The TAS Test project: a prospective longitudinal validation of new online motor-cognitive tests to detect preclinical Alzheimer’s disease and estimate 5-year risks of cognitive decline and dementia

The Lancet Commission [1] has described dementia as “the greatest global challenge for health and social care in the 21st century.” The progressive brain degeneration and associated loss of cognitive function have devastating effects on quality of life for people with dementia and their families, and enormous economic impacts on health and social care systems. The worldwide prevalence of dementia is rapidly rising, driven by ageing populations, and current figures of 50 million people living with dementia globally are predicted to triple to more than 152 million by 2050 [2].

The World Health Organization’s key strategy to reduce dementia prevalence is through prevention [3]. Research indicates that up to 40% of dementia cases worldwide could be prevented or delayed by addressing modifiable lifestyle and/or medical risk factors such as physical inactivity, smoking and hypertension [1, 2]. To maximise the effectiveness of dementia prevention strategies, we urgently need readily-accessible population-level screening tools – to facilitate targeted interventions for high-risk individuals or those early in the disease course.

The most common cause of dementia is Alzheimer’s disease (AD), that clinically manifests as progressive loss of memory, language and other cognitive functions. The underlying brain pathology is characterised by accumulation of abnormal amyloid beta (Aβ) and phosphorylated tau (p-tau) proteins followed by neurodegeneration. New data show that silent and progressive brain changes occur 10–20 years before cognitive symptoms emerge, with two stages of AD occurring before dementia: preclinical AD and mild cognitive impairment (MCI) [4, 5]. This critical period of preclinical AD, when there are no detectable clinical symptoms, is the optimal time to intervene with targeted dementia prevention interventions early in the disease [1].

Positron emission tomography (PET) brain scans and cerebrospinal fluid (CSF) tests [6] can measure preclinical AD pathology but they are too invasive, specialist or costly to be used at the population level; currently they are only available in a small number of hospitals and rarely in low- and middle-income countries. Standard pen and paper cognitive tests lack sensitivity in preclinical AD [7]. However, over the past 3 years, a range of new blood-based biomarker assays have been developed that can quantify AD pathology with minimal invasiveness, including in the preclinical phase.

One of the most promising blood-based biomarkers is plasma p-tau181, as it has been shown that levels increase across the AD clinical continuum and correlate strongly with CSF and PET measures of AD pathology [8, 9]. Furthermore, antemortem plasma p-tau181 levels 5–8 years before death predict AD brain pathology at post-mortem, including in those with unimpaired cognition [10, 11]. Baseline plasma p-tau 181 levels in individuals with unimpaired cognition (as well as those with impaired cognition) are predictive of prospective cognitive decline and neurodegeneration [12]. In summary, the development of blood-based biomarkers is a major advance towards identifying preclinical AD in clinical and research settings, and plasma p-tau181 is considered both sensitive and specific to AD pathology. However, the practicalities and cost of obtaining a blood sample and accessing highly-specialist analytic equipment limit wide accessibility to most populations around the world, as well as for large-scale population intervention studies. Very few research centres have access to the specialist and expensive highly sensitive analytical infrastructure required to analyse the blood samples and each biomarker test typically costs at least $50–100 USD.

Thus, there remains urgent need for low-cost non-invasive and accessible tests that can pre-screen individuals for their likelihood of AD pathology prior to more specialist and expensive biomarker tests, or simply to identify individuals at higher risk for targeted risk reduction. A population-level screening test for preclinical AD would have wide applications in public health initiatives and clinical trials. Similar to screening programs for other chronic diseases such as diabetes, cancer and heart disease, a test that allows people to identify likely preclinical AD pathology would provide more time to reduce dementia risk before cognitive decline and before the brain degenerates; it would also enable opportunities for early recruitment to clinical trials. This would transform the effectiveness of dementia secondary prevention strategies in reducing dementia incidence.

There is a growing body of research that shows movement analysis is an effective method to identify the preclinical AD phase; several studies have shown that patterns of human movements change in preclinical AD and decline across the AD continuum. For more than a decade, research studies have found that slowed walking, especially when a dual motor-cognitive task was performed, predicted cognitive decline and dementia years later [13, 14]. However, the scalability of detailed gait analysis is limited by the need for wearable sensors, in person measures, or gait laboratories, and there are also substantial risks of falls.

More recently, several different types of studies have provided strong evidence that hand movement analysis is a sensitive biomarker of preclinical AD. First, functional magnetic resonance imaging (MRI) research has revealed that the human posterior cingulate cortex is specifically involved in control of hand movements [15] and this is highly relevant because the posterior cingulate cortex is known to be one of the first areas of the brain to show abnormalities in preclinical AD – in terms of both hypometabolism [16] and Aβ deposition [17]. This new imaging discovery accords with well-established single-unit microelectrode studies in monkeys that show high activity in the posterior cingulate cortex during forelimb self-paced tapping movements via projections to the primary motor cortex [18]. Taken together, these studies provide evidence that hand movement tests can detect the earliest pathological changes associated with preclinical AD.

Further, in a recent study of older adults with preclinical AD (defined by having elevated CSF Aβ indicative of AD, but with unimpaired cognition) Mollica et al. (2019) found that the speed and variability (a measure of rhythm) of simple finger tapping tests on a computer keyboard are abnormal in preclinical AD, and the degree of variability correlates with CSF Aβ levels [19]. These findings support earlier studies that determined hand reaction times correlate highly with CSF Aβ biomarker levels in preclinical AD [20] and progressively worsen over the AD continuum [20‐22]. Additionally, it has been shown that even with unimpaired cognition ApoE4 carriers have been found to have delayed hand reactions compared to non-carriers, suggesting hand movements deteriorate at the very earliest stages of preclinical AD [20].

Based on this mounting evidence, and the wide accessibility of computers, including among older adults [23], we have devised TAS Test (or ‘Tasmanian Test’¹), a new 20-minute online platform of digital motor-cognitive tests that are designed to be completed in people’s own homes without any specialist supervision. Specifically, TAS Test collects a range of hand movement data using standard computer cameras, keyboards and mouse-click data. It builds upon our work that found computer vision and other cutting edge artificial intelligence (AI) technologies can accurately measure hand movements using standard computer equipment [24‐27], and that hand movements deteriorate with cognitive decline [28, 29].

For the first time, we will use AI video technologies in a health-related test. Similar technologies have already successfully automated cancer screening from static images (eg breast and lung from mammograms and radiographs, respectively), but we will analyse much richer video images, and collect these in the participant’s own home, or preferred remote setting. Our prior research shows that computer vision accurately measures hand movements [24, 25, 27, 30] and TAS Test combines these advanced technologies with hand-movement tests that are known to be affected in preclinical AD [19‐22]. Using plasma p-tau181 test results as ground-truth, this novel remote method has high potential to reliably predict preclinical AD risk. In addition, it also includes computerized tests of cognition that are based on principles of other neuropsychological tests. This will allow exploration of motor-cognitive associations and the development of multi-modal predictive algorithms using a range of different types of data.

There are several practical advantages to using hand movement as the core assessment of a population-level test - including the ease with which it translates across other countries, cultures and languages, and the speed and ease of assessment that can be performed while sitting in front of a home computer. With wide accessibility to computers, we envisage that TAS Test may be an inexpensive first-line population-level screening test that can identify high-risk individuals, including those in remote and rural communities, for further evaluation (including with blood biomarkers, where available) and/or risk modification.

The planned outcome of this project is TAS Test, a new inexpensive computer screening test to estimate the risk of preclinical AD, cognitive decline and AD dementia. If validated, this new scalable tool can potentially transform dementia prevention and research globally. The significant advantages of using a hand-movement based protocol, along with a broad measure of oromotor, visuospatial, memory and language functions, are sensitivity to early preclinical AD and a protocol that has minimal language or cultural barriers. The advantages of using an online test with standard computer equipment is the global reach of the internet crossing geographical barriers and providing accessibility for people in rural and remote communities and those in low-income countries.

Potential risks of the study are acknowledged and strategies to mitigate these are now discussed. There is a risk of inadequate recruitment but we will mitigate this by recruiting existing participants in established research cohorts, namely the longitudinal cohorts of THBP, HBA and ISLAND. There is a risk that TAS Test is not sufficiently accurate but we have mitigated this risk by selecting component test items based on evidence of sensitivity to preclinical AD, combining multiple tests to amplify the multivariable model input data [> 10,000 data points], and electing well-established, transparent statistical modelling approaches that reveal the most discriminatory components of motor-cognitive data, allowing further refinement. We have also devised a study protocol that plans to collect a sufficiently large dataset to employ multiple modelling methods, including feature-agnostic deep learning. A further risk is that participants lack a computer camera to provide video-recorded hand movement data, or do not wish to do so; we have mitigated this risk by recruiting participants from studies that have online assessments already, offering the opportunity for participants to attend the clinical research centre in person if preferred, and providing alternate methods (such as keyboard tapping test and mouse click reaction time tests) within TAS Test that collect movement data. It is conceivable that some participants will not want to know their dementia risk and this will hinder selection of participants for the clinical subset assessments, but we have minimised this risk by recruiting from studies that focus on reducing dementia risk, where participants have already undergone multiple tests of dementia risk. It is also important to acknowledge that, as the bulk of this project relies on self-report of known neurological diagnoses, we have limited ability to make distinctions between other neurodegenerative disorders, including other forms of dementia and disorders which are correlated with dementia (for example, Parkinson’s disease). Finally, there are risks around COVID-19 pandemic restrictions limiting recruitment or progress of the study; as most of the study is based around online movement and cognitive tests that can be completed at home, there are likely to be minimal effects and we will be able to collect blood for p-tau181 levels using personal protective equipment or at a later date.

In summary, this study directly addresses the critical need for population-level screening tests to detect the earliest stages of dementia. ‘We can’t manage what we can’t measure’ sums up a global critical struggle against the rising tide of dementia and the odds are stacked against drug trials and preventative strategies as AD is typically detected so late in the disease course - when cognitive symptoms arise. Without a screening test to identify preclinical AD risk early, our preventive strategies will remain blunted and belated. There is high potential that TAS Test may provide a scalable screening approach, where the test equipment is already in our homes and health centres.