved by the Yale School of Medicine Human Investigation Committee, the Human Subjects Subcommittee of the Veterans Affairs Connecticut Healthcare System, the Yale University Radiation Safety Committee, the Yale-New Haven Hospital Radiation Safety Committee, and the Yale MRI safety committee. Subjects were recruited from the New Haven and surrounding areas by advertisement, word of mouth and referrals, and provided written informed consent for the study after full explanation of the study procedures.

Radiochemistry

[11C]P943 was prepared by N-methylation of the precursor with [11C]methyl triflate, as previously described [9]. Both the P943 standard and the N-desmethyl precursor were provided by Pfizer, Inc (Groton, CT). The average radiochemical yield was approximately 10 % with radiochemical purity of >99 %.

PET imaging

All subjects underwent two PET scans with [11C]P943 using the same procedure for each, on the same day. Female patients were scanned in the follicular phase of the menstrual cycle. There was a mean gap of approximately 4 h between the two PET injections (interval between injections mean ± SD 237 ± 45 min, range 173 – 323 min). In preparation for the scans, the antecubital vein and the radial artery were cannulated for radiotracer injection and arterial blood sampling, respectively. PET imaging was performed using the High Resolution Research Tomograph (HRRT) (Siemens/CTI, Knoxville, TN) which acquires 207 slices (1.2-mm slice separation) with a reconstructed image resolution of about 3 mm. Following a 6-min transmission scan (with a 137Cs rotating point source) used for attenuation correction, 622 ± 155 MBq (scan 1) and 555 ± 141 MBq (scan 2) of [11C]P943 was injected over 1 min by an infusion pump (Harvard PHD 22/2000; Harvard Apparatus, Holliston, MA). Radiochemical parameters are shown in Table 1. List-mode data were acquired for a total duration of 120 min. Head motion was measured using an optical detector (Vicra, NDI Systems, Waterloo, Ontario, Canada) with a rigid tool attached to a swim cap. Calibrations between the HRRT and the well counters are performed on a regular basis and show variability of only 1 – 2 %. Five of seven subjects had injections at approximately 0930 and 1330 hours. One subject’s scans were at 1100 and 1630 hours and another subject’s scans were at 1230 and 1530 hours.Arterial input functions were measured as previously described [16].
Table 1

Injected dose, injected mass and specific activity at the time of the injection for scans 1 and 2

Measure

Scan 2

Scan 1

p value

Mean

SD

Mean

SD

Injected dose (MBq)

630.2

143.3

570.2

134.1

0.43

Injected mass (nmol)

4.7

1.6

4.9

2.9

0.83

Specific activity at the time of injection (MBq/nmol)

158

86.9

146.3

71.9

0.79

Magnetic resonance imaging

MR imaging was performed on a Trio 3-T whole-body scanner (Siemens Medical Systems, Erlangen, Germany) with a circularly polarized head coil using an MPRAGE sequence. The dimensions and pixel size of MR images were 256 × 256 × 176 mm and 0.98 × 0.98 × 1.0 mm, respectively.

Image reconstruction and motion correction

Dynamic scan data were reconstructed with all corrections (attenuation, normalization, scatter, randoms, deadtime, and motion, using the information from the optical detector) using the MOLAR algorithm [18] with the following frame timing: 6 × 30 s, 3 × 1 min, 2 × 2 min; 22 × 5 min. A second step of motion correction was performed by smoothing with a Gaussian filter with a full-width at half-maximum (FWHM) of three pixels and coregistering each frame image to an early summed image (0 – 10 min after injection) using a six-parameter mutual information algorithm (FLIRT, FSL 3.2; Analysis Group, FMRIB, Oxford, UK). Each PET dataset was registered independently to the subject’s MR data. All methods were implemented with IDL 8.0 (ITT Visual Information Solutions, Boulder, CO)

Cerebellum time–activity curve computation

For analyses without arterial input function, the cerebellum TAC was computed. This region of interest (ROI) was defined from the Anatomical Automatic Labeling (AAL) template [19] delineated on a MR template [20]. Using immunoreactivity or autoradiography, 5-HT1B receptors have not been detected in the cerebellum, except in the deep nuclei of the cerebellum in rodents [21, 22], and autoradiographic studies do not show specific binding in the cerebellar gray matter in humans [10, 23]. Based on these findings, the cerebellum was chosen as the reference region in this study. However, in order to reduce the observed spill-over/partial volume effect from surrounding regions of varying receptor binding, we made an a priori decision to manually adjust the cerebellum ROI in the AAL template by excluding the top four and bottom four slices. All subsequent references to the cerebellar ROI refer to this specific region.

To apply the cerebellum ROI to the PET data, a summed image (0 – 10 min after injection) was created from the motion-corrected data and registered to the subject’s T1-weighted 3-T MR image (six-parameter registration), which in turn was registered to the MR template using a 12-parameter affine transformation. The two transformations were combined so that interpolation of the PET data was only performed once.

Quantification

Parametric images of [11C]P943 VT were computed using the metabolite-corrected arterial input function data and the multilinear analysis [15]. The MA1 operational equation is:
$$ {C}_{\mathrm{T}}(t)=-\frac{V_{\mathrm{T}}}{b}{\displaystyle \underset{0}{\overset{t}{\int }}{C}_{\mathrm{P}}}(u)du+\frac{1}{b}{\displaystyle \underset{0}{\overset{\mathrm{t}}{\int }}{C}_{\mathrm{T}}}(u)du,t>t\ast $$
(1)
Parametric images of [11C]P943 BPND were also computed, without relying on input function data, using the second version of the MRTM2 analysis [16], and the cerebellum as the reference region. The MRTM2 operational equation is:
$$ {C}_{\mathrm{T}}(t)=-\frac{1+B{P}_{\mathrm{ND}}}{b}{\displaystyle \underset{0}{\overset{t}{\int }}{C}_{\mathrm{Ref}}}(u)du+\frac{1}{b}{\displaystyle \underset{0}{\overset{t}{\int }}{C}_{\mathrm{T}}}(u)du+\frac{b^{\hbox{'}}}{b}\left(1+B{P}_{ND}\right){C}_{\mathrm{Ref}}(t),t>t\ast $$
(2)

In the above equation, b′ is a parameter linked to the reference region TAC only. Thus, it should ideally be the same in all regions [16]. For each subject, this common b′ value was estimated using Eq. 2 and three-parameter fits, then computing the median of b′ estimates from selected brain voxels (4 > BPND >0.5, and −1,000 < b < −1). Then, Eq. 2 with two-parameter fits was used to compute the final BPND parametric images. For both MA1 and MRTM2, the dynamic images were smoothed with a Gaussian filter (FWHM three voxels). The parameter t* was set to 20 min.

Regional mean parameter value computation

Gray matter ROIs were taken from the AAL template. We selected ten ROIs to cover a range of binding, size and relevance to clinical pathophysiology: amygdala (3.7 cm3); hippocampus (15.0 cm3); pallidum (4.6 cm3); putamen (16.6 cm3); insula (29.0 cm3); frontal (255.9 cm3), anterior cingulate (21.7 cm3), parietal (64.6 cm3), temporal (172.4 cm3) and occipital (80.6 cm3) cortices; and cerebellum (28.9 cm3). Individual parametric images were resliced in template space using the PET to MR and the MR to template transforms (see above). Then, the mean parameter values were computed for each region.

Outcome measures

Based on MA1 estimates, three binding potential outcome measures [24] were considered, each being derived from VT:
$$ \begin{array}{l}\mathrm{M}\mathrm{A}1{\mathrm{BP}}_{\mathrm{ND}}=\left({\mathrm{V}}_{\mathrm{T}}-{\mathrm{V}}_{\mathrm{ND}}\right)/{\mathrm{V}}_{\mathrm{ND}}\hfill \\ {}\mathrm{M}\mathrm{A}1{\mathrm{BP}}_{\mathrm{P}}={\mathrm{V}}_{\mathrm{T}}-{\mathrm{V}}_{\mathrm{ND}}\hfill \\ {}\mathrm{M}\mathrm{A}1{\mathrm{BP}}_{\mathrm{F}}=\left({\mathrm{V}}_{\mathrm{T}}-{\mathrm{V}}_{\mathrm{ND}}\right)/{\mathrm{f}}_{\mathrm{P}}\hfill \\ {}\mathrm{M}\mathrm{RTM}2{\mathrm{BP}}_{\mathrm{ND}}\mathrm{is}\kern0.5em \mathrm{calculated}\kern0.5em \mathrm{a}\mathrm{s}\ \mathrm{s}\mathrm{hown}\kern0.5em \mathrm{in}\kern0.5em \mathrm{Equation}\kern0.5em 2\hfill \end{array} $$

Metrics for comparison of modeling strategies

Between-scan percent variability and percent difference were used to assess test–retest reliability, which were calculated as follows:
$$ \begin{array}{l}\%\mathrm{variability}=\left(\mathrm{Scan}2\mathrm{B}\mathrm{P}\hbox{-} \hbox{-} \mathrm{Scan}1\mathrm{B}\mathrm{P}\right)/\left[\left(\mathrm{Scan}1\mathrm{B}\mathrm{P}+\mathrm{Scan}2\mathrm{B}\mathrm{P}\right)/2\right]\ast 100\kern1em \\ {}\%\mathrm{difference}=\left(\mathrm{Scan}1\mathrm{B}\mathrm{P}\hbox{-} \hbox{-} \mathrm{Scan}2\mathrm{B}\mathrm{P}\right)/\mathrm{Scan}1\mathrm{B}\mathrm{P}\ast 100\kern1em \end{array} $$

Intraclass correlation coefficient (ICC) was used to measure within-subject variability relative to between-subject variability as follows:

ICC = (BSMSS − WSMSS)/(BSMSS + (k − 1)WSMSS)

where BSMSS is the mean sum of squares between subjects, WSMSS is the mean sum of squares within subjects, and k is the number of repeated observations.

A power analysis was also performed to determine the method that would allow the least number of subjects for both the within-subject and between-subject study designs to detect an a priori determined 15 % difference among these ten ROIs. Finally, a voxel-wise ICC analysis was performed using the Statistical Parametric Mapping imaging suite SPM5 (http://www.fil.ion.ucl.ac.uk/spm/). Parametric images for MRTM2 BPND in MNI space were smoothed with a Gaussian kernel of 8 mm (FWHM). ICC maps were computed using the Reliability Toolbox for SPM5 [25]. The voxel-wise analysis was performed for the whole brain and all the brain regions taken from the AAL template (amygdala, hippocampus, pallidum, putamen, insula, frontal, anterior cingulate, parietal, temporal and occipital cortices, and cerebellum) for the ROI-based ICC analysis. Reliability was calculated as the median of the ICC values in the whole brain and within the respective ROIs [26].

Statistics

The data were analyzed using SPSS version 17. Two-tailed t tests were used to compare plasma-free fractions and the injected mass of P943 in scan 1 and scan 2. Repeated measures ANOVA was used to determine differences among four outcome measures (MA1 BPND,MA1 BPP, MA1 BPF, MRTM2 BPND) across regions between scans. Pearson correlations were used to determine the relationship between fP and MA1 BPP.. Power analyses were performed using an online Java-based statistics tool [27].

Results

Input function

Six subjects had usable input function data for both scans and were included in all subsequent analyses. The seventh subject’s arterial cannulation failed in the second scan and hence there was no input function data for that scan. However, the MRTM2 data included all seven subjects. There were no significant differences in injected dose, injected mass or specific activity at the time of injection between scan 1 and scan 2 (Table 1). The free fraction of [11C]P943 in the plasma (fP) was 5.0 ± 0.7 % for scan 1 and 4.3 ± 1.2 % for scan 2 (N = 6). There was a statistically significant decrease of −14 ± 12 % (range = −30 % to +9 %; p = 0.03) in fP values between scan 1 and scan 2. No significant differences in total plasma activity (p = 0.31), parent fraction (p = 0.20) or final parent activity (p = 0.27) were found between 20 and 40 min after injection in scan 1 and scan 2. However, significant differences (p = 0.02) in parent activity were present between 60 and 90 min after injection. This was due to lower average parent fraction in the retest scans (44 ± 11 %) as compared with the test scans (50 ± 10 %).

Cerebellum

Values of cerebellar VT were 4.6 ± 1.1 mL/cm3 (mean ± SD) for scan 1 and 4.5 ± 0.9 mL/cm3 for scan 2. The minimal change in mean VT between scan 1 and scan 2 (percent variability −0.7 ± 6.7 %, range −9.5 % to 7.1 %) was not statistically significant (p = 0.89).

Graphical analysis with input function

MA1 BPND

The mean (± SD) percent variability in MA1 BPND between scan 1 and scan 2 ranged from 0.4 ± 12.3 % to 12.4 ± 13.8 %. The mean (± SD) percent difference in MA1 BPND between scan 1 and scan ranged from −1.0 ± 12.4 % to −14.3 ± 16.6 % (Table 2). There was no significant difference in MA1 BPND values across the regions between scan 1 and scan 2 (F(1,95) = 0.06, p = 0.81). The anterior cingulate cortex showed the lowest percent variability (0.4 ± 12.3 %), and the temporal cortex showed the highest percent variability (12.4 ± 13.8 %).
Table 2

Percent variability, percent difference and absolute percent difference values for MA1 BPND, MA1 BPP, MA1 BPF, MRTM2 BPND in the ten regions of interest

Outcome measure

Insula

Amygdala

Frontal cortex

Parietal cortex

Occipital cortex

Temporal Cortex

Putamen

Pallidum

Hippocampus

Anterior cingulate cortex

MA1 BPND

 Scan 1

0.96

0.79

1.02

0.81

0.90

0.84

0.83

1.60

0.40

1.05

 Scan 2

0.99

0.88

1.11

0.84

0.96

0.97

0.87

1.69

0.42

1.04

 Percent variability, mean (SD)

2.9 (7.7)

11.0 (13.8)

8.0 (13.4)

4.2 (14.0)

7.2 (5.8)

12.4 (13.8)

4.6 (6.1)

5.6 (15.4)

5.7 (13.9)

0.3 (12.2)

 Percent difference, mean (SD)

−3.2 (7.8)

−12.5 (14.9)

−9.2 (14.3)

−5.2 (15.)

−7.6 (6.3)

−14.3 (16.6)

−4.9 (6.5)

−6.9 (16.4)

−6.7 (14.5)

−1.0 (12.4)

MA1 BPP

 Scan 1

4.29

3.63

4.51

3.61

3.98

3.77

3.73

7.20

1.80

4.65

 Scan 2

4.40

4.01

4.89

3.74

4.23

4.27

3.90

7.67

1.87

4.65

 Percent variability, mean (SD)

2.0 (7.6)

10.1 (13.8)

7.1 (15.4)

3.3 (13.5)

6.3 (5.9)

11.5 (16.0)

3.7 (8.1)

4.7 (12.6)

4.8 (13.5)

−0.5 (13.0)

 Percent difference, mean (SD)

−2.3 (7.7)

−11.6 (15.7)

−8.5 (16.8)

−4.1 (13.7)

−6.7 (6.1)

−13.6 (20.0)

−4.1 (8.7)

−5.6 (13.5)

−5.8 (14.8)

−0.2 (13.1)

MA1 BPF

 Scan 1

87.65

74.30

91.74

73.37

81.26

76.89

76.09

146.89

36.62

94.51

 Scan 2

108.39

97.60

121.22

91.91

104.12

106.52

96.19

187.56

46.13

114.49

 Percent variability, mean (SD)

19.2 (19.9)

26.9 (24.9)

24.1 (25.9)

20.4 (23.3)

23.4 (19.9)

28.3 (25.4)

20.8 (19.8)

21.9 (14.1)

21.8 (20.3)

16.6 (23.0)

 Percent difference, mean (SD)

−23.2 (22.2)

−35.3 (35.2)

−31.2 (31.3)

−25.4 (24.2)

−28.7 (23.8)

−37.4 (35.7)

−25.3 (22.6)

−25.7 (17.2)

−27.3 (30.2)

−20.7 (24.1)

MRTM2 BPND

 Scan 1

0.91

0.69

0.96

0.77

0.86

0.79

0.79

1.42

0.36

0.99

 Scan 2

0.89

0.72

0.94

0.77

0.87

0.80

0.80

1.42

0.36

0.94

 Percent variability, mean (SD)

−1.9 (6.1)

5.4 (13.9)

−1.3 (9.1)

0.3 (8.5)

1.7 (5.1)

1.3 (7.8)

0.8 (6.4)

0.9 (19.1)

0.2 (16.4)

−3.7 (8.3)

 Percent difference, mean (SD)

1.8 (6.0)

−6.4 (14.6)

1.0 (8.7)

−0.6 (8.2)

−1.8 (5.3)

−1.5 (7.8)

−1.0 (6.5)

−2.7 (21.5)

−1.4 (17.6)

3.4 (7.9)

MA1 BPP

The mean (± SD) percent variability in MA1 BPP between scan 1 and scan 2 ranged from −0.5 ± 13.0 % to 11.5 ± 16.0 %. The mean (± SD) percent difference in MA1 BPP between scan 1 and scan ranged from −0.2 ± 13.1 % to −13.6 ± 20.0 % (Table 2). There was no significant main effect of time on MA1 BPP values between scan 1 and scan 2 (F(1,95) = 0.17, p = 0.69). There were no statistically significant correlations between fp and BPP. This suggests that correction of VT data by fP (i.e., to calculate BPF) may not be useful. The anterior cingulate cortex showed the lowest percent variability (−0.5 ± 13.0 %), and the temporal cortex showed the highest percent variability (11.5 ± 16.0 %).

MA1 BPF

The mean (± SD) percent variability in MA1 BPF between scan 1 and scan 2 ranged from 16.7 ± 23.0 % to 28.3 ± 25.4 %. The mean (± SD) percent difference in MA1 BPF between scan 1 and scan 2 ranged from −20.7 ± 24.1 % to −37.4 ± 35.7 % (Table 2). There was a trend towards a significant increase in MA1 BPF values in the regions between scan 1 and scan 2 (F(1,95) = 3.52, p = 0.06). This trend was likely caused by the mean decrease in fP between scan 1 and scan 2. The anterior cingulate cortex showed the lowest percent variability (16.7 ± 23.0 %), and the temporal cortex showed the highest percent variability (28.3 ± 25.4 %).

Graphical analysis using the reference region

MRTM2 BPND

The mean (± SD) percent variability in MRTM2 BPND between scan 1 and scan 2 ranged from 0.2 ± 16.4 % to 5.4 ± 13.9 %. The mean (± SD) percent difference in MRTM2 BPND between scan 1 and scan 2 ranged from −0.6 ± 8.2 % to −6.4 ± 14.6 % (Table 2). There was no significant difference in MRTM2 BPND values in the regions between scan 1 and 2 (F(1,114) = 0.09, p = 0.76). The hippocampus showed the lowest percent variability (0.2 ± 16.4 %), and the amygdala showed the highest percent variability (5.4 ± 13.9 %). Figure 1 shows parametric images of BPND obtained in the same healthy control subject with MRTM2 for test and retest scans.
Fig. 1

MR anatomic images (a) and parametric images of BPND obtained from the test scan (b) and retest scan (c) in the same healthy control subject with MRTM2. The slices were selected to display regions with high binding: the pallidum and occipital cortex (top row) and substantia nigra (bottom row). Images were postsmoothed with a 3-mm Gaussian kernel for display

Comparison

Table 2 shows the percent variabilities and differences using the four outcome measures MA1 BPND, MA1 BPP, MA1 BPF, and MRTM2 BPND. The variabilities and differences were the highest with MA1 BPF and were less than 10 % in almost all regions with MA1 BPND, MA1 BPP and MRTM2 BPND. The mean percent variability values in the regions were 0.4 ± 4.7 % for MRTM2 BPND, 5.3 ± 3.6 % for MA1 BPP, 6.2 ± 3.6 % for MA1 BPND, and 22.4 ± 3.6 % for MA1 BPF. The mean percent difference values in the regions were −0.9 ± 2.7 % for MRTM2 BPND, −6.2 ± 4.1 % for MA1 BPP, −7.1 ± 4.0 % for MA1 BPND, and 28.0 ± 5.2 % for MA1 BPF.

Power analyses

The minimum numbers of subjects to detect an arbitrarily chosen difference of 15 % with a power of 0.8 and alpha of 0.05 in at least five of the ten regions studied were calculated. Using a within-subject design, the numbers required were 13, 15, 17 and 19, and using a between-subject design the numbers were 23, 26, 31 and 35 for MA1 BPP, MRTM2 BPND , MA1 BPND and MA1 BPF, respectively (Table 3).
Table 3

Power analysis for the four outcome measures (MA1 BPND, MA1 BPP, MA1 BPF, and MRTM2) in the ten ROIs to detect a 15 % difference (alpha = 0.05)

Region

Total numbers based on within-subject design

Numbers per group based on between-subject design

MA1 BPND

MA1 BPP

MA1 BPF

MRTM2 BPND

MA1 BPND

MA1 BPP

MA1 BPF

MRTM2 BPND

Insula

18

11

19

15

33

19

35

26

Amygdala

15

32

45

15

28

62

90

26

Frontal cortex

19

13

16

16

35

23

29

30

Parietal cortex

21

15

18

17

39

26

34

31

Occipital cortex

19

11

20

15

36

19

37

28

Temporal cortex

13

11

19

12

24

19

35

20

Putamen

8

14

21

8

13

25

40

13

Pallidum

10

13

18

14

17

23

34

26

Hippocampus

14

13

18

16

26

24

34

30

Anterior cingulate cortex

17

8

12

14

31

13

20

26

Intraclass correlation coefficient

Table 4 shows ICC values in the different regions. The lowest and highest ICC values in the ten ROIs were obtained using MA1 BPF and MRTM2 BPND, respectively.
Table 4

ICC values for the four outcome measures in the ten ROIs

 

Outcome measures

Region

MA1 BPND

MRTM2 BPND

MA1 BPP

MA1 BPF

Insula

0.91

0.94

0.90

0.55

Amygdala

0.53

0.55

0.88

0.58

Frontal cortex

0.78

0.89

0.61

0.30

Parietal cortex

0.81

0.92

0.76

0.44

Occipital cortex

0.93

0.96

0.86

0.41

Temporal cortex

0.60

0.88

0.47

0.25

Putamen

0.81

0.87

0.88

0.54

Pallidum

0.41

0.32

0.78

0.53

Hippocampus

0.60

0.69

0.68

0.31

Anterior cingulate cortex

0.75

0.83

0.66

0.41

Voxel-wise ICC analysis

The median ICC value for the whole brain was 0.71 (standard error 0.001). Figure 2 illustrates the voxel-wise reliability for the parametric images of MRTM2 BPND across the whole brain. Table 5 shows median ICC values within the ten ROIs used in the ROI-based analysis. The median ICC value of the whole brain indicates good reliability across all brain voxels. The distribution of binding in the ten ROIs demonstrates that high-binding regions show greater reliability than low-binding regions. This supports the results of the ROI-based ICC analysis.
Fig. 2

ICC values for MRTM BPND greater than 0.70 are overlain on T1-weighted MR images in MNI space in six axial slices from a single subject

Table 5

Voxel-wise ICCs (median, SE) for parametric images of MRTM2 BPND

Region

ICC

Whole brain

0.71 (0.001)

Insula

0.78 (0.003)

Amygdala

0.72 (0.001)

Frontal cortex

0.67 (0.002)

Parietal cortex

0.70 (0.003)

Occipital cortex

0.79 (0.002)

Temporal cortex

0.76 (0.002)

Putamen

0.77 (0.005)

Pallidum

0.70 (0.013)

Hippocampus

0.46 (0.012)

Anterior cingulate cortex

0.73 (0.004)

Discussion

This is the first study of the test–retest reliability of [11C]P943, a 5-HT1B PET radioligand in humans using two multilinear methods (MA1 and MRTM2). Absolute values of BP obtained in this study were consistent with those found in other studies using the same radioligand [16, 28]. The data show that [11C]P943 can be used to reliably measure 5-HT1B receptor availability in humans. The observed test–retest variabilities for each outcome measure were consistent with those obtained in other studies with different PET radioligands [29, 30]. There was minimal variation (<10 %) in the measure of receptor availability using [11C]P943, with MRTM2 providing the least variability. In the power analysis MA1 BPP provided the lowest number of subjects required to determine a 15 % difference in binding using either a within-subject or between-subject design. ICC values were highest when using MRTM2 BPND. Note that ICC values cannot be directly compared when applied to different measures (e.g., BPP vs. BPND); however, these values are useful to compare the same measure (BPND) determined by different methods (MA1 vs. MRTM2). While making decisions on which specific model will be used in a study, one has to acknowledge that MA1 requires arterial input functions with concomitant logistical issues and metabolite analyses, while MRTM2 does not require either (kinetic modeling of [11C]P943 has been discussed by Gallezot et al. [16]).

As shown in Table 2, the mean test–retest reliability was less than 10 % for all except two medial temporal regions using MA1 BPND, MA1 BPP, and MRTM2 BPND; reliability using MA1 BPF was appreciably worse. The most parsimonious explanation for this was the significant decrease of 14 % in fp. Six of the seven subjects showed a decrease in fp from the morning to the afternoon scan, while the seventh showed an increase. Given the small number of subjects, it may not be useful to speculate on the reason for this apparent decrease. Further, we expect that this apparent effect was artifactual, since there was no significant (or even trend level) change in cerebellar volume of distribution between the two scans. With a low free fraction of approximately 4 %, even a small change in the free fraction measurement compromises the ability of BPF to accurately quantify 5-HT1B receptor availability. Thus, this change in fp explains the differences in quantification of radioligand binding using MA1 BPND, MA1 BPP, and MRTM2 BPND compared with MA1 BPF.

Overall the mean variability across the ten regions studied was numerically lowest with MRTM2 BPND suggesting that this method may be the most appropriate to quantify 5-HT1B receptor availability when small changes are a concern. In a recent test–retest reliability study of the novel 5HT1B receptor radioligand [11C]AZ10419369, Nord et al. found similar variability across cortical regions using the simplified reference tissue model [31] for BPND. Nevertheless MA1 BPP and MA1 BPNDalso provided comparable data, and it may well be that the appropriateness of the use or acceptability of an arterial line in the patient population to be studied is the key difference in choosing between methods. The power analysis showed that to detect a change in BP of 15 %, MA1 BPP was the appropriate algorithm to use, with MRTM2 BPND following closely. As mentioned above, the presence or absence of an arterial line may be the deciding factor as the difference between the two best methods in the power analysis had a difference of only two and three subjects for the within-subject and between-subject study designs, respectively.

The radioactivity in the supernatant and pellet was measured with high precision and the percent radioactivity recovered in the supernatant after plasma protein precipitation was very similar for the test and retest scans (e.g., 100.2 ± 11.8 % and 97.5 ± 6.9 % at 90 min after injection, respectively). Overall, 101 ± 14 % of the radioactivity was recovered at the output of the HPLC system. The HPLC fractions were measured with very good precision, e.g., measurement errors of 6.3 ± 1.4 % for the 90-min fraction. All labeled metabolites were more polar (less lipophilic) than the parent compound, and therefore unlikely to have crossed the blood–brain barrier and contributed to the brain activity [32].

The limitations of the study include a small, though very carefully screened cohort of healthy volunteers; extrapolation of the results observed to a larger population should be done with caution although the numbers studied are in keeping with test–retest studies of other radioligands [33, 34]. Therefore the near systematic decrease in free fractions in the second scan compared with the first is a complex finding that needs further explanation. These data also do not allow one to ascribe the change in [11C]P943 binding to either a true change in receptor number or merely the ability of [11C]P943 to detect changes in receptor availability as a consequence of changing synaptic 5-HT. In this context nonhuman primate studies have shown sensitivity of 5-HT1B receptors to changes in 5-HT with two different ligands [30, 35]. Circadian variations in endogenous 5-HT and neuroreceptor expression might also affect [11C]P943 binding [36, 37].

Conclusion

It is possible to obtain reliable measures of 5-HT1B receptor binding using the novel PET radioligand [11C]P943. Quantification of 5-HT1B receptor binding with MRTM2 BPND and MA1 BPP both provided the least variability and optimal power for within-subject or between-subject designs using this radioligand and could be used in future studies.

Acknowledgments

Supported in part by a NARSAD Young Investigator Award (A.S.), Daniel X. and Mary Freedman Fellowship (A.S.), and National Institutes of Health (NIH, K23-MH077914; Z.B.). This publication was also made possible by CTSA grant number UL1 RR024139 from the National Center for Research Resources (NCRR), a component of the NIH, and NIH Roadmap for Medical Research. Its contents are solely the responsibility of the authors and do not necessarily represent the official view of NCRR or NIH.

Conflicts of interest

None.

Copyright information

© Springer-Verlag Berlin Heidelberg 2014

Authors and Affiliations

  • Aybala Saricicek
    • 1
    • 2
    • 3
  • Jason Chen
    • 1
  • Beata Planeta
    • 4
  • Barbara Ruf
    • 1
  • Kalyani Subramanyam
    • 1
    • 2
  • Kathleen Maloney
    • 1
    • 2
  • David Matuskey
    • 1
    • 2
    • 4
  • David Labaree
    • 4
  • Lorenz Deserno
    • 5
    • 6
  • Alexander Neumeister
    • 1
    • 7
    • 8
  • John H. Krystal
    • 1
    • 2
    • 8
  • Jean-Dominique Gallezot
    • 4
  • Yiyun Huang
    • 4
  • Richard E. Carson
    • 2
  • Zubin Bhagwagar
    • 1
    • 2
    • 9
  1. 1.Department of PsychiatryYale UniversityNew HavenUSA
  2. 2.Abraham Ribicoff Research FacilitiesConnecticut Mental Health CenterNew HavenUSA
  3. 3.Department of PsychiatryIzmir Katip Celebi UniversityIzmirTurkey
  4. 4.PET Center, Department of Diagnostic RadiologyYale UniversityNew HavenUSA
  5. 5.Department of Psychiatry and Psychotherapy, Campus Charité MitteCharité - Universitätsmedizin BerlinBerlinGermany
  6. 6.Max-Planck-Institute for Human Cognitive and Brain SciencesBerlinGermany
  7. 7.Department of PsychiatryMount Sinai School of MedicineNew YorkUSA
  8. 8.Clinical Neuroscience Division, VA National Center for PTSDVA Connecticut Healthcare SystemWest HavenUSA
  9. 9.Bristol-Myers SquibbWallingfordUSA

Personalised recommendations