Biomarkers Predict Future Cognitive Impairment

A blood test can predict whether an asymptomatic older adult will develop cognitive impairment or Alzheimer’s disease within two to three years, a study shows.

By Tracy Vence | March 9, 2014

WIKIMEDIA, CALLE EKLUND/V-WOLFA panel of 10 metabolites from peripheral blood could be used to predict future cognitive impairment in asymptomatic older adults, according to a study published today (March 9) in Nature Medicine. Georgetown University Medical Center’s Howard Federoff and his colleagues found that these biomarkers indicate whether an elderly person who showed no signs of cognitive problems would go on to develop either mild memory impairment or Alzheimer’s disease within two to three years, with greater than 90 percent accuracy. Their work adds to a growing body of literature implicating aberrant lipid metabolism in the pathophysiology of Alzheimer’s disease.

“This is the first time that blood-based lipidomics has been used to describe an at-risk population in a neurodegenerative disease area,” Federoff told The Scientist.

This study “highlights the enormous potential of lipid biomarkers and their ability to predict memory impairment at a preclinical stage of Alzheimer’s disease,” said Columbia University’s Gilbert Di Paolo, an associate professor of pathology and cell biology, who was not involved in the work. That these biomarkers can be detected in the blood is particularly promising, he added. “The scientific community working on Alzheimer’s disease has explored this body fluid from many different angles to try and find biomarkers predicting the onset of this disorder, unfortunately without much success.”

The team followed 525 healthy participants, aged 70 and older, for five years. During that time, 74 of the participants met the clinical criteria for amnestic mild cognitive impairment or Alzheimer’s disease. Forty-six of the 74 were incidental cases at the outset, but 28 developed symptoms over the course of the experiment. Federoff and his colleagues focused on whether there were any metabolites that seemed to distinguish this latter group, so-called “converters,” in their metabolomic biomarker discovery efforts. And indeed, the researchers found that certain amino acids and phospholipids—which play key roles in the integrity and functionality of cell membranes—were, as they put it, “potent discriminators” of converters versus non-converters.

In a validation cohort of 40 participants, the team confirmed that this 10-metabolite panel was predictive of developing amnestic mild cognitive impairment or Alzheimer’s disease.

“[This study] is very interesting and certainly deserves more follow-up work and attention,” said Lesley Jones, a professor of psychological medicine and clinical neuroscience at Cardiff University in the U.K., who was not involved in the work. “It’s pretty clear that lipids play a role in Alzheimer’s disease and susceptibility, and there has been some genetic evidence [of this] as well, but we don’t really understand—mechanistically—how this operates at the moment.”

Because apolipoprotein E (APOE) is involved in lipid metabolism and its ε4 allele a known genetic risk factor for Alzheimer’s disease, the team sought to test the effects of APOE genotype on their lipidomic classification of converters versus non-converters. But the effect of the APOEε4 allele on this test appeared insignificant. And judging by APOE status alone, the researchers were unable to distinguish converter versus non-converters any better than they might have by chance.

“The result was interesting, insofar as APOE does not change the predictive accuracy of this test,” said Federoff.

The study builds upon researchers’ recent interests in unraveling the biochemical pathways behind neurodegeneration, said Rima Kaddurah-Daouk, an associate professor of pharmacometabolomics at the Duke Institute for Brain Sciences in Durham, North Carolina.

“We have speculated that [changes in] the lipids that are important for proper [cell] membrane structure and communication could really be a critical step in triggering the cascade of events leading to cognitive impairment and Alzheimer’s disease,” she said. “All of these observations—along with this wonderful new study that goes earlier in the disease process—confirm that it’s time to come back to the biochemistry in a very serious way.”

Federoff’s team would now like to see whether the results hold up in younger adults and more ethnically diverse participants. In the meantime, the researchers are analyzing genomic information in an effort to connect any apparent differences in gene expression to changes in lipid metabolism. “We’ve sequenced everyone, and although we haven’t fully analyzed [that data], we’d like to take [the metabolomics information] back to the genome, to see whether there any new pieces of information that could be helpful that would only be understood at the genomic level.”

“The tools have become available for us in metabolomics and lipidomics over the last decade to dive deep into the metabolic understanding of Alzheimer’s disease,” said Kaddurah-Daouk. It’s increasingly clear, she added, that “genetics alone are not sufficient.”

M Mapstone et al., “Plasma phospholipids identify antecedent memory impairment in older adults,” Nature Medicine, doi:10.1038/nm.3466, 2014.

Add a Comment

Avatar of: You



Sign In with your LabX Media Group Passport to leave a comment

Not a member? Register Now!

LabX Media Group Passport Logo


Avatar of: Hugh-F-61


Posts: 75

March 10, 2014



This test is 90%  accurate. Whether it misses 10% of cases, or diagnoses 10% of normals as having a memory-impaired future isn't clear. Around 5% of the people in the study "converted". If the error rate is 10% false positives, then twice as many unaffected people will get positive diagnoses as affected people. Put another way, only one third of diagnoses of future impairment within 2-3 years will be correct.


Avatar of: rferguson


Posts: 1

March 10, 2014

Are you sure that it does not refer to the accuracy in determining the 5% of the cases (converters)? In that case, the range is 10% of the 28 positive cases out of 525 total. The 10% range of 28 is 2.8 (3 people). This does not track 1/3 correct diagnosis.

Let’s say I have a sample set of 100 people and I can predict 10 of them will have trait X with 90% accuracy. In other words, of the 10 I have selected, 10% could be incorrect. That leaves 1 person predicted incorrectly. Whether this is a false positive or negative is another subject. Not enough information is available, but this does not triple the error.

Avatar of: john5


Posts: 2

March 10, 2014

maybe this helps expliain the negative neurocognitive effects of statin drugs

Avatar of: JohnC


Posts: 13

March 11, 2014

This is another publication that illustrates that genomic screening giving correlations delivers relatively little. A few percentage points of genetic predispostion (association) is not clinically very useful. The doctor and the patient need a reliable test and that means studying the patients at the physiological, cellular and biochemical level. Charles Yanofsky was one of many who criticized the 'blind' statistical GWA genetics approach. Barbara McClintock told us to 'know your organism' - she used a microscope and found, forty years before anyone else believed her and long before gene cloning that genes could transpose. I vote with them. Use common sense and try to guide granting bodies rather than have your research guided by granting-area descriptors.

Avatar of: tvence


Posts: 1052

Replied to a comment from Hugh-F-61 made on March 10, 2014

March 12, 2014

Hi Hugh,

Thanks very much for reading and for your comment. Today we discuss statistical analyses and false positives over at The Nutshell.

Tracy Vence

News Editor, The Scientist


Popular Now

  1. Could Rapamycin Help Humans Live Longer?
  2. Renowned Physicist Stephen Hawking Dies
  3. John Sulston, Human Genome Project Leader, Dies
  4. Pupil Response to an Optical Illusion Tied to Autistic Traits