Our main goal was to investigate whether impulsivity and BE may interact with the TDt and to explore the underlying brain activity in obese or overweight outpatients who are starting treatment to achieve weight loss. In particular, the aims of the study were: (a) to investigate respectively whether a group of obese patients with higher impulsivity and a group with BE showed preference for immediate gratification (i.e., decreased delay discounting), respectively, in different conditions of rewards available sooner (Now condition) or later (Not-now condition), as compared with patients with lower impulsivity or without BE behavior; (b) to evaluate whether the groups with different levels of impulsivity and BE behavior showed different brain activity through high-density electroencephalogram (hd-EEG) recording during TDt. Based upon the previous literature, we expected that: (a) both groups of patients with higher impulsivity or with BE would exhibit more TD in Now and Not-now conditions than patients with lower levels of impulsivity and without BE behavior, (b) different levels of impulsivity and BE behavior would be accompanied by a modulation of frontal cerebral rhythms during TDt.
2.1 Participants
A consecutive and unselected sample of treatment-seeking obese and overweight outpatients enrolled at the Clinical Nutrition Centre of the University Clinical Hospital of Chieti (Italy), was recruited from referrals to a dietary control program for any medical reason during a 12 months-period (January to December 2019). Patients were evaluated during their first medical examination. All the participants were started on a non-surgical, non-medication weight loss program, which was aimed at dietary change, weight control, adequate daily food intake, paced eating, and healthy lifestyle maintenance, tailored on individual basis.
To maximize ecological validity, patients aged 18 to 65 and with a BMI ≥ 25 were included. Subjects were selected for inclusion only if their main reason for consultation was being overweight and had no significant medical comorbidity. Documented or self-reported psychiatric disorders, cognitive impairment, pregnancy, severe medical comorbidity (e.g., thyroid dysfunction, diabetes, chronic liver disease, and any other physical diseases which could interfere with eating behavior), or inability to perform/understand the self‐rating scales were considered exclusion criteria. The sample included 40 outpatients and, after removing those who did not meet the inclusion criteria or provided incomplete data, 26 (65%) outpatients formed the final sample. A team of expert physicians and psychologists evaluated patients for their medical history and past or current psychopathology.
Participants were categorized as either with (BE group) or without BE (NBE group) based on BED or subthreshold BED status (less than 1-weekly episode and/or for less than 3 months) assessed with the eating disorders module of the Structured Clinical Interview for DSM5 – Research Version (SCID-5-RV) [48]. Thirteen participants met criteria for BE and 13 for NBE.
All patients gave written informed consent to participate. The study was designed and carried out in accordance with the World Medical Association Declaration of Helsinki and its subsequent revisions [49] and was approved by the Ethics Committee of the Department of Psychological, Health, and Territorial Sciences of the University G. d’Annunzio of Chieti-Pescara (Prot. n. 254 of 03/14/2017).
2.2 Procedures
2.2.1 Temporal discounting task
The experimental paradigm of Temporal Discounting Task (TDt) was adapted from a standard procedure previously used in literature (see [50; 51; 52; 53]) and completely automated by means of a homemade software written in Microsoft Visual Basic v6.0. Before starting the task, participants were informed that in each trial they had to press “Esc'' for the left-side option and “Enter” for the right-side option on the computer keyboard. The left-side “Esc” key was to be used if they chose the option tag located at the left side of the screen (corresponding to a lower amount of money available sooner) and the right-side “Enter” key if they chose the option tag located at the right side of the screen (corresponding to a larger amount of money available later). They were informed that there were no right or wrong choices and that all choices were fictitious, namely, that they will not receive the actual consequences of their choice.
The option stimuli consisted of 2 labels reporting a monetary amount and a temporal delay (e.g., 20 euros now / 40 euros tomorrow). Each trial began with a 1 s fixation, followed by a screen depicting the two available options. The two options appeared on the left and on right side of the screen, indicating the amount (e.g., a lower amount on the left-side choice tag compared to the amount indicated on the right-side choice tag) and the delay of delivery of the reward (e.g., ‘now’ on the right, ‘later’ on the left). The positions of the labels reporting the amount of money (smaller amount vs larger amount) and the temporal delays (immediately, after a delay) were balanced across conditions.
In the TDt, two temporal ‘Now’ and ‘Not-now’ conditions were included:
-
In the “Now” condition, participants had to choose between a smaller amount of money available immediately (e.g., right-side choice) or a larger amount of money available after a variable delay (e.g., left-side choice). There were six possible delays: 2, 14, 30, 90, 180, and 365 days, which were presented in different blocks wherein for each block five choices had to be made.
-
In the “Not-now” condition, participants had to choose between a smaller amount of money that could be obtained in 60 days (e.g., right-side choice) and a greater amount of money that could be obtained after a variable delay of more than 60 days (e.g., left-side choice). As in the "Now" condition, there were six possible variable delays: 62, 74, 90, 150, 240, and 425 days, which were presented in different blocks wherein for each block five choices had to be made.
Within each block, the amount of the sooner reward was adjusted based on the participant’s previous choice, using a titration procedure that converged towards the amount of the sooner reward that was equal, in subjective value, to the later reward. For example, if participants were presented with a choice between either 40€ later (i.e., later reward or larger-later) or 20€ sooner (i.e., sooner reward or smaller-sooner), whenever the participant picked the sooner reward, the subsequent trial presented an amount that was smaller (e.g., 10€) than the one selected in the previous trial (i.e., 20€). On the other hand, whenever the later reward was chosen (i.e., 40€), the subsequent trial increased the amount in the sooner condition (e.g. 30€). The size of the adjustment in the sooner reward decreased with successive choices; the first adjustment was half of the difference between the sooner and the later reward (in the example above, 10€) whereas, for subsequent choices, it was half of the previous adjustment (e.g., 5€; 2.50€; 1.25€; etc.). This procedure was repeated until the participant made five choices at one specific delay of delivery of the later option, after which, a new series of choices began at another delay of delivery of the later option (of the same or different temporal condition).
There was a control condition, in which participants had to choose between 40€ and a smaller amount of money (adjusted based on the titration procedure described above), both immediately available or both available in 365 days. This condition allows us to verify participant’s attention and engagement levels during the task.
The order of delay blocks, as well as the different temporal conditions (Now, Not-now, and control condition), was randomized within subjects.
== Insert Fig. 1 about here ==
Figure 1. The three experimental conditions of the study.
2.2.2 Self-reported questionnaires
Barratt Impulsiveness Scale (BIS-11): the Italian version of the Barratt Impulsiveness Scale (BIS-11) [54; 55] was used to measure impulsivity. The scale consists of 30 items that evaluate motor, attention, and planning components. Participants were asked to rate how often impulsive traits were descriptive for them (e.g., “I act on impulse”, “I get easily bored when solving thought problems”, “I say things without thinking”). Each item is scored on a 4-point Likert scale ranging from 1 (= rarely or never) to 4 (= almost always or always). Scores range from 30 to 120, with higher scores indicating a higher level of impulsivity. The total score of the BIS-11 is an internally consistent measure and is widely used for the assessment of impulsiveness among the general population and selected patients [56]. Within this sample, Cronbach’s α was .83.
Binge Eating Scale (BES): the Italian version of the 16-item Binge Eating Scale (BES) [57; 58] was used to assess the severity of BE behavior. Scores range from 0 to 46, with a score ≥ 27 having conventionally served as a threshold value for identifying the presence of severe BE, ≥ 18 for moderate BE, and ≤ 17 for minimal or no BE [59]. This instrument has been widely used as a screening tool [59; 60]. It has good internal consistency reliability and high sensitivity and specificity for discriminating between binge eaters and non-binge eaters, presenting similar results to those obtained by reliable and supported semi-structured interviews [59; 60; 61; 62]. Within this sample, Cronbach’s α was .87.
2.2.3 EEG recordings
Participants filled out the psychological scales previously described before having the EEG electrodes attached to their scalp. Participants sat on a chair at about 60 cm from the computer monitor measuring 34 x 27 cm (15.4 inches), and they were instructed to assume and maintain a relaxed position for the entire duration of the session. EEG data were continuously recorded (bandpass: 0.01–100 Hz, sampling rate: 1024 Hz; EB-Neuro Be-plus) from 56 scalp cup electrodes positioned according to a standard 10–10 system (electrical reference between AFz and FCz; ground electrode between Pz and Oz). Electrode impedance was kept smaller than 5 kΩ. Signals were stored on a computer for offline analysis. The EEG recording started 3 min before the TDt and ended 3 min after it. During the registration, the participant was seated comfortably and was asked to maintain open eyes, to reduce the number of blinks, to stay still, and to focus on the TDt.
2.3 Preliminary EEG data analysis
EEG data were pre-processed offline, by using the NPXLab 2016 software (available at http://www.brainterface.com; [63]). Raw EEG data were bandpass filtered between 0.5 and 100 Hz, notch filtered at 50 Hz and sampled at 512 Hz. Data were processed with Independent Component Analysis (ICA) to remove the eye-blinks artifacts. Remaining artifacts were removed by visual inspection. Cleaned data were segmented in single 1-s epochs (1000 ms post-stimulus) with respect to the stimulus onset (response options display) and analyzed in the frequency domain with respect to a baseline period of 1s chosen in the rest period before the task. The frequency bands including delta (δ, 1–4 Hz), theta (θ, 4.5-8 Hz), alpha (α, 8.5–13 Hz), beta (β, 13.5–29.5 Hz), and gamma (γ, 30–40 Hz) were computed for each condition, using average Fourier cross-spectral matrices, with the LORETA (LOw-Resolution brain Electromagnetic TomogrAphy; [64]) KEY software package (v20181101; http://www.uzh.ch/keyinst/loreta#_Toc391372608).
Source localization of the EEG frequency bands was obtained using the sLORETA technique [65]. sLORETA employs the current density estimate given by the Minimum Norm solution and the localization inference based on standardized values of the current density estimates. In conditions of high signal-to-noise ratio, sLORETA has a zero-localization error. sLORETA solutions are computed using a realistic head model [66] within the source space (6239 voxels at 5 mm spatial resolution; [67]), and they are restricted to cortical grey matter and hippocampi, as determined by the probabilistic Talairach atlas [68].
In order to identify intracerebral electrical sources, we used the LORETA software package ‘ROI-maker 2’ to construct the region of interests (ROIs). We selected a list of 33 ROIs, including all voxels with coordinates corresponding to the respective Brodmann Areas (BA; Table 1). Current densities in the region of interest were then computed using the ‘sLORETA to ROI’ function.
Table 1
Regions of interest (ROIs) for the EEG LORETA analysis. Abbreviations: G = gyrus, L = lobe, Inf = inferior, Sup = superior, Fro = frontal, Par = parietal, Temp = temporal, Occ = occipital, Cing = cingulate.
ROI n) Name | Brodmann areas | Lobe | ROI n) Name | Brodmann areas | Lobe |
1) Angular G | 39 | Par, Temp L | 18) Paracentral Lobule | 3–7,31 | Fro, Par L |
2) Anterior Cing | 10,24–25,32–33 | Limbic L | 19) Parahippocampal G | 19–20,27–28,30,34–37 | Limbic, Occ L |
3) Cingulate G | 6,23–24,31–32 | Limbic L | 20) Postcentral G | 1–5,7,40,43 | Fro, Par L |
4) Cuneus | 7,17–19,23,30–31 | Occ L | 21) Posterior Cing | 18,23,29–31 | Limbic L |
5) Extra Nuclear | 13,47 | Fro L, Sub-lobar | 22) Precentral G | 4,6,9,43–44 | Fro, Par L |
6) Fusiform G | 18–20,36–37 | Temp, Occ L | 23) Precuneus | 7,18–19,23,31,39 | Par, Occ L |
7) Inf Fro G | 6,9–11,13,44–47 | Fro, Temp L | 24) Rectal G | 11 | Fro L |
8) Inf Occ G | 17,18,19 | Occ L | 25) Subcallosal G | 11,13,25,34,47 | Fro L |
9) Inf Par Lobule | 7,34,35,49,76,77 | Par L | 26) Sub-Gyral | 2,6–8,10,13,19–21, 31,37,39–40 | Fro, Limbic, Temp, Par L |
10) Inf Temp G | 15,16,17,32,57,58,59,74 | Limbic, Temp, Occ L |
11) Insula | 13,22,40,41,45,47 | Temp L, Sub-Lobar | 27) Sup Fro G | 6,8–11 | Fro L |
12) Lingual G | 17–19 | Occ L | 28) Sup Occ G | 19,39 | Temp L,, L |
13) Medial Fro G | 6,8–11,25,32 | Fro L | 29) Sup Par Lobule | 5,7,40 | Par L |
14) Middle Fro G | 6,8–11,46–47 | Fro L | 30) Sup Temp G | 13,21–22,38–39,41–42 | Temp L |
15) Middle Occ G | 18–19,37 | Occ L | 31) Supramarginal G | 39–40 | Temp, Par L |
16) Middle Temp G | 19–22,37–39 | Temp, Occ L | 32) Transverse Temp G | 41–42 | Temp L |
17) Orbital G | 11,47 | Fro L | 33) Uncus | 20,28,34,36,38 | Limbic L |
2.4 Data analysis
Preliminarily, we performed a power analysis to determine the sample size needed to detect a medium effect size. Power analysis was conducted using: (1) an estimated mean partial eta square of 0.08 for the mixed model ANOVA's repeated measures; (2) an alpha level of 0.05; and (3) a power of 0.95. According to statistical computing, a sample size of n = 26 was required for ANOVA repeated measures. Power calculation was performed using the program G*Power 3.1 [69].
Statistical data analysis was performed using SPSS 26.0 for Windows. Descriptive statistics were reported in terms of mean and standard deviation [Mean (SD)] or absolute frequencies. The level of significance was set at 95%. Independent and paired-sample Student’s t tests or chi-square tests (χ2) were used to compare between- and within-group differences in socio-demographic and clinical variables between BE and NBE patients. Cohen’s d was used as a measure of effect size. A standardized effect size of 0.20–0.50 is considered small, 0.50–0.80 moderate, and > 0.80 large [70]. The Cramer’s φ was also used as a measure of the strength of association for the χ2 test [71]. The three magnitudes of effects of .10–.30, .30–.50, and > .50 are considered as small, medium, and large, respectively. Pearson’s correlation coefficient was used for the associations between BIS and BES scores and TD rate. The four magnitudes of Pearson’s coefficient of .20–.40, .40–.60, .60–.80, and > .80 are considered as low, moderate, marked, and high, respectively.
TDt was assessed through the temporal discounting parameter (k) [72; 73; 74]. This is the rate at which the subjective value of a future reward decays with delay (TD rate), for each temporal condition (Now, Not-now). The hyperbolic function SV = 1/(1 + kD), where SV = subjective value (expressed as a fraction of the delayed amount), and D = delay between the two options at stake (in days), was fit to the data to determine the k constant of the best fitting TD function using a nonlinear, least squares algorithm. The larger the value of k (the steeper the discounting function), the more participants were inclined to choose smaller-sooner (SS) rewards over the larger-later (LL) rewards. The hyperbolic k constants were normally distributed after log-transformation, and the comparisons were performed using parametric statistical tests. Log-transformed k near to 0 describes a prevalent SS pattern of choice (i.e., higher delay discount) while very negative log-transformed k describes a prevalent LL pattern of choice (i.e., lower delay discount).
For this, we first carried out a correlation analysis between BIS scores, BES scores, and the log-transformed k of the Now and Not-now conditions. Due to the multiple comparisons, we applied the Bonferroni correction at a significance threshold of p = .006.
Subsequently, we divided our sample according to presence of BE (according to SCID-5-RV) and impulsivity trait (BIS median score) and carried-out a mixed analysis of variance (ANOVA) on log-transformed k values with BE Group (NBE, BE) and BIS Group (Low BIS, High BIS) as between-subjects factors, and Temporal condition (Now, Not-now) as a within-subjects factor.
Regarding the EEG data, our focus was on assessing the EEG power spectra after the options display (before the response) in the two different conditions of the TDt (Now and Not-now, see the Methods section). We carried out 5 separate mixed ANOVAs, one for each of the 5 frequency bands, on the current density values with BIS Group (Low BIS, High BIS) and BE Group (NBE, BE) as between-subjects factors, Temporal condition (Now, Not-now), Response (SS, LL), and the 33 ROIs (Table 1) as within-subjects factors.
Due to the multiple comparisons, we applied the Bonferroni correction at a threshold of p < 0.01. In all analyses, the post-hoc analyses were performed with the Newman-Keuls test.