The Use of Racial, Ethnic, and Ancestral Categories in Human Genetics Research

Abstract

The global dispersal of anatomically modern humans over the past 100,000 years has produced patterns of phenotypic variation that have exerted—and continue to exert—powerful influences on the lives of individuals and the experiences of groups. The recency of our common ancestry and continued gene flow among populations have resulted in less genetic differentiation among geographically distributed human populations than is observed in many other mammalian species. Nevertheless, differences in appearance have contributed to the development of ideas about “race” and “ethnicity” that often include the belief that significant inherited differences distinguish humans. The use of racial, ethnic, and ancestral categories in genetics research can imply that group differences arise directly through differing allele frequencies, with little influence from socially mediated mechanisms. At the same time, careful investigations of the biological, environmental, social, and psychological attributes associated with these categories will be an essential component of cross-disciplinary research into the origins, prevention, and treatment of common diseases, including those diseases that differ in prevalence among groups.

Introduction

Human genetics research is generating unprecedented amounts of data about the genetic differences among individuals and groups. Investigation of these differences will transform our understanding of the origins and nature of human diseases (Collins et al. 2003).

Research into human genetic differences also has the potential to generate great controversy. In the past, concepts drawn from genetics have been used—both by geneticists and by individuals outside the field—to justify and perpetuate racial and ethnic discrimination (Kevles 1985; Provine 1986). The belief that racial and ethnic groups have substantial, well-demarcated biological differences and that these differences are important has contributed to many of the great atrocities of the 20th century and continues to shape personal interactions and social institutions (Mosse 1985; Shipler 1997). Because of the history of misuse of genetics ideas, geneticists have a special responsibility to examine carefully their use of racial and ethnic categories in their research. Investigations that fail to recognize and acknowledge the full range of mechanisms through which designations of race, ethnicity, and ancestry can correlate with personal traits and health outcomes threaten to reinforce widely held stereotypes. Yet genetics research also has the potential, by delineating the complex origins of traits and the close biological affinities between human groups, to help dispel these stereotypes.

The sequencing of the human genome (International Human Genome Sequencing Consortium 2001; Venter et al. 2001) and the ongoing international effort to catalog common haplotypes in several populations (International HapMap Consortium 2003) make this an opportune time to examine the complex relationships between genetics research and the categories of race, ethnicity, and ancestry. Although such a review inevitably draws on very different academic disciplines and literatures, a cross-disciplinary conversation is essential for reconciliation of the promise of genetics research with the historical and potential abuses of ideas drawn from genetics. This review summarizes what is known about patterns of human genetic variation; the historical development of widely held conceptions about race, ethnicity, and ancestry; and the interactions between these conceptions and human genetics research.

The Origins, Patterns, and Physical Manifestations of Human Genetic Variation

The Origins of Modern Humans

Information about the history of our species comes from two main sources: the paleoanthropological record and historical inferences based on current genetic differences observed in humans. Although both sources of information are fragmentary, they have been converging in recent years on the same general story.

The existing fossil evidence suggests that anatomically modern humans evolved in Africa, within the last ∼200,000 years, from a pre-existing population of humans (Klein 1999). Although it is not easy to define “anatomically modern” in a way that encompasses all living humans and excludes all archaic humans (Lieberman et al. 2002), the generally agreed-upon physical characteristics of anatomical modernity include a high rounded skull, facial retraction, and a light and gracile, as opposed to heavy and robust, skeleton (Lahr 1996). Early fossils with these characteristics have been found in eastern Africa and have been dated to ∼160,000–200,000 years ago (White et al. 2003; McDougall et al. 2005). At that time, the population of anatomically modern humans appears to have been small and localized (Harpending et al. 1998). Much larger populations of archaic humans lived elsewhere in the Old World, including the Neandertals in Europe and an earlier species of humans, Homo erectus, in Asia (Swisher et al. 1994).

Fossils of the earliest anatomically modern humans found outside Africa are from two sites in the Middle East and date to a period of relative global warmth, ∼100,000 years ago, though this region was reinhabited by Neandertals in later millennia as the climate in the northern hemisphere again cooled (Lahr and Foley 1998). Groups of anatomically modern humans appear to have moved outside Africa permanently sometime >60,000 years ago. One of the earliest modern skeletons found outside Africa is from Australia and has been dated to ∼42,000 years ago (Bowler et al. 2003), although studies of environmental changes in Australia argue for the presence of modern humans in Australia >55,000 years ago (Miller et al. 1999). To date, the earliest anatomically modern skeleton discovered from Europe comes from the Carpathian Mountains of Romania and is dated to 34,000–36,000 years ago (Trinkaus et al. 2003).

Existing data on human genetic variation support and extend conclusions based on the fossil evidence. African populations exhibit greater genetic diversity than do populations in the rest of the world, implying that humans appeared first in Africa and later colonized Eurasia and the Americas (Tishkoff and Williams 2002; Yu et al. 2002; Tishkoff and Verrelli 2003). The genetic variation seen outside Africa is generally a subset of the variation within Africa, a pattern that would be produced if the migrants from Africa were limited in number and carried just part of African genetic variability with them (Cavalli-Sforza and Feldman 2003). Patterns of genetic variation suggest an earlier population expansion in Africa followed by a subsequent expansion in non-African populations, and the dates calculated for the expansions generally coincide with the archaeological record (Jorde et al. 1998).

Aspects of the relationship between anatomically modern and archaic humans remain contentious. Studies of mtDNA (Ingman et al. 2000), the Y chromosome (Underhill et al. 2000), portions of the X chromosome (Kaessmann et al. 1999), and many (though not all) autosomal regions (Harpending and Rogers 2000) support the “Out of Africa” account of human history, in which anatomically modern humans appeared first in eastern Africa and then migrated throughout Africa and into the rest of the world, with little or no interbreeding between modern humans and the archaic populations they gradually replaced (Tishkoff et al. 2000; Stringer 2002). However, several groups of researchers cite fossil and genetic evidence to argue for a more complex account. They contend that humans bearing modern traits emerged several times from Africa, over an extended period, and mixed with archaic humans in various parts of the world (Hawks et al. 2000; Eswaran 2002; Templeton 2002; Ziętkiewicz et al. 2003). As a result, they say, autosomal DNA from archaic human populations living outside Africa persists in modern populations, and modern populations in various parts of the world still bear some physical resemblance to the archaic populations that inhabited those regions (Wolpoff et al. 2001).

However, distinguishing possible contributions to the gene pool of modern humans from archaic humans outside Africa is difficult, especially since many autosomal loci coalesce at times preceding the separation of archaic human populations (Pääbo 2003). In addition, studies of mtDNA from archaic and modern humans and extant Y chromosomes suggest that any surviving genetic contributions of archaic humans outside Africa must be small, if they exist at all (Krings et al. 1997; Nordborg 1998; Takahata et al. 2001; Serre et al. 2004). The observation that most genes studied to date coalesce in African populations points toward the importance of Africa as the source of most modern genetic variation, perhaps with some subdivision in the ancestral African population (Satta and Takahata 2002). Sequence data for hundreds of loci from widely distributed worldwide populations eventually may clarify the population processes associated with the appearance of anatomically modern humans (Wall 2000), as well as the amount of gene flow among modern humans since then.

The Distribution of Variation

A thorough description of the differences in patterns of genetic variation between humans and other species awaits additional genetic studies of human populations and nonhuman species. But the data gathered to date suggest that human variation exhibits several distinctive characteristics. First, compared with many other mammalian species, humans are genetically less diverse—a counterintuitive finding, given our large population and worldwide distribution (Li and Sadler 1991; Kaessmann et al. 2001). For example, the chimpanzee subspecies living just in central and western Africa have higher levels of diversity than do humans (Ebersberger et al. 2002; Yu et al. 2003; Fischer et al. 2004).

The distribution of variants within and among human populations also differs from that of many other species. The details of this distribution are impossible to describe succinctly because of the difficulty of defining a “population,” the clinal nature of variation, and heterogeneity across the genome (Long and Kittles 2003). In general, however, 5%–15% of genetic variation occurs between large groups living on different continents, with the remaining majority of the variation occurring within such groups (Lewontin 1972; Jorde et al. 2000a; Hinds et al. 2005). This distribution of genetic variation differs from the pattern seen in many other mammalian species, for which existing data suggest greater differentiation between groups (Templeton 1998; Kittles and Weiss 2003).

Our history as a species also has left genetic signals in regional populations. For example, in addition to having higher levels of genetic diversity, populations in Africa tend to have lower amounts of linkage disequilibrium than do populations outside Africa, partly because of the larger size of human populations in Africa over the course of human history and partly because the number of modern humans who left Africa to colonize the rest of the world appears to have been relatively low (Gabriel et al. 2002). In contrast, populations that have undergone dramatic size reductions or rapid expansions in the past and populations formed by the mixture of previously separate ancestral groups can have unusually high levels of linkage disequilibrium (Nordborg and Tavare 2002).

Many other geographic, climatic, and historical factors have contributed to the patterns of human genetic variation seen in the world today. For example, population processes associated with colonization, periods of geographic isolation, socially reinforced endogamy, and natural selection all have affected allele frequencies in certain populations (Jorde et al. 2000b; Bamshad and Wooding 2003). In general, however, the recency of our common ancestry and continual gene flow among human groups have limited genetic differentiation in our species.

Substructure in the Human Population

Although the genetic differences among human groups are relatively small, these differences nevertheless can be used to situate many individuals within broad, geographically based groupings. For example, computer analyses of hundreds of polymorphic loci sampled in globally distributed populations have revealed the existence of genetic clustering that roughly is associated with groups that historically have occupied large continental and subcontinental regions (Rosenberg et al. 2002; Bamshad et al. 2003).

Some commentators have argued that these patterns of variation provide a biological justification for the use of traditional racial categories. They argue that the continental clusterings correspond roughly with the division of human beings into sub-Saharan Africans; Europeans, western Asians, and northern Africans; eastern Asians; Polynesians and other inhabitants of Oceania; and Native Americans (Risch et al. 2002). Other observers disagree, saying that the same data undercut traditional notions of racial groups (King and Motulsky 2002; Calafell 2003; Tishkoff and Kidd 2004). They point out, for example, that major populations considered races or subgroups within races do not necessarily form their own clusters. Thus, samples taken from India and Pakistan affiliate with Europeans or eastern Asians rather than separating into a distinct cluster. However, samples from the Kalash, a small population living in northwestern Pakistan, form their own cluster on a level comparable with those of the major continental regions (Rosenberg et al. 2002).

Sampling design can have a critical influence on the results of such studies. Studies of genetic clustering often have relied on samples taken from widely separated and socially defined populations. When samples were analyzed from individuals who were more evenly distributed geographically, clustering was far less evident (Serre and Pääbo 2004). Furthermore, because human genetic variation is clinal, many individuals affiliate with two or more continental groups. Thus, the genetically based “biogeographical ancestry” assigned to any given person generally will be broadly distributed and will be accompanied by sizable uncertainties (Pfaff et al. 2004).

In many parts of the world, groups have mixed in such a way that many individuals have relatively recent ancestors from widely separated regions. Although genetic analyses of large numbers of loci can produce estimates of the percentage of a person’s ancestors coming from various continental populations (Shriver et al. 2003; Bamshad et al. 2004), these estimates may assume a false distinctiveness of the parental populations, since human groups have exchanged mates from local to continental scales throughout history (Cavalli-Sforza et al. 1994; Hoerder 2002). Even with large numbers of markers, information for estimating admixture proportions of individuals or groups is limited, and estimates typically will have wide CIs (Pfaff et al. 2004).

Physical Variation in Humans

The distribution of many physical traits resembles the distribution of genetic variation within and between human populations (American Association of Physical Anthropologists 1996; Keita and Kittles 1997). For example, ∼90% of the variation in human head shapes occurs within every human group, and ∼10% separates groups, with a greater variability of head shape among individuals with recent African ancestors (Relethford 2002).

A prominent exception to the common distribution of physical characteristics within and among groups is skin color. Approximately 10% of the variance in skin color occurs within groups, and ∼90% occurs between groups (Relethford 2002). This distribution of skin color and its geographic patterning—with people whose ancestors lived predominantly near the equator having darker skin than those with ancestors who lived predominantly in higher latitudes—indicate that this attribute has been under strong selective pressure. Darker skin appears to be strongly selected for in equatorial regions to prevent sunburn, skin cancer, the photolysis of folate, and damage to sweat glands (Sturm et al. 2001; Rees 2003). A leading hypothesis for the selection of lighter skin in higher latitudes is that it enables the body to form greater amounts of vitamin D, which helps prevent rickets (Jablonski 2004). However, the vitamin D hypothesis is not universally accepted (Aoki 2002), and lighter skin in high latitudes may correspond simply to an absence of selection for dark skin (Harding et al. 2000).

Because skin color has been under strong selective pressure, similar skin colors can result from convergent adaptation rather than from genetic relatedness. Sub-Saharan Africans, tribal populations from southern India, and Australian Aborigines have similar skin pigmentation, but genetically they are no more similar than are other widely separated groups. Furthermore, in some parts of the world in which people from different regions have mixed extensively, the connection between skin color and ancestry has been substantially weakened (Parra et al. 2004). In Brazil, for example, skin color is not closely associated with the percentage of recent African ancestors a person has, as estimated from an analysis of genetic variants differing in frequency among continent groups (Parra et al. 2003).

Considerable speculation has surrounded the possible adaptive value of other physical features characteristic of groups, such as the constellation of facial features observed in many eastern and northeastern Asians (Guthrie 1996). However, any given physical characteristic generally is found in multiple groups (Lahr 1996), and demonstrating that environmental selective pressures shaped specific physical features will be difficult, since such features may have resulted from sexual selection for individuals with certain appearances or from genetic drift (Roseman 2004).

The Social Interpretation of Physical Variation

The Development of the “Ideology of Race”

Given our visual acuity and complex social relationships, humans presumably have always observed and speculated about the physical differences among individuals and groups. But different societies have attributed markedly different meanings to these distinctions. Classical civilizations from Rome to China tended to invest much more importance in family or tribal affiliations than in physical appearance (Dikötter 1992; Goldenberg 2003). Some Roman writers adhered to an environmental determinism in which climate could affect the appearance and character of groups (Isaac 2004). But in many ancient civilizations, individuals with widely varying physical appearances could become full members of a society by growing up within that society or by adopting the society’s cultural norms (Snowden 1983; Lewis 1990).

The English word “race” (possibly derived from the Spanish raza, meaning “breed” or “stock”), along with many of the ideas now associated with the term, were products of the European era of exploration (Smedley 1999). As Europeans encountered people from different parts of the world, they speculated about the physical, social, and cultural differences between human groups. The rise of the African slave trade, which gradually displaced an earlier trade in slaves from throughout the world, created a further incentive to categorize human groups to justify the barbarous treatment of African slaves (Meltzer 1993). Drawing on classical sources and on their own internal interactions—for example, the hostility between the English and Irish was a powerful influence on early thinking about the differences between people (Takaki 1993)—Europeans began to sort themselves and others into groups associated with physical appearance and with deeply ingrained behaviors and capacities. A set of “folk beliefs” took hold that linked inherited physical differences between groups to inherited intellectual, behavioral, and moral qualities (Banton 1977). Although similar ideas can be found in other cultures (Lewis 1990; Dikötter 1992), they appear not to have had as much influence on social structures as they did in Europe and the parts of the world colonized by Europeans.

In the 18th century, the differences between human groups became a focus of scientific investigation (Todorov 1993). Initially, scholars focused on cataloging and describing “The Natural Varieties of Mankind,” as Johann Friedrich Blumenbach entitled his 1775 text (which established the five major divisions of humans still reflected in some racial classifications). But as the science of anthropology took shape in the 19th century, European and American scientists increasingly sought explanations for the behavioral and cultural differences they attributed to groups (Stanton 1960). For example, they measured the shapes and sizes of skulls and related the results to group differences in intelligence or other attributes (Lieberman 2001). Both before and after the 1859 publication of On the Origins of Species, a debate raged in Europe over whether different human groups had the same origin or were the product of separate creations or evolutionary lineages (Wolpoff and Caspari 1997).

From the 17th through the 19th centuries, the merging of folk beliefs about group differences with scientific explanations of those differences produced what one scholar has called an “ideology of race” (Smedley 1999). According to this ideology, races are primordial, natural, enduring, and distinct. Some groups might be the result of mixture between formerly distinct populations, but careful study can distinguish the ancestral races that had combined to produce admixed groups.

The concept of race found wide application in many societies. The eugenics movement of the late 19th and early 20th centuries asserted as self-evident the biological inferiority of particular groups (Kevles 1985). In many parts of the world, the idea of race became a way of rigidly dividing groups by use of culture as well as physical appearances (Hannaford 1996). Campaigns of oppression and genocide often used supposed racial differences to motivate inhuman acts against others (Horowitz 2001).

The Incongruities of Racial Classifications

Even as the idea of “race” was becoming a powerful organizing principle in many societies, the shortcomings of the concept were apparent. In the Old World, the gradual transition in appearances from one group to adjacent groups emphasized that “one variety of mankind does so sensibly pass into the other, that you cannot mark out the limits between them,” as Blumenbach observed in his writings on human variation (Marks 1995, p. 54). In parts of the Americas, the situation was somewhat different. The immigrants to the New World came largely from widely separated regions of the Old World—western and northern Europe, western Africa, and, later, eastern Asia and southern Europe. In the Americas, the immigrant populations began to mix among themselves and with the indigenous inhabitants of the continent. In the United States, for example, most people who self-identify as African American have some European ancestors—in one analysis of genetic markers that have differing frequencies between continents, European ancestry ranged from an estimated 7% for a sample of Jamaicans to ∼23% for a sample of African Americans from New Orleans (Parra et al. 1998). Similarly, many people who identify as European American have some African or Native American ancestors, either through openly interracial marriages or through the gradual inclusion of people with mixed ancestry into the majority population. In a survey of college students who self-identified as “white” in a northeastern U.S. university, ∼30% were estimated to have <90% European ancestry (Shriver et al. 2003).

In the United States, social and legal conventions developed over time that forced individuals of mixed ancestry into simplified racial categories (Gossett 1997). An example is the “one-drop rule” implemented in some state laws that treated anyone with a single known African American ancestor as black (Davis 2001). The decennial censuses conducted since 1790 in the United States also created an incentive to establish racial categories and fit people into those categories (Nobles 2000). In other countries in the Americas where mixing among groups was more extensive, social categories have tended to be more numerous and fluid, with people moving into or out of categories on the basis of a combination of socioeconomic status, social class, ancestry, and appearance (Mörner 1967).

Efforts to sort the increasingly mixed population of the United States into discrete categories generated many difficulties (Spickard 1992). By the standards used in past censuses, many millions of children born in the United States have belonged to a different race than have one of their biological parents. Efforts to track mixing between groups led to a proliferation of categories (such as “mulatto” and “octoroon”) and “blood quantum” distinctions that became increasingly untethered from self-reported ancestry. A person’s racial identity can change over time, and self-ascribed race can differ from assigned race (Kressin et al. 2003). Until the 2000 census, Latinos were required to identify with a single race despite the long history of mixing in Latin America; partly as a result of the confusion generated by the distinction, 42% of Latino respondents in the 2000 census ignored the specified racial categories and checked “some other race” (Mays et al. 2003).

Ethnicity as a Way of Categorizing People

As the problems surrounding the word “race” became increasingly apparent during the 20th century, the word “ethnicity” was promoted as a way of characterizing the differences between groups (Huxley and Haddon 1936; Hutchinson and Smith 1996). Ethnicity typically emphasizes the cultural, socioeconomic, religious, and political qualities of human groups rather than their genetic ancestry. It may encompass language, diet, religion, dress, customs, kinship systems, or historical or territorial identity (Cornell and Hartmann 1998).

However, as a way of understanding human groups, ethnicity also suffers from several shortcomings. First, ascribing an ethnic identity to a group can imply a much greater degree of uniformity than is actually the case. In the United States, the ethnic group “Hispanic or Latino” contains such subgroups as Cuban Americans, Mexican Americans, Puerto Ricans, and recent immigrants from Central America (Hayes-Bautista and Chapa 1987). Combining these groups into a single category may serve useful bureaucratic or political ends but does not necessarily result in a better understanding of these groups.

Also, ethnicity, like race, is a malleable concept that can change dramatically in different times or circumstances (Waters 1990; Smelser et al. 2001). Ethnic groups may come into existence and then dissipate as a result of broad historical or social trends. Individuals might change ethnic groups over the course of their lives or identify with more than one group. A researcher, clinician, or government official might assign an ethnicity to an individual quite different from the one that person would acknowledge (Kressin et al. 2003).

Finally, despite attempts to distinguish “ethnicity” from “race,” the two terms often are used interchangeably (Oppenheimer 2001). Ethnic groups can share a belief in a common ancestral origin (Cornell and Hartmann 1998), which also can be a defining characteristic of a racial group. Furthermore, ethnic groups tend to promote marriage within the group, which creates an expectation of biological cohesion regardless of whether that cohesion existed in the past.

Ancestry as a Way of Categorizing People

An alternative to the use of racial or ethnic categories in genetics research is to categorize individuals in terms of ancestry. Ancestry may be defined geographically (e.g., Asian, sub-Saharan African, or northern European), geopolitically (e.g., Vietnamese, Zambian, or Norwegian), or culturally (e.g., Brahmin, Lemba, or Apache). The definition of ancestry may recognize a single predominant source or multiple sources. Ancestry can be ascribed to an individual by an observer, as was the case with the U.S. census prior to 1960; it can be identified by an individual from a list of possibilities or with use of terms drawn from that person’s experience; or it can be calculated from genetic data by use of loci with allele frequencies that differ geographically, as described above. At least among those individuals who participate in biomedical research, genetic estimates of biogeographical ancestry generally agree with self-assessed ancestry (Tang et al. 2005), but in an unknown percentage of cases, they do not (Brodwin 2002; Kaplan 2003).

Despite its seemingly objective nature, ancestry also has limitations as a way of categorizing people (Elliott and Brodwin 2002). When asked about the ancestry of their parents and grandparents, many people cannot provide accurate answers. In one series of focus groups in the state of Georgia, 40% of ∼100 respondents said they did not know one or more of their four grandparents well enough to be certain how that person(s) would identify racially (Condit et al. 2003). Misattributed paternity or adoption can separate biogeographical ancestry from socially defined ancestry. Furthermore, the exponentially increasing number of our ancestors makes ancestry a quantitative rather than qualitative trait—5 centuries (or 20 generations) ago, each person had a maximum of >1 million ancestors (Ohno 1996). To complicate matters further, recent analyses suggest that everyone living today has exactly the same set of genealogical ancestors who lived as recently as a few thousand years in the past, although we have received our genetic inheritance in different proportions from those ancestors (Rohde et al. 2004).

In the end, the terms “race,” “ethnicity,” and “ancestry” all describe just a small part of the complex web of biological and social connections that link individuals and groups to each other.

Racial, Ethnic, and Ancestral Categories in Genetics Research

The Effects of Racial and Ethnic Identities on Health

Racial and ethnic groups can exhibit substantial average differences in disease incidence, disease severity, disease progression, and response to treatment (LaVeist 2002). In the United States, African Americans have higher rates of mortality than does any other racial or ethnic group for 8 of the top 10 causes of death (Hummer et al. 2004). U.S. Latinos have higher rates of death from diabetes, liver disease, and infectious diseases than do non-Latinos (Vega and Amaro 1994). Native Americans suffer from higher rates of diabetes, tuberculosis, pneumonia, influenza, and alcoholism than does the rest of the U.S. population (Mahoney and Michalek 1998). European Americans die more often from heart disease and cancer than do Native Americans, Asian Americans, or Hispanics (Hummer et al. 2004).

Considerable evidence indicates that the racial and ethnic health disparities observed in the United States arise mostly through the effects of discrimination, differences in treatment, poverty, lack of access to health care, health-related behaviors, racism, stress, and other socially mediated forces. The infant mortality rate for African Americans is approximately twice the rate for European Americans, but, in a study that looked at members of these two groups who belonged to the military and received care through the same medical system, their infant mortality rates were essentially equivalent (Rawlings and Weir 1992). Recent immigrants to the United States from Mexico have better indicators on some measures of health than do Mexican Americans who are more assimilated into American culture (Franzini et al. 2001). Diabetes and obesity are more common among Native Americans living on U.S. reservations than among those living outside reservations (Cooper et al. 1997). Rates of heart disease among African Americans are associated with the segregation patterns in the neighborhoods where they live (Fang et al. 1998). Furthermore, the risks for many diseases are elevated for socially, economically, and politically disadvantaged groups in the United States, suggesting that socioeconomic inequities are the root causes of most of the differences (Cooper et al. 2003; Cooper 2004).

However, differences in allele frequencies certainly contribute to group differences in the incidence of some monogenic diseases, and they may contribute to differences in the incidence of some common diseases (Risch et al. 2002; Burchard et al. 2003; Tate and Goldstein 2004). For the monogenic diseases, the frequency of causative alleles usually correlates best with ancestry, whether familial (for example, Ellis–van Creveld syndrome among the Pennsylvania Amish), ethnic (Tay-Sachs disease among Ashkenazi Jewish populations), or geographical (hemoglobinopathies among people with ancestors who lived in malarial regions). To the extent that ancestry corresponds with racial or ethnic groups or subgroups, the incidence of monogenic diseases can differ between groups categorized by race or ethnicity, and health-care professionals typically take these patterns into account in making diagnoses.

Even with common diseases involving numerous genetic variants and environmental factors, investigators point to evidence suggesting the involvement of differentially distributed alleles with small to moderate effects. Frequently cited examples include hypertension (Douglas et al. 1996), diabetes (Gower et al. 2003), obesity (Fernandez et al. 2003), and prostate cancer (Platz et al. 2000). However, in none of these cases has allelic variation in a susceptibility gene been shown to account for a significant fraction of the difference in disease prevalence among groups, and the role of genetic factors in generating these differences remains uncertain (Mountain and Risch 2004).

The Allelic Architecture of Disease

The genetic architecture of common diseases is an important factor in determining the extent to which patterns of genetic variation influence group differences in health outcomes (Reich and Lander 2001; Pritchard and Cox 2002; Smith and Lusis 2002). According to the common disease/common variant hypothesis, common variants present in the ancestral population before the dispersal of modern humans from Africa play an important role in human diseases (Goldstein and Chikhi 2002). Genetic variants associated with Alzheimer disease, deep venous thrombosis, Crohn disease, and type 2 diabetes appear to adhere to this model (Lohmueller et al. 2003). However, the generality of the model has not yet been established and, in some cases, is in doubt (Weiss and Terwilliger 2000; Pritchard and Cox 2002; Cardon and Abecasis 2003). Some diseases, such as many common cancers, appear not to be well described by the common disease/common variant model (Kittles and Weiss 2003; Wiencke 2004).

Another possibility is that common diseases arise in part through the action of combinations of variants that are individually rare (Pritchard 2001; Cohen et al. 2004). Most of the disease-associated alleles discovered to date have been rare, and rare variants are more likely than common variants to be differentially distributed among groups distinguished by ancestry (Risch et al. 2002; Kittles and Weiss 2003). However, groups could harbor different, though perhaps overlapping, sets of rare variants, which would reduce contrasts between groups in the incidence of the disease.

The number of variants contributing to a disease and the interactions among those variants also could influence the distribution of diseases among groups. The difficulty that has been encountered in finding contributory alleles for complex diseases and in replicating positive associations suggests that many complex diseases involve numerous variants rather than a moderate number of alleles, and the influence of any given variant may depend in critical ways on the genetic and environmental background (Risch 2000; Weiss and Terwilliger 2000; Altmüller et al. 2001; Hirschhorn et al. 2002). If many alleles are required to increase susceptibility to a disease, the odds are low that the necessary combination of alleles would become concentrated in a particular group purely through drift (Cooper 2004).

Population Substructure in Genetics Research

One area in which racial and ethnic categories can be important considerations in genetics research is in controlling for confounding between population substructure, environmental exposures, and health outcomes. Association studies can produce spurious results if cases and controls have differing allele frequencies for genes that are not related to the disease being studied (Cardon and Palmer 2003; Marchini et al. 2004), although the magnitude of this problem in genetic association studies is subject to debate (Thomas and Witte 2002; Wacholder et al. 2002). Various methods have been developed to detect and account for population substructure (Morton and Collins 1998; Hoggart et al. 2003), but these methods can be difficult to apply in practice (Freedman et al. 2004).

Population substructure also can be used to advantage in genetic association studies. For example, populations that represent recent mixtures of geographically separated ancestral groups can exhibit longer-range linkage disequilibrium between susceptibility alleles and genetic markers than is the case for other populations (Hoggart et al. 2004; Patterson et al. 2004; Smith et al. 2004; McKeigue 2005). Genetic studies can use this admixture linkage disequilibrium to search for disease alleles with fewer markers than would be needed otherwise. Association studies also can take advantage of the contrasting experiences of racial or ethnic groups, including migrant groups, to search for interactions between particular alleles and environmental factors that might influence health (Chaturvedi 2001; Collins et al. 2003).

Conclusions

When deciding whether and how to use racial, ethnic, and ancestral categories in research, geneticists face conflicting demands. On the one hand, many observers have made powerful arguments in favor of reducing or even eliminating the use of racial or ethnic categories in genetics research (Fullilove 1998; Goodman 2000; Lee et al. 2001; Braun 2002; Duster 2003, 2005; Stevens 2003; Kahn 2004; Sankar et al. 2004; Ossorio and Duster 2005). These observers point out that the use of these categories reinforces the widespread impression that health inequities arise through the action of genetic differences and independent of socially mediated mechanisms. In this way, genetics research that involves making population comparisons can inaccurately stereotype racial and ethnic groups, both by implying that such groups are clearly delineated and by associating health outcomes with all individuals in those groups rather than with only those individuals who exhibit the outcome. Furthermore, according to critics, an overemphasis on the genetic component of health differences shifts attention and resources away from established contributors to health disparities—in particular, the differences in treatment and socioeconomic disadvantages that disproportionately affect minority groups (Sankar et al. 2004). Genetics research offers no evidence that any one group is superior or inferior to any other, although some individuals continue to try to distort genetic findings to buttress prejudiced outlooks. Biomedical research that accentuates genetic differences among groups, say critics of this research, is as conceptually flawed as the race science of the 19th century (Bhopal 1997).

On the other hand, race and ethnicity are such prominent aspects of many societies that it is difficult, and often inadvisable, to ignore them in genetics research. The members of these groups can have widely disparate economic, social, and psychological experiences and can be exposed to very different environments as a consequence of their membership in a particular group. These differential experiences and environmental exposures can be used to investigate the biological mechanisms that contribute to health disparities among groups (LaVeist 1996). In addition, self-identified race, ethnicity, or ancestry can provide measures of population substructure that help avoid false-positive results in association studies.

One way for geneticists to ease the dilemma they face is to try to move beyond racial, ethnic, or ancestral categories in their work (Ota Wang and Sue 2005; Shields et al. 2005). Rather than using racial, ethnic, or ancestral labels as proxies for much more detailed social, economic, environmental, biological, or genetic factors, researchers can try to measure these factors directly. For example, controlling for socioeconomic status by use of census tract data can substantially reduce the excess mortality risk observed in disadvantaged minority populations (Krieger et al. 2005). Similarly, genotyping to estimate biogeographical ancestry can be a better control for population substructure than self-identified race, ethnicity, or ancestry (Shields et al. 2005).

When the use of racial or ethnic categories in research is deemed necessary, researchers can avoid overgeneralization by using labels that are as specific as possible. Today many genetic investigations label populations with the same loose terms used by the public (Sankar and Cho 2002; Clayton 2003; Collins 2004; Comstock et al. 2004). But labels such as “Hispanic,” “Black,” “Mexican American,” “White,” “Asian,” “European,” or “African” can have ambiguous or contradictory meanings among researchers, research subjects, and the general public. Use of such broad labels without careful definitions can impair scientific understanding and imply that distinctions between socially defined populations are genetically well established. Genetics researchers often rely on the categories specified in the U.S. census—encouraged by regulations that urge diversity of study populations—but these categories are used today mainly for administrative and social purposes and were not designed for genetics research. Even when the census categories are used to select research subjects to ensure diversity, researchers can analyze their results using more-specific labels that are closely tied to the scientific questions being asked (Kaplan and Bennett 2003). For example, labels based on biogeographical ancestry may be suited for many genetics studies, socially based labels may be more appropriate for health disparities and clinical research, and both types of information may be valuable for studies of some gene-environment interactions.

Individuals can be assigned to specific population categories in a number of ways, with the most appropriate way, again, depending on the research question being investigated (Foster and Sharp 2004; International HapMap Consortium 2004). Research subjects can be asked to identify themselves with geographical or cultural populations, which may be defined by the researcher or by the local communities within which the research is being conducted. Communities and researchers can choose categories together through a consultative or engagement process between researchers and the community (Foster et al. 1999; Condit et al. 2002). Categorical systems also can include the possibility of simultaneous, multiple-group memberships in groups at higher or lower levels of organization.

A number of journals, including Nature Genetics (Anonymous 2000), Archives of Pediatrics & Adolescent Medicine (Rivara and Finberg 2001), and the British Medical Journal (Anonymous 1996), have separately issued guidelines stating that researchers should carefully define the terms they use for populations, and some journals have asked researchers to justify their use of racial or ethnic groups in research. But enforcement of these guidelines has been uneven, and compliance will continue to be spotty without greater awareness among researchers of the difficulties and risks involved in defining populations (Sankar and Cho 2002; Anonymous 2004).

Efforts to move past the use of racial and ethnic categories in genetics research often will require consideration of a very broad range of additional variables (Chakravarti and Little 2003). These variables will differ from study to study, but even a partial list includes racism and discrimination, socioeconomic status, social class, personal or family wealth, environmental exposures, insurance status, age, diet and nutrition, health beliefs and practices, educational level, language spoken, religion, tribal affiliation, country of birth, parents’ country of birth, length of time in the country of residence, and place of residence along with genetic variation (Kaplan and Bennett 2003). Research that successfully integrates such a wide range of variables will require the collaboration of individuals with many different disciplinary backgrounds (Bonham et al. 2005).

A particular challenge for interdisciplinary teams will be designing their studies and reporting their results in ways that convey to the public the complexities of biological systems (Weiss 1998; Clark 2002; Chakravarti and Little 2003). Within the highly interconnected network of factors involved in complex diseases, the influence of any given allele likely will depend on past and current biological and environmental contexts, which often will make it difficult to demonstrate that a given variant directly “causes” a particular condition (Weatherall 1999; Page et al. 2003). Growing appreciation of the ways in which gestational influences (Sallout and Walker 2003), childhood illnesses (Gluckman and Hanson 2004), obesity (Calle and Kaaks 2004), exposure to toxins (Whyatt et al. 2004), stress (Wallace et al. 2004), and other factors influence later illnesses highlights the multiple interconnections among biological mechanisms, environmental influences, and chance events (Shostak 2003).

Despite this complexity, genetics researchers have a unique opportunity to reduce at least some of the confusion and controversy surrounding the issues of race, ethnicity, ancestry, and health. They can demonstrate the irrelevance of racial and ethnic labels for pursuing many research questions and health improvement objectives—for example, by clarifying the many ways in which environmental factors that extend across groups interact with biological processes to produce common diseases (Lin and Kelsey 2000; Rotimi 2004). By emphasizing the close genetic affinities between members of different groups, researchers can reduce the widespread misconception that substantial genetic differences separate groups (Wilson et al. 2001; Olson 2002; Jorde and Wooding 2004). As the complex origins of human traits, behaviors, and diseases slowly are unraveled, how genetics research is conducted could influence whether racial and ethnic discrimination increases or decreases over time.