Color-blind? The value of racial data in medical research

Late last year, VaxGen Inc. announced the much-anticipated results of its phase III AIDS vaccine trial. But the data released by the Brisbane, Calif., biopharmaceutical company ultimately caused more confusion than conclusions.

Though the vaccine showed no efficacy within the study population, a subset analysis seemed to suggest that African-Americans and Asian-Americans experienced some benefit.

So did the vaccine work or not?

At the time, the data seemed to indicate that it might for some people. Subsequent results from a trial in Thailand exclusively among Asians found no efficacy. And similar conclusions from a National Institutes of Health analysis are expected this year. A VaxGen spokeswoman said the company would like to do a trial focused on an African-American study population to settle the question, but does not have the funding.

Now, as the dust is settling, critics point to the episode as an example of what can happen when selected data are given too much weight too soon.

"It was a desperate act by a company that was trying to save a failed product," said Mark Feinberg, MD, PhD, a professor of medicine and HIV specialist at Emory University School of Medicine in Atlanta. "If they really cared about racial and ethnic differences, they would have structured a very different trial."

But there is also a scientific contingent that sees value in this kind of numbers crunching.

Thus, the VaxGen story is not only about a vaccine and a certain degree of puzzlement. It also highlights the debate about the role race and ethnicity play in clinical trials.

Studies are released almost daily finding minor, large or no differences between various racial or ethnic groups. In some instances, this type of analysis is required. The NIH requires data about minority participation but not necessarily about how one group does in comparison with another. The Food and Drug Administration requires such analysis along with age, gender and, just recently, genetic information.

"There are many, many factors that play roles in drug response, and one day we will have many more ways to characterize subpopulations than we do now," said Katherine Hollinger, DVM, MPH, senior supervisory health promotions officer at the FDA Office of Women's Health. "What we want to be able to do is to consistently report race and ethnicity across the lifespan of a product. If there are issues in a subgroup of the population, we need to know that so that we make decisions fully informed."

Still, analyzing clinical trial data in this manner is a matter of great debate. Numerous editorials have appeared in the scientific literature, including the British Medical Journal and the New England Journal of Medicine, arguing both for and against.

On one hand, critics argue that race and ethnicity, unlike gender, are social constructs that change according to time and place and thus should not be relied upon by science. Most studies use the federal government's Office of Management and Budget standards to define racial and ethnic categories, but those standards have changed over the years. A 1997 revision, for instance, shifted Filipino-Americans from the Pacific Islander category to Asian. Hispanic used to be a race but is now an ethnic group within African-American or white classifications.

Racial and ethnic identity is almost always self-reported and usually determined by the subject checking off a box from a presented list. Many scientists maintain that a person's genetic makeup is more specific and more accurate, and in the modern genetic age, that is the more scientific approach.

"There isn't a biological basis to race," said Saif S. Rathore, MPH, author of an Oct. 4 British Medical Journal editorial and a medical student at Yale University School of Medicine in New Haven, Conn. "How valuable is a construct that's fundamentally mutable over time and location? There is no evidence that race itself modifies the effect of any drug."

Meanwhile, though advocates acknowledge that race and ethnicity are social constructs, some say these factors still correlate well enough with genetics to be useful. They also relate to other factors, such as socioeconomic status, that can affect the course of treatment. Besides, genetics is not yet at a point where such markers are always valuable in clinical trials.

"Anyone who practices clinical medicine has made the observation that certain tendencies appear to show up in certain populations," said Clyde W. Yancy, MD, associate dean for clinical affairs and associate professor of medicine and cardiology at the University of Texas, Southwestern Medical School in Dallas. "The question becomes, do you simply ignore these or do you try to uncover whether or not there are real dilemmas here with regard to the expression of disease and the therapy of disease that appear to line up on the basis of race."

Critics also charge that subset analysis, such as in the VaxGen case, which has much less statistical power than the general study population, can lead researchers into dead-end trials.

"It's been a huge distraction and created a lot of bad science," said Richard S. Cooper, MD, author of several papers on the subject and head of the Dept. of Preventive Medicine and Epidemiology at the Loyola Stritch School of Medicine in Maywood, Ill. "And it can generate a lot of cost if you end up with an implausible hypothesis that you then pursue. It's never produced anything useful."

A case in point

Its supporters agree that subset analysis is not to be taken as gospel. Nonetheless, their position is that it generates useful hypotheses worth following, and that collecting this kind of trial data is simple and doesn't add much to the cost.

"You ask one more question up front, which is basically a check box, and on your back end you're going to do a couple more tables and maybe a few more analyses, but the cost is minimal," said Kenneth Borow, MD, CEO of The Covalent Group, a Wayne, Pa., clinical research organization. "If you try to make a broader statement on the subpopulation analysis -- that's when you can start getting yourself into trouble. But it can lead to hypotheses for subsequent studies."

To date, no such questions have generated results with great clinical significance. Hepatitis C drugs are thought of as less effective in African-Americans. Physicians tend to use that information to advise patients about the possible effectiveness of treatment, although there is no evidence that patients are turning it down because of the possibility of reduced efficacy.

Beta-blockers and ACE inhibitors also are labeled as less effective in this group, but they're still offered, although some cardiologists say that they may prescribe them at slightly higher doses. The International Society on Hypertension in Blacks published a consensus statement in the March 10 Archives of Internal Medicine encouraging the use of these medications to lower the blood pressure of African-Americans. Additionally, several studies have suggested that these drugs are as effective at reducing mortality in African-Americans as they are in whites.

"In terms of efficacy, there is no rationale for using race as a reason to avoid certain classes of agents in African-American patients with high blood pressure," the society said.

When race matters

But race-based prescribing soon could be a part of medicine, in at least one instance. The "poster child" for race-based hypotheses is expected to be answered in 2005. Experts say the African-American Heart Failure trial, now recruiting participants, will, if findings are positive, explode the field of race-based medicine.

The study was conceived when retrospective analysis of trial data regarding a particular heart failure therapy suggested it might work better in African-Americans than in Caucasians. The FDA in 2001 gave the go-ahead for a study of this heart-failure drug, BiDil -- a combination of hydralazine hydrochloride and isosorbide-- exclusively in this population.

"You never get adequate information on the subgroups in large mega trials," said Jay N. Cohn, MD, professor of medicine at the University of Minnesota Medical School in Minneapolis who led initial trials that suggested further study in African-Americans. "It's appropriate to look at a population that is likely to get a better response from your therapy."

Researchers are hoping the study will find a successful treatment for a medical condition that disproportionately burdens African-Americans, and that the results might have some generalizability to non-African-Americans who have similar heart failure characteristics.

"We're not doing this just so that at the end of the day we can say, if a patient comes through your office and they happen to be black, this is how you should treat their heart failure," said Dr. Yancy, a member of the trial's steering committee and one of the principal investigators. "Anyone in the population who has this abnormality in nitric oxide homeostasis is likely to have the same benefit, and there's nothing to say that abnormality is exclusive to African-Americans, although it may be more common."

But even if the results are not earth-shattering, those who favor using race and ethnicity factors in clinical trials say the study is significant for recruiting a study population that has traditionally been hesitant to participate.

Some even maintain that analyzing data in this way may make minorities more eager to participate.

"It raises awareness for patients because they recognize that they are welcome participants," Dr. Yancy said.

Advocates say the only way to determine drug response variations among subgroups is to explore the concept. There indeed may not be a difference in how African-Americans and Caucasians respond to heart failure drugs. Nonetheless, that question must be asked to reap a definitive answer. Not pursuing it would be the mistake, and not having data for questions that come up later would hamper future research.

"There appear to be some subgroup differences in clinical trials," said Keith C. Ferdinand, MD, past board chair of the Assn. of Black Cardiologists and professor of clinical pharmacology at Xavier University in New Orleans. "If we don't define the true scientific significance of those subgroup differences, we may miss opportunities."