Skip to main content

First Steps for Personalized Medicine


Intuitively, pharmacogenomics and personalized medicine make sense. After all, every patient is different, and one person's tumor may not be like the average tumor, or a drug that works well for a particular type of depression might be ineffective for another. "It's very appealing," says Lecia Sequist, a physician at Massachusetts General Hospital. "It makes sense that you don't want to just be treated under an umbrella diagnosis; you really want to have the most specific, personalized treatment that exists."

In the past few years since the rise of Herceptin and Gleevec, more and more personalized treatments are coming out. With them come companion diagnostics to see if a particular patient is the right candidate for the drug — or if that patient should stay away.

"Pharmacogenomics is a powerful idea. It makes so much intuitive sense that a good pharmacogenomic test could be extremely useful in some clinical circumstances," says Al Berg, a family medicine physician at the University of Washington. And the tests, which are particularly taking aim at cancer treatment and eliminating adverse reactions, keep on coming.

According to one study, adverse drug reactions cause or contribute to six to seven percent of hospitalizations and 100,000 deaths annually in just the United States. If those effects could be avoided by a diagnostic test, why not use it? The catch is just how valid and clinically useful pharmacogenomic tests and companion diagnostics are. Since these tests have fallen into a regulatory no-man's-land, their utility is less clear than perhaps it should be.

"I think in general the hype has exceeded the reality, but there is reason to be excited and optimistic," says Kathy Hudson, director of the Genetics and Public Policy Center at Johns Hopkins University.

Personalizing cancer

Much of the ongoing work in developing pharmacogenomics-based medicine is taking place in a branch of the field that is already somewhat personalized. "Cancer treatment has always been: get a biopsy and give the right treatment for that type of cancer," says MGH's Sequist. "I think that it's not really a change in the way that we take care of cancer. It's more an expansion of our abilities of understanding what causes cancer and how to treat it." Now, she adds, that no longer ends with peering at the tumor under a microscope, but extends to examining the tumor's genetics to see where and how it's gone awry.

Before they can further divvy up types of cancer, researchers need biomarkers that indicate diagnosis, prognosis, and targets for therapies. Some biomarkers already exist; there are the BRCA genes and HER2 for breast cancer, and KRAS and EGFR are implicated in various types of cancer. At Johns Hopkins, Nickolas Papadopoulos is currently looking for more. "We're trying to identify all genetic alteration that might be responsible for the development and progression of cancer," he says. "The reason we do that is because we believe that's going to give [a] good target for therapeutics and, we hope, a lot of good biomarkers for diagnostics and prognostics."

There are also many cancer drugs on the market that can work well some of the time, adds Monogram Biosciences' Alf Merriweather. The trick there is knowing which population of patients will respond.

Some tests on the market are purely diagnostic. Clinical Data's PGxPredict: Rituximab was developed as a diagnostic to determine if treatment with rituximab is appropriate for a patient with non-Hodgkin's lymphoma. The test genotypes a SNP in the Fc gamma IIIa receptor found on natural killer cells, the receptor to which rituximab attaches. A certain amino acid change found in about 20 percent of the population affects the binding affinity of rituximab to the receptor. People with the mutation, says Carol Reed, chief medical officer of Clinical Data, respond better to the drug. "They get better binding affinity. They get lysis of their lymphoma cells," she says. This FCG3ra variant and its pathway are also emerging as important in understanding responses to other drugs, such as Herceptin and Erbitux.

Genomic Health, on the other hand, created a test that not only determines whether a particular treatment will be effective, but also gives an estimate of how likely cancer recurrence is for that patient. Oncotype Dx, Genomic Health's test for breast cancer, is an assay for 21 genes that determines whether patients should have chemotherapy and the likelihood that their cancer will recur. The test also focuses on newly diagnosed patients with node negative, estrogen receptor positive disease.

"Although chemotherapy is offered to the vast majority of women with node negative, estrogen receptor positive breast cancer, only about four out of 100 benefit from the addition of chemotherapy," says Steve Shak, chief medical officer of Genomic Health. "It is very important to better identify which women at the time of diagnosis are likely to recur, but even more importantly, which women are more likely to benefit from chemotherapy."

Others are hoping that targeting a drug to a specific population can resurrect a weaker-than-expected product. One drug that may make a comeback is AstraZeneca's tyrosine kinase inhibitor called Iressa. The drug, whose generic name is gefitinib, first came on the market as a lung cancer treatment in 2003. But it was taken off the market when a large study determined that it did not lengthen patient survival time.

Then, in 2004, researchers noticed that a few lung cancer patients have mutated epidermal growth factor receptor genes. MGH's Sequist says this is a mutation mostly associated with nonsmokers and it appears to be another way of developing lung cancer. The result is actually a different type of lung cancer — one that responds to treatment with gefitinib or a similar drug called erlotinib, marketed by OSI Pharmaceuticals as Tarceva.

In a recent study funded by AstraZeneca, Sequist and her colleagues investigated whether adding a genetic analysis step into the workflow of diagnosis and therapy was feasible and how people with EGFR mutations responded to gefitinib. The patients in the study were screened for EGFR mutations by directly sequencing the gene. The most common mutations were a point mutation, L858R, and a series of overlapping deletions in exon 21. Patients with the mutation were treated with gefitinib rather than the standard chemotherapy. Results were encouraging, says Sequist. The study reported a response rate of 55 percent; in a general lung cancer population, chemotherapy has a response rate of about 20 percent. More trials are currently underway to study how patients with this type of lung cancer respond to tyrosine kinase inhibitors versus chemotherapy.

Other conditions, too

Pharmacogenomic testing, of course, isn't limited to cancer. Other areas of interest include CYP450 testing for use in determining proper drug doses in conditions as varied as blood clots and depression, as well as determining how to best treat HIV. CYP450 testing has recently been of particular interest as it is thought to be involved in the metabolism of more than 75 percent of drugs, including warfarin.

In fact, last year FDA changed the warfarin label to say that people with CYP2C9 and VKORC1 gene variants may respond to warfarin differently than patients without those variations. Many labs, including Clinical Data, LabCorp, and Quest Diagnostics, offer genetic testing for warfarin. "Warfarin is a difficult drug to use. It would be nice to have all the evidence you can get to design a good strategy for patients," says UW's Berg.

Gloria Grice has been using CYP2C9, VKORC1, and other variants to study anticoagulation pharmacogenomics at Barnes-Jewish Hospital and Washington University. CYP2C9, she says, dictates how quickly someone metabolizes warfarin, and VKORC1 determines if someone is sensitive or resistant to treatment. "Knowing their CYP2C9 gives about 25 percent of the picture, and we know that VKORC1 can give roughly about that much," Grice says. "You add the two together and you are just starting warfarin, you have a little less than 50 percent of what dictates the dose that they are going to need. That's a big percentage of how we decide that." The rest of the decision, she says, is based on information such as the patient's age, body size, gender, and other medications.

Another area of interest for developing companion diagnostics is HIV treatment. Monogram Biosciences developed a test that tells clinicians if a particular drug is right for an HIV patient. The test, called Trofile, determines whether the patient's HIV strain is using the CCR5 or CXCR4 receptor to hijack cells. "The test provides a yes/no answer to the physician that the patient's virus is using the CCR5 receptor or not," says Merriweather, Monogram's chief financial officer. "It's a yes/no test in terms of whether a particular drug should be applied to a patient or not." Only one drug on the market — Pfizer's Selzentry — blocks the CCR5 receptor, and that drug is inextricably linked to the Trofile test. When Selzentry went through clinical trials, Monogram provided Trofile to Pfizer to select patients. "The testing is required commercially as part of FDA approval for the drug to determine the right patients," Merriweather says.

Other companies are developing both the drug and the diagnostic. Clinical Data's vilazodone, an anti-depressant that is both an SSRI and a 5-HT1A partial agonist, is in its second set of phase III trials and the company is hoping to find the best target population from the get-go. "If you think about how drugs are developed now, they are developed for a broad population of patients. … Depending on the drug and the therapeutic area, a smaller or larger percent of patients will respond positively," says Clinical Data's Carol Reed. "We think that by developing biomarkers in order to target these drugs specifically to responder, or to have patients avoid the drugs if they are at risk for an adverse event, that we can make drug prescription not only more effective in the clinic but more cost-effective overall."

As vilazodone wends its way through clinical trials — in which scientists hope to show the drug is effective in the general population — Clinical Data is also looking for biomarkers correlated with response to the drug. One marker identified so far is associated with a doubling of response to vilazodone compared to people without the marker. That and any other biomarkers the company finds will then be incorporated into a companion diagnostic to determine which patients will respond well to the drug.

Murky regulations

If all a test does is determine which patients will respond well to a drug, it can be difficult to appreciate why there's so much regulatory angst related to pharmacogenomics.

"On the face of it," says UW's Berg, "it seems like it's just a test, how can it cause any harm?" The problem is that no test is perfectly accurate, and the long-term effects of a false negative or positive are at this point poorly understood. "You may mislead the clinician and the patient into a clinical strategy that actually is ineffective," Berg adds. "So you use the test and it leads to the wrong outcome or maybe a worse outcome than if you didn't use the test."

Furthermore, Hopkins' Hudson adds, people tend not to think about having a genetic test repeated, as they would a cholesterol test, so bad results aren't overturned. Sometimes, clinicians are only relying on a genetic test as a diagnosis, particularly for in utero diagnoses, or for predicting breast cancer. "People are making decisions about continuing or terminating a pregnancy based solely on the genetic test. Similarly, cancer risk information might lead to prophylactic surgery or increased screening — or a false negative might lead to losses of opportunities for those prophylactic, preventative interventions," Hudson says. "In the pharmacogenetic context, there's no other clinical information that's going to tell you you're going to have an adverse reaction to a drug."

Indeed, a handful of recent studies highlight the pitfalls of testing. One presentation at the 2007 American Society of Clinical Oncology meeting pointed out that variable cutoff criteria for Her2 levels exclude women from receiving Herceptin who may actually benefit from it. Also, a 2006 study found that 14 percent to 16 percent of Her2-positive tumors are really Her2-negative, and that 18 percent to 23 percent of Her2-negative tumors are actually positive.

In short, if a test is going to determine whether a patient receives a drug, "you want to know that they're detecting the right sequence variant or detecting the right expression pattern or whatever it is that they are detecting, and that there is a direct and valid association between that and drug response," Hudson says.

In the United States, genetic testing labs are certified by the Centers for Medicaid and Medicare Services under the Clinical Laboratory Improvement Act. With that act, Hudson says, Congress intended to ensure that clinical laboratories have the appropriate training to run labs, maintain sample integrity, and are proficient in the tests that they perform. "Sadly, the Department of Health and Human Services in its implementation of CLIA sort of didn't really fulfill Congress's expectations," Hudson says. "We've fallen quite short of ensuring that genetic tests in the United States are analytically valid when they are offered to consumers."

"There are CLIA labs that are doing a lot of diagnostics, and those are the ones that not every doctor trusts," says Papadopoulos, adding that companion diagnostics should be subject to the same rigorous validation that drugs endure.

Over the years, the FDA has gone back and forth on the question of whether it had jurisdiction over these kinds of tests. Recently, though, the debate on authority has largely been put to rest, but many say that the agency still has not developed a firm stance on how these tests will be regulated. "FDA has said, 'Yes, yes, yes, we do have the authority to regulate laboratory-developed tests and we are using our enforcement discretion, so we are going to pick and choose which of these laboratory-developed tests we feel warrant our oversight,'" Hudson says.

Last spring, HHS assembled an advisory committee on genetics, health, and society to make recommendations about genetic testing. The committee, which included Hudson, concluded that the agency's oversight was inadequate and needed improvement. In particular, the committee said that a mandatory genetic testing registry should be created so consumers, physicians, and watchdog groups could know a test's false-positive and false-negative rate, its positive predictive value, and its evidence for clinical validity. The group also said that diagnostic labs should have to participate in proficiency tests. "As of today, the department has not said a word about whether it's going to accept those recommendations, if it's going to adopt those recommendations, if it agrees with those recommendations," says Hudson. She points out that a US presidential election is coming up, which could mean a shake-up in priorities and personnel at FDA.

Still, FDA has shown interest in companion diagnostics through its voluntary genomic data submission policies and has included genetic testing information on multiple drug labels. Genomic data voluntarily submitted to FDA is not, according to the guidance document, taken into consideration during the approval process.

Not only has FDA changed warfarin labeling to encourage genetic testing, but it also added a warning to the label for the HIV drug abacavir, marketed by GlaxoSmithKline as Ziagen. A small percentage, between five and eight, of people have a certain human leukocyte antigen allele, HLA-B*5701, that leads them to have a serious and even fatal reaction to the drug. GSK did large prospective clinical trials to look at HLA testing, abacavir hypersensitivity reactions, and whether testing for the allele would decrease the number of reactions. "During that period of time, because the AIDS community, both patients and providers, are sophisticated, people started routinely testing for this before giving abacavir anyway. The number of patient deaths and adverse reactions went plummeting," Hudson says.

Helping clinicians

For now, clinicians are left to try to figure out on their own if a test is valid or useful. Despite the lack of clarity, medical professional groups have staked out a claim in the genetic testing arena. Both the American Medical Association and the American College of Medical Genetics have released policy statements saying that healthcare professionals are needed to order and interpret genetic tests.

In 2004, the CDC National Office of Public Health Genomics launched a pilot project called Evaluation of Genomic Applications in Practice and Prevention to apply evidence-based medicine principles to assess genetic tests and genomic technology. "I think clinicians want an authoritative source of advice on whether the tests do what they say that they do and whether they are a good idea," says Berg, who leads EGAPP. "This is a model project to try to put together some methods to see whether we could adapt evidence-based processes to this new area of medicine and make some recommendations that are relevant to clinicians."

The group began several reviews of genetic tests for breast cancer, Lynch syndrome and hereditary nonpolyposis colorectal cancer, ovarian cancer, and depression. The review of CYP450 testing for depression treatment is the first to be completed.

The group combed through the literature on CYP450 testing for SSRI use and reported its findings in last January's Genetics in Medicine. In the end, the group found "insufficient evidence to support a recommendation for or against use of CYP450 testing in adults beginning SSRI treatment for non-psychotic depression."

"All we can say is that the evidence was not clear on lots of potential benefits and harms. We weren't able to reach a conclusion on what those harms would be," Berg says. "Basically, we ended up with more question marks than answers."

To support the use of a test, Berg says the group would have needed to see high-quality clinical trials. "One of the reasons why CYP450 and SSRIs are such a potentially good thing to look at is it's exactly the kind of condition and the kind of test that should lend itself extremely well to high-quality, randomized controlled trials. It's a common condition and the decision about the use of SSRIs is a very common decision," he says. He has no doubt, he says, that if a high-quality, randomized trial comes along with sound evidence that CYP450 testing is effective in cutting down the time that patients take to respond or in cutting costs, that "physicians will only be too happy to apply it."

But first …

To get to that point of physicians happily using pharmacogenomics to help their patients, more diagnostics — and well-validated ones at that — will have to be developed.

Monogram's Merriweather expects more and more diagnostics to come into the marketplace. "I think you'll see a proliferation of companion diagnostics for drugs, both currently approved drugs and drugs coming through the pipeline, where there is some targeting of the drug with diagnostics, to get the right drug to the right patient," he says.

Kathy Hudson takes a more cautious view. "Can we expect a continuous drip, drip, drip and maybe even a little flow? I hope so, but it's not going to be quite as grand and revolutionary as I think we would sometimes lead the public to believe," she says.

Either way, many see the progression to personalized medicine as necessary and even inevitable. "If we want to save a lot of lives, I think in the end, it has to be personalized medicine in place at some level because my tumor — knock on wood — is going to be different than somebody else's," says Hopkin's Papadopoulos. "That right there, that's personalized medicine."

The Scan

Call to Look Again

More than a dozen researchers penned a letter in Science saying a previous investigation into the origin of SARS-CoV-2 did not give theories equal consideration.

Not Always Trusted

In a new poll, slightly more than half of US adults have a great deal or quite a lot of trust in the Centers for Disease Control and Prevention, the Hill reports.

Identified Decades Later

A genetic genealogy approach has identified "Christy Crystal Creek," the New York Times reports.

Science Papers Report on Splicing Enhancer, Point of Care Test for Sexual Transmitted Disease

In Science this week: a novel RNA structural element that acts as a splicing enhancer, and more.