Executive functions describe a wide variety of higher order cognitive processes that allow the flexible modification of thought and behaviour in response to changing cognitive or environmental contexts. Their impairment is common in neurodegenerative disorders. Executive deficits negatively affect everyday activities and hamper the ability to cope with other deficits, such as memory impairment in Alzheimer's disease or behavioural disorders in frontotemporal lobar degeneration. Our study aimed to characterize the neural correlates of executive functions by relating respective deficits to regional hypometabolism in early dementia. Executive functions were assessed with two classical tests, the Stroop and semantic fluency test and various subtests of the behavioural assessment of the dysexecutive syndrome test battery capturing essential aspects of executive abilities relevant to daily living. Impairments in executive functions were correlated with reductions in brain glucose utilization as measured by [18F]fluorodeoxyglucose positron emission tomography and analysed voxelwise using statistical parametric mapping in 54 subjects with early dementia, mainly Alzheimer's disease and frontotemporal lobar degeneration, and its prodromal stages: subjective and mild cognitive impairment. Although the analysis revealed task-specific frontoparietal networks, it consistently showed that hypometabolism in one region in the left lateral prefrontal cortex—the inferior frontal junction area—was related to performance in the various neuropsychological tests. This brain region has recently been related to the three component processes of cognitive control—working memory, task switching and inhibitory control. Group comparisons additionally showed hypometabolism in this area in Alzheimer's disease and frontotemporal lobar degeneration. Our study underlines the importance of the inferior frontal junction area for cognitive control in general and for executive deficits in early dementia.
inferior frontal junction
Executive functions describe a wide variety of higher order cognitive processes that allow the flexible modification of thought and behaviour in response to changing cognitive or environmental contexts (Stuss, 1992; Rabbitt, 1997). They are crucial for coping with the changing demands of everyday life. Executive functions have been related to the frontal lobes, namely the anterior cingulate/posterior medial frontal cortex and lateral prefrontal cortices (Carter et al., 1998; Schroeter et al., 2002, 2007a; Derrfuss et al., 2004; Ridderinkhof et al., 2004). Work by Miyake et al. (2000) using a latent variable approach analysing data from a large sample of young adults on several complex cognitive tasks suggests that executive control may be best conceptualized as comprising at least three different component processes: working memory, task switching and inhibitory control. One comprehensive systematic and quantitative meta-analysis on studies with functional MRI using the activation likelihood estimate approach identified the inferior frontal junction area, a border zone between Brodmann Area (BA) 6/8/44, as essential for task switching and inhibitory control as measured by Stroop task performance (Brass et al., 2005; Derrfuss et al., 2005, 2009). Another wide-ranging activation likelihood estimate meta-analysis confirmed this finding and also demonstrated the relevance of this brain region for working memory in addition to a relevant overlap of the three cognitive component processes in the posterior medial frontal cortex (Christ et al., 2009).
A range of tests is used to measure executive functions. A ‘classical’ test is the Stroop colour-word interference task (MacLeod, 1991), where subjects are requested to name the colour of a word with an incongruent meaning. Subjects have to inhibit an overlearned response (reading) in favour of a novel response (colour naming). Interference resolution and response inhibition may be examined with this task. The verbal fluency task is another classical test involving executive functions. Here, subjects have to produce as many words as possible beginning with a specific letter or from a specific category (Crawford and Henry, 2005).
Since many patients with noticeably impaired adaptive behaviour in daily activities still score within the norms in neuropsychological testing, it has been discussed that classical executive tests do not capture essential aspects of executive abilities that are particularly relevant to daily living (Shallice and Burgess, 1991; Burgess, 1997; Rabbitt, 1997). These tests mostly have a specific goal, rules to be followed and are generally of a relatively clear structure—there may even be direct feedback concerning mistakes—all of which are generally uncommon for the demands of daily life. More recent test batteries such as the Behavioural Assessment of the Dysexecutive Syndrome (BADS; Wilson et al., 1996) incorporate these demands to a greater extent by tapping everyday executive functions. Indeed, the BADS is a better predictor of everyday problems than, for instance, the Wisconsin Card Sorting Test (Strauss et al., 2006).
Our current study aimed to systematically explore the neural correlates of executive functions by relating deficits in this domain to regional hypometabolism as measured with [18F]fluorodeoxyglucose-PET in early dementia. Neuropsychological tests included in this study were chosen so as to cover different aspects of executive functions, namely inhibition of interfering information (Stroop Test), semantic or category fluency (Supermarket Test), problem solving (subtest Action Program from the BADS), planning (subtest Zoo Map Test) and strategy building (subtest Key Search from the BADS).
The cohort involved 54 subjects with various kinds of early dementia, mainly Alzheimer's disease and FTLD, and in the prodromal stages of subjective and mild cognitive impairment (Winblad et al., 2004; Jessen et al., 2010). Whereas subjective cognitive impairment is characterized by complaints of cognitive impairment that cannot be confirmed by neuropsychological testing, such impairments are detectable in mild cognitive impairment. We decided to involve such a broad group of subjects for several reasons: (i) we were interested in the neural correlates of executive functions per se, and this approach may increase variance and accordingly, statistical power; (ii) this method was applied successfully in recent comprehensive studies (Rosen et al., 2005; Rankin et al., 2006); (iii) Alzheimer's disease and FTLD have recently been discussed as occurring in a spectrum of neurodegenerative brain diseases (van der Zee et al., 2008); and (iv) previous studies indicated that executive deficits have similar severity, and are related to regionally similar neural correlates in different kinds of dementia, such as Alzheimer's disease and FTLD (Hutchinson and Mathis, 2007). Although the latter view has been questioned, at least for fluency and regional atrophy in FTLD subtypes (Laisney et al., 2009; Libon et al., 2009), our approach may prevent the possibility that we just replicate disease-specific regional structural or functional alterations in the correlation analyses, because we correlate deficits across a spectrum of neurodegenerative diseases.
We hypothesized that executive deficits in early dementia and its pre-stages are mainly associated with hypometabolism in prefrontal regions (Carter et al., 1998; Schroeter et al., 2002, 2007a; Ridderinkhof et al., 2004; Derrfuss et al., 2005, 2009). Because it is well known that behavioural disorders are also influenced by impairments in frontal lobe function (Rosen et al., 2005; Schroeter et al., 2007c, 2008, 2011; Raczka et al., 2010), we compared our results with the neural correlates of behavioural disorders as measured with the Neuropsychiatric Inventory in the same cohort as published in a previous paper (Schroeter et al., 2011). Finally, we identified brain regions affected in dementia by comparing regional metabolism in Alzheimer's disease and FTLD with subjective cognitive impairment as a control group. We chose this control group, because subjects with subjective cognitive impairment are not impaired in neuropsychological testing, and accordingly show a normal age-related decrease in cognitive performance. This control group is particularly suitable for diagnostic reasons, because in clinical practice it is crucial to discriminate between these subjective cognitive impairment subjects and patients with an early stage of dementia (Dukart et al., 2011). The subjective cognitive impairment cohort offers the unique opportunity to regard it both as a pre-stage of dementia and a control group—an advantage that is underlined by the ethical problems of conducting [18F]fluorodeoxyglucose-PET in healthy subjects due to radiation exposure.
Materials and methods
Fifty-four subjects who were admitted to the Day Clinic of Cognitive Neurology at the University of Leipzig were included in this study. They had presented with complaints of cognitive and/or behavioural alterations by their own account or by the account of caregivers. Investigations included a high-quality [18F]fluorodeoxyglucose-PET scan, a comprehensive neurological and psychiatric history and examination, neuropsychological rating of behavioural deficits (Neuropsychiatric Inventory), testing of memory, executive function, attention and language (Schroeter et al., 2005, 2007a, b, 2010), structural MRI or CT scans, and lumbar puncture. Symptom severity was measured with the Clinical Dementia Rating Scale (Morris, 1997). Although this instrument has been developed for Alzheimer's disease, to date no instrument is available that considers all the different dementia syndromes equally. Patients were excluded if structural imaging revealed lesions due to stroke, traumatic head injury, brain tumour or inflammatory diseases. All data were acquired for diagnostic purposes. Informed consent was obtained from subjects to analyse these data retrospectively. The research protocol was approved by the ethics committee of the University of Leipzig, and was in accordance with the latest version of the Declaration of Helsinki.
Groups consisted of patients with probable Alzheimer's disease (n = 19; McKhann et al., 1984; Dubois et al., 2007), FTLD (n = 13; FTD = 3, semantic dementia = 6, mixed type = 4; Neary et al., 1998) and ‘other dementias’ (n = 10; posterior cortical atrophy n = 3, dementia with Lewy bodies n = 2, corticobasal degeneration n = 1, logopenic aphasia n = 1, dementia with symptoms of Alzheimer's disease/FTD n = 2 and with symptoms of Lewy body dementia/posterior cortical atrophy n = 1; McKeith et al., 1996; Mendez et al., 2002; Boeve et al., 2003; Gorno-Tempini et al., 2004). Two other patients suffered from mild cognitive impairment, presenting with complaints of cognitive impairment and were at least 1.5 SD below average performance in neuropsychological testing (Winblad et al., 2004). Another group presented with complaints of cognitive impairment that could not be confirmed by neuropsychological testing (subjective cognitive impairment; n = 10).
Group differences of clinical characteristics (Table 1) were examined with one-way ANOVAs followed by unpaired two-tailed post hoc Student's t-tests (Bonferroni corrected and adjusted for inequality of variance if necessary). As expected, Clinical Dementia Rating Scale as a measure of symptom severity differed between groups. Mean Clinical Dementia Rating Scale was ∼1 in all three dementia groups, indicating early stages of disease. Post hoc Student's t-tests revealed higher values in dementia than subjective cognitive impairment (Alzheimer's disease/FTLD/other dementias versus subjective cognitive impairment T = −3.9, −3.6, −4.3; df = 27, 21, 18; P < 0.05). Mean age differed significantly between groups, in particular, between subjective cognitive impairment and Alzheimer's disease/other dementias (T = −3.4, −3.6; d = 27, 18; P = 0.02), whereas there were no differences regarding education, duration of symptoms and sex between groups.
↵b As tested with two-tailed test: χ2, P. Four degrees of freedom for each test. Mean ± standard deviation.
Significant values (P < 0.05) are written in bold.
Measurement of executive functions and behavioural impairment
In the Stroop Test, subjects are shown a list of names of colours (e.g. ‘red’, ‘blue’, etc.). In the critical incongruent condition, the print colour never corresponds to the colour of the written word. The subjects are asked to name the colour the word is written in. Here, they have to inhibit the overlearned response, reading the word (MacLeod, 1991; Schroeter et al., 2002, 2003, 2004, 2007a, b; Derrfuss et al., 2005). Response times for the critical condition per word were included in the analysis (Oswald and Fleischmann, 1986; Wolfram et al., 1986). The Supermarket Test is a test of semantic or category fluency that has recently been shown to be a very sensitive marker of cognitive decline in dementia (Cerhan et al., 2002). Subjects are asked to name as many different things that are available in a supermarket within a given time limit. This test is contained in the DemTect, a German language screening test for early dementia (Kalbe et al., 2004). The number of correct words given in 1 min was included in the analysis.
The BADS (Wilson et al., 1996) is a battery consisting of six different subtests that are all intended to tap everyday executive functions. In each subtest, performance is indicated in raw scores, which are then transformed into profile scores. Since profile scores are much less fine-grained compared with raw scores, we used the latter in all subtests for the correlation analysis. The Action Program task is conceptualized to test practical problem-solving abilities. Patients are instructed to extract a small cork from a tube using a set of different materials in a five step procedure. Scoring is based on how much help the patient requires from the examiner; raw scores range from 0 (no solution) to 5 (perfect solution without help). The Zoo Map Test taps into abilities of planning. In this test, a route has to be found in order to visit six designated locations on a map of a zoo, following certain rules. It consists of two parts: the Zoo Map Test 1 is highly demanding since little external structure for solving the task is provided; the Zoo Map Test 2 demands are much lower, since patients are simply asked to follow a pre-specified order of steps. Each part is scored according to certain criteria (number of locations visited, breaking rules, etc.) ranging from 0 to 8. In the Key Search subtest, subjects need to generate a strategy to search a field for a lost key, marking their path on a sheet using a pen. The efficiency of their strategy is scored by means of a set of criteria from which a total raw score ranging from 0 to 16 is computed.
Behavioural impairment was assessed with the Neuropsychiatric Inventory, a validated, caregiver-based behavioural rating system (Cummings, 1997). This inventory evaluates 12 major behavioural disorders in dementia, which include delusions, hallucinations, depression, anxiety, aggression/agitation, euphoria, apathy, irritability/lability, disinhibition, aberrant motor behaviour, sleep disturbances, and changes in appetite and eating behaviour. A composite score is calculated for each domain by multiplying the values for frequency (1–4/occasionally to very frequently) and severity (1–3/mild–severe).
For some subjects, test results for all tests were not available as, for practical reasons, tests were adapted to clinical needs. Some tests could not be carried out due to lack of time and/or reduced patients’ capability. For instance, the Zoo Map Test cannot be performed in subjects with remarkable impairments in the understanding of instructions. Potential relationships between executive deficits and behavioural impairments were examined with Pearson's correlation coefficients.
[18F]fluorodeoxyglucose-PET imaging and analysis
Dynamic PET scans were acquired after intravenous injection of 370 MBq [18F]fluorodeoxyglucose using an ECAT EXACT HR+ scanner (CTI/Siemens) in 2D mode simultaneously collecting 63 slices with an axial resolution of 5 mm full width at half maximum and an in-plane resolution of 4.6 mm. Subjects fasted for one night before PET imaging. Basic image processing and voxel-based data analysis were performed using Statistical Parametric Mapping (SPM5) routines (Wellcome Department of Cognitive Neurology; http://www.fil.ion.ucl.ac.uk/spm) implemented in Matlab 6.0 (Mathworks Inc.).
The 30–60 min post-injection time frames of the dynamic PET scans were realigned and a mean image was created for each subject. To enable an interindividual comparison of the subjects’ regional brain [18F]fluorodeoxyglucose uptake, each voxel of the mean images was normalized to the average intensity value of the cerebellum, which is one of the least affected regions in dementia (Yakushev et al., 2008; Dukart et al., 2010, 2011). After this activity normalization, PET image data were non-linearly spatially normalized into Montreal Neurological Institute (MNI) space using the SPM5 standard PET brain template. Normalized images were represented with 2 × 2 × 2 mm voxel size. As a final preprocessing step, images were smoothed with a Gaussian kernel of 12 mm full width at half maximum.
Correlation between glucose utilization and executive functions
[18F]fluorodeoxyglucose-PET images and the scores of the various executive tests of each of the 54 subjects were entered into a design matrix and the relationship between changes in glucose utilization and executive deficits was analysed using the general linear model. The significance of each effect of interest was determined using the theory of Gaussian fields. More precisely, test scores of each test/subtest were entered into separate ‘multiple regression’ design matrices. Age was always included as an additional covariate, as patterns of glucose utilization have been shown to change with increasing age and age differed between groups (Loessner et al., 1995; Dukart et al., 2010).
The relationship between voxel values of glucose utilization and the executive impairment of interest was examined with a-1 t-contrast, with the assumption that increasing severity of executive deficits would be associated with decreased glucose utilization. Diagnosis was generally not included in the model so as to provide greater variability of the data, since we were interested in the neural correlates of executive deficits across the various dementia subtypes, and different subtypes overlap with regard to these impairments. For display purposes illustrating the relevant clusters, we used P < 0.001, uncorrected for multiple comparisons. We identified brain regions that remained significant after correction for multiple comparisons on the cluster level with a statistical threshold of P < 0.05. Data were analysed at the whole brain level to avoid any a priori assumptions.
Conjunction analysis across various executive deficits
As discussed above, several aspects of executive functions have been suggested to be associated with frontal lobe function. To identify the most important and consistent brain regions involved in executive deficits, we conducted a conjunction analysis across the neural correlates of the various executive tasks (for this analysis, results of the correlation analysis with P < 0.001, uncorrected for multiple comparisons, were used). The first conjunction analysis investigated the overlapping brain regions. The respective t-maps were transformed into maps of z-scores. Then, for each voxel, we computed the number of z-maps (executive functions) whose value in this voxel was above the threshold 3.1 (P < 0.001). In a second analysis, we generated a map which shows the minimum z-value for all executive functions for each voxel. This map corresponds to a conjunction analysis in the sense of a logical ‘AND’ of all input images (Nichols et al., 2004).
Dissociation between executive and behavioural deficits, and group comparisons
To investigate the relationship between executive and behavioural impairments, we compared the results of the present conjunction analysis to our previous data from a similar study on behavioural deficits, which systematically investigated their neural correlates in exactly the same cohort (Schroeter et al., 2011). The comparison was again performed using a conjunction analysis. Finally, we analysed regional hypometabolism in patients with Alzheimer's disease and FTLD in comparison with subjects with subjective cognitive impairment as a control group in our cohort. For details of the methods and results, see Dukart et al. (2010). Age was again included as a covariate, and cerebellar normalization of [18F]fluorodeoxyglucose-PET data was applied.
Executive functions in early dementia and its pre-stages
Table 1 illustrates neuropsychological test results for the various subject groups. Groups differed significantly with respect to the Stroop task, fluency, Zoo Tests and Key Search of the BADS (one-way ANOVA). Performance in the Action Program of the BADS did not show differences between groups. A post hoc analysis was conducted for subjects with subjective cognitive impairment, Alzheimer's disease and FTLD groups with unpaired two-tailed post hoc Student's t-tests, corrected for multiple comparisons according to Bonferroni, and adjusted for inequality of variance if necessary. We excluded patients with other dementia subtypes due to the lack of specificity of the dementia syndrome and subjects with mild cognitive impairment due to their small number from this analysis.
Both dementia groups showed impairments in the Stroop task in comparison with subjects with subjective cognitive impairment (Alzheimer's disease: T = −3.7, df = 8.3, P = 0.02; FTLD: T = −3.2, df = 7.2, P = 0.04), whereas there were no significant differences between patients with dementia (T = −0.3, df = 15, P = 0.78). The same was the case for fluency (subjective cognitive impairment versus Alzheimer's disease: T = 4.0, df = 15, P = 0.003; versus FTLD: T = 3.2, df = 11, P = 0.03; Alzheimer's disease versus FTLD: T = 0.7, df = 20, P = 1.0) and Key Search (subjective cognitive impairment versus Alzheimer's disease: T = 3.0, df = 23.9, P = 0.02; versus FTLD: T = 5.2, df = 17, P = 0.001; Alzheimer's disease versus FTLD: T = 2.2, df = 23, P = 0.12). Regarding both Zoo Tests, patients with Alzheimer's disease were more impaired than subjects with subjective cognitive impairment (Zoo 1: T = 7.3, df = 11.1, P = 0.001; Zoo 2: T = 3.3, df = 14, P = 0.02), whereas there was no difference between subjective cognitive impairment and FTLD (Zoo 1: T = 1.2, df = 14.1, P = 0.78; Zoo 2: T = 1.0, df = 16, P = 0.99). For the Zoo 1 test, there was a significant difference between Alzheimer's disease and FTLD (T = −3.6, df = 8.8, P = 0.02), and for the Zoo 2 test, there was a trend regarding this group difference (T = −2.4, df = 20.7, P = 0.08). Summing up the behavioural results, executive deficits were higher in both kinds of early dementia in comparison to subjects with subjective cognitive impairment for five out of the six tests applied, whereas the only difference between Alzheimer's disease and FTLD was worse performance for the Zoo Test of the BADS in Alzheimer's disease.
Although almost all results for executive tests correlated with each other (P < 0.05; correlation according to Pearson across the whole cohort of 54 subjects, two-tailed; except Zoo 1 and Key Search P = 0.16, Zoo 1 and Action Program, Zoo 2 and Stroop P = 0.4), executive and behavioural deficits were dissociated. Out of apathy, disinhibition and eating disorders, the most frequent behavioural disorders in early FTLD, Alzheimer's disease and also in our cohort (Schroeter et al., 2011), only apathy correlated with performance in the Stroop task (r = 0.49, P = 0.005) and fluency (Supermarket Test) (r = −0.37, P = 0.02). All other correlations did not reach significant values (P > 0.1). The same was the case for depression.
Correlation between executive deficits and reductions in brain glucose utilization
Results for the correlation analysis, including all 54 subjects, are illustrated in Figs 1 and 2, and Tables 2 and 3.
Correlation between reductions in brain glucose utilization and executive dysfunctions—Stroop Test (Stroop), Category Fluency (Fluency) and subtest Action Program (Act. Pr.) of the BADS. Brain sections (location illustrated by crosshairs) visualize relevant neural clusters. Age as covariate. P < 0.001 uncorrected. Left is left. Colour scales represent t-values.
Correlation between reductions in brain glucose utilization and executive dysfunctions—Subtests Zoo Map (Zoo; Part 1 and 2) and Key Search (Key S.) of the BADS. Brain sections (location illustrated by crosshairs) visualize relevant neural clusters. Age as covariate. P < 0.001 uncorrected. Left is left. Colour scales represent t-values.
Age was included as a covariate in the analysis. Cluster size is reported in voxels, voxel size = 2 × 2 × 2 mm. P(uncorrected) < 0.001, equivalent to Zoo Map (Part 1) T = 3.32, Zoo Map (Part 2) T = 3.33, Key Search T = 3.30, Action Program T = 3.33.
a P < 0.01 corrected for multiple comparisons on the cluster level. Clusters >30 voxels are reported. Coordinates are in MNI space.
Impairment in behavioural performance related to Stroop interference was correlated with hypometabolism in the left inferior frontal junction and posterior superior and middle frontal gyrus (BA8/9).
Reduced semantic fluency was associated with hypometabolism again in the left hemisphere, namely the inferior frontal junction (BA6/8/44), all frontal gyri (BA8/9, 44–47), anterior and middle insula and deep frontal operculum (BA14/15). Moreover, the analysis revealed clusters in the inferior temporal gyrus and sulcus (BA20/21), caudate head and lentiform nucleus.
Behavioural Assessment of the Dysexecutive Syndrome
Impairments on the Action Program of the BADS were correlated with reduced brain glucose metabolism in a left hemispheric network (Fig. 1, Table 3). This contained the inferior frontal junction (BA6/8/44), pars orbitalis and triangularis of the inferior frontal gyrus (BA10/45–47), the anterior insula (BA15/16), anterior and lateral orbital gyrus (BA11/12), inferior temporal pole and sulcus (BA20/21/37/38) and the parahippocampal gyrus (BA20). Deficits in the subtest Key Search of the BADS were associated with regional hypometabolism in the left inferior frontal junction (BA6/8/44), anterior insula (BA15/16), inferior frontal gyrus, namely the pars triangularis and orbitalis (BA11/45/47), lateral and anterior orbital gyrus (BA11/12), left inferior temporal pole and gyrus (BA20/21/38) (Fig. 2, Table 3).
In contrast to the aforementioned tests that were mainly related to the left lateral prefrontal lobe, both Zoo Map Tests showed a different pattern, mainly involving the parietal cortex bilaterally. More specifically, the Zoo Map Test Part 1 was related to hypometabolism bilaterally in the posterior inferior parietal lobule (BA 7/39), inferior precuneus (BA7), posterior cingulate gyrus/retrosplenial cortex/parieto-occipital fissure (BA18/19/23/29/30), right temporoparietal junction area (BA22), posterior middle temporal gyrus (BA21/37), left posterior inferior temporal sulcus/anterior occipital sulcus (BA19/37/39) and right lateral and superior occipital gyri (BA19). Additionally, as in the former tasks, the inferior frontal junction (BA6/8/44) and other frontal regions, namely the posterior middle frontal gyri/frontal eye fields (BA6/8), were bilaterally involved. In contrast, areas showing hypometabolism in correlation with the Zoo Map Test Part 2 were limited to parietal areas. The clusters spread across the left posterior inferior parietal lobule (BA 7/39), bilateral inferior precuneus (BA7), parieto-occipital fissure (BA18/19/23/31), left temporoparietal junction area (BA22), right angular gyrus (BA39), the anterior horizontal part of the intraparietal sulcus (BA7/39) and superior occipital gyrus (BA18/19).
After correction for multiple comparisons on the cluster level with P < 0.05, the analysis confirmed relevant clusters for fluency, both Zoo Map Tests, and the Key Search subtest of the BADS (Tables 2 and 3). The clusters included the inferior frontal junction with respect to fluency and Key Search of the BADS, whereas for the Stroop test, Action Program and Zoo Map Test Part 1, the cluster in the inferior frontal junction was presumably too small.
Executive deficits and glucose utilization: conjunction analysis
To isolate the brain region most consistently involved in executive deficits, we conducted a conjunction analysis across the neural correlates of the various executive tasks. This analysis included the Stroop and fluency test, and the subtests Action Program, Key Search and Zoo Map Test Part 1 of the BADS. Part 2 of the Zoo Map Test was excluded from this analysis, since patients are simply asked to follow a pre-specified order of steps, and accordingly, executive functions are less involved. This assumption was supported by the absence of a correlation with frontal regions for the latter test. Figure 3 illustrates the results of the conjunction analysis. All five executive tests were related to hypometabolism in one prefrontal region, namely the left inferior frontal junction (BA6/8/44). This finding was observed for both types of conjunction analyses. In the map of overlapping regions, there was z > 3.1 (P < 0.001) for all five executive functions in (maximum) −44, 17, 29 (Fig. 3A). The map of minimum z-values for all executive functions (Nichols et al., 2004) showed its maximum in the same voxel (−42, 20, 30) (Fig. 3B). Most strikingly, these results coincided anatomically with a recent systematic and quantitative meta-analysis (Derrfuss et al., 2005), which used the activation likelihood estimate method to investigate the neural correlates of Stroop and switching in functional MRI studies (Fig. 3C; respective coordinates of maxima after transformation to MNI space −43, 7, 32 and −43, 7, 38).
Conjunction analyses. Brain regions that showed consistently reduced glucose metabolism in association with executive deficits. (A) Overlap of t-maps of all five tests. Maximum is located in the vicinity of the inferior frontal junction area (i.e. in this region, t-values of all statistical tests are significant with voxelwise P < 0.001). Scale represents number of overlapping tests. (B) Minimum z-score for all five executive function tests which corresponds to a conjunction analysis in the sense of a logical ‘AND’. Maximum of the resulting map of z-values is again located in the vicinity of the inferior frontal junction. (C) Results of the meta-analysis of studies investigating Stroop (Str) and Switch (Sw) tasks with functional MRI transformed into the MNI space using the unified segmentation approach (Ov = overlap; Derrfuss et al., 2005). (D) Results of our previous study with exactly the same cohort investigating the neural correlates of behavioural disorders in dementia—correlation between reduced glucose metabolism and apathy (Ap), disinhibition (Dis) and eating disorders (Ea) (Schroeter et al., 2011).
Dissociation between executive and behavioural deficits in early dementia
We compared the results of our conjunction analysis with the results of our previous study that systematically investigated the neural correlates of behavioural disorders in exactly the same cohort with the same method ([18F]fluorodeoxyglucose-PET) (Schroeter et al., 2011). This study isolated three relevant behavioural disorders, namely apathy, disinhibition and eating disorders, out of all components of the Neuropsychiatric Inventory that correlated with regional brain hypometabolism. The comparison of the results from our multiple regression analysis including all three behavioural disorders (Fig. 3D) with the results of our present conjunction analysis (Fig. 3A and B) revealed no overlapping brain regions. Hypometabolism in the inferior frontal junction seems to be specifically related to executive dysfunctions, but not related to behavioural disorders in our study cohort. Results agree with the observation that executive deficits hardly correlated with behavioural disorders in the study group.
Regional reductions in glucose metabolism in early dementia: group comparisons
Finally, we analysed regional hypometabolism in patients with Alzheimer's disease and FTLD in our cohort by comparing them separately with control subjects (subjective cognitive impairment) (Fig. 4). Interestingly, the analysis revealed regionally overlapping reductions in glucose utilization for both Alzheimer's disease and FTLD mainly in the left inferior frontal junction (cluster size 1477 voxels; maximum in MNI space −44, 4, 60), and additionally, but to a smaller extent, in the left inferior temporal (514; −62, −28, −18) and angular gyrus (228; −56, −62, 36).
Group comparison—brain regions that showed reduced glucose metabolism in Alzheimer's disease (AD, red) and frontotemporal lobar degeneration (FTLD, blue) in comparison with subjects with subjective cognitive impairment (SCI). Overlap (purple) includes the left inferior frontal junction area. Age as covariate, P < 0.001 uncorrected, extent threshold of 30 voxels. Left is left. Note that coordinates in MNI space are identical to Fig. 3.
Executive functions and the inferior frontal junction area
Our study systematically investigated the neural correlates of executive functions by correlating reductions in glucose utilization with executive deficits in early dementia and its pre-stages as measured by classical tests, such as Stroop and fluency tasks and recently developed tasks investigating executive functions in a context more relevant to daily life, namely the BADS. In agreement with our hypothesis, executive deficits were mainly associated with prefrontal hypometabolism. Most interestingly, we identified the inferior frontal junction (BA 6/8/44) as the most consistently relevant region across all executive tasks. Miyake suggested that executive control may be best conceptualized as comprising at least three different component processes: working memory, task switching and inhibitory control. Indeed, two comprehensive systematic and quantitative meta-analyses of functional MRI studies using the activation likelihood estimate approach have identified the inferior frontal junction as being relevant for the three component processes (Brass et al., 2005; Derrfuss et al., 2005; Christ et al., 2009). In sum, our study indirectly supports Miyake et al's (2000) assumption that executive control comprises the three component processes and the relevance of the inferior frontal junction.
The inferior frontal junction is located at the junction of the inferior frontal sulcus and the inferior precentral sulcus (Derrfuss et al., 2009). If one follows the assumption that the border zone between the dorso- and ventrolateral prefrontal cortex is constituted by the inferior frontal sulcus, the inferior frontal junction covers both areas (Sakai et al., 2008). Others, based on monkey studies, have argued that the border zone is constituted by the middle frontal sulcus (as the analogue of the monkey's sulcus principalis)—a region within the middle frontal gyrus (Petrides and Pandya, 1999; see Fig. 3 in Petrides and Pandya, 2002). In this case, the inferior frontal junction would be located in the ventrolateral prefrontal cortex and one might conclude from our data that executive functions in early dementia are not related to the dorsolateral prefrontal cortex. However, if the results for the single executive tests are examined instead the conjunction analysis, test performance also correlated with hypometabolism in the dorsolateral prefrontal cortex according to its second definition (at least for three out of the five relevant ‘frontal’ neuropsychological tests: Stroop, fluency, Zoo 1 task; Figs 1–3, Tables 2 and 3). These effects disappeared in the conjunction analysis, presumably due to a high regional variability.
Executive deficits are related to the inferior frontal junction area in early dementia
In the following, we discuss our results in the context of executive deficits in dementia. Interestingly, inhibitory functions that are relevant in the Stroop task are the first component of the attentional system to become impaired after amnestic symptoms in early Alzheimer's disease (Amieva et al., 2004; Hodges, 2006). Thereafter, divided attention as tested with switching tasks is affected (Baddley et al., 2001). Breakdown in higher order cognitive (executive) abilities is virtually always present in patients with established Alzheimer's disease, in agreement with impairments of the frontal lobes (Perry and Hodges, 1999; Hodges, 2006; Schroeter et al., 2009).
Although most of the executive tests showed a main focus in the left lateral prefrontal lobe, the Zoo Map Test mainly involved the parietal cortex bilaterally. This test poses considerable demands on visuospatial constructional abilities and spatial navigation, which are impaired in patients with Alzheimer's disease (Hodges, 2006) and have been related to the inferior parietal lobules extending into the intraparietal sulcus and the precuneus—regions that are particularly affected in Alzheimer's disease (Cavanna and Trimble, 2006; Schroeter et al., 2009). This explanation is further supported by the observation that patients with Alzheimer's disease were more impaired in both parts of the Zoo Map Test than the other dementia subgroups in our cohort—whereas for the other executive tests, we did not find group-specific differences. Interestingly, the difference between Alzheimer's disease and FTLD disappeared when we compared the difference between performance in Part 1 and Part 2 of the Zoo Map Test, isolating more strictly executive functions (as discussed above) than those involved in the single tests (−5.1 ± 3.4 versus −2.6 ± 3.5; T = −1.7, df = 21, P > 0.1; two-tailed unpaired Student's t-test).
Executive deficits in early dementia and the frontomedian cortex
Surprisingly, we did not detect a correlation between hypometabolism in frontomedian regions and executive functions, although the posterior medial frontal cortex is involved in performance monitoring in relation to anticipated reward (Ridderinkhof et al., 2004), and activation in this brain region has been related to the three component processes of executive control (Miyake et al., 2000; Christ et al., 2009). This result might be related to the fact that the focus of frontomedian impairments in Alzheimer's disease and FTLD/FTD as shown in comprehensive activation likelihood estimate meta-analyses (Schroeter et al., 2007c, 2008, 2009; Schroeter and Neumann, 2011) is located more anteriorly (anterior medial frontal cortex and pregenual anterior cingulate gyrus) than functional neural correlates for the detection of response conflict, errors and unfavourable outcomes (Ridderinkhof et al., 2004) and the three component processes of executive control (Christ et al., 2009).
In sum, our study shows that patients with both early Alzheimer's disease and FTLD are impaired in executive functions, and that executive dysfunctions are related to the inferior frontal junction. Inferior frontal junction dysfunction might impair the three component processes of cognitive control, working memory, task switching and inhibitory control (Miyake et al., 2000; Derrfuss et al., 2005; Christ et al., 2009), and finally lead to impairment in all executive tests as applied in our study. The mechanisms of this impairment, namely the specifically involved cognitive processes, need to be investigated in future studies.
It is well known that the BADS captures essential aspects of executive abilities that are relevant to daily living (Wilson et al., 1996; Strauss et al., 2006). Indeed, results in this test battery were, on average, more closely related to dementia severity (Clinical Dementia Rating Scale) than the other classical executive tests in our cohort (Action Program r = −0.56, Zoo Map Test 1 and 2 r = −0.53, −0.65, P < 0.001; Key Search r = −0.38, P = 0.01; Stroop and fluency task r = 0.48, −0.36, P = 0.005, 0.03, respectively). One might also conclude that impairment of the inferior frontal junction influence, deficits in every day life in early dementia.
Dissociation between executive and behavioural deficits in early dementia
Another aim of the present study was to compare the neural correlates of executive deficits and behavioural disorders in early dementia, as it is widely known that both are related to the frontal lobes (Carter et al., 1998; Derrfuss et al., 2004, 2005; Ridderinkhof et al., 2004; Rosen et al., 2005; Raczka et al., 2010). Accordingly, we compared our results with those of our previous study systematically investigating the neural correlates of behavioural disorders in exactly the same subjects with the same imaging method (Schroeter et al., 2011). Our results suggest that the inferior frontal junction is specifically related to executive dysfunctions and not behavioural deficits in early dementia, as the second conjunction analysis did not reveal any overlapping brain regions and executive deficits hardly correlated with behavioural disorders in our cohort.
Limitations of the study
Finally, the limitations of the study must be acknowledged. As already discussed in the ‘Introduction’, we included a diverse group of subjects ranging from subjects with various kinds of early dementia to its prodromal stages, subjective cognitive impairment and mild cognitive impairment. We decided to involve such a study cohort, because we were interested in the neural correlates of executive functions per se, and we wanted to include a large variance of symptom severity and accordingly increase statistical power—an approach that has been applied successfully in other studies (Rosen et al., 2005; Rankin et al., 2006). In fact, we were able to systematically reveal the correlates of executive impairments in early dementia. Since both FTLD and Alzheimer's disease showed comparable deficits in executive functions, regional hypometabolism in the inferior frontal junction and the dementia subgroups did not differ regarding severity, education and disease duration, one might assume that our results are relevant for both early Alzheimer's disease and FTLD. However, the comparison with the subjective cognitive impairment group could not be performed for patients with dementias other than Alzheimer's disease/FTLD, because this dementia subgroup consisted of diverse dementia subtypes. Regarding the trade-off between specificity for dysexecutive syndromes and for dementia subtypes, we elected to give preference to the former, as we were interested in the specific neural correlates for executive deficits per se. One might suggest conducting the same kind of study to determine the role of the inferior frontal junction in executive function in healthy subjects. However, performing [18F]fluorodeoxyglucose-PET in this cohort is almost impossible due to ethical issues (radiation exposure). Furthermore, for the BADS one might expect ceiling effects in control subjects, because they are not impaired in executive functions. One might further criticize the fact that we did not correct for atrophy. However, most studies that have performed voxel-based atrophy correction of resting glucose metabolism have reported a relative independence of metabolism from brain atrophy, particularly for localization of maxima (Bracco et al., 2007; Drzezga et al., 2008)—a finding that could be replicated in our cohort (Dukart et al., 2010).
Our study examined the neural correlates of executive functions by correlating respective test scores with reductions in glucose utilization in early dementia and its prodromal stages. The analysis consistently isolated one region in the left lateral prefrontal cortex, namely the inferior frontal junction area, across various neuropsychological tests. Results underline the importance of this brain region for cognitive control processes in general, and for executive deficits in early dementia in particular.
LIFE - Leipzig Research Centre for Civilization Diseases at the University of Leipzig (to M.L.S., H.B., A.V. and O.S.). European Union, European Regional Development Fund (ERDF) and Free State of Saxony within the framework of the Excellence Initiative to LIFE. German Federal Ministry of Education and Research (BMBF) - German FTLD consortium (to M.L.S.).
The authors thank the PET and cyclotron staff at the Department of Nuclear Medicine and the staff at the Day Clinic for Cognitive Neurology at the University of Leipzig for help in data acquisition, and Jan Derrfuss and Jürgen Dukart for providing the results of their studies. The authors are most grateful to all subjects participating in the study.
behavioural assessment of the dysexecutive syndrome
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0), which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
. Neural activations at the junction of the inferior frontal sulcus and the inferior precentral sulcus: interindividual variability, reliability, and association with sulcal morphology. Hum Brain Mapp 2009;30:299-311.
. Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA work group under the auspices of department of health and human services task force on Alzheimer's disease. Neurology 1984;34:939-44.