Skip to main content
Premium Trial:

Request an Annual Quote

Takeda, Zinfandel Evaluate Genetic Risk Algorithm Ahead of Alzheimer's Delay Study with Diabetes Rx


Originally published July 23.

In preparing to launch a phase III study to investigate whether low doses of a diabetes drug will delay the onset of Alzheimer's disease, researchers from Takeda, Duke University, and Zinfandel Pharmaceuticals have established the performance characteristics of a genetic risk algorithm that they will use to stratify participants in the trial.

At the Alzheimer's Association International Conference in Boston last week, researchers presented data establishing the analytical validity of an algorithm that factors an individuals' age, APOE genotype, and TOMM40 polymorphisms to gauge whether the person is at high or low risk of mild cognitive decline due to Alzheimer's. According to data from a simulation study, the risk algorithm has a positive predictive value and a negative predictive value between 70 percent and 80 percent. Furthermore, the genomic risk algorithm "compares favorably" with imaging-based assays and cerebrospinal fluid biomarkers in determining whether individuals have cognitive decline due to Alzheimer's, researchers led by Michael Lutz of Duke University reported in a poster presented at AAIC.

The algorithm, developed by neuroscientist Allen Roses and his research team, will be used to stratify people in a Phase III trial that will assess the ability of the diabetes drug pioglitazone – marketed by Takeda for type 2 diabetes under the brand name Actos – to delay Alzheimer's-associated cognitive decline. In the same trial, researchers are hoping to qualify the biomarkers' ability to determine the risk of Alzheimer's onset in cognitively normal patients between the ages of 65 and 83, as well as to validate the algorithm for use as a blood-based commercial diagnostic test.

This trial is slated to start enrolling participants soon. Nearly 6,000 normal subjects will enter a high-risk or a low-risk arm based on the genetics-based algorithm. In the high-risk arm, study participants will be randomized to receive either low-dose pioglitazone or placebo. In the low-risk arm, study subjects will only receive placebo. The two placebo arms are intended to qualify the algorithm and validate the companion test; the two high-risk cohorts, randomized to either pioglitazone to placebo, will evaluate the safety and efficacy of the drug.

Study subjects will receive treatment for four years, researchers estimate. The phase III trial will end once 410 cognitively normal participants in the high-risk arm develop impaired cognition due to Alzheimer's. After this many events, the study will have 90 percent power to detect a difference between pioglitazone and placebo arms and measure the ability of the drug to delay Alzheimer's onset with a significance level of less than 0.01.

According to the World Alzheimer's Report, 36 million people were living with dementia worldwide in 2010 and the incidence could balloon to 115 million by 2050. The rate of Alzheimer's occurrence is doubling every five years for people between 65 and 85 years old. The Alzheimer's Association estimates that a drug that could delay the onset of the disease would reduce the cost of patient care by $40 billion in 2020.

The company that brings to market a safe drug that substantially delays the symptoms of Alzheimer's will not only ease a growing disease burden, but will also hit on a blockbuster. In the phase III trial, researchers are using a low-dose formulation of Takeda's once highly profitable type 2 diabetes drug, which in recent years has made headlines for increasing risk of heart failure and bladder cancer. However, in the Alzheimer's setting, researchers will administer a 0.8 mg/day formulation of pioglitazone – significantly lower than the up to 45 mg/day Actos dose diabetes patients receive.

The patent for pioglitazone has expired, but because the drug will be given with the help of a diagnostic that may prove to be informative for the majority of the at-risk Alzheimer's population, Takeda still stands to garner blockbuster-sized profits if it can launch the drug at a branded-generic price as a treatment for delaying disease onset.

The genomic risk algorithm to be used in this phase III study is the culmination of more than 20 years of Alzheimer's genetics research conducted by Roses, Jefferson-Pilot professor of neurobiology and neurology at Duke and CEO of Zinfandel, a company focused on using pharmacogenetics to improve Alzheimer's risk prediction and treatments. In 1992, Roses' lab at Duke identified the association between the APOE ε4 allele and heightened risk of late-onset Alzheimer's. More recently, Roses' team has published research showing that varying lengths of a TOMM40 poly-T polymorphism — located at intron 6 of the TOMM40 gene and linked to APOE ε3 and APOE ε4 polymorphisms — can be used to craft a three-allele risk prediction system for gauging the age of onset for Alzheimer's (PGx Reporter 1/12/2011).

It has been estimated that approximately 30 percent of Alzheimer's patients are APOE4 carriers. Past studies have shown that depending on age, a person who carries two copies of the APOE4 allele has a higher risk of late-onset Alzheimer's than a non-carrier. With the addition of the TOMM40 genotype to APOE4 genotype and age, researchers at Takeda and Zinfandel are hoping to show that the biomarker risk algorithm will be applicable to a much larger swath of the population and be more accurate in determining the risk for Alzheimer's-linked cognitive decline.

"Our biomarker algorithm provides a genotype for all individuals and the set of age-of-onset distribution covers 97 percent of the population," Roses said. "The APOE4/4 genotype [alone] is informative for 2 percent of the population, but now there are TOMM40 age-of-onset curves that account for all APOE genotypes except for individuals with two exceedingly rare APOE genotypes, APOE2/2 and APOE2/4, representing less than 3 percent of the population." To date, researchers haven't been able to collect enough Alzheimer's cases with these genotypes to determine at what age they are likely to develop Alzheimer's. Roses questioned whether any verified Alzheimer's cases exist for APOE2/2 genotypes.

“We have an algorithm that is highly informative for gauging risk in the population … at the same accuracy as one would get using more expensive and, as yet, not technically validated, cerebrospinal fluid markers or extremely expensive imaging biomarkers," Roses said.

Practical biomarker

The development of treatments to delay Alzheimer's symptoms has been dogged by limitations of available tools that can accurately identify who has the disease. Autopsies have revealed that as many as 25 percent of study participants enrolled in clinical trials evaluating treatments for mild to moderate Alzheimer's and thought to have the disease as defined by amyloid-related biomarkers don't actually have it.

There is currently no test for determining with certainty whether a person has Alzheimer's. Doctors may issue a battery of neurocognitive tests and brain scans to try to hone in on whether a person's dementia is due to Alzheimer's, but ultimately confirmation only comes after death, by evaluating the histopathological features and plaque deposits in brain tissue from autopsies.

In an attempt to improve diagnosis of the disease, most drug developers have embraced the so-called "amyloid hypothesis," which holds that the accumulation of the protein amyloid beta causes Alzheimer's disease. Some drug firms have tried to develop treatments that bind to amyloid plaque and flush it out. Other companies are working to develop agents that block an enzyme involved in producing A-beta and stop its buildup in the brain. Within these development programs, drug makers have tried to identify more accurately those with cognitive decline who are at risk for Alzheimer's by testing for APOE4 genotypes; by gauging protein biomarker levels in cerebrospinal fluid, such as amyloid-beta42 and total tau protein; and by assessing amyloid markers via imaging tests.

However, regulatory bodies have been cautious about recommending the use of such markers to assess Alzheimer's progression and determine treatment outcomes in clinical trials, particularly when involving cognitively normal study subjects who may not progress to dementia.

In its draft guidance on developing drugs for early-stage Alzheimer's, issued in February, the US Food and Drug Administration said it supports the use of biomarkers to enrich trials with individuals most likely to progress to more "overt" dementia. "However, the need for an assessment of sensitivity and specificity in identifying patients who do have actual Alzheimer's in clinical trials … does not allow the FDA to formally endorse any specific diagnostic frameworks at this time," the agency wrote.

Roses expects that if the phase III study is successful in qualifying the genetic algorithm's use to determine Alzheimer's-related cognitive decline risk, then that will also provide a template for others looking to validate various biomarkers for predicting Alzheimer's onset.

The FDA further expressed in its draft guidance that for approval, drugs must be shown to improve clinical outcomes, such as cognition and functioning, in study participants. Although drug makers are eager to apply surrogate biomarkers as measures of Alzheimer's treatment efficacy, there is no reliable evidence backing the use of such a biomarker, the agency said. "Until there is widespread evidence-based agreement in the research community that an effect on a particular biomarker is reasonably likely to predict clinical benefit, we will not be in a position to consider an approval based on the use of a biomarker as a surrogate outcome measure in Alzheimer's disease," the guidance states.

Takeda's and Zinfandel's phase III trial will determine the efficacy of pioglitazone based on its impact on clinical outcomes – by measuring the drug's ability to delay the onset of mild cognitive decline due to Alzheimer's in those at high risk of developing impaired cognition in a five-year span. Furthermore, there are extensive processes in place for determining whether a study participant has progressed from being cognitively normal to being cognitively impaired as a result of Alzheimer's.

During the course of the study, an investigator at each site will track participants' cognitive function during regular clinic visits. If, based on various functional tests during such a visit and after an extensive medical examination, the investigator deems someone to have impaired cognition, the subject's data will be forwarded to an adjudication committee for a formal determination. The participant will continue on the study, and be evaluated again by the investigator and adjudication committee after six months.

If the subject is twice found to have abnormal cognition in a six-month period, then that individual will count toward the 410 endpoint events needed to end the trial. Patients deemed to have progressed to mild cognitive decline due to Alzheimer's will have the option to join an extension study.

Because Takeda and Zinfandel aren't using imaging or cerebrospinal fluid biomarkers to track Alzheimer's onset in the phase III trial, they made sure to show that their biomarker strategy was as good as these other markers, which are gaining broader acceptance in the field, despite their limitations.

For example, the FDA last year approved Eli Lilly's PET imaging agent Amyvid to detect beta amyloid plaque in the brains of individuals with signs of cognitive impairment. If the diagnostic reveals moderate to frequent amyloid plaque, it could mean the patient has Alzheimer's disease, but it could also be indicative of other neurologic conditions, the agency said.

Although the FDA hasn't yet qualified any specific cerebrospinal fluid markers, the situation is somewhat different in Europe. Bristol-Myers Squibb prevailed last year in getting the European Medicines Agency to qualify the use of Aβ1-42 and T-tau in cerebrospinal fluid and amyloid-related PET signals to identify patients with mild to moderate Alzheimer's. BMS is evaluating the effect of the investigational agent BMS-241027 on these biomarkers in those with mild Alzheimer's disease.

The EMA decision "created a real difference between US and European regulators" with regard to the use of these biomarkers, Roses said.

In the study presented last week at AAIC, researchers from Takeda, Zinfandel, and Duke tested their genetic risk algorithm using patient data from the Alzheimer’s Disease Neuroimaging Initiative, which the EMA considered in qualifying the cerebrospinal fluid and imaging biomarkers. ADNI is an effort through which researchers are investigating the progression of Alzheimer's using a variety of biomarkers gauged through imaging, proteomic, and genomic tools.

“What we've shown is that our risk algorithm is comparable to anything that they've shown," Roses said. "The reason we can't be more exact than be comparable is because the data documenting age of onset that we have determined prospectively in our study wasn't well established" in the ADNI study.

The Alzheimer's Association considers biomarkers assessed using MRI and PET imaging and in cerebrospinal fluid as the "strongest biomarker candidates for Alzheimer's disease." Still, at least with regard to cerebrospinal fluid markers, Alzheimer's researchers generally agree that there is substantial variation in gauging these markers across labs, as well as differences in preclinical procedures, such as lumbar punctures. Recognizing that this variability complicates the use of such markers in multi-center studies, the Alzheimer's Association launched a quality control program in 2009 to standardize their measurement.

Some other drawbacks are cost and the invasive nature of some of these diagnostic procedures. In order for researchers to investigate biomarkers in cerebrospinal fluid, study participants must endure spinal taps. The use of PET or MRI scans in the research setting costs several hundred dollars, and has a price tag of several thousands of dollars outside the context of a clinical trial.

An algorithm that incorporates data from relatively simple blood-based genetic tests may be more practical than these other biomarker tests, Roses believes. For preclinical studies that Takeda, Zinfandel, and Duke researchers used to develop the biomarker risk algorithm, genetic testing services firm Polymorphic DNA Technologies performed APOE and TOMM40 genotyping for around $30 per subject.

Quest Diagnostics will perform APOE and TOMM40 genotyping for the phase III trial in which Takeda and Zinfandel will be evaluating pioglitazone and the risk algorithm. If, after the phase III trial, Takeda and Zinfandel are able to launch a commercial diagnostic for gauging risk of cognitive decline due to Alzheimer's, it will likely cost more than $30. The APOE genotyping test sold by Quest subsidiary Athena Diagnostics has a list price of $475, while complex algorithm-based molecular diagnostics can cost several thousands of dollars.

However, Roses believes that the price of a commercial test in this instance will not be prohibitive. "We want to see that [people] are tested before they get the drug," Roses emphasized. "In order to do that, it probably doesn't make sense to jack up the price of the test. So, what we want to do is make it a very reasonably priced test for clinical use."

A losing race

The race to develop treatments that delay Alzheimer's by targeting beta-amyloid have mostly yielded a handful of disappointing candidates that collapse well before the finish line.

Johnson & Johnson and Pfizer's bapineuzumab, a monoclonal antibody designed to bind to and clear beta amyloid, failed late-stage clinical trials last year. The drug was being studied as a treatment for mild to moderate Alzheimer's disease in APOE4 carriers. The companies halted research on the drug based on the study results.

Then in May, Baxter International's Gammagard yielded lackluster results in a late-stage clinical trial. Gammagard was made using antibodies that researchers had hoped would be able to remove amyloid plaque from the brain. A small early study had suggested the drug might work for those with moderate Alzheimer's who were carriers of APOE4. But when the drug failed to significantly impact cognitive decline in patients enrolled in a larger study, Baxter discontinued development of the drug.

These high-profile failures don't appear to have deterred the pharma industry from targeting beta-amyloid, however. Eli Lilly, for example, has already stumbled with solanezumab, designed to bind to the beta-amyloid protein and remove it from the brain before it forms amyloid plaque. Even though the drug failed to slow the rate of cognitive decline in two separate studies, Lilly recently announced it is moving ahead with a 2,000-patient trial investigating solanezumab in those with mild cognitive decline with evidence of amyloid plaque in their brains. In its latest attempt, Lilly is hoping to better define the population demonstrating symptoms of Alzheimer's with the help of its Amyvid PET imaging agent.

Other companies are approaching the amyloid hypothesis by looking at cerebrospinal fluid biomarkers. For example, Roche is conducting a 770-patient study, called Scarlet Road, in which researchers will measure Tau/Aβ levels in study participants' spinal fluid to identify early onset or prodromal Alzheimer's patients and treat them with gantenerumab. Roche is developing a companion test to gauge Tau/Aβ levels in trial participants (PGx Reporter 9/12/2012).

Merck, meanwhile, is developing a drug that inhibits an enzyme involved in the production of the beta amyloid peptide and is enriching clinical trials with people at high risk of Alzheimer's, based on whether they have mild cognitive impairment and high levels of Aβ42 and t-tau in their cerebrospinal fluid. Merck and Luminex are collaborating on a companion test that gauges these protein markers in drug development trials (PGx Reporter 3/20/2013).

Several prominent neuroscientists believe that drugs targeting beta amyloid haven't panned out because these agents cleared the brain of the protein too indiscriminately. Research has shown that the brain produces amyloid for a variety of purposes, not just in the case of Alzheimer's, so completely eliminating the protein from the brain could cause harm.

Keeping this in mind, Roche subsidiary Genentech has set out to prove that A-beta causes Alzheimer's by studying the residents of a remote Colombian region who share a common ancestor. Many in the town of Medellin and surrounding regions also carry mutations in the presenelin 1 gene, which impairs the ability of certain organelles in brain cells to digest unwanted proteins. People with this mutation are guaranteed to get Alzheimer's; they start to lose their memory in their mid-40s and their cognitive functions deteriorate by age 50.

Genentech is studying crenezumab, a drug that targets A-beta, in a study involving 300 patients from this region. By studying the drug in a genetically homogenous population, the company is hoping to show that a drug targeting amyloid beta can indeed halt the symptoms of Alzheimer's (PGx Reporter 5/23/2012).

Minority view

Roses has refused to join the majority of the research community and drug development industry in betting on amyloid plaque, arguing that buildup of amyloid plaque is part of the pathogenesis of Alzheimer's but doesn’t cause the disease. As such, he has maintained that beta amyloid is not the right drug target.

Based on his work, Roses believes that certain APOE and TOMM40 genotypes have increased damaging effects on mitochondrial function within brain cells, which limits their ability to utilize oxygen and glucose. As mitochondrial energy production fails, amyloid and other aggregating proteins accumulate in the brain. Consequently, damaged mitochondria can't reach distant synapses of neurons and provide the energy needed to support new neurite formation. This results in the kind of cognitive decline characteristic in Alzheimer's.

As Roses has diverged from the amyloid hypothesis, many in the research community likewise doubt his work describing the influence of TOMM40 on the age of Alzheimer's onset and disease risk. Since TOMM40 is located in the same region as APOE, several scientists have asserted that the disease risk and age-of-onset associations reported by Roses and his team are really due to the influence of APOE genotypes.

"TOMM40 and APOE are in complete linkage disequilibrium. Functionally, only one of the genes matters for Alzheimer's risk, not both," said Rudolph Tanzi, neurology professor at Harvard Medical School. Tanzi's lab was involved in discovering several genetic markers linked to Alzheimer's, including presenilin 1, presenilin 2, and amyloid precursor protein. Those who inherit mutations in these genes are sure to get Alzheimer's, but these markers only account for between 1 percent and 2 percent of cases.

A report in the Archives of Neurology last year by Boston University's Gyungah Jun and others analyzed more than 20,000 cases and controls from multiple genome-wide association studies and found that after adjusting for APOE status, the TOMM40 poly-T polymorphism was not significantly associated with disease risk or age of onset. In this analysis TOMM40 genotypes were determined retrospectively for around 3,000 people.

In reviewing this study in the same journal, University College London's Rita Guerreiro and John Hardy wrote that "it is very difficult" based on the analysis by Jun et al. "to attribute a genetic and APOE-independent role of TOMM40 in the risk and age of onset of Alzheimer's disease development."

Another study, published in the Archives of Neurology in 2011 by Carlos Cruchaga and colleagues, found a significant association between very long poly-T repeats in those with two copies of the APOE3 allele and very late onset of Alzheimer's – something also seen in Roses' own research. However, Cruchaga and colleagues failed to find any effect of TOMM40 variants on age of disease onset (PGx Reporter 1/12/2011).

"The overwhelming data indicate that the relevant gene of the two is APOE, not TOMM40, based on strength of genetic signals … across studies," Harvard's Tanzi said. "So, I see no relevance for TOMM40 as a functional marker with the pioglitazone trial results. Only APOE should be necessary as a genetic risk factor for this [Phase III] trial, as it is for other clinical trials in Alzheimer's."

The trouble with these retrospective analyses, according to Roses, is the manner in which age of onset has been determined, since age is a critical part of the genetic algorithm developed by his group. In an email to PGx Reporter, Boston University's Lindsay Farrer appeared confident that his team's analysis of data from more than 20,000 subjects was accurate. Farrer felt it sufficient that several of the original studies included in the analysis prospectively followed subjects to gather age-at-onset information. "Our evidence based on an extremely large sample is fairly conclusive," said Farrer, a lead author on the paper published by Jun et al.

According to Roses, another consideration in replicating his TOMM40 findings is the specific genetic test used to gauge APOE and TOMM40 genotypes. In a paper published in Alzheimer's and Dementia earlier this year, Roses and colleagues highlighted the need to use standardized assessment tools across studies when determining biomarkers, age of Alzheimer's onset, and disease symptoms.

In the meta-analysis by Jun et al, for example, Farrer said that TOMM40 genotype was determined using the same test that Roses' team used. However, the researchers incorporated the APOE status of participants as established by different tests in the original studies.

"APOE and TOMM40 are adjacent genes located on the same region of linkage disequilibrium – they travel together over evolution with all their accumulated mutations," Roses said. "The genetic effect of APOE and TOMM40 is not measured accurately by genome-wide SNP screening data because the mutations that define APOE are not even contained on … the popular screening platforms."

Russell Swerdlow, director of the University of Kansas Alzheimer's Disease Center, is in the process of replicating Roses' work around TOMM40 and has had initial success in a small cohort. Swerdlow and his colleagues are prospectively establishing the age of onset of Alzheimer's-linked cognitive problems in study participants in the manner described by Roses' team and have used the same lab to gauge APOE and TOMM40 status. The researchers plan to enroll 400 people in their study, so it may take some time to complete the analysis and publish the results.

In the data generated to date in approximately 150 subjects, "we find, as Dr. Roses has reported, that certain TOMM40 alleles predict the onset of cognitive decline at younger ages than certain other TOMM40 alleles," Swerdlow said. He acknowledged that given the close proximity of TOMM40 and APOE, it is "tricky" for researchers to unravel whether APOE is driving the association seen with Alzheimer's risk, if TOMM40 is the culprit, or if both are responsible.

"If you approach it from the position that APOE is it, and all you're looking for is to see whether TOMM40 is adding something on top of APOE, you may end up restricting your [study] population quite a bit," Swerdlow said. "For example, if you're only looking for people who have the long TOMM40 allele and an APOE3 backbone … then things start getting very confusing."

To counter those who would attribute the impact of TOMM40 genotypes in Alzheimer's to APOE status alone, Roses points to how TOMM40-mediated normal mitochondrial function changes once APOE mutations are at play. "At the protein level, there are different protein interactions that occur at the site of the outer mitochondrial membrane through which the only peptide/protein import channels are the TOMM40 proteins," he said, explaining this as a well-understood biological process. "All proteins that are necessary to form new mitochondria by … fission enter the mitochondria through the TOMM40 channel."

However, when APOE4 and APOE3 mutations attach to TOMM40-associated peptide recognition molecules, it impacts the normal functioning of these outer mitochondrial membrane proteins, causing cell death at different rates. With the varying rates of apoptosis, cognitive decline sets in for individuals at different ages during the period of Alzheimer's risk (generally after 60), according to Roses.

In the deepening divide in Alzheimer's research – between the amyloid devotees and the mitochondrial function trackers – Swerdlow has more faith in the latter movement. "If you think mitochondrial function plays an important role in Alzheimer's, then you might be more receptive to the idea that at least part of the action is coming from TOMM40," he said. "Personally, I very much believe mitochondria play a critical role, and that's why our center has decided to look at this issue."

The TOMM40/APOE genetic and metabolic hypothesis is further bolstered, according to Roses, by research showing that the heterogeneity of TOMM40 genotypes is "far greater" than the three polymorphisms seen in APOE, and by his research findings that a risk algorithm containing TOMM40 and APOE is informative for a larger portion of the population in gauging Alzheimer's onset than just APOE genotypes.

"Diseases are the result of biological processes, and it is the complex interactions of the APOEs with TOMM40s that have been inherited by each individual that leads to decreased ability to replace damaged mitochondria over time and determines the rate of mitochondrial failure leading to late-onset AD," Roses said. "The overwhelming genetic and cell biology evidence strongly support this mode of pathogenesis."

Diabetes of the brain

Despite the naysayers, Roses is not only standing behind his hypothesis, but has found a pharma partner to back his vision.

The approach being explored by Takeda and Zinfandel conceptualizes Alzheimer's as a form of diabetes of the brain. Studies have shown that individuals with impaired ability to metabolize cerebral glucose exhibit cognitive impairment of the kind seen with Alzheimer's onset. In line with Roses' view that amyloid plaque seen in Alzheimer's patients forms in the brain due to diminished oxygen and glucose utilization in neurons, Takeda and Zinfadel are using a diabetes drug that has been shown to increase these key sources of energy in the brain in order to try to stave off cognitive decline due to Alzheimer's.

As GlaxoSmithKline's senior VP of genetics research and pharmacogenetics from 1997 to 2007, Roses was involved in testing the impact of the diabetes drug Avandia (rosiglitazone) on mild to moderate Alzheimer's patients who were negative for APOE4. The drug, at doses normally given to diabetics, didn't improve cognition or behavior in these patients.

However, when the drug was given at very low doses in both a phase II and a phase III study, Avandia appeared to increase glucose utilization in study subjects' brains, but just missed reaching statistical significance in the phase III trial. The studies enrolled people from different ethnic populations, which likely confounded findings given the variability of TOMM40 and APOE genotypes across ethnicities, according to Roses.

For example, in Caucasians the frequency of the APOE4 allele is 15 percent, while the frequency of the same allele is 5 percent among Japanese people. Roses noted that allele frequencies of poly-T lengths of TOMM40 are even more variable across ethnic groups.

The primary analysis in the phase III trial for pioglitazone will be performed in Caucasians, but anyone between the ages 65 and 83 can be tested using the genetics-based algorithm and, if cognitively normal, can enroll in the trial. This open-label design will allow researchers to collect exploratory data on other ethnic groups. "One of the things we eventually want to establish is that your risk [for Alzheimer's onset], no matter what your ethnic background is, depends on our algorithm," Roses said. "What differs between ethnic populations is how many people in that population would have a particular TOMM40 genotype."

GSK ultimately decided against Avandia's development in Alzheimer's for safety concerns. Avandia was a blockbuster diabetes drug until 2007, when studies showed that the drug increased risk of heart attacks, leading health regulators to place restrictions on its use.

However, rat studies conducted within the Zinfandel/Takeda collaboration have shown that pioglitozone at extremely small doses – 0.04, 0.08, 0.16, or 0.32 mg/kg/day – improved oxygen metabolism in key regions of the brain after two days of treatment with the drug (PGx Reporter 7/25/2012). Researchers at Duke have followed up these findings in humans by testing low doses of pioglitazone during functional cognitive testing.

The data from a functional MRI study involving 40 people showed that when participants received a pioglitazone dose of 0.6 mg and then participated in a memory challenge exercise, they had the best oxygen utilization in the left hippocampus region of the brain, as compared to other higher dosages. This, researchers hypothesize, suggests that pioglitazone could regulate a number of pathways in the brain – involved in energy and lipid metabolism, insulin sensitivity, inflammation, and amyloid beta stabilization – and delay Alzheimer's related cognitive decline.

"What we have shown is that in humans when you … give a high dose of pioglitazone, the metabolic effect is spread all across the brain. You get activation of other things, and you don't see the effect you want," Roses said. "But as you go down lower and lower in dose … you see the specificity and intensity of the signal increase for the appropriate regions most affected in Alzheimer's."

Actos' label includes a black box warning for congestive heart failure and the FDA has warned of the risk of bladder cancer for diabetics who have taken the drug for a long time. By using such a low pioglitazone dose in the phase III Alzheimer's trial, Takeda and Zinfandel are hoping to avoid side effects diabetes patients have experienced when taking the thiazolidinedione class of drugs, Avandia and Actos. The safety of the drug given in this trial is particularly critical since with the help of the biomarker algorithm researchers will be enrolling cognitively normal individuals at risk for Alzheimer's onset.

"Drug companies usually try to push the dose of a drug as high as they can to demonstrate efficacy in their clinical trials,” Roses said. “At doses this low, and given the good safety record of pioglitazone from more than 22 million man-years of use, one would not expect there to be a problem."