Researchers have set a threshold for Aβ in the cerebrospinal fluid of 192 pg/ml—anything less indicates Alzheimer’s disease pathology has begun. Now, a report suggests that people who are just above that cutoff are highly likely to dip below it within three years. The study, led by Michael Weiner of the University of California, San Francisco, and reported in the March 16 JAMA Neurology, indicates that moderately low Aβ42 is not a benign or random occurrence, but instead an early harbinger of AD. “Using a higher cutoff would likely increase the sensitivity for detecting very early stage amyloid pathology,” first author Niklas Mattsson, now at Lund University in Sweden, wrote to Alzforum. “Trials specifically focused at the first stages of amyloid pathology could possibility benefit from this.” Mattsson added, however, that the findings need to be confirmed in larger studies first.

Amyloid pathology in the brain can precede the onset of cognitive decline by decades. However, by the time amyloid plaques light up in a PET scan, it may be too late for therapies aimed at Aβ to work, as has been suggested by the failures of anti-Aβ trials. To help push back the start date of interventions, researchers have scoured the CSF for antecedent biomarkers. Previous studies using data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) determined that the 192pg/mL cutoff correlated strongly with pathologically confirmed AD, and later studies found that this threshold also correlated with positive amyloid PET scans, even for cognitively normal people (see Shaw et al., 2009; Aug 2010 news; and Weigand et al., 2011).

What about people with concentrations just above that cutoff? Were their moderately low concentrations of Aβ42 unrelated to amyloid pathology, or could they define a group on the brink of decline? To find out, Mattsson and colleagues measured CSF Aβ42, total tau, and phospho-tau at baseline and at least once more over a three-year period in 35 cognitively healthy ADNI enrollees. Volunteers were also screened for baseline hippocampal volume, cognition, and ApoE genotype. Eight people also received amyloid PET scans. At baseline, all of the volunteers had CSF Aβ42 concentrations above 192pg/mL, and those who were tested for amyloid PET were negative.

After three years, CSF Aβ42 dropped below the threshold in 11 people (designated as “decliners”), whereas 24 others maintained levels above the threshold (“non-decliners”). The researchers found that the only significant difference between the two groups was lower baseline levels of Aβ42 in decliners. Of 12 people with CSF Aβ42 baseline concentrations below 225pg/mL, 10 of them became decliners upon follow-up. On the other hand, 22 of the 23 people with higher baseline concentrations were non-decliners. This translated into an 83 percent positive and a 96 percent negative predictive value over the course of the study for a cutoff of 225 pg/ml. Alone, cognitive scores, hippocampal volume, ApoE status, and CSF levels of total or phospho-tau did not vary significantly between the two groups. Ultimately, the most powerful predictor of decline emerged when researchers considered p-tau, Aβ42, and ApoE status together.

None of the eight participants given follow-up PET scans tested positive for amyloid at the end of the study, despite the fact that six of them were decliners. This could support the idea that low Aβ42 levels could precede PiB-positivity, commented Stephen Salloway of Brown University in Providence, Rhode Island.

“The initial Aβ CSF cutoffs were based on autopsies of patients with dementia and may be too high to determine risk for AD in individuals in earlier stages of the disease process,” Salloway wrote.

Henrik Zetterberg of the University of Gothenburg in Sweden commented that it is crucial to consider the “gray zone” surrounding biomarker cutoff points such as CSF Aβ42, which are dictated in part by the analytical variations among different experiments. For example, in the current study, the researchers calculated that such variability was 6.5 percent, which translates into a 95 percent confidence interval of around 25pg/mL on either side of the 192pg/mL cut-point. Zetterberg pointed out that most of the people who eventually declined had baseline levels within that gray zone, so perhaps they should not have been labeled “amyloid negative” in the first place. He added that this study confirms that such numbers within the gray zone are unlikely to be benign. “When selecting individuals for prevention trials on the basis of a biomarker result, it’s very important to realize the cut point is a bit artificial, and it comes with an uncertainty,” he said.

Jonathan Schott of University College London agreed with Zetterberg, and added that adopting biomarker gray zones, rather than sharp cut-points, could be a useful strategy in study designs. “This may be both a more biologically plausible way of classifying individuals in the border zone between normal and abnormal, and, as this study demonstrates, a potential means of highlighting those at greatest risk of future decline,” he wrote.—Jessica Shugart


  1. These findings by Mattsson and colleagues are consistent with recent reports that cognitively normal elderly with PIB retention close to the cutoff at baseline were more likely to show higher rates of retention on follow-up scans than those with low baseline retention. The initial Aβ CSF cutoffs were based on autopsies of patients with dementia and may be too high to determine risk for AD in individuals in earlier stages of the disease process. Further, higher p-tau values predict decline in CSF Aβ, and the combination of low baseline Aβand high p-tau was the best predictor of decline. It is not clear if they tested the total tau/Aβ ratio or p-tau/Aβ ratio to predict decline as measured by Shaw et al. (Shaw et al., 2009). High baseline CSF tau and p-tau may be important predictors of clinical decline in Aβ-positive amnestic MCI (Buchhave et al., 2012). The findings from the Mattsson study have implications for developing risk profiles and determining eligibility for intervention in preclinical AD. Study limitations include a small sample size, low E4 carrier rate, and the term “decliner” probably should be changed to “who declines below cut-off,” as some individuals with high baseline Aβ also demonstrated decline in CSF values. Finally, it is interesting to note that six decliners and two non-decliners had amyloid PET during the follow-up period, and none were PET positive.

    View all comments by Stephen Salloway
  2. Whilst the numbers are relatively small, this is a valuable study providing a rare opportunity to study the CSF profiles of healthy elderly individuals as they change over time. The principal finding—that individuals with initially low-normal CSF Aβ42 were most likely to decline to “pathological” levels of CSF Aβ42—is of considerable interest, suggesting a potential strategy for enriching clinical trials aiming to recruit amyloid accumulators. It will be of particular interest to follow these individuals to see if, as expected, they develop cognitive impairment; and if so, over what time frame. This study also highlights the potential value of adopting biomarker “gray zones” either in place of, or as well as, a single cutoff point for CSF Aβ42, or other biomarkers. This may be both a more biological plausible way of classifying individuals in the border zone between normal and abnormal, and, as this study demonstrates, a potential means of highlighting those at greatest risk of future decline.

    View all comments by Jonathan Schott

Make a Comment

To make a comment you must login or register.


News Citations

  1. Triple Confirmation: AD Footprint in CSF of Cognitively Normal People

Paper Citations

  1. . Cerebrospinal fluid biomarker signature in Alzheimer's disease neuroimaging initiative subjects. Ann Neurol. 2009 Apr;65(4):403-13. PubMed.
  2. . Transforming cerebrospinal fluid Aβ42 measures into calculated Pittsburgh Compound B units of brain Aβ amyloid. Alzheimers Dement. 2011 Mar;7(2):133-41. PubMed.

Further Reading


  1. . Alzheimer's disease - Recent biomarker developments in relation to updated diagnostic criteria. Clin Chim Acta. 2015 Feb 7; PubMed.
  2. . Determining cut-points for Alzheimer's disease biomarkers: statistical issues, methods and challenges. Biomark Med. 2012 Aug;6(4):391-400. PubMed.

Primary Papers

  1. . Predicting Reduction of Cerebrospinal Fluid β-Amyloid 42 in Cognitively Healthy Controls. JAMA Neurol. 2015 May;72(5):554-60. PubMed.