Drug researchers must select from a diverse biomarker palette to find the best blend for drug development and diagnostics.
For all of the attention biomarker research receives these days, it would be easy to conclude that biomarkers are a new idea. In fact, they are simply a re-branded concept, known to physicians of yore as “tests.” Beta-hemolysis is a biomarker for strep throat; cholesterol and blood pressure are biomarkers for heart disease.
The renaming reflects a deeper change in the way basic researchers and drug developers think about disease. Traditionally, the specific biomarker was largely irrelevant to a drug developer, as long as it worked. Whether a clinical trial uses blood agar plates or a new microarray-based test for strep throat, the process of developing an antibiotic remains the same. Today, the complex diseases drug companies are trying to treat present a more nuanced problem, blurring the boundaries between biomarker discovery, drug development, and diagnosis.
The new generation of biomarkers is unlikely to produce the kinds of unambiguous results regulatory agencies like. “The Food and Drug Administration (FDA) has a very strict definition of [the kinds of] biomarkers [that] can be used as a surrogate,” says Orest Hurko, MD, assistant vice president for translational research at Wyeth in Collegeville, Pa. Hurko adds that such surrogates “are the minority—most of the biomarkers are simply tools that one can use not as a substitute, but rather a tool to get to the final distance.”
Seeking stratification
Part of the problem is that discovering and validating a true surrogate biomarker is more like a career-long dream than a feasible project. To accept a surrogate as valid, the FDA needs proof that the biomarker tracks disease severity reliably, and that any treatment that improves the biomarker also treats the underlying disease. Validating cholesterol as a surrogate marker for heart disease took three decades.
That’s not to say the search is futile. “There is plenty of use in developing biomarkers and tools for individual decision-making, not to satisfy the needs of the US Food and Drug Administration, but to let our … executives make intelligent decisions,” says Hurko. Like a growing number of other drug developers, Hurko and his colleagues are taking the approach of “different biomarkers for different purposes.” In this view, a biomarker doesn’t have to be diagnostically relevant to be useful.
Stratification biomarkers could help companies place better bets on which leads to pursue. “Frequently [for] patients even in one diagnostic category, some patients will respond nicely to the drug, others won’t,” says Hurko. In a now-famous example, breast cancer patients whose tumors carry the HER-2 receptor respond astonishingly well to Genentech’s Herceptin (trastuzumab). “Unfortunately, that’s only one out of ten breast cancer patients, the other nine out of ten you might as well be giving them tap water,” says Hurko.
Similar phenomena are probably involved with other drugs that show highly variable results, but clinicians often lack the tools to determine which patients will respond. That’s especially true in psychiatry, where diagnosis of conditions such as depression remains highly imprecise, and medications can seem almost capricious, providing miracle cures for some patients while having no effect on others. For companies organizing new drug trials, it’s a nightmare.
|
“Some centers are so eager to enroll patients that they take patients at the fringe, where it’s not really clear that they have an affective illness,” says Wayne Drevets, MD, a senior investigator in the mood and anxiety disorders program at the National Institute of Mental Health in Bethesda, Md. As an example, Drevets cites a study that tracked rates of unipolar depression in several cities across the US. Data from some centers, such as Washington University, pegged the prevalence of depression at around 7%, while other centers, such as Yale University, found rates closer to 25%.
But New Haven is not necessarily sadder than St. Louis. “This difference between 7% and 25% was probably not a real one, but it was just the sort of thresholds … that people were approaching cases with to decide whether they had the clinical syndrome or not,” says Drevets. He adds that “one thing that a biomarker might do is to help homogenize the subjects that are going into a study.”
Indeed, evidence is mounting that depression, like cancer, is not a single disease at all. “It’s something that’s more like … pneumonia, where once you have the tools to really subtype an illness like pneumonia, you realize that it’s got lots of different etiologies and some differences in pathophysiology depending upon the etiology,” says Drevets. Giving the same drugs to every depressed patient might be as unreliable as prescribing the same antibiotic to everyone with pneumonia.
Stratifying biomarkers for psychiatric conditions may involve a combination of imaging and molecular markers. Drevets and his colleagues have found that, with careful pre-screening, they can detect subsets of depressed patients with specific volumetric changes in particular brain regions. In bipolar depression, the team has also discovered a genetic marker; patients with different alleles of the muscarinic-2 receptor may develop slightly different forms of the disease. One genotype correlates with specific abnormalities in imaging studies, and also with more severe cognitive impairment.
For drug developers, the implications are obvious. “Identifying, at least for early trials, those patients who are likely to respond and only enrolling those … is not only economic, but more ethical,” says Hurko. He advocates using biomarkers to narrow the population of patients for an initial clinical trial: “In a sense you start off by doing the experiment and saying if the drug’s going to work, it’s going to work in these people, then if it works you say ‘great, maybe it’ll work in some more people,’ and then you expand from that.”
Importantly, none of this requires waiting for regulators to approve a biomarker as a true surrogate. “We choose what dose we give, we choose what patients we enroll in our study, we don’t have to go through this 30-year procedure … with the FDA giving us a blessing that ‘oh yes you can use that as a primary endpoint’. We can use that right away; it’s an internal decision,” says Hurko.
|
Not-so-rapid diagnosis
That doesn’t mean that the next generation of biomarkers will remain inside pharmaceutical companies’ walls. For some chronic diseases, even biomarkers that seem highly specialized today could turn out to have broader uses in the future. “I actually don’t think anybody has enough information about most of the biomarkers to actually know whether what they’re working on will be useful for diagnosis or not,” says William Thies, PhD, chief medical and scientific officer for the National Alzheimer’s Association in Chicago, IL.
In Alzheimer’s disease research, the boundaries between diagnosis and treatment are not just blurred, they’re gone. That’s because the absence of good biomarkers for early diagnosis of the disease has made it nearly impossible for drug developers to test treatments. “There is a general sense that if we can find good biomarkers that will be surrogates for disease progression, and we can identify those with some surety, that should shorten drug development time a great deal,” says Thies. Conversely, he adds, “you can’t really prove your biomarker without an effective therapy, and certainly having the biomarker there will reinforce the effectiveness of the therapy.”
To cut this Gordian knot, drug companies, academic researchers, and nonprofit organizations have formed the Alzheimer’s Disease Neuroimaging Initiative, or ADNI. “While it’s titled as a neuroimaging initiative, it also has a very robust chemical indicator [effort] that will look at blood and CSF, and … begin to point to what might be the best possible biomarkers,” says Thies.
The initiative is already well underway, and its scale indicates why pharmaceutical companies needed to band together for it. “We’re now $82 million and about 1,200 patients into it … no one company could possibly afford that kind of investment, even someone like Wyeth where we’ve put a major stake in the ground on Alzheimer’s disease,” says Hurko.
Because ADNI’s projects focus on “pre-competitive” research, preceding work on specific compounds, companies can participate without fearing that they will give their competitors an edge. That also means there’s no reason to keep the data under wraps, so ADNI releases all of its results to the public online. “We all decided it’s probably best to share the risk and enormous cost of developing that surrogate marker and also enjoy the blessing of the Food and Drug Administration … this is not an undertaking for a single company,” says Hurko.
Nor is it likely to yield quick results. An earlier effort to characterize the progress of Alzheimer’s disease, the UK Medical Research Council’s Cognitive Function and Aging Study, or CFAS, has been running for more than 15 years. Involving nearly 18,000 patients at six research centers, the mammoth project has provided substantial new insights into the disease, but few clues for new biomarkers.
“We haven’t found any particular set of criteria that work very well in a population setting, certainly in terms of its ability to identify most people who dement; none of them are much good,” says Carol Brayne, PhD, director of the Institute of Public Health at the University of Cambridge in Cambridge, UK. However, the effort’s postmortem analysis of elderly brains has helped illuminate many features of Alzheimer’s disease pathogenesis, as well as normal aging.
Those studies have also provided a cautionary tale for would-be biomarker hunters. “What we find when we look at the brains of people in our study, who have a median age of death of well over 80, is that there’s a big mixture of pathology, and that there’s a lot of pathology in those people who do not express dementia in their lifetime,” says Brayne. She adds that “you can go to your grave with the pathologies … without developing dementia.” In order for a diagnostic biomarker to be useful, researchers will need to ensure that it accounts for such vast discrepancies in outcomes.
When they do finally achieve validation, the next generation of diagnostic biomarkers will also require some changes in the clinic. Drevets and his colleagues, for example, have found promising leads for depression biomarkers, but doing so requires carefully calibrated, research-grade MRI equipment. The requirement we have for being able to get the magnets with a certain reliability and a certain precision and sensitivity is much different than you need for clinical imaging, and so I think it would really put a greater burden on clinical radiology groups, because right now they’ve gotten used to using MRI scanners that are pretty much turnkey,” says Drevets.
Fortunately—or unfortunately, depending on your perspective—radiologists have plenty of time to get ready. “As genetics is advancing, as imaging is advancing, and as some of the postmortem work is advancing, I think we’re increasingly homing in to things that have bigger effect sizes. Maybe after another decade or so we’ll have a number of these kinds of abnormalities that have been worked out into meaningful biomarkers,” says Drevets.
About the Author
Originally trained as a microbiologist, Alan Dove has been writing about science and its interfaces with industry and goverment for more than a decade.
This article was published in Drug Discovery & Development magazine: January, 2008, pp. 12-16.
Filed Under: Drug Discovery
