Working memory assessment using cambridge neuropsychological test automated battery can help in the diagnosis of mild cognitive impairment: a systematic review and meta-analysis

ABSTRACT Mild cognitive impairment (MCI) is an interstitial state between normal aging and dementia. Objective: In this study, we investigated working memory (WM) profiles of MCI patients using the Cambridge Neuropsychological Test Automated Battery (CANTAB). We also examined the diagnostic accuracy and possible associated factors as secondary outcomes of the study. Methods: We conducted an electronic search on EMBASE, PubMed, and ScienceDirect databases. Studies with MCI participants and using CANTAB battery subtests for the assessment of WM were included. Meta-analysis was conducted using the CMA2 software. Results: Out of 1537 records, 14 studies were covered in this systematic review, and 7 of them were included in the meta-analysis. There was a significant difference between MCI patients and healthy controls in spatial working memory (SWM) (SDM: 0.535; 95%CI 11–96; p-value=0.014), spatial span (SSP) (SDM: 0.649 95%CI 0.297–0.100; p-value<0.01), and rapid visual information processing (RVP) (SDM: 0.52; 95%CI 0.386–0.654; p-value<0.01). WM function of MCI patients was associated with the cerebrospinal fluid (CSF) levels of tau-protein and amyloid-beta (Aβ). Conclusions: WM is an impaired cognitive domain in MCI. CANTAB WM subtests including SSP, SWM, and RVP are accurate enough to be used as a proper assessment tool for the diagnosis of MCI in clinical settings. Tau-protein and Aβ are associated with lower WM scores in MCI patients; however, sex, age, psychiatric disorders, apolipoprotein 4 allele, and functional activity scores cannot affect WM.


INTRODUCTION
M ild cognitive impairment (MCI) is known as a transitional state between normal aging and dementia in the age continuum in which patients experience memory loss more than healthy age-matched older adults, but do not fulfill defined criteria for dementia diagnosis 1 . Based on manifestations and disease course, MCI includes different subtypes: amnestic or non-amnestic MCI and single-or multiple-domain MCI 2 . The amnestic MCI is typically associated with an increased risk of conversion to Alzheimer's disease (AD); however, non-amnestic subtypes, which may progress to non-AD dementias, may also evolve to AD 3 .
With the global increase in life expectancy, early diagnosis and precise application of disease-modifying treatments for MCI have turned into a priority for the health systems 9,10 . Previous studies following MCI patients for 6 years found that 80% of patients progress to AD with an annual rate of 10-15% 1,11 , which is 10-fold higher than the conversion rate in the normal population 12 .
Several cognitive domains such as learning, shortand long-term memory, social cognition, language, perceptual motor, complex attention, or executive functioning are characteristically affected by the pathogenesis of AD along with disease progression 13,14 . Working memory (WM) can be defined as a component of shortterm memory with a restricted capacity that depends on central executive functions and attention, utilizing stored information and linking them to long-term memory 15 . Unlike short memory which provides shortterm storage of information, WM has been proposed as a multicomponent structure that stores incoming information and operates them to a more complicated cognitive function [16][17][18] . WM is highly associated with daily functioning abilities 19 and has shown an explicit linear decreasing relationship with age 20,21 so it can be used as a measure for early diagnosis of dementia 22 .
Previous studies have shown impairment of WM in the early stage of dementia [23][24][25][26] , which makes it a good factor for early diagnosis of the disease and prevention of disease progression. Classic paper-pencil tests like Montreal Cognitive Assessment (MoCA) and Mini-Mental State Examination (MMSE) are widely being used for the assessment of MCI 27 ; however, these tests have shown some serious drawbacks with standardization of administration, the accuracy of response measurement, and demographic factors, importantly years of education and illiteracy [28][29][30][31] .
The Cambridge Neuropsychological Test Automated Battery (CANTAB) is a computerized neuropsychological test with a game-like and non-verbal environment that assesses the different cognitive domains like memory, attention, executive functions, learning, and problem-solving 32 . Among various subtests of CANTAB, spatial span (SSP) and spatial working memory (SWM) account for the assessment of WM [33][34][35] . Also, rapid visual processing (RVP) accounts for sustained attention and target detection that has a small WM component that is sensitive to parietal and frontal lobe dysfunction 36 .
In this systematic review and meta-analysis study, we aimed to study the WM function in MCI patients using CANTAB to determine the severity of WM impairment in MCI patients, as the primary outcome, and compare it with healthy matched older adults, to define the diagnostic accuracy of WM profiles of CANTAB in the detection of MCI. Also, as another secondary outcome, we investigated the associated factors of WM function in MCI patients.

METHODS
This study was conducted following the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement 37 . This systematic review was designed to assess the WM function in MCI patients using the CANTAB, and the meta-analysis was conducted to compare the differences between MCI and healthy participants in WM subtests of the CANTAB.

Search
Two independent researchers (Z.S. and A.N.) conducted a systematic literature search on EMBASE, PubMed, and ScienceDirect databases combining the keywords "Cognitive dysfunction, cognitive decline, cognitive impairment, mental deteriorations, mild cognitive impairment, CANTAB, Cambridge Neuropsychological Test Automated Battery, neuropsychological test, working memory, immediate memory, short-term memory," on December 16, 2020. For the sake of comprehensiveness, the references of each included study were checked for any additional related papers.

Study selection
Search results were imported to the EndNote reference manager. After deleting duplicated studies, two independent authors (Z.S. and A.N.) started screening and selecting papers by title/abstract in the first stage and full text in the second stage. In case of any conflicts, investigators tried to convince each other or ask for a third expert researcher's comment (M.T. or M.F.).
Inclusion criteria were as follows: 1. Original journal articles, Data extraction Data were extracted by two independent authors (Z.S. and A.N.) in a pre-specified format using a data extraction table, including the name of the first author of the study, publication year, study design, the overall number of participants as well as the number of patients in each group of the study, mean age, years of education, diagnostic criteria, MMSE score, mean and standard deviations (SD) of CANTAB WM subtests in MCI group and the healthy control group, and finally associated and non-associated factors with WM function. We could not examine amnesic and non-amnesic subtypes separately since they were not described in most of the included articles. The online version of Web Plot Digitizer was used for extracting the exact values from the graphs. Extracted data were reviewed by a third author (M.T. or M.F.) and, in case of any disagreements about results, it was determined between authors or by a judgment of a third author.

Risk of bias in individual studies
The risk of bias (RoB) and methodological quality were evaluated (by Z.S. and A.N. separately) with Joanna Briggs Institute (JBI) checklist that contains eight questions, evaluating inclusion criteria, detailed study subjects and setting, the validity of exposure, the standard measurement of the condition, and the outcome, identifying and dealing with confounding factors and statistical analysis 38 .

Statistics
In this study, meta-analysis was performed using comprehensive meta-analysis (CMA) version 2.0. The confidence interval was considered at 95% and 0.05 level of significance for the p-value. Studies that used SWM total errors, SSP length, as well as A' or latency measures of RVP subtest of CANTAB in MCI patients and healthy control group were included in the quantitative analysis. The I 2 model was also utilized for assessing the level of heterogeneity among included studies. Whenever any of the studies had reported data for MCI by subgroups (subjective MCI, amnestic MCI, single-domain MCI, multiple-domain MCI), we merged them using an excel code. The mean, SD, and the number of the individuals in each group were imported into CMA, and both the random-effect model (REM) and fixed-effect model (FEM) were utilized for assessing the difference between the groups. Also, the results of the study were reported in funnel plots in Supplementary Material.

RESULTS
Search results and selection process The electronic search identified 1,235 records through databases and 655 records added from other resources. After removing duplicates, 1,537 records were screened, and 1,434 records were excluded. Out of 66 studies that were assessed in the full-text stage, 14 studies were included in this systematic review, and 7 of them met our inclusion criteria for the meta-analysis. The PRISMA flow diagram is presented in Figure 1. Table 1 10,33,39-50 is a summary of the characteristics and findings of included studies.

Characteristics of the studies and participants
Five of included studies were cross-sectional and nine were cohorts. Only baseline data of the cohort studies are taken into account. In sum, 930 out of 1670 participants were diagnosed with MCI, and 527 were healthy controls. The mean age of the participants was between 55 and 75 years. The years of education varied from 7 to 14, and the male ratio varied between 8 and 56%.

MCI diagnosis
In this study, most of the researchers used MMSE for the diagnosis of MCI, and the rest of the studies used the other tests or criteria, such as Petersen criteria, MOCA, Rey Auditory Verbal Learning Test, and Dementia Rating Scale.

CANTAB tests for WM
Regarding the tests for WM in CANTAB, 11 of the included studies reported SWM, and 9 of them reported SSP for assessing WM. As mentioned before, RVP has a small WM component and was used in nine of our included studies. Only one study reported that used delayed matching to sample (DMS) subtest of CANTAB as an assessment tool for WM. depression, apolipoprotein 4 (ApoE4), and functional activity scores were not significantly correlated to CANTAB WM scores, while a higher cerebrospinal fluid (CSF) levels of tau-protein and amyloid-beta (Aβ) were associated with a lower function in WM tests.

Meta-analysis
Out of 14 included studies, 3 of them did not include any healthy participants for control group and 4 others did not report our intended component of CANTAB subtests to be included in the quantitative synthesis; hence, they were excluded from the meta-analysis. Seven remaining studies were included in the meta-analysis. The forest plots of the meta-analyses are shown in Figures  Studies included in quantitative synthesis (meta-analysis) (n = 7)

DISCUSSION
This study assessed the WM function of patients with MCI and compared it between MCI patients and healthy people using the CANTAB. Also, influencing factors on WM were considered. SWM, SSP, and RVP were the most commonly used subtests of CANTAB for assessing the WM. The results of quantitative synthesis revealed a significant difference between healthy controls and patients with MCI regarding the CANTAB-based WM assessments. Also, the available evidence suggested a significant correlation between CSF levels of tau-protein and Aβ with WM function in patients with MCI.
One of the preclinically deteriorated domains in AD and MCI is WM 21,51,52 . WM comprises a cognitive spectrum from attention allocation to specific stimuli to complex decision-making. Some studies have suggested WM as an early predictor of AD 53 . Regardless of the method of assessment, WM function is found to significantly deteriorate in MCI 24,54 . WM is subdivided into verbal and visual components 55 . Emrani et al. found that the visual component of WM is more sensitive than verbal WM, for distinguishing between MCI patients and healthy older adults 56 . Align with the aforementioned study, our quantitative synthesis reveals that the WM of MCI patients based on SWM, SSP, and RVP is impaired significantly, so it can be suggested as a proper diagnostic evaluation for MCI.
CANTAB is a novel neuropsychological battery for evaluating cognitive state. This battery has shown promising outcomes in the diagnosis of cognitive function in healthy older adults, MCI, AD, or any other possible diseases that may compromise cognition 32,57 . It has several benefits over traditional paper-pencil tests, such as reducing the risk of human error and data noise, recording reaction times precisely, lowering data storing problems, easing task scoring, and having access to normative comparison 32,58 . Also, CANTAB has a non-verbal structure that makes it more convenient for people with different languages 59,60 . Regarding the disadvantages, CANTAB is a time-consuming test, and providing the test instruments, imposes an extra cost to the clinicians, which limits its usage in resource-limited settings. The accuracy of WM tests of CANTAB battery in distinguishing between MCI patients and healthy older adults was studied in our review and CANTAB has shown to be a proper battery for MCI diagnosis. As a secondary outcome of the study, we assessed related factors with WM function in MCI patients. Aging is one of the confirmed predictors of cognitive decline 61 . Although WM function is found to be affected by age 62 , in most of our included studies, age was not associated with the WM scores of the patients. This may be because most of the participants in our study were older people while there is a need for the participation of patients with a wider age range to survey the age differences.
The relation between CSF biomarkers and cognitive state is one of the interest areas for research. Soldan et al. in a cohort study investigated the performance of cognitively healthy adults on CANTAB-PAL and found that it was associated with CSF p-tau levels 63 . This study suggested that the AD-related CSF biomarker can predict specific cognitive dysfunctions. In our included studies, Aβ and tau-protein were associated biomarkers with WM functions of MCI patients. On the contrary, ApoE4, which is one of the most studied genetic factors associated with human cognition and one of the wellknown predictors of AD 64 , was not associated with WM function of MCI patients, as reported in two studies 39,40 .
This study is a novel and unprecedented review of WM assessment of MCI patients with CANTAB. One of the challenges related to this study was that the included studies did not report the sensitivity and specificity of CANTAB for the diagnosis of WM deficits in MCI patients. This should be considered in future studies. The other related limitation was that the included studies used heterogeneous criteria for baseline diagnosis of MCI; thus, the results cannot be generalized to all of the considered populations. Nevertheless, a comprehensive review of available evidence with a systematic approach was the main strength of this study.
This study reveals that WM is an impaired cognitive domain at MCI. Based on our assessment, WM subtests of CANTAB, including SWM, SSP, and RVP, can pinpoint deficits in MCI patients, so CANTAB-based WM assessment can help the clinicians in the diagnosis of MCI. Also, WM functions of MCI patients are associated with some of the AD-associated biomarkers, such as tau-protein and Aβ. There is a need for future well-designed studies on this topic to reach a comprehensive conclusion in terms of both diagnostic accuracies of WM profiles of CANTAB battery and factors that can affect the WM in MCI patients.