Aberrant Food Choices after Satiation in Human Orexin-Deficient Narcolepsy Type 1

Participants

Patients were recruited from the outpatient clinics of Sleep Medicine Center Kempenhaeghe (Heeze, the Netherlands), SEIN Sleep-wake Center ‘Heemstede’ (Heemstede, the Netherlands) and through advertisement by the Dutch narcolepsy patients' organization. All patients met the diagnostic criteria of the International Classification of Sleep Disorders, Third Edition (ICSD-3).

We included 24 patients with narcolepsy type-1 (also known as narcolepsy with cataplexy [NC]). All had clear-cut cataplexy, a low mean sleep latency (< 8 min) measured with the Multiple Sleep Latency Test (MSLT) and ≥ 2 sleep onset REM periods (SOREMPs) during MLST naps and the previous night's diagnostic sleep study. In 13 patients, orexin cerebrospinal fluid levels were known and shown to be ≤ 110 pg/mL.

Patients with idiopathic hypersomnia (n = 14, [IH]) all had clear excessive daytime sleepiness and a mean sleep latency at the MSLT ≤ 8 min, and the symptoms could not be explained by another sleep disorder.

Healthy control (HC) participants were recruited via poster and word-of-mouth advertisements in Nijmegen and surrounding areas. Healthy controls (n = 19) were matched to the narcolepsy patients on age, gender, BMI, and level of education.

Inclusion criteria were age 18–60 years old, BMI 20–35, and right-handedness. Exclusion criteria were diabetes mellitus, (a history of) clinically significant hepatic, cardiac, renal, cerebrovascular, endocrine, metabolic, or pulmonary disease, uncontrolled hypertension, (a history of) clinically significant neurological or psychiatric disorders and current psychological treatment other than for narcolepsy or idiopathic hypersomnia, deafness, blindness, or sensory-motor handicaps, history of taste or smell impairments, drug, alcohol or gambling addiction in the past 6 months, inadequate command of Dutch language, current strict dieting (i.e. specific diet and/or in treatment with dietician), or food allergy to one of the ingredients used in food rewards.

All participants were recruited on a voluntary basis and gave written informed consent before the start of the study. The study was approved by the Ethical Committee of the Radboud University Medical Center and reported in the acknowledged Dutch Trial register (www.trialregister.nl: TC = 4508).

Procedure

All patients stopped taking medication a week before the study day (see Table 1 for description of medication use). All participants fasted for 5 h before the test session to increase the motivation to win snacks in the food-choice satiety test. The test session took place between 09:00 and 18:00. Timing of the session was matched between groups. Participants visited the Donders Institute for a session that lasted for approximately 3.5 h. The session included 1 h of MRI scanning (results reported elsewhere), the food-choice satiety test and completing questionnaires while having excess to ad libitum snacks.

Questionnaires

The Pittsburgh Sleep Quality Index (PSQI²⁴) and the Epworth Sleepiness Scale (ESS²⁵) were administered to all participants to assess nocturnal sleep quality and disturbances, as well as daytime sleepiness over the last month respectively. Attention span was tested with the forward and backward Digit Span test (subscales from the WAIS-IV²⁶). Measures of height and weight were also collected on the day of testing to assess BMI. The short Dutch Healthy Diet–Food Frequency Questionnaire (DHD-FFQ, modified version: 2 components assessing risk for dental caries were dropped²⁷) was administered to assess how well participants followed the Dutch healthy eating guidelines.

Behavioral Tasks

The food-choice satiety task, described by Hogarth and colleagues,²³ was employed to measure changes in snack choices after satiety on one of two snacks (Figure 1). The task was presented on a laptop using Presentation software. The task consisted of three phases. During the first phase (training), participants could choose and win sweet and salty snacks by making button presses. During the second phase (devaluation by satiety), participants were sensory-specific satiated on either the sweet or the salty snack. The third phase (test) was identical to the first phase, i.e., participants would make button presses to choose the same snacks, except that during this phase no direct feedback was delivered (i.e., nominal extinction) to prevent that reward outcomes would influence behavior.

Participants first selected their preferred salty snack (crisps, red bell pepper salty biscuits or cocktail nuts) and sweet snack (wine gums, skittles or chocolate M&M's). Participants received the following instructions explaining the first phase of the task: “you can win your preferred salty or sweet snack by pressing the left or right arrow button on the laptop. You will learn which button represents which snack by trying, but you won't win a snack each time. When the computer screen shows “choose a button” you can push the button of the snack you want. After the task you will receive 1/5 of the snacks you won.” The (deterministic) key-reward assignment was counterbalanced between participants. Each key had only a 50% chance of yielding its respective outcome (a key press can be interpreted as representing the preference of winning the corresponding snack). On non-rewarded trials, the text “You win nothing” was presented. The snack outcomes (i.e., pictures) were presented for 1,500 ms, followed by a random intertrial interval between 1,000 and 2,500 ms prior to the next trial. Training comprised five 16 trial blocks, and each block contained 2 cycles of 8 trials, with a short break (30 s) in between.

After reading the instructions, participants indicated how hungry they were and how much they wanted the snacks on visual analog scales (VAS). After completing the first task phase, participants received and consumed the snacks they won; participants won 5 sweet and 5 salty snacks on average.

In the second task phase, participants were randomly assigned to either satiety on the sweet or salty snack with the following instruction: “we would like you to eat as much of this snack as you can until you are full or don't want to eat it anymore.”

After sensory-specific satiety, the third phase started. This phase was identical to the first phase except that outcomes were omitted from trials and there were 72 trials in total. The following instruction was given: “you will play the same game as before, you can still win the same snacks and the buttons will still be associated with the same snacks. However, the computer will no longer show you whether you won a snack or not. The computer still counts the number of snacks you won and you will receive the snacks after you finished the task.” After the task participants received their snacks (average around 4.5 sweet and 4.5 salty snack) and rated their hunger and snack wanting on VAS.

Spontaneous Snack Intake

After the food-choice satiety task, participants were asked to fill out questionnaires while 4 bowls with a variety of snacks were placed in front of them. The 4 bowls contained crisps, raisins, wine gums and cocktail nuts. Subjects were told that they could eat the snacks if they felt like it.

Data Analysis

Statistical analyses were performed in SPSS version 21 (SPSS, Inc. Chicago, IL). Outliers, identified using the Grubb test (also known as the maximum-normed residual test²⁸), were excluded for further analyses. Hence, 2 HCs were excluded from data analyses because they preferred one snack considerably more than the other snack in the food-choice satiety task. One NC patient was excluded from data analysis because this patient's devaluation effect was considered an outlier by the Grubb test. One idiopathic hypersomnia patient was excluded from analysis because this patient did not want to participate in the sensory-specific satiety procedure.

All outcome measures were tested for normal distribution and homogeneity of variance; all assumptions of parametric data were met. Percentage of button presses for the devalued snack before and after sensory-specific satiety were calculated to test for devaluation effects. We used a Repeated Measurement ANOVA with pre- and post-satiety button presses for the devalued snack as a within-subjects factor and the two groups as a between-subjects factor to test for the main effect of Satiety, main effect of Group and the interaction effect: Satiety * Group(2). Moreover, hunger and wanting ratings for sweet and salty snacks before versus after sensory-specific satiety were entered into a repeated measurement ANOVA as within-subject variables, and Group as between-subjects variables. Correlations between devaluation effect (post-pre) and ratings were conducted with Pearson correlations, two-tailed. Significant differences in correlations between groups were tested with one-sided Fisher r-to-z transformation.

Spontaneous snack intake was measured by weighing the bowls containing a variety of different snacks before and after presenting them to the participants and calculating the number of calories consumed. Group differences in spontaneous snack intake were tested with an ANOVA with spontaneous snack intake as dependent measure and group as independent factor.

Significance level for all statistical tests was set at P = 0.05 and effect sizes (Cohen's d or partial eta squared [η²] are noted). All data are reported as mean and standard deviation.

Our analyses focus on comparing narcolepsy patients with HC. The additional comparisons with our extra control group, i.e., the IH patients, are shortly described under the heading “Control comparisons” in the Results section.

Participant Characteristics

Table 1 summarizes the demographic and clinical characteristics of the participants who were included in the data analysis. NC patients and HC were well matched on sex, age, BMI and education level. Consistent with previous studies,⁵ the NC patients had a significantly higher BMI than the idiopathic IH patients. Patient groups did not differ in medication type used, daytime sleepiness, and quality of sleep and were—as expected—significantly sleepier than the HC. The digit span did not show differences in working memory between the groups. The patient groups were not different in dietary food patterns, nor impulsivity. However, the HC did differ from the patient groups in that HC followed the Dutch healthy eating guidelines more (measured with the short DHD-FFQ) and were less impulsive.

Devaluation Effect in the Food-Choice Satiety Task

Across both NC and HC groups, our sensory-specific satiety procedure made participants choose the devalued snack less after satiety than before (main Satiety, Table 2). However, this devaluation effect was significantly reduced in NC patients relative to HC (Figure 2; Satiety * Group, Table 2). Testing the devaluation effect (i.e., main Satiety) separately per group revealed that devaluation by satiety led to significantly less button presses for the satiated snack in HC (t₁₈ = 3.391, P = 0.003, d = 0.778), but not in NC (t₂₃ = 1.655, P = 0.111, d = 0.385). This indicates that NC patients did not significantly change their choice for the devalued snack after satiety. These group differences in the devaluation effect were present despite the fact that the number of calories consumed during the sensory-specific satiety procedure was not significantly different between HC and NC patients (Table 2).

Ratings in the Food-Choice Satiety Task

Across HC and NC, hunger was reported lower after than before the sensory-specific satiety procedure (main Satiety, Table 2). We did observe a larger effect of satiety on hunger feelings in HC than NC (Satiety * Group, Table 2), but both groups reported significant reductions in hunger feelings after satiety (HC: t₁₈ = 7.111, P < 0.001, d = 1.631; NC: t₂₃ = 3.035, P = 0.006, d = 0.617).

Next, we assessed whether changes in hunger ratings during the task would be associated with the devaluation effect. Indeed, the satiety-induced decrease in hunger feelings (before – after satiety) correlated with the decrease in choices for the devalued snack in the HC group (r = 0.478, P = 0.038), but did not significantly correlate in the NC group (r = −0.015, P = 0.945). The correlation showed a trend for differences between the HC and NC (z = 1.61, P = 0.053). Thus, less hunger after satiety was associated with reduced responding for the satiated snack in HC, but this association did not hold for the NC group (Figure 3A).

Across and within both groups, reported wanting of the sweet snack was lower after than before satiety (main Satiety, Table 2; HC: t₁₈ = 5.490, P < 0.001, d = 1.261; NC: t₂₃ = 2.439, P = 0.002, d = 0.713). Reported wanting of the salty snack was also lower after than before satiety across groups (main Satiety, Table 2). However, between groups, we did observe more wanting of the salty snack after satiety in NC compared with HC (Satiety * Group, Table 2), although both groups reported significant reductions in wanting of the salty snack after satiety (HC: t₁₈ = 6.249, P < 0.001, d = 1.432; NC: t₂₃ = 5.481, P < 0.001, d = 1.117).

We also tested whether changes in wanting ratings during the task would be associated with the devaluation effect. Correlations between the difference (before – after satiety) in wanting ratings and the devaluation effect showed a similar result as for the hunger ratings. Specifically, in HC, satiety-induced decreases in wanting (across sweet and salty snacks) were associated with decreased choices for the devalued snack (r = 0.679, P = 0.001), but did not significantly correlate in the NC group (r = 0.176, P = 0.411) (Figure 3B). The correlations were significantly different between HC and NC (z = 1.96, P = 0.026). Thus, less wanting after satiety was associated with reduced responding for the satiated snack in HC, and this was significantly different from that in NC who did not show this association (Figure 3).

Spontaneous Snack Intake

After the food-choice satiety task, participants filled out questionnaires and could consume snacks from the bowls placed in front of them. The NC group consumed significantly more than the control group (Table 2 and Figure 4).

Correlations with Trait Measures

Between the groups, we observed differences in impulsivity, and BMI (Table 1). However, none of these trait measures correlated with the devaluation effect or caloric intake during the food-choice satiety task, not across groups and neither within groups. However, we did observe that impulsivity was positively correlated with spontaneous snack intake in the NC group (r = 0.527, P = 0.008) and not in HC (r = −0.173, P = 0.480); this association was significantly different between the groups (z = 2.27, P = 0.023).

Control Comparisons

To test for possible sleep disorder-related processes (such as alertness and medication-withdrawal), we also compared HC with IH patients. Overall, we found no differences between HC and IH patients (for details see Table S1 in the supplemental material).

Additionally, we tested for specific orexin-deficiency effects by comparing NC patients with IH patients (for details see Table S2 in the supplemental material). We observed that satiety-induced decreases in hunger were to a larger extent associated with an increased devaluation effect in IH than in NC (z = 1.68, P = 0.046), similar to the effects observed for HC versus NC. Also, spontaneous snack intake was significantly higher in NC patients compared with IH patients (F_1,37 = 8.929, P = 0.005, d = 1.109), as observed for NC versus HC.

We repeated our analyses in a subset of narcolepsy patients (n = 13) in which orexin levels were available and deficient, and found similar results versus HC (see Data S3 in the supplemental material).