Q&A with Natalia Drosopoulou: Important Implications of NIA-AA Framework for Alzheimer’s Disease Research – Part 2 of 3

By Editors of Talking Trials,


The common Alzheimer’s disease research framework proposed in 2017 by the National Institute on Aging and Alzheimer’s Association (NIA-AA) pivots upon the understanding and acknowledgment that Alzheimer’s disease (AD) refers to an aggregate of pathophysiologic processes and therefore can be uniquely defined in vivo by biomarkers and, of course, by post mortem pathologic changes but conspicuously not by clinical symptomatology (1, 2).

National Institute of Aging and Alzheimer’s Association Framework for Alzheimer’s Disease Research

Welcome to Part 2 of our three-part series featuring Worldwide’s Natalia Drosopoulou, Ph.D., Worldwide’s Senior Director of Project Management, Neuroscience, on the updated NIA-AA research framework. In Part 1 of “Implications of NIA-AA Framework for Alzheimer’s Research,” we addressed the new framework’s definition of Alzheimer’s disease, its guidance for using biomarkers in Alzheimer’s disease research, and how biomarkers help to categorize the disease.

To further understand the framework, Talking Trials asked Dr. Drosopoulou to discuss its strong emphasis on biomarkers and any potential problems.

Do you see an overreliance on biomarkers in this Alzheimer’s disease research framework?

Clinical symptoms are not totally disregarded because the updated AD research framework also provides two options for staging of clinical symptoms. The first option utilizes the terms “cognitively unimpaired”, “mild cognitive impairment,” and “dementia” to indicate the severity of cognitive impairment but not to infer underlying pathology (1). Each of these terms has been described in detail in the research framework and is worthy of further examination. Notably, each can be combined with the eight Alzheimer’s biomarker profiles outlined in Part 1 of this series to produce twelve independent categories that range from “normal AD biomarkers, cognitively unimpaired” to “non-Alzheimer’s pathophysiology contributing to dementia.”

For the most part, there is an effort to follow the nomenclature developed in the 2011 criteria – with some notable exceptions. For example, “Alzheimer’s disease contributing to MCI” is used for subjects who are positive for both amyloid and tau biomarkers but who are positive or negative for a neurodegeneration biomarker, rather than “MCI due to Alzheimer’s disease” (as in the 2011 criteria) (3). The authors suggest that an alternative approach to naming is to simply combine an ATN biomarker profile with a cognitive stage (without the use of any descriptive phrases at all) such as “A+T-N- MCI” instead of “Alzheimer’s pathophysiology contributing to MCI” or “A+T+N+ dementia” instead of “Alzheimer’s disease contributing to dementia” (1).  

Furthermore, the updated research framework also provides for a numeric neurocognitive staging option to be used exclusively for those subjects in the Alzheimer’s pathophysiologic continuum. This staging option avoids syndromal labels by using the numbers one through six to reflect the evolution of AD from an initial stage in which asymptomatic patients have abnormal AD biomarkers to the final stage reflecting severe dementia (1). Stage 1 is defined by biomarker evidence only; Stage 2 describes the earliest detectable neurocognitive consequence of Alzheimer’s pathophysiology; Stage 3 describes neurocognitive impairment that is not severe enough to result in significant functional loss; Stages 4 through 6 describe progressively greater functional loss (1). It is important to note that the two staging systems (numeric and categorical) reflect different needs and therefore vary in several important respects as well as that the numeric staging is only applicable to those in the Alzheimer’s pathophysiologic continuum while syndromal categorical staging includes all Alzheimer’s biomarker profiles (1).

Are there any missing pieces in this updated Alzheimer’s disease research framework?

There is no doubt that great effort was demonstrated in this initial draft of a research framework to guide AD drug development with the lofty goal of facilitating a more precise approach to therapeutic intervention trials in which specific pathways can be targeted at specific points in the disease process and, importantly, to the appropriate subjects (1).  

The authors proceeded quickly and decisively, building on the growing body of biomarker evidence while clearly making an attempt to build in flexibility to allow for future biomarker inclusion. The research framework is built upon the assumption that imaging and CSF biomarkers are valid proxies for pathophysiologic changes associated with the AD continuum.  

Arguments regarding the appropriate definition and use of the term “pathophysiology” by different medical fields aside, it is widely accepted that CSF beta amyloid levels (Aβ 42/40 ratio) and tau PET binding are valid indicators of the abnormal pathophysiologic state associated with brain fibrillar beta-amyloid and tau deposition, respectively. However, other prominent biomarkers associated with proteinopathies, such as alpha-synuclein and TDP43, that play a crucial role in AD have been largely overlooked.

Can Alzheimer’s disease clinical investigators incorporate other biomarkers into the new framework? 

It is acknowledged that biomarkers such as alpha-synuclein and TDP43 have only recently become available for investigation and unfortunately may not be ready for incorporation into this framework due to uncertain reliability. Although the contribution of these and other possible biomarkers (e.g., vascular biomarkers) can reportedly be easily incorporated into the current ATN framework, it appears that the addition of even two to three additional biomarkers would render the framework very complex and cumbersome with a resultant large number of biomarker categories and syndromal labels, arguably limiting the overall utility of the general model – even when a dichotomous approach to cutoff values are in place for each biomarker.  

The addition of other potential Alzheimer’s biomarkers also increases the possibility of having an incomplete biomarker profile (especially in regard to PET imaging) and of having a higher frequency of discordance among biomarkers purportedly measuring related constructs. Discordance among biomarkers can be anticipated, as evidenced by a recent study of 144 clinical patients along the AD continuum (MCI and AD patients) who had data from both PET-PiB and Aβ42 values from CSF.  The percentage of discordance was reported at roughly 14% overall but fell to roughly 6% for AD patients only. Discordance in this case was mostly caused by beta amyloid positive patients who were negative for PET-PIB (80%) (4). Although low rates of discordance can be easily addressed using simple rules for categorization or more complex weighting algorithms, having five to ten biomarkers may prove sufficiently difficult to resolve, especially in terms of assignment to cognitive status when all biomarkers are not expected to have equal contribution. 

Finally, and maybe most importantly, there is no reliable and valid biomarker for what many neuropathologists believe to be the most relevant outcome measure related to neurodegeneration, specifically micro-hemorrhages. Neuroimaging is too crude at this point to serve as a biomarker for this important measure, and other biomarkers such as neurogranin and neurofilament light chain that measure synaptic degeneration/loss and axonal injury have not been sufficiently investigated to date. Like almost all biomarkers of neurodegeneration, these reflect nonspecific indicators of damage across many neurodegenerative diseases.

Thank you for reading Part 2 of “Implications of NIA-AA Framework for Alzheimer’s Research.” In our final blog of this series, we explore other practical implications of the research framework and certain cases where it would not be applicable.