Putting the New Alzheimer Disease Amyloid, Tau, Neurodegeneration (ATN) Diagnostic System to the Test

The field of neurodegenerative dementias, particularly Alzheimer disease, has been limited by challenges in accurate diagnosis, but has recently been potentially revolutionized by the development of imaging and cerebrospinal fluid biomarkers. These biomarkers have influenced the diagnostic evaluation of symptomatic patients with cognitive impairment or dementia, particularly in dementia subspecialty practice. The primary biomarker modalities include MRI, PET, and cerebrospinal fluid. MRI has widely accepted clinical utility for the evaluation of structural brain lesions of a variety of types, including evidence for cerebrovascular disease and atrophy patterns consistent with, but not specific for, neurodegenerative pathologies. FDG PET has strong evidence and a recent practice guideline (1- Nobili, Arbizu et al. 2018) supports its use as a marker of functional brain abnormalities suggestive of a variety of neurodegenerative pathologies associated with dementia. Amyloid PET is an FDA approved biomarker that is sensitive and specific for fibrillar amyloid plaques, a fundamental pathologic feature of Alzheimer disease; an appropriate use guideline specified how amyloid PET could be usefully deployed in subspecialty clinical practice (2 - Johnson, Minoshima et al. 2013). A recent large study also provided evidence supporting the utility of amyloid PET in dementia subspecialty clinical practice (3- Rabinovici, Gatsonis et al. 2019). In addition, several PET tracers that appear to bind to tau-based neurofibrillary tangles (NFTs), the other pathological hallmark of AD, have emerged (4 - Villemagne, Doré et al. 2018). Alternatively, cerebrospinal fluid can be analyzed for levels of amyloid Beta, as well as tau proteins suggestive of NFTs. A recent practice guideline supports the value of CSF AD biomarkers in the subspecialist evaluation of patients with cognitive impairment or dementia (5- Shaw, Arias et al. 2018). Thus, these biomarkers are increasingly affecting clinical practice for the evaluation of symptomatic patients with cognitive impairment, and are being used extensively in research. While it is clear that these varied tests improve diagnostic accuracy and treatment planning now, their full potential to impact outcomes will likely grow with the emergence of more effective therapies.

In parallel, remarkable developments have taken place demonstrating the capacity to measure these biomarkers of key pathological features of AD in cognitively normal individuals. These individuals have been classified as having preclinical AD and the assumption is that a high percentage of those with this pathology will ultimately develop symptomatic disease. Research diagnostic constructs to define preclinical AD were first established in 2011 (6 - Sperling, Aisen et al. 2011) and have been refined using the so-called amyloid, tau, neurodegeneration (ATN) system (7- Jack, Bennett et al. 2016) with the recent proposal of a new research framework defining AD using these three categories of biomarkers, dichotomously classified as positive or negative, and proposing a separation between the definition of the neuropathological disease and clinical syndromes of cognitive impairment (8 - Jack, Bennett et al. 2018). The ATN framework defines AD biologically, as requiring the presence of amyloid plaques (i.e. A+) and neurofibrillary tangles (T+) akin to neuropathological definitions and research is beginning to evaluate this new framework in older adults without dementia. N represents neurodegneration that is typically measured with MRI and FDG PET.

In this issue of JAMA, an important new contribution on this topic is reported by some of the originators of this framework (9 - Jack et al. 2019). In this report, the authors describe the first study to examine longitudinal clinical outcomes associated with ATN classification among 480 individuals without dementia, including mostly cognitively normal adults (92%) and some with mild cognitive impairment. The investigators compared the predictive value of current clinical and genetic measures, including demographic characteristics, apolipoprotein e status ApoE, the strongest genetic risk factor for AD), and cardiovascular and metabolic conditions, with adding the ATN classification for prediction of longitudinal decline in memory. The ATN framework is still a research construct, and its validation and ultimate utility depend on data supporting its link to clinical outcomes. The authors report that inclusion of ATN biomarkers enhanced prediction of cognitive loss beyond clinical and ApoE4 data alone although this difference was relatively small (r-squared of 0.31 vs 0.26). However, several ATN groups were associated with substantially higher rates of annual memory decline (i.e. A+T+N+, A+T+N-, A+T-N+) that were equivalent to being 20 years older in age. As these groups of non-dementia patients with more extensive AD pathology had faster rates of decline, these findings indicate a potentially useful role of AD biomarkers in forecasting clinical course deterioration among pre-dementia patients.

Another important contribution of this study is that approximately 50% of the memory change with older age was associated with underlying AD pathology in pre-dementia individuals (estimated based on changes in the prevalence of more AD-enriched ATN groups with older age). This supports the concept that a substantial portion of the age-related changes observed in “normal” aging are actually related to the presence of AD-related pathological changes. Yet it remains unclear what factors may account for the other 50% and whether they are linked to distinct other pathologies or reflect non-specific effects of aging.

The findings of the study by Jack et al may be most immediately relevant for use in clinical trials, allowing patients to be categorized into distinct prognostic groups that are more or less informative regarding therapeutic agent effects. Several current and proposed trials in cognitively normal adults enroll study participants on the basis of a “positive” amyloid PET scan (10 - Touchon, Rosenbaum et al. 2017). However, the data in the report by Jack et al suggest that amyloid alone (A+T-N-) does not seem to be related to an increased rate of decline (at least over this relatively short follow up interval) compared to non-AD continuum groups. It was only in the presence of concomitant tau pathology, neurodegeneration, or both that the rate of decline was increased. As such, these latter groups may be more likely to show progression in control groups that could be modified in actively treated groups in prevention studies of only a few years duration.

However, the analysis by Jack also has some potential limitations,, many of which were considered by the authors. For instance, the clinical measures included ApoE, which is highly associated with the presence of cerebral amyloid and may diminish the predictive value of the amyloid PET measure. Also, the clinical variables did not include cognitive measures, which are often linked to the presence of NFTs and neurodegeneration. Their lack of inclusion may overvalue the predictive value of T and N. Indeed, most N+ groups tended to have poorer memory at baseline. Further, the groups that displayed the most decline also tended to be associated with lower baseline cognition. Thus, the added value of ATN may be smaller with inclusion of cognitive measures, which are generally easier and less expensive to obtain in the clinical context.

Moreover, creating distinct cut-offs (and categories) remains a major hurdle for standardization of ATN and dichotomous decisions may be especially challenging along the natural continuum for tau PET and MRI changes. The selection of regions of interest (ie, xxx AU: explain what you mean by ‘reporter regions of interest) for tau and neurodegeneration also will influence findings, as different regions will have variable sensitivity and specificity to different disease stages. In addition, there is no agreement currently on the degree to which these biomarkers, particularly N, should be controlled for age. Age may have non-specific effects on morphometric measures of a number of brain regions that may be independent of distinct “pathology”. Not controlling for age may reduce specificity to AD related changes (11 - Dickerson, Wolk et al. 2013). Notably, age was not fully controlled for in the current analysis and more than 73% of study participants older than 80 years were N+, compared with 24% of those aged 60–69 years who were N+. Many laboratory diagnostic test values have age-adjusted ranges, which likely would make MRI measures more specific for neurodegeneration as defined by neuropathologists.

An interesting finding in the study by Jack et al was related to the A+T-N+ category and the fact that patients in this group demonstrated faster cognitive decline than those in the A-T-N+ group. In both cases, N is thought driven by non-AD pathology. The underlying etiology in the A-T-N+ group is unclear. Enriching the framework with meausres of other important contributors to age-related cognitive decline such as vascular, TDP-43 and alpha synuclein could increase the diagnostic and prognostic precision of the framework. (12 - Wilson, Yang et al. 2019).

In addition, application of ATN neuroimaging (an MRI scan and two PET scans) is expensive and will not be practical to add to many research projects. The added value of each specific ATN measure has not been fully delineated, and requires evaluation in future work, as does whether the findings of the current study will generalize to other biomarkers (CSF, FDG) and more diverse patient populations.

Despite these caveats, the study by Jack et al represents an important contribution not only to advancing the conceptualization of AD, but also for putting this new framework to the test rapidly in a relatively large sample of participants.