Background

A growing amount of evidence indicate that anorexia nervosa (AN) is associated with impaired or inefficient neuropsychological performance in relation to healthy control subjects, regarding attention [1, 2], memory [1,2,3,4], processing speed [4], and especially the executive functions [5] central coherence [6], decision-making [6, 7], and cognitive flexibility [8, 9]. It has been debated whether this is related to state (due to factors such as malnutrition) or trait (a premorbid trait or endophenotype of the disorder [10]). Some studies have found that patients who recovered from AN have impaired cognitive performance compared to healthy control subjects [11, 12], supporting the trait theory of the disorder. However, longitudinal studies have found that executive functions can be normalized following weight stabilization in patients with AN [13, 14], supporting the state theory.

Research on cognitive performance before and after re-nutrition in adult patients with extreme and chronic AN is sparse. Some studies have examined cognitive performance in patients with AN with a mean body mass index (BMI) below 15 kg/m2 (e.g. [10]), corresponding to extreme AN severity according to the Diagnostic and Statistical Manual of Mental Disorders 5 (DSM-5) [15]. However, it is unclear if patients with AN with BMI below 10 kg/m2 will display the same cognitive profile.

It has been suggested that malnutrition might affect cognitive performance since the classic Minnesota Semi-Starvation Experiment [16], where cognitive functions were studied in 36 healthy military objectors with normal weight before and after semistarvation with 25% weight-loss over a 24-week period. The men reported decline in concentration. However, the standardized tests that were administered did not confirm measurable alterations. Newer research on healthy subjects, although somewhat inconclusive, indicates affected psychomotor speed and executive functions following short-term semi-starvation [17].

However, other factors than malnutrition or weight-loss have been suggested to affect cognitive performance in patients with AN, such as long illness duration [18] and age [18]. This could explain a difference in results for children/adolescents and adults with AN mentioned in the literature [19, 20], which cannot be explained by the trait theory.

The current case report was part of an ongoing longitudinal research project investigating the effect of re-nutrition on cognitive performance in patients with severe AN. The aim of the case study was to present the neuropsychological performance of a patient with chronic AN and extremely low BMI in order to discuss whether extremely low weight and long duration of illness are associated with cognitive impairment and if cognitive adaptation takes place. No study to our knowledge has previously reported on the cognitive profile of a patient with AN with a BMI as low as 7.7 kg/m2.

We want to introduce the idea of cognitive adaptation to severe malnutrition as a supplement to the discussion on cognitive impairment in AN. However, this idea should not be confused with Taylor’s Theory of Cognitive Adaptation [21]. The presented idea of cognitive adaptation is the idea that cognitive functions can adapt to persisting low weight in AN, i.e. cognitive performance can remain normal or regain normality in severe and enduring AN. The adaptation does not exclude specific cognitive impairment.

Case presentation

The current case report investigates the cognitive profile of a 35-year-old Caucasian woman with extremely severe and enduring AN who was diagnosed at the age of 10 years. The patient’s weight loss is accomplished through fasting. According to the DSM-5 [15], the patient’s symptoms are in accordance with the restricting type and the severity of AN for the patient is categorized as extreme. The patient has had low body weight since the onset of the disease 25 years ago. Consequently, she is still prepubescent.

The patient’s extreme malnutrition, the medical complications, and the refeeding treatment has previously been described in a case report [22]. Since the previous report [22], she has survived another 5 years, living in her own residence with several stabilizing hospitalizations. Her nadir BMI, defined as the lowest registered BMI, has decreased further to 7.2 kg/m2. To our knowledge, this is the lowest BMI reported in AN in the literature. During her long and severe illness course, she has participated in psychotherapy for years. However, during the past few years, she has refused to participate in psychotherapy, while she has continued the harm-reducing treatment in the nutrition department. No cognitive profile has been assessed before the current report.

She has continuously been provided supplementation with vitamins and minerals. At the present admission, she weighed 20.2 kg, including edema corresponding to at least 2 kg, and her height was 1.55 m, corresponding to a BMI of 8.41 kg/m2. After life-saving and stabilizing fluid and electrolyte correction, and refeeding according to guidelines [23] during 2 weeks of hospitalization, we tested her with a neuropsychological test battery (2 weeks after admission: T0). After an additional 2 months of hospitalization, she could not be motivated to continue the treatment any longer. Due to years of history with rapid relapse after prolonged forced treatment, she was allowed to be discharged to outpatient follow-up. She was re-tested in the outpatient clinic 6 days following dropout from inpatient treatment and approximately 3 months after admission, (re-test: T1) with a weight of 22.4 kg (BMI: 9.3 kg/m2), and again at 12 months from T0, during a re-hospitalization, 7 days after admission (follow-up: T2), with BMI 7.7 kg/m2. Thus, T0 and T2 were done at the hospital after initial stabilizing glycemic, fluid- and electrolyte correction, whereas T1 was done in an outpatient setting, where she was in a clinically stable condition, but without the initial stabilizing treatment.

The psychopathological profile of the patient

The patient scored 21 on the Beck Depression Inventory II (BDI-II [24];) indicating moderate depression at 2 weeks after admission (T0). Her scores on the Eating Disorder Inventory 3 (EDI-3 [25];) at T0 are presented in Table 1 below. Compared to the Danish validation of EDI-3 for patients with AN ( [26]; Table 1), her low scores on the Drive for Thinness, the Interoceptive Deficits, the Perfectionism, and the Asceticism subscales are of interest.

Table 1 Eating Disorder Inventory 3 (EDI-3) scores for the patient and from the Danish validation of EDI-3a on a sample of patients with anorexia nervosa (AN)

Qualitative observations

During the first 2 weeks after admission, the patient was unable to participate in the neuropsychological assessment due to fatigue. Two weeks after admission, when the baseline assessment took place (T0), the patient was lying down during the assessment and was noticeably tired. This was neither the case at retest (T1) nor at follow-up (T2) where the patient was sitting at a table. Her alertness and energy level at follow-up (T2) were notable in light of her low BMI. The patient was calm during all three assessments (divided into six sessions) and expressed that the tests were fun. The aim of the study was explained to the patient before the first administration. However, only information written in the test manuals was given during each assessment.

Measures

The following validated neuropsychological tests were selected in cooperation with an experienced neuropsychologist to examine a wide range of cognitive functions: the Wechsler Memory Scale III (WMS-III) [27]; the d2-R Test of Attention – Revised [28, 29]; the Processing Speed Index (PSI) of the Wechsler Adult Intelligence Scale IV (WAIS-IV) [30]; the Delis-Kaplan Executive Function System (D-KEFS) [31], Verbal Fluency Test, Design Fluency Test and Trail Making Test; and the Wisconsin Card Sorting Test Revised and Expanded (WCST) [32] (only administered at T0). Information on each test variable, including internal consistency and test-retest reliability, are presented in Table 2. The test battery can be administered in approximately 2 h. For all three administrations, the test battery was divided into two sessions (1 h per session) 1 day apart.

Table 2 The neuropsychological test battery

Neuropsychological findings

Table 3 gives an overview of the timeline of the patient’s raw scores and scaled scores on the test battery. Table 4 presents the patient’s norm scores and percentiles on the WMS-III, the WAIS-IV PSI, and the d2-R. Table 5 presents the patient’s WCST scores at 2 weeks after admission (T0). Information on scoring are presented below each of the tables.

Table 3 The patient’s neuropsychological performance raw scores and scaled scores at two weeks after admission (T0), at re-test (T1), and at follow-up (T2)
Table 4 The patient’s norm scores and percentiles on WMS-III, WAIS-IV Processing Speed Index (PSI), and d2-R at two weeks after admission (T0), at re-test (T1), and at follow-up (T2)
Table 5 The patient’s scores on the WCST at two weeks after admission T0

Memory performance on WMS-III

The patient’s scores on WMS-III indicate average to very superior auditory, visual, immediate and general memory performance (108 to 142; Mean: 100), and low average to average working memory (Table 4). The technical manual for WMS-III reports adequate test – retest reliability for all indexes in the age group 16–54 years, except for the Auditory Recognition Delayed Index ( [33]; Table 2). Estimated standard error of difference (SDiff) scores were calculated based on Iverson and Grant ( [34]; Table 2). Differences between the three assessments are outlined here. Her scores on the Auditory Delayed Index decreased more than Sdiff: 6.70 from 132 (very superior) at 2 weeks after admission (T0) to 108 (average) at re-test (T1) and increased again to 132 (very superior) at follow-up (T2). Her scores on the Visual Immediate Index increased slightly more than Sdiff: 6.70 from 118 (high average) at re-test (T1) to 127 (superior) at follow-up (T2). Her scores on the Visual Delayed Index decreased more than Sdiff: 7.65 from 125 (superior) at re-test (T1) to 109 (average) at follow-up (T2). Her scores on the Immediate Memory Index increased more than Sdiff: 3.17 from 134 (very superior) at re-test (T1) to 142 (very superior) at follow-up (T2). Her scores on the Working Memory Index decreased more than Sdiff: 8.22 from 102 (average) at 2 weeks after admission (T0) to 88 (low average) at re-test (T1). The scores on the rest of the indexes did not change more than the estimated Sdiff scores between time points.

Cognitive flexibility on D-KEFS and WCST

Overall, she performed above average on the Verbal Fluency Test (Table 3) at all three test times compared to the normative population for age, except for her performance at re-test (T1) on the switching condition, which was decreased more than Sdiff: 2.42 to average, and the high number of repetition errors (7; below average) at re-test (T1) and (3; average) at follow-up (T2).

She performed average to above average on the Design Fluency Test at all three test sessions (Table 3). However, the switching condition score was lower [6] at follow-up (T2) compared to 8 at 2 weeks after admission (T0) and re-test (T1), though still average.

During follow-up (T2) on the Trail Making Test (Table 3), her performance on the Number-Letter Sequencing test, measuring cognitive flexibility, was below average (111 s), in spite of being average at 2 weeks after admission (T0; 90 s) and re-test (T1; 79 s). The numbers condition was very low at T0 (55 s; below average), improving somewhat at re-test (T1; 46 s; below average) and follow-up (T3; 41 s; below average). We have no explanation for this result. On the other conditions, her performance was average at all three test times on the Trail Making Test.

Her scores on the WCST (Table 5) 2 weeks after admission (T0) place her in the mild to moderately-to-severely range of impairment on cognitive flexibility according to this task. She completed one out of six categories (< 1st percentile). She made 52 perseverative responses (< 1st percentile; standard score 55; moderately-to-severely impaired range). She committed 50 errors (8th percentile; standard score 79: mildly impaired range), of which 36 were perseverative errors (1st percentile; standard score 55: moderately impaired range).

WAIS-IV processing speed

The scores on the Processing Speed Index (Table 4) were average compared to the normative population for age at all three test times. There were no relevant differences between time points. She scored 93 at admission (T0) and re-test (T1) and 98 at follow-up (T2).

d2-R test of attention

At 2 weeks after admission (T0) and re-test (T1), she had a small number of processed targets (426 and 420), 18th to 21st percentile (Tables 3 and 4), her concentration performance was 175 and 176 corresponding to the 42nd percentile and she committed three and no errors respectively (> 90th percentile). At follow-up (T2), her concentration performance was above the mean (185; 54th percentile) but not increased more than Sdiff: 24.89. The total processed targets score was still low (451; 34th percentile), and she committed few errors (four; 90th percentile).

Discussion and conclusions

The patient exhibited average to very superior performance on verbal fluency, design fluency, processing speed, and memory. However, her working memory performance was low average. Her attention and concentration performance were below average to average, and her performance on cognitive flexibility tasks were average to moderately-to-severely impaired.

The present case report demonstrates surprisingly good cognitive performance in a patient with severe and enduring AN with extremely low BMI varying between 7.7 and 9.3 during the study period of 1 year. However, some of her executive functions seem to be impaired. This is in line with previous research on patients with AN [5, 8]. The present results suggest that her working memory was normal (low average) in line with previous studies [35, 36]. However, her working memory performance was lower compared to the rest of her memory performance, which was average to very superior. The results from the D-KEFS indicate average to above-average performance with perhaps somewhat weaker cognitive flexibility (below average to average). On the other hand, the results from the WCST indicate impairment in cognitive flexibility. The overall differences in performance between the three assessments were minimal. This indicates that the minor differences in BMI between the test assessments did not significantly affect her cognitive performance, as expected.

Impaired cognitive flexibility

It could be that impaired cognitive flexibility existed prior to the illness as a premorbid trait as suggested previously [10], or that the malnutrition has affected the patient’s cognitive flexibility. Since we are missing data on her premorbid level, we cannot draw any firm conclusions.

Impaired cognitive flexibility has previously been reported in patients with AN with higher BMI [37], indicating that impairments in cognitive flexibility do not necessarily relate to undernutrition. In patients with AN who had recovered from the illness, cognitive flexibility was in the normal range in this study. However, other studies found that individuals who recovered from AN exhibited more or less impaired executive functioning [10]. Longitudinal research on the relationship between different BMI states and cognitive performance is highly needed.

Impaired cognitive flexibility may also play a role in the perpetuation of AN. Impaired cognitive flexibility has been suggested as a maintenance factor [38] and a factor related to lack of illness insight characteristic of patients with restrictive AN [39]. Lack of illness insight could be related to treatment resistance [40]. The patient’s low scores on EDI-3 subscales also reflect a discrepancy between illness severity and self-reported symptoms. This discrepancy or ambivalence is part of the nature of the disorder reflected in the low motivation for recovery and high number of dropouts from treatment alongside an expressed desire to change [41].

Cognitive adaptation in anorexia nervosa

Survival of long-term starvation is only possible due to extensive adaptive endocrine and metabolic alterations [42]. How these alterations affect cognitive functions still remains to be clarified. Well-designed longitudinal studies on severely underweight patients with a long illness duration are lacking. However, the present case report suggests that essential preservation of some cognitive functions occurs even in extreme chronic semi-starvation.

The mechanisms allowing for such preservation remains a subject of speculation. Links can be made to research on neuroplasticity and functional reorganization of cognitive functions after brain injury since patients with AN have white matter alterations [43]. Research shows that brain maturation processes of especially the prefrontal cortex continue until people are approximately 25 years old [44]. Nutritional status seems to impact this brain maturation [44]. Executive functions associated with the prefrontal cortex could therefore be affected by undernutrition during development of prefrontal connections in the brain in adolescence and young adulthood. Thus, impairment on executive functions may not arise until adulthood in patients with AN. This is in line with research that found no cognitive flexibility impairment in children and adolescents with AN but impairments in adults with AN [19, 20]. The literature indicates that other cognitive functions associated with the prefrontal cortex, such as memory, are also impaired in adults with AN [3]. However, overall, this literature is not as explicit as the literature showing cognitive flexibility impairment in adults with AN. The ambiguity in the literature indicates differences between cognitive functions related to the prefrontal cortex in patients with AN. It might be that some prefrontal connections potentially being affected during low weight in adolescence could be reorganized or “compensated for” with time as is possible with reorganization or apparent functional recovery after brain injury [45]. In that case, cognitive performance could be regained after impairment has occurred. Some dimensions of cognitive flexibility might, however, be more difficult to compensate for. This could explain specific cognitive flexibility impairment in patients recovered from AN [10] and explain that the patient in the present case report performed normal and superior on some functions associated with prefrontal connections (memory and verbal fluency) but poorer on cognitive flexibility. We therefore suggest that reorganization of some cognitive functions can occur in spite of persisting low weight in patients with AN. In line with the possibility of cognitive reorganization in AN, Cognitive Remediation Therapy seems to improve executive functioning in patients with AN [46]. The suggested theory of cognitive adaptation may therefore not be specific to persisting low weight in AN. However, fast, substantial weight-loss could affect cognitive performance differently than persisting low weight. Therefore, studies on starving healthy subjects, including the Minnesota Semi-Starvation Experiment [16], could show different results than studies on patients with severe and enduring AN. Likewise, studies on patients with short illness duration might find different results than studies of patients with enduring AN. It is also unclear if patients developing AN in adulthood will display the same cognitive impairments. In line with these reflections, a case report of a 27-year-old Japanese woman in a coma, with BMI of 8.5 kg/m2 at admission, describes a patient with AN where the outcome of severe malnutrition was persistent neurologic sequelae [47]. The woman developed AN at the age of 21 years where the patient in the present case report was diagnosed at the age of 10 years. The difference in age of onset, duration of illness, and/or manner of weight-loss (fast, substantial weight-loss compared to persisting low weight) may have resulted in different outcomes for the women. It is, however, also a possibility that the patient in the present case report might have an extreme phenotype which enables her to perform well in spite of her being extremely underweight.

We cannot say how high the patient’s scores on the neuropsychological test battery might be if she had not been as malnourished. We assume the patient would perform better on cognitive flexibility tasks, that her processing speed and working memory would be higher, and that she would be able to concentrate better had she not been malnourished. This is somewhat supported by previous research. Although the literature suggests impaired cognitive performance in patients with AN, the reported impairments were limited compared to healthy subjects [8, 48]. Furthermore, it may be that severely underweight patients with AN have a higher verbal IQ [49], which does not, however, exclude the possibility of specific cognitive impairments [50]. This could explain the patient’s high memory performance (and probably global IQ) alongside specific impairment in cognitive flexibility on the WCST. This case may therefore not differ from other patients with severe AN regarding cognitive performance. It may be that the superior performance related to some cognitive functions is a trait of severely underweight patients with AN and/or that a cognitive adaptation to enduring AN increases performance to the premorbid level. In this case, (regained) superior performance of some cognitive functions (i.e. memory and verbal IQ) can exist alongside cognitive impairment in others (i.e. cognitive flexibility). This may change our view of the cognitive profile and its development in patients with severe and enduring AN.

Regardless, the fact that we were able to test the patient in the present case, raises a discussion as to whether she and others with extremely low weight may be responsive to psychotherapy as well. In the present case, the patient underwent psychotherapy for several years albeit without any impact on her weight. More research focusing on the validation of neuropsychological tests including investigation of the practice effect in this patient population is needed.

The individual scores on neuropsychological tests should always be interpreted with care. Factors other than persisting low weight may affect neuropsychological performance (e.g. dehydration, stress, depression, and anxiety). In the present case, the patient did express depressive symptoms corresponding to moderate depression, which might have influenced results on impairment in cognitive flexibility. Furthermore, the patient might experience other issues related to cognitive performance in daily life, which cannot be discovered in a neuropsychological assessment context.

Obviously, conclusions can never be drawn from one case. However, since the neuropsychological testing included a broad range of tests and was repeated three times during a year, the present case report is valid as a basis for reflecting on the affected individual’s cognitive performance at this stage. The present case report demonstrates that cognitive functions may be largely preserved under extreme chronic malnutrition or that cognitive functioning may be regained (reorganized) in spite of extreme chronic malnutrition. More research on patients with AN with extremely low BMI (< 10) is needed to determine whether cognitive performance is affected by starvation and malnutrition.