- Email: [email protected]
The Role of Genetic and Environmental Factors in Cardiovascular Disease in African Americans
r The Role of Genetic and Environmental Factors in Cardiovascular Disease in African Americans RICHARD COOPER, MD
ABSTRACT: Considerable interest has been focused over the years on estimating the relative importance of genetic and environmental factors on differences in rates of cardiovascular disease in blacks and whites. However, recent advances in molecular science have helped to illuminate the underlying complexity of this problem. Attempts to impute the genetic component from "what was left over" after control for a limited set of environmental exposures is increasingly recognized as naive. The requirements for a model that could account for interactions between genetic and environmental factors far exceeds the precision of our measurements. Although it is obvious that blacks experience not only unique environmental exposures, such as white racism, but more intense exposure to common factors, such as obesity, current methods make it very difficult to summarize these effects. Simpler models, using data from large samples, could provide greater precision and might illuminate the exposure-outcome relationships common to all groups. Meanwhile, efforts to identify genetic underpinnings of complex disorders will have to reach a much higher level of development before useful conclusions can be reached about the magnitude and variation of effects between racial and ethnic groups. KEY INDEXING TERMS: Race; Gene-environment interactions; Hypertension. [Am J Med Sci 1999; 317(3):208-13.]
From the Department of Preventive Medicine and Epidemiology, Loyola University, Stritch School of Medicine, Maywood, Illinois. This work was supported in part by Grants HIA550B and HL47910 from the National Heart, Lung, and Blood Institute. Correspondence: Richard Cooper, M.D., Department of Preventive Medicine and Epidemiology, Loyola University Stritch School of Medicine, 2160 S. First Ave, Maywood, IL 60153 {E-mail: [email protected]}.
208
T
he structure of reasoning in biological science requires that genes and the environment determine the phenotype. An important goal of epidemiological research is to obtain increasing precision and sophistication in assessing the contribution of these fundamentally different dimensions to disease risk, either in isolation or jointly. In practice, however, it is difficult to meet this challenge. Conceived of as a supporting discipline of public health, epidemiology excels at the identification of common disease-promoting exposures. Quantification of the impact of these exposures, and the unraveling of mechanisms, however, are ends to which it is less well suited. Fortunately, medical practice does not always require an understanding of how disease processes work, and as the basis for interventions, the categorical designation of risk factors is a sufficient outcome. Cardiovascular (CV) epidemiology has achieved great success with this method, and the reduction in coronary heart disease (CHD) and stroke represents the signal advance toward control of chronic disease this century. As the requirement for understanding moves to the mechanistic level, however, an entirely new set of demands must be met, and this dichotomy between identifYing risk factors and elucidating causal processes is nowhere more evident than in the study of health differentials among racial and ethnic groups. Few acts of data perusal are more disconcerting than a simple comparison of death rates among African Americans and other US population groups. Among whites, age-adjusted, all-cause mortality was 477 per 100,000 in 1995, whereas among blacks, the rate was 767 per 100,000_ 1 Life expectancy at birth for Hispanic men has been estimated to be 81 years; the comparable figure for black men is 65 years.2 Clearly, widespread exposures of moderate to great intensity must be at work here, and one would assume that public health science could provide an unambiguous explanation. Unfortunately, although the manifestations of the black health disadvantage have been documented in great detail, the process itself has never received the sophisticated treatment afforded such individual disease syndromes as atherosclerosis and AIDS. March 1999 Volume 317 Number 3
Cooper
Table 1. Death Rates and Years of Productive Life Lost from Cardiovascular Diseases, By Race and Sex Men
Death rate* Heart disease CHD Stroke YPLLt Heart disease CHD Stroke
Women White
Black
Ratio BfW
1.38 1.05 1.97
96.1 62.4 22.8
158.0 83.6 40.1
1.64 1.34 1.76
1.83 1.28 3.13
658.9 364.9 155.5
1681.8 699.9 446.7
2.55 1.92 2.87
White
Black
Ratio BfW
183.8 128.9 26.6
254.1 134.8 52.4
1709.1 1108.9 195.5
3120.8 1418.2 612.1
YPLL, years of productive life lost; CHD, coronary heart disease. * Ref l. t Ref 3.
These arguments are not intended to imply that epidemiology has provided no insight into the causes ofCV disease in general, or the excess among blacks. To the contrary, as the mortality declines eloquently testify, much of value has been learned3 and an enormous and rich literature on the environmental influences on CV health is available. It is equally obvious that the principles that have proven so effective in disease prevention apply equally to all groups. Yet we continue to be confronted with very substantial excess risk among blacks, and we will require new information if we are to better understand this dilemma. 4 ,5 Part of the explanation for the lack of progress can be attributed to the approach we have taken to disentangling the threads that bind nature to nurture. The Pattern of CV Disease in African Americans
As is well known, Mrican Americans suffer higher mortality rates from all major forms of CV disease (Table 1). In addition, the earlier age of onset of CV diseases among blacks results in a much larger differential when the measure is "years of productive life lost" (YPLL)(Table 1). Likewise, an enormous body of evidence suggests that the differential in incidence of CV diseases among blacks can be attributed primarily, if not exclusively, to hypertension. 5,6 Given the improvement in medical therapyboth preventive and curative-and the differentials in medical care received between blacks and whites, some proportion of the differences in survival after the onset of CV disease at the present time can be ascribed to differential access to both medical and surgical therapies. 7 The effect of in terventionaI therapies is likely to be particularly important for CHD. It is equally apparent that the social class gradient in incidence and survival from CV diseases accounts for much of the black disadvantage. 5 Although the impact of social class is a relatively straightforward problem when formulated in genTHE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
eral terms, as discussed below, the literature on this question is filled with controversy. The interpretation of adjustment for measures of socioeconomic status (SES) is particularly difficult.5,8 Although it is reasonable to use insights from the study of social class to inform the study of heart disease among blacks, the outcome may not always meet our expectations. For example, black men with higher SES have been found to have greater levels of obesity, contrary to whites, and smoking rates among young black women are relatively low. 1 It is likely, however, that a set of more generalized influences, from psychosocial factors to reduced medical care access, contribute to higher rates of morbidity and mortality. Likewise, although the descriptive information regarding SES differentials in risk factors has obvious utility for public health interventions, the use of these data in studies of causality has been questioned. 5,8,9 When it is not linked to specific research questions or accompanied by measurements that can provide insights into mechanisms, the emphasis on patterns of outcomes organized by SES has generated few new insights. A number of questions related to social factors can be identified that are uniquely relevant to minority populations, particularly African Americans. For example, it would be important to know whether specific cultural beliefs or practices influenced CV risk. Unique social stressors are postulated for minority groups, but not well characterized. Clearly racism and other forms of discrimination are important considerations here. On the other hand, nontraditional support systems might exist within these communities. Targeted research will be necessary to elucidate these relationships. Environmental Aspects of CV Risk among Blacks
A number of important areas for the study of environmental risk factors for CV disease among blacks can be identified. First, it must be acknowl-
209
Role of Genetic and Environmental Factors
Table 2. Multivariable Coefficients Associated with Major Coronary Risk Factors among Blacks and Whites, the NHANES I and II Cohorts
40 CI)
35
'u;
30
>
Women
Men Risk Factor Age Cholesterol Systolic BP Smoking
White .932* .029 .117 .453
Black .769 .048 .134 .009
White 1.188 .020 .152 .736
C
Black .906 .117 .113 .397
* Estimated with the proportional hazards model. Increment in risk for + 10 years, +10 mg/dL, +10 mm Hg, Yes/No smoking, respectively.
edged that the vast majority of disease-causing processes are common to all ethnic groups. Although an exhaustive list may not be available, most important risk factors that are thought to be important in whites have been verified in blacks,1° A more complicated question, however, relates to the relative impact of the individual factors and the manner in which they interact. Ultimately, any attempts to identify ethnic-group specific aspects of CV risk must be placed in the social context of these populations. Obviously, for African Americans, full accounting must be given to the enduring impact of racism on all aspects of health. We used the probability samples ofthe US population available through the First and Second National Health and Nutrition Examination Surveys to examine the predictive significance of the traditional risk factors (Table 2). Reasonable consistency exists, with the exception of smoking, which is notoriously hard to compare across populations. It should be recognized, however, that these cohorts provide information on only a combined total of 265 fatal CHD events for black men and women. Precision greater than that provided by this study is required to identify novel interactions or outcomes. As noted above, because hypertension can account for the great majority of the black excess in heart disease, the crucial question in this field reduces to an effort to understand the cause and potential prevention of hypertension in US blacks. An enormous literature exists on this problem, yet no solution is in sight. 6 It seems relevant at this juncture to introduce emerging findings from international comparisons among black populations. The prevalence of hypertension has been shown to vary, from 8-10% in rural Africa to 33% in the United States (Figure 1),11 Expressed against a common genetic background, this variation in phenotype emphasizes the determining role of environmental exposures. The relationship between known risk factors, such as obesity, closely parallels the cross-population trends in hypertension (Figure 2),11 Similar pat-
210
~CI)
25
C.
:>. 20 ::I: 15 C
CI)
(,) ~
10
CI)
C.
5
o Figure 1. Prevalence of hypertension among six populations of West African origin. The ICSHIE study, 19,95. Hypertension is defined as blood pressure 2 140/90 mm Hg, or antihypertension medication.
terns are observed for non -insulin-dependent diabetes mellitus, a CV risk factor of growing imp ortance. 12 An important challenge for descriptive epidemiology will be to elucidate more fully within populations the mechanisms by which risk factor exposures result in their physiological consequences. A Framework for Studying Racial/Ethnic Variation in Disease Risk
A paradox emerges in the analysis of racial differentials. The data available for descriptive purposes is overwhelming in its breadth and complexity, as suggested above, although the methods required for most comparative research are relatively straightforward. By contrast, the literature devoted to etiologic or explanatory analyses is impoverished and
35
Maywood
CI)
> 'u; c
30
~ 25 CI)
• S t. L·uCla ~arbados
Co
-
Jamaica
>-
J:
cCI)
20
• Cameroon
(,) ~
CI)
C.
15
o
Nigeria
10~-'---'----'----'----'----'
22
24
26
28
30
32
Body Mass Index Figure 2. Prevalence of hypertension by mean body mass index in six populations of West African origin. ISCHIE, 1995. March 1999 Volume 317 Number 3
Cooper
restricted, and often defaults to generalities about the "interplay of genes and environment." The logical construct, however, through which the "ethnic paradigm" can be pursued to arrive at causal explanations is exceedingly complex and is not well served by these simplistic generalities. Consequently, journals are filled with descriptions of shades of difference in health status, most of which are derivative effects from a few fundamental processes, with little attempt to define the underlying processes. The vascular sequelae of hypertension, for example, have been dissected in countless publications. The conclusions, more often than not, are restricted to speculation about the potential "genetic differences" that caused these secondary changes. The structure of the inquiry that could lead one to differentiate the bona fide effect of a cause from mere epiphenomenon is rarely defined, however. On balance, we have accumulated a large number of descriptive articles that seek to persuade us, by repetition more often than analysis, of "intrinsic differences" between blacks and whites. Because genetic and environmental influences are the focus of this presentation, it is important to pursue this problem to obtain further precision on the question of how we determine a causal relationship. Under what circumstances can something be said to be a cause? Although this problem is an everyday concern to working epidemiologists, it receives relatively little formal attention. The "Bradford Hill criteria" are familiar enough. Likewise, the properties of the randomized experiment, which make it unique as the basis for inference, are intuitively understood within the field. As on other counts, however, the work on racial differentials exposes important limitations in the application of causal inference. Informed primarily by the statistical literature, a modest body of work on causal inference exists in epidemiology.9,13-15 Crucial to this reasoning is the concept of "exchange ability" or "the alternative state."14 Formalized within the construct of the counterfactual, a framework is created that asks, "What would the outcome have been if this individual had never been exposed to the putative cause?" In a randomized experiment, of course, this means that we are justified in inferring that the outcome observed in the treated group would have occurred among the control subjects, had they been treated. With this generalization, we are equipped to make conclusions about the value of this intervention for similar patients in the future. For observational studies, however, the criterion of "exchange ability" is much harder to define. Control for confounding, through statistical adjustment, is the only recourse at hand. Although we can be reasonably comfortable with some forms of adjustment (for example, age, because it is plausible for THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
individuals to assume various ages), this method faces serious challenges when applied to attributes that define the essential character of an individual, such as gender or race. Thus, it is difficult to work with an analytic sequence that asks, "What would the risk of depression have been for this person if she had not been a woman?" A question naturally arises: can we speak of a woman as the same individual if she were not a woman? In fact, we would have difficulty defining what it means for a woman "to not be a woman." Practically, of course, the alternative is to be a man, but what does that tell us about "woman-ness" as a cause of depression? Although an adequate explication of this argument requires substantially more detail, it suffices in this context to make the observation that causal inference related to essential attributes, rather than discrete events, raises complex problems for the epidemiologic method. Similar, perhaps more difficult, problems confront racial analyses. 9 Although it is easy to accept that biological differences between men and women might serve as a reservoir of potential "causes," racial comparisons rely directly on the hypothesis that allele frequencies vary among the population groups and cause the observed contrasts in disease risk. Although it is acknowledged that some genetic factors do vary across populations, causal inference about the role of "essential" attributes is at least as complex for racial as for gender comparisons. To estimate genetic effects, one must control for environmental confounding. The infinitude of social influences, some explicit, many buried from view by unconscious prejudice, are more than a match for the techniques of statistical control. We recognize the risk of overweening inference; however, given the complete inability of studies on racial differences to yield useful etiologic insights, it is logical to ascribe this inability to the weakness of the pathway to causal reasoning. In effect, it follows from the argument outlined here that no valid inferences about "essential" differences between blacks and whites can be derived from observational studies. No generalized "genetic effects"-which constitute the only plausible meaning of essentialism in the biological context-can be detected using the methods of observational epidemiology. Genetic Aspects of CV Risk among Blacks
Although genetic effects have played a key role in etiologic reasoning for many years, useful measurements were not possible until the last decade. This technological revolution is slowly altering the theoretical approach to studying disease in human populations. In practice, however, the field of CV genetic epidemiology is still in its infancy. Many new markers were found to be associated with increased disease risk, but the findings often proved to be 211
Role of Genetic and Environmental Factors
Causal Effects Are Often Assumed to Fit an Additive Model:
Genes + Environment
=
Organism
In the Attempt to Isolate "Genetic Effects", this Model is Rewritten:
Figure 3. Modeling the contribution of genes and environment.
Observed Phenotypic Variation - Environmental Contribution
= Genetic Effect
However, Because Interactions Occur, Additional Terms are Required: IJ
Observed Phenotypic Variation
=
Genes + Environment + Genes
* Environment + Genes * Genes +
Environment
* Environment
r
I
t ! l
t
inconsistent in subsequent studies. For example, a number of polymorphisms in the renin-angiotensin system (RAS) have been linked to risk of hypertension and other forms of CV disease.1 6 - 19 Unfortunately, the totality of evidence on the RAS is contradictory. The approach using candidate genes has now been superseded by the genome-wide screen. These studies seek to identify, based on markers scattered across all chromosomes, regions of the genome that are consistently found among persons with the health condition of interest. After this stage, the painstaking effort of locating nearby genes must take place. This work has just gotten under way, however, and empirical data are not yet available. Because the methodology in this field is developing rapidly, both in terms ofthe laboratory methods and statistical analysis, it is difficult to predict the immediate future. It is clear, nonetheless, that genetic effects for complex traits, such as blood pressure, will be exceedingly subtle, thus necessitating very large sample sizes. Genetic epidemiology, therefore, may continue to become a more specialized field. In general, an obvious dichotomy exists between the approach to "genetic effects" that was standard in epidemiology for the last several decades and the molecular approach. Genetic effects in the past have been 212
•
modeled from measurements of phenotypes, typically as heritability. Statistical methods have been used to estimate overall effects (for example, comparing monozygotic and dizygotic twins). In epidemiology, genetic and environmental effects were often conceived of in an additive model, which together accounted for 100% of the phenotypic variation (Figure 3). Given this model, it was possible to rearrange the terms and isolate the quantity designated as "genetic effects." Assuming, however, that many interactions occur, it is obvious that a fully specified model does not permit the isolation of "genetic effects" in this manner (Figure 3). By the same token, with the advent of molecular methods, interest in modeling overall quantities has diminished markedly. The challenge now becomes cloning specific susceptibility genes and defining their physiological roles. These developments are particularly salutary in studies of ethnic variation, because they reduce the potential to confuse genetic effects and residual confounding. An opportunity now exists to seek reproducible, stable effects of specific molecular markers within individual populations. In an important way, the focus has shifted to identification of pathways by which DNA variants directly influence phenotypic traits. With these findings in hand, it will become possible to attempt comparisons across groups. March 1999 Volume 317 Number 3
I
I
! '
Cooper
Conclusions
Experience to date suggests that earlier attempts to model "genes and environment" as global concepts were almost certainly naive. In practice, the degree of precision required to make sense of a model at this level of generality far exceeded the capacity of the investigational methods available. The tendency to desist from the effort to make judgments about the underlying cause of "black-white" differences, when the causal process within neither group was understood, is a welcome change. In addition, the advent of molecular techniques has given research into genetic determinants a material basis, and an objective criteria for the evaluation of success. The requirement for those who wish to speak of genes as causes is now to name the gene and the sequence variation. Reasoning backwards from differences in observed phenotypes to hypothetical differences in genotype should no longer receive credence from the scientific audience. Racial categories imply genetic differences, and explanations of racial differences inevitably respond to this formulation. The conflation of genes and environment in the explanatory model, when both are poorly characterized, makes it impossible to define clearly their independent contribution. In particular, the search for specific susceptibility genes may be compromised by the desire to ascribe effects to global, yet undefined, "genetic" influences. What is required is an attempt to simplify the working model, eliminating those elements that are grounded in tenuous and, for the present, untestable assumptions. Perhaps as an outcome, on both counts, therefore, it is possible to argue that "less is more"-simpler descriptive models are needed first to define effects free of interactions; ethnic comparisons are hopelessly confounded and putative genetic effects should be judged by the strength of the molecular evidence. What confronts us is the task of unraveling the complex pathway from genes, as impacted by environment, to phenotype. In our desire to offer explanations, we clearly exaggerated the strength of the claims that could be offered on the basis of what is currently known. References 1. National Center for Health Statistics. Health, United States, 1996-97 and injury chartbook. Hyattsville (MD): Na-
THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
2. 3.
4. 5. 6. 7. 8. 9.
10.
11. 12. 13. 14. 15. 16. 17. 18. 19.
tional Center for Health Statistics, Centers for Disease Control, US Department of Health and Human Services; 1997. Publication no. (PHS) 96-1232. Liao Y, Cooper RS, Cao G, et al. Mortality patterns among adult Hispanics: findings from the NHIS, 1986 to 1990. Am J Public Health, 1998;88:227-32. National Heart, Lung, and Blood Institute. NHLBI Morbidity and Mortality Chartbook (1996). Bethesda (MD): National Heart, Lung, and Blood Institute, National Institutes of Health, US Department of Health and Human Services; 1996. Cooper RS. Ethnicity and disease prevention. Hum BioI 1993;5:387-98. Cooper RS. Health and the social status of blacks in the United States. Ann Epidemiol 1993;3:137-44. Cooper R, Rotimi C. Hypertension in African Americans. Am J Hypertens 1997;10:804-12. Ford ES, Cooper RS. RaciaVethnic differences in health care utilization of cardiovascular procedures: A review of the evidence. Health Services Res 1995;30:237-52. Moss N. What are the underlying sources of racial differences in health? Ann EpidemioI1997;7:320-1. Kaufman JS, Cooper RS, McGee D. Socioeconomic status and health in blacks and whites: The problem of residual confounding and the resiliency of race. Epidemiology 1997; 30:1511-7. Francis CK, Grant AO, Cooper RS et al. NHLBI report of the working group on research in coronary heart disease in blacks. Bethesda (MD): National Heart, Lung, and Blood Institute, National Institutes of Health, US Department of Health and Human Services; 1994. Publication no. (PHS) 95-129037. Cooper R, Rotimi C, Ataman S, et al. Hypertension prevalence in seven populations of African origin. Am J Public Health 1997;87:160-8. Cooper R, Rotimi C, Kaufman JS, et al. Prevalence of NIDDM among populations of the African diaspora. Diabetes Care 1997;20:343-8. Holland PW. Statistics and causal inference. J Am Stat Assoc 1986;81:945-61. Greenland S, Robins JM. Identifiability, exchange ability and confounding. Int J Epidemiol 1986;15:413-9. Greenland S. Randomization, statistics and causal inference. Epidemiology 1990;1:421-9. Jeunemaitre X, Soubrier F, Kotelevtsev YV, et al. Molecular basis of human hypertension: role angiotensinogen. Cell 1992;71:169-80. Caulfield M, Lavender P, Farrall M, et al. Linkage of the angiotensinogen gene to essential hypertension. N Engl J Med 1994;330:1629-33. Rotimi C, Cooper R, Ogunbiyi 0, et al. Hypertension, serum angiotensinogen and molecular variants of the angiotensinogen gene among Nigerians. Circulation 1997;95:2348-50. Rotimi C, Puras A, Cooper RS, et al. Polymorphisms of the genes in the renin-angiotensin system among Nigerians, Jamaicans and African Americans. Hypertension 1996;27:558-63.
213
4
ABSTRACT: Considerable interest has been focused over the years on estimating the relative importance of genetic and environmental factors on differences in rates of cardiovascular disease in blacks and whites. However, recent advances in molecular science have helped to illuminate the underlying complexity of this problem. Attempts to impute the genetic component from "what was left over" after control for a limited set of environmental exposures is increasingly recognized as naive. The requirements for a model that could account for interactions between genetic and environmental factors far exceeds the precision of our measurements. Although it is obvious that blacks experience not only unique environmental exposures, such as white racism, but more intense exposure to common factors, such as obesity, current methods make it very difficult to summarize these effects. Simpler models, using data from large samples, could provide greater precision and might illuminate the exposure-outcome relationships common to all groups. Meanwhile, efforts to identify genetic underpinnings of complex disorders will have to reach a much higher level of development before useful conclusions can be reached about the magnitude and variation of effects between racial and ethnic groups. KEY INDEXING TERMS: Race; Gene-environment interactions; Hypertension. [Am J Med Sci 1999; 317(3):208-13.]
From the Department of Preventive Medicine and Epidemiology, Loyola University, Stritch School of Medicine, Maywood, Illinois. This work was supported in part by Grants HIA550B and HL47910 from the National Heart, Lung, and Blood Institute. Correspondence: Richard Cooper, M.D., Department of Preventive Medicine and Epidemiology, Loyola University Stritch School of Medicine, 2160 S. First Ave, Maywood, IL 60153 {E-mail: [email protected]}.
208
T
he structure of reasoning in biological science requires that genes and the environment determine the phenotype. An important goal of epidemiological research is to obtain increasing precision and sophistication in assessing the contribution of these fundamentally different dimensions to disease risk, either in isolation or jointly. In practice, however, it is difficult to meet this challenge. Conceived of as a supporting discipline of public health, epidemiology excels at the identification of common disease-promoting exposures. Quantification of the impact of these exposures, and the unraveling of mechanisms, however, are ends to which it is less well suited. Fortunately, medical practice does not always require an understanding of how disease processes work, and as the basis for interventions, the categorical designation of risk factors is a sufficient outcome. Cardiovascular (CV) epidemiology has achieved great success with this method, and the reduction in coronary heart disease (CHD) and stroke represents the signal advance toward control of chronic disease this century. As the requirement for understanding moves to the mechanistic level, however, an entirely new set of demands must be met, and this dichotomy between identifYing risk factors and elucidating causal processes is nowhere more evident than in the study of health differentials among racial and ethnic groups. Few acts of data perusal are more disconcerting than a simple comparison of death rates among African Americans and other US population groups. Among whites, age-adjusted, all-cause mortality was 477 per 100,000 in 1995, whereas among blacks, the rate was 767 per 100,000_ 1 Life expectancy at birth for Hispanic men has been estimated to be 81 years; the comparable figure for black men is 65 years.2 Clearly, widespread exposures of moderate to great intensity must be at work here, and one would assume that public health science could provide an unambiguous explanation. Unfortunately, although the manifestations of the black health disadvantage have been documented in great detail, the process itself has never received the sophisticated treatment afforded such individual disease syndromes as atherosclerosis and AIDS. March 1999 Volume 317 Number 3
Cooper
Table 1. Death Rates and Years of Productive Life Lost from Cardiovascular Diseases, By Race and Sex Men
Death rate* Heart disease CHD Stroke YPLLt Heart disease CHD Stroke
Women White
Black
Ratio BfW
1.38 1.05 1.97
96.1 62.4 22.8
158.0 83.6 40.1
1.64 1.34 1.76
1.83 1.28 3.13
658.9 364.9 155.5
1681.8 699.9 446.7
2.55 1.92 2.87
White
Black
Ratio BfW
183.8 128.9 26.6
254.1 134.8 52.4
1709.1 1108.9 195.5
3120.8 1418.2 612.1
YPLL, years of productive life lost; CHD, coronary heart disease. * Ref l. t Ref 3.
These arguments are not intended to imply that epidemiology has provided no insight into the causes ofCV disease in general, or the excess among blacks. To the contrary, as the mortality declines eloquently testify, much of value has been learned3 and an enormous and rich literature on the environmental influences on CV health is available. It is equally obvious that the principles that have proven so effective in disease prevention apply equally to all groups. Yet we continue to be confronted with very substantial excess risk among blacks, and we will require new information if we are to better understand this dilemma. 4 ,5 Part of the explanation for the lack of progress can be attributed to the approach we have taken to disentangling the threads that bind nature to nurture. The Pattern of CV Disease in African Americans
As is well known, Mrican Americans suffer higher mortality rates from all major forms of CV disease (Table 1). In addition, the earlier age of onset of CV diseases among blacks results in a much larger differential when the measure is "years of productive life lost" (YPLL)(Table 1). Likewise, an enormous body of evidence suggests that the differential in incidence of CV diseases among blacks can be attributed primarily, if not exclusively, to hypertension. 5,6 Given the improvement in medical therapyboth preventive and curative-and the differentials in medical care received between blacks and whites, some proportion of the differences in survival after the onset of CV disease at the present time can be ascribed to differential access to both medical and surgical therapies. 7 The effect of in terventionaI therapies is likely to be particularly important for CHD. It is equally apparent that the social class gradient in incidence and survival from CV diseases accounts for much of the black disadvantage. 5 Although the impact of social class is a relatively straightforward problem when formulated in genTHE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
eral terms, as discussed below, the literature on this question is filled with controversy. The interpretation of adjustment for measures of socioeconomic status (SES) is particularly difficult.5,8 Although it is reasonable to use insights from the study of social class to inform the study of heart disease among blacks, the outcome may not always meet our expectations. For example, black men with higher SES have been found to have greater levels of obesity, contrary to whites, and smoking rates among young black women are relatively low. 1 It is likely, however, that a set of more generalized influences, from psychosocial factors to reduced medical care access, contribute to higher rates of morbidity and mortality. Likewise, although the descriptive information regarding SES differentials in risk factors has obvious utility for public health interventions, the use of these data in studies of causality has been questioned. 5,8,9 When it is not linked to specific research questions or accompanied by measurements that can provide insights into mechanisms, the emphasis on patterns of outcomes organized by SES has generated few new insights. A number of questions related to social factors can be identified that are uniquely relevant to minority populations, particularly African Americans. For example, it would be important to know whether specific cultural beliefs or practices influenced CV risk. Unique social stressors are postulated for minority groups, but not well characterized. Clearly racism and other forms of discrimination are important considerations here. On the other hand, nontraditional support systems might exist within these communities. Targeted research will be necessary to elucidate these relationships. Environmental Aspects of CV Risk among Blacks
A number of important areas for the study of environmental risk factors for CV disease among blacks can be identified. First, it must be acknowl-
209
Role of Genetic and Environmental Factors
Table 2. Multivariable Coefficients Associated with Major Coronary Risk Factors among Blacks and Whites, the NHANES I and II Cohorts
40 CI)
35
'u;
30
>
Women
Men Risk Factor Age Cholesterol Systolic BP Smoking
White .932* .029 .117 .453
Black .769 .048 .134 .009
White 1.188 .020 .152 .736
C
Black .906 .117 .113 .397
* Estimated with the proportional hazards model. Increment in risk for + 10 years, +10 mg/dL, +10 mm Hg, Yes/No smoking, respectively.
edged that the vast majority of disease-causing processes are common to all ethnic groups. Although an exhaustive list may not be available, most important risk factors that are thought to be important in whites have been verified in blacks,1° A more complicated question, however, relates to the relative impact of the individual factors and the manner in which they interact. Ultimately, any attempts to identify ethnic-group specific aspects of CV risk must be placed in the social context of these populations. Obviously, for African Americans, full accounting must be given to the enduring impact of racism on all aspects of health. We used the probability samples ofthe US population available through the First and Second National Health and Nutrition Examination Surveys to examine the predictive significance of the traditional risk factors (Table 2). Reasonable consistency exists, with the exception of smoking, which is notoriously hard to compare across populations. It should be recognized, however, that these cohorts provide information on only a combined total of 265 fatal CHD events for black men and women. Precision greater than that provided by this study is required to identify novel interactions or outcomes. As noted above, because hypertension can account for the great majority of the black excess in heart disease, the crucial question in this field reduces to an effort to understand the cause and potential prevention of hypertension in US blacks. An enormous literature exists on this problem, yet no solution is in sight. 6 It seems relevant at this juncture to introduce emerging findings from international comparisons among black populations. The prevalence of hypertension has been shown to vary, from 8-10% in rural Africa to 33% in the United States (Figure 1),11 Expressed against a common genetic background, this variation in phenotype emphasizes the determining role of environmental exposures. The relationship between known risk factors, such as obesity, closely parallels the cross-population trends in hypertension (Figure 2),11 Similar pat-
210
~CI)
25
C.
:>. 20 ::I: 15 C
CI)
(,) ~
10
CI)
C.
5
o Figure 1. Prevalence of hypertension among six populations of West African origin. The ICSHIE study, 19,95. Hypertension is defined as blood pressure 2 140/90 mm Hg, or antihypertension medication.
terns are observed for non -insulin-dependent diabetes mellitus, a CV risk factor of growing imp ortance. 12 An important challenge for descriptive epidemiology will be to elucidate more fully within populations the mechanisms by which risk factor exposures result in their physiological consequences. A Framework for Studying Racial/Ethnic Variation in Disease Risk
A paradox emerges in the analysis of racial differentials. The data available for descriptive purposes is overwhelming in its breadth and complexity, as suggested above, although the methods required for most comparative research are relatively straightforward. By contrast, the literature devoted to etiologic or explanatory analyses is impoverished and
35
Maywood
CI)
> 'u; c
30
~ 25 CI)
• S t. L·uCla ~arbados
Co
-
Jamaica
>-
J:
cCI)
20
• Cameroon
(,) ~
CI)
C.
15
o
Nigeria
10~-'---'----'----'----'----'
22
24
26
28
30
32
Body Mass Index Figure 2. Prevalence of hypertension by mean body mass index in six populations of West African origin. ISCHIE, 1995. March 1999 Volume 317 Number 3
Cooper
restricted, and often defaults to generalities about the "interplay of genes and environment." The logical construct, however, through which the "ethnic paradigm" can be pursued to arrive at causal explanations is exceedingly complex and is not well served by these simplistic generalities. Consequently, journals are filled with descriptions of shades of difference in health status, most of which are derivative effects from a few fundamental processes, with little attempt to define the underlying processes. The vascular sequelae of hypertension, for example, have been dissected in countless publications. The conclusions, more often than not, are restricted to speculation about the potential "genetic differences" that caused these secondary changes. The structure of the inquiry that could lead one to differentiate the bona fide effect of a cause from mere epiphenomenon is rarely defined, however. On balance, we have accumulated a large number of descriptive articles that seek to persuade us, by repetition more often than analysis, of "intrinsic differences" between blacks and whites. Because genetic and environmental influences are the focus of this presentation, it is important to pursue this problem to obtain further precision on the question of how we determine a causal relationship. Under what circumstances can something be said to be a cause? Although this problem is an everyday concern to working epidemiologists, it receives relatively little formal attention. The "Bradford Hill criteria" are familiar enough. Likewise, the properties of the randomized experiment, which make it unique as the basis for inference, are intuitively understood within the field. As on other counts, however, the work on racial differentials exposes important limitations in the application of causal inference. Informed primarily by the statistical literature, a modest body of work on causal inference exists in epidemiology.9,13-15 Crucial to this reasoning is the concept of "exchange ability" or "the alternative state."14 Formalized within the construct of the counterfactual, a framework is created that asks, "What would the outcome have been if this individual had never been exposed to the putative cause?" In a randomized experiment, of course, this means that we are justified in inferring that the outcome observed in the treated group would have occurred among the control subjects, had they been treated. With this generalization, we are equipped to make conclusions about the value of this intervention for similar patients in the future. For observational studies, however, the criterion of "exchange ability" is much harder to define. Control for confounding, through statistical adjustment, is the only recourse at hand. Although we can be reasonably comfortable with some forms of adjustment (for example, age, because it is plausible for THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
individuals to assume various ages), this method faces serious challenges when applied to attributes that define the essential character of an individual, such as gender or race. Thus, it is difficult to work with an analytic sequence that asks, "What would the risk of depression have been for this person if she had not been a woman?" A question naturally arises: can we speak of a woman as the same individual if she were not a woman? In fact, we would have difficulty defining what it means for a woman "to not be a woman." Practically, of course, the alternative is to be a man, but what does that tell us about "woman-ness" as a cause of depression? Although an adequate explication of this argument requires substantially more detail, it suffices in this context to make the observation that causal inference related to essential attributes, rather than discrete events, raises complex problems for the epidemiologic method. Similar, perhaps more difficult, problems confront racial analyses. 9 Although it is easy to accept that biological differences between men and women might serve as a reservoir of potential "causes," racial comparisons rely directly on the hypothesis that allele frequencies vary among the population groups and cause the observed contrasts in disease risk. Although it is acknowledged that some genetic factors do vary across populations, causal inference about the role of "essential" attributes is at least as complex for racial as for gender comparisons. To estimate genetic effects, one must control for environmental confounding. The infinitude of social influences, some explicit, many buried from view by unconscious prejudice, are more than a match for the techniques of statistical control. We recognize the risk of overweening inference; however, given the complete inability of studies on racial differences to yield useful etiologic insights, it is logical to ascribe this inability to the weakness of the pathway to causal reasoning. In effect, it follows from the argument outlined here that no valid inferences about "essential" differences between blacks and whites can be derived from observational studies. No generalized "genetic effects"-which constitute the only plausible meaning of essentialism in the biological context-can be detected using the methods of observational epidemiology. Genetic Aspects of CV Risk among Blacks
Although genetic effects have played a key role in etiologic reasoning for many years, useful measurements were not possible until the last decade. This technological revolution is slowly altering the theoretical approach to studying disease in human populations. In practice, however, the field of CV genetic epidemiology is still in its infancy. Many new markers were found to be associated with increased disease risk, but the findings often proved to be 211
Role of Genetic and Environmental Factors
Causal Effects Are Often Assumed to Fit an Additive Model:
Genes + Environment
=
Organism
In the Attempt to Isolate "Genetic Effects", this Model is Rewritten:
Figure 3. Modeling the contribution of genes and environment.
Observed Phenotypic Variation - Environmental Contribution
= Genetic Effect
However, Because Interactions Occur, Additional Terms are Required: IJ
Observed Phenotypic Variation
=
Genes + Environment + Genes
* Environment + Genes * Genes +
Environment
* Environment
r
I
t ! l
t
inconsistent in subsequent studies. For example, a number of polymorphisms in the renin-angiotensin system (RAS) have been linked to risk of hypertension and other forms of CV disease.1 6 - 19 Unfortunately, the totality of evidence on the RAS is contradictory. The approach using candidate genes has now been superseded by the genome-wide screen. These studies seek to identify, based on markers scattered across all chromosomes, regions of the genome that are consistently found among persons with the health condition of interest. After this stage, the painstaking effort of locating nearby genes must take place. This work has just gotten under way, however, and empirical data are not yet available. Because the methodology in this field is developing rapidly, both in terms ofthe laboratory methods and statistical analysis, it is difficult to predict the immediate future. It is clear, nonetheless, that genetic effects for complex traits, such as blood pressure, will be exceedingly subtle, thus necessitating very large sample sizes. Genetic epidemiology, therefore, may continue to become a more specialized field. In general, an obvious dichotomy exists between the approach to "genetic effects" that was standard in epidemiology for the last several decades and the molecular approach. Genetic effects in the past have been 212
•
modeled from measurements of phenotypes, typically as heritability. Statistical methods have been used to estimate overall effects (for example, comparing monozygotic and dizygotic twins). In epidemiology, genetic and environmental effects were often conceived of in an additive model, which together accounted for 100% of the phenotypic variation (Figure 3). Given this model, it was possible to rearrange the terms and isolate the quantity designated as "genetic effects." Assuming, however, that many interactions occur, it is obvious that a fully specified model does not permit the isolation of "genetic effects" in this manner (Figure 3). By the same token, with the advent of molecular methods, interest in modeling overall quantities has diminished markedly. The challenge now becomes cloning specific susceptibility genes and defining their physiological roles. These developments are particularly salutary in studies of ethnic variation, because they reduce the potential to confuse genetic effects and residual confounding. An opportunity now exists to seek reproducible, stable effects of specific molecular markers within individual populations. In an important way, the focus has shifted to identification of pathways by which DNA variants directly influence phenotypic traits. With these findings in hand, it will become possible to attempt comparisons across groups. March 1999 Volume 317 Number 3
I
I
! '
Cooper
Conclusions
Experience to date suggests that earlier attempts to model "genes and environment" as global concepts were almost certainly naive. In practice, the degree of precision required to make sense of a model at this level of generality far exceeded the capacity of the investigational methods available. The tendency to desist from the effort to make judgments about the underlying cause of "black-white" differences, when the causal process within neither group was understood, is a welcome change. In addition, the advent of molecular techniques has given research into genetic determinants a material basis, and an objective criteria for the evaluation of success. The requirement for those who wish to speak of genes as causes is now to name the gene and the sequence variation. Reasoning backwards from differences in observed phenotypes to hypothetical differences in genotype should no longer receive credence from the scientific audience. Racial categories imply genetic differences, and explanations of racial differences inevitably respond to this formulation. The conflation of genes and environment in the explanatory model, when both are poorly characterized, makes it impossible to define clearly their independent contribution. In particular, the search for specific susceptibility genes may be compromised by the desire to ascribe effects to global, yet undefined, "genetic" influences. What is required is an attempt to simplify the working model, eliminating those elements that are grounded in tenuous and, for the present, untestable assumptions. Perhaps as an outcome, on both counts, therefore, it is possible to argue that "less is more"-simpler descriptive models are needed first to define effects free of interactions; ethnic comparisons are hopelessly confounded and putative genetic effects should be judged by the strength of the molecular evidence. What confronts us is the task of unraveling the complex pathway from genes, as impacted by environment, to phenotype. In our desire to offer explanations, we clearly exaggerated the strength of the claims that could be offered on the basis of what is currently known. References 1. National Center for Health Statistics. Health, United States, 1996-97 and injury chartbook. Hyattsville (MD): Na-
THE AMERICAN JOURNAL OF THE MEDICAL SCIENCES
2. 3.
4. 5. 6. 7. 8. 9.
10.
11. 12. 13. 14. 15. 16. 17. 18. 19.
tional Center for Health Statistics, Centers for Disease Control, US Department of Health and Human Services; 1997. Publication no. (PHS) 96-1232. Liao Y, Cooper RS, Cao G, et al. Mortality patterns among adult Hispanics: findings from the NHIS, 1986 to 1990. Am J Public Health, 1998;88:227-32. National Heart, Lung, and Blood Institute. NHLBI Morbidity and Mortality Chartbook (1996). Bethesda (MD): National Heart, Lung, and Blood Institute, National Institutes of Health, US Department of Health and Human Services; 1996. Cooper RS. Ethnicity and disease prevention. Hum BioI 1993;5:387-98. Cooper RS. Health and the social status of blacks in the United States. Ann Epidemiol 1993;3:137-44. Cooper R, Rotimi C. Hypertension in African Americans. Am J Hypertens 1997;10:804-12. Ford ES, Cooper RS. RaciaVethnic differences in health care utilization of cardiovascular procedures: A review of the evidence. Health Services Res 1995;30:237-52. Moss N. What are the underlying sources of racial differences in health? Ann EpidemioI1997;7:320-1. Kaufman JS, Cooper RS, McGee D. Socioeconomic status and health in blacks and whites: The problem of residual confounding and the resiliency of race. Epidemiology 1997; 30:1511-7. Francis CK, Grant AO, Cooper RS et al. NHLBI report of the working group on research in coronary heart disease in blacks. Bethesda (MD): National Heart, Lung, and Blood Institute, National Institutes of Health, US Department of Health and Human Services; 1994. Publication no. (PHS) 95-129037. Cooper R, Rotimi C, Ataman S, et al. Hypertension prevalence in seven populations of African origin. Am J Public Health 1997;87:160-8. Cooper R, Rotimi C, Kaufman JS, et al. Prevalence of NIDDM among populations of the African diaspora. Diabetes Care 1997;20:343-8. Holland PW. Statistics and causal inference. J Am Stat Assoc 1986;81:945-61. Greenland S, Robins JM. Identifiability, exchange ability and confounding. Int J Epidemiol 1986;15:413-9. Greenland S. Randomization, statistics and causal inference. Epidemiology 1990;1:421-9. Jeunemaitre X, Soubrier F, Kotelevtsev YV, et al. Molecular basis of human hypertension: role angiotensinogen. Cell 1992;71:169-80. Caulfield M, Lavender P, Farrall M, et al. Linkage of the angiotensinogen gene to essential hypertension. N Engl J Med 1994;330:1629-33. Rotimi C, Cooper R, Ogunbiyi 0, et al. Hypertension, serum angiotensinogen and molecular variants of the angiotensinogen gene among Nigerians. Circulation 1997;95:2348-50. Rotimi C, Puras A, Cooper RS, et al. Polymorphisms of the genes in the renin-angiotensin system among Nigerians, Jamaicans and African Americans. Hypertension 1996;27:558-63.
213
4