functionbar_help
Font Size
functionbar_contact
Aboutus
Search Results for ""
Share
Top Stories

Published On September 22, 2012

CLINICAL RESEARCH

The Problem with Biomarkers

The best signposts--from blood pressure readings to genetic tests--can personalize diagnosis and treatment. Most don't help.

IN JUNE, THE NEW YORK TIMES MAGAZINE TOUTED Eva Redei’s discovery—of 18 genetic markers in the blood that appear to identify major depression and anxiety in adolescents—as an innovation “that will change your tomorrow.” It could certainly help solve an acute problem. Only a quarter of depressed and anxious young people get treated, many self-medicate with drugs and alcohol, and suicide is all too common.

“Adolescents must overcome horrendous hurdles just to be diagnosed,” says Redei, professor of psychiatry, behavioral sciences and physiology at Northwestern University’s Feinberg School of Medicine. “Parents have to recognize signs of depression, and teens have to be willing to talk about it—and adolescents aren’t communicative. They’re also dealing with significant peer pressure and having to make major decisions about their lives.” A blood test could ease the stigma of depression by objectively diagnosing the disease, says Redei.

But translating her findings, 12 years in the making, into a diagnostic test could take several more years, might cost hundreds of thousands of dollars, and may never happen. Yet she and many other scientists continue their quest for biological clues—often in the form of changes in blood, urine or body tissues—that signal the risk or presence of disease. Such “biomarkers” are at the root of every routine screening or diagnostic test. Fasting glucose is a biomarker for diabetes, high readings for cholesterol or blood pressure indicate an elevated risk of cardiovascular disease, and liver enzymes that spill into the blood suggest liver disease.

All of those biomarkers, however, are fairly blunt tools, and science has long been anticipating much more sensitive measurements. In the wake of the Human Genome Project and other advances in molecular biology—and with technology now able to quantify thousands of changes involving proteins, enzymes, DNA, RNA and antibodies—there have been more than 150,000 published papers heralding thousands of new biomarkers. Yet there are only about 100 biomarkers in routine use, and many have been around for decades. “In cancer, there are about 3,000 papers every year proposing biomarkers,” says John Ioannidis, director of the Stanford Prevention Research Center. “But we use only about a dozen—because all the others either have failed or have not yet shown they’re really helpful.”

And recently, even some long-established biomarkers have been discredited. In separate announcements this year, the U.S. Preventive Services Task Force recommended against PSA testing, which measures levels of prostate-specific antigen in the blood, for routine screening of prostate cancer, and said that women without symptoms should no longer get the cancer antigen-125 (CA-125) blood test, which detects a protein that ovarian cancer cells produce. The task force said the tests weren’t sensitive or specific enough to distinguish between aggressive cancer and benign changes.

That judgment has been controversial, with many physicians and patients insisting the tests often lead to early, lifesaving cancer detection. And the search for valuable new biomarkers continues for good reason—because effective finds can do a world of good. After the HIV virus was discovered in 1982, scientists worked with unprecedented speed and funding to come up with an HIV antibody blood test just two years later. That was an invaluable step in controlling the spread of AIDS, in part by preventing HIV-infected donors from contaminating the global blood supply. And the test let people know they were at risk well before they had symptoms—early enough to benefit from the effective treatments that eventually arrived.

To facilitate the hunt for the next great biomarker, a few nonprofit organizations have established collaborative research networks, while other scientists are urging higher standards in biomarker research. Further advances in technology should also help. But it may take all of these developments, and more, to reverse the recent pattern of disappointment.

BIOMARKERS HAVE ALWAYS BEEN INTEGRAL TO MEDICINE, even though the term didn’t show up in the scientific literature until 1977. The first person to record a patient’s temperature, for example, was using a biomarker of fever and infection. As early as the 1940s, physicians measured rheumatoid factors in the blood to diagnose rheumatoid arthritis. In the 1950s, scientists identified cardiac troponin, an enzyme released by dying heart cells; in the 1990s, when they developed a lab test to measure it, cardiologists and emergency physicians were handed a superbly accurate biomarker. “It’s an amazing test,” says Eleftherios Diamandis, head of clinical biochemistry at the University of Toronto, who says that a high troponin score is a definitive sign that someone has had a heart attack.

Many other well-established biomarkers, along with a handful of recent arrivals, provide crucial information. Several, including simple blood tests that measure cholesterol and C-reactive protein, a marker for inflammation of the arteries, help gauge the risk of heart disease. And for kidney disease there’s serum creatinine, which measures kidney function.

Important cancer biomarkers are the BRCA1 and BRCA2 mutations on genes that normally inhibit tumor growth. Analyzing DNA in blood samples can detect those alterations, which put women at higher risk of breast cancer. Other cancer tests look for proteins found in blood, urine, stool or body tissues. The blood markers AFP and beta-HCG can indicate testicular germ-cell cancer and can help monitor a patient’s response to treatment. “But only half of germ-cell tumors make those proteins,” says Timothy Gilligan, an oncologist at the Cleveland Clinic who develops biomarker guidelines for his specialty.

There are also a few biomarkers that can identify who’s most likely to respond to treatment. Women with breast cancer who produce an excess of a protein called HER2 may benefit from the drug Herceptin, while Tarceva is active against lung cancer cells with an EGFR mutation; Gleevec, meanwhile, targets the abnormal proteins of chronic myelogenous leukemia. There have also been exciting discoveries to identify key mutations in kidney cancer, says Gilligan.

For now, however, very few biomarkers can identify early-stage cancer, and a few do more harm than good. For example, although a blood test showing a high level of PSA may point to prostate cancer, an elevated reading can also stem from benign disease or infection. “PSA really can’t tell you what you most want to know—who needs to be treated,” says Carolyn Compton, president and CEO of the Critical Path Institute, a nonprofit group created by the FDA and the University of Arizona to reduce the time it takes to develop medical products.

That situation may improve with the introduction of the Prostate Health Index, a follow-up test for men with high PSA readings that won FDA approval in June after clinical trials showed a 31% reduction in unnecessary biopsies—thanks to the test’s ability to distinguish prostate cancer from benign conditions. And earlier this year, the FDA gave the go-ahead to a gene-based urine test, PCA3, that is more sensitive than PSA in detecting whether a slow-growing cancer is progressing.

Compared with other disorders, psychiatric disease may have the fewest revealing biomarkers, in part because criteria for diagnosing depression, schizophrenia, autism and other mental illnesses have shifted radically during the past 50 years. “If the experts can’t agree on how to recognize a disease, it’s very hard to come up with a surrogate to help measure whether it’s present,” says Compton.

REDEI'S INDICATORS FOR TEEN DEPRESSION COULD HELP CHANGE THAT, and there may be other landmark biomarkers waiting to be discovered. But research flaws often impede progress. Consider a seemingly momentous study published in The Lancet in 2002, in which investigators from the FDA and the National Cancer Institute announced they had discovered proteins in the blood that could be used to diagnose ovarian cancer with near-perfect accuracy. “That was a huge claim, and it caused the NIH and other institutions to spend hundreds of millions on proteomics technology,” including mass spectroscopy machines that could analyze almost any protein or peptide, says David Ransohoff, professor of medicine and epidemiology at the University of North Carolina School of Medicine. Yet as remarkable as such machines may be, problems with research design and analysis can fool researchers into thinking they’ve discovered an important association to disease that isn’t really there.

With mass spectroscopy, researchers analyze blood samples from people with and without a particular disease, and by examining complicated patterns of protein expression and interaction, try to discern differences. But when you try to fit a huge amount of data on many variables into a mathematical model that predicts a single outcome, you may find a pattern valid only for blood samples you’re analyzing, says Ransohoff. You could end up “overfitting” the data, identifying chance patterns that won’t reappear in samples from other people. Or you might detect a spurious signal that has to do with an extraneous factor—such as the calibration of a mass spectroscopy machine, which may wander from one day to the next—and says nothing about a disease process.

That’s what happened in the 2002 Lancet study. “The different signals the researchers found weren’t due to cancer at all,” says Ransohoff. Two years after publication, with no independent validation of the study’s findings, plans to develop a blood test for ovarian cancer had to be abandoned.

Other research errors also continue to impede progress. One rampant problem occurs when tissue samples are poorly matched—when, for example, because of the difficulty of obtaining tissue for testing, a researcher compares 10-year-old specimens from one clinic with freshly taken samples from another institution. Or some biopsies may have been taken under general anesthesia while others were harvested under local anesthesia—a difference that can alter proteins and lead to flawed comparisons.

In a study of 35 novel studies widely considered influential, Ioannidis found that associations between biomarker and disease were exaggerated 83% of the time. Even in cases in which there was a true association between biomarker and disease, very few ended up having much value for physicians. Says Diamandis: “If you discover a biomarker that improves prediction of disease by another 5% to 10%, that’s not enough. Most markers we find today offer marginal improvements over what we already have.”

That was also the upshot of a 2011 study of more than 1,000 women to validate 28 biomarkers, also for ovarian cancer. “The study found that none of the assays was any better than the one we had known about for years—the blood test CA-125, which doesn’t work very well,” says Ransohoff, who worked on the study. “Ten years ago, we thought that if we did enough measurements and math, all of nature’s secrets about prognosis, prediction and early detection would be unveiled, but it hasn’t worked out that way.”

Ioannidis advocates testing biomarkers in randomized controlled trials that compare new tests with the best existing options. Such studies would consider whether labs get consistent results as well as whether physicians can correctly interpret findings and use them to make treatment decisions.

EVEN WHEN NEW BIOMARKERS SHOW PROMISE, funding to explore their potential can be hard to come by. And while it may be somewhat cheaper to create a new test than to come up with a novel cancer drug, the treatment will provide a much greater return on investment, with patients often spending $50,000 to $100,000 per year. The financial upside of creating a $60 diagnostic test isn’t great, and few biomarkers get strong financial support.

To help find and develop biomarkers, the U.S. government recently formed the Early Detection Research Network, a program at the National Cancer Institute that funds work at dozens of academic labs. Yet while Joshua LaBaer, chair of personalized medicine and director of personalized diagnostics at the Biodesign Institute at Arizona State University, welcomes that commitment, he wishes it were larger. LaBaer, who is on the NCI’s board of scientific advisors, notes that the funds earmarked for biomarker research amount to less than 10% of the institute’s budget for cancer drug research. “Yet discoveries of cancer therapeutics aren’t coming out any faster.”

George Poste, chief scientist and professor in health innovation at Arizona State University and former president of research and development at SmithKline Beecham, believes the solution is to form large collaborative research networks that include scientists from academia and pharmaceutical companies, as well as experts in clinical trial design, data analysis, epidemiology and regulatory compliance. Such a network could also set standards for biomarker research.

One such private-public partnership, the Critical Path Institute, brings together more than 1,000 scientists and 41 pharmaceutical companies to share data, and the FDA and other countries’ regulatory agencies to sanction candidate biomarkers for speeding drug development. Critical Path takes credit, for example, for the decision of the European equivalent of the FDA to approve the use of magnetic resonance imaging to measure the volume of the hippocampus, a brain region, as a biomarker to select people likely to develop Alzheimer’s disease. Those individuals can then be recruited for clinical trials to test drugs that might arrest the disease before symptoms occur. “We want everyone to be able to use these biomarkers for research instead of having to develop their own,” says Critical Path’s Compton.

DESPITE FUNDING ISSUES AND OTHER PROBLEMS, plenty of researchers still believe in the potential of biomarkers. LaBaer thinks a test to detect early gastrointestinal cancers may be a real possibility, because cells from an emerging cancer often break off and travel through the gut and might be detectable in DNA studies on stool samples. Other potentially important biomarkers might test immune system responses. “Type 1 diabetes, rheumatoid arthritis, inflammatory bowel disease and multiple sclerosis all involve an overactive immune system that causes trouble by attacking ‘self,’” says LaBaer. “If we can identify these so-called auto-antibodies in the blood and determine what they are attacking, we might be able to detect early chronic diseases.”

Biomarkers’ predictive ability may also improve as researchers develop panels of markers rather than today’s single-molecule indicators. To find such biomarker “signatures,” researchers will need technologies such as mass spectroscopy to become even more sensitive—so they can analyze up to 1,000 proteins in the blood rather than the 100 proteins they can find today. And technologies such as nanotechnology or microfluidics, which can monitor individual cells and perform complex analyses on minute amounts of fluid, may facilitate analysis of biological changes that can’t be detected now. “Beyond proteomics and genomics, multiple new ‘omics’ fields—transcriptomics, metabolomics, epigenomics and ribonomics—will also be explored for cancer and for other diseases,” says Ransohoff.

“The quest to discover more important biomarkers is certainly not over,” says Diamandis. “But the biomarker of the future may be defined very differently than the classical biomarkers we use today.”

 

DOSSIER

1. “Comparison of Effect Sizes Associated with Biomarkers Reported in Highly Cited Individual Articles and in Subsequent Meta-Analyses,” by John P.A. Ioannidis and Orestis A. Panagiotou, JAMA, June 1, 2011. Highly promising initial biomarker studies appearing in scientific journals often turn out to have exaggerated associations to disease, according to these researchers, who urge greater skepticism in interpreting published reports and higher standards for claiming biomarker success.

2. “So, You Want to Look for Biomarkers,” by Joshua LaBaer,Journal of Proteome Research, June 2005. Biomarker researcher LaBaer outlines the basic rules for discovering clinically useful biomarkers, including understanding the consequences of being wrong, such as developing a biomarker that is so sensitive it generates unacceptable false-positive results.

3. “Bias as a Threat to the Validity of Cancer Molecular-Marker Research,” by David F. Ransohoff, Nature Reviews Cancer, February 2005. Researchers can unwittingly compromise the specimens they are analyzing for possible biomarkers, making their discoveries impossible to validate. Ransohoff explains how bias occurs in biomarker research and how to avoid it.

Share