This study “contributes to the growing evidence that we may be able to find a plasma biomarker, which would be, from a population perspective, much more usable than cerebrospinal fluid (CSF) or brain imaging measures,” Nicole Schupf of Columbia University, New York, told ARF. Though the latter measures, at this point, offer more predictive and diagnostic value, spinal taps and brain scans are expensive and invasive, fueling interest in AD biomarkers that could be evaluated from a simple blood draw.

Thus far, the search for blood-based biomarkers can be characterized by sprinkles of promise awash with confusion. Plasma Aβ concentrations rise with age and are elevated in people with pathogenic mutations in presenilin or amyloid-β precursor protein (APP) (Ringman et al., 2008). However, data from studies of sporadic AD have been all over the map, with various groups reporting that disease risk associates with high plasma Aβ42 levels (e.g., Schupf et al., 2008 and ARF related news story) or with low Aβ1-42/Aβ1-40 ratios (e.g., van Oijen et al., 2006), or that correlations with plasma Aβ were too weak to be useful (e.g., Lopez et al., 2008; Hansson et al., 2008 and ARF related news story).

The current paper was an epidemiological tour de force involving 8,414 non-institutionalized elderly (ages 65 and up) in Bordeaux, Montpellier, and Dijon. Funded primarily by the French government as well as several non-government organizations and pharmaceutical companies, the Three-City (3C) Study was designed to determine the influence of vascular factors in dementia, Lambert said. Scientists collected baseline blood samples from participants randomly chosen from electoral rolls of each city between 1999 and 2001, and did follow-up assessments, including neuropsychological testing, two and four years later.

From the initial cohort, 257 developed incident dementia during the four-year follow-up period, and 1,185 were selected randomly as non-demented controls. Lambert and colleagues used a new xMAP-based assay developed by Innogenetics, Ghent, Belgium, to measure full-length and truncated Aβ peptides in the blood samples. Controlling for various factors including age, gender, education, diabetes, cholesterol, and APOE genotype, they found that people in the upper Aβ1-42/Aβ1-40 tertile had half the risk of developing AD and a 2.9-fold lower risk of mixed/vascular dementia, compared to those in the lowest tertile. However, the upper tertile for truncated Aβ, i.e., n-42/n-40, was only associated with reduced (3.7-fold) risk of mixed/vascular dementia but not AD.

“Maybe these truncated species are more involved in vascular amyloidosis than they are in forming plaques,” Schupf said. In a recent study led by Richard Mayeux at Columbia University, New York, Schupf and colleagues analyzed more than 1,100 seniors in northern Manhattan and found that those with high baseline levels of plasma Aβ42 were three times as likely to develop AD over four to six years of follow-up (Schupf et al., 2008).

At a glance, these findings seem to clash with the 3C Study, which found that high Aβ1-42/Aβ1-40 ratios correlated with lower AD risk. However, the New York study found that conversion to AD coincided with a drop in plasma Aβ42, which Schupf sees as “highly consistent” with the association Lambert and colleagues report at two years of follow-up. “You don't really know where those guys were five years ago. Two years is quite a short period of time. They may have been en route [to dementia/AD],” she told ARF. “It would be of interest to know how many of those who converted to AD had mild cognitive impairment (MCI) at their initial examination.”

On this point, Lambert noted that MCI was still a relatively new concept at the time the 3C Study was designed, making systematic diagnoses extremely challenging for large epidemiological studies. “However, we can postulate that some of the individuals who developed dementia at two or four years of follow-up likely presented MCI at inclusion,” he wrote in an email to ARF. In the future, baseline MCI status could likely be addressed in one of the three cities to make the sample size more manageable, he noted.

While large sample size is clearly a strength of the 3C Study, it can also cause problems for interpretation. “Usually more is better, but I wonder if it holds any meaning for individual diagnosis,” Anne Fagan of Washington University, St. Louis, Missouri, said of the new work. “If you study enough people, you'll find an association somewhere.” She suggested that timing may be critical in studies of plasma Aβ and dementia. “I found it curious that the associations were restricted to individuals diagnosed at two years of follow-up and not at four years,” Fagan told ARF. She notes previous work showing that CSF levels of the β-secretase enzyme BACE1 are elevated in MCI, relative to healthy people, but dip during conversion to dementia (see ARF related news story). “This plasma story could be the same thing,” Fagan said. “Maybe you have to catch it at a certain timepoint in the course of disease in order for it to be meaningful diagnostically and predictively.” Schupf agrees, stressing that plasma Aβ studies need to be longitudinal, with repeated peptide measures in elders who start off free of cognitive impairment. “If Aβ was high and then goes down with onset, it’s not going to look different from non-demented in a cross-sectional study,” she said.

Another problem with plasma Aβ studies is that variations in the techniques for measuring plasma Aβ make it hard to compare results from different labs. Such challenges were highlighted in a recent study showing highly variable recovery rates for Aβ40 and Aβ42 in spiked samples sent to various U.S. labs using different protocols to measure Aβ peptide concentrations (Okereke et al., 2009 and ARF related news story). The use of different assays may not necessarily cause problems, though. Lambert noted that the 3C findings essentially reproduced those of a similar, large Dutch study that did not use xMAP-based technology to measure plasma Aβ (van Oijen et al., 2006).

Barring methodological concerns, a fundamental problem with plasma Aβ studies is the uncertainty about what is being measured. “We don't know, in fact, whether the plasma Aβ peptides are coming directly from brain, from vascular cells, or from other organs in the body,” Lambert said. New data by Fagan and colleagues seem to suggest that plasma Aβ measurements may not reflect changes in brain Aβ as closely as CSF measures or positron emission tomography (PET) using the amyloid tracer Pittsburgh compound-B (PIB). Consistent with past work by her group (Fagan et al., 2006) and others, “everybody who was PIB-positive had low CSF Aβ42 levels, but the majority of PIB-negative people had high CSF Aβ42,” Fagan said of her new study, which is currently in press at EMBO Molecular Medicine. However, plasma Aβ concentrations did not seem to correlate with brain amyloid load assessed by PIB-PET, or with CSF Aβ, she said.

These data would seem to strengthen the case for relying on brain scans and CSF samples for AD biomarkers. “But when the day comes where we have treatments, we're going to want to have large-scale screening of predictors that allow us to intervene before people develop symptoms,” Schupf said. “On a population basis, that really is a blood test. I don't think we're there yet. I think this [study] helps to suggest that we might get there.”

Masood Kamali-Moghaddam of Uppsala University, Sweden, expects the new data to intensify efforts “to detect and identify other toxic isomers of Aβ, such as Aβ oligomers in blood.” (See full comment below.) Scientists described some attempts in this direction at the ICAD meeting in July (see ARF related conference story).—Esther Landhuis.

Reference:
Lambert JC, Schraen-Maschke S, Richard F, Fievet N, Rouaud O, Berr C, Dartigues JF, Tzourio C, Alpérovitch A, Buée L, Amouyel P. Association of plasma amyloid beta with risk of dementia: the prospective Three-City Study. Neurology. 2009 Sep 15;73(11):847-53. Abstract