In this Webinar, Clifford Jack gave a slide talk summarizing a staged biomarker model of the Alzheimer’s cascade, followed by a panel discussion with Chet Mathis, David Holtzman, Henrik Zetterberg, Paul Aisen, Keith Johnson, Giovanni Frisoni, Douglas Galasko, and Larry Altstiel.
Please see comment by John Trojanowski.
- View/Listen to the Webinar
Click on this image to launch the recording.
By Gabrielle Strobel
The driving ambition of the past decade of Alzheimer disease research—human research, at least—has been the quest for validated biomarkers. But for most people, it was easy to lose track of a rapidly growing, sometimes contradictory literature. Which marker changes first? Second? Last? Which predicts disease? Tracks with cognitive decline? Flags if a drug works? Drug trials can’t employ all markers—how to choose? The answers boil down to how each marker matches up with stages of the disease. Clifford Jack, a leading MRI researcher at the Mayo Clinic in Rochester, Minnesota, has devoted much thought on integrating what is currently known about the five best-validated markers into a dynamic model of how things change over time on the long road to full-blown AD dementia. (Teaser tidbit: Aβ is first but far from be-all, end-all.) Jack presented his thinking at last summer’s ICAD conference in Vienna, Austria. There, it was uniformly welcomed as a fair-minded synthesis that captures an emerging consensus on how current biomarkers fit into the AD cascade. In the past, some scientists have naturally tended to favor the biomarker they’d been studying the most at their respective institutions. Representatives from each of these fields noted in Vienna that Jack’s model orders the markers objectively, moving the debate forward for everyone.
The model appeared as a “Personal View” essay in this month’s Lancet Neurology (Jack et al., 2010). Jack’s coauthors include clinicians David Knopman and Ron Petersen at Mayo; FDG-PET expert William Jagust at University of California, Berkeley; ADNI CSF core leaders Leslie Shaw and John Trojanowski of University of Pennsylvania in Philadelphia; ADCS leader Paul Aisen of UC San Diego; and ADNI chief Michael Weiner of UC San Francisco. Jack himself heads the MRI core of ADNI. In essence, their article states that of the five most studied fluid and imaging markers to date, Aβ changes first, as early as 20 years before symptoms appear. It also suggests that by the time a person has dementia, the disease has become quite disconnected from Aβ itself. MRI begins to change much later in the disease process, but then stays closely tied to it as dementia worsens. The authors place the synaptic dysfunction marker FDG-PET and the neurodegeneration marker CSF tau in between Aβ and MRI. The Alzforum editors recommend this article. For Webinar attendees without access to this journal, those without deep background in AD biomarker studies, or for those who simply had no time to absorb the 10-page article and its 115 citations, below is your editor’s summary.
The five biomarkers Jack et al. discuss include two for brain Aβ plaque deposition (CSF Aβ42 and PET amyloid imaging) and three for neurodegeneration (CSF tau, FDG-PET, and structural MRI). These markers are validated enough to be used in active therapeutic trials or large multisite observational studies such as ADNI. They have helped the field move past the black-and-white view of Alzheimer disease held some decades ago, when researchers thought, essentially, that people with AD pathology had dementia and people without AD pathology did not. This simple division started blurring when pathology studies revealed plaques and tangles in a sizable fraction of elders who had died with their cognition intact, and biomarker and longitudinal aging studies of the past 20 years subsequently swept aside this old binary view in favor of a more nuanced, dynamic picture. Now, the majority view holds that both pathologic and clinical changes occur gradually over time. Only after decades of early molecular change leading to neural dysfunction and degeneration, which in turn lead to clinical change and eventually functional loss, does the patient reach dementia as the end stage of the disease process. Importantly, the authors point out, each of these individual aspects develops on its own time course. At any given point in time, a given brain area may be at a different point along this process while the disease slowly eats its way through the brain.
Image credit: Cliff Jack
Such a long process cries out for a sequential ordering, in other words, a way of staging Alzheimer disease by means of individual biomarker change. Clearly, individual biomarkers do not become abnormal and peak simultaneously, Jack et al. write.
Ordering biomarkers in time is important not only because it expresses the disease process in terms of a series of testable biological indicators, but also because clinical trial sponsors are searching for the best way to apply biomarkers for selecting patients and to measure drug effects. This is true both for a growing number of trials of disease-modifying drugs and for the concept of future prevention trials that is building steam across the field. Furthermore, understanding the temporal order of each biomarker then makes it possible to use a given marker for staging AD in vivo.
For background, Jack and colleagues first summarize what’s known about each of the five markers. They write that evidence from many labs strongly supports the notion that amyloid PIB-PET and low CSF Aβ42 are valid biomarkers for a person’s brain Aβ plaque load. Increased CSF tau is an indicator of tau pathological changes and neuronal injury, and it correlates with clinical disease severity. CSF tau is not specific for AD in that it also changes in stroke and brain trauma (e.g., after a boxing match) and, curiously, it does not change in pure tauopathies such as supranuclear palsy. FDG-PET measures brain metabolism and probably indicates synaptic activity. FDG-PET decreases in the AD pathogenic process from normal to AD in a regional pattern that matches affected areas and also correlates with cognitive impairment. According to Jack and colleagues, FDG-PET is a valid indicator of the synaptic dysfunction that accompanies neurodegeneration in AD. Last but not least, structural MRI measures cerebral atrophy as synapses disappear and neurons die. Like tau, it is not specific to AD, but it correlates closely with the severity of a person’s clinical impairment even late into the disease and with Braak staging and tangle pathology at autopsy. (It actually correlates for longer into the disease than CSF tau does, see below.)
So how to order these top five? In a nutshell, Jack and colleagues suggest that biomarkers of Aβ deposition become abnormal first, long before neurodegeneration and clinical symptoms appear. Biomarkers of neuronal dysfunction, injury, and neurodegeneration turn abnormal later. Cognitive symptoms are tied to biomarkers of this second category, not to biomarkers of Aβ deposition.
What are the main points of evidence the scientists cite for their model? On Aβ deposition happening first among the five markers, the authors note that autopsy, PIB-PET, and CSF Aβ42 studies all agree that up to 40 percent of cognitively normal elderly have a sufficient plaque burden to meet diagnostic criteria for AD. So amyloid alone does not equal dementia; however, it precedes and predicts dementia. By the time the first subtle symptoms appear some years later, all the other biomarkers have already turned as well. “There is strong evidence that MRI, FDG-PET, and CSF tau biomarkers are already abnormal in patients who are in the MCI phase of AD,” Jack and colleagues write. (This sentence appears to advance a related debate about the continued need for keeping mild cognitive impairment, or MCI, as a separate diagnostic category. In recent years, a shift away from using MCI for diagnosis in research settings has occurred along with the movement toward biomarkers [Dubois et al., 2007], particularly in Europe. While the present article also contains a statement about the continued usefulness of the MCI category and the term “dementia” in clinical practice, the article’s main thrust of staging the AD process by means of biomarker changes portrays the MCI stage of AD as part of a seamless disease continuum.)
The essay cites numerous studies, all supporting the conclusion that the abnormality of Aβ plaque biomarkers approaches a plateau before MRI atrophy and cognitive symptoms appear. After that point, the Aβ plaque load remains quite static. By and large, Aβ deposition is said and done before the person notices anything is amiss. In contrast, abnormalities in neurodegenerative biomarkers pick up speed as symptoms appear and from then on keep pace with a person’s cognitive decline. This is most strongly true for MRI.
In terms of what changes first and last, Aβ and MRI form the extremes, but how about the markers in between, i.e., CSF tau and FDG-PET? It is important to understand where on the disease continuum each of these is most dynamic, i.e., changes the most, and where it reaches its maximum, the authors note. On CSF tau, they cite a number of studies suggesting that this marker is most informative just prior to and during the earliest clinical symptoms, but—unlike MRI—no longer changes appreciably as the patient advances deep into dementia.
With regard to FDG-PET, the authors emphasize an interesting point. In general, each biomarker becomes abnormal on its own, non-linear time scale, and this puts in-vivo staging using biomarkers within reach. Beyond that general insight, FDG-PET in particular offers an additional advantage for staging in that it gives good anatomical resolution early on. Imaging abnormalities spread from one brain area to the next, such that at any given point in time, different brain areas are at different stages of disease. In other words, in any given area of a person’s brain there might be an amyloid phase, a neuronal dysfunction phase, an atrophy phase, and this will occur in an anatomical order. In this way, imaging is more informative than CSF, the authors write. Both FDG-PET and MRI volumetry give this staged anatomical information, but the former does so earlier in the disease process. The sigmoid curves of declining FDG-PET uptake take off first in the posterior cingulate, then in the lateral temporal, and later still in frontal areas. In graphs plotting biomarker abnormality across disease stage, these curves all originate to the left of—i.e., earlier than—otherwise similar regional MRI curves.
What does all this mean for clinical trials? Jack et al. confirm that selecting patients for trials of anti-amyloid drugs on the basis of Aβ biomarker evidence makes sense. One trial of prodromal AD is currently doing so; it requires abnormal CSF Aβ42 values and mild cognitive symptoms as inclusion criteria and excludes people with diagnosed dementia. More trials are being planned in similar ways. Biomarkers for neural dysfunction or neurodegeneration are not specific to AD and should not take precedence over Aβ markers as inclusion criteria, Jack et al. write. On the other hand, change in Aβ load relates poorly to improved cognition. Hence, it may hold little promise as an outcome measure of clinical importance beyond showing narrowly that the drug has engaged its target in the expected way. The “later” biomarkers could be used to show if the drug treats the neurodegeneration that marks the AD process, possibly as co-primary outcome measures along with the clinical ones that are in use today.
Any serious attempt at synthesizing a large literature into a central model is wont to draw criticism as well as praise, and the authors of this model hasten to point out some caveats themselves. For one, they write, they are fully aware that every published observation does not fit their model. For another, the model is generic, meaning there is room for different disease courses in individual people.
Moreover, ongoing research continues to generate data for a finer-grained separation of when which biomarker changes. For example, early indications are that CSF Aβ precedes PIB-PET by a bit; likewise, data emerging from ADNI suggest that the FDG-PET curve might begin to take off earlier, i.e., lie to the left, of the CSF tau curve.
Importantly, the model has big gaps where robust biomarkers simply do not exist yet. Would if there were chemical biomarkers for Aβ oligomers or imaging markers for diffuse amyloid deposits! Similarly, the field urgently awaits PET ligands for tau abnormalities, microglial activation, and α-synuclein deposits, the latter to assess overlapping pathologies on the spectrum from AD to Parkinson disease. Seeing TDP-43 deposits with PET scanning would be nice, too. “Thus, our biomarker model…is just that—a model of the stages of disease that can be assessed with currently validated biomarkers, and not a comprehensive model of all pathological processes in AD,” the authors acknowledge.
They do, however, point a finger at one widely accepted notion that is simply not borne out by this biomarker model of AD. A much-cited postmortem study that describes entorhinal neurofibrillary tangles as preceding cortical plaques (Braak and Braak, 1997; Braak and Braak, 1991) has led to the conclusion that tangle accumulation is the initiating event in AD (Duyckaerts and Hauw, 1997). This view finds no support in AD genetics, either, the authors write. On the question of whether Aβ or tau becomes abnormal first in humans, the answer appears to be in.
Other observations do not fit the model at present but may need to be incorporated as scientists learn more. For example, some PET studies show abnormally low FDG uptake in cognitively normal ApoE4 carriers even at young adult ages, decades before AD symptoms would be expected (Reiman et al., 2004; Scarmeas et al., 2003). It’s a puzzle at present if this is a developmental consequence of growing up with ApoE4 or a very early harbinger of AD. Likewise, other imaging modalities may find a place in this model as they mature and are becoming more widely used, for example, connectivity imaging using functional MRI, the authors write.
In all, the new model offers scientists a rich framework for testing various hypotheses. For example, because ApoE4 carriers tend to get AD earlier, the sigmoid curves of their biomarker changes should be shifted to the left (i.e., earlier in time) than those of non-carriers. Also, modifying factors that lengthen the lag time between Aβ accumulation and symptoms, such as cognitive reserve, should spread the Aβ and neurodegeneration curves farther apart on the time axis; this would represent people who function well for quite some time despite having a head full of amyloid. Au contraire, factors thought to shorten this lag time, such as cerebrovascular disease, would be expected to move the curves closer together; this would represent people in double trouble who become symptomatic soon after amyloid settles in their brains. Through studies that combine genotyping with longitudinal biomarker panel measurements, for example, ADNI, genes influencing AD risk can be tested for whether they affect this lag phase as well.
- Jack CR, Knopman DS, Jagust WJ, Shaw LM, Aisen PS, Weiner MW, Petersen RC, Trojanowski JQ. Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade. Lancet Neurol. 2010 Jan;9(1):119-28. PubMed.
- Dubois B, Feldman HH, Jacova C, Dekosky ST, Barberger-Gateau P, Cummings J, Delacourte A, Galasko D, Gauthier S, Jicha G, Meguro K, O'brien J, Pasquier F, Robert P, Rossor M, Salloway S, Stern Y, Visser PJ, Scheltens P. Research criteria for the diagnosis of Alzheimer's disease: revising the NINCDS-ADRDA criteria. Lancet Neurol. 2007 Aug;6(8):734-46. PubMed.
- Braak H, Braak E. Frequency of stages of Alzheimer-related lesions in different age categories. Neurobiol Aging. 1997 Jul-Aug;18(4):351-7. PubMed.
- Braak H, Braak E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol. 1991;82(4):239-59. PubMed.
- Duyckaerts C, Hauw JJ. Prevalence, incidence and duration of Braak's stages in the general population: can we know?. Neurobiol Aging. 1997 Jul-Aug;18(4):362-9; discussion 389-92. PubMed.
- Reiman EM, Chen K, Alexander GE, Caselli RJ, Bandy D, Osborne D, Saunders AM, Hardy J. Functional brain abnormalities in young adults at genetic risk for late-onset Alzheimer's dementia. Proc Natl Acad Sci U S A. 2004 Jan 6;101(1):284-9. PubMed.
- Scarmeas N, Habeck CG, Stern Y, Anderson KE. APOE genotype and cerebral blood flow in healthy young individuals. JAMA. 2003 Sep 24;290(12):1581-2. PubMed.
- Li M, Ona VO, Guégan C, Chen M, Jackson-Lewis V, Andrews LJ, Olszewski AJ, Stieg PE, Lee JP, Przedborski S, Friedlander RM. Functional role of caspase-1 and caspase-3 in an ALS transgenic mouse model. Science. 2000 Apr 14;288(5464):335-9. PubMed.