Adverse drug reactions are an underestimated but serious problem in medicine, and are the fourth leading cause of death in the USA after cancer, coronary heart disease and stroke (Lazarou et al, 1998). Although any given drug, if prescribed correctly, is beneficial for most patients, it may have unexpected results in a small subset of users. This can range from simply having no effect at all to causing serious, sometimes life-threatening, side effects. This variation stems—at least in part—from metabolic and genetic differences between patients. It would be a big leap for medicine if physicians could identify these differences and treat patients based on their individual genetic makeup. Now that the human genome sequence is complete, several follow-up genomic projects, most notably the HapMap initiative to identify and catalogue haplotypes, are focusing on identifying such genetic variants. But it is not yet clear whether this knowledge will lead to personalized medicine. One hindrance is the fact that, for the time being, physicians and scientists have found a comfortable interim solution: racial profiling. Although, at first, this approach could help to use drugs more efficiently, in the long term, it could hinder the exploitation of genetic information in medicine.

In fact, physicians increasingly use their patients' skin colour or other physiological features as a first step towards ‘individual' treatment, under the assumption that specific traits cluster by race. “When I prescribe Prozac to a patient who is African-American, I start at a lower dose, […] in part because clinical experience and pharmacological research show that blacks metabolize antidepressants more slowly than Caucasians and Asians,” wrote Sally Satel, a psychiatrist and resident scholar at the American Enterprise Institute for Public Policy Research (Washington, DC, USA; Satel, 2002). She is certainly not alone: from 1995 to 1998, the prescription information for 15 new drug products contained a statement about their differing effectiveness depending on race (Tate & Goldstein, 2004). Claims have been made in peer-reviewed journals that at least 29 medications have racial or ethnic differences in safety or efficacy, although most of these claims are controversial (Tate & Goldstein, 2004). In June 2005, the US Food and Drug Administration (FDA; Rockville, MD) approved the first drug labelled for a racially identified population: BiDil® (NitroMed; Lexington, MA, USA) for the treatment of chronic heart failure in African-Americans. And in July 2005, the European Patent Office (Munich, Germany) renewed a patent for the BRCA2 gene test “for diagnosing a predisposition to breast cancer in Ashkenazi-Jewish women”, because mutations of this gene are frequently found in that population.

Race has been a dominant concept in our society for centuries and has concerned scientists as much as anyone else. However, most scientists agree that race makes a bad scientific concept (Rotimi, 2004; Tishkoff & Kidd, 2004; Wilson et al, 2001). Not only do humans share 99.9% of their genetic makeup, regardless of race, but most of the variants that occur in the remaining 0.01% of the genome—although they vary between individuals—are shared between whole populations. Moreover, a social or physiological concept of race is not the best strategy to identify underlying genetic differences. Relying on a few obvious characteristics, such as skin colour or eye shape, tells something, but not everything, about a person's ancestry. “It doesn't mean that if we look at these various groups as we define them now, that some of these genetic variants are not going to be more frequent in some groups, but there is overlap,” said Charles Rotimi, a biochemist and genetic epidemiologist at the National Human Genome Center at Howard University (Washington, DC, USA). “What is important to recognize however, is that genetic variants cut across social demographic groups in ways that make it difficult for us to consistently say who is black or Hispanic, for example.”

Although it may be reasonable to use the occurrence of genetic variants in medical or scientific settings, using race as a category in an uncritical way, and interpreting it as reflecting distinct groups, may lead to incorrect conclusions. “I've been trying to track how ‘Asian' becomes a category in some of the conclusions in scientific articles. When you go back and try to find out what exactly has been used, you may find for example that 90% of the samples come from one specific Chinese population and then is extrapolated to the broader category of Asian,” commented Sandra Soo-Jin Lee, an anthropologist at the Center for Biomedical Ethics at the Stanford University School of Medicine (Palo Alto, CA, USA). The interpretation that genetic variations in one small group of people with Asian origin automatically reflect the genetic makeup of the whole continent's inhabitants is definitely misleading and just one example of a flawed use of race.

There is no question that people differ from one another, but the interpretation of how they are different and whether certain differences cluster with race has a long history of prejudice, discrimination and even genocide. Race as a criterion in science and medicine is therefore frequently met with scepticism, and scientists expect that race will eventually become obsolete once the underlying genetic mechanisms are identified (Duster, 2005; Lee, 2005). But will it? Will there be personalized medicine and will it help to eradicate racial thinking in medicine? “My fear is that the coarse way we are looking at and using the information on genetic variation may make it more difficult to realize the promise of pharmacogenomics or in some cases really compromise it,” said Rotimi.

Large-scale genome projects to map genetic variances in and across populations will inevitably change the current view of race in both science and medicine. Of these projects, the International HapMap Project is the most comprehensive, sampling data from four populations with African, Asian and European ancestry to map single nucleotide polymorphisms (SNPs) and define blocks of SNPs that are usually inherited together—so-called haplotypes (Altshuler et al, 2005). The HapMap Project might provide the knowledge to better understand mankind's historical patterns of sequence variation and to identify genes that are responsible for differential drug responsiveness or predisposition to diseases. However, although insights from the HapMap could have the potential to render obsolete the use of race as a proxy for genetic variance, there are some obstacles in its way.

The success of the HapMap depends on whether most common diseases are caused by a small set of variations that occur relatively often in the human population. With the data available, the HapMap can detect haplotypes with a frequency of 5% or higher. But if several, rare variants contribute to a disease or drug responsiveness, this resolution is not sufficient to detect them (Rotimi, 2004; Couzin, 2004). In addition to genetics, social factors will contribute to the pace at which racial medicine is left behind—if at all. Race is such a prominent concept in our society that it has a strong impact on scientists and their interpretation of data. As long as racial thinking influences science, science will not necessarily be the best venue to overcome racial thinking and profiling.

Depending on how the data of the HapMap are used, they may reinforce racial thinking, rather than eliminate it, because the HapMap and other genome projects provide the tools to search for genetic differences between races (Lee, 2005). “If you look for differences, you find differences. What becomes the central question is whether a difference is significant,” said Lee. The danger is that scientists might uncover genetic differences with disproportional frequency between different groups and then incorrectly extend these results to broader racial groups; thereby strengthening the misconception that race is a biological category. This could have an impact on how differences in health status between racial groups are interpreted. If the idea of inherent genetic factors is reinforced, non-genetic factors—such as discrimination or racism in clinical encounters, social status, or environmental or behavioural factors—might be ignored.

The question then is whether race as a category should be eliminated completely, from science as well as society. But this may not be the best approach either: paradoxically, ‘race' is required as a concept to overcome the disparities in health status that it once helped to create. “If we eliminate race, how are we going to track differences such as access to health care, discrimination in the clinical setting and exposure to harmful environmental agents that often travel with race?” asked Lee. One of the public health policy goals in the USA is to eliminate differences in populations' health status. But if clinical trials include only Caucasian males—as they once did—drug development will be tailored to this group and will put other groups at a disadvantage (Lee, 2003). Furthermore, databases for genetic variations might be biased towards populations of European descent, which could have consequences for research on disease susceptibility and drug responses that are more frequent among African-Americans (Carlson et al, 2003). To address such inequalities, the US government enacted the National Institutes of Health Revitalization Act in 1993, and in January 2003, the FDA issued guidelines advocating the collection of race and ethnicity data in all clinical trials. Accordingly, the HapMap and other genome projects must sample data from different populations to avoid such disparities. How to do this while avoiding the dangers of racial stereotyping is a tricky issue.

In June 2005, the FDA approved BiDil for the treatment of chronic heart failure in “self-identified black patients”, thus setting a precedent for approving drugs exclusively for a racially identified population. The procedure that led to BiDil's approval reflects how racial thinking is still part of the zeitgeist and how market incentives facilitate classification into racial groups. BiDil did not start out as a racially targeted drug. Two studies carried out in the 1980s by Jay N. Cohn, a cardiologist at the University of Minnesota (Minneapolis, MN, USA), and his colleagues found some beneficial effects from the drug for people suffering heart failure (Cohn et al, 1986, 1991). But the FDA denied its approval in 1997 owing to statistical uncertainties in the trial data (Bloche, 2004; Kahn, 2004). It was soon after this rejection that BiDil was reborn as a race-specific drug. While re-analysing the data from the previous clinical trials, Cohn's team found that BiDil was more effective in black patients (Carson et al, 1999). Another study carried out in self-identified black patients only—which found that the medication decreased their death rates by 43% (Taylor et al, 2004)—finally led to the FDA's approval.

It is not clear why race was chosen in the search for a subgroup that responded better to the medication; any other criterion, such as the patient's medical history, ventricular geometry or hormonal markers, would have seemed more appropriate from a scientific point of view (Kahn, 2004). But in retrospect, race produced a commercial benefit—extended patent protection and FDA approval—even without a biological reason for higher efficacy of BiDil in one race over another. Critics fear that if it is so easy to market a drug by using race as a surrogate marker for a presumed genomic difference, the incentive for drug companies to sponsor research aimed at explaining relevant genetic variations might be reduced (Bloche, 2004).

With FDA approval for race-specific use, non-African Americans can only receive BiDil off-label. “This practice should be unacceptable and serious attempts have to be made to adequately characterize individuals who are likely to benefit from this drug,” commented Rotimi. BiDil is the first medicine for which the cost depends on the consumer's skin colour, but this distributive injustice is not the only problem. As long as there is no understanding of why BiDil helps some people and not others, it will be difficult to identify patients whose lives could be saved by the drug, and its use will not be optimal. “If we truly want to make sure that all humans profit from the genomic revolution, we will have to understand the underlying biology,” said Rotimi. But as long as companies are not encouraged to search for the biological reasons for differential drug reactions, scientific progress will be slow.

Racial thinking hinders scientific progress in many ways. Researchers might ask incorrect questions and define populations in clinical trials in an uncritical way. Data on genetic variation could be used to confirm racial prejudice rather than overcome it. “It is important to think about how scientists are affected by ideas of what race means and how that becomes incorporated into their research,” said Lee.

But even if research progresses and the genetic basis for differential drug reactions or predisposition to diseases are uncovered, these data might not be used to overcome racism in medicine. In fact, the renewal of the patent for the BRCA2 gene test for Ashkenazi-Jewish women shows how genetic data can be used to discriminate people from a particular ancestral group. Alterations in the BRCA2 gene, which are known to enhance breast cancer risk, frequently occur in Ashkenazi-Jews—but with the patent regulation, those who most need the test would pay, whereas it is free for others (Abbott, 2005). More research is needed to overcome racially defined medicine, but the results must also be interpreted and used in the right way. “Those who believe that we will get rid of race through the study of individual genetic variants tend to think that science happens in a vacuum. I have a different perspective. I think that science is deeply embedded by the social values and the historical and political environment in which it is conducted,” said Lee.