Background
Anorexia nervosa (AN) is an eating disorder (ED) characterized by food restriction, inappropriate eating habits, obsession with having a thin figure, an irrational fear of weight gain, and a distorted body self-perception [
1]. AN is increasingly recognized as a serious disease that affects many young individuals. However, its etiology is complex and treatment effect is limited [
1]. Problems with homeostasis, drive, and self-regulation are biological factors that are known to be associated with AN [
1]. Although findings of magnetic resonance imaging studies of subjects with AN are inconsistent [
2], other functional neuroimaging modalities, including single-photon emission tomography, positron emission tomography (PET), and functional magnetic resonance imaging (fMRI), suggest that patients with AN exhibit functional abnormalities in the frontal, parietal, and cingulate cortices [
3‐
5]. In addition, a recent fMRI study in an adolescent population reported that activation of the medial prefrontal cortex during performance of a theory of mind task was lower in patients with AN than in controls (CTLs) [
6]. These results may indicate that decreased activation of the prefrontal cortex underlies neural malfunction related to social cognition and behavior.
Although patients with ED with extremely low body weight often pose a serious clinical problem, there are few studies of this population. In addition, patients with ED require careful physical management and behavioral suppression in the acute phase of therapy. The portability, compactness, and non-invasive features of near-infrared spectroscopy (NIRS) make it an ideal tool with which to study functional brain activity in patients with ED. NIRS allows the measurement of functional brain activity under near-natural conditions [
7]. It is based on the principle that near-infrared light is preferentially absorbed by oxygenated hemoglobin (oxy-Hb) and de-oxygenated hemoglobin (deoxy-Hb) compared with other body tissues [
8]; furthermore, it quantifies regional oxy- and deoxy-Hb concentrations with a high time resolution. Patients with depression, schizophrenia, and bipolar disorder have been studied using NIRS, and their characteristic time courses of oxy-Hb changes in the frontal lobe have been investigated [
8‐
10]. Four NIRS studies in patients with ED have been conducted [
11‐
14], three of which reported lower oxy-Hb concentrations in the frontal cortex during the letter fluency task (LFT) in patients with ED than in healthy CTLs [
11‐
13]. However, the patient population and methodology varied across these studies. For example, Uehara et al. (2007) did not evaluate the relevance of clinical symptoms [
11], Suda et al. (2010) excluded patients with body mass index (BMI) less than 14.5 kg/m
2 to exclude the effect of malnutrition [
12], Nagamitsu et al. (2011) studied children [
13], and Sutoh et al. (2013) studied patients with AN with relatively high BMI (mean ± SD, 17.0 ± 3.1 kg/m
2) [
14].
Some patients with ED experience feelings of social self-doubt and unhappiness, which may have implications for treatment [
15]. Many studies have reported that social skills and interpersonal difficulties are strongly associated with the psychopathology of patients with ED [
16,
17]. It is speculated that patients with ED tend to have interpersonal sensitivity, low self-esteem, social anxiety, poor emotional support, and social inhibition, all of which are associated with ED psychopathology [
16]. A recent study reported that patients with AN showed impaired cognitive flexibility as well as hypo-activity in the ventrolateral prefrontal cortex [
18]. Furthermore, patients with ED had alterations in the frontal cortex that contribute to reward and anxiety processing [
19]. Moreover, it was shown that the frontal cortex is involved in reward-guided learning and decision-making [
20], while the reward system is associated with prosocial behavior (i.e., helping, sharing, donating, cooperating, and volunteering) [
21]. For these reasons, it is hypothesized that patients with ED, especially ones with severe weight loss, have a neural abnormality that influences social cognition and behavior and prevents them from adapting well to society.
In summary, there are several studies showing that patients with AN have both feelings of social insecurity (SI) and functional abnormalities in the frontal cortex that are associated with prosocial behavior. Therefore, the correlation between SI and frontal activity may be altered in patients with AN compared to CTLs. Although a previous study showed that the correlation between frontal cortex oxy-Hb concentrations during the LFT and Eating Attitudes Test scores differed between the ED and CTL groups [
13], no neuroimaging studies have directly examined the relation between frontal cortex activity and SI in subjects with ED and CTLs. The aim of the present study was to investigate brain activity and its association with social relationships in patients with ED with extremely low body weight. We tested the hypothesis that frontal cortex oxy-Hb concentrations during the LFT would be lower in patients with ED than in healthy subjects, and that the correlation between frontal cortex oxy-Hb concentrations during the LFT and Eating Disorder Inventory-2 (EDI-2) score, which includes the SI subscale, would differ between groups.
Discussion
Mean change in oxy-Hb during the LFT and the post-task period
In the present study, we examined frontal cortex activity in patients with ED with extremely low body weight using hemodynamic changes measured by NIRS during an LFT. Although mean change in oxy-Hb concentration during LFT in several channels, including the bilateral OFC, did not show significant differences between the ED group and the CTL group, their effect sizes were medium-to-large. This could be due to a low statistical power to detect an effect when one is present (type II error), which is in part explained by adjustment for multiple testing and small sample size. This result suggests that the oxy-Hb concentration during LFT in the bilateral OFC tended to be lower in the ED group than in the CTL group. The result that oxy-Hb concentration during LFT in the bilateral OFC tended to be lower in the ED group than in the CTL group is consistent with our first hypothesis and previous reports [
11‐
13].
Nagamitsu et al. [
13] suggested that impairment of regional cerebrovascular reactivity might be caused by prolonged starvation or abnormal eating behavior during the illness, and that the unchanged or less fluctuating response pattern of oxy-Hb in the prefrontal area might indicate abnormal cortical processing during cognitive activation. In the following sections, we focus on the relevance of the activity of the OFC and LFT performance and EDI-2 score to discuss whether this suggestion is consistent with our results or not.
In terms of the post-task period, we cannot compare our results with a previous report directly because the previous report use different methods for data processing called linear fitting [
13]. In our study, there were no significant differences in the mean change in oxy-Hb concentration during the post-task period between the ED group and the CTL group after correction of significance levels by the false discovery rate method, and their effect sizes were small to large. It is difficult to evaluate these results because of a lack of robustness.
All subscales of the EDI-2 were significantly higher in the ED group than in the CTL group, which is consistent with a previous report except for the subscale of Body Dissatisfaction [
40]. Our finding that the ED group showed significantly higher scores on the BDI than the CTL group is also consistent with a previous report [
41]. Several studies reported that LFT performance was significantly positively related to level of education [
42‐
47], and that the performance of LFT fell in subjects with damage to the frontal lobe [
48,
49]. It is conflicting that the ED group in our study showed significantly higher LFT performance than the CTL group despite the fact that years of education were significantly lower, and oxy-Hb concentration of the bilateral OFC during the LFT tended to be smaller in the ED group than in the CTL group.
Previous NIRS reports that adopted a similar LFT protocol to our study showed that the ED group had lower frontal oxy-Hb concentration but showed almost exactly same performance on the LFT [
11‐
13]. Nagamitsu et al. [
13] mentioned that the specific patterns of oxygenation changes might indicate less supply and less demand of cerebral blood volume. A meta-analysis reported that patients with AN performed better on the LFT than CTL subjects [
50]. One author suggested that this finding may be because patients with AN patients showed a higher intelligence quotient (IQ) than the CTL group [
51], and LFT has been shown to have a strong relationship with IQ [
52]. Even so, we cannot explain the reason that the ED group showed better performance than the CTL group on the LFT yet had a lower oxy-Hb concentration during the LFT.
Two hypotheses may help explain these findings. One hypothesis is that as a result of a malfunction in the OFC, patients with ED might have partial overactivity in other cortical brain regions such as the thalamus, parietal lobes, or temporal lobes, which are also reported to be activated during the LFT [
28]. Another hypothesis is that as a result of a malfunction of the mechanisms that coordinate work and energy supply, so called “neurovascular coupling” may occur [
53], and neural overactivity might occur despite low blood perfusion in the OFC in patients with ED. It was reported that in patients with ED, OFC volume was higher compared to CTL, which, in general, is supposed to reflect anxiety and high frontal activation in patients with ED [
19].
Altogether, we hypothesize that overactivity in other cortical brain areas as a result of a malfunction in the OFC or neural overactivity despite low blood perfusion in the OFC might be related to high performance on the LFT and inadequate feelings of SI, for example, over-optimistic expectations. To gain evidence of the hypofrontality and better performance in patients with ED, a further study enabling measurements of the entire cortex or of neuroimaging signals of the OFC is required.
Correlation between oxy-Hb concentration and demographic characteristics
The mean BMI of patients with ED in the present study was equivalent to the lowest BMI of patients with ED included in previous NIRS studies [
11‐
14], and lower than the BMI of patients with ED included in many studies using fMRI [
54‐
64] or PET [
65‐
72]. To the best of our knowledge, this study represents the first report of both brain activity and clinical features of patients with ED with extremely low body weight. As such, the results may be influenced by malnutrition. However, the BMI of patients with ED was not significantly correlated with oxy-Hb concentration during the LFT of the bilateral dorsolateral prefrontal cortex and bilateral frontopolar areas, and the mean change in oxy-Hb concentration during the LFT was significantly correlated with SI score, even after adjusting for BMI and after correction of significance levels by the false discovery rate method. These results suggest that the BMI of patients with ED may not affect the frontal activity and SI score. Further study of recovered patients with ED is needed to examine whether the observed correlations are a trait of patients with ED that remains after recovery or are associated with the state of malnutrition.
Correlation between oxy-Hb concentration and clinical characteristics
The mean change in oxy-Hb concentration during the LFT in channels 20 and 21, which include the bilateral OFC, had a strong, positive correlation with SI score in the ED group even after adjusting for age, years of education, BMI, and BDI score as well as correction of significance levels by the false discovery rate method. In contrast, the mean change in oxy-Hb concentration during the LFT in channels 17 and 22, which includes the left OFC, had a weak negative correlation with SI score in the CTL group before adjusting for age, years of education, BMI, and BDI score, but lacked significance after adjusting for these variables. Consistent with our hypothesis, oxy-Hb concentration during the LFT of the OFC in the ED group was lower than in the CTL group, and correlations between oxy-Hb concentration during the LFT in OFC and SI were different in the ED and CTL groups.
Recent work has emphasized the role of the OFC both in value-based decision-making [
73] and in signaling outcome expectancies that are crucial for changing established behavior in the face of unexpected outcomes [
74]. Signaling expected outcomes could be considered a general property of the OFC [
73]. The reason that oxy-Hb concentration during the LFT in OFC correlated with only SI but not with other EDI-2 subscales may be that, in our opinion, SI is directly related to expectations and other subscales are not. The CTL group had high oxy-Hb concentration in the OFC during the LFT, and there is a tendency that the higher the oxy-Hb concentration of the OFC, the lower the SI score.
We propose both that the CTL group, which showed increased oxy-Hb concentration in the OFC during the LFT, behave adaptively and exhibit value-based decision-making in the face of unexpected outcomes in complex human relationships, and that this adaptive behavior may enable them to solve problems and form good human relationships, thus enabling them to integrate in society; as a result, their SI score becomes lower.
In contrast, the ED group had low oxy-Hb concentration in the OFC during the LFT and a high SI score, and the lower the oxy-Hb concentration of the OFC, the lower the SI score. In other words, the patients with ED who have low oxy-Hb concentration of the OFC during the LFT tend not to feel SI. We propose that low oxy-Hb concentration during the LFT of the OFC means that patients with ED neither behave adaptively nor exhibit value-based decision-making in the face of unexpected outcomes in complex human relationships, and that this maladaptive behavior may inhibit the formation of human relationships, thus isolating patients with ED from society. In support of this proposal, it has been shown that patients with ED have a non-assertive interpersonal style, greater social skill difficulties, less socially effective behavior, a smaller social support network, and more difficulties using this network than CTL subjects [
16]. Furthermore, OFC malfunction may mean that patients with ED are not aware of their isolation. This is supported by a report finding that although AN patients had significantly less social support than BN patients, they were satisfied with the support they received [
75]. This phenomenon may relate to denial of illness that most patients with AN have, which is also associated with resistance to treatments observed in these patients [
76].
A solid therapeutic relationship is recommended to overcome treatment resistance [
76,
77], and it might be also recommended for patients with ED who have low oxy-Hb concentration of the OFC during the LFT. Interpersonal psychotherapy that improves interpersonal functioning by enhancing communication skills in significant relationships has been reported to be an effective therapy for AN [
78].
Therefore, malfunction of the OFC may underlie the maladaptive behavior of patients with ED and may represent a biological cause of the psychopathological factors of ED. To support this hypothesis, methodological improvements that can investigate relations between the function of the OFC and performance of tasks that directly induce SI are needed in future studies. In addition, a comparison of AN and BN using the same method would be of interest because there is evidence that there are functional and structural cerebral differences between BN and AN [
79].
Regional hemodynamic changes in the left dorsolateral prefrontal cortex (Channel 18) were positively correlated with the drive for thinness score, and regional hemodynamic changes in the left frontopolar area (Channel 12) were positively correlated with the bulimia score. This is inconsistent with a previous study using the Japanese version of the Eating Attitude Test (EAT-26) [
80,
81], which reported that regional hemodynamic changes in the right frontotemporal regions negatively correlated with dieting tendency scores on the EAT-26, and regional hemodynamic changes in the left OFC negatively correlated with binge eating scores in patients with ED [
12]. These discrepancies may be due to differences in the methods used to evaluate ED symptoms and/or differences in ED patient characteristics such as BMI. Further study is needed to test these possibilities. In the present study, the higher the drive for thinness score, the larger the increase in oxy-Hb concentration in the left prefrontal cortex during LFT in the ED group. However, a previous PET study reported that [18 F]-altanserin binding potential in several cortical regions, including the prefrontal cortex, was negatively related to the drive for thinness in patients with AN [
82]. It is interesting that oxy-Hb concentration increased in the left prefrontal cortex of patients with ED during LFT, whereas metabolism in the same region was decreased.
Study limitations
This study has several limitations. First, the number of participants was small, and further study with more participants is required to increase the statistical power. There was heterogeneity of ED subtypes, but each subtype had too few participants to analyze intra-group differences. Additional subjects are needed for future studies. Second, the ED group was not homogeneous in terms of comorbidity, psychotherapy, and medications, and this may have influenced the results. Third, there may be a selection bias in the CTL group such as years of education, because they were recruited from the hospital staff. Fourth, NIRS has several methodological limitations. The exact measurement point over the cortex differs across subjects according to the size of the skull and brain; therefore, the point of measurement can only be determined in a probabilistic manner. In addition, determining the exact distance of the near-infrared light emitters from light detectors that is required to calculate the change in oxy-Hb remains difficult. As brain atrophy has been seen in patients with AN [
2,
83‐
85], data from the NIRS was possibly affected due to path length factors of near-infrared light. Moreover, evaluating deep structures of the brain is not possible. The validity of NIRS measured on the forehead as a measure of functional brain activity is unknown; however, according to Takahashi et al. (2011), NIRS signals measured on the forehead during the LFT would reflect task-related changes in subcutaneous blood flow [
86].
Acknowledgements
We wish to offer our sincere thanks to both the patients and the healthy volunteers who participated in our study. This work was done at Department of Psychiatry, Nagoya University Graduate School of Medicine. The summary of this study was presented at the 11th World Congress of Biological Psychiatry, in 2013, Japan. Funding for this study was provided by research grants from the Ministry of Education, Culture, Sports, Science and Technology of Japan (Grant-in-Aid for Scientific Research on Innovative Areas 23118004 "Adolescent Mind & Self-Regulation"); JSPS KAKENHI Grant Number 22591283; the Ministry of Health, Labor and Welfare of Japan and The Academic Frontier Project for Private Universities, Comparative Cognitive Science Institutes, Meijo University.
Competing interests
The authors declare that they have no conflicts of interest.
Authors’ contributions
HK, ST, and NO conceived and designed the experiments. HK, KK, ST, MI, NK, and KN performed the experiments. HK, KK, NK, MA, TI, and NO analyzed the data. YN contributed reagents/materials/analysis tools. HK, KK, NK, BA, TI, and NO wrote the paper. All authors read and approved the final manuscript.