Twenty-four women volunteers were recruited from the University of Birmingham. Posters advertised the study as an “Appetite & fMRI study”, and participants were compensated with cash or course credits upon completion. The sample size was based on the results of a previous study (Thomas et al. 2014). Ethical approval was provided by South Birmingham Research Ethics Committee (National Research Ethics Service number 11/WM/0411) and informed consent was provided by all participants. Participants were screened to exclude the following: under 18 or over 65 years old, body mass index (BMI) under 18.5 or over 24.9 kg/m2, English not the first language, taking any psychotropic medication or recreational drugs, past or current Axis 1 disorder (determined by the Structured Clinical Interview for DSM-IV Axis I Disorders; SCID-I/P; Spitzer et al. 2004), pregnant or breastfeeding, smoker, dyslexic, food allergies, diabetic, cognitive dietary restraint score higher than 10 as measured by the Three-Factor Eating Questionnaire (TFEQ; Stunkard and Messick 1985). Participants were also excluded if they had previously taken part in a mCPP study, were left-handed or had any contraindications to fMRI scanning. Women were asked to participate in test days that fell outside their premenstrual week.
In a double-blind, placebo-controlled, crossover design, participants were randomized immediately after the screening days to receive oral mCPP (30 mg) (Thomas et al. 2014) in a single morning dose, or placebo, in a counterbalanced order. mCPP and the matched placebo were supplied by the Guy’s and St Thomas’ NHS Foundation Trust Pharmacy Manufacturing Unit. The washout period between test sessions was 7 days. To maintain blinding, mCPP and placebo were prepared in identical capsules and unblinding occurred on study completion. Peak plasma levels of mCPP are observed 120–180 min after oral administration, which was timed to coincide with the second fMRI scan.
Universal eating monitor
Food was served on a UEM consisting of a balance (Sartorius Model CP4201, Sartorius Ltd., Epsom, UK; 0.1 g accuracy) placed underneath the surface of a table and connected to a laptop computer. A placemat hid the balance from view (Thomas et al. 2014).
Dishes filled with 220 g of pasta were provided. Each time the participant ate 50 g of pasta, the Sussex Ingestion Pattern Monitor (SIPM) software (version 2.0.13) interrupted the participant to complete computerised visual analogue scale (VAS) ratings (hunger, fullness and pleasantness of the pasta). After consuming 150 g, participants were interrupted and provided with a fresh dish of 220 g of pasta. Participants were asked to eat in this manner until they felt ‘comfortably full’. The lunch consisted of pasta shells in a tomato and herb sauce served at 55–60 °C (207 kcal per 220-g serving).
Bowls containing 80 g of cookie pieces were provided. Each time the participant ate 10 g of cookie pieces, the SIPM software interrupted the participant to complete VAS ratings as described above. After consuming 60 g, participants were interrupted and provided with a fresh bowl containing 80 g of cookie pieces. Participants were asked to eat until they felt ‘comfortably full’. The cookies were Maryland Chocolate Chip Cookies, with each cookie being broken into 6–7 pieces (390 kcal per 80-g serving).
Salivary cortisol assessment
Salivary cortisol was collected to confirm a pharmacological response to mCPP administration (Meltzer and Maes 1995) and was measured by liquid chromatography–mass spectrometry (LC–MS/MS) as described previously (Thomas et al. 2014).
The experimental procedure is summarised in Fig. 1.
Participants who met the study criteria were invited to a screening day at which they completed: a medical screening sheet, the Eysenck Personality Questionnaire (EPQ; Eysenck and Eysenck 1975) and a questionnaire to determine whether they usually consume lunch. Height and weight were also taken to calculate BMI. Participants returned after a week for two practice sessions, both a week apart, with the UEM.
Participants arrived having consumed their normal breakfast. If they passed a medical examination (first day only), they were breathalysed and completed a pregnancy test, before completing the first batch of questionnaires to assess what and when they had eaten that morning, and several measures to assess mood: Beck Depression Inventory (BDI; Beck et al. 1961); Befindlichskeit Scale of mood and energy (BFS; von Zerssen et al. 1974); Positive and Negative Affect Schedule (PANAS; Watson et al. 1988); and State-Trait Anxiety Inventory (STAI; Spielberger 1983). They also completed the Power of Food Scale (PFS, Lowe et al. 2009), and the Barratt Impulsivity Scale (BIS—Patton et al. 1995) and baseline VAS to assess the following: ‘alertness’; ‘disgust’; ‘drowsiness’; ‘light-headed’; ‘anxiety’; ‘happiness’; ‘nausea’; ‘sadness’; ‘withdrawn’; ‘faint’; ‘hungry’; ‘full’; ‘desire to eat’; and ‘thirst’.
A baseline fMRI scan was then conducted, after which participants completed the VAS, provided a saliva sample and took either mCPP or placebo. At 30 min post-dosing, they completed another set of VAS. Thirty minutes later, participants provided a saliva sample and completed VAS. They were scanned again and then completed a set of VAS and provided another saliva sample.
Immediately before lunch, participants completed VAS and were given ad libitum access to a pasta lunch via the UEM. After lunch, participants completed VAS followed by a 20-min break, after which a further set of VAS was completed before participants were given ad libitum access to a cookie snack. Immediately after the snack, participants filled out VAS. Approximately 30–40 min later, participants completed a second batch of questionnaires: VAS, BDI, BFS, PANAS, STAI, PFS and BIS and provided a final saliva sample, rated the scanner task food images and had a single blood sample taken. At the end of their second session, participants were fully debriefed, thanked for their time and reimbursed for participation.
The scanner task was based on that used by Allen et al. 2016. There were three separate blocks in which participants viewed 40 food pictures (20 high-calorie food images and 20 low-calorie food images), 40 non-food control pictures and 5 smiley face images in three separate blocks. The high-calorie images (mean number of calories per 100 g of food shown = 365) and low-calorie images (mean number of calories per 100 g of food shown = 79) were comparable with previously published tasks (Goldstone et al. 2009). The pictures were displayed for 2500 ms, fixation points were displayed for 3500 ms and smiley face images were displayed for 1500 ms. Participants were asked to pay attention to all images, but to imagine eating the foods they saw during the task, and to press a button on a button box when they saw a smiley face to ensure they were maintaining attention.
fMRI data acquisition and analysis
An event-related design was used, in which the stimuli were presented in pseudo random sequence. The scanner was a 3.0 T Achieva (Philips) whole body scanner with an eight-channel head coil. T2*-weighted echo planar imaging (EPI) slices were acquired every 2.5 s (TR = 2.5). Thirty-three axial slices with an in-plane resolution of 2.5 × 2.5 × 3 mm2 and slice thickness of 3 mm (no gap) were acquired, with a matrix size of 96 × 96 and field of view of 240 × 240 mm. Acquisition angulation was consistently AC-PC. Two hundred volumes were acquired for each block of the task with two dummy scans which were discarded prior to any analysis. A whole brain T2*-weighted EPI volume (resting state) was also acquired (AC-PC angulation), along with an anatomic T1-weighted volume acquired in the sagittal plane slice thickness of 1 mm and in-plane with a reconstructed resolution of 1.0 × 1.0 × 1.0 mm. The FMRIB software library (FSL; FMRIB, Oxford, www.fmrib.ox.ac.uk/fsl) was used for pre-processing and data analyses. Pre-processing involved high-pass filter cutoff of 60 s; motion correction using FMRIB’s Linear Image Registration Tool (MCFLIRT); motion parameters as regressors of no interest; interleaved slice timing correction; spatial smoothing with a 6-mm full-width-half-maximum kernel; high-pass temporal filtering and FILM pre-whitening. Functional data were registered to their corresponding structural images and transformed to Montreal Neurological Institute (MNI) space using a reference brain (12 DOF linear transformation). Multivariate Exploratory Linear Optimized Decomposition into Independent Components (MELODIC) was used to remove artefacts (these comprised 2% of all MELODIC components).
Analysis 1: effect of task
MELODIC filtered data were entered into a first-level analysis, to produce contrast of parameter estimate (COPE) images for blood oxygen level-dependent (BOLD) response to food and control images separately. Mean BOLD % signal change (unthresholded) was extracted with Featquery using masks of a priori regions of interest (ROI) selected from standard templates from WFUPickatlas (Maldjian et al. 2003). The ROIs chosen were based on previous neuroimaging work (see Supplemental Table 1 for list of ROIs and supporting references). The BOLD % signal change to food and control images was compared using paired sample t tests using IBM SPSS (version 23). Bonferroni correction was applied to control for the family-wise error (FWE).
Analysis 2: effect of placebo versus mCPP on BOLD signals to high- and low-calorie foods
MELODIC filtered data were entered into a first-level analysis, to produce COPE images for high-calorie and low-calorie food images, minus the BOLD response to the corresponding control images. These COPEs were averaged across each of the scanning blocks for each participant. A subsequent analysis was run on these outputs for each participant to subtract baseline scans from post-dosing scans. The outputs were entered into the final mixed effects (FLAME 1 + 2) group analysis, producing contrasts between placebo and mCPP conditions for BOLD signal activity in response to the high-calorie food images and the low-calorie food images separately.
Group Z statistic images were corrected for multiple comparisons by FWE correction using AlphaSim, part of the AFNI toolkit (Cox 1996) (AFNI Version 16.1.16—May 25, 2016). With a voxel-wise threshold of p < 0.005 (Z > 2.6), only clusters with more than 24 contiguous voxels were significant with a FWE rate corrected p < 0.05.
Analysis 2 above was repeated with mean centred VAS ratings (taken immediately prior to each scan) of nausea, light-headed and faint entered as covariates (separately) to account for any non-specific effects of mCPP on the BOLD response to food images.
Main effects and interactions with condition were examined with analysis of variance (ANOVA). Bonferroni correction was used on all follow-up t tests unless otherwise stated.
Data were lost for seven pasta sessions and two cookie sessions due to technical issues, such as participants leaning on the balance. In addition, one participant did not return for the second session.