Alzheimer’s Disease Revealed By Reliable Digital Memory Assessments
Alzheimer’s disease (AD) research has advanced in detecting pathological changes through biomarkers, but cognitive measures lag in capturing early, subtle impairments. Traditional assessments designed for overt deficits are now outpaced by discoveries in episodic memory networks within the medial temporal lobe, which are crucial in AD progression. With limitations in conventional neuropsychological tools, especially for decentralized clinical trials, remote digital cognitive assessments via smartphones and tablets are emerging as promising alternatives. This study explores the feasibility and validity of these digital assessments, aiming to improve early detection and monitoring of AD.
THE STUDY BACKGROUND
Alzheimer’s disease (AD) is a progressive neurodegenerative disorder marked by the gradual decline of cognitive functions, particularly memory. Despite significant advances in fluid and neuroimaging biomarkers for detecting pathological changes associated with AD, cognitive measures used in healthcare and clinical trials have not kept pace. Traditional cognitive assessments were primarily designed to detect overt cognitive impairments, making them less effective for identifying the early, subtle cognitive changes that occur in the preclinical and early stages of AD [1,2]. This gap is especially problematic given recent research on episodic memory networks within the medial temporal lobe (MTL) and neocortex, which are crucial in the progression of AD [3,4].
Episodic memory relies on complex pattern separation and completion processes, primarily mediated by specific subregions within the MTL. The dentate gyrus plays a crucial role in pattern separation, reducing memory interference between similar events. At the same time, the hippocampal Cornu Ammonis 3 (CA3) is involved in pattern completion, working in conjunction with neocortical regions [3,5]. Additionally, distinct pathways within the hippocampal-entorhinal circuitry process objects and spatial information, further emphasizing the complexity of memory functions that are disrupted in AD [4]. Evidence suggests that both short-term and long-term memory impairments occur during the predementia stages of AD, yet traditional neuropsychological assessments often fail to capture these subtle deficits [6].
These limitations include high participant burden and the impracticality of frequent test repetitions or long-term delay formats, essential for assessing memory function over time [7]. Such constraints are particularly challenging in clinical trials targeting preclinical and early-stage AD populations, where decentralized trial structures are increasingly adopted for case-finding and monitoring [8]. For example, the TRAILBLAZER-ALZ 3 trial underscores the need for practical and scalable assessment tools that can be implemented remotely [9].
In response to these challenges, remote and unsupervised digital cognitive assessments via smartphones and tablets have emerged as a promising solution. Studies, including those from our group, have demonstrated the feasibility of these assessments in both healthy older adults and individuals at risk for AD [10]. These digital tools reduce participant burden and enable more frequent and accessible monitoring, which is crucial for identifying mild cognitive impairment (MCI) and even cognitively unimpaired individuals who are β-amyloid (Aβ) positive. The present study aims to evaluate these digital memory assessments’ feasibility, usability, and construct validity. It examines their relationship with traditional neuropsychological assessments and AD biomarkers, offering a potential avenue for improved early detection and monitoring in clinical trials and healthcare settings.
THE STUDY METHOD
This research methodology involved two smartphone-based memory tasks: the Mnemonic Discrimination Test for Objects and Scenes (MDT-OS) and the Objects-In-Room Recall (ORR) test. The MDT-OS assessed participants’ ability to distinguish between similar memories by focusing on subregions in the medial temporal lobe (MTL), which is critical for this function. Participants were evaluated based on hit rates, false alarm rates, and corrected hit rates, with the primary outcome being the average corrected hit rate across both object and scene conditions. Additionally, object- and scene-specific corrected hit rates were analyzed to understand memory performance further.
The ORR test required participants to memorize the spatial arrangement of objects within a room and later recall the correct object in both immediate and delayed retrieval phases. This task was designed to tax the hippocampal pattern completion process, a key memory mechanism that helps restore full memories from partial cues. The ORR test provided scores for immediate recall (ORR-IR) and delayed recall (ORR-DR), with the study focusing on the ORR-DR score as the primary measure of episodic recall accuracy.
Quality control procedures were implemented to ensure the reliability of the data. The study included participants from the Swedish BioFINDER-2 study, recruited between February 2019 and February 2022. Only participants who completed at least one session of both tasks were included in the analysis. Sessions with excessive missing responses or exceeding the maximum delay period were excluded. The final dataset consisted predominantly of participants’ first or second test sessions.
In addition to the digital memory assessments, the study also included cerebrospinal fluid (CSF) and plasma analysis and imaging techniques such as MRI and PET scans to assess participants’ amyloid and tau status. Traditional cognitive measures were also employed, including the Alzheimer’s Disease Assessment Scale (ADAS), Mini-Mental State Examination (MMSE), Symbol Digit Modalities Test (SDMT), and verbal fluency tests. These assessments were used to calculate a composite score similar to the Preclinical Alzheimer Cognitive Composite (PACC), which was then compared with the digital memory task results to evaluate the construct validity of the remote assessments.
ANALYSIS
The research analysis employed multiple regression models to explore the relationships between positron emission tomography (PET) imaging measures, medial temporal lobe (MTL) subregional atrophy, and performance on unsupervised digital remote memory tests. These models were adjusted for age, sex, and intracranial volume, with corrections for multiple comparisons to ensure result accuracy. Robust regression techniques were used to enhance the reliability of the findings. The study also analyzed participant-specific slopes from the modified Preclinical Alzheimer Cognitive Composite (mPACC) over time. It examined their relationship with plasma phosphorylated tau (p-tau217) levels and remote memory performance while accounting for demographic factors.
Also, the study calculated intraclass correlation coefficients (ICCs) for individual and averaged test sessions to assess the reliability of the digital memory assessments. These analyses provided insight into the test-retest reliability of the evaluations over short-term (7 weeks) and medium-term (13 weeks) intervals, with ICC values indicating varying levels of reliability, from poor to excellent. This comprehensive approach to statistical analysis ensured robust findings, contributing valuable insights into using digital assessments in Alzheimer’s disease research.
RESULTS
Participant Sample:
- Out of 160 recruited participants, 100 completed at least one valid test session of the Mnemonic Discrimination Task for Objects and Scenes (MDT-OS), and 66 completed both MDT-OS and the Object-In-Room Recall Task (ORR).
- Participants were categorized into cognitively unimpaired (CU) and mild cognitive impairment (MCI) groups, each with positive or negative amyloid beta (Aβ) status.
Acceptability of Smartphone-based Assessments:
- 88% of participants owned a smartphone, and 64% could download apps without assistance.
- Of the 120 participants surveyed, most reported positive mobile app experiences, citing ease of use and clear instructions.
- 76% preferred smartphone-based tests over paper-and-pencil tasks, and 86% rated their experience highly.
Relationship Between Unsupervised Mobile Assessments and In-clinic Cognitive Scores:
- Performance in unsupervised memory assessments was significantly associated with traditional in-clinic cognitive measures, indicating that smartphone-based assessments could reflect in-clinic cognitive function.
- Age, sex, and device type were key predictors of performance, with older age, female sex, and tablet use associated with lower performance.
- Memory task performance on mobile devices showed strong correlations with in-clinic measures such as the Alzheimer’s Disease Assessment Scale (ADAS), the Symbol Digit Modalities Test (SDMT), and the Mini-Mental State Examination (MMSE).
Association with Alzheimer’s Disease Pathology:
- Remote memory task performance was linked to Alzheimer’s Disease-related biomarkers, particularly tau protein accumulation and medial temporal lobe (MTL) atrophy.
- Strong correlations were observed between digital memory task scores and tau-PET scans in the hippocampus and area 35, regions linked to memory deficits in Alzheimer’s Disease.
- These associations persisted even after accounting for amyloid-beta deposition in the brain.
Prediction of Future Cognitive Decline:
- Remote cognitive assessments and blood-based biomarkers, such as plasma p-tau217, predicted future cognitive decline.
- A model combining digital cognitive markers and plasma p-tau217 outperformed individual markers in predicting cognitive decline over five years, mainly when using MDT-S and ORR-DR task scores.
This study demonstrates that smartphone-based cognitive assessments are well-accepted by older adults and reflect in-clinic cognitive measures. These digital assessments also show strong associations with Alzheimer’s Disease biomarkers and can be used with blood-based markers to predict future cognitive decline.
DISCUSSION ON THE DIGITAL MEMORY TESTS STUDY
The study demonstrates that remote and unsupervised digital memory assessments have better construct validity than traditional in-clinic neuropsychological assessments. These digital assessments, including the Mnemonic Discrimination Task for Scenes (MDT-S) and the Object-In-Room recall-delayed recall (ORR-DR), were shown to be strongly associated with established cognitive markers like the modified Preclinical Alzheimer’s Cognitive Composite (ACC)—the most robust model predicting cognitive decline combined plasma p-tau217 levels with these digital memory assessments. Retest reliability was moderate-to-good, indicating that these tests offer cognitively reliable evaluations over time, even in unsupervised, remote settings [1].
Additionally, these digital assessments demonstrated sensitivity to measures of Alzheimer’s disease (AD) pathology, particularly tau-PET imaging, and medial temporal lobe (MTL) atrophy. The relationship between digital assessments and AD-related biomarkers supports their utility in tracking cognitive decline linked to neurodegenerative changes. However, associations with amyloid burden weakened some of these relationships, reflecting the complexity of the interaction between amyloid and tau in AD progression. Despite this, the strong association between remote assessments and MTL atrophy further underscores the role of these tasks in identifying early cognitive impairment [1].
One of the most important findings was the positive reception and high usability ratings from participants, especially older adults. Although stereotypes exist about older adults’ unfamiliarity with technology, most participants completed the remote tasks independently or with minimal assistance. The majority rated the memory tasks as challenging but straightforward and accessible, with more than 85% giving the app a score of 7 or higher out of 10. However, older participants or those with mild cognitive impairment (MCI) were less likely to participate or complete the assessments, suggesting that targeted support may enhance participation in this group [1].
These findings suggest that remote digital assessments could be valuable for case finding, cognitive monitoring, and prognosis in AD. The combination of plasma biomarkers and remote cognitive assessments shows promise for identifying individuals at risk of cognitive decline and monitoring disease progression. The moderate-to-good retest reliability and the absence of significant practice effects suggest that these assessments can reliably capture cognitive changes over time, making them useful for research and clinical applications in Alzheimer’s disease [1].
STUDY LIMITATIONS
- Sample Size
- The study had a modest sample size.
- A limited number of participants completed both memory paradigms.
- Adherence to Retrieval Delay Intervals
- Implementing the Object-In-Room Recall (ORR) did not strictly enforce the planned retrieval delay intervals.
- Some individuals performed recall assessments after extended delays, which may affect performance accuracy.
- Exclusion of Sessions
- Sessions with delays exceeding 240 minutes were excluded from the analysis.
- This exclusion resulted in a substantial reduction in the number of test sessions analyzed.
- Recommendations for Future Studies
- Future implementations of the ORR and similar assessments should facilitate participant integration of remote tests into daily life.
- There is a need to enforce minimized delay periods while accommodating participant needs to enhance the study design.
CONCLUSION
The results suggest that unsupervised and remote digital memory assessments hold significant potential as valuable tools for diagnosing and predicting Alzheimer’s disease (AD), mainly when used alongside plasma biomarkers. These digital assessments offer a promising and innovative approach to enhancing the accuracy and convenience of AD evaluation, transforming how cognitive decline is monitored and predicted in clinical settings.
BIBLIOGRAPHY
- [1 Hansson O. Biomarkers for neurodegenerative diseases. Nat Med. 2021;27(6):954-963. (https://doi.org/10.1038/s41591-021-01382-x)
- [2] Hansson O, Blennow K, Zetterberg H, Dage J. Blood biomarkers for Alzheimer’s disease in clinical practice and trials. Nat Aging. 2023;3(5):506-519. (https://doi.org/10.1038/s43587-023-00403-3 )
- [3] Maass A, Berron D, Harrison TM, et al. Alzheimer’s pathology targets distinct memory networks in the aging brain. Brain. 2019;142(8):2492-2509. (https://doi.org/10.1093/brain/awz154)
- [4] Berron D, Vogel JW, Insel PS, et al. Early stages of tau pathology and its associations with functional connectivity, atrophy, and memory. Brain. 2021;144(9):awab114. (https://doi.org/10.1093/brain/awab114)
- [5] Grothe MJ, Barthel H, Sepulcre J, et al. In vivo staging of regional amyloid deposition. Neurology. 2017;89(20):2031-2038. (https://doi.org/10.1212/wnl.0000000000004643)
- [6] Ritchey M, Cooper RA. Deconstructing the posterior medial episodic network. Trends Cogn Sci. 2020;24(6):451-465. (https://doi.org/10.1016/j.tics.2020.03.006)
- [7] Ranganath C, Ritchey M. Two cortical systems for memory-guided behavior. Nat Rev Neurosci. 2012;13(10):713-726. (https://doi.org/10.1038/nrn3338)
- [8] Berron D, Schütze H, Maass A, et al. Strong evidence for pattern separation in the human dentate gyrus. J Neurosci. 2016;36(29):7569-7579. (https://doi.org/10.1523/jneurosci.0518-16.2016)
- [9] Bakker A, Kirwan BC, Miller M, Stark CE. Pattern separation in the human hippocampal CA3 and dentate gyrus. Science. 2008;319(5870):1640-1642. (https://doi.org/10.1126/science.1152882)
- [10] Grande X, Berron D, Horner AJ, Bisby JA, Düzel E, Burgess N. Holistic pattern-completion recollection involves hippocampal sub-field CA3. J Neurosci. 2019;39(41):8100-8111. (https://doi.org/10.1523/jneurosci.0722-19.2019)
- [1] Berron D, Olsson E, Andersson F, et al. Remote and unsupervised digital memory assessments can reliably detect cognitive impairment in Alzheimer’s disease. Alzheimer’s Dement. 2024;20:4775–4791. (https://doi.org/10.1002/alz.13919)
Oncology Related Tools
Other
Latest Research
