Two new longitudinal analyses suggest that even small elevations in brain amyloid, far below the threshold of brain-wide amyloid positivity, can be associated with subtle memory decline and cortical tau deposition. In people with low but detectable amyloid, the rate of accumulation, more than their absolute Aβ level, predicted worsening memory and upped the chance of an abnormal tau scan later on, say the studies, both from the lab of William Jagust at the University of California Berkeley.

  • Two studies look at changes in amyloid in people with sub-threshold levels.
  • In both, the rate of amyloid accumulation, not baseline levels, best predicts declining memory and tau accumulation.
  • Researchers need better ways to identify these amyloid accumulators for clinical trials.

The work sheds light on what happens in the amyloid-negative phase of the decades-long Alzheimer’s continuum. One study, published April 24 in Neurology, tied a five-year course of low-level amyloid accumulation to cognitive changes, while the second, in the April 28 Journal of Neuroscience, showed abnormal tau PET emerging even in the amyloid-negative group. In both cases, faster rate of Aβ increase foretold more dramatic changes.

Amyloid accumulation is a slow, creeping process, but with the advent of amyloid PET measures for diagnosis and the need to group participants in clinical trials, scientists decided to divide people into just two categories, amyloid-positive and -negative. Amyloid load above a much-debated cutoff of ~1.1 SUVR on a global brain scan associates with greater cognitive decline, even in people who start out cognitively normal (Landau et al., 2012Sep 2014 newsJun 2017 news), but what are the risks with lower amyloid levels, and when does danger start? Jagust uses the trove of longitudinal data scientists have gathered over the last 10 years to address those questions. “We want to know when Alzheimer’s disease starts,” he told Alzforum. “We want to find out when is it too early or too late for a successful therapeutic intervention.”

In the Neurology paper, lead author Susan Landau reports analysis of data from 142 cognitively normal, healthy older people in ADNI who were tracked over time with multiple florbetapir PET scans and with tests of memory or executive function. With an average age of 75, all participants were classified as amyloid-negative at baseline.

Each had between two and four florbetapir scans over an average of four years, with annual cognitive testing. To reduce noise in the data, Landau used a composite reference region comprising brainstem, whole cerebellum, and white matter instead of the standard cerebellum-only reference. The 0.79 standard uptake value ratio cutoff for positivity corresponds to a cerebellum-standardized SUVR of 1.11. Thirteen people, or 9 percent, crossed the threshold to amyloid-positivity in the course of the study.

Accumulators: Repeat PET scans, shown in different-color spaghetti lines, identify people who test negative for brain amyloid at baseline, and either stay stable or accumulate Aβ over time. The dotted line indicates the brainwide SUVR threshold for positivity. [Courtesy of Landau et al. Republished with permission © 2018 American Academy of Neurology.]

In the group, 60 percent of people saw their amyloid going up with time. The faster a person’s amyloid increased, the more likely they were to have declining scores on repeat tests of memory. Executive function was not affected.

“If you looked at a single amyloid scan for these people, you would see a negative scan and assume they’re fine. But if their amyloid was rising on multiple scans, they would be likely to show memory decline,” Jagust told Alzforum. “You couldn't see that at a single time point,” he said.

Landau emphasized that the declines in memory are very subtle: “We are measuring very small changes within both the normal amyloid range and the normal cognition range—we’re not talking about people who are amyloid-negative losing the ability to do basic tasks. The small memory deficits would likely not even be apparent to family members.” The important message is that people did not have to be amyloid positive to see consequences of amyloid accumulation, she said.

The memory decline is a very interesting finding, said Oskar Hansson, Lund University, Sweden. “That picks up on what we have been seeing in our cohort—an association mainly between amyloid and memory, and not executive function. If that is reproduced in other cohorts, it has implications for the outcomes we use in trials,” he said.

From the analysis, it wasn’t clear why some people accumulated amyloid faster than others. Their slope did not correlate with any of a host of variables, including age, sex, education, ApoE4 status, baseline memory or executive function, brain glucose metabolism, or hippocampal volume. The only association Landau found was with baseline amyloid levels—people with higher baseline Aβ tended to increase faster over the course of the study. That suggests that in these people, Aβ was already depositing at baseline, and then continued to build apace over the course of the study. The lowest of the low had the slowest increase over time, with many showing flat or even negative slopes for accumulation. Those negative slopes likely stem from noise in the measurements, and not amyloid clearance, the authors said. It is unknown why some people have higher baseline levels than others.

In the second study, first author Stephanie Leal analyzed PiB PET data from the Berkeley Aging Cohort, a different longitudinal study. The 51 cognitively normal older people with a PiB SUVR of less than 1.07 also averaged 75 years old, and were followed for an average of 4.5 years with multiple scans. In 27 of them, the researchers also collected tau PET data using AV-1451, three months after the last PiB scan.

In this cohort, both baseline amyloid and the rate of increase correlated with tau PET levels. “We found the same as we saw for memory—the faster the amyloid was going up over time, the more tau there was in the brain,” Jagust said. “I was really surprised to see that—at baseline, everyone was within a narrow range of negative amyloid values, and still, the more amyloid they had, the more tau they ended up with five years later. That stunned me—we’ll have to see if it replicates,” he told Alzforum. To be sure, tau levels were quite low, with scans for the majority of subjects suggesting the mildest, Braak I/II stage for neurofibrillary tangles, but the investigators were able to detect tau signals in higher level cortical regions characteristic of AD, as well.

In this small cohort, the investigators detected no correlation between baseline amyloid and the rate of its accumulation or episodic memory performance. Jagust thinks this may be because the sample size means they lack power to detect the subtle cognitive effects they picked up in the larger ADNI cohort. Nonetheless, the results suggest the tau effect is larger than the memory effect, which is consistent with the idea that tau deposition precedes memory issues, Jagust said.

The data help clarify relationships between Aβ accumulation and the AV-1451 tau signal, Victor Villemagne and Colin Masters of the University of Melbourne in Australia, wrote in an email to Alzforum. “These findings strongly support the current concept of AD pathogenesis, in which Aβ drives cortical tau, and that cognitive impairment develops only when both Aβ and tau reach critical levels,” they wrote. As in ADNI, Leal identified a subset of non-accumulators—people whose amyloid did not increase over the course of four years. “It would be extremely interesting to follow these non-accumulators over time in terms of tau deposition, together with markers of neuronal injury as well as cognition,” Villamagne and Masters wrote.

But Val Lowe, Mayo Clinic, Rochester, Minnesota, urges caution on this point. The association is interesting and calls for more longitudinal research on amyloid and tau imaging, he wrote to Alzforum. “It is important to emphasize that this doesn’t demonstrate causation of tau accumulation but only association or ‘prediction.’ That’s a very important distinction. We don’t know if these people had tau of the same amount when the first amyloid scans were performed … These answers await more data.”

The study highlights that while a binary system for scoring amyloid PET works in symptomatic people, in early disease it may be too simplistic, said Hansson. But, he pointed out, the recently updated NIA-AA research diagnostic framework (Apr 2018 news) allows for the possibility of an intermediate stages, or even continuous measures. “These two studies support the idea of stratifying the data into three different categories and also using continuous data when needed,” he told Alzforum.

In an editorial accompanying the Neurology paper, Corey McMillan of University of Pennsylvania, Philadelphia, and Gael Chételat of the University of Caen-Normandie in France consider the implications for clinical trials. “The Landau et al. report provides support for the possibility that Aβ accumulators, rather than Aβ-positive individuals, may be the optimal candidates for early stage Aβ-targeted clinical trials,” they wrote. That’s the idea behind the latest prevention trial, the Ante-Amyloid Prevention of AD, or A3, study headed by Reisa Sperling at Brigham and Women’s Hospital, in Boston (Dec 2017 conference news).

But how to pick those people? A3 will enroll some people who were turned away from other trials because their amyloid scans fell below the cutoff for positivity. However, investigators can’t yet predict who will turn out to be accumulators, outside of serial scans, a slow and expensive prospect. One goal is to identify additional biomarkers that identify accumulators based on a single scan.

Recent work suggests measuring regional accumulation of amyloid may help, said Michael Schöll, University of Gothenburg, Sweden. Schöll and Hansson recently showed that, in people in the earliest stages of amyloid deposition, pockets of the default mode network were among the first to accumulate amyloid and associated with weakening connectivity long before amyloidosis was widespread (Nov 2017 news). And Michel Grothe of the German Center for Neurodegenerative Diseases in Rostock has proposed a staging system based on spatial amyloid deposition (Oct 2017 news). “The subthreshold designation only refers to global amyloid, but if we look at where the signal is, these subjects are probably not going to be subthreshold anymore,” he said.—Pat McCaffrey


  1. The data from Leal et al. help to bring clarity to the relationships between Aβ accumulation and the AV-1451 tau signal. On a relatively small number of cognitively normal 75 years old subjects, followed over five years, those who were accumulating Aβ, even from sub threshold levels, could be seen to have developed an abnormal tau signal. These findings strongly support the current concept of AD pathogenesis in which Aβ drives cortical tau, and that cognitive impairment develops only when both Aβ and tau reach critical levels.

    One issue that might need to be addressed when comparing markers of pathologic change is to account for the disparity in the concentration of Aβ and tau aggregates, given that tracer retention reflects the density of binding sites, and the concentration of tau aggregates in the brain is about one order of magnitude lower than Aβ. Other issues to be considered are tracer stoichiometry, binding affinity, and non-specific binding, which are unlikely to be the same for different tracers. Perhaps, we should qualify the statements about tau (and Aβ) by adding “detectable”, because there might be tau deposits in the brain that while present, are too low to be detected by PET.

    We also noted that the analysis included both accumulators and non-accumulators. The advantage of having rates of Aβ accumulation is that they allow the separation of those subthreshold individuals who are on an AD trajectory from those who are not. It would have been preferable, in order to assess the true relationship between Aβ accumulation and tau burden, to examine Aβ accumulators and non-accumulators separately. That said, and as the authors acknowledge, even when non-accumulators were excluded from the analyses, it had little effect on the results. It would be extremely interesting to follow up these non- accumulators over time in terms of tau deposition, together with markers of neuronal injury as well as cognition.

    Overall, coupling the longitudinal with cross-sectional data are most informative, and adds to the push to develop therapeutic strategies that, to be successful, will need to aim at those subjects who have yet to cross the threshold.

  2. Evaluating the impact of various levels of amyloid or tau signal seen on PET imaging that may have some clinical associations is important for the field.  The determination of cutpoints or thresholds that define characteristics of certain groups, such as people with AD dementia and normals, are also important. It is not that one approach is better than the other. Different approaches have different applications.

    We can think of PET signal in terms of continuous change, to be most precise. At some point along the continuum, the level of protein may have an impact on a person. That level of “impact” could reasonably differ from one person to another. For example, amyloid and tau levels that cause symptomatic disease in one person may be distinct as compared to another person. These possibilities become complex and are important to consider as they may lessen our complete commitment to group level “cutoffs.” Having said that, cutoffs are what we live with in medicine, whether it is a blood pressure value or blood sugar measurement. There is a range and a cutoff that we all accept, while working within those guidelines to manage individuals who maybe don’t fit the mold perfectly.

    In these recent papers, the investigators from the Jagust lab provide an intriguing look at modest elevations of the amyloid PET signal. It is very interesting and believable that cognition is affected at levels of amyloid that are “subthreshold”. We have to remember however that many of these “thresholds” were designed to characterize normal people vs. people with AD dementia. There isn’t really an applicable threshold in terms of detecting very subtle memory change in relatively cognitively unimpaired people and they therefore describe measuring subtle increases in amyloid. Using the term “threshold,” communicates the idea that the levels of amyloid being tested are lower than those typically used to distinguish AD dementia patients.

    The researchers also demonstrate that subtle baseline elevation and increasing levels of amyloid are associated with tau signal at the end of the described amyloid imaging observation period.  It is important to emphasize that this doesn’t demonstrate causation of tau accumulation but only association…or “prediction.” A very important distinction. We don’t know if these people had elevated tau when the first amyloid scans were performed. The tau scan was not done at that time.  If tau was present first, maybe tau could have caused amyloid to increase. This can’t be determined with the data presented. Serial tau imaging in unison with serial amyloid imaging would be the best way to evaluate the interaction of amyloid and tau relative to causation. These answers await more data.  To this paper’s point, the association of the two is an important finding and provides for hypotheses generation for future work.

  3. These are fascinating studies. They remind us there is a lot we do not know about the biology of AD. They reinforce that we need to solve the mystery of how Aβ talks to tau. They also show that kinetics matter. There has been some literature (e.g. Wogulis et al, 2005) linking amyloid growth to toxicity in cultured neurons. Though I am not necessarily a fan of such artificial studies, it is worth considering that rate of accumulation might be more deterministic of toxicity than absolute amounts. This could further explain why overall amyloid levels in cross sectional studies tend to correlate poorly with symptoms. Perhaps if you are a slow accumulator you can reach high amyloid loads without symptoms whereas if you accumulate quickly, regardless of overall level, things are not so good. Enlisting these individuals for brain autopsy would be highly useful as that would provide even more information. Indeed, another possibility (speculative though it may be) is that the signal is low because there is efficient local clearance, which is actually causing bystander damage.


    . Nucleation-dependent polymerization is an essential component of amyloid-mediated neuronal cell death. J Neurosci. 2005 Feb 2;25(5):1071-80. PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Together, Aβ and Neurodegeneration Spell Cognitive Decline in Three Years
  2. At Risk, or Already Alzheimer’s? Elevated Aβ Predicts Cognitive Decline
  3. New Definition of Alzheimer’s Hinges on Biology, Not Symptoms
  4. 10th CTAD: Finally, Alzheimer’s Field Is Serious About Prevention Trials
  5. Daydreaming Network Serves as Ground Zero for Aβ Deposition
  6. PET Staging Charts Gradual Course of Amyloid Deposition in Alzheimer’s

Paper Citations

  1. . Amyloid deposition, hypometabolism, and longitudinal cognitive decline. Ann Neurol. 2012 Oct;72(4):578-86. PubMed.

Further Reading

No Available Further Reading

Primary Papers

  1. . Memory decline accompanies subthreshold amyloid accumulation. Neurology. 2018 Apr 24;90(17):e1452-e1460. Epub 2018 Mar 23 PubMed.
  2. . Amyloid "accumulators": The next generation of candidates for amyloid-targeted clinical trials?. Neurology. 2018 Apr 24;90(17):759-760. Epub 2018 Mar 23 PubMed.
  3. . Subthreshold Amyloid Predicts Tau Deposition in Aging. J Neurosci. 2018 May 9;38(19):4482-4489. Epub 2018 Apr 23 PubMed.