An ancient Indian subcontinent parable tells a story in which a group of blind men each touch a different part of an elephant. They cannot agree on the nature of the elephant because none of them observed the elephant as a whole. If, in this scenario, the elephant was dynamically evolving from a juvenile to an adult, it would be even more difficult for them to characterize the elephant. This parable reminds me of the challenges encountered when studying the Alzheimer’s disease (AD) continuum with a variety of multimodal biomarkers. To understand the AD continuum, we must focus not only on each of the biomarkers, but on how they relate to, and can be used to link events in, the preclinical phase of AD through overt AD dementia.
So far, we have no curable treatment for AD. Epidemiology studies suggest that delaying AD development by five years could reduce the incidence of AD by more than 50% . In order to have a window of opportunity to intervene with strategies to prevent or slow the AD continuum, we must target the preclinical and prodromal disease population. However, this is challenging because, in a preventative AD clinical trial, we do not know how to assess whether the AD developmental processes, if any, have been delayed or stopped. The AD development processes, from the insidious preclinical phase to clinical AD dementia, take decades; in fact, the amount of time it takes for a subject to evolve to AD dementia is often longer than the duration of a preventive AD clinical trial. At the preclinical phase, the lack of reliable tools to accurately predict whether preclinical subjects are likely to evolve to AD during the course of a preventive clinical trial has significantly impaired advancement in the search for effective interventional strategies.
There is a strong need to develop a practical tool that can closely link the evolution of AD from its preclinical phase to overt AD dementia [2, 3]; such a tool will be invaluable in evaluating interventional studies for delaying AD onset. For this purpose, my research team and I have developed the CARE (characterizing Alzheimer’s disease risk events) index, as described in our 2017 award-winning article published in the Journal of Alzheimer’s Disease . The CARE index takes advantage of the event-based probability (EBP) model to integrate a variety of biomarkers that represent disease events at different phases of the disease continuum. The sequential evolution of these disease events links the initial preclinical phase of AD to clinical AD dementia, therefore providing a tool to assess disease states and the risk of AD onset for each subject.
We can use this new CARE index to test a series of hypotheses. For example, the CARE index score can help to test the evolution of disease from one stage to another and be used to assign a specific state, among the 10 biology-based states, to an individual subject. The CARE index does not rely on a few discrete clinical states to characterize cognitively normal (CN), early and later mild cognitive impairment (MCI), and AD subjects. The CARE index has high state resolution and, thus, can accurately predict disease evolution and quantify the time to AD onset. For example, MCI has been conceptualized as a transitional clinical state between CN and AD dementia. MCI has been considered as a key prognostic and therapeutic target in the management of AD. However, the clinically assessed MCI subjects are CARE index-heterogeneous: As shown in our published article, the disease states of MCI subjects ranged from scores of 1 to 10 on the CARE index. Clinical observation showed that not all MCI subjects convert to AD, and many individuals remain cognitively stable or even revert to normal status. We hypothesize that those MCI subjects with higher CARE index scores will have a higher rate of AD onset than those with lower CARE index scores. In general, using CARE index scores, we can select a homogeneous subject population to sensitively predict the probability of moving from one state to other states and the time of AD onset on an individual subject level. This is in place of the “conversion rate” typically used to describe conversion from the MCI to AD state at the group level. When evaluating the outcome of a particular drug in delaying AD progression, a change in CARE index score from 5 to 6, for example, may be detectable many years prior to the detection of a change in the conversion rate from MCI to AD dementia; therefore, use of the CARE index could shorten the duration of clinical trials.
The CARE index could quantitatively differentiate between clinical AD and the normal aging processes at the whole-continuum level. Differentiating between the CN aging and AD processes at the preclinical phase has been a significant challenge because these two processes significantly share diverse neurodegenerative events measured by various biomarkers; for example, gradually decreased Aβ and increased tau levels in cerebrospinal fluid, medial temporal lobe atrophy, default mode network disruption, and cognitive deterioration were detected in both CN and AD elderly cohorts. We can generate two hypotheses based on the fact that the CN and AD cohorts share these common abnormalities: 1) Many of the abnormalities that occur in normal aging represent the preclinical stage of AD; 2) These abnormalities reflect the vulnerability of the same brain systems to various detrimental factors including neurodegenerative diseases and advancing age. How to mechanistically disentangle the association between the CN and AD processes is much debated; however, technically separating them to facilitate disease prevention and early treatment is imperative. With the CARE index, we hypothesize that the CN and AD cohorts have different trajectories of sequential neurodegenerative states over time. For example, say two CN subjects are different ages, 60 years old and 80 years old, for example, and both have the same Aβ and tau positivity . Assuming that the time from Aβ and tau positivity to AD onset is 15–20 years, the former subject may have a high risk for AD onset in their life time, whereas the later may have no AD onset in life and die from other causes.
In comparison with the many multimodal biomarker studies in the literature, the CARE index has several advantages. For example, in classification studies, separation of disease states often relies on a biomarker threshold or cut-off point as well as supervised analysis using clinical diagnostic information. In the CARE index analysis, the problematic dichotomy of selecting a biomarker cut-off point can be avoided. The CARE index score mathematically represents the probability distribution of a series of sequential disease events during AD evolution. The assessment of a CARE index score for an individual subject is an unsupervised analysis that is independent of clinical diagnostic information. Such a characteristic will have a superior advantage when analyzing interventional outcome in a double-blind, placebo-controlled clinical trial. The CARE index certainly can be applied to enrich subject selection or stratification for disease-delaying or -modifying therapies, to assess response from one state to the other, and to quantify the lifetime risk of time to AD onset.
There are several limitations in the current version of the CARE index. The current version of the CARE index involved three sets of measurements: cognitive, CSF, and MRI/fMRI. It is anticipated that information from local blood perfusion and diffusion tensor imaging measurements may help closely link the evolution of the disease. In addition, because the CSF information on Aβ and tau is often not available, equivalent events that can be used in place of Aβ and tau events should be explored.
I would like to emphasize that the CARE index, which describes AD evolution processes from the preclinical phase to AD onset, is derived from the cross-sectional dataset in ADNI. It is necessary to analyze longitudinal datasets to further confirm these evolutionary disease processes. The remaining question is how do we determine the probability of transitioning from one state to the other, the amount of time it takes to transition from one to the other, and the approximate age at which AD onset might occur for an individual subject.
 Brookmeyer R, Gray S, Kawas C (1998) Projections of Alzheimer's disease in the United States and the public health impact of delaying disease onset. Am J Public Health 88, 1337-1342.
 Jack CR Jr, Knopman DS, Jagust WJ, Petersen RC, Weiner MW, Aisen PS, Shaw LM, Vemuri P, Wiste HJ, Weigand SD, Lesnick TG, Pankratz VS, Donohue MC, Trojanowski JQ (2013) Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol 12, 207-216.
 Young AL, Oxtoby NP, Daga P, Cash DM, Fox NC, Ourselin S, Schott JM, Alexander DC; Alzheimer’s Disease Neuroimaging Initiative (2014) A data-driven model of biomarker changes in sporadic Alzheimer's disease. Brain 137, 2564-2577.
 Chen G, Shu H, Chen G, Ward BD, Antuono PG, Zhang Z, Li SJ; Alzheimer’s Disease Neuroimaging Initiative (2016) Staging Alzheimer’s disease risk by sequencing brain function and structure, cerebrospinal fluid, and cognition biomarkers. J Alzheimers Dis 54, 983-993.
 Dubois B, Hampel H, Feldman HH, Scheltens P, Aisen P, Andrieu S, Bakardjian H, Benali H, Bertram L, Blennow K, Broich K, Cavedo E, Crutch S, Dartigues JF, Duyckaerts C, Epelbaum S, Frisoni GB, Gauthier S, Genthon R, Gouw AA, Habert MO, Holtzman DM, Kivipelto M, Lista S, Molinuevo JL, O’Bryant SE, Rabinovici GD, Rowe C, Salloway S, Schneider LS, Sperling R, Teichmann M, Carrillo MC, Cummings J, Jack CR Jr; Proceedings of the Meeting of the International Working Group (IWG) and the American Alzheimer’s Association on “The Preclinical State of AD”; July 23, 2015; Washington DC, USA (2016) Preclinical Alzheimer's disease: Definition, natural history, and diagnostic criteria. Alzheimers Dement 12, 292-323.