Background
Anorexia nervosa (AN) has consistently been associated with reduced global gray (GM) and white matter (WM) brain volumes, alongside enlarged cerebrospinal fluid (CSF) cavities [
1,
2]. Furthermore, studies have shown focal GM reductions in widespread subcortical and neocortical regions in patients with AN [
2].
Considerable research effort has been devoted to determine whether the structural brain alterations persist following weight-restoration and recovery from AN, but extant findings are equivocal. Longitudinal studies of AN patients have shown normalization of WM but persistent loss of GM at follow-up [
3‐
5]. Normal WM, but reduced GM volumes have also been reported in cross-sectional studies of individuals recovered from AN [
6‐
9], with one study even reporting
increased GM volumes in several regions [
10]. This suggests that while WM normalize, GM reductions persist following weight-restoration and recovery.
In contrast, several studies have failed to detect any GM alterations in recovered AN patients [
11‐
14], indicating that GM reductions observed during the acute phase of the illness are reversible, and thus secondary to emaciation. The inconsistent findings regarding the presence of GM volume reduction in recovered patients may be due to heterogeneity in study design, for example, differences in the operationalization or duration of recovery, or variations in the length of follow-up interval in longitudinal studies. Also, several studies are limited by small sample sizes. In sum, it remains unclear if individuals recovered from AN are characterized by reduced GM mass, and more studies are warranted [
2].
To increase our understanding of the neurobiological underpinnings of AN, it is important to ascertain the presence of GM alterations in recovered AN individuals.
GM reductions following recovery may reflect a trait-like neurobiological risk factor for developing AN, and thus be important in the etiology of the disorder. Alternatively, such reductions may reflect irreversible changes subsequent to emaciation. In the present study, we used magnetic resonance imaging (MRI) to investigate global cerebral mass and regional GM volumes in women recovered from AN.
Discussion
In the present study, regional GM and global GM, WM, and CSF volumes were similar for women long-term recovered from AN and age-matched healthy controls. There were no cerebral tissue volume differences between individuals with a history of restricting AN subtype and individuals with a history of binging-purging AN subtype. This indicates that the failure to detect differences in tissue volumes between healthy controls and RAN women were not due to effects of AN subtype.
Our results support findings from previous studies, showing normal cerebral tissue volumes in individuals recovered from AN [
11‐
14]. In a longitudinal study, Mainz and colleagues [
13] showed that while adolescent AN patients had reduced GM volume when ill, these changes fully normalized at discharge. Moreover, the GM gain observed was positively correlated with weight and inversely correlated with cortisol levels, providing evidence that GM mass reverse in conjunction with hormonal and weight restoration. Wagner and colleagues [
14] also found normal global and regional GM volumes in normal-weight women long-term recovered from AN.
These findings suggest the loss of GM in patients do not reflect atrophy or AN-related neurobiological risk factors, but rather temporary and reversible changes in brain microstructure. Although the mechanisms behind these changes remain unclear, they could be related to hormonal imbalances, for instance excessive cortisol levels [
3,
13,
27]. Cell-shrinkage due to dehydration might also be a contributing factor, as dehydrated healthy individuals show transient reductions of GM and WM [
30].
In contrast, others report persistent GM loss following recovery from AN [
3‐
7,
10]. The conflicting results between studies showing GM reductions in recovered AN individuals and the present study may be attributed to differences in study design, for example sample size, inclusion criteria, and operationalization and duration of recovery. For example, Mühlau and colleagues [
7] reported reduced global and regional GM volumes among recovered AN patients. However, liberal criteria for recovery were used, requiring only a minimum BMI ≥ 17.0 and regular menses for 6 months. One challenge is that there is no universal definition of recovery from AN [
31]. Subsequently, there is great variability in how investigators define recovery, which hampers cross-study comparisons. In the present study, we emphasized weight-restoration and absence of pathological behaviors (e.g. binging, restrictive food-intake) over a 1-year period in our definition of recovery, similar to other studies [
14]. Adequate weight restoration and ample duration of recovery may be crucial for brain normalization to occur. Indeed, longitudinal studies have shown that GM and WM volumes increase upon weight restoration [
3,
5,
13,
32] and reports of persistent GM loss at follow-up may be due to short follow-up intervals [
3,
5]. In our study, the duration of recovery was minimum 1 year, and on average 4 years, which may be sufficient for normalization of cerebral tissue.
One exception is the Friedrich et al. (2012) study, which reported reduced regional GM volumes in women who had been recovered for many years [
6]. Our study failed to detect similar regional GM reductions, even though participants’ average duration of recovery and BMI was comparable between studies. The reasons for these conflicting results are unclear. However, it is worth noting that the sample size in the study by Friedrich et al. was small (
n = 13).
In the present study, lowest lifetime weight was positively correlated with GM volumes in the precuneus and insula. This suggests that GM volumes in these areas are influenced by the severity of AN, even years after weight-restoration and recovery. Similar associations between lowest lifetime weight and GM volumes have been reported in a few previous studies [
7,
27], but not specifically in the precuneus and insula. However, Joos and colleagues [
8] reported that recovered women with a history of severe AN showed decreased GM volumes in the precuneus, and suggested that certain irreversible pathophysiological processes can occur when patients’ weight drops below a critical threshold. In a similar vein, there is some evidence to suggest that longer illness duration also exacerbates GM reductions in AN [
25,
28]. Interestingly, Bär and colleagues [
25] showed that illness duration was inversely correlated with GM volume in the precuneus among AN patients, suggesting this area of the brain is particularly susceptible to the effects of AN. However, no association between illness duration and regional GM volumes was found in the present study, and similar null-findings have been reported by others [
10,
33].
The severity of AN may be an important factor when considering the reversibility of regional GM reductions. It is possible that certain subgroups of recovered AN patients, for instance those with a history of extreme emaciation or long illness duration, are characterized by persistent regional GM reductions following recovery. This could account for why the present study found significant associations between lowest lifetime weight and GM volumes in the RAN group, but no significant between-group differences in GM volume. The impact of the GM reductions incurred by severe AN are likely subtle, and not easily detected in case–control studies where there typically is considerable sample heterogeneity. These issues could partly be the source of contrasting evidence regarding the presence of GM reductions in recovered AN individuals. Selectively investigating individuals with a history of severe AN, as some have already done [
8,
28], may be one fruitful avenue for future studies, to ensure sufficient power is achieved.
The present study also found a significant inverse association between lowest lifetime weight and GM volume in the extrastriate cortex. The direction of this relationship is puzzling considering previous research and the issues discussed above, and its nature is unclear. Of note, the cluster in the extrastriate cortex showing this association was small, and like the other significant associations, did not survive correction for multiple comparisons. It is important to acknowledge the possibility of spurious associations, particularly when results are uncorrected for multiple comparisons.
Limitations of the present study include the cross-sectional design, which precludes making longitudinal conclusions regarding the reversibility of cerebral mass. This design also entailed that lifetime AN diagnosis had to be established retrospectively. There was also considerable variability in the clinical characteristics of the RAN group, which might have influenced our results. Lastly, it is important to note that absence of evidence does not constitute evidence of absence [
34], and the present study’s failure to detect regional GM alterations in AN does not imply that there is none. The sample size in the present study was modest, and it cannot be ruled out that small reductions in regional or global brain volumes in women recovered from AN would be evident with larger sample sizes. However, previous studies with sample sizes comparable to or smaller than our own have reported GM alterations in recovered AN individuals [
6,
8‐
10], indicating the present study was sufficiently powered to detect true group differences of similar magnitude. As discussed, it is possible that GM reductions only are evident in certain subgroups of recovered AN individuals, for example those with severe clinical histories. Unless specifically recruiting such cases, moderate sample sizes might not be sufficiently powered to detect such patterns.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
LB conceived the study, participated in its design, collected the data, performed the statistical analyses, and drafted the manuscript. ØR participated in the design and coordination of the study, and helped draft the manuscript. TE participated in the design of the study, contributed to the statistical analyses, and helped draft the manuscript. All authors read and approved the final manuscript.