Skip to main content
Premium Trial:

Request an Annual Quote

Duke Researchers Present New Data on APOE/TOMM40 Algorithm for Alzheimer's MCI Risk


NEW YORK (GenomeWeb) – APOE-TOMM40 '523 haplotypes are more informative than just APOE genotypes in predicting the age when healthy people are likely to start experiencing mild cognitive impairment (MCI) due to Alzheimer's disease, a group of researchers from Duke and Johns Hopkins University reported at a major medical conference this week.

The analysis, led by JHU's Marilyn Albert, is in line with the earlier discovery by Allen Roses' team at Duke that different lengths of the TOMM40 rs10524523 variant, combined with APOE genotypes and a person's age, can predict whether a cognitively normal person is at risk for AD-related dementia.

Although the Albert group's study examined a relatively small cohort of around 50 Caucasian subjects with MCI or dementia, its agreement with the Roses findings is important in light of the fact that two large, published meta-analyses did not find an association between TOMM40 '523 and age of onset for MCI.

Roses has criticized these two earlier attempts to measure TOMM40 '523 variants and replicate his team's work as having flawed technical methods and retrospective assessments of age of MCI onset. Meantime, Albert's group prospectively followed participants in the Biomarkers for Older Controls at Risk for Dementia (BIOCARD) study, which between 1995 and 2005 enrolled more than 250 middle-aged healthy volunteers with a family history of Alzheimer's disease. Researchers have performed a battery of neuro-psychological tests on these volunteers and have been following them for 19 years.

"The long duration of follow-up … of the BIOCARD cohort is ideal for investigating the genetic effects of APOE and TOMM40 '523 on the age of onset of MCI and dementia," Albert and colleagues wrote in a poster presented at the Alzheimer's Association International Conference in Washington, DC, this week. "Duration of follow-up is a critical factor in determining age-specific genetic effects of these variants."

In the latest analysis, Albert's group identified 54 Caucasians in the BIOCARD cohort who were diagnosed with MCI or dementia after 64 years of age and who had MCI for less than four years, and compared them to 122 cognitively normal people. They then compared these two groups' survival in terms of APOE and TOMM40 genotypes.

In the early 90s, Roses' team at Duke discovered that people carrying the APOE4 allele were at greater risk for developing late-onset Alzheimer's earlier than non-carriers. Subsequently, Roses' research has suggested that different lengths of the TOMM40 '523 variant adds additional prognostic information on top of people's APOE genotypes with regard to their risk for developing Alzheimer's-related cognitive decline — and the latest work by Albert et al. corroborates this hypothesis.

The APOE/TOMM40 algorithm developed by Roses' team factors in people's age, the different copies of the APOE gene they have inherited — APOE2, APOE3 or APOE4 — as well as whether they have short (S), long (L), or very long (VL) copies of the TOMM40 '523 variant.

According to Albert's group, earlier onset of MCI, dementia, or Alzheimer's was significantly impacted by the APOE 4/4 genotype inherited with the TOMM40 '523 L/L genotype; or the APOE 3/4 genotype with the TOMM40 '523 L/VL or S/L genotype. In fact, among study subjects with the APOE 3/4 genotype, those with TOMM40 '523 L/VL on average developed MCI or dementia a decade earlier than those with TOMM40 '523 S/L.

Moreover, in the BIOCARD cohort, subjects with the APOE 3/3 genotype and TOMM40 '523 S/S or VL/VL versions "had a markedly later age of [MCI] onset distribution" than those with the S/VL genotype. Based on this analysis, Albert and colleagues concluded that APOE-TOMM40 '523 haplotypes as a genetic predictor are "potentially more informative than using the APOE genotypes alone for quantitative phenotypes including age of onset of MCI/dementia and duration of MCI."

Since BIOCARD subjects were followed for such a long time, researchers were also able to identify subjects who had "stable MCI" and hadn't gotten dementia for more than four years. In this subset, most subjects had the TOMM40 '523 VL/VL variant, while fewer exhibited TOMM40 '523 S/S. This cohort of "stable MCI"' subjects "may provide additional insight into the heterogeneity of cognitive decline," Albert and colleagues stated in their poster.

Roses told GenomeWeb that the TOMORROW trial, a global, Phase III study that his startup Zinfandel Pharmaceuticals is conducting with Takeda Pharmaceutical, has also identified MCI patients with the TOMM40 '523 VL/VL variant who are older but seem to stabilize without progressing to Alzheimer's. "The VL/VL '523 haplotype group will be confirmed in the TOMMORROW clinical trial in which they are considered as low risk with the APOE2/3 haplotypes," Roses said.

Comparing biomarkers

In another study, researchers from Duke and two of Roses' wine-themed startups (Zinfandel and Cabernet Pharmaceuticals) gauged how well the APOE/TOMM40 algorithm assessed the risk of MCI due to Alzheimer's compared to neurocognitive tests, cerebrospinal fluid (CSF) levels of amyloid-beta 42 and phosphorylated tau, and beta-amyloid PET imaging.

Led by Roses and Duke School of Medicine's Michael Lutz, researchers used data from the Alzheimer's Disease Neuroimaging Initiative cohort. They showed that when the APOE/TOMM40 algorithm separated healthy controls and patients diagnosed with MCI or Alzheimer's in to high- and low-risk categories, there were statistically significant differences between these groups in their performance of neurocognitive tests, CSF levels, and imaging markers.

Importantly, Lutz's group was able to demonstrate that the risk stratifications by the APOE/TOMM40 algorithm went in the same direction as these other methods of assessing MCI onset. "By examining the correlation between the risk assessment provided by the [genetic algorithm] and these biomarkers, we are showing that the predictions of the algorithm are supported by the gold-standard markers," Lutz told GenomeWeb.

Although, CSF and neuroimaging markers have been well studied in patients with MCI or Alzheimer's disease, the US Food and Drug Administration hasn't yet qualified them as markers that researchers can use in studies to predict disease progression. The agency in February issued a letter supporting the Critical Path Institute Coalition Against Major Diseases' plan to study CSF measures of a-beta 42, tau, and phosphorylated tau as potential markers for enriching Alzheimer's trials. "Greater experience with the use of these exploratory biomarkers in clinical trials is needed to more accurately determine their clinical utility for prognostic enrichment, impacting drug development decisions, and study design considerations," the FDA stated in the letter.

Based on FDA's 2013 guidance on developing drugs for early-stage Alzheimer's, researchers have been employing CSF markers and neuroimaging in clinical trials to identify patients with mild cognitive decline (before clinical symptoms of Alzheimer's set in), so treatments have the best chance to stave off the mind-robbing disease. However, Lutz pointed out some downsides to these approaches.

For example, analyzing CSF markers requires invasive lumbar punctures, is impacted by lab variability, and currently lacks international reference standards. Meanwhile, neuroimaging by PET or MRI is expensive, requires specially trained personnel, and the field hasn't agreed upon clinically meaningful measurement thresholds.

Blood-based biomarkers — such as APOE and TOMM40 variants as detected by a genetic test — "have the potential to be easier [and] more economical, and platforms to test these are widely available at medical facilities around the world," Lutz said.

Separately at AAIC, Albert's group from JHU presented data from BIOCARD showing that neurocognitive tests, CSF levels of a-beta and phosphrylated tau, and MRI scans to characterize the thickness of the right entorhinal cortex and the volume of the hippocampus were useful in predicting which cognitively normal subjects developed MCI in five years. "Our study shows that — up to five years before any Alzheimer's symptoms appear — a small set of factors can tell us, with significant accuracy, which cognitively normal individuals will develop MCI due to Alzheimer's," Albert said in a statement.

A few years ago, her team had published data showing that CSF markers can assess MCI onset. "We hope that this information will be useful for designing clinical trials aimed at delaying the onset of symptoms among cognitively normal individuals," Albert added. "An approach such as ours could be used for determining which people might be most likely to benefit."

Although Lutz's group was able to show correlation between the APOE/TOMM40 algorithm and some of these other markers, the field won't know if the genetic test is better until a larger trial qualifies the algorithm's prognostic ability. The APOE/TOMM40 algorithm will be qualified as a prognostic biomarker at the end of the Phase III TOMMORROW trial that is currently going on at multiple sites around the world. "Once qualified, the biomarker can be used as a companion pharmacogenetic test for a therapeutic to delay the onset of Alzheimer's," Lutz said.


In the TOMMORROW trial, Zinfandel and Takeda are seeking to qualify the APOE/TOMM40 algorithm as a test to identify cognitively normal subjects between 65 and 83 years who are at risk for MCI due to Alzheimer's in five years. When the first 410 subjects convert to diagnosed MCI, researchers will not only qualify this algorithm but also determine whether a very low dose of pioglitazone (0.8 mg sustained release) is effective in delaying MCI in patients that the APOE/TOMM40 test deems to be at high risk for the phenotype.

In the summer of 2013, researchers enrolled the first subject into the study, which is expected to go on for four years or until 410 events have occurred. According to an update presented at AAIC this week, researchers screened more than 21,400 individuals, assayed around 19,000 samples, and enrolled approximately 2,600 subjects into TOMMORROW as of mid-2015.

According to Roses, investigators enrolled the majority of participants (more than 2,000 subjects) just in the first half of this year. "To date, after randomization, subject retention within TOMMORROW has met expectations," researchers led by Roses noted in a poster. So far, around 9 percent of randomized study volunteers have terminated from the study early.

TOMORROW has launched at multiple sites in the US, UK, and Australia, and sites in Germany and Switzerland are recruiting volunteers. In order to enroll the sufficient number of subjects, researchers estimated they will need 60 sites each recruiting approximately 100 cognitively normal subjects, who will be followed every six months to track whether they develop MCI.

As an extension to TOMMORROW, researchers will also conduct a long-term evaluation of the safety of pioglitazone and follow those who develop MCI to further study their cognitive decline for an additional two years.