Functional MRI investigation of verbal working memory in adults with anorexia nervosa

European Psychiatry 29 (2014) 211–218

Available online at

www.sciencedirect.com

Original article

Functional MRI investigation of verbal working memory in adults with anorexia nervosa N.P. Lao-Kaim a,1, V.P. Giampietro b,1, S.C.R. Williams b,c, A. Simmons b,c, K. Tchanturia a,* a

King’s College London, Institute of Psychiatry, Department of Psychological Medicine, London, United Kingdom King’s College London, Institute of Psychiatry, Department of Neuroimaging, SE5 8AF London, United Kingdom c NIHR Biomedical Research Centre for Mental Health at South London, Maudsley NHS Foundation Trust, Institute of Psychiatry, King’s College London, London, United Kingdom b

A R T I C L E I N F O

A B S T R A C T

Article history: Received 11 April 2013 Received in revised form 10 May 2013 Accepted 22 May 2013 Available online 11 July 2013

Literature regarding verbal working memory (vWM) in anorexia nervosa (AN) has been inconsistent due to a misunderstanding of the key components of vWM and introduction of confounding stimuli. Furthermore, there are no studies looking at how brain function in people with AN relates to vWM performance. The present study used functional magnetic resonance imaging (fMRI) with a letter n-back paradigm to study the effect of increasing vWM task difficulty on cortical functioning in the largest AN sample to date (n = 31). Although the AN group had low BMI and higher anxious and depressive symptomology compared to age-matched controls (HC), there were no between-group differences in accuracy and speed at any task difficulty. fMRI data revealed no regions exhibiting significant differences in activation when groups were compared at each difficulty separately and no regions showing group x condition interaction. Although there was a trend towards lower accuracy as duration of illness increased, this was not correlated with activity in regions associated with vWM. These findings indicate that vWM in AN is as efficient and performed using the same cognitive strategy as HC, and that there may not be a need for therapies to pursue remediation of this particular neurocognitive faculty. Crown Copyright ß 2013 Published by Elsevier Masson SAS. All rights reserved.

Keywords: Eating disorders Anorexia Working memory Neuropsychology fMRI Executive function

1. Introduction Anorexia nervosa (AN) is a disorder characterised by extreme dietary restraint, abnormal psychosocial functioning and poor weight-related physical morbidity. With a lack of empirical support for behavioural and pharmacological treatment of AN, further research is needed to improve understanding and development of successful interventions [65]. People with AN exhibit inefficient visuo-spatial memory and processing [36], cognitive flexibility [63], long-term memory [48], attention [17] and central coherence [38]. However, the results of studies assessing verbal working memory (vWM) in AN have been inconsistent (Table 1). WM is a limited capacity store through which information is held, maintained and manipulated in order to plan and carry out behaviour [43] and to facilitate complex processes such as comprehension, learning and reasoning [3] (Fig. 1).

* Corresponding author. PO59, King’s College London, Institute of Psychiatry, Department of Psychological Medicine, De Crespigny Park, SE5 8AF London, United Kingdom. Tel.: +44 0 207 848 0134; fax: +44 0 207 848 0182. E-mail address: [email protected] (K. Tchanturia). 1 Both authors contributed equally to the manuscript.

Specific cortical regions within the fronto-parietal brain network have been found to control these vWM mechanisms [49]. With relation to Baddeley and Hitch’s model [3], the left posterior parietal cortex (PP) has been implicated in phonological storage, the left inferior frontal gyrus (IFG) in sub-vocal rehearsal and the central executive is thought to reside in the dorsolateral prefrontal cortex (DLPFC) [60]. Focal lesions in these areas give rise to vWM process-specific deficits [45] and recent neuroimaging experiments have found these same regions to show consistent abnormalities in people with AN. Voxel-based morphometric analysis of volumetric magnetic resonance imaging (MRI) [1], has localised decreases in cortical grey matter to the parietal cortex [14,25], and cerebellum [8]. The cerebellum is traditionally associated with fine motor co-ordination but has also been implicated in facilitating initial phonological encoding of verbal information [53]. Conversely, one study found a significant increase in grey matter volume in the DLPFC [8]. A timely review of positron emission tomography (PET), single photon emission computer tomography (SPECT) and functional MRI (fMRI) studies, has demonstrated a consensus of functional ‘‘disturbance’’ in the IFG, DLPFC and inferior parietal lobule (IPL) [50]. Effects between binge/purge and restrictor subtypes were similar, but laterality towards the left hemisphere for IPL

0924-9338/$ – see front matter . Crown Copyright ß 2013 Published by Elsevier Masson SAS. All rights reserved. http://dx.doi.org/10.1016/j.eurpsy.2013.05.003

N.P. Lao-Kaim et al. / European Psychiatry 29 (2014) 211–218

212

Table 1 Results of a systematic review of vWM in people with anorexia nervosa (AN). PubMed, ScienceDirect and ISI Web of Knowledge were searched using the terms ‘Anorexia’ and ‘memory’. Only cross-sectional studies including at least 1 AN group and 1 healthy control (HC) group and which provided test results pertaining specifically to verbal shortterm or verbal working memory rather than ‘‘global working memory’’ were included. Tests of ‘‘working memory’’ were scrutinised and where appropriate, reappraised for classification within this table. In particular, the 1-Back task was classified as a measure of short-term retention rather than vWM as it lacks any requirement for manipulation of information [60]. Further details of each study can be found in 1 (R) = AN Restrictive subtype; (B/P) = Binge-Purge subtype.

Verbal short-term memory (vSTM)

Verbal working memory (vWM)

Direction

No. of studies

Authors

Combined number of participants

AN better than HC

1

Hatch et al. (2010) [29]

No difference between AN and HC

7

AN worse than HC

3

Castro-Fornieles et al. (2010) [15] Bosanac et al. (2007) [5] Key et al. (2006) [35] Kemps et al. (2006) [34] Mathias and Kent (1998) [41] Szmukler et al. (1992) [61] Witt et al. (1985) [66] Castro-Fornieles et al. (2009) [14] Green et al. (1996) [27] Kingston et al. (1996) [36]

AN = 37 HC = 45 AN = 133 HC = 130

AN better than HC

2

No difference between AN and HC

1

Brooks et al. (2012) [9] Dickson et al. (2008) [17] Nikendei et al. (2011) [48]

AN worse than HC

1

Seed et al. (2002) [57]

hyper-responsiveness and abnormal activity in IFG was mainly attributable to the restrictive subtype, with the binge/purge subtype showing bilateral disturbance in the parietal cortices and no evidence of abnormally functioning IFG. In light of these abnormalities, it is surprising that some AN studies report superior vWM performance in comparison to healthy control participants (HC) who have no prior history of psychiatric illness (Table 1). One may hypothesise that the central executive compensates for the impaired phonological store, either directly or via the visuo-spatial sketchpad [54], the WM ‘‘slave’’ system dealing with visual, spatial and kinaesthetic information [3]. This would be in line with increased DLPFC grey matter volume [8] and fMRI studies showing that the DLPFC and IFG exhibit higher activation when healthy people undertake dieting behaviour [28,32]. It may also be that AN participants view vWM tasks as more of a ‘‘challenge’’ than simple verbal short term memory (vSTM) tasks and delegate more attentional resources to their completion [55]. For example, Dickson et al. [17] observed significantly fewer errors by AN participants compared to HC at higher loads of the N-Back

Fig. 1. Schematic diagram of the working memory model. The term ‘‘working’’ implies involvement of both short-term stores (phonological loop, visuo-spatial sketchpad and episodic buffer) and mechanisms that process the stored items [16]. With respect to vWM (highlighted in black), short-term retention is sub-served by the phonological loop, which is comprised of a phonological store that holds verbal information and a sub-vocal rehearsal component that maintains this information by preventing trace decay. The central executive is the attentional control mechanism that manipulates items held within the phonological loop [3]. Adapted from Baddeley, 2000 [2].

AN = 70 HC = 67

AN = 37 HC = 44 AN = 34 (R), 19 (B/P) HC = 30 AN = 20 HC = 20

task in conjunction with subliminally presented food, aversive and neutral stimuli, but no difference at lower WM loads. It would be useful to elaborate on these possible theories by examining associations between cortical activation and accompanying clinical demographics and performance data. To our knowledge, only one study has used fMRI with a vWM task in people with AN, without the inclusion of disorder-relevant stimuli. Castro-Fornieles et al. [15] found that whilst adolescent AN showed similar performance to age-matched controls in a verbal digit N-Back task, they exhibited increased activation in the left superior parietal lobule, an additional vWM area reportedly involved in the manipulation of information [37] and left inferior temporal gyrus, during fMRI examination. Despite the importance of adolescent data, problems exist in analysing images from young cohorts, due to continuation of functional maturation [19]. This could explain failure to find differences in frontal executive and rehearsal regions and increased activation in the inferior temporal lobe. This area is not commonly activated in this task [49] but responds in WM paradigms during maintenance of visual objects [52], matching of an actively maintained object to a current object [18] and presentation of novel stimuli [42]. In addition, Castro-Fornieles et al. [15] only used the 1-Back condition, which is more representative of a vigilance task assessing attention and storage mechanisms and does not adequately tax prefrontal manipulation mechanisms [60]. Whereas Braver et al. [7] found a positive linear relationship between task load and hyperactivation in the DLPFC, left IFG, ACC and bilateral parietal cortices others indicate that increased activation of the DLPFC only occurs at higher cognitive loads, such as the 2-Back and 3-Back [60]. The current study aims to extend the findings of CastroFornieles et al. [15] by including conditions of increasing difficulty to comprehensively assess the effect of cognitive load on vWM. We use an exploratory whole-brain approach and include the largest AN fMRI sample to date, which will increase statistical robustness and enable a comparison of adult and adolescent groups with Castro-Fornieles et al. [15]. In light of the current literature regarding the N-Back task [17,9], we firstly hypothesise that the AN group will show superior

N.P. Lao-Kaim et al. / European Psychiatry 29 (2014) 211–218

213

performance compared to the HC group and that this difference will be associated with functional brain alterations in vWM regions. We secondly hypothesise activity in independently defined vWM loci to be correlated with relevant clinical measures [10]. 2. Subjects and methods 2.1. Participants Sixty-six participants took part in this study. Two HC participants were excluded due to anti-depressant use and one for scoring above the clinical threshold (> 10) on the anxiety subscale of the Hospital Anxiety and Depression Scale (HADS-a). Remaining HC participants were propensity matched [31,30] by age, making the final sample thirty-one AN (Restrictive = 24; Binge/Purge = 7) (age range, 18–46; years of illness, 1–35) and thirty-one HC (age range, 22–34) (Table 2). AN participants were recruited from the South London and Maudsley eating disorder service and from a community sample via the B-EAT website (http://www.b-eat.co.uk/) (inpatient = 10; outpatient = 10; daycare = 6; community = 5) and had been diagnosed by clinicians as fulfilling DSM-IV criteria for AN. Sixteen AN were on one or more types of psychoactive medication at the time of study (Supplementary Table 2). A further three participants were regularly taking osteoporotic, anti-asthmatic, anti-muscarinic and thyroid hormone replacing medications. Exclusion criteria for both groups included a history of brain trauma or neurological problems (e.g. epilepsy), suspected/known pregnancy, claustrophobia, inadequate English proficiency, colour blindness, non-corrected visual impairment and presence of metallic implants. HC participants were further excluded if they had a personal history or first degree relative with an eating disorder or psychiatric illness, if body mass index (BMI) was less than 18.5 and/or if EDE-Q score was greater than 3. This study was carried out in accordance with the Declaration of Helsinki under approval of the National Research Ethics Committee, London (Ref 11/LO/0952). All participants gave written informed consent and were reimbursed for their time. 2.2. Clinical and behavioural assessment All participants were measured for height and weight in order to calculate their BMI. IQ was estimated using the National Adult Reading Test Revised (NART-R) [47]. General screening of current psychopathology was performed using the Structured Clinical Interview for DSM-IV (SCID-I) [21]. The Hospital Anxiety and Depression Scale (HADS-a/HADS-d) [67] indicated the presence of anxious and/or depressive symptoms. To comprehensively profile eating-disordered behaviour, the Eating Disorder Examination Questionnaire (EDE-Q) [20] was administered, which has been shown to have good concurrent and criterion validity [44]. The 5-item Work and Social Adjustment Scale (WSAS) [46] was used as a measure of day-to-day functional status and the Cognitive Flexibility Scale (CFS) [40] was used to assess the ability to disengage with an irrelevant task and re-engage with a relevant task. 2.3. N-Back task The current study used a verbal variant of the N-Back task, modified from the original paradigm described by Braver et al. [7], which has been shown to exhibit good test-retest reliability [51]. Stimuli were letters and the input modality was visual. A block design was used with three levels of increasing difficulty (1-Back, 2-Back, 3-Back) and a vigilance baseline condition (0-Back) (Fig. 2).

Fig. 2. Participants were presented with a series of white capital letters (A–Z) on a black background on a projection screen, viewed using a prism mounted on the headcoil and were required to press a button with their right index finger when the letter they saw was identical to the letter they saw n trials previously (where n = 1, 2 or 3). For the control condition (0-Back), participants were required to press the button when they saw the letter ‘‘X’’.

Three blocks of each condition were presented (12 blocks in total) in the following pseudorandom order: 2-Back/1-Back/3Back/0-Back/1-Back/3-Back/0-Back/2-Back/1-Back/2-Back/0Back/3-Back. Each block consisted of 14 letters presented in pseudorandom order for 2000ms each. Within each block, there were three ‘‘target’’ stimuli requiring a response from the participant. The remaining 11 letters (‘‘non-targets’’) did not require a response. Written instructions on the screen preceded each block and lasted for 3 s, making the duration of each block 31 s and the total duration of the N-Back task, 372 s. Reaction times (RT) were recorded in milliseconds for trials for which a response was made and whether these responses corresponded to a target or non-target letter. RT was calculated as the latency between stimulus onset and participant response. 2.4. fMRI acquisition fMRI data were acquired on a 1.5-Tesla GE Signa HDx system running 14m5 software (General Electric Medical Systems, Wisconsin) at the Centre for Neuroimaging Sciences of the Institute of Psychiatry, King’s College London. A body coil was used for radio frequency (RF) transmission and an 8-channel head coil used for RF reception. T2*-weighted gradient echo echoplanar images (GE-EPI) depicting blood-oxygen-level-dependent (BOLD) contrast were acquired on an axial plane, parallel to the anterior commissure–posterior commissure (AC–PC) line, with the following parameters: repetition time (TR) = 2000 ms, echo time (TE) = 40ms, flip angle = 908, slice thickness = 5 mm, slice gap = 0.5 mm, field of view (FoV) = 24  24 cm, matrix size = 64  64. Whole-brain coverage was acquired with 25 slices and 186 T2*-weighted whole-brain volumes were acquired for each participant. To facilitate normalisation to standard space, we acquired a highresolution GE-EPI (TR = 3000 ms, TE = 40 ms, flip angle = 908, slice thickness = 3 mm, slice gap = 0.3 mm, FoV = 24  24 cm, matrix size = 128  128) with slices parallel to the AC-PC line. One wholebrain volume was acquired consisting of 43 slices. Data quality was assured using an automated quality control procedure [59]. 2.5. Demographic, clinical and performance analysis Three N-Back performance measures were derived for each participant in each vWM condition. Accuracy scores were

214

N.P. Lao-Kaim et al. / European Psychiatry 29 (2014) 211–218

calculated as the number of errors made per condition and were further subdivided into omission (no response to a target) and commission errors (response to a non-target). Mean RT was calculated using only RTs for trials for which a correct response was made. Demographic, clinical and performance data were analysed using Statistical Package for the Social Sciences (SPSS 20) [33]. Shapiro-Wilk tests were first used to assess normality in the demographic, clinical and performance data. In cases where normality for both groups could be assumed, parametric between-groups comparisons were conducted using Student’s ttests. When the assumption of normality was violated, nonparametric between-groups comparisons were conducted using Mann-Whitney U and within-groups comparisons using Friedman’s and Wilcoxon tests. Correlations were conducted for the AN group to test associations between performance measures and demographic/ clinical scores. Spearman’s P was used for correlations WSAS, CFS, HADS-a, HADS-d and EDE-Q due to the ordinal characteristics of Likert scale scoring methods. All other correlations were conducted using Pearson’s r. 2.6. fMRI analysis The fMRI data were analysed with software developed at the King’s College London Institute of Psychiatry (XBAM version 4.1), which utilises a non-parametric permutation-based approach to minimize assumptions and reduce the effect of outliers (c.f. http:// brainmap.it). Data were first processed [11] to minimize motion related artefacts. Following realignment, images were smoothed using an 8.8 mm full-width half-maximum Gaussian filter, chosen to improve signal-to-noise ratio over the spatial neighbourhood of each voxel. Responses to each vWM condition were then detected by timeseries analysis using a linear model in which each component of the experimental design was convolved separately with a pair of Poisson kernels (l = 4 and 8 s) to allow variability in the haemodynamic delay. The best fit between the weighted sum of these convolutions and the time-series at each voxel was computed using the constrained BOLD effect model [23]. A goodness of fit statistic was then computed as the ratio of the sum of squares of deviations from the mean image intensity resulting from the model (over the whole time-series) to the sum of squares of deviations resulting from the residuals (SSQ ratio). Following computation of the observed SSQ ratio at each voxel, the data were permuted by the wavelet-based method described in Bullmore et al. [13]. The observed and permuted SSQ ratio maps for each individual were transformed into the standard space of Talairach and Tournoux [62] using a two-stage warping procedure. Group maps of activated clusters were then computed using the median SSQ ratio at each voxel (over all individuals) in the observed and permuted data map [6]. Computing intra and inter participant variations in effect separately constitutes a mixed effect approach, which is desirable in fMRI. Detection of activated voxels was extended from voxel to 3D cluster-level using the method described by Bullmore et al. [12]. Resulting cluster-level maps were then thresholded to ensure less than 1 expected type I error cluster per map. Unlike Bonferroni-based multiple comparison correction procedures, use of the false discovery rate (FDR) ensures adequate control over the type I error rate whilst preventing over-inflation of type II error [26]. 2.6.1. Group and cognitive load comparisons Comparisons of responses between-groups at each vWM condition and comparisons of responses between conditions for

each group separately were performed by fitting the data at each intracerebral voxel at which all participants have non-zero data using the linear model: Y ¼ a þ bX þ e where ‘Y’ is the vector of SSQ ratios for each individual, ‘X’ is the contrast matrix for the inter-group/inter-condition contrast, ‘a’ is the mean effect across all individuals in the groups/conditions, ‘b’ is the computed group/condition difference and ‘e’ is a vector of residual errors. The model is fitted by minimising the sum of absolute deviations to reduce outlier effects. The null distribution of ‘b’ is computed by permuting data between-groups/conditions and refitting the above model 50 times at each voxel and combining the data over all intracerebral voxels.The interaction between group membership and vWM condition was analysed using a split-plot analysis of variance (ANOVA) model in order to ascertain whether any brain regions exhibit group-dependent differences in within-group monotonic trends. All resulting 3D cluster-level maps were then computed to yield less than 1 expected type I error cluster by appropriate thresholding of the null distribution.Correlational analysis Correlational analyses were performed on three main regions of interest (ROI), identified as being integral to vWM functioning [60]. ROIs for the right DLPFC [41,33,29], right PP [12, –60, 43] and left IFG [–48, 9, 10] were defined independently from a large activation-likelihood estimation (ALE) meta-analysis of N-Back literature [49]. Others have indicated a hemispheric bias towards the left PP during vWM tasks, however, Owen et al. [49] only reported likely activation in the corresponding right homolog. An additional ROI was therefore included, based on the coordinates of a left IPL [–34, –46, 26]. Of the two inferior frontal clusters reported by Owen et al. [49], the one with the highest ALE was chosen as an ROI. Masks with spherical ROIs of 5 mm diameter constructed around each of the coordinates were then created to extract the mean SSQ ratios for each participant in each condition. SPSS was then used to perform correlations between mean ROI SSQ ratios at each load level and clinical and demographic measures that significantly correlate with performance. 3. Results 3.1. Demographic and clinical measures AN and HC groups did not differ significantly in age; AN had significantly lower BMI, IQ and years of education in comparison to HC. For the questionnaire measures, AN subjects had significantly higher levels of depression, anxiety, work and social adjustment difficulties, eating-disordered behaviour and cognitive inflexibility (Table 2). 3.2. Task performance To test the hypothesis that the AN group would perform better than the HC group on the N-Back task, between-groups MannWhitney U tests were conducted on the performance data. Table 3 shows no significant difference between groups on all performance measures on any of the vWM conditions. Friedman’s tests were carried out to determine the effect of increasing vWM load on performance. There were significant differences in mean RT (HC: x2 (3) = 58.02, P < 0.001; AN: x2 (3) = 46.33, P < 0.001), commission (HC: x2 (3) = 25.44, P < 0.001; AN: x2 (3) = 30.59, P < 0.001) and omission (HC: x2 (3) = 64.6, P < 0.001; AN: x2 (3) = 58.99, P < 0.001) error scores across vWM conditions.

N.P. Lao-Kaim et al. / European Psychiatry 29 (2014) 211–218

215

Table 2 Demographic and clinical measures of AN and HC groups. AN (n = 31)

Current BMI IQ (NART)a WSAS totalb HADS-a

SD

Mean

SD

16.0 110.6 26.8 15.0

1.6 5.3 6.4 3.8

21.9 115.3 – 4.1

1.8 6.9 – 2.9

Median Age Years of Educationc EDE-Q Global HADS-d CFS total Years of illnessd

HC (n = 31)

Mean

23.0 16.0 4.0 11.0 42.0 6.0

Inter-quartile range 9.0 3.0 1.4 8.0 11.0 9.0

Statistic (df)

P-value

t (60) = 13.735 t (56.313) = 3.000 n/a t (60) = 12.747

< 0.001 .004 n/a < 0.001

Median

Inter-quartile range

Statistic (df), Z-score

P-value

25.0 18.0 0.5 0.0 61.0 –

3.0 3.0 0.6 1.0 5.0 –

U (1) = 369.5, Z = –1.569 U (1) = 227.5, Z = –3.592 U (1) = 3.0, Z = –6.723 U (1) = 10.5, Z = –6.710 U (1) = 55.5, Z = –5.989 n/a

.117 < 0.001 < 0.001 < 0.001 < 0.001 n/a

Score Ranges: HADS (0–21); EDE-Q (0–6); CFS (0–72); WSAS (0–40). –: all participants scored 0 in this field; AN: anorexia nervosa; HC: healthy control. a One AN participant did not have an entry for IQ, therefore an entry was made by taking the mean IQ value of the AN group. b Statistic not calculated as all HC participants scored 0 on the total WSAS. c One HC participant did not have an entry for years of education, therefore an entry was made by taking the mean of all HC participants of the same education level (postgraduate). d Four AN participants did not report their duration of illness, therefore statistics are reported for the remaining 27 participants.

For both groups, as vWM load increased, mean RT and accuracy increased (Table 3). Pair-wise Wilcoxon post-hoc comparisons between adjacent vWM conditions specified common significant differences in which both groups had longer RTs and made more omission errors for the 2-Back compared to the 1-Back and for the 1-Back compared to the 0-Back (Table 4). Whereas the AN group showed no further differences between conditions, the HC group additionally showed significantly longer mean RT and made significantly more commission errors on the 3-Back compared to the 2-Back condition (Table 4). To test whether presence of psychoactive medication affected task performance, the AN group were first split into medicated (n = 16) and non-medicated (n = 15) sub-groups. Sub-groups were then compared on mean RT, omission and commission error measures at each level of the N-Back task using Mann-Whitney U tests. All tests failed to reach significance at the uncorrected threshold of P < 0.05. 3.3. Correlations with demographic/clinical measures For AN, no significant correlations were found between demographic and clinical scores (Age, BMI, IQ, years of education,

years of illness, WSAS, HADS-a, HADS-d, CFS, EDE-Q) and task performance measures (mean RT, omission, commission errors) at any level of the N-Back after correction for multiple comparisons (corrected thresholds; mean RT = 0.00125, omission errors = 0.00167, commission errors = 0.00167). However, there was a trend towards a positive correlation between length of illness and commission error scores on the 3-Back condition (r = 0.525, n = 27, P = 0.005). 3.4. fMRI results 3.4.1. Task effects Cortical regions that followed a linear trend in activation as task difficulty increased were identified by performing trend analyses for each group independently. Both groups exhibited clusters showing positive trends (3-Back > 2-Back > 1-Back) in the bilateral IPL (BA 40), bilateral middle and superior frontal gyri extending into the DLPFC (BA 6/9/10), left precuneus (BA 7) and right insula (BA 13). The AN group additionally showed positive trends in the left middle temporal gyrus (BA 21), right precuneus (BA 7) and left IFG (BA 45) (Supplementary Table 3).

Table 3 Between-groups (AN, HC) Mann-Whitney U comparisons of mean RT and accuracy scores for all vWM conditions (0-Back, 1-Back, 2-Back and 3-Back). The threshold for statistically significant results was adjusted for multiple comparisons using Bonferroni correction to 0.0125 (0.05/4). N-back condition

AN (n = 31)

HC (n = 31)

U (1) statistic

Z-score

Uncorrected P-value

Median

IQR

Median

IQR

Mean RT 0-Back 1-Back 2-Back 3-Back

537.2 575.1 630.0 714.0

106.8 200.8 203.9 216.8

503.8 535.1 619.1 708.2

101.6 122.0 134.4 213.5

359.5 364.0 413.0 437.0

–1.704 –1.640 –0.950 –0.612

.088 .101 .342 .540

Commission errors 0-Back 1-Back 2-Back 3-Back

0 – 0 1

0 – 0 1

– 0 0 0

– 0 0 1

449.5 449.5 418.0 426.0

–1.426 –1.426 –1.717 –0.853

.154 .154 .086 .394

Omission errors 0-Back 1-Back 2-Back 3-Back

– 0 1 2

– 0 2 3

– 0 0 1

– 0 1 2

480.5 418.0 335.5 379.0

.00 –1.717 –2.222 –1.490

1.00 .086 .026 .136

–: all participants scored 0 in this field; RT: reaction time (ms).

N.P. Lao-Kaim et al. / European Psychiatry 29 (2014) 211–218

216

Table 4 Within-groups Wilcoxon pair-wise comparisons of mean RT and accuracy scores between experimentally adjacent working memory conditions (0-Back, 1-Back, 2-Back, 3-Back). The threshold for statistically significant results was adjusted for multiple comparisons using Bonferroni correction to 0.01667 (0.05/3). Group

Measure

Comparison

Z-score

P-value (uncorrected)

AN

Mean RT

0-Back 1-Back 2-Back 0-Back 1-Back 2-Back 0-Back 1-Back 2-Back

– – – – – – – – –

1-Back 2-Back 3-Back 1-Back 2-Back 3-Back 1-Back 2-Back 3-Back

–4.076 –3.978 –1.205 –1.414 –2.121 –2.372 –2.121 –3.146 –3.606

< 0.001a < 0.001a .228 .157 .034 .018 .034 .002a < 0.001a

0-Back 1-Back 2-Back 0-Back 1-Back 2-Back 0-Back 1-Back 2-Back

– – – – – – – – –

1-Back 2-Back 3-Back 1-Back 2-Back 3-Back 1-Back 2-Back 3-Back

–3.253 –3.861 –2.508 –1.414 –0.577 –3.002 –1.000 –3.066 –4.044

.001a < 0.001a .012a .157 .564 .003a .317 .002a < 0.001a

Commission errors

Omission errors

HC

Mean RT

Commission errors

Omission errors

AN: anorexia nervosa; HC: healthy control; RT: reaction time (ms). a Comparisons yielding P-values that fall below the corrected threshold.

Clusters showing negative trends (1-Back > 2-Back > 3-Back) were found in both groups in the right medial frontal gyrus (BA 10) and left cingulate gyrus (BA 30/31). For AN, negative trends were also found in the left postcentral gyrus (BA 2) and left superior frontal gyrus (BA 8/9) whereas for HC, negative trends were found in the right postcentral gyrus (BA 3) and bilateral superior frontal gyri (BA 9). 3.4.2. Between-groups comparisons To test the hypothesis that areas of the vWM network show differential activation between AN and HC, between-groups comparisons were conducted at each level of the N-Back task. There were no cortical regions showing significant differences in activity between-groups at any level of the N-Back (1-Back, 2Back, 3-Back). These results remained unchanged after adding years of education, IQ and illness duration as covariates. 3.4.3. Group x vWM condition interaction To test whether cortical regions respond differently between groups to a parametric increase in vWM load, a 2  3 interaction, with group as a between-groups measure and condition as a repeated measure, was performed. There were no regions exhibiting a significant interaction effect. As with the previous between-groups comparisons, co-varying years of education, IQ and illness duration had no effect on the results. 3.4.4. Correlational analysis To test whether vWM-related activity in the right DLPFC, right PP, left IFG and left IPL correlated with years of illness in AN, mean SSQ values were first extracted from the interaction analysis for each individual using the ROI masks as described in methods, then subjected to Pearson’s r tests. After correction for multiple comparisons (corrected threshold = 0.0167), no significant correlations were found in any of the four ROIs at any level of the N-Back task. Of note however, is that the only two tests yielding P-values below the uncorrected threshold (P < 0.05) were between years of illness and DLPFC activity in the 2-Back (r = –0.402, n = 27, P = 0.038) and 3-Back (r = –0.386, n = 27, P = 0.047) conditions.

4. Discussion The present study found that despite adult AN participants having low BMI and high anxious and depressive symptoms, vSTM and vWM performance was similar to age-matched HC on measures of both speed and accuracy. Furthermore, there was no difference in whole-brain functional activation during the vWM task and although illness duration was closely related to error scores, this factor did not correlate significantly with brain regions widely accepted to be involved in vWM processes. Other studies have shown that people with AN are just as capable at temporary retention of verbal information (vSTM) as HC participants [5,15,34,35,41,61,66]. Kemps et al. [34] also showed that AN exhibit poor central executive function, as they are unable to combine verbal and visuo-spatial STM to the same level as controls. When components were tested individually, people with AN performed as well as HC participants when recalling the names of objects successively shown on a monitor whereas they were poorer at recalling their locations. Inefficient visuo-spatial STM could therefore be contributing to the observed inefficient integrative function of the central executive. Key et al. [35] found no difference in global digit span, although others have indicated lower performance in AN compared to HC [14]. Many studies regularly report global digit span but in doing so lose important data pertaining to separable WM components, thereby making their contributions difficult to interpret. CastroFornieles et al. [15] used a similar task to ours in adolescents and reported no difference in performance, although the efficacy of the 1-Back task in assessing vWM is debateable [60]. Taken alongside the current study, both adolescent [15] and adult AN perform to normal standards with regards to vSTM, but are accompanied by different functional profiles. AN adolescents exhibit increased activation in the left inferior temporal and right superior parietal cortices whereas our study in adults found no difference on the 1-Back task, suggesting that adolescents require additional resources to perform to the normal standard. However, Castro-Fornieles et al. [15] employed an ROI approach to the analysis, which lowers the threshold for statistically significant results and limits the regions one wishes to assess. Since there are very few cognitive imaging studies in AN, not involving disorderrelevant stimuli, we felt that our whole-brain approach was necessary to inform the field of global function albeit limiting the parallels drawn between the two studies. With regards to vWM and in accordance with the current study, a recent neuropsychological assessment found no difference on isolated backwards digit scores between restricting, binge-purging, weight-gained AN subtypes and HC [48]. Others have found slower performance on a memory scanning task requiring a response to pre-specified numbers when they appear within a sequence [57] and conversely, two recent studies have reported superior vWM performance in AN. Dickson et al. [17] administered the N-Back task (1-Back, 2-Back) whilst presenting aversive, neutral or food stimuli either subliminally or supraliminally and found a two-way interaction in which, regardless of NBack image valence, AN participants made less errors than HC participants when images were presented subliminally and more errors when presented supraliminally. Brooks et al. [9] extended these findings using a similar protocol and found that people with AN made less errors compared to controls when subliminal presentation included either neutral or aversive stimuli but that this superiority was abolished when presenting subliminal food stimuli. These findings are important for understanding how cognition is affected by implicit visual processes and present a paradigm that has useful diagnostic value, however, vWM performance without the presence of additional stimuli was not assessed. This

N.P. Lao-Kaim et al. / European Psychiatry 29 (2014) 211–218

complicates interpretation as there is no neutral comparison against which to test performance. In combination with the current study, it appears that efficiency of ‘‘raw’’ vWM is equal between AN and HC groups and the differences in subliminal studies may be driven by a greater negative effect on the HC group. This raises questions regarding implicit processing, a relatively unexplored area that merits attention given that people with AN exhibit deficits in implicit learning [58]. Our imaging findings only strengthen the conclusion that in the absence of distracting stimuli, vWM functioning in AN is similar to HC, as there is no evidence of any compensatory mechanism by which vWM performance could have been sustained. With this basis for comparison, the field would benefit from future studies looking at how implicit processing biases using disorder-relevant/irrelevant subliminal/supraliminal imagery affect brain function. Brooks et al. [9] also found that AN participants made more errors when the task demanded more resources but only when food stimuli were presented. Although there were no betweengroup differences, the current study indicates that AN appear not to perceive the increase in difficulty on conditions requiring the highest cognitive effort whereas HC participants do. Indeed, the literature has pointed to a phenomenon in which people with AN display enhanced performance when tasks require more effort [24]. This performance enhancement comes at the expense of automatic processing and learning and is perhaps related to the fact that people with AN are known to be cognitively inflexible [63]. 5. Conclusions, limitations and future research The current study informs the development of current therapeutic intervention strategies. In particular, cognitive remedial therapy (CRT) aims to improve daily functioning by increasing neurocognitive ability in several key domains [64]. This study, benefitting from a large study sample, informed approach to a robust vWM paradigm, rigorous non-parametric statistical methods and the combination of behavioural and functional imaging data, suggests that it may not be necessary to pursue remediation of vWM in CRT. Although more work is required to assess the neural correlates of poor visuo-spatial WM ability [34,35], CRT should focus on other neurocognitive domains known to be affected in AN, such as cognitive flexibility [63] and attention to detail [22]. Although a recent systematic review suggests that AN have above average intelligence [39], the HC group in this study exhibited an IQ and education bias. However, the fMRI results did not change after taking these variables into account, therefore we conclude that their influence is relatively minor in the context of vWM. Additionally, AN participants exhibited significantly higher comorbid anxious and depressive symptoms whereas we excluded HC participants reporting such symptoms. The use of ‘‘wellcontrols’’ is well documented and introduces a confounding bias towards a significant effect whether it be associated with the disorder of interest or a combination of multiple [56]. However, as many as 97% of AN cases have at least 1 comorbid diagnosis [4], thus begging the question as to whether comorbidity should be treated as part of AN. In our analysis, we found no association between performance and anxiety and depression scores. Nevertheless, future studies would benefit from comparisons with other affected cohorts, to elucidate how comorbid diagnoses interact with AN. Disclosure of interest The authors declare that they have no conflicts of interest concerning this article.

217

Acknowledgements This work was supported by the Swiss Anorexia Foundation, the Psychiatry Research Trust, the NIHR Biomedical Research Centre for Mental Health at the South London and Maudsley NHS Foundation Trust and the Institute of Psychiatry, King’s College London. We would also like to thank Dr. Helen Davies and Naima Lounes for assisting in data collection and Dr. Owen O’Daly for feedback on statistical analysis helpful suggestions in preparation of this manuscript. Appendix. Supplementary data Supplementary data (Tables 1–3) associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.eurpsy.2013.05.003. References [1] Ashburner J, Friston KJ. Voxel-based morphometry – the methods. NeuroImage 2000;11:805–21. [2] Baddeley AD. The episodic buffer: a new component of working memory? Trends Cogn Sci 2000;4:417–23. [3] Baddeley AD, Hitch G. Working memory. In: Bower GH, editor. The psychology of learning and motivation. New York: Academic Press; 1974. p. 47–87. [4] Blinder BJ, Cumella EJ, Sanathara VA. Psychiatric comorbidities of female inpatients with eating disorders. Psychosom Med 2006;68:454–62. [5] Bosanac P, Kurlender S, Stojanovska L, Hallam K, Norman T, McGrath C, et al. Neuropsychological study of underweight and ‘‘weight-recovered’’ anorexia nervosa compared with bulimia nervosa and normal controls. Int J Eat Disord 2007;40:613–21. [6] Brammer MJ, Bullmore ET, Simmons A, Williams SC, Grasby PM, Howard RJ, et al. Generic brain activation mapping in functional magnetic resonance imaging: a nonparametric approach. Magn Reson Imaging 1997;15:763–70. [7] Braver TS, Cohen JD, Nystrom LE, Jonides J, Smith EE, Noll DC. A parametric study of prefrontal cortex involvement in human working memory. NeuroImage 1997;5:49–62. [8] Brooks SJ, Barker GJ, O’Daly OG, Brammer M, Williams SC, Benedict C, et al. Restraint of appetite and reduced regional brain volumes in anorexia nervosa: a voxel-based morphometric study. BMC Psychiatry 2011;11:179. [9] Brooks SJ, O’Daly OG, Uher R, Schioth HB, Treasure J, Campbell IC. Subliminal food images compromise superior working memory performance in women with restricting anorexia nervosa. Conscious Cogn 2012;21:751–63. [10] Brooks SJ, O’Daly O, Uher R, Friederich HC, Giampietro V, Brammer M, et al. Thinking about eating food activates visual cortex with reduced bilateral cerebellar activation in females with anorexia nervosa: an fMRI study. Plos One 2012;7:e34000. [11] Bullmore ET, Brammer MJ, Rabe-Hesketh S, Curtis VA, Morris RG, Williams SCR, et al. Methods for diagnosis and treatment of stimulus-correlated motion in generic brain activation studies using fMRI. Human Brain Mapp 1999;7:38–48. [12] Bullmore ET, Suckling J, Overmeyer S, Rabe-Hesketh S, Taylor E, Brammer MJ. Global, voxel, and cluster tests, by theory and permutation, for a difference between two groups of structural MR images of the brain. IEEE Trans Med Imaging 1999;18:32–42. [13] Bullmore E, Long C, Suckling J, Fadili J, Calvert G, Zelaya F, et al. Colored noise and computational inference in neurophysiological (fMRI) time series analysis: Resampling methods in time and wavelet domains. Human Brain Mapp 2001;12:61–78. [14] Castro-Fornieles J, Bargallo N, Lazaro L, Andres S, Falcon C, Plana MT, et al. A cross-sectional and follow-up voxel-based morphometric MRI study in adolescent anorexia nervosa. J Psychiatr Res 2009;43:331–40. [15] Castro-Fornieles J, Caldu X, Andres-Perpina S, Lazaro L, Bargallo N, Falcon C, et al. A cross-sectional and follow-up functional MRI study with a working memory task in adolescent anorexia nervosa. Neuropsychologia 2010;48:4111–6. [16] Cowan N. What are the differences between long-term, short-term, and working memory? Prog Brain Res 2008;169:323–38. [17] Dickson H, Brooks S, Uher R, Tchanturia K, Treasure J, Campbell IC. The inability to ignore: distractibility in women with restricting anorexia nervosa. Psychol Med 2008;38:1741–8. [18] Druzgal TJ, D’Esposito M. A neural network reflecting decisions about human faces. Neuron 2001;32:947–55. [19] Ernst M, Mueller SC. The adolescent brain: insights from functional neuroimaging research. Dev Neurobiol 2008;68:729–43. [20] Fairburn CG, Beglin SJ. Assessment of eating disorders – interview or selfreport questionnaire. Int J Eat Disord 1994;16:363–70. [21] First M, Spitzer R, Gibbon M, Williams J. Structured clinical interview for DSM-IV axis I disorders. Washington, DC: American Psychiatric Press; 1997: e63964.

218

N.P. Lao-Kaim et al. / European Psychiatry 29 (2014) 211–218

[22] Fonville L, Lao-Kaim NP, Giampietro V, Van den Eynde F, Davies H, Lounes N, et al. Evaluation of enhanced attention to local detail in anorexia nervosa using the embedded figures test. Plos One 2013;8:e63964. [23] Friman O, Borga M, Lundberg P, Knutsson H. Adaptive analysis of fMRI data. NeuroImage 2003;19:837–45. [24] Galderisi S, Mucci A, Monteleone P, Sorrentino D, Piegari G, Maj M. Neurocognitive functioning in subjects with eating disorders: the influence of neuroactive steroids. Biol Psychiatry 2003;53:921–7. [25] Gaudio S, Nocchi F, Franchin T, Genovese E, Cannata V, Longo D, et al. Gray matter decrease distribution in the early stages of anorexia nervosa restrictive type in adolescents. Psychiatry Res Neuroimaging 2011;191:24–30. [26] Genovese CR, Lazar NA, Nichols T. Thresholding of statistical maps in functional neuroimaging using the false discovery rate. NeuroImage 2002;15: 870–8. [27] Green MW, Elliman NA, Wakeling A, Rogers PJ. Cognitive functioning, weight change and therapy in anorexia nervosa. J Psychiatr Res 1996;30:401–10. [28] Hare TA, Camerer CF, Rangel A. Self-control in decision-making involves modulation of the vmPFC valuation system. Science 2009;324:646–8. [29] Hatch A, Madden S, Kohn MR, Clarke S, Touyz S, Gordon E, et al. In first presentation adolescent anorexia nervosa. Do cognitive markers of underweight status change with weight gain following a refeeding intervention? Int J Eat Disord 2010;43:295–306. [30] Ho DE, Imai K, King G, Stuart EA. Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Polit Anal 2007;15:199–236. [31] Ho DE, Imai K, King G, Stuart EA. MatchIt: Nonparametric preprocessing for parametric causal inference, version 2.4-20. J Stat Softw 2007 [http://gking.harvard.edu/matchit/ (Accessed 10 January, 2013)]. [32] Hollmann M, Hellrung L, Pleger B, Schlogl H, Kabisch S, Stumvoll M, et al. Neural correlates of the volitional regulation of the desire for food. Int J Obesity 2012;36:648–55. [33] IBM Corp. SPSS Statistics for Windows, Version 20.0. Armonk, NY: IBM Corp; 2011. [34] Kemps E, Tiggemann M, Wade T, Ben-Tovim D, Breyer R. Selective working memory deficits in anorexia nervosa. Eur Eat Disord Rev 2006;14:97–103. [35] Key A, O’Brien A, Gordon I, Christie D, Lask B. Assessment of neurobiology in adults with anorexia nervosa. Eur Eat Disord Rev 2006;14:308–14. [36] Kingston K, Szmukler G, Andrewes D, Tress B, Desmond P. Neuropsychological and structural brain changes in anorexia nervosa before and after refeeding. Psychol Med 1996;26:15–28. [37] Koenigs M, Barbey AK, Postle BR, Grafman J. Superior parietal cortex is critical for the manipulation of information in working memory. J Neurosci 2009; 29:14980–6. [38] Lopez C, Tchanturia K, Stahl D, Treasure J. Central coherence in eating disorders: a systematic review. Psychol Med 2008;38:1393–404. [39] Lopez C, Stahl D, Tchanturia K. Estimated intelligence quotient in anorexia nervosa: a systematic review and meta-analysis of the literature. Ann Gen Psychiatr 2010;9:40. [40] Martin MM, Rubin RB. A new measure of cognitive flexibility. Psychol Rep 1995;76:623–6. [41] Mathias JL, Kent PS. Neuropsychological consequences of extreme weight loss and dietary restriction in patients with anorexia nervosa. J Clin Exp Neuropsychol 1998;20:548–64. [42] Miller EK, Desimone R. Parallel neuronal mechanisms for short-term memory. Science 1994;263:520–2. [43] Miller GA, Galanter E, Pribram K. Plans and the structure of behaviour. New York: Holt; 1960. [44] Mond JM, Hay PJ, Rodgers B, Owen C, Beumont PJ. Validity of the Eating Disorder Examination Questionnaire (EDE-Q) in screening for eating disorders in community samples. Behav Res Ther 2004;42:551–67.

[45] Mu¨ller NG, Knight RT. The functional neuroanatomy of working memory: contributions of human brain lesion studies. Neuroscience 2006;139:51–8. [46] Mundt JC, Marks IM, Shear MK, Greist JH. The Work and Social Adjustment Scale: a simple measure of impairment in functioning. Br J Psychiatry 2002;180:461–4. [47] Nelson HE, Willison JR. The revised National Adult Reading Test (NART): Test Manual, 2nd edition, Windsor, UK: NFER Nelson; 1991. [48] Nikendei C, Funiok C, Pfuller U, Zastrow A, Aschenbrenner S, Weisbrod M, et al. Memory performance in acute and weight-restored anorexia nervosa patients. Psychol Med 2011;41:829–38. [49] Owen AM, McMillan KM, Laird AR, Bullmore E. N-back working memory paradigm: a meta-analysis of normative functional neuroimaging studies. Human Brain Mapp 2005;25:46–59. [50] Pietrini F, Castellini G, Ricca V, Polito C, Pupi A, Faravelli C. Functional neuroimaging in anorexia nervosa: a clinical approach. Eur Psychiatry 2011;26:176–82. [51] Plichta MM, Schwarz AJ, Grimm O, Morgen K, Mier D, Haddad L, et al. Testretest reliability of evoked BOLD signals from a cognitive-emotive fMRI test battery. NeuroImage 2012;60:1746–58. [52] Ranganath C, D’Esposito M. Directing the mind’s eye: prefrontal, inferior and medial temporal mechanisms for visual working memory. Curr Opin Neurobiol 2005;15:175–82. [53] Ravizza SM, McCormick CA, Schlerf JE, Justus T, Ivry RB, Fiez JA. Cerebellar damage produces selective deficits in verbal working memory. Brain 2006;129:306–20. [54] Salmon E, Van der Linden M, Collette F, Delfiore G, Maquet P, Degueldre C, et al. Regional brain activity during working memory tasks. Brain 1996;119: 1617–25. [55] Sattler JM, Ryan JJ. Assessment with the WAIS-IV. San Diego, California: J.M. Sattler Publishing Company; 2009. [56] Schwartz S, Susser E. The use of well controls: an unhealthy practice in psychiatric research. Psychol Med 2011;41:1127–31. [57] Seed JA, McCue PM, Wesnes KA, Dahabra S, Young AH. Basal activity of the HPA axis and cognitive function in anorexia nervosa. Int J Neuropsychopharmacol 2002;5:17–25. [58] Shott ME, Filoteo JV, Jappe LM, Pryor T, Maddox WT, Rollin MDH, et al. Altered implicit category learning in anorexia nervosa. Neuropsychology 2012;26: 191–201. [59] Simmons A, Moore E, Williams SCR. Quality control for functional magnetic resonance imaging using automated data analysis and Shewhart charting. Magn Reson Med 1999;41:1274–8. [60] Smith EE, Jonides J, Marshuetz C, Koeppe RA. Components of verbal working memory: evidence from neuroimaging. Proc Natl Acad Sci USA 1998;95: 876–82. [61] Szmukler GI, Andrewes D, Kingston K, Chen L, Stargatt R, Stanley R. Neuropsychological Impairment in anorexia nervosa: before and after refeeding. J Clin Exp Neuropsychol 1992;14:347–52. [62] Talairach J, Tournoux P. Co-planar stereotactic atlas of the human brain. Stuttgart: Thieme; 1988. [63] Tchanturia K, Davies H, Roberts M, Harrison A, Nakazato M, Schmidt U, et al. Poor cognitive flexibility in eating disorders: examining the evidence using the wisconsin card sorting task. Plos One 2012;7:e28331. [64] Tchanturia K, Lloyd S, Lang K. Cognitive remediation therapy for anorexia nervosa: current evidence and future research directions. Int J Eat Disord 2013;46:492–6. [65] Treasure J, Claudino AM, Zucker N. Eating disorders. Lancet 2010;375:583–93. [66] Witt ED, Ryan C, Hsu LK. Learning deficits in adolescents with anorexia nervosa. J Nerv Ment Dis 1985;173:182–4. [67] Zigmond AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatr Scand 1983;67:361–70.