Executive function (EF) refers to “higher-level” cognitive functions involved in the control and regulation of “lower-level” cognitive processes and goal-directed, future-oriented behaviors.1 EFs include planning, abstract reasoning, processing speed, working memory, cognitive flexibility, set-shifting, inhibitory control, and generative fluency.1 Executive dysfunction is a crucial feature of several neurodegenerative disorders such as fronto-temporal dementia (FTD), Parkinson’s disease (PD), and progressive supranuclear palsy (PSP).2

For many years, Alzheimer’s disease (AD) had been known to have memory dysfunction as an initial and prominent symptom, with the executive function being relatively spared in the early stage of the disease.3 In recent years, however, researchers reported that even mild AD patients exhibit impairments in a variety of EF tests, and over a 2-year period, mild AD with dysexecutive clinical phenotype progresses more rapidly than AD with amnestic phenotype.4 Previous studies have also shown that deficits in EF are common even in amnestic mild cognitive impairment (aMCI).56789 A growing body of evidence has shown that aMCI combined other cognitive deficits leads to greater risk for conversion into AD, than in the case of memory dysfunction alone.10 Rozzini et al.11 found that aMCI patients with worsening of EF and functional status but not of memory, were more likely to progress to AD at a 1-year follow up.

Some researchers demonstrated that the Clinical Dementia Rating-Sum of Boxes (CDR-SB) is a more detailed quantitative general index for staging of severity of dementia than the CDR-Global Score (CDR-GS), and is more accurate for tracking changes across time.1213 O’Bryant et al.14 suggested that CDR-GS of 0.5 does not represent a homogeneous group, and CDR-SB score of 2.5 may have potential for discriminating between MCI (questionable impairment) and very early AD with CDR-GS of 0.5. Moreover, the Alzheimer’s Disease Neuroimaging Initiative (ADNI Go and ADNI 2) sub-categorized MCI into early MCI (EMCI) and late MCI (LMCI), to predict conversion into AD and to develop intervention strategies for the earliest stages of the disease.15

Although the aforementioned literatures consistently indicate the relationship between executive dysfunction and aMCI as well as AD, there is a lack of research that indicates how early the deficits of EF exist in aMCI. The present study investigated whether executive dysfunctions occur in the earlier stage of aMCI (EMCI), and which sub-domains of EF are disproportionately impaired earlier than others.

The present study examined whether impairment of EF exists in MCI as well as mild AD, and studied which sub-domains of EF are differentially affected at the earlier stages. The results indicated significant declines of EF in either EMCI or LMCI for 7 out of the 9 measures in 5 EF tests, except in K-CWST: Word and K-TMT-E: Part A.

With regard to the COWAT: Animal and Supermarket, the results indicated that semantic fluency was impaired at the early stage of MCI (EMCI), which gradually worsens up to mild AD. Semantic fluency has been recognized to be dependent not only upon the integrity of semantic memory, but also on executive functioning; semantic fluency is more impaired following focal temporal and frontal lobe damage.21 Previous studies suggest that semantic memory impairment is often present in aMCI as well as in dementia of the Alzheimer’s type. It is said to be associated with injury of the temporal lobe structure.21 In line with the findings of recent literature, the results of the present study indicated that semantic fluency deficits exist even in the earliest stage of the MCI, with progression through AD, suggesting that semantic fluency is a sensitive measure to detect the early stage of aMCI, and could be a good index to monitor the progression to AD.

Results for post-hoc analysis for the COWAT: Phonemics and DSC showed that phonemic fluency and psychomotor speed were impaired at the early stage of aMCI relative to NE, but were maintained at the same level to mild AD, unlike semantic fluency. Phonemic fluency measures the generating of as many words as possible based on orthographic criteria. It requires effortful self-initiated retrieval processing.22 In addition, phonemic fluency imposes greater demands on executive skills, than semantic fluency, and many authors have reported that phonemic fluency is more sensitive to frontal, as opposed to non-frontal, lesions.22 The results of the present study indicated that the effortful self-initiated retrieval processing for words based on lexical representation was already impaired even in the early stage of aMCI. DSC is associated with visuospatial perception, processing speed, sustained attention, visuomotor coordination, and incidental memory. It is especially recognized for use in assessing the psychomotor speed.23 Several authors have emphasized that the marked characteristic of normal aging is the slowing of processing speed,2425 which itself is a very sensitive measure of brain abnormality.26 Unlike the DSC, however, we could not find any differences on K-TMT: Part A which is a known measure of processing speed among all groups. This result indicated that DSC is a more sensitive test to detect the impairments of psychomotor speed than TMT: Part A. It also showed that psychomotor slowing was already present at the early stage of aMCI. Recent studies reported that psychomotor slowing is a crucial determinant of performance on verbal fluency tests in normal elderly as well as dementia of the Alzheimer’s type.27 Our results showed the same pattern of decline both in the COWAT: Phonemics and in the DSC, supporting the common role of psychomotor speed for both tests.

With regard to the Go-No go Test, K-CWST: Color naming, and K-TMT-E: Part B, we found that EMCI performed similarly to NE, though inhibitory control and mental set-shifting progressively worsened from MCI (EMCI and LMCI) to mild AD. Go-No go Test and K-CWST (Stroop test) assess the ability to inhibit the automatized or previously learned responses, and select the appropriate responses.56 Some authors have reported that inhibitory control, which is measured by the Hayling and Stroop tests, was the most frequently and severely impaired in aMCI compared to normal elderly controls, relative to other EF tests.5 In contrast, other researchers have shown that tests for inhibition of prepotent responses (Go-No go Test and Stroop tests) failed to uncover significant group differences between normal control and aMCI.6 In our results, we observed that EMCI performed similarly to NE, but LMCI exhibited a poorer performance than NE on the Go-No go Test and K-CWST: Color. K-TMT-E: Part B measures the working memory, divided attention, and processing speed; however, it mainly requires mental set-shifting ability.28 Some authors reported that MCI performed similarly to control group on TMT: Part B,7 whereas other authors found that TMT: Part B was impaired in MCI compared to normal control.6 In the present study, we found that EMCI performed similarly to NE; in contrast, LMCI was more impaired than NE on the K-TMT-E: Part B. Many reasons could explain the inconsistent conclusions of previous studies indicating the use of different tasks, but we believe it is primarily because of different criteria for the recruitment of aMCI patients. In the current study, we subdivided aMCI patients with a CDR-GS of 0.5 into EMCI and LMCI, according to CDR-SB. As a post-hoc analysis, we re-analyzed after merging the two MCI groups (EMCI and LMCI) into one group (aMCI). We found group differences between the aMCI and NE in the Go-No go Test, K-CWST: Color, and K-TMT-E: Part B (Table 3). Since the EMCI, but not LMCI, still maintained the inhibitory control and mental set-shifting abilities, these results supported that subdividing aMCI into EMCI and LMCI based on CDR-SB would help in understanding the spectrum of impairments of EF in the disease, and to identify the progression of dementia.

To summarize, there were no group differences in simple processing speed among the NE, aMCI, and mild AD. We found that the EMCI performed worse than the NE in verbal fluency and psychomotor speed tests; however, there were no differences between the EMCI and NE in inhibitory control and set-shifting abilities. Also, no differences were observed between the EMCI and LMCI in any of the 9 measures in 5 EF tests; but the LMCI revealed worse performance than the NE in 7 measures. The EMCI exhibited better performance on semantic fluency, Go-No go Test, K-CWST: Color, and K-TMT: Part B than the mild AD; the LMCI did not show any differences with the mild AD on any measures in 5 EF tests.

These results indicated that EF impairments exist even in the earlier stage of the aMCI in several sub-domains, with some sub-domains being similar to NE. This means that the semantic and phonemic fluency and psychomotor speed decline earlier, whereas inhibitory control and mental set-shifting abilities decline later. Functional neuroimaging studies showed that patients with left dorsolateral prefrontal cortex (DLPFC) or inferior frontal gyrus lesions are impaired on verbal fluency,2930 whereas psychomotor speed is associated with superior frontal lobe (middle frontal gyrus).26 It has known that anterior cingulate cortex (ACC) and orbitofrontal cortex (OFC) play an important role in Stroop and Go-No go Test performances.313233 Set-shifting or cognitive flexibility are also known to be related with increased activation in the DLPFC and medial prefrontal cortex (PFC).34 In addition, some researchers reported that the most substantial age-related decline is the volume of prefrontal grey matter including both DLPFC and OFC areas.3536 Several authors, however, found a relative preservation of the OFC and ACC, although they found the strongest volume reduction within the DLPFC regions.353738 Taking together these findings from functional and structural neuroimaging with the present study results, we can assume that functions of the PFC which are more related to lateral regions decline at earlier stage of aMCI, whereas the EFs associated with the more orbital or medial part of PFC are impaired at later stage of aMCI.

Our results indicated that executive dysfunction is an important feature of aMCI. Thus, it should be noted that if a clinical setting cannot afford to administer a comprehensive neuropsychological assessment for aMCI, we suggest that at least two kinds of EF tests should be administered for staging aMCI. The results of the present study indicated that the COWAT and DSC are more sensitive for the detection of an early-stage of aMCI, whereas the Go-No go Test, K-CWST: Color, and K-TMT: Part B are useful for the detection of late-stage aMCI.

There are several limitations to the current study. First, this study was retrospective with a relatively small sample size. Second, we could not control the drug effects on each patient group. Third, we did not have biomarker evidences such as hypometabolism, β-amyloid deposition, and cerebrospinal fluid (CSF) proteins which would prove that our aMCI is the prodromal stage of AD, although we excluded patients who had a stroke, significant ischemic changes on brain MRI, movement problems, or salient personality changes. Fourth, all the EF tests included in our study were traditional “paper-and-pencil” tests used in clinical settings. We suggest this study should be replicated using more sensitive experimental tasks that measure the EF in detail. Finally, this study was a cross-sectional study; thus, future research should be conducted as a longitudinal base.