ABSTRACT
Aim
Identifying high‐risk individuals for future cognitive decline is key for timely and efficient interventions against dementia. Although imaging and biofluid biomarkers have been developed to detect neuropathological changes and predict future disease progression, their high costs and low accessibility limit their widespread use. We previously developed an eye‐tracking‐based cognitive assessment (ETCA) as a novel screening tool for dementia and demonstrated its utility in detecting mild cognitive decline with high accuracy. This study aimed to investigate the ETCA's performance in predicting future cognitive decline in a community‐based longitudinal study.
Methods
Community‐dwelling older adults (n = 55, mean age: 57.8 (SD, 12.6) years) without a formal diagnosis of dementia were enrolled and underwent the ETCA and neuropsychological tests, including the Mini‐Mental State Examination (MMSE), Addenbrooke's Cognitive Examination‐III (ACE‐III), and Rivermead Behavioral Memory Test (RBMT), both at baseline and at a 2‐year follow‐up point.
Results
Approximately half (54.5%, 35/55) of participants showed a decline in RBMT scores during the 2‐year follow‐up and were defined as the memory decline group. The baseline ETCA scores were significantly lower in the memory decline group than in the memory stable group. The baseline ETCA composite subscores achieved an AUC‐ROC of 0.709 in detecting future memory decline.
Conclusions
The ETCA offers a rapid, objective, and low‐burden approach for screening for the risk of future cognitive decline, which could make a substantial contribution to the prevention of dementia.
Keywords: dementia, eye‐tracking, mild cognitive impairment, prediction, prognosis
Eye‐tracking‐based cognitive assessment (ETCA) is a novel screening tool for dementia. In this community‐based longitudinal study, we demonstrated the ETCA's performance in predicting future cognitive decline, which could make a substantial contribution to the prevention of dementia.
1. Introduction
A growing body of evidence suggests that early intervention can prevent or delay the onset of dementia [1, 2]. Managing modifiable risk factors for dementia could potentially prevent up to 40% of dementia cases [3, 4]. Initiating treatment with anti‐amyloid antibody therapy during the preclinical phase of Alzheimer's disease offers an optimal opportunity to slow the disease's progression [5]. These findings suggest that detecting mild cognitive changes at the early stage and identifying high‐risk individuals for future cognitive decline are crucial for timely interventions.
However, predicting subtle cognitive decline before the onset of noticeable symptoms remains a significant challenge. Currently, the first step toward detecting cognitive impairment involves interview‐ or paper‐based neuropsychological tests, such as the Mini‐Mental State Examination (MMSE) [6]. Although the MMSE has been widely used as a screening tool for dementia, its sensitivity is insufficient for detecting mild cognitive changes and predicting future decline [6, 7, 8, 9]. More comprehensive neuropsychological tests, such as the Addenbrooke's Cognitive Examination‐III (ACE‐III) and the Rivermead Behavioral Memory Test (RBMT), have shown better screening performance for detecting mild cognitive decline [10, 11, 12]. Nevertheless, these cognitive tests are primarily designed to evaluate current cognitive status and their performances in predicting future cognitive decline have not been fully examined. In addition, these tests require long administration times and trained examiners, which prevents their widespread use as screening tools [13, 14].
Various biomarkers for detecting early signs, or predicting future risks, of dementia have been investigated, including imaging and biofluid biomarkers [15, 16]. These biomarkers are helpful for detecting the cerebral accumulation of pathological proteins and neurodegeneration even at the preclinical stage; therefore, they enable a prediction of future disease progression. While these biomarkers offer the advantage of an accurate assessment of underlying pathological changes that eventually lead to developing dementia, their high cost and low accessibility limit their routine use, making them unsuitable for large‐scale screening. Recently, digital technologies have offered novel approaches for cognitive assessment and the prediction of a future risk of dementia [17, 18]. There has been growing progress in the development of digital biomarkers that leverage digital technologies to enable disease diagnosis and monitoring at a low cost.
We previously developed a novel digital cognitive assessment tool based on eye‐tracking technology (Eye‐Tracking‐based Cognitive Assessment, ETCA) [19, 20, 21]. The ETCA quantitatively evaluates cognitive function by tracking gaze points while subjects view task movies designed to assess major cognitive domains such as memory, judgment, attention, and visuospatial function. The ETCA's cognitive scores are determined by measuring the percentage of viewing time spent on the correct answer among multiple options presented on a screen. Notably, the ETCA can be performed in a semi‐automated manner and completed in 3 min. The ETCA scores are well correlated with traditional neuropsychological test scores and can detect dementia with high accuracy in clinical settings [20]. More recently, we have demonstrated its potential as a screening tool for detecting mild cognitive impairment (MCI) in a community‐based cohort [19]. However, its performance in predicting future cognitive decline has not yet been fully investigated.
This study aimed to investigate the ETCA's potential utility to predict the future risk of cognitive decline in a community‐based longitudinal study. We enrolled older adults without a prior diagnosis of dementia at the baseline and followed up for 2 years. Cognitive assessments were performed with the ETCA and traditional neuropsychological tests, including the MMSE, ACE‐III, and RBMT, both at the baseline and at the 2‐year follow‐up time point. As slight memory decline can be one of the earliest signs of dementia, and detecting changes in memory scores provides a beneficial window of opportunity for early intervention, we used the score changes in the RBMT, a specialized and sensitive memory test, over 2 years to define individuals at risk of future memory decline. Then, we investigated if more concise assessments like the ETCA could predict the risk of future memory decline and compared its performance with the MMSE and ACE‐III.
2. Methods
2.1. Study Design and Participants
This longitudinal study was conducted at the Osaka Psychiatric Research Center, Osaka Psychiatric Medical Center, Osaka, Japan. Participants were recruited from a community‐based volunteer population between October 1, 2020, and August 31, 2022. Individuals aged 40 years or older without formally diagnosed dementia were enrolled. The main exclusion criteria were: (1) participants with active neurological or psychiatric disorders, (2) individuals for whom gaze detection was impossible because of severe visual impairments (e.g., low visual acuity or visual field defects). All eligible participants were invited for a re‐examination after 2‐year follow‐up period.
All participants provided written informed consent during the study enrollment phase. The Osaka Psychiatric Medical Center's institutional review board approved this study. All study procedures conformed to the relevant guidelines and regulations, including the Declaration of Helsinki.
2.2. Procedures
At baseline, all participants underwent physical and neurological examinations conducted by geriatricians and psychiatrists and neuropsychological assessments (MMSE, RBMT, and ACE‐III) and ETCA conducted by neuropsychologists. The ETCA was conducted before the neuropsychological assessments. At the 2‐year follow‐up, participants undertook the same physical and neurological examinations, neuropsychological assessments, and ETCA as the baseline. These assessments were performed in the same examination rooms as those used for the baseline tests to minimize environment‐related factors' influence on the results.
2.3. Neuropsychological Tests
The MMSE scores range from 0 to 30 points; the lower the score, the greater the cognitive impairment [6]. The ACE‐III is used for assessing major cognitive domains including attention, orientation, memory, language, and visuospatial functions [10, 22]. The ACE‐III scores range from 0 to 100, with lower scores indicating greater impairment [22]. The RBMT is used for assessing everyday memory performance [23]. We used standardized profile scores (range: 0–24); the lower the score, the greater the memory impairment [23]. In this study, memory decline was defined as a decrease of at least one point in the RBMT score over the 2‐year observation period. Test administrators used the same standardized protocols for each neuropsychological test to minimize inter‐rater variability and obtain reliable scores.
2.4. The Eye‐Tracking‐Based Cognitive Assessment
Participants' gaze points were recorded using a high‐performance eye‐tracking device, as described previously [24, 25, 26]. Briefly, the device uses infrared light sources and cameras to detect the individual's eye position using a corneal‐reflection technique. The gaze points were recorded at a frequency of 50 Hz while the task movies were presented on the monitor.
The task movies were presented on the screen in the following order (running for about 3 min in total): task 1‐a, memory encoding; task 2, deductive reasoning (odd‐one‐out task); task 3, visual working memory (pattern matching); tasks 4–5, calculation; task 6, visuospatial function; task 7, visual working memory task (intersecting double pentagon); and task 1‐b, memory recall. Each task movie included instructions or problem text and multiple choices that appeared simultaneously on the screen. One of the choices was the correct answer. A region of interest (ROI) was set on the correct answer, and the percentage of the participant's viewing time (% of total time) spent in the ROI was recorded as their score. These tasks were categorized into four subdomains, including delayed recall (task 1‐b), judgment (task 2), working memory (tasks 3–5 and 7), and visuospatial function (task 6). The mean of scores from tasks 3–5 and 7 was used as a working memory score. The mean of the four subscores was used as the total ETCA score [19, 20].
2.5. Statistical Analysis
We used the Wilcoxon signed‐rank test to compare participants' scores on the MMSE, ACE‐III, RBMT, and ETCA between baseline and follow‐up. The Mann–Whitney U test was applied to compare the RBMT scores at follow‐up between participants classified as RBMT decliners and non‐decliners, and to compare MMSE, ACE‐III, total ETCA scores, and ETCA subscores at baseline depending on whether RBMT scores had declined. The performances of the ETCA composite score, total ETCA score, MMSE, and ACE‐III for predicting future memory decline were determined via ROC analyses. The area under the ROC curve (AUC) was used as an index of performance for discriminating whether RBMT scores had declined. The 95% confidence interval (CI) for the AUC was determined using the bootstrap method. All statistical analyses were performed using JMP Student Edition 18.2.1 (SAS Institute Inc., Cary, NC, USA) and GraphPad Prism 5 (GraphPad Software Inc., San Diego, CA), and p < 0.05 was considered statistically significant.
3. Results
3.1. Participants' Characteristics
This study recruited 83 participants without a prior diagnosis of dementia who completed a baseline evaluation. Among them, 55 (66.3%) were followed up and completed the 2‐year follow‐up assessment (Figure 1). Demographic characteristics of the 55 participants are presented in Table 1. The mean age at baseline was 57.8 (SD, 12.6) years (range 40–84), and 34 (61.8%) were female. The mean scores for each neuropsychological test at the baseline were 28.8 (SD, 1.6) for the MMSE, 92.4 (SD, 6.8) for the ACE‐III, and 20.3 (SD, 3.8) for the RBMT. Most participants were cognitively normal or had slight cognitive decline (Figure S1). At the 2‐year follow‐up, neither the MMSE (mean 28.9 (SD, 1.4)) nor the ACE‐III (mean 92.9 (SD, 6.8)) scores had changed significantly compared to the baseline levels (Table 1). In contrast, the RBMT scores showed a significant decline at the 2‐year follow‐up (mean 19.0 (SD, 4.6)). The baseline score for the ETCA was 76.3 (SD, 11.2) and did not change significantly at the 2‐year follow‐up.
FIGURE 1.
Flow diagram of participant selection and follow‐up, and schematic overview of the study. Participants' cognitive functions were assessed using the ETCA and neuropsychological tests on the same day. ACE‐III, Addenbrooke's cognitive examination III; ETCA, eye‐tracking‐based cognitive assessment; MMSE, Mini‐Mental State Examination; RBMT, Rivermead Behavioral Memory Test.
TABLE 1.
Participant characteristics.
| All participants (n = 55) | |
|---|---|
| Age (years) | 57.8 ± 12.6 |
| Female | 34 (61.8) |
| Neuropsychological test scores | |
| MMSE | |
| Baseline | 28.8 ± 1.6 |
| Follow‐up | 28.9 ± 1.4 |
| ACE‐III | |
| Baseline | 92.4 ± 6.8 |
| Follow‐up | 92.9 ± 6.8 |
| RBMT | |
| Baseline | 20.3 ± 3.8 |
| Follow‐up | 19.0 ± 4.6 a |
| ETCA | |
| Baseline | 76.3 ± 11.2 |
| Follow‐up | 76.0 ± 14.9 |
Note: Values are presented as mean ± standard deviation for age and neuropsychological scores or as number (%) for gender.
Abbreviations: ACE‐III, Addenbrooke's cognitive examination III; ETCA, eye‐tracking‐based cognitive assessment; MMSE, Mini‐Mental State Examination; RBMT, Rivermead Behavioral Memory Test.
Significant difference between baseline and follow‐up (p < 0.05) based on the Wilcoxon signed‐rank test.
3.2. Changes in Memory Performance Assessed by the RBMT Over 2 Years of Follow‐Up
As we observed a significant decline in memory performance assessed by the RBMT, we divided the participants into two groups, “RBMT decliners,” who showed a decline in the RBMT score of at least one point, and “RBMT non‐decliners,” whose test scores were stable or increased over the follow‐up period. Figure 2 shows the changes in each participant's RBMT scores. We observed that 54.5% (30/55) of participants showed a decline in RBMT scores and were classified as decliners. Mean age did not differ significantly between the decliner (60.0 (SD 13.0) years) and non‐decliner groups (55.2 (SD 11.7) years) (p = 0.15, Welch's t‐test). The difference in the RBMT scores between the decliners and non‐decliners was statistically significant (decliners, 17.0 (SD, 5.2); non‐decliners, 21.4 (SD, 2.1), p < 0.001) at 2‐year follow‐up.
FIGURE 2.
Changes in memory performance assessed by the RBMT over the 2‐year follow‐up. Each dot and dashed line (red, decliners; blue, non‐decliners) represents an individual participant's score changes in the RBMT between the baseline and 2‐year follow‐up (n = 55). Solid lines represent mean changes in RBMT score of decliners (red) and non‐decliners (blue). Error bars represent standard errors. Mann–Whitney U test. ****p < 0.001. RBMT, Rivermead Behavioral Memory Test.
3.3. Baseline Scores From Neuropsychological Tests and the ETCA in the Memory Decline Group
Aiming to evaluate an association between the baseline test scores and future memory decline, we next compared the baseline scores of the MMSE, ACE‐III, and ETCA between the RBMT decliners (memory decline group) and non‐decliners (memory stable group) (Figure 3). There were no differences in the baseline MMSE (Figure 3A) or ACE‐III (Figure 3B) scores between the two groups (RBMT non‐decliners, mean 28.9 (SD, 1.5) vs. RBMT decliners, mean 28.7 (SD, 1.8) for the MMSE, and RBMT non‐decliners, mean 92.9 (SD, 6.6) vs. RBMT decliners, mean 92.0 (SD, 7.1) for the ACE‐III). In contrast, the memory decline group had lower ETCA scores at the baseline, compared with the memory stable group (RBMT non‐decliners, mean 79.0 (SD, 2.2) vs. RBMT decliners, mean 74.1 (SD, 2.2), p = 0.05) (Figure 3C). These results suggest that low ETCA scores may be associated with future memory decline.
FIGURE 3.
Baseline scores on the MMSE, ACE‐III, and ETCA in memory stable and decline groups. Participants were divided into two groups (memory stable and decline) based on changes in their RBMT scores over the 2‐year follow‐up. Baseline total scores for the MMSE, ACE‐III, and ETCA, and ETCA subscores from each cognitive domain (delayed recall, working memory, judgment, and visuospatial function) were compared between the memory stable (blue, RBMT non‐decliners, n = 25) and decline (red, RBMT decliners, n = 30) groups. Error bars represent standard errors. Mann–Whitney U test. *p < 0.05. ACE‐III, Addenbrooke's cognitive examination III; ETCA, eye‐tracking‐based cognitive assessment; N.S., not significant; RBMT, Rivermead Behavioral Memory Test.
We then looked at the ETCA's subscores at the baseline and compared them between the memory stable and decline groups (Figure 3D). The memory decline group had significantly lower scores than the memory stable group in working memory (RBMT non‐decliners, mean 73.7 (SD, 3.0) vs. RBMT decliners, mean 69.0 (SD, 2.7), p < 0.05) and visuospatial function (RBMT non‐decliners, mean 71.6 (SD, 9.1) vs. RBMT decliners, mean 61.3 (SD, 3.1), p < 0.05). This suggests that these subscores can be a good indicator for predicting future memory decline.
3.4. Performance of the ETCA Scores in Predicting Future Memory Decline
Given that ETCA scores, especially the working memory and visuospatial function subscores, were associated with future memory decline (Figure 3), we next evaluated the performance of the ETCA for future memory decline (Figure 4). We used the ETCA's total and composite subscores (the means of the working memory and visuospatial function subscores) to distinguish memory‐stable (RBMT non‐decliners) and decline (RBMT decliners) groups, comparing them with MMSE and ACE‐III scores. The ETCA composite subscores demonstrated the highest performance for predicting memory decline and achieved an AUC‐ROC of 0.709 (95% CI: 0.55–0.83), followed by the total ETCA score, which achieved an AUC of 0.654 (95% CI: 0.49–0.79), outperforming the MMSE (AUC = 0.522, 95% CI: 0.38–0.66) and the ACE‐III (AUC = 0.494, 95% CI: 0.34–0.64). Statistical comparisons of the AUC values using DeLong's test with Bonferroni correction are provided in Table S1.
FIGURE 4.
Performance of the ETCA for predicting future memory decline. ROC curve analyses showing the performance of the ETCA composite subscores [working memory + visuospatial function] (orange line), total ETCA scores (dashed orange line), MMSE scores (dashed black line), and ACE‐III scores (dashed gray line) for distinguishing the memory stable (RBMT non‐decliners, n = 25) and decline (RBMT decliners, n = 30) groups. The area under the ROC curve (AUC) was used to compare the performance of each test. ACE‐III, Addenbrooke's cognitive examination III; ETCA, eye‐tracking‐based cognitive assessment; MMSE, Mini‐Mental State Examination; RBMT, Rivermead Behavioral Memory Test; ROC, receiver operating characteristic.
4. Discussion
We previously reported the utility of the ETCA as a screening tool for detecting dementia and MCI [19, 20, 21]; however, the ETCA's performance for predicting future memory decline has not been investigated. In this community‐based longitudinal study, we observed memory decline in approximately half (54.5%) of participants after a 2‐year follow‐up period, when evaluated by highly sensitive memory assessment tools like the RBMT. ETCA scores at the baseline were associated with a future memory decline (Figure 3) and showed a good AUC‐ROC value for distinguishing memory‐stable and decline groups (Figure 4). These results highlight the ETCA's potential as a screening tool for predicting a risk of future cognitive decline.
This study enrolled participants who had not been diagnosed as having dementia and whose activities of daily living (ADL) were not significantly impaired. Most participants were regarded as cognitively unimpaired, since they scored higher than the traditional cutoff values of the neuropsychological tests for diagnosing dementia and MCI (Table 1 and Figure S1). Among these relatively healthy participants, RBMT scores declined significantly during the 2‐year follow‐up period, while the MMSE and ACE‐III scores were stable. Memory decline can be the earliest manifestation of dementia, and detecting slight memory impairment is crucial for predicting the risk of developing dementia. Our observations indicate that specialized tests such as the RBMT are necessary to detect slight memory decline, while global cognitive tests such as the MMSE and ACE‐III are not sensitive enough to detect them (Table 1). However, specialized memory tests such as the RBMT require a long administration time and high proficiency of the administrators, which prevents their widespread use as screening tools.
In this study, the memory decline group was defined as at least a one‐point decrease in the RBMT score over the 2‐year follow‐up. The rationale for this strict cutoff is as follows. The mean RBMT scores at 2‐year follow‐up (19.0 (SD, 4.6)) showed a significant decrease, by 1.3 points, compared to the baseline level (20.3 (SD, 3.8)) (Table 1). In contrast, the mean scores of the MMSE and ACE‐III at 2‐year follow‐up (MMSE, 28.9 (SD, 1.4); ACE‐III (92.9 (SD, 6.8))) were almost unchanged, or even slightly higher, compared to the baseline levels (MMSE, 28.8 (SD, 1.6); ACE‐III (92.4 (SD, 6.8))). Based on this observation, we assumed that even a one‐point reduction in the RBMT score could have clinical significance, so we used it as a cutoff for defining the memory decline group in this study. Further, we strictly standardized the protocols for the neuropsychological tests, and all assessments were performed under the same conditions, including the examination room, time, and administrators assigned, in order to minimize environment‐related factors and to obtain reliable scores. Nevertheless, it should be noted that such a sensitive threshold may capture measurement errors or physiological fluctuations, rather than true pathological decline. Correlation analyses of the scores between each RBMT subcategory and the ETCA in larger sample size cohort should be performed in a future study.
The baseline scores of the MMSE and ACE‐III did not differ between the memory‐stable and decline groups (Figure 3), meaning that those scores are unable to predict a future memory decline. In contrast, the baseline scores of the ETCA, especially subscores in working memory and visuospatial function tasks, were significantly lower in the memory decline group. The composite subscores consisting of the two ETCA subscores (working memory and visuospatial function) achieved an AUC‐ROC of 0.709 in detecting the memory decline group (Figure 4), suggesting the utility of the ETCA for predicting the risk of future cognitive decline. The performances of each ETCA subscore were AUC‐ROC of 0.695 (95% CI: 0.55–0.84) for working memory and 0.688 (95% CI: 0.54–0.84) for visuospatial function, respectively. Unexpectedly, the ETCA's delayed recall subscores did not differ between the memory‐stable and decline groups (Figure 3D). This might be attributed to the fact that only a single task was assigned for delayed recall in the ETCA and the difficulty level was not high enough to detect a potential risk of memory decline. Indeed, the mean scores of the delayed recall were the highest among the ETCA subscores, and most participants obtained full scores, creating a ceiling effect in this subscore (Figure 3D).
One potential explanation for the superior sensitivity of the ETCA in predicting future cognitive decline, compared to the MMSE and ACE‐III, is its robustness against ceiling effects, which preserves the variance necessary for identifying subtle or potential cognitive declines. In the present cohort, which primarily consists of cognitively unimpaired individuals, the baseline mean scores for the MMSE and ACE‐III were 28.8/30 (SD 1.6) and 92.4/100 (SD 6.8), respectively (Table 1). These scores, which approach the maximum possible values, indicate a pronounced ceiling effect that limits the capacity for effective risk stratification. In contrast, the baseline mean ETCA score was 76.3/100 (SD 11.2) (Table 1), demonstrating greater score variability and a minimal ceiling effect. This characteristic facilitates a more granular stratification, even within a cognitively unimpaired population, where conventional scales such as the MMSE and ACE‐III often fail to differentiate subtle differences. This advantage likely stems from the ETCA's scoring methodology, which uses the percentage of the viewing time (% to total time) spent in the correct ROI within a specified timeframe. Because this continuous metric is less prone to ceiling effects and can capture intermediate performance levels across tasks, it may more accurately reflect the subtle variations in the participants' underlying cognitive status.
A range of biomarkers, including neuroimaging and plasma‐ and cerebrospinal fluid (CSF)‐based proteomics, has been explored as predictive tools for assessing the future risk of cognitive decline [27, 28, 29]. Lower hippocampal and greater white matter hyperintensity volume are associated with an increased risk of developing MCI in the future [16, 30]. Plasma levels of pathological proteins such as amyloid‐β (Aβ) and tau have been associated with the increased risk of MCI [31, 32]. CSF biomarkers can predict rapid cognitive decline [33] and dementia onset [34]. However, the cost‐effectiveness and accessibility of such imaging and biofluid biomarkers are not high enough, which has been a bottleneck for their widespread use aiming for dementia prevention. The ETCA offers a low‐burden, cost‐effective, and easily implementable approach for evaluating the risk of cognitive decline. Simple assessment with the ETCA could make it possible to screen for the risk of cognitive decline on a large scale even among individuals who are still cognitively unimpaired, allowing for earlier interventions targeting modifiable lifestyle factors [3]. Further, this does not mean that the ETCA can replace established biomarkers like imaging and biofluid biomarkers. A future study should examine the relationship between the ETCA and the established biomarkers to further validate the ETCA's potential clinical utility.
This study has potential limitations. First, we did not consider a clinical diagnosis of MCI or dementia based on the diagnostic criteria at the baseline because most participants visited the facility on their own, and we could not obtain detailed information necessary for clinical diagnosis. Although most participants scored higher than the traditional cutoff values in three neuropsychological tests (MMSE, ACE‐III, and RBMT) for MCI and dementia, a small portion of the participants might have had MCI or dementia at the baseline. Second, along the same line, we did not determine whether participants who exhibited memory decline at the 2‐year follow‐up eventually progressed to MCI or dementia. This study was a 2‐year longitudinal design, and a longer follow‐up period is necessary to clarify the relationship between the baseline ETCA performance and the future development of clinically defined MCI or dementia. Third, the study's relatively small sample size (n = 55), and especially the number of RBMT decliners (memory decline group, n = 30), restricts the generalizability of the conclusion. Although no statistically significant difference in age was observed between the RBMT decliner and non‐decliner groups, the limited sample size precluded the use of multivariate analysis with age as a covariate. Consequently, further validation is required to definitively establish the ETCA score as a strictly independent predictor of future cognitive decline. A future study with a larger sample size would further validate the reliability of the results and applicability for large‐scale screening.
Overall, we evaluated the ETCA's performance to predict future cognitive decline among community‐dwelling individuals without a prior diagnosis of dementia. The baseline scores of the ETCA were associated with future memory decline. The ETCA offers a rapid, objective, and low‐burden approach for screening for the risk of future cognitive decline, which could make a substantial contribution to the prevention of dementia.
Author Contributions
M.K., A.O., and S.T. (Shuko Takeda) designed the work, acquired the data, performed data analyses, and wrote the manuscript. S.T. (Shin Teshirogi) and S.Y. performed data analysis and wrote the manuscript. Y.I. (Yuki Ito), Y.I. (Yuya Ikegawa), M.O., and T.N. acquired the data and performed data analyses. Y.N., H.B., S.T. (Sayaka Tanaka), M.H., and K.I. acquired the data. All authors have approved the final version of the article.
Funding
This work was supported by JSPS KAKENHI Grant Number 22K19500 [Grant‐in‐Aid for Challenging Exploratory Research] (S.T.), 22K15701 [Grant‐in‐Aid for Early‐Career Scientists] (A.O.), 25K19008 [Grant‐in‐Aid for Early‐Career Scientists] (A.O.), and 25K24185 [Grant‐in‐Aid for Research Activity Start‐up] (M.K.), and research grants from The Osaka Medical Research Foundation for Intractable Diseases [23‐2‐1] (S.T.). The funders had no role in the design, data collection, data analysis, or reporting of this study.
Disclosure
The authors have nothing to report.
Ethics Statement
The Osaka Psychiatric Medical Center institutional review boards approved this study (approval number 2020‐10). All study procedures conformed to the relevant guidelines and regulations, including the Declaration of Helsinki.
Consent
All participants provided written informed consent during study enrollment.
Supporting information
Figure S1: Distributions of the MMSE, ACE‐III, and RBMT scores at baseline.
Table S1: Statistical comparison of AUC values among the ETCA, MMSE, and ACE‐III.
Acknowledgments
We would like to thank all participants of the study.
Contributor Information
Akane Oyama, Email: oyamaak@mh-opho.jp.
Shuko Takeda, Email: takedashuk@mh-opho.jp, Email: takeda@cgt.med.osaka-u.ac.jp.
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.
References
- 1. Anderson N. D., “State of the Science on Mild Cognitive Impairment (MCI),” CNS Spectrums 24 (2019): 78–87. [DOI] [PubMed] [Google Scholar]
- 2. Morris J. C., “Mild Cognitive Impairment and Preclinical Alzheimer's Disease,” Geriatrics, Suppl 6 (2005): 9–14. [PubMed] [Google Scholar]
- 3. Livingston G., Huntley J., Liu K. Y., et al., “Dementia Prevention, Intervention, and Care: 2024 Report of the Lancet Standing Commission,” Lancet 404 (2024): 572–628. [DOI] [PubMed] [Google Scholar]
- 4. Livingston G., Huntley J., Sommerlad A., et al., “Dementia Prevention, Intervention, and Care: 2020 Report of the Lancet Commission,” Lancet 396 (2020): 413–446. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Rafii M. S. and Aisen P. S., “Detection and Treatment of Alzheimer's Disease in Its Preclinical Stage,” Nature Aging 3 (2023): 520–531. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Folstein M. F., Folstein S. E., and McHugh P. R., ““Mini‐Mental State”. A Practical Method for Grading the Cognitive State of Patients for the Clinician,” Journal of Psychiatric Research 12 (1975): 189–198. [DOI] [PubMed] [Google Scholar]
- 7. Jia X., Wang Z., Huang F., et al., “A Comparison of the Mini‐Mental State Examination (MMSE) With the Montreal Cognitive Assessment (MoCA) for Mild Cognitive Impairment Screening in Chinese Middle‐Aged and Older Population: A Cross‐Sectional Study,” BMC Psychiatry 21 (2021): 485. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Hoops S., Nazem S., Siderowf A. D., et al., “Validity of the MoCA and MMSE in the Detection of MCI and Dementia in Parkinson Disease,” Neurology 73 (2009): 1738–1745. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Diniz B. S. O., Yassuda M. S., Nunes P. V., Radanovic M., and Forlenza O. V., “Mini‐Mental State Examination Performance in Mild Cognitive Impairment Subtypes,” International Psychogeriatrics 19 (2007): 647–656. [DOI] [PubMed] [Google Scholar]
- 10. Takenoshita S., Terada S., Yoshida H., et al., “Validation of Addenbrooke's Cognitive Examination III for Detecting Mild Cognitive Impairment and Dementia in Japan,” BMC Geriatrics 19 (2019): 123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Johansson M. and Wressle E., “Validation of the Neurobehavioral Cognitive Status Examination and the Rivermead Behavioural Memory Test in Investigations of Dementia,” Scandinavian Journal of Occupational Therapy 19 (2012): 282–287. [DOI] [PubMed] [Google Scholar]
- 12. Mitchell A. J., “A Meta‐Analysis of the Accuracy of the Mini‐Mental State Examination in the Detection of Dementia and Mild Cognitive Impairment,” Journal of Psychiatric Research 43 (2009): 411–431. [DOI] [PubMed] [Google Scholar]
- 13. Chan J. Y. C., Yau S. T. Y., Kwok T. C. Y., and Tsoi K. K. F., “Diagnostic Performance of Digital Cognitive Tests for the Identification of MCI and Dementia: A Systematic Review,” Ageing Research Reviews 72 (2021): 101506. [DOI] [PubMed] [Google Scholar]
- 14. Rasmussen L. S., Larsen K., Houx P., et al., “The Assessment of Postoperative Cognitive Function: Assessment of Postoperative Cognition,” Acta Anaesthesiologica Scandinavica 45 (2001): 275–289. [DOI] [PubMed] [Google Scholar]
- 15. Yakoub Y., Gonzalez‐Ortiz F., Ashton N. J., et al., “Plasma p‐tau217 Identifies Cognitively Normal Older Adults Who Will Develop Cognitive Impairment in a 10‐Year Window,” Alzheimer's & Dementia 21 (2025): e14537. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Prosser L., Macdougall A., Sudre C. H., et al., “Predicting Cognitive Decline in Older Adults Using Baseline Metrics of AD Pathologies, Cerebrovascular Disease, and Neurodegeneration,” Neurology 100 (2023): e834–e845. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Chen C., Ding S., and Wang J., “Digital Health for Aging Populations,” Nature Medicine 29 (2023): 1623–1630. [DOI] [PubMed] [Google Scholar]
- 18. Kourtis L. C., Regele O. B., Wright J. M., and Jones G. B., “Digital Biomarkers for Alzheimer's Disease: The Mobile/ Wearable Devices Opportunity,” npj Digital Medicine 2 (2019): 2, 10.1038/s41746-019-0084-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Katsuhisa M., Oyama A., Ito Y., et al., “Eye‐Tracking‐Based Cognitive Assessment Efficiently Detects Mild Cognitive Decline in the Predementia Stage,” Dementia and Geriatric Cognitive Disorders 1–22 (2024): 1–11. [DOI] [PubMed] [Google Scholar]
- 20. Oyama A., Takeda S., Ito Y., et al., “Novel Method for Rapid Assessment of Cognitive Impairment Using High‐Performance Eye‐Tracking Technology,” Scientific Reports 9 (2019): 12932. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Takami Y., Miki T., Takeya Y., Rakugi H., Morishita R., and Takeda S., “Eye‐Tracking‐Based Tablet Application for Enhanced Cognitive Assessment: A Clinical Trial for Software as a Medical Device (SaMD),” GeroScience (2025): 1–13, 10.1007/s11357-025-01941-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Bruno D. and Schurmann Vignaga S., “Addenbrooke's Cognitive Examination III in the Diagnosis of Dementia: A Critical Review,” Neuropsychiatric Disease and Treatment 15 (2019): 441–447. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Wilson B., Cockburn J., Baddeley A., and Hiorns R., “The Development and Validation of a Test Battery for Detecting and Monitoring Everyday Memory Problems,” Journal of Clinical and Experimental Neuropsychology 11 (1989): 855–870. [DOI] [PubMed] [Google Scholar]
- 24. Yamasue H., Okada T., Munesue T., et al., “Effect of Intranasal Oxytocin on the Core Social Symptoms of Autism Spectrum Disorder: A Randomized Clinical Trial,” Molecular Psychiatry 25 (2020): 1849–1858. [DOI] [PubMed] [Google Scholar]
- 25. Nishizato M., Fujisawa T. X., Kosaka H., and Tomoda A., “Developmental Changes in Social Attention and Oxytocin Levels in Infants and Children,” Scientific Reports 7 (2017): 2540. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Fujioka T., Inohara K., Okamoto Y., et al., “Gazefinder as a Clinical Supplementary Tool for Discriminating Between Autism Spectrum Disorder and Typical Development in Male Adolescents and Adults,” Molecular Autism 7 (2016): 19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Yarbro J. M., Shrestha H. K., Wang Z., et al., “Proteomic Landscape of Alzheimer's Disease: Emerging Technologies, Advances and Insights (2021–2025),” Molecular Neurodegeneration 20 (2025): 83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Yoong S. Q., Lu J., Xing H., Gyanwali B., Tan Y. Q., and Wu X. V., “The Prognostic Utility of CSF Neurogranin in Predicting Future Cognitive Decline in the Alzheimer's Disease Continuum: A Systematic Review and Meta‐Analysis With Narrative Synthesis,” Ageing Research Reviews 72 (2021): 101491. [DOI] [PubMed] [Google Scholar]
- 29. Ewers M., Sperling R. A., Klunk W. E., Weiner M. W., and Hampel H., “Neuroimaging Markers for the Prediction and Early Diagnosis of Alzheimer's Disease Dementia,” Trends in Neurosciences 34 (2011): 430–442. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Woodward M., Bennett D. A., Rundek T., Perry G., and Rudka T., “The Relationship Between Hippocampal Changes in Healthy Aging and Alzheimer's Disease: A Systematic Literature Review,” Frontiers in Aging Neuroscience 16 (2024): 1390574. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Ossenkoppele R., Pichet Binette A., Groot C., et al., “Amyloid and Tau PET‐Positive Cognitively Unimpaired Individuals Are at High Risk for Future Cognitive Decline,” Nature Medicine 28 (2022): 2381–2387. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Strikwerda‐Brown C., Hobbs D. A., Gonneaud J., et al., “Association of Elevated Amyloid and Tau Positron Emission Tomography Signal With Near‐Term Development of Alzheimer Disease Symptoms in Older Adults Without Cognitive Impairment,” JAMA Neurology 79 (2022): 975–985. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Degerman Gunnarsson M., Lannfelt L., Ingelsson M., Basun H., and Kilander L., “High Tau Levels in Cerebrospinal Fluid Predict Rapid Decline and Increased Dementia Mortality in Alzheimer's Disease,” Dementia and Geriatric Cognitive Disorders 37 (2014): 196–206. [DOI] [PubMed] [Google Scholar]
- 34. Oh H. S.‐H., Urey D. Y., Karlsson L., et al., “A Cerebrospinal Fluid Synaptic Protein Biomarker for Prediction of Cognitive Resilience Versus Decline in Alzheimer's Disease,” Nature Medicine 31 (2025): 1592–1603. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Figure S1: Distributions of the MMSE, ACE‐III, and RBMT scores at baseline.
Table S1: Statistical comparison of AUC values among the ETCA, MMSE, and ACE‐III.
Data Availability Statement
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.