Caseras, Xavier PhD; Mataix-Cols, David PhD; Giampietro, Vincent MS; Rimes, Katharine A. PhD; Brammer, Michael MD; Zelaya, Fernando PhD; Chalder, Trudie PhD; Godfrey, Emma L. PhD
Chronic fatigue syndrome (CFS) is a complex and medically unexplained illness, which is characterized by severe and prolonged disabling fatigue and a combination of symptoms, including sleep disturbances, musculoskeletal pain, and impairments in concentration and short-term memory (1). These cognitive symptoms are common among patients with CFS and considered some of the most disabling and troubling symptoms (2,3). Despite being reported by 90% of patients with CFS, neuropsychological studies on memory and attention have produced inconsistent results regarding cognitive impairment (for a review, see (4)).
Working memory (WM) can be defined as the capacity to store information in short-term registers and to simultaneously manipulate it online. It is a crucial cognitive function for human thought processes such as reasoning, learning, and comprehension (5), facilitating the momentarily maintenance and manipulation of task-relevant information. This important cognitive function is severely impaired in debilitating psychiatric disorders such as schizophrenia (6) and depression (7). A wide number of tasks have been used to evaluate WM in patients with CFS, but inconsistent results were obtained. For instance, DeLuca et al. (8), using the Paced Auditory Serial Addition Test (PASAT) task, found no differences between patients with CFS and control subjects on WM accuracy, but did find differences on speed of information processing. In another study, Dobbs et al. (9) reported clear between-group differences in performance as task difficulty increased using the Digit Span and the Dobbs and Rule WM tasks (10).
One possible explanation for these inconsistent results could be a lack of sensitivity of behavioral tasks to detect subtle neuropsychological changes in CFS. Functional neuroimaging techniques are capable of detecting subtle differences in brain function between groups of individuals even in the absence of behavioral performance differences. Indeed, preliminary neuroimaging research in CFS suggests that subtle functional neural changes may be implicated in the mediation of the cognitive (i.e., WM) difficulties experienced by these patients. Specifically, two recent studies have shown increased widespread brain activation during performance on verbal WM tasks despite intact behavioral performance. Schmaling et al. (11) used single photon emission computed tomography (SPECT) during performance on the PASAT and found that although patients showed less perfusion in the anterior cingulate region, the signal change in this area during the test compared with rest was greater for patients than for control subjects. In a recent functional magnetic resonance imaging (fMRI) study using a similar task, Lange et al. (12) found that patients displayed greater activations in parts of the WM network (bilateral supplementary and premotor and left superior parietal regions). Because of the lack of behavioral differences in performance between the groups, the authors concluded that patients needed to exert greater effort (i.e., to allocate more neural resources) to process auditory information as effectively as controls.
Building on the initial work by Schmaling et al. (11) and Lange et al. (12), the present research aims to further investigate the neural correlates of verbal WM function in CFS. We used a widely used verbal WM task called n-back (13). Unlike other WM tasks such as the PASAT, the n-back allows us to parametrically manipulate the level of difficulty and may therefore be more sensitive to subtle differences in WM function between patients with CFS and control subjects. Note that this task does not manipulate the participant's speed of information processing (8) but the ability to handle increasing WM loads. Based on previous research using verbal versions of the n-back in healthy populations (14), we expected both patients with CFS and control subjects to engage bilateral dorsolateral and ventrolateral prefrontal cortices, anterior cingulate gyrus, premotor and parietal cortices while performing the task. We also predicted that the magnitude of activation in these brain regions would lineally increase with task difficulty. We tested the null hypothesis that there would be no differences between patients with CFS and control subjects either on behavioral performance or on brain activation during the n-back task.
Eighteen right-handed patients with CFS (11 females; age range, 22–45 years) and 12 right-handed healthy control subjects (eight females; age range, 24–45 years) agreed to participate in the study. The patients with CFS were recruited between February and October 2005 from the CFS Research and Treatment Units of King's College London (n = 17) and St. Bartholomew's Hospital (n = 1) and fulfilled strict Centers for Disease Control and Prevention (CDC) criteria for CFS (1). The patients were referred by their general practitioner to the unit with a suspected diagnosis of CFS. At the unit, a physician conducted a complete clinical interview to confirm the diagnosis of CFS. Finally, a clinical psychologist independently confirmed the diagnosis using a checklist of the CDC (1) criteria. All patients reported subjective short-term memory and concentration problems on the CDC checklist.
Exclusion criteria included presence of major depression, history of brain injury, psychosis, bipolar disorder, claustrophobia, suspected pregnancy, and presence of metallic implants. Five patients with CFS were on antidepressants at the time of their inclusion in the study. These patients did not differ from the rest of the CFS group on any sociodemographic or clinical measures and had comparable performances and patterns of brain activation on the n-back and were therefore included in the study. Five patients met diagnostic criteria for panic disorder at the time of their inclusion in this study. Because these patients had comparable sociodemographic and clinical characteristics (only a trend to show higher anxiety was detected) and their exclusion from the fMRI analyses did not substantially modify the results a part from an expected decrease in statistical power, they were kept in the analysis. One patient obtained an accuracy score on average two standard deviations below the mean of the CFS group and was therefore considered an outlier and excluded from the analyses.
Control participants were recruited by advertisement and included ancillary staff at the Institute of Psychiatry and members of the wider community. Exclusion criteria were present or past history of CFS or any mental disorder, history of brain injury, claustrophobia, suspected pregnancy, presence of any metallic implants, and regular physical exercise. We reasoned that it was important to select control subjects with a low-to-medium level of physical fitness to make them as comparable as possible to the CFS group; thus, this variable was taken into account when recruiting control participants. Regarding performance on the n-back task, there were no outliers in this group. The final sample was of 29 participants (17 patients with CFS and 12 control subjects).
The study was approved by the ethics committees of the South London and Maudsley NHS Trust and of St. Bartholomew's Hospital. All individuals gave written informed consent to participate.
After confirming the presence of CFS, an experienced clinician administered the Mini International Neuropsychiatric Interview (MINI (15)) to rule out the presence of major psychiatric comorbidity.
All participants completed the Physical Functioning Scale from the SF-36 (16), the Chalder Fatigue scale (17), the Work and Social Adjustment Scale (18,19), and the Hospital Anxiety and Depression Scale (20).
A verbal version of the n-back originally described by Braver et al. (13) was used. In a block design, three levels of difficulty (1-back, 2-back, and 3-back) and one control condition (0-back) were included. Participants were presented with a series of capital letters on a projection screen (which they saw through a mirror above their head) and were required to press a button whenever the letter presented was the same as that presented n trials previously (1-, 2-, or 3-back). During the control condition, participants were required to press the button whenever they saw the letter “X.” All blocks consisted of a pseudorandom sequence of 21 letters presented for 1 second each and separated by an interstimulus interval of 1 second. Three blocks of each condition were presented, totaling 12 blocks in the following order: 1-back/2-back/0-back/3-back/2-back/0-back/1-back/3-back/2-back/3-back/0-back/1-back. The task had a total duration of 9 minutes.
A computer automatically recorded the participants' performance (accuracy and reaction time to target) during the task. All participants received a training session before scanning to ensure that they understood the task.
Single-shot, gradient echo, echoplanar imaging was used to acquire 180 T2*-weighted image volumes on a 1.5-Tesla GE Excite MRI scanner (General Electric, Milwaukee, WI) at the Centre for Neuroimaging Sciences of the Institute of Psychiatry, King's College London. For each volume, 37 noncontiguous axial planes parallel to the intercommissural plane were collected with the following parameters: repetition time (TR) = 3000 ms, echo time (TE) = 40 ms, slice thickness = 3 mm, slice skip = 0.3 mm, field of view = 24 cm, image acquisition matrix 64 × 64, yielding an in-plane spatial resolution of 3.75 × 3.75 mm. Four dummy acquisitions were also made at the beginning of each run to set the longitudinal magnetization into steady state. The start of the n-back paradigm was triggered by the scanner immediately at the end of the dummy acquisition.
To facilitate coregistration of the fMRI data in standard space, we also acquired during each session, a 43-slice, high-resolution inversion recovery echoplanar image of the whole brain in the same intercommissural plane with the following parameters: TR = 3000 ms, TE = 40 ms, slice thickness = 3 mm, slice skip 0.3 mm, NEX = 8, image acquisition matrix 128 × 128, yielding an in-plane spatial resolution of 1.875 × 1.875 mm.
Age, years of education, and clinical measures were compared between groups by means of Student t test. The association between gender and group was calculated by means of a χ2 statistic in a 2 × 2 contingency table.
Two behavioral measures were calculated from the participants' performance in the n-back task. Accuracy responding was calculated as the percentage of correctly detected repetitions during each WM condition (i.e., 1-, 2-, and 3-back) and as the number of correctly identified “X” during the control condition (i.e., 0-back). Because this measure was not normally distributed, comparisons between groups and across load conditions were done by means of nonparametric tests (Mann-Whitney U tests between patients with CFS and control subjects, and Wilcoxon test between load conditions). Reaction time was computed as the latency between the target onset and the participant's response. This measure was normally distributed and was subjected to an analysis of variance with one between-groups factor, “group” (CFS vs. control) and one within-groups factor, “load” (0-, 1-, 2-, 3-back). Comparisons across load levels were conducted by means of related samples Student t test.
Functional Magnetic Resonance Imaging Data Analysis
The fMRI data were analyzed with software developed at the Institute of Psychiatry (XBAM) using a nonparametric approach to minimize assumptions. Data were first corrected for subject motion (21) and then smoothed using a Gaussian filter (FWHM 7.2 mm) chosen to improve signal-to-noise ratio over the spatial neighborhood of each voxel. Responses to the experimental paradigms were then detected by time-series analysis using a linear model in which each component of the experimental design was convolved separately with two gamma variate functions (peak responses at 4 and 8 seconds, respectively) to permit variability in the hemodynamic delay. The method of Friman et al. (22) was used to constrain model fits to those deemed physiologically plausible. After computation of the model fit, a goodness-of-fit statistic was computed. This consisted of 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 (SSQratio). This addresses the problem inherent in the use of the F statistic that the residual degrees of freedom are often unknown in fMRI time series as a result of the presence of colored noise in the signal. After computation of the observed SSQratio at each voxel, the data were permuted by the wavelet-based method described and extensively characterized in Bullmore et al. (23), which permits the data-driven calculation of the null distribution of SSQratios under the assumption of no experimentally determined response. This distribution can then be used to threshold the activation maps at any desired type I error rate. In addition to the SSQratio, the percentage blood oxygen level dependent (BOLD) change was also calculated from the model fit at each voxel.
The detection of activated regions was extended from voxel to cluster level using the method described in detail by Bullmore et al. (23). The observed and randomized SSQratio data for each individual was transformed into standard space of Talairach and Tournoux (24) and group maps of activated regions were computed using the median observed and randomized SSQratio data as described by Brammer et al. (25). Permutation methods and median statistics were used to allow exact computation of p values with minimal assumptions and the minimization of outlier effects. The hierarchical method of analysis used previously also allows separate treatment of intra- and interindividual variance. After extension of inference from voxel to cluster level (26), the resulting cluster maps were thresholded to give <1 expected type I error cluster per whole brain volume to make interpretation of maps as intuitive as possible.
For each level of the task, comparisons of responses between groups was performed by fitting the data at each intracerebral voxel at which all subjects have nonzero data using a linear model of the type:
Equation (Uncited)Image Tools
where Y is the vector of BOLD effect sizes for each individual, X is the contrast matrix for the particular intergroup contrasts required, a is the mean effect across all individuals in the various groups, b is the computed group difference, and e is a vector of residual errors. The model is fitted by minimizing the sum of absolute deviations rather than the sums of squares to reduce outlier effects. The null distribution of b is computed by permuting data between groups (assuming the null hypothesis of no effect of group membership) and refitting this model 50 times at each voxel and combining the data over all intracerebral voxels. Group difference maps are at any desired voxel or cluster-wise type I error rate can then be computed by appropriate thresholding of this null distribution.
Trend Analyses (task load effects)
For each the CFS and control groups, we first fitted linear and quadratic models to the estimated BOLD responses at each level of task for all subjects in each group. A further analysis was then conducted as described previously (under “Between-Group Comparisons”) to assess whether any brain regions showed group-dependent differences in these relationships.
In no case was a significant quadratic component detected, that is, even when the slopes appeared nonlinear, the departure from linearity did not reach significance. Thus, all results shown relate to estimated linear trends.
Demographic and Clinical Measures
As shown in Table 1, patients and control subjects did not differ in terms of age, sex distribution, or years of education. Regarding clinical measures, patients reported significantly higher levels of fatigue, work and social disability, depression, and anxiety and lower scores on the physical functioning scale. The patients' scores on these scales were in the severe range and their reported high levels of functional impairment (Table 1).
The mean and standard deviations for accuracy and reaction time across load levels for each group are presented in Table 1. The groups did not differ in accuracy at any level of the n-back. As the task difficulty increased, there was a linear increase in the number of errors; all pairwise comparisons (Wilcoxon tests) across load levels were significant (smaller z = 2.81, p < .01), with the exception of the comparison between the 0-back and 1-back conditions (z = 1.34, p > .1).
Regarding reaction times, neither the group nor the group × load interaction effects were significant (p > .1). There was, however, a significant load effect (F [3, 81] = 12.07, p < .001); post hoc pairwise comparisons revealed slower reaction times during the higher load (2- and 3-back) compared with the lower load (0- and 1-back) conditions (Table 1).
Functional Magnetic Resonance Imaging Data
Generic Brain Activation Maps
As predicted, patients with CFS and control subjects showed a qualitatively similar pattern of neural activation during performance on the 1-, 2-, and 3-back conditions contrasted with the 0-back condition. Indeed, across all levels of difficulty, both groups showed strong activations in bilateral brain regions previously implicated in WM tasks, including the dorsolateral (Brodmann's areas [BA] 6/9/10/44/46/47) and dorsomedial (BA 9/10) prefrontal cortex, the precentral gyrus (BA 4/6), the dorsal anterior cingulate gyrus (BA 24/32), the occipitoparietal cortex (BA 7/19/40), and the occipitotemporal cortex (BA 19/20/21/37).
However, there were two regions that were strongly activated in the CFS group but were absent in the control group. In the 1-back condition, patients significantly activated bilateral medial (BA 10; right: x = 7, y = 52, z = 13, 58 voxels; left: x = −7, y = 48, z = 10, 28 voxels) and lateral (BA 10; right: x = 29, y = 56, z = 7, 20 voxels; left: x = −29, y = 56, z = 3, 19 voxels) prefrontal regions, including the anterior cingulate gyrus (BA 24/32; x = −14, y = 30, z = −3, 24 voxels). In the 2-back and 3-back conditions, patients but not control subjects significantly activated a cluster in the right inferior/middle temporal gyrus (BA 21/37; x = 51, y = −48, z = −7, 48 voxels).
Comparisons Between Groups
On the 1-back condition, the CFS group showed greater activation than control subjects in a large cluster in the medial prefrontal cortex covering the medial and lateral parts of BA 10 and extending toward the anterior cingulate gyrus (BA 24/32) (peak activation x = 29, y = 56, z = 7; cluster size 144 voxels; clusterwise p = .0005) (Fig. 1A).
On the 2-back, the CFS group showed reduced activations in left lateral BA 10 extending to dorsolateral prefrontal cortex (BA 45/46; peak activation: x = −33, y = 56, z = 3; 112 voxels; clusterwise p = 0.0018) and in the left parietal lobule (BA 7/19; peak activation: x = −18, y = −78, z = 33; 112 voxels; clusterwise p = .0019) (Fig. 1B).
Finally, during the 3-back condition, patients with CFS showed reduced activation in superior parietal regions bilaterally (BA 7; peak activation: x = 18, y = −70, z = 43; 151 voxels; clusterwise p = .0009) and increased activation in the right inferior temporal gyrus (BA 20; peak activation: x = 54, y = −37, z = −17; 79 voxels; clusterwise p = .0028) (Fig. 1C).
Trend Analysis Within Groups
With increasing task difficulty, the healthy control subjects showed a significant linear increase in activation in bilateral parietal (BA 7/19/40) cortex, right dorsolateral prefrontal cortex (BA 8/9/45/46), right superior frontal gyrus (BA 6), right middle frontal gyrus (lateral BA 10), bilateral dorsal anterior cingulate gyrus (BA 32), and bilateral cerebellum (Table 2).
With increasing task difficulty, the patients with CFS showed significant increases in activation in the right middle frontal gyrus (lateral BA 10), bilateral parietal cortex (BA 7/19/40), right dorsolateral (BA 45/46) and right ventrolateral (BA 47) prefrontal cortices. They also showed a significant negative linear trend across task load in a large cluster in the medial prefrontal cortex including the cingulate gyrus (BA 10/32) (Table 2).
Trend Comparisons Between Groups
We next tested for any significant between-group differences on the slopes resulting form the trend analyses. There were statistically significant differences between the groups in three clusters (Fig. 2).
The first cluster was situated in the bilateral parietal cortex (BA 7/19; peak activation in x = 0, y = −63, z = 50; 67 voxels; clusterwise p = .0010) (Fig. 2A). Post hoc comparisons at each level of the n-back revealed no statistically significant differences between the groups (all p > .1), suggesting that a greater positive slope in control subjects than in patients drove the significant interaction effect.
The second cluster was situated in the ventromedial/ ventrolateral prefrontal cortex (BA 11/47) extending toward the anterior cingulate cortex (BA 32). Its most activated voxel was in the ventrolateral prefrontal cortex (x = 40, y = 41, z = −7; 39 voxels; clusterwise p = .0033) (Fig. 2B). Post hoc comparisons revealed significant differences between patients and control subjects in the 2-back and 3-back conditions (U = 57 and U = 56, both p < .05). Visual inspection of Figure 2B shows a less pronounced negative trend in activation across task loads in patients with CFS compared with control subjects.
The third significant cluster was in the right middle frontal gyrus (lateral BA 10; peak activation at x = 33, y = 56, z = 17; 34 voxels; clusterwise p = .0031) (Fig. 2C). Visual inspection of Figure 2C suggests an interaction between group and load in this region, whereby patients show higher activations during the 1-back condition and lower activations during the 2-back and 3-back conditions than control subjects. However, this only reached statistical significance for the 1-back condition (U = 58, p = .05).
We next extracted the effect size values (percent change in BOLD signal) of the brain regions that separated patients and control subjects in the previous analyses and correlated them with the relevant psychometric measures within the patient group. There were no significant correlations between effect size values and any of the scales (SF-36, CFS, WSAS, HADS) (all p > .05).
To our knowledge, this is the first fMRI study of WM in CFS using a parametric verbal WM task, which allowed us to examine the effects of load on performance and brain activation. During the low load (1-back) condition, patients with CFS showed greater activation than control subjects in medial prefrontal regions, including the anterior cingulate gyrus. However, on the high load conditions (2-back and 3-back), patients with CFS demonstrated reduced activation in working memory-related brain regions (dorsolateral prefrontal and parietal cortices) compared with healthy control subjects. Furthermore, on the 2-back and 3-back conditions, patients but not control subjects significantly activated a large cluster in the right inferior/medial temporal cortex. Finally, trend analyses of task load demonstrated statistically significant differences in brain activation between the two groups as the difficulty of the task increased. These results were uncorrelated with any demographic or clinical variables. Each of these findings are now discussed in more detail.
The first important finding of this study was that patients with CFS showed greater activation than control subjects in medial prefrontal regions (BA 10), including the rostral anterior cingulate gyrus (BA 24/32), during the lower load condition (1-back). This finding partially replicates earlier work. In a SPECT study, Schmaling et al. (11) found greater change in anterior cingulate perfusion while performing the PASAT compared with rest. Lange et al. (12) found that patients displayed greater activations in parts of the WM network (bilateral supplementary and premotor and left superior parietal regions) during the PASAT. These results led the authors to suggest that patients with CFS require additional neural resources (i.e., more activation) to achieve the same level of behavioral performance on the PASAT than control subjects. This interpretation is only partially supported by the current data because, in this study, as the task demand increased, patients with CFS showed reduced rather than increased recruitment of working memory-related regions.
The second key finding of this study was that the CFS group showed reduced activation in dorsolateral prefrontal (BA 10/45/46) and parietal (BA 7/19) cortices during the more demanding levels of the task. These regions are key nodes of the WM system and our findings suggest that at high task demands, patients with CFS fail to recruit these regions to the same extent as healthy control subjects. These findings appeared to be independent from behavioral performance, which did not separate the two groups on any level of WM load. However, it is possible that the 2-back and 3-back conditions are not demanding enough to produce between-group differences at the behavioral level but enough to show differences at the neurophysiological level. Future research including further load levels (i.e., 4-back) could help clarify the relationship between brain activation and behavioral performance on WM tasks in patients with CFS.
The third important finding of this study was that during the 2-back and 3-back conditions, patients activated a large cluster in the right inferior/medial temporal cortex (BA 21/37), which was not activated by the control subjects. After applying our strict statistical threshold, the patients with CFS showed significantly greater activation in this region only in the 3-back condition. Similar findings have been reported in some neurological conditions and autism (e.g., (27,28)) and interpreted as compensatory strategies when the WM network is dysfunctional or saturated. It is therefore plausible that patients with CFS recruit this region to compensate for their WM problems and that this somehow allows them to maintain an adequate behavioral performance even on the most demanding conditions of the task. Future studies using dual-task paradigms designed to obstruct the use of compensatory strategies might reveal statistically significant differences on the n-back at the behavioral level.
Task Load Effects
Trend analysis allowed us to examine which brain regions increased or decreased in activation as the WM load increased. The results showed that as the load increased, both groups showed a typical pattern of increased activation in parietal and dorsolateral prefrontal regions and decreased activation in medial prefrontal regions. The latter decrease in activation was strongly significant in the CFS group but did not reach statistical significance in the control group. However, when our strict statistical threshold was relaxed, this region also showed a clear decrease in the control group (data not shown). When we tested for any significant differences on the slopes resulting from these analyses, we found that 3 clusters separated the CFS and control groups: the bilateral parietal cortex (BA 7/19), the right middle frontal gyrus (lateral BA 10), and the ventromedial prefrontal cortex, including the anterior cingulate gyrus (BA 11/32/47).
Regarding the parietal and lateral prefrontal regions, both groups showed a linear increase in activation as the memory demand of the task increased as reported by many previous studies (e.g., (29–31)). However, patients with CFS showed smaller increases in activation in these regions (see Fig. 2A, 2C), confirming the results of the between-group analyses.
Conversely, both groups showed a negative linear trend in activation across task load in the ventromedial prefrontal cortex. This effect has also been reported previously (e.g., (30,32,33)), and although a fully satisfactory interpretation has not been provided, it has been hypothesized to reflect an “emotional gating” mechanism aimed at inhibiting adverse emotional signals to maximize the level of performance (32). In the present study, patients with CFS showed a less prominent decrease in activation of this region as the task demand increased compared with the control subjects (see Fig. 2B). If the “emotional gating” hypothesis is correct, these results could suggest greater emotional interference in the CFS group, although this was not associated with reduced performance on the task. Interestingly, similar findings were recently reported in depressed patients (33,35). We carefully excluded any patients with CFS who met criteria for major depression, but the CFS group still had higher anxiety and depression scores than the control subjects; therefore, it cannot be ruled out that this particular finding is related to low mood in our CFS group. However, our correlation analyses failed to find an association between percentage of change in BOLD signal in this region and depression/anxiety scores (data not shown). Furthermore, Harvey et al. (34) found that depressed patients show higher rather than lower activation of dorsolateral and parietal regions compared with control subjects, suggesting important differences between CFS and depressed patients during the n-back task. Future studies would benefit from using psychiatric (i.e., depressed or anxious) control groups to fully rule out the potential effects of psychiatric comorbidity. Taken together, our findings of a) reduced increases in activation in parietal and dorsolateral prefrontal regions and b) reduced decreases in activation in ventromedial prefrontal regions in CFS are consistent with the notion that the WM system is compromised in CFS.
The Potential Role of Sleep Disturbances
It is important to notice the close resemblance between our results and those observed when sleep-deprived healthy adults perform similar tasks (30,35). Because as many as 58% to 79% of patients with CFS report sleep disorders (e.g., (36,37)), this raises the possibility that our findings may be a consequence of disrupted sleep patterns in this population. However, the relationship between sleep disturbances and impaired WM may be modulated by an underlying vulnerability to sleep disturbances and fatigue. Indeed, a recent fMRI study by Mu et al. (38) compared brain activity during a WM task in sleep-deprivation-vulnerable versus sleep-deprivation-resilient individuals both at baseline and after sleep deprivation. They found that sleep-deprivation-vulnerable individuals showed reduced patterns of brain activation during the WM task even at baseline (i.e., when they were not sleep-deprived). In a similar study, Caldwell et al. (39) found that cortical brain activation at baseline was related to fatigue vulnerability in military pilots during simulator flight performance. Therefore, our findings could be interpreted either as a consequence of our patients' sleep problems or as a preexisting vulnerability that would explain the reduced brain activation even in the presence of good restoring sleep. We are currently planning to rescan some of the study participants after treatment to examine these alternative hypotheses. Future studies would benefit from carefully recording the patients' sleep patterns and examining whether these correlate with brain activation.
Some limitations of the present study should be considered. Five patients were on antidepressant medication and five met diagnostic criteria for panic disorder (two patients were both on medication and had panic disorder) at the time of the study. However, these patients did not differ from the rest of the patients with CFS on any sociodemographic or clinical variables. Furthermore, their exclusion from the fMRI analyses did not substantially modify the results. We carefully excluded patients with severe depressive disorders, because this has implications for the differential diagnosis of CFS, but deliberately did not exclude comorbid anxiety disorders to increase the representativeness of the sample. We reasoned that because anxiety disorders are not infrequent in CFS (e.g., (40,41)), it would be artificial to recruit a “pure” noncomorbid sample. The high prevalence of panic disorder in our sample was similar to that reported in other clinical studies (41). Future studies would benefit from recruiting psychiatric control groups to control for the putative effects of psychiatric comorbidity on brain imaging results.
In summary, the current findings suggest that patients with CFS do not engage the WM network in the same way as healthy control subjects do, perhaps reflecting compensatory strategies to counteract their underlying cognitive difficulties and achieve a comparable behavioral performance. These results are consistent with the subjective feelings of memory impairment that many patients report. Whether these abnormal patterns of brain activation can be ameliorated with successful treatment is a crucial question that will need to be addressed in future longitudinal studies.
Professor Peter White and Giselle Whiters helped with the recruitment of one participant at St. Bartholomew's Hospital.
1.Fukuda K, Straus SE, Hickie I, Sharpe MC, Dobbins JG, Komaroff A. The chronic fatigue syndrome: a comprehensive approach to its definition and study. International Chronic Fatigue Syndrome Study Group. Ann Intern Med 1994;121:953–9.
2.Afari N, Buchwald D. Chronic fatigue syndrome: a review. Am J Psychiatry 2003;160:221–36.
3.Tiersky LA, DeLuca J, Hill A, Dhar SK, Johnson SK, Lang G, Rappolt G, Natelson BH. Longitudinal assessment of neuropsychological functioning, psychiatric status, functional disability and employment status in chronic fatigue syndrome. Appl Neuropsychol 2001;8:41–50.
4.Michiels V, Cluydts R. Neuropsychological functioning in chronic fatigue syndrome: a review. Acta Psychiatr Scand 2001;103:84–93.
5.Baddeley A. Working memory: looking back and looking forward. Nat Rev Neurosci 2003;4:829–39.
6.Aleman A, Hijman R, De Haan EH, Kahn RS. Memory impairment in schizophrenia: a meta-analysis. Am J Psychiatry 1999;156:1358–66.
7.Rose EJ, Ebmeier KP. Pattern of impaired working memory during major depression. J Affect Disord 2006;90:149–61.
8.DeLuca J, Christodoulou C, Diamond BJ, Rosenstein ED, Kramer N, Natelson BH. Working memory deficits in chronic fatigue syndrome: differentiating between speed and accuracy of information processing. J Int Neuropsych Soc 2004;10:101–9.
9.Dobbs BM, Dobbs AR, Kiss I. Working memory deficits associated with chronic fatigue syndrome. J Int Neuropsych Soc 2001;7:285–93.
10.Dobbs AR, Rule BG. Adult age differences in working memory. Psychol Aging 1989;4:500–3.
11.Schmaling KB, Lewis DH, Fiedelak JI, Mahurin R, Buchwald DS. Single-photon emission computerized tomography and neurocognitive function in patients with chronic fatigue syndrome. Psychosom Med 2003;65:129–36.
12.Lange G, Steffener J, Cook DB, Bly BM, Chirstodoulou C, Liu WC, DeLuca J, Natelson BH. Objective evidence of cognitive complaints in chronic fatigue syndrome: a BOLD fMRI study of verbal working memory. Neuroimage 2005;26:513–24.
13.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.
14.Owen AM, McMillan KM, Laird AR, Bullmore E. N-back working memory paradigm: a meta-analysis of normative functional neuroimaging studies. Hum Brain Mapp 2005;25:46–59.
15.Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, Hergueta T, Barker R, Dunbar GC. The Mini-International Neuropsychiatric Interview (MINI): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry 1998;59(suppl 20):22–33.
16.Ware JE, Sherbourne C. The MOS 36-item Short Form health survey (SF-36). Med Care 1992;30:473–83.
17.Chalder T, Berelowitz G, Pawlikowska T, Watts L, Wessely S, Wright D, Wallace EP. Development of a fatigue scale. J Psychosom Res 1993;37:147–53.
18.Mundt JC, Marks IM, Greist JH, Shear K. Work and Social Adjustment Scale: a simple accurate measure of impairment in functioning. Br J Psychiatry 2002;180:461–4.
19.Mataix-Cols D, Cowley AJ, Hankins M, Schneider A, Bachofen M, Kenwright M, Gega L, Cameron R, Marks IM. Reliability and validity of the Work and Social Adjustment Scale in phobic disorders. Compr Psychiatry 2005;46:223–8.
20.Zigmond AS, Snaith RP. The Hospital Anxiety and Depression Scale. Acta Psychiatr Scand 1983;67:361–70.
21.Bullmore ET, Brammer M, Rabe-Hesketh S, Curtis V, Morris R, Williams SCR, Sharma T, McGuire PK. Methods for the diagnosis and treatment of stimulus correlated motion in generic brain activation studies using fMRI. Hum Brain Mapp 1999;7:38–48.
22.Friman O, Borga M, Lundberg P, Knutsson H. Adaptive analysis of fMRI data. Neuroimage 2003;19:837–45.
23.Bullmore ET, Long C, Suckling J, Fadili J, Calvert GA, Zelaya F, Carpenter TA, Brammer MJ. Coloured noise and computational inference in neurophysiological (fMRI) time series analysis: resampling methods in time and wavelet domains. Hum Brain Mapp 2001;12:61–78.
24.Talairach J, Tournoux P. Co-planar Stereotactic Atlas of the Human Brain. Stuttgart: Thieme, 1988.
25.Brammer M, Bullmore ET, Simmons A, Williams SCR, Grasby PM, Howard RJ, Woodruff PWR, Rabe-Hesketh S. Pi-by-no brain activation mapping in functional magnetic resonance imaging: a nonparametric approach. Magn Reson Imaging 1997;15:763–70.
26.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.
27.Wishart HA, Saykin AJ, McDonald BC, Mamourian AC, Flashman LA, Schuschu KR, Ryan KA, Fadul CE, Kasper LH. Brain activation patterns associated with working memory in relapsing-remitting MS. Neurology 2004;62:234–8.
28.Koshino H, Carpenter PA, Minshew NJ, Cherkassky VL, Keller TA, Just MA. Functional connectivity in an fMRI working memory task in high-functioning autism. Neuroimage 2005;24:810–21.
29.Jaeggi SM, Seewer R, Nirkko AC, Eckstein D, Schroth G, Groner R, Gutbrod K. Does excessive memory load attenuate activation in the prefrontal cortex? Load-dependent processing in single and dual tasks: functional magnetic resonance imaging study. Neuroimage 2003;19:210–25.
30.Chee MW, Choo WC. Functional imaging of working memory after 24 hr of total sleep deprivation. J Neurosci 2004;24:4560–7.
31.Mattay VS, Fera F, Tessitore A, Hariri AR, Berman KF, Das S, Meyer-Lindenberg A, Goldberg TE, Callicott JH, Weinberger DR. Neurophysiological correlates of age-related changes in working memory capacity. Neurosci Lett 2006;392:32–7.
32.Pochon JB, Levy R, Fossati P, Lehericy S, Poline JB, Pillon B, Le Bihan D, Dubois B. The neural system that bridges reward and cognition in humans: an fMRI study. Proc Natl Acad Sci U S A 2002;99:5669–74.
33.Rose EJ, Simonotto E, Ebmeier KP. Limbic over-reactivity in depression during preserved performance on the n-back task. Neuroimage 2006;29:209–15.
34.Harvey PO, Fossati P, Pochon JB, Levy R, Lebastard G, Lehericy S, Allilaire JF, Dubois B. Cognitive control and brain resources in major depression: an fMRI study using the n-back task. Neuroimage 2005;26:860–9.
35.Mu Q, Nahas Z, Johnson KA, Yamanaka K, Mishory A, Koola J, Hill S, Horner MD, Bohning DE, George MS. Decreased cortical response to verbal working memory following sleep deprivation. Sleep 2005;28:55–67.
36.Farmer A, Jones I, Hillier J, Llewelyn M, Borysiewicz L, Smith A. Neuraesthenia revisited: ICD-10 and DSM-III-R psychiatric syndromes in chronic fatigue patients and comparison subjects. Br J Psychiatry 1995;167:503–6.
37.Morriss R, Sharpe M, Sharpley AL, Cowen PJ, Hawton K, Morris J. Abnormalities of sleep in patients with the chronic fatigue syndrome. BMJ 1993;306:1161–4.
38.Mu Q, Mishory A, Johnson KA, Nahas Z, Kozel FA, Yamanaka K, Bohning DE, George MS. Decreased brain activation during a working memory task at rested baseline is associated with vulnerability to sleep deprivation. Sleep 2005;28:433–46.
39.Caldwell JA, Mu Q, Smith JK, Mishory A, Caldwell JL, Peters G, Brown DL, George MS. Are individual differences in fatigue vulnerability related to baseline differences in cortical activation? Behav Neurosci 2005;119:694–707.
40.Fischler B, Cluydts R, De Gucht Y, Kaufman L, De Meirleir K. Generalised anxiety disorder in chronic fatigue syndrome. Acta Psychiatr Scand 1997;95:405–13.
41.Manu P, Lane TJ, Matthews DA. Chronic fatigue and chronic fatigue syndrome: clinical epidemiology and aetiological classification. Ciba Found Symp 1993;173:23–31.