Hasty generalisation and exaggerated certainties: reporting genetic findings in health disparities research

Abstract

In the United States, research that examines potential genetic contributions to health disparities often relies on racial categories. Some see benefit in this research especially for conditions where disparities in health status seem strongly associated with racial identity, such as heart disease and prostate cancer. But this research calls for close scrutiny. First, despite common optimism about genetic research, it may not be the most productive way to examine health disparities. And second, this research has the potential to contribute to racial stereotypes, arguably a prime cause of the health disparities the genetic research actually seeks to ameliorate. Two articles reporting results about genetics and heart disease are used to illustrate these concerns. Both report strong correlations between increased vulnerability to heart disease and black racial identity. Despite serious sampling and analysis problems in these articles, the findings rapidly entered the scientific and popular literature. A possible reason for their ready acceptance is their congruence with stereotypes that attribute poor health and genetic inferiority to minority US populations.

Point of departure: racial stereotyping and health disparities

Disparities in health status persist in nearly all nations, and most everywhere poverty is the root cause. In the United States, however, while poverty is a consistent predictor of poor health (Gwatkin, Guillot & Heuveline, 1999; World Health Organization, 2001), research that controls for economic status shows that people of color receive health care inferior to that given to whites (with only limited exceptions (Palloni & Morenoff, 2001)) (Gee, 2002; Geiger, 2003; Hogue, Hargraves & Collins, 2000; Smedley et al., 2003). In an effort to explain the persistence of these disparities, the United States Institute of Medicine recently conducted a thorough review of hundreds of studies about health care delivery and health status in the US. The resulting report found that direct or overt racial discrimination did not explain fully the persistence of these disparities. The report suggested instead that racial stereotyping might provide the important mechanism that perpetuates unequal treatment in the US.

Stereotyping is a thought process that uses social categories, such as age, gender or race, to ‘acquire, process, and recall information about others …. [Stereotypes] help to organize and simplify complex and uncertain situations’ (Balsa & McGuire, 2003; Dovidio, 1999). The capacity to reduce complexity and interpret messages based on prior knowledge is a functional and fundamental feature of human thought and interaction. In this sense, stereotyping is not by definition negative. However, it can sustain social biases when the simplified idea informing the stereotype is itself biased and biased in a way that forestalls genuine interactions that might contradict the stereotype.

Research has shown that beliefs about race constitute a common stereotype relied upon by US physicians and other health care workers, and that practitioners are even more likely to rely on stereotypes in situations that are stressful or that have high cognitive demands, such as when seeing patients. For example, research has shown that a stereotype operative in clinical encounters is that blacks and Hispanics are less likely to comply with treatment and that additional problems such as drug abuse, lack of health insurance, or difficulty scheduling appointments might undercut the success of certain approaches to treatment (van Ryn, 2002). As a result of these sterotypes, physicians sometimes do not offer such treatments or offer them tentatively. Patients can interpret this hesitation in a variety of ways, some possibly leading the patient to distrust the physician's recommendations or to act in ways that confirm the physician's stereotype. Research also has shown that stereotypes are subtle and operate typically in health care without a practitioner being aware of their presence (Rathore et al., 2000). This feature makes racial stereotypes difficult to identify and change, and powerfully influential in personal interactions.

The dogged persistence of racially patterned health disparities in the US, despite at least two decades of serious reform efforts (Epstein & Ayanian, 2001), suggests the need for a broader view of the problem. Here, attention is directed toward how the use of race as a variable in genetic research, particularly research about conditions associated with health disparities, might inadvertently help to perpetuate racial stereotypes and thus contribute to health disparities.

Types of health disparities

Health disparities refers broadly to differences in the incidence and ‘burden of disease and other adverse conditions’ that exist among populations (Green et al., 2003). These disparities can be conceptually distinguished in numerous ways. Here inequalities in health care, which refers to the quality of health care controlling for differences in access or need (Smedley et al., 2003), are distinguished from inequalities in health status, which are actual health outcomes measured as morbidity and mortality.

Disparities in health care (race as a social construct)

Disparities in health care are evident in the US in the type and quality of care provided to blacks, Hispanics, Native Americans and Pacific Islanders, and to a lesser extent, Asians, relative to whites. These inequalities have been documented for conditions requiring technologically advanced treatments such as organ transplant or coronary bypass grafting, as well as for conditions requiring routine surgeries and for common chronic diseases, such as diabetes and asthma. Furthermore, several meta-analyses of heath disparities research have confirmed the persistence of these inequalities, even after health insurance status, age, sex, income, education, and hospital type have been taken into account (The Henry J. Kaiser Family Foundation, 2002).

Limited data on the use of genetic health care services suggest a similar pattern. Concern about discrimination in medical settings seems likely to limit the number of minorities seeking genetic testing for conditions such as breast and colon cancer. If they do seek testing, research suggests that these patients might not benefit effectively from genetic testing services due to inadequate knowledge of cancer risks and of the relationship between family history and genetic cancer risk (Durfy et al., 1999; Lipkus et al., 1999). More data is needed to fully understand how practitioners and institutions might also contribute to the current trends in genetic health services delivery. However, absent an effective intervention, it seems likely that race as a social construct will influence a patient's benefit from genetic heath services in the same manner that it influences the delivery of other health services: For a combination of factors relating to the beliefs and conduct of patients, practitioners, and health care institutions, people of color are likely to receive care inferior to that of whites.

Disparities in health status (race as a social construct)

Disparities in health status are differences in actual health outcomes measured as rates of mortality and morbidity. Disparities in health status are caused by unequal medical treatment (or, disparities in health care), but also result from long-standing, pervasive racial and ethnic discrimination in the US (Smedley et al., 2003). This discrimination has set in place patterns of residential segregation that relegate minority populations to inferior housing, neighborhoods, and jobs. In these settings, minority populations are more likely to encounter health hazards such as lead paint, unsafe wiring, rodent and insect infestations, air pollution, and toxic chemicals (Anonymous, 1996; Berkowitz et al., 2003; Bowler, Gynsens & Hartney, 2003; Istre et al., 2001; Parker, et al., 1993; Shalat et al., 2003; Soliman et al., 1993). Furthermore, this segregation reproduces inequality by confining minority populations, especially blacks, to inferior schools and employment opportunities (Williams & Collins, 2001).

Inadequate health care and unsafe living conditions contribute to high psychological stress (Halpern, 1993) and, combined, result in shorter, less healthy lives for blacks, Hispanics, and American Indian and Alaskan Native populations (The Henry J. Kaiser Family Foundation, 1999; World Health Organization, 2001). These inequalities are evident from birth. Neo-natal mortality rates remain elevated among children born to black mothers relative to those born to whites, in part accounted for by a continued high rate of low birth weight babies among black mothers (Alexander et al., 1999). Inequalities persist throughout the life span with people of color reporting higher rates of many serious chronic illnesses, including hypertension, diabetes, and depression (Exner et al., 2001; Noh & Kaspar, 2003; Rotimi et al., 1999). The mortality rate for blacks is 1.6 times higher than that for whites, a ratio that is unchanged since 1950 (Williams & Rucker, 2000). The American Indian male's average life expectancy remains in the mid-50s, exactly where it was in 1960 (The Henry J. Kaiser Family Foundation, 1999), while the average life expectancy of white males has advanced nearly ten years to 74 over the same period. The average life expectancy of black males is now 66, also an advance since 1950, although less of one than for white males.

To the extent that disparities in health status are caused by unequal treatment based on race, the relationship of race and racism to disparities in health status is the same as the relationship of race and racism to disparities in health care. Here too race as a social construct—what people perceive to be the racial identity of a person with whom they are dealing—strongly influences how people are treated and the overall state of their health.

Disparities in health status (race as a biological construct)

Analyses of disparities in health status are not confined to research that treats race as a social construct, however. Some recent genetic research suggests that the poorer health status of people of color relative to whites depends not only on pervasive racism that creates an unhealthy physical and emotional environment, but also on genetic differences that make members of some populations more likely to develop certain conditions than members of other populations (Risch et al., 2002). In other words, this research posits that disparities in health status are due not only to racism but to race, in the sense of categories describing human groups that differ genetically. This is a controversial but increasingly widespread premise for research (Cooper, Kaufman & Ward, 2003; Sankar & Cho, 2002).

Were this research directed at the few single gene disorders, such as sickle cell disease (SCD) or cystic fibrosis (CF), that are often identified with specific populations, the trend would be of little interest. The history of human population migration makes inevitable the concentration of single gene disorders in populations that were once, or that remain, geographically adjacent, although even these disorders do not correspond precisely with groups popularly referred to as races. Thus SCD is one of several hemaglobinopathies common across western Africa, the Mediterranean, and India, while CF in contrast, is common primarily among certain Northern European populations and among eastern Europeans. However, research seeking to identify the genetic contribution to health disparities is not concerned with single gene disorders. Rather it seeks to identify genetic variations that interact with the environment in such a way to increase a person's vulnerability to common chronic conditions, such as heart disease and diabetes.

It certainly is possible that genetic variations do contribute to the onset and severity of these conditions, if only because all vulnerability to illness is related at some level to genetics. Not everyone from the same population, region, or even family, will develop the same diseases in response to the same environmental conditions. Who does and who does not can depend on subtle, contingent differences, which sometimes are genetic. If it is the case that genetic differences might make members of one population more likely to develop conditions associated with health disparities than members of another population, as some research suggests (Exner et al., 2001; Iacoviello et al., 2001; Makridakis et al., 1999; Risch et al., 2002), what are the costs of not pursuing this research?

Risks and benefits of genetic research on conditions associated with health disparities

Three arguments favour genetic research on race and disparities in health status. First, some diseases associated with health disparities have apparent genetic contributions. For example, research indicates that after controlling for socio-economic factors, blacks still have higher rates than whites of breast and prostate cancer, diabetes, and hypertension (Anonymous, 2004; Reynolds, 2003; Rotimi et al., 1999; Sellers, 1998). In the effort to understand these conditions—as with any serious health problem—all avenues that might lead to better diagnosis, prevention, treatment, or cure should be pursued, including nutrition, chemical exposure, behavior, and genetics. Second, the legacy of racial discrimination that persists today in genetics research—as elsewhere—has meant that conditions of importance to the health of people of color have been understudied relative to conditions of interest to whites. Research on health disparities and race could be viewed as addressing this imbalance. Third, a consequence of these biased priorities is that most of the major DNA repositories do not reflect the genetic variation of the US population (Pollack, 2003; Risch et al., 2002). More diverse DNA collections created by this research would allow not only for certain kinds of medical research but also for studies on population history and genetic variation of social and cultural interest (Kaiser, 2003; Reynolds, 2003).

These arguments support continuing and perhaps expanding health disparities research dedicated to determining genetic contributions. However, there are equally compelling reasons to question it. Elsewhere I have argued that seeking genetic solutions to conditions associated with health disparities deflects attention away from environmental causes and from recognised, if socially challenging, solutions (Sankar et al., 2004). This paper addresses not what is ignored if genetic research of this sort is pursued, but what is emphasised: specifically, genetic distinctions among US populations.

Social risks of research

Research risks are typically thought of as potential physical harms to subjects or as breaches of subject confidentiality. Concern over genetic research has expanded the idea of research risk to incorporate the possibility of broader social harms to individuals and to groups, such as stigmatisation or discrimination (Sharp & Foster, 2000, 2002). Discrimination might result from negative beliefs associated with a condition, such as cancer, or with a social group, such as gypsies. But at a more fundamental level, the harm results not from a one-to-one correspondence of a genetic test or research result associating a genotype with a social group. Rather, the greater harm comes from legitimating clear, self-evident, natural (or genetic) boundaries between groups. The harm is most threatening in settings such as the US where social framing based on an implied racial hierarchy is pervasive.

Genetic research associated with health disparities is particularly worrisome because asserting that genetic differences might make members of one population more likely to develop certain diseases than members of another also reifies and magnifies genetic difference in a way that too easily might be made to serve assertions about the inherent (genetic) inferiority or superiority of different populations. Based on what is known about the influential role of racial stereotyping in perpetuating health disparities, such research might inadvertently contribute to the conditions it seeks to alleviate (Doescher et al., 2000; Smedley et al., 2003).

Research risks: the limits of genetic solutions to health disparities

Arguably, even if there were no potential for reinvigorating racist claims, genetic research would not necessarily be the most productive line of research to follow to address health disparities (Holtzman & Marteau, 2000). This is the case for two related reasons. First, disease-causing mutations are extremely rare and will not account for most cases of common diseases. For example, prostate cancer among black men is often cited as one of the potentially most useful targets for genetic research. Between 1995 and 2000, statistics show that there were 73 prostate cancer deaths per 100,000 black men and only 30 per 100,000 white men. Approximately 10% to 15% of prostate cancer deaths are believed to have hereditary causes (Reynolds, 2003). If scientists discovered the genes for prostate cancer risk among black men tomorrow, and discovered a way to use this information to eliminate the morbidity due to genetic prostate cancer (the latter being far more difficult than the former), the number of prostate cancer deaths per year per 100,000 black men would decrease by 7 to 10 cases, to approximately 63 to 66 per 100,000, a rate still more than double that of white men, thus still leaving in place a large unexplained imbalance.

Second, genetic contributions to conditions associated with health disparities are complex and highly dependent on environmental factors. New findings seem only to add to this complexity (Doris, 2002; Pritchard, 2001). Recent articles report finding that scores rather than a handful of genes might be responsible for a single disease phenotype and that numerous unidentified confounders, such as mutations that confer protection rather than vulnerability or the age of research subjects, particularly in cardiology studies, might play important roles in shaping findings about risk profiles for common conditions (Doris, 2002; Pritchard, 2001). Recent reviews on the genetics of hypertension, breast and prostate cancer, and asthma suggest that a full understanding of the molecular bases for these diseases is still years in the future (Bradley, Given & Roberts, 2002; Bratt, 2002; Hakonarson & Wjst, 2001; Iacoviello et al., 2001).

This complexity likely accounts for why genetic association studies so often fail to demonstrate statistically significant findings. A recent analysis of these studies identified 600 articles that reported positive associations between common gene variants and diseases. Of these, the authors found that 166 of the putative associations had been reported three or more times. Of this group only six were consistently replicated (Hirschhorn et al., 2002). The day when gene-environment interactions for common conditions can be reliably described seems to be retreating rather than drawing nearer. Despite recent methodological innovations to overcome these problems (Dahlman et al., 2002), this research is not likely to bear fruit any time soon enough to address the current crisis in health disparities.

Research risks: social consequences of race-related genetic research

The complexities of genetic research and the provisional status of many of its recent findings about common conditions, such as hypertension, asthma, and cancer (Bratt, 2002; Hakonarson & Wjst, 2001; Iacoviello et al., 2001) however, are not always consistently recognised by genetic researchers. Research conducted without fully recognising the provisional or preliminary status of findings is particularly likely to inadvertently contribute to racial stereotyping. To illustrate how this might happen, I want to examine two scientific articles that report results about genetics and heart disease.

Disparities in health care and in health status related to heart disease in the US have been well documented (Krieger, 1990; Schulman et al., 1999). Some research suggests that after controlling for a wide range of socio-economic factors and co-morbidities, blacks still experience a higher prevalence of heart disease than other US populations. The two articles analyzed here pursue the idea that genetic variation might account for these differences. The first article reports differences between whites and blacks in their reaction to a drug to treat heart disease, Enalapril (Exner et al., 2001). The second reports findings of a difference between the reaction of blacks and whites to the presence of an allele coding for a glycolprotein that influences blood coagulation reactions in the gene Ala455Val (Wu et al., 2001). Both of these studies were conducted by prominent scientists working at major research universities and were funded by federal dollars. The results were published in prestigious science journals, and both were covered by the popular press.

Study 1: Enalapril

Enalapril is an angiotensin-converting enzyme (ACE) inhibitor, a drug widely endorsed as part of a regimen to manage heart failure. Using data originally collected to compare Enalapril to a placebo, a 2001 study of Enalapril, published in the New England Journal of Medicine (Exner et al., 2001), sought to determine if the drug's benefit varied by the patient's race. Despite efforts in the re-analysis to match black subjects to white subjects on risk factors for heart failure, in the final sample the black subjects were more likely than whites to have left ventricular dysfunction unrelated to ischemia, a history of hypertension, or a history of diabetes. Blacks and whites also differed on demographic factors including gender and age. The re-analysis combined data from two separate arms of the original study, one that had been designed to study the drug's influence on prevention of heart failure, the other on treatment. The endpoints in the re-analysis—deaths from any cause and number of hospitalisations—did not match the primary endpoints from the prevention arm of the original placebo-controlled study. These limitations, however, did not temper the certainty of the article's conclusion:

Enalapril therapy is associated with a significant reduction in the risk of hospitalization for heart failure among white patients with left ventricular dysfunction, but not among similar black patients.

Study 2: Ala455Val

Another 2001 article reported results of a study on the relationship between polymorphisms for a gene associated with acute myocardial infarction, and the race of subjects with these polymorphisms. The polymorphism of interest, Ala455Val, is a variation of the TM, or thrombomodulin, gene (Wu et al., 2001). Three genotypes, AA, AV, and VV, were identified in the study population, some of whom were known to have heart disease. Analysis found that genotype AA was more likely to occur in people without heart disease and therefore was interpreted as being protective against the disease. This appeared to be the case regardless of the subject's racial identity. Analysis also found that AV or VV (in other words the presence of a V allele) also was not associated with an increased likelihood of heart disease; however, this finding only applied to whites. In blacks the presence of a V allele was associated with a significant increase in heart disease risk, even after adjusting for age, sex, and other coronary heart disease risk factors. Of the study's 246 subjects with a V allele (AV or VV genotype), 21 were black, 225 were white. Of the 21 blacks with a V allele, 14 were categorised as having coronary heart disease and 7 were not. The authors interpreted this finding to mean that the presence of a V allele has significantly different consequences for whites than for blacks.

In this study, investigators also examined a hypothesised mechanism that they thought accounted for the V allele's influence on coronary heart disease (which seems to be related to the presence of a glycoprotein involved in coagulation reactions). However, this analysis showed no statistically significant difference between samples from people with AA genotypes versus those with AV or VV. In other words, investigators could not isolate the predicted effect of having a V allele. Thus while researchers found a difference in disease incidence between those with and those without a V allele, the hypothesised explanation that would link the presence of the V allele to disease was not supported. Nonetheless, the article concludes:

Our results show that this single nucleotide polymorphism of the TM gene is an independent risk factor for CHD [coronary heart disease] in blacks.

The design limitations and sampling problems in this study, and in the Enalapril study, are not exceptional nor do they necessarily invalidate the studies' basic findings. The question is whether these findings support the broader inferences made by the authors. While it might have been accurate to report that twice the number of blacks with coronary heart disease also had a V allele (14 versus 7 subjects), it is well beyond the scope of a finding based on 21 people, in a study that also fails to reproduce the proposed mechanism for translating this difference into an elevated CHD risk, to assert unequivocally any generalisation about the study populations is the case.

While even the most sober scientific reports are known to succumb to overly grand conclusions, that doesn't make it all right. Journal editors and the peer review process are meant to weed out what writing specialists call the ‘hasty generalization’ (Alley, 1998), that is, one ‘based on too little evidence or on exceptional or biased evidence.’ It is important to note that genetics researchers themselves have called attention to overstated results and emphasised the need for care when interpreting results concerning linkages between genetic variation and disease, especially those that claim to report differences by race (Gambaro, Anglani & D'Angelo, 2000; Ioannadis, Ntzani & Trikalinos, 2004). Listed below are examples of generalisations that are drawn from studies that are not about genetics but that are similar to those analyzed here in that they concern population-based causes for common diseases (such as cancer or heart disease) and they demonstrate a far greater restraint.

The association between EMF [electromagnetic field] exposure and prostate cancer mortality warrants further investigation. (Charles et al., 2003, italics added)

This was consistent with the previous report from this group. Interestingly, the cohort demonstrated a statistically significant risk of overall cancer incidence and specific increased incidence of gallbladder cancer. CONCLUSIONS: This study contributes further evidence to the growing body of literature indicating the carcinogenic properties of mineral oils used in occupational settings, in particular those used prior to 1970s. (Yassi, Tate & Routledge, 2003, italics added)

Agricultural exposures were associated with significant risk increases among white women and white men. Further work is required to investigate in more detail. (Cocco, Dosemeci & Heineman, 1998, italics added)

The difference between the tenor of these conclusions and the genetics studies’ conclusions might be explained away by pointing out that statements of proof or the rhetoric of evidence vary across sub-disciplines of science. This however simply begs the question: what accounts for these differences? For example, what relation if any does a conclusion's certainty have with how closely it matches the researcher's expectations?

From finding to fact: examples from secondary literature

Despite design weaknesses in both the Enalapril and Ala455Val studies, and despite later research that contradicts the claims about racial differences in the Enalapril study (Dries et al., 2002), these articles entered the scientific and popular literature with little challenge. The Ala455Val findings were reported in the Houston Chronicle as the discovery of a genetic mutation that ‘increases sixfold the risk of heart disease in black carriers’ (Ackerman, 2001). While some hyperbole might be blamed on the journalist, the newspaper also quoted the scientist who conducted the study as saying: ‘This discovery underscores the importance of studying genetic risk factor by ethnic group’ (Ackerman, 2001).

Exaggerated certainty about genetic research is not limited to newspaper accounts. A recent scholarly article, entitled ‘Constitutional issues in the use of pharmacogenomic variations associated with race’ (Robertson, 2003), set as its task assessing whether conducting genetic research using racial categories is in itself discriminatory and thus unconstitutional. Its author states that part of this task will require assessing the quality of research making these claims. Yet despite his emphasis on quality, the author relies on a popular press account, the Houston Chronicle's report that Ala455Val is a gene associated ‘with a sixfold increase in serious heart disease’ among blacks (Ackerman, 2001), to support his assertion that genetic research that uses racial categories is useful and that failure to pay attention to its findings might ‘deprive racial and ethnic minorities of good health care’ (Robertson, 2003).

The consequences of insufficient skepticism concerning the original Enalapril study might be even greater than those concerning the Ala455Val findings because its findings imply a clear departure from standard of treatment. The Enalapril study received broad attention in the popular press and has been cited 147 times by other articles published in scientific journals (ISI Web of Science, 2005). The majority of these citations refer to the article as evidence of the racial differences in response to heart medications it purports to identify.

However, results that contradict these findings were reported by a 2002 study that analyzed the same data set (Dries et al., 2002). The 2002 study did not combine the treatment and prevention arms of the original study and matched its endpoints to those of the original study. One of the authors of the 2002 study is also an author of the 2001 Enalapril study. The 2002 study found no difference by race in response to Enalapril. The 2002 study received no media coverage and has been cited only twice in scientific journals since its publication (Cooper et al., 2003; Taylor et al., 2004). The first citation appeared in an article about problems with the claims of genetic research generally, not in an article that was addressing differential responses to Enalapril. The second is an article that reports significant racial differences in response to heart medications, but it cites the 2002 study as background without addressing how its findings might contradict the new results. The difference in citation patterns for the 2001 Enalapril study and its 2002 re-analysis suggests a direction for future research, which is to determine whether genetic findings claiming response differences by race are of greater interest to scientists and are more likely to be used as the basis for subsequent work than findings that contest racial differences, or report none.

The inconsistencies between the 2001 and the 2002 Enalapril studies are not atypical in science, and under certain circumstances might eventually sort themselves out through clinical trials. When new drugs are developed, FDA requirements demand clinical trials to show efficacy, so perhaps the possibly exaggerated claims of difference between blacks and whites do little damage because subsequent research will either uphold or disprove them.

However, there are limits to this reassurance. While FDA oversight is required to bring a new treatment to market, because Enalapril has already been approved there is no demand to subject the Enalapril findings to more clinical trials simply to propose that it be used differently for blacks and whites. Nor does recent history suggest that the public is likely to demand a change to this policy. Public fascination with genetics seems to divert critical attention away from assertions about new genetic technologies, which allows ideas that deserve careful scrutiny to avoid challenge and gain wide acceptance. Such was the case with research labeled gene ‘therapy’ that received much acclaim in the 1990s and yet still has not achieved therapeutic results (Holtzman & Marteau, 2000). And such appears to be the case with genetic research claiming to identify contributors to health disparities. Reaction to the recent approval of a heart medication putatively developed for Blacks seems to confirm this (Sankar & Kahn, 2005).

From finding to fact to racial stereotyping

But more to the point here is how exaggerated claims might contribute to racial stereotypes. How might the 2001 Enalapril and the Ala455Val articles be used as evidence of broader claims, claims that both rely on and reproduce racial stereotyping? Consider for example, a newspaper article published originally in the Los Angeles Times in August, 2003 (Healy, 2003a) and since reprinted in 4 city newspapers across the US, including Phoenix, Fort Wayne, Miami, and St. Louis, and in one national quarterly business journal, The Black College Quarterly, also available on the web as BlackVoice.com.

The establishment of a DNA bank (the Genomic Research on the African Diaspora biobank, or GRAD) at historically black Howard University supplies the opening news hook to launch the article's general discussion of race and genetics. After describing GRAD and its particular research focus on medical conditions that are common among blacks, the author generalises this interest in genetics to a broader trend. Researchers who had ‘long believed that the social, environmental and economic stresses of lower-income and minority status’ explained the higher rates of ‘hypertension, heart disease, prostate and breast cancer, asthma, glaucoma and obesity’ among blacks, are no longer able to ‘ignore a growing mountain of research’ that shows that ‘inheritance plays a key role in the development, course and outcome of certain diseases.’ One might challenge even these assertions (Khan, 2003), but the more troubling claims come in the following paragraph. It appears that the ‘mountains of research’ supported by a vague ‘welter of other initiatives’ have not only reoriented thinking about disparities in cancer incidence and heart disease away from ‘social, environmental and economic stresses of lower-income and minority status’, but also have redirected research into causes of socially constructed attributes, such as intelligence or criminality. Back on the table, suggests the author, rehabilitated from the trash heap of bad science, are claims that link together criminality and intelligence with genetics and thus, in the United States, to race. The article reads:

Howard's biobank—along with a welter of other initiatives—points toward a provocative conclusion: that some racial differences are encoded in the genes, and those differences can make people of one skin color inherently more or less susceptible to certain diseases than people whose complexion is different … Suddenly, they say, it might be fair game to look into the genes for more controversial racial divides in areas such as intelligence, criminality and addiction. (Healy, 2003b)

There need not be a direct link between the Enalapril and Ala455Val studies and the slant of the Los Angeles Times piece to see the relationship between them. Studies such as those on Enalapril and Ala455Val, if not these studies themselves, are part of the ‘growing mountain of research’ that the Los Angeles Times article cites to justify the assertion that a paradigm shift is underway to re-invigorate genetic studies of race. The intent of the genetic research that is creating the mountain need not have been to re-open the field to such research, but once published it is available to be appropriated by others as they choose, for example to extend that interest to studies of intelligence and criminality.

The point here is not to suggest that geneticists or other scientists are responsible for how the popular press interprets scientific findings. Nor is it even to assert that geneticists should not conduct research using racial categories. Perhaps using racial categories is the best way to uncover connections needed to treat and prevent conditions associated with health disparities. Certainly many prominent researchers argue that this is the case (Collins et al., 2003; Risch et al., 2002). The point rather is that close attention must be paid to the categories that this research uses and to the generalisations it makes. Findings from these studies diffuse rapidly into the general public. And because they often support rather than contradict widely shared social assumptions of fundamental differences between populations, they diffuse without notice.

Acknowledgements

Earlier versions of this paper benefited from comments during peer review by this journal and by organisers and participants in the Conference on Genetics and Health Disparities, Institute for Social Research. Ann Arbor, Michigan, 20–21 March 2004. In particular, I would like to thank discussant David R. Williams for his comments.

This work was supported by NHGRI R01 HG02189-01.