A Selective Contemporary History of Cardiovascular Epidemiology
Abstract and Keywords
This chapter discusses the contemporary social and historical development of cardiovascular epidemiology. Given that epidemiology is at a historic moment in its development—social and genetic epidemiologists are mobilizing new conceptual models, methodological tools, and explicit scientific and political commitments to further their respective research agendas. Cardiovascular epidemiology has become a credible and consequential source of scientific and popular knowledge about heart disease, and a key site in which conceptions of race, class, and sex/gender are shaped and invoked. In particular, the conceptual framework of the multifactorial model in cardiovascular epidemiology shows the biomedical relevance of racial, socioeconomic, and sex/gender classifications, and enables their inclusion as flattened, reductionist variables in epidemiologic research.
This chapter provides a contemporary history of cardiovascular epidemiology, drawing mostly on secondary sources, supplemented with ethnographic data collected from epidemiologic conferences and interviews with epidemiologists. It emphasizes the contemporary social and historical developments most pertinent to the conceptualization and study of individual and population differences, particularly those of race, class, and sex/gender. I also include conceptual, methodological, and epistemological debates circulating both within and outside the world of epidemiology that are relevant to research on human differences. This history underscores the social and cultural shaping of epidemiologic practices and their significance for our current state of official knowledge on cardiovascular disease causation and risks.
As a discipline, epidemiology is at a historic moment in its development: Social and genetic epidemiologists are mobilizing new conceptual models, (p.49) methodological tools, analytic techniques, greater interdisciplinary collaborations, and explicit scientific and political commitments to further their respective research agendas. In so doing, they are reshaping the relationships among the profession of epidemiology, other disciplines, individual scientists, and the public. The existence of such theoretical, methodological, epistemological, and political heterogeneity complicates any exploration of the discipline of epidemiology. As Dora Roth explains, epidemiology “is regarded as a distinct and independent science (or discipline) not because it deals with special problems or because it has acquired a unique content, but because it has developed specialized procedures of investigation and application.”1 Departments of epidemiology exist throughout academic institutions in the United States, numerous professional associations and academic journals have been established, and funding agencies have provided broad support for epidemiologic research. All of these indicate the taken-for-granted assumption that epidemiology is a legitimate scientific approach to an ever-larger body of problems. Epidemiology is widely considered to be an authoritative mode of knowledge production on health risks and disease as well as a tool for policymaking. Cardiovascular epidemiology in its own right has become a credible and consequential source of scientific and popular knowledge about heart disease, and a key site in which conceptions of race, class, and sex/gender are shaped and invoked.
In the following section, I lay out the argument that the epidemiology of cardiovascular inequalities is a contemporary manifestation of biopower and domain of power-knowledge, but one that must work to maintain its public authority and credibility in the face of scientific and social disputes. I next provide a chronicle of the more recent history of cardiovascular disease epidemiology in the United States, focusing on the period from 1947—widely considered to be the debut of modern cardiovascular epidemiology—to the present. Specifically, I examine the emergence and consequences of the basic theoretical paradigm of epidemiology—the multifactorial model of disease causation—and its implications for the routinization of the measurement of race, socioeconomic status, and sex.2 In the final sections of this chapter, I describe the subsequent burst of activity in (p.50) cardiovascular epidemiology that served to place millions of individuals and multiple communities under the epidemiologic gaze and that provided the scientific basis for the making of claims about the nature of cardiovascular risks. I explore debates over the credibility of epidemiologic science that mark its relationship with the public and configure how the discipline and its practitioners consider research on health inequalities. This analysis illuminates the social conditions that structure contestations over the conceptualization, measurement, and interpretation of racial, class, and sex/gender inequalities in heart disease incidence and distribution.
Biopower and the Epidemiologic Gaze
The development of epidemiology more broadly and that of cardiovascular epidemiology in particular are part of a unique set of historical, social, political, cultural, and technoscientific contexts. Epidemiology has roots that reach as far back as Hippocrates3 but came of age approximately in the past century and a half. It emerged out of the same crucible that produced the development of bureaucracies and technologies for classifying and enumerating, the rise of statistical thinking and the social authority of quantification, the emergence of a scientifically rooted medical profession, and concerns to intervene in the public’s health.4 However, according to Olga Amsterdamska, noted sociologist and historian of science and medicine, “the first explicit and extensive efforts to define epidemiology as a science and to demarcate it from other neighboring fields date from the period immediately following World War I.”5
Epidemiology’s core concerns are with the patterns of health conditions in human populations and with the factors that influence these patterns. A fundamental project of epidemiology is to develop predictive models of health status.6 As such, epidemiology ultimately aims to apply its quantitative technologies and products to the governance of collective and individual health and the factors and behaviors that determine them. That is, epidemiology performs cultural work by tying together statistics about individuals and populations with particular conceptions of the “problem” (p.51) of health, disease, morbidity and mortality.7 In a word, the kind of cultural work that epidemiology does is what Michel Foucault calls biopower.8
Biopower, as described in chapter 1, is the mutually productive combination of power and knowledge, in which apparatuses and technologies exert diffuse yet constant forces of surveillance and control on human bodies and their behaviors, sensations, physiological processes, and pleasures. New disciplinary bodies of knowledge concerned with problems of population such as birthrates, longevity, and public health9 began to construct and establish what was considered normal and pathological across wide-ranging spheres of life.10 Such knowledge fixed and rendered individual differences “scientific,” arraying individuals around constructed notions of the norm and pinning them down in their particularity. Individuals became cases to be studied, enabling the refinement of more diverse and efficient techniques of control based upon the regulation, quantification, judgment, and hierarchization of human bodies.
Reflective of the preoccupations of modernity, epidemiology thus came to serve as an instrument of normalization; that is, it served not only as a form of representation but also as a means of instrumental control.11 As the historian Theodore Porter observes, “[M]easures succeed by giving direction to the very activities that are being measured. In this way individuals are made governable. … Numbers create and can be compared with norms, which are among the … most pervasive forms of power in modern democracies.”12 Weighted with the power of scientific reason and rhetoric, epidemiology became a significant tool for the disciplining of individuals and populations, for the exercise of power over behaviors and bodies, and for arbitrating contestations over different forms of knowledge and their legitimacy.
Dominant Paradigms of Disease Causation: The Multifactorial Model
By the turn of the twentieth century, the germ theory—the notion that diseases had specific, singular causes in the form of microorganisms—had (p.52) become the ruling paradigm for understanding disease causation in the West.13 Throughout the first half of the twentieth century, however, the credibility of the germ theory was gradually eclipsed by the epidemio-logic transition: the shift from the predominance of infectious to chronic diseases like stroke, cancer, and most notably, heart disease. The germ theory’s monocausal understanding of disease proved insufficient to account for the increased etiologic complexity and long latency of these new leading causes of death. Also, at the time of this transition, these chronic illnesses were understood to be degenerative, inevitable manifestations of the natural aging process.14 However, rising rates of chronic diseases in the industrialized West suggested that they were instead the outcome of multiple, specifiable, and changeable factors that might therefore be prevented. In epidemiology, such hypotheses eventually coalesced into the multifactorial model of disease causation.
In the multifactorial model, the incidence of chronic illnesses like heart disease is viewed not as a random phenomenon, nor as the inevitable outcome of aging, but as linked to identifiable factors of susceptibility and exposure. The multifactorial model posits that most of these illnesses are the result of multiple causes, determinants, and risks, involving complex “webs” of interactions among agent, environmental, and host factors.15 This is the dominant conceptual framework—the foundational assumption—underwriting the epidemiology of chronic disease.16 Thus, contemporary epidemiology understands the health status of individuals to be the outcome of their particular constellations of health risks and exposures, not in a deterministic but rather a probabilistic fashion.
The shift in focus from infectious to chronic disease after World War I occurred at the same time that governments in the West began taking on a greater role in health care and medical research. The state’s increased need for information on population health, and for ongoing surveillance of morbidity and mortality, in turn benefited epidemiologists whose proficiency with statistical models proved crucial to understanding population health.17 Thus the interwar period saw the synergistic development of and dependency upon epidemiology, with an emergent emphasis on (p.53) the complex web of causes contributing to disease incidence. The rise of a statistically minded epidemiology provided techniques through which multiple “risk factors”—social, environmental, and behavioral variables statistically associated with disease incidence—could be studied and analyzed.18 These new methods were critical to identifying, establishing, and legitimating associations between predisposing factors and disease incidence. The epidemiologic study of heart disease in particular emerged as a central player, shaping the development of this maturing discipline and both reflecting and contributing to the normalization of the notion of the “risk factor.”
The Emergence of Cardiovascular Epidemiology: The Framingham Study
The era of the epidemiologic study of heart disease in many ways began in 1947, when researchers at Harvard Medical School joined with officers of the U.S. Public Health Service to initiate a study in the town of Framingham, Massachusetts, located about twenty miles outside of Boston.19 The initial objectives of the Framingham Study were to develop and test methods for the early detection of heart disease and to screen an asymptomatic population to determine rates and incidence of the disease.20
In 1949, the just-created National Heart Institute (now the National Heart, Lung, and Blood Institute) took over the Framingham Study.21 With this transfer came several revisions to the study objectives and methods.22 First, the goals of the project shifted to identifying factors, including “bodily traits, [and] life habits,”23 that predisposed individuals to coronary heart disease. Specific hypotheses were formulated for testing, including the increase of coronary heart disease with age, male sex, hypertension, elevated blood cholesterol levels, tobacco smoking, habitual use of alcohol, lack of physical activity, increased body weight, and diabetes. Based upon these newly articulated hypotheses, data collection procedures and instruments were revised to gather information including various bodily measurements; blood pressure; consumption of (p.54) coffee, tea, alcohol, and tobacco; other dietary factors; physical activity; medical history; blood and urine chemistries; electrocardiograms and X-rays; measurements of pulmonary function; work status and other sociodemographic data.
The investigators subsequently had to reconsider the basis for sample selection. Felix Moore, who became director of biometrics at the National Heart Institute, concluded that the study sample needed to be more representative of the town’s population if incidence rate estimates and associations of heart disease to hypothesized risk factors were to be reliable and applicable to other regions and populations in the United States. Consequently, a random sample was drawn up and participants were recruited accordingly. Volunteers were also accepted, but their data were kept in a special category, as their characteristics might differ systematically from those of participants who were randomly selected. All told, 5,209 Framingham residents, both men and women, were recruited into the sample. (Several additional cohorts have subsequently been recruited to the study, including the original participants’ children [the Offspring Cohort] in 1971, their spouses [New Offspring Spouse Cohort] in 2003, and the grandchildren of the original participants [Generation Three Cohort] in 2001.24)
Over the past six decades, the Framingham Study has produced some of the most notable medical findings related to cardiovascular risk and the progression of heart disease. It identified the major heart disease risk factors—high blood pressure, high blood cholesterol,25 smoking, obesity, diabetes, and physical inactivity—and the influence of age, sex, meno-pause, and psychosocial factors.26 The contributions of the Framingham Study to the theoretical and methodological development of epidemiology are also highly significant. The epidemiologist Mervyn Susser argues that it was one of two major historic epidemiologic events (the other being the linking of lung cancer to cigarette smoking) that firmly established the multifactorial paradigm with its attention to potentially modifiable environmental factors.27 The Framingham Study is often held up as an exemplar of epidemiologic research and as the model and prototype for a prospective cohort study,28 a specific and (from that point forward) (p.55) widely used study design in which a group of individuals is followed over a period of time to observe who develops the condition in question.
The prospective cohort study was a critical conceptual and methodological development. As the science and technology studies scholar Anne Pollock observes, “[S]tudying healthy populations—rather than ones selected because they already had a particular disease of interest to investigators—was an innovation that reversed the lens of pathology … [a] shift from seeing the body as inherently healthy in need of occasional intervention … to one always at multiple risk for disease.”29 Indeed, as the first site in which the term “risk factor” was used instead of “cause,”30 the Framingham Study served to verify and integrate the very concept of risk factors into modern medicine.31 Thus while people were no longer seen as vulnerable to inevitable processes of degeneration, they were viewed as always potentially at risk and therefore in need of surveillance, screening, preventive services, and other medical interventions. Crucially, as the historian Robert Aronowitz points out, the risk factor approach did not so much displace existing models of disease as supplement them with new procedures for quantifying variables and establishing their relationships to illness.32 Thus Framingham served to demonstrate the utility of new methods of epidemiologic research, the ability of such studies to produce knowledge that could enhance the public’s health, and the legitimacy of a new perspective on disease progression. As a pioneer of a new paradigm of clinical surveillance and intervention, the Framingham Study stretched research much earlier into the formation of disease than ever before.
However, upon closer examination, the story of the Framingham Study is not so neat. In many ways, the study breached many of the foundational principles and rules for the conduct of a study that epidemiologists now take to be inviolate.33 For example, the basic objectives of Framingham were revised after the study and data collection had begun.34 As originally envisioned by the study’s founders in 1947, its purpose was to establish the incidence of heart disease among the general population. However, when the National Heart Institute took over the study in 1949, its rationale shifted to determining factors influencing the development (p.56) of heart disease. Only at this point was a required sample size estimated and data collection protocols and instruments accordingly revised; these continued to be modified over the years. However, recruitment efforts were not successful in gaining the participation of sufficient numbers of potential respondents selected at random, and the sample was therefore supplemented by volunteers. Yet, most descriptions of the Framingham Study in epidemiology textbooks and other accounts elsewhere omit these alterations, while others defend the methodological integrity of the study. For example, Thomas Dawber, a principal investigator of the Framing-ham Study in its earlier years, argues that with the modified objective of determining which differences among groups played a role in incidence, random sampling and high participation rates were no longer imperative:
Random sampling is not essential if the purpose of an epidemiologic study is to compare subgroups of the population determined by specific characteristics. The primary concern should be that the population contain sufficient numbers of subjects of these characteristics to enable comparison.35
In part because of these methodological workarounds, conclusions drawn from the Framingham Study have always been dogged by questions about their generalizability and the representativeness of the sample. While the study did include almost equal numbers of women and men and some attempts were made to measure indicators of socioeconomic status, concerns were raised about refusal and retention rates as well as the inclusion of only white individuals. Indeed, many questioned whether the study’s estimates of incidence and risk were reliable enough to be applied to other populations. Yet, according to Dawber,
there appears to be good reason to accept the Framingham findings as a reliable estimate of the actual incidence of the various disorders, with some obvious exceptions: there were too few black residents of Framingham to provide sufficient incidence data; the makeup of the white population was not completely representative of the U.S. white population; and there were (p.57) more participants of Italian extraction than would be found in most communities in this country. However, unless national origin plays an important role (which apparently it does not), the data reported may be considered reasonably representative of the North American white population.36
This conclusion seems curious, in that “national origin” (or “Italian extraction,” as it was termed in data collection instruments and some publications) among whites was summarily dismissed as playing a role in coronary heart disease incidence,37 whereas the importance of “race,” defined here as a black–white binary, was in comparison unquestioned. Given the relative lack of knowledge about racial disparities in heart disease incidence and risk at that time, the taken-for-granted significance of race is striking.38
Given these problems, along with the initial lack of recruitment procedures including eligibility criteria, data collection methods, and details of how the data were to be analyzed, Susser argues that the Framingham Study would never have been funded in today’s political, economic, scientific, and regulatory contexts.39 Indeed, in 1995 the Framingham investigators had to add the Omni Study, a cohort of minority residents, in order to comply with the new 1994 mandate that National Institutes of Health–funded research on human subjects include representation by minorities and women.40 Nonetheless, the “successes” of Framingham served to establish the now-dominant conviction that the determinants of heart disease can legitimately and usefully be thought of as constellations of measurable risk factors, that the varying prevalence of these risk factors across populations explains the varying incidence of heart disease, and that prospective studies produce sound epidemiologic data and scientifically defensible results.
Expanding Epidemiologic Surveillance and Biomedicalizing Difference
Since the Framingham Study, numerous cohort studies, involving scores of communities and millions of individuals, have been launched to (p.58) investigate etiologic factors of cardiovascular disease.41 These massive epidemiologic projects were part of a larger phenomenon in the United States to produce quantitative criteria for public decisions, aided by the success of quantification in the social, behavioral, and medical sciences alongside the emergence of a “culture of evidence.”42 Also, in the postwar period in the United States, the rise of laboratory-based medical sciences instigated efforts within epidemiology to redefine its own bases of scientificity, in part by presenting its statistical research designs as experimental and reframing communities as metaphorical laboratories.43 Long-term studies of communities therefore afforded a way for epidemiology to remain scientifically and socially relevant.
Cohort investigations of communities other than Framingham began in Tecumseh, Michigan, in 1957 and in Evans County, Georgia, in 1960. Other community-based studies include the Honolulu Heart Study, started in 1965 with a sample of Japanese American men (N=8,000)44 who enabled the investigation of cultural, dietary, and immigration factors in the development of heart disease.45 Beginning in the 1980s, several cohort studies were initiated that involved multiple recruitment and research sites. For example, the Atherosclerosis Risk in Communities Study (ARIC) began in 1987 (N=16,000), with sites across the United States, and a significant subsample of African Americans.46 The Coronary Artery Risk Development in Young Adults (CARDIA) study was initiated in 1984 with a sample fairly evenly divided among blacks and whites, and women and men from four urban areas (N=5,100).47 In 1989, the Cardiovascular Health Study began among a randomized sample of elderly people (N=5,000) in four clinical sites. More recently, the Multi-Ethnic Study of Atherosclerosis (MESA) was initiated in 1999 to observe and identify subclinical characteristics and risk factors of cardiovascular disease; approximately 28 percent of the cohort is African American, 22 percent Hispanic, and 12 percent Chinese American (N=6,500).48 Other efforts to focus more exclusively on the cardiovascular risk factors among women and people of color have accelerated and include the Nurses’ Health Study, begun in 1976 with a sample of female nurses (p.59) (N=122,000);49 the Strong Heart Study, launched in 1989 among Native Americans (N=4,500);50 the Women’s Health Initiative, which in 1993 started recruiting women from multiple racial and ethnic backgrounds (N=205,000);51 and the Black Women’s Health Study, initiated in 1995 (N=65,000).52
Some of the core epidemiologic claims about cardiovascular risk produced by these observational studies helped establish key risk factors.53 For instance, in the 1960s through the early 1970s, epidemiologic research first indicated links between diet, serum cholesterol levels,54 Type A behavior,55 and coronary heart disease. In the American Heart Association (AHA) Pooling Project, scientists from the AHA Committee of Epidemiological Studies–Subcommittee on Criteria and Methods decided to minimize the perceived uncertainties of an increasing number of small epidemiologic studies on coronary heart disease by conducting a statistical summary of risk factors and individual coronary heart disease risk from several select cohort studies. The Pooling Project firmly ascertained the quantitative relationships between cholesterol, blood pressure, smoking, and coronary heart disease risk, and its publications56 “had a major strengthening effect on the risk factor concept as the basis for preventive action.”57
The following decade was marked by the emerging understanding of the contributory role of lipids and lipid fractions,58 insulin, and alcohol consumption in cardiovascular risk. In the mid-1980s through the 1990s, interest in Type A behavior shifted to attempts to dissect component effects of hostility and anger59 and the risks associated with restricted social networks and social support.60 The importance of obesity and body fat distribution, and the role of diet, particularly the effects of antioxidants and different kinds of foods, were further elaborated.61 Now, genetic epidemio-logic research into genetic polymorphisms potentially linked to heart disease has begun in many of these cohort studies, including ARIC, CARDIA, MESA, and the Strong Heart Study. This time period also witnessed the growing understanding of the importance of risk factor clustering—that certain susceptibilities often travel together and have both independent (p.60) and synergistic effects on one another and on the cumulative risk for heart disease.62 Metabolic syndrome is probably the most commonly recognized of these clusters. It consists of central obesity (excess weight around the middle and upper parts of the body) and insulin resistance (wherein the body uses insulin less effectively, leading to elevated levels of fat and blood sugar), along with other risks such as physical inactivity and inflammatory factors in the blood. Insulin resistance is itself a constellation of interrelated risk factors and markers that include abnormal cholesterol levels, hyper-tension, and obesity; metabolic syndrome is also referred to as insulin resistance syndrome or Syndrome X.63
What are now considered to be the main, established risk factors for coronary heart disease include cigarette smoking, high blood cholesterol, high blood pressure, diabetes, sedentary lifestyle, and obesity. These factors also constitute the primary elements that go into clinical risk assessments. Male sex, family history of heart disease, and increasing age are also widely recognized as risk factors. However, these known risk factors altogether account for only approximately 40 percent of the cases of heart disease; that is, more than half of the individuals who have heart disease have none of these factors.64 Thus ongoing epidemiologic research is also examining a slew of newer risk factors—such as homocysteine, inflammatory factors, C-reactive protein, and fibrinogen (all chemicals or proteins found in the blood), among others—for their effects on the incidence and progression of cardiovascular disease, and for their predictive value in identifying who will eventually develop heart disease.
Under the intense surveillance of epidemiologic cohort studies, individuals classified as distinct groups and populations characterized by particular risk factors become sites for the further production of epidemiologic knowledge on cardiovascular risk. Indeed, often asymptomatic risk factors are seen as diseases in and of themselves, producing new classes of patients and changing our norms of what constitutes legitimate targets for medical intervention to include those that derive their meaning from their probabilistic associations with disease.65 When demographic and behavioral factors are implicated in producing more (p.61) frequent adverse outcomes, as is often the case, standards of conformity and deviance are created.
Moreover, the power of these knowledge claims and acts of judgment about health risks is magnified by the mask of scientific neutrality, generated through the constitution of epidemiologic knowledge and the rational, statistical arbitration of risk. As Porter notes, the validity of epidemiologic calculations, embodied in their seeming abstractedness and scientific objectivity, serves as “an agency for acting on people, exercising power over them. … Numbers turn people into objects to be manipulated.”66 Individual bodies are constructed not as the potential objects of medical control, but as the de facto objects of epidemiologic surveillance, under the current assumption that almost all bodies have one or more health risk factors.67 Epidemiology’s classificatory practices thus confer scientific legitimacy on the enterprise of risk assessment and management, imparting an aura of rationality to what are thoroughly social, power-laden, and ultimately hierarchical discourses, institutions, and practices.
These governing and disciplining knowledges and technologies of epidemiology reflect and sustain longstanding public health and biomedical concerns to categorize and pathologize populations by race, class, and gender.68 Such efforts do not go unresisted, however; there is a strong though often obscured history of counter-hegemonic understandings of the health effects of race, class, and gender. Many social movements have contested the dominant politics of disease causation and health care, even going so far as to set up alternative health systems that they believe took better account of the role of social difference and inequalities in the production of health.69 But even against the backdrop of this resistance, the epidemiologic gaze continued to sharpen its focus, such that scrutiny has converged upon select groups whose members manifest or embody a disproportionate share of the “problem” of heart disease. Based on the multifactorial model, epidemiologic research seeks to identify characteristics of the “host” which increase the likelihood that a category of individuals defined by those characteristics will develop (p.62) some condition or disease. As the social studies of biomedicine scholar Catherine Waldby explains,
This conceptualisation of disease aetiology means that epidemiological science can only proceed through the specification and classification of sub-populations. If the social topography of disease is taken to indicate a pattern of disease aetiology constituted at least in part through host factors, then hosts must be categorised according to these factors.70
Thus, the conceptual framework of the multifactorial model, which posits that differences in host characteristics contribute to disease, enables the biomedical relevance of racial, socioeconomic, and sex/gender classifications and their inclusion as flattened, reductionist variables in epidemiologic research. Conventional epidemiologic practices categorize the population at large into more specific classifications of higher- and lower-risk groups. The focus is thereby narrowed on sub-populations—often characterized by sociodemographic dimensions of race, socioeconomic status, and sex—that represent apparently more pressing concerns. (Indeed, Pollock argues that the very emergence of modern cardiology was deeply intertwined with efforts to differentiate and construct distinct racial and class groups.71) In so doing, however, epidemiology has run into a series of issues that I explore next.
Conceptual, Methodological, and Political Contestations within Epidemiology
Currently, the production of epidemiologic knowledge continues to escalate, and the public and policymakers increasingly turn to its findings for guidance on risk identification, assessment, and management. At the same time, the reliability, utility, and relevance of epidemiology has been subject to ever closer scrutiny both within the epidemiologic community and from without. In particular, debates over the front and back ends of the epidemiologic research process—setting the agenda and (p.63) research questions, and the interpretation and application of results—can be especially intense.
However, epidemiologists, like many other scientists, feel that public doubts about the utility of their discipline stem from the public’s fundamental misunderstanding of the objectives, interpretations, and inherent limitations of the scientific arbitration of risk. For example, the then-editors of the New England Journal of Medicine, Marcia Angell and Jerome Kassirer, found fault with the press for its reporting of epidemiology and with members of the public for their “unrealistic expectations” of what modern medical research can do for their health. They assert that “the public at large needs to become much more sophisticated about clinical research, particularly epidemiology.”72
Such representations of a misguided public echo the science studies scholars Alan Irwin and Brian Wynne’s observation that the dominant ideology regarding the public’s understanding of science assumes that lay people desire and expect certainty. The public is therefore seen as incapable of confronting “science’s ‘grown-up’ recognition that risk and uncertainty are intrinsic to everything.”73 However, within critical social studies of science, normative assessments of the public’s “misinterpretation” and “ignorance” of scientific objectives, methods, and results are considered to be framings of the “problem” that are socially constructed and power-laden. Such claims serve to fortify the demarcation between “expert” understandings and analyses of risk and disease and what the “lay” public is able to grasp, thereby sustaining the social credibility and authority of science.74 Indeed, qualitative field research indicates no such naïveté and instead shows that individuals and groups are quite sophisticated and sometimes more aware of the ambiguities and contingencies of scientific knowledge than scientists are forthcoming about them.75
In this vein, public doubts about the utility and credibility of epidemiology implicitly and explicitly question the ability of cardiovascular epidemiology, as currently practiced, to address the kinds of issues that some argue matter most for public health. The taken-for-granted nature of the multifactorial model, the risk factor approach, and other aspects of (p.64) the contemporary epidemiologic paradigm seem to have been disrupted to some extent. There is no longer the undisputed confidence that standard techniques will provide plausible and reliable answers to questions of disease etiology and risk reduction. Moreover, challenges such as the persistence of social inequalities in heart disease incidence and outcomes are proving increasingly pivotal to the public’s as well as scientists’ assessments of the efficacy of epidemiology. At their core, these contestations speak to the changing social conditions within which epidemiology plays an increasingly visible role yet paradoxically also comes under increasing fire. As such, epidemiology as a discipline constitutes a significant site of both public and professional surveillance, participation, and intervention.
In the following sections, I examine some of these contestations, using ethnographic and interview data, as well as findings from a content analysis of literature on epidemiologic theories, methods, and the state of the discipline. These debates involve some of the key concepts and discourses in epidemiology, including notions of causal inference, validity, and units of analysis. Although on their surface these debates appear to be highly technical, they go to the core of how scientific and social credibility are intimately intertwined, and they prove to be particularly consequential for the study of social inequalities in heart disease.
The Problem of Causal Inference
A major fissure within epidemiology and source of critique from without has to do with resolving the issue of causal inference, or the ability of a study to make claims about potential disease determinants or causes from the data. The choices of study design and methods are absolutely critical. Evolving alongside advances in basic and clinical research, epidemiology has developed a well-understood hierarchy of research models and kinds of data based upon the perceived relative validity of the answers each provides. This hierarchy embodies, in a sense, the epidemiologic conventional wisdom on the ability of various kinds of study designs to make scientifically supportable claims about causality. At the apex of this hierarchy (p.65) is the randomized controlled trial, in which the exposure to some suspected causal factor is applied to a randomly selected experimental group of research participants, while it is replaced with some kind of placebo or alternate intervention for the other, control group.76 This design enables the assumption that the experimental and control groups are quite likely to be “otherwise equal” with the sole exception of the exposure or risk factor under study. This assumption in turn permits the inference that any observed effect must be due to the independent contribution of the experimental intervention. This ability to experimentally and deliberately control the exposure independently of other factors and the random selection of who is to be exposed are widely represented as making the randomized trial the most powerful research tool for ascertaining causation. Indeed, I found that in scientific meetings, public lectures on heart disease risks and causes, interviews with epidemiologists, and in epidemiology textbooks, these features of clinical trials are repeatedly noted in order to bolster the legitimacy of their conclusions.
On the other hand, observational data of the kind generated in large quantities by heart disease cohort studies,77 like those described earlier in this chapter, are regularly depicted as coming in a poor second. In fact, observational data are often viewed as suspect in that they can lead to misconceptions about disease etiology and risk that sometimes prove to be obdurate, even in the face of later, “more accurate” clinical trial data that counter observational findings. For example, epidemiologists I interviewed frequently note that data from observational studies of the effects of hormone replacement therapy on heart disease risk led to the belief that it reduced cardiovascular risk for postmenopausal women.78 Yet this conventional wisdom was subsequently contradicted by data emerging from several large, highly reputed clinical trials. Respondents thus repeatedly caution that findings from observational studies must be viewed as provisional answers and not definitive, “conclusive” ones, which can be obtained only through randomized controlled trials.
However, assessing causation in chronic conditions like heart disease involves substantial uncertainties on pivotal issues such as the timing and (p.66) nature of exposures, long incubation periods, the progressive nature of the disease, and the multiple co-factors involved in pathogenesis.79 For example, many of the risk factors under study in cardiovascular epidemiology, such as sedentary lifestyle, obesity, and high cholesterol, take years to produce results large enough to be measured. Moreover, these risk factors often interact with one another, complicating attempts to ascertain causes, mediators, and effects. Such uncertainties further undermine the ability of studies to make causal claims in the case of complex, chronic diseases.
Moreover, statistical association is not the same as a causal relationship. Although historically in the development of epidemiology, statistical methods enhanced the scientific status of the discipline and the plausibility of its flagship disease model—multifactorial causation—critics point out that at the end of the day, epidemiology rarely is able to say that risk factor X is a cause of disease Y. In fact, exactly what is meant when the term “risk factor” is invoked is ambiguous; it can refer to cause, association, predisposition, or susceptibility. This embodies, as Aronowitz notes, both the limitations and the attraction of the risk factor logic80 that characterizes a great deal of cardiovascular epidemiology. While the precise meaning of a statistical association is highly ambiguous, scientists can avoid having to equivocate among one meaning or another. Instead, risk factors can be defined “by utilitarian or empiric criteria,”81 simply by their demonstrated statistical relationship to heart disease. Indeed, the multifactorial model could be construed as “an objective model of disease causation with no beginning or end—just multiple, interacting associations … an empirically driven and often mechanismless multicausality. … Whatever worked in a model was potentially causal. … It was left to statistical techniques to sort out the relationship among factors.”82 Thus, in settling questions of causality—the very reason why such studies are presumably launched in the first place—epidemiologic research frequently falls short.
In addition, the factors of most interest here—racial, class, and gender differences—cannot be randomly “assigned” to individuals as can a drug or a low-fat diet. Moreover, even if they could, it could not be done so (p.67) independently of other factors that would make the experimental and control groups “otherwise equal.”83 One’s race, class, and gender permeate and indelibly shape so many other aspects of life potentially related to cardiovascular health that it would be impossible for groups to be “otherwise equal.” This thereby violates a key assumption that helps to validate inferences of causality. Thus, not only feasibility but also epidemiologic validity becomes problematic in the quest to address issues of social inequalities and their consequences for cardiovascular health through clinical trial research. Observational research, then, represents the next best solution.
But in relying on observational studies, epidemiologic research on the roles of sociodemographic inequalities in heart disease incidence and distribution encounters other methodological quandaries that have hampered its ability to state with confidence the conclusions of a study and its implications. Such research depends upon the reliable collection of myriad factors that influence cardiovascular health. Through the application of multivariate statistics (an analytic analogue to the theoretical framework of the multifactorial model84), it is assumed that the contributions of each epidemiologic variable—including categorical race, socioeconomic status, and sex—to the outcome can be independently calculated. However, the impossibility of disentangling the effects of one variable from those of another, and from other aspects of life that may affect heart disease incidence, means that the assumption of “independent” variables is again violated.
Disputes about Validity
A second methodological quandary that epidemiology runs into, particularly when investigating inequalities in heart disease, is the problem of validity. In designing a study, there is always some amount of tradeoff between what are termed “internal” and “external” validity. Internal validity is closely related to the issues of causal inference discussed above. It refers to the extent to which conclusions drawn from the data in fact (p.68) reflect the normative “reality” of the group being studied per se and are not the outcomes of biased or chance observations. This aspect of epidemiologic research is often encapsulated in calculations of a study’s statistical significance, confidence intervals, and so on. External validity, in contrast, refers to the degree to which the conclusions drawn about the group under study can be applied to other groups and populations; this characteristic bears on concerns about a study’s generalizability.85 Thus the selection of a study sample that is heterogeneous, while enhancing its external validity because the sample mirrors more closely the natural variability within the general population, may threaten its internal validity because so many differences exist within the sample that it is difficult to tease out which play a role in the outcome of interest. On the other hand, the selection of a more restricted sample has the potential to strengthen its internal validity, but at the expense of its generalizability.
Because of this seemingly unavoidable tradeoff, research on samples that are diverse along sociodemographic lines may be viewed as scientifically suspect in terms of their internal validity. On the other hand, studies of groups that are limited by racial category, socioeconomic status, or sex—samples, in effect, defined by their “difference”—are viewed as circumscribed in their applicability to other groups. Working on the assumption that results without internal validity are of little use to either the group being researched or to other populations, many epidemiologists tend to give up external validity for internal validity, creating a body of knowledge based on fairly circumscribed and homogeneous samples.
The use of complex social variables like racial categories and socioeconomic status86 will always, to some extent, be plagued by questions of scientific validity (the extent to which they measure what they set out to measure), and reliability (usually judged by the degree to which measurements can be replicated across space and time). Within epidemiology, stable and quantifiable markers that reside in the biological body are constructed as more definitive and accurate indicators of risk and effect, or, as one epidemiologist put it, as more likely to be “really measuring what you want [them] to measure,” and therefore less threatening to internal (p.69) validity. Causal claims are thus seen as far more legitimate when they emerge from studies using biologically manifested measures. Epidemiologists I interviewed who use or had considered using measures of racialized or gendered experience or class-based exposures that went beyond the conventional indicators recounted the subtle pressures they felt to “prove” the scientific acceptability of such measures. These epidemiologists perceived being “disciplined” by their colleagues and by epidemio-logic convention, and socialized as scientists to choose instead biological indicators perceived to be more reliable because they are viewed as more replicable.87 Epidemiologic research on the causes of social inequalities in heart disease, then, often falls short of ideal standards of measurement and validity in this regard.
Critiques of Conceptual and Methodological Individualization
In addition to these methodological concerns of validity in the epidemiology on racial, socioeconomic, and sex differences, a host of broader, more fundamental epistemological debates figure prominently in the often fractious world of epidemiology. A primary point of contention is the devolution of the focus of epidemiologic concepts and practices to the individual. Because most epidemiologic findings emerge from research in which the unit of analysis is the individual, processes and dynamics of disease risk and causation are systematically reduced to the level of the individual, simultaneously simplifying a complex world into smaller, presumed independent units of observation.88
The image of a complex and interconnected “web” of both causal and protective factors that together determine an individual’s health status tends to concentrate attention on those risk factors closest to the outcome of interest. These typically translate to the direct biological or behavioral risks addressable at the individual level.89 Nancy Krieger, a prominent social epidemiologist, for example, finds that at critical shifts in the historical development of epidemiology, practitioners were exhorted to focus on identifying “causes” most amenable to medical intervention as (p.70) close to the specified outcome as possible, given that “‘even knowledge of one small component may allow some degree of prevention.’”90 Even so-called psychosocial and social-behavioral variables—such as health-seeking behaviors, social support, and lifestyle factors like exercise, diet, and smoking—are measured at the individual level. Underwriting such methodological practices is the notion that these factors, despite their occasional designation as “social” variables, are individually mediated, the independent characteristics and behavioral choices of decontextualized individuals. Some epidemiologists thus argue that these practices implicitly assume that characteristics of one’s social environment, as incorporated in the multifactorial model, are exogenous to the individual—that one’s circumstances are taken as a given, as if individuals were dropped into a set of conditions that are not socially constructed or patterned.
Moreover, the kinds of interpretations conventionally given to statistical associations between risk factors and measures of race, socioeconomic status, and sex serve to further reinforce the “individualization of risk,” perpetuating the notion that risk is individually, rather than socially, determined.91 For example, the recognition that rates of sedentary lifestyle or physical activity vary systematically by race, income, education, and sex92 is often interpreted to be the consequence of the lifestyle choices and risk management routines of individuals.93 At other times, studies examining rates of physical activity adjust or control for racial category and socioeconomic status or compare across racial and socioeconomic categories,94 often with little discussion or explicit study of the complex social contexts and causes leading to sedentary lifestyles. The first convention clearly individualizes the risk of physical inactivity, fostering “blame the victim” assumptions, while the second often can leave the impression that such associations are to be expected given the populations in question, playing into common racial and class stereotypes.
Conventional epidemiology therefore remains predominantly concerned with the identification of individual-level risk factors,95 earning it the label of “risk factor epidemiology.” Such consequences further (p.71) concretize the notion that non-intersectional and individualistic constructions, such as racial classifications, socioeconomic status, and sex, additively help “explain” the distribution of health and illness, pushing aside uncertainty and ambiguity over what exactly about one’s race, class, sex, or gender shapes risk for chronic illness and how these mechanisms operate in concrete practice.96 There is an inherent contradiction here: Race, class, and gender are not attributes of individuals but refer to relations among socially defined and differentiated groups. Thus an epidemiology of individuals and their risk factors will not be able to account for such group-based processes and relationships. All of these slippages and conceptual imprecision, some epidemiologists argue, produce an overly simplified, reductionist, and therefore inaccurate and scientifically invalid picture of disease causation and risk vis-à-vis such categories.
Thus, many social epidemiologists—who represent an important subfield within the larger discipline—advocate fundamental conceptual and methodological shifts that require the incorporation of variables at the group level into the analysis of individual-level health outcomes.97 Social epidemiologists argue that social relations may affect individual health outcomes independently from individual factors,98 an epidemiologic analogue to the sociological proposition that individuals are shaped not only by their personal characteristics but also by the characteristics of the social groups to which they belong. As Krieger asserts,
the essential claim is that understanding patterns of health and disease among persons in these groups requires viewing these patterns as the consequence of the social relationships between the specified groups. … This perspective … asks how individuals’ membership in a society’s historically-forged constituent groups shapes their particular health status, and how the health status of these groups in turn reflects their position within the larger society’s social structure.99
Social epidemiologists thus promote the thesis that disease is socially produced and that the relative positions of socially designated groups (p.72) and the structural processes and institutions that maintain such positions are highly consequential for their health. A more comprehensive understanding of disease etiology must therefore include the investigation of social, cultural, and political forces as causes of illness as well as the biological ones, and of group-level dynamics as well as individual characteristics.100
Though the inclusion of both individual- and group-level factors in epidemiologic studies may appear at first mention to be a relatively innocuous practice, it in fact challenges many of the discipline’s standard operating procedures and assumptions about the scientific validity and legitimacy of different types of research and data. With the “individualization of risk”101 and the focus on individual factors and outcomes, group-level data is perceived by traditional epidemiologists to be less useful in contributing to etiologic understanding. Further, such data can even be viewed as somewhat suspect in that they can indicate associations between some variable and outcome that may not bear out at the individual level. The lore of epidemiology is in fact littered with famous examples of such “ecological fallacies.”102 Common epidemiologic practice usually proscribes mixing individual-level with community- and macro-level variables, because numerous statistical complications can arise, rendering problematic the drawing of causal inferences from such multilevel data.103
However, social epidemiologists counter that while such concerns are indeed valid, “the complexity of developing theoretical formulations that relate multiple levels … is likely to be a better reflection of reality than the simpler multicausal models prevalent today.”104 The epidemiologists Mervyn and Ezra Susser,105 for example, argue for moving beyond the multifactorial framework of disease causation to a causal model based on the metaphor of Chinese boxes, a set of boxes of differing sizes that nest within one another. The essence of this new paradigm is that disease causation occurs on multiple levels, and the integration of these levels is critical to investigating and solving a designated problem. In using this conceptual framework then, “hypothesis, design, and analysis (p.73) would always keep in focus the object of viewing all the relevant levels as a whole. Each level is seen as a system in itself that interacts with those above and below it.”106 Social epidemiologists also express great interest in plumbing mathematical modeling and statistical procedures designed to account for complex, fluid, and dynamic systems that are being used in other disciplines, including econometrics, climatology, and the study of global warming.107
The epidemiologist Ana Diez-Roux also proposes that alternate types of causation be considered:
These include not only causal determination (determination of the effect by an external cause, as in “among susceptible individuals, smoking causes lung cancer”) and statistical determination (as in “x percent of persons with high cholesterol will develop a myocardial infarction”), which are the types of determination commonly implicit in epidemiologic research, but also other types of determination such as reciprocal causation and structural or holistic determination. … Reciprocal determination (determination of the consequent by mutual action) would be present if, for example, a person’s consumption of “unhealthy” foods is influenced by the types of foods available where he or she lives, and if in turn food availability is influenced by consumption in the area. Holistic determination (determination of the parts by the whole) would be present if a person’s risk of adopting a certain behavior were influenced by the prevalence of that behavior in the social group to which he or she belonged, or if a person’s risk of disease depended on the degree of social inequality in his or her society.108
The concepts of reciprocal and holistic determination aim to consider the contextual and social conditions and situations within which people live and work and allow for synergistic and more complex relationships between individuals’ traditional risk factors and their social, cultural, and material communities and environments. Such new definitions of causality are potentially important to the study of sociodemographic (p.74) inequalities in heart disease, as they provide fresh conceptualizations of those cardiovascular risks that are frequently mobilized in explanations for racial, class, and gender differences, as discussed in the Introduction and in chapters 3 through 5.
Finally, social epidemiologists emphasize the distinction between the question of why some individuals are at higher or lower risk than others and that of why some populations as a whole are at higher or lower risk. The work of the epidemiologist Geoffrey Rose109 has been theoretically influential in formulating a different set of etiologic questions based upon this distinction. Rose points out that the set of factors which explains why a particular individual has a disease may be very different from society- or population-wide forces that produce a populace with a particular distribution of risk for that disease. Relatively rare factors may explain the risk distribution of individuals while widespread and common factors can account for the distribution among populations. The former approach is characteristic of conventional risk factor epidemiology, while the latter articulates a population-based strategy.
As a basic science of public health, epidemiology has had to deal with a public that is increasingly diverse, more demanding that science acknowledge and account for its diversity, and at times openly skeptical of the utility and credibility of epidemiology as a relevant and beneficial scientific discipline. Critiques and commentaries on mainstream epidemiology, both from within and from outside the profession, have occasioned much contestation and debate, even earning the label of “the epidemiology wars.”110 Great divides exist on what is considered causally “fundamental”: Some argue that social, political, and structural influences underlie individual and biological causes of disease.111 Others argue that all levels of causation—social structural, individual, and genetic—matter, and that it is meaningless to designate one as more important than another.112 And finally, still others simply observe that the balance in (p.75) the epidemiologic knowledge currently being produced and in the body of epidemiologic literature as a whole needs to be tipped more toward the consideration of societal and structural factors, as currently it tends to favor the individual and micro levels.113
Certainly there is much contestation within the epidemiologic community over whether more, better, or radically different theoretical and methodological explications can strengthen the discipline and perceptions of its utility and credibility. Simultaneously, there is considerable deliberation about how epidemiology can achieve its goal of determining disease mechanisms and improving public health, and what kinds of scientific, methodological, and/or conceptual endeavors might accomplish this.114 Debates over whether deep and fundamental changes to the practice of epidemiology may be in order have filled the pages of leading epidemiology journals, the hotel hallways and ballrooms of epidemio-logic conferences, and the narratives of cardiovascular epidemiologists interviewed.
Mervyn Susser, for example, attributes the present “stagnation and inertia” of epidemiologic science to the ultimate failure of the dominant conceptual paradigm—the taken-for-granted multifactorial model of causation—to further illuminate chronic disease etiology and stimulate new thinking in epidemiology.115 Instead, he argues, “The dominant risk factor black box obscures our vision and impedes our capacity to deal with the near future.”116 As indicated by the deficiency of current epidemiologic knowledge and procedures, “the signs are ominous that we are nearing its displacement by a new era. … [W]e need either to adopt a new paradigm or face a sort of eclipse.”117 Raj Bhopal, in his review of epidemiology textbooks, argues that their heterogeneous vocabulary to describe concepts and methods, and differences in perspectives on the purposes and scope of epidemiology, can be read as Kuhnian indications of an evolving discipline or signs of an imminent paradigm shift.118 Others suggest that epidemiology may already be in transition “from a science that identifies risk factors for disease to one that analyzes the systems that generate patterns of disease in populations … from relationships (p.76) between exposure and disease variables to the analysis of systems that give rise to exposures and through which those exposures act to cause disease.”119 But the fact that at least within the epidemiologic community, these debates largely revolve around issues of science—its concepts, models, methods, and the like—reflects Porter’s observation that
scientific knowledge is most likely to display conspicuously the trappings of science in fields with insecure borders, communities with persistent boundary problems. … Science is indeed made by communities, but communities that are often troubled, insecure, and poorly insulated from outside criticism. … The enormous premium on objectivity in science is at least partly a response to the resultant pressures.120
Whether epidemiology is to be transformed by approaches infused with new theoretical frameworks or by renewed commitments to conventional definitions of scientific rigor and validity, or by both, or by altogether other forces, social and cultural analyses of science argue that scientific disciplines do not “progress” on some linear path or by intellectual considerations alone.121 Ultimately, no amount of methodological reform can “rid” epidemiologic knowledge of the social and political concerns of the actors who construct it and the broader social, cultural, and political situations within which they must do so.122 Instead, social, cultural, political, and economic forces and the content and conduct of science must be viewed and analyzed as mutually shaping and constituting one another. This co-production of science and society can be especially visible and acute when considering the nature and influence of human differences on human health, as the chapters that follow will show.
(2.) As mentioned in the Introduction, I use the terms “race,” “class,” and “gender” for the most part to refer to social relations of power, and the terms “racial category/ies,” “socioeconomic status,” and “sex” to refer to their transposition and devolution into individualized, discrete variables typically used in epidemiologic research.
(9.) Foucault (1978). Indeed, Schweber (2006: 25) argues that “Foucault’s identification of ‘population’ as one of the primary objects of modern political power gave a (p.231) new currency to population statistics in particular”—that is, to epidemiology and other allied disciplines.
(15.) Host factors include, for example, immunologic status, genetic background, socioeconomic level, hygienic practices, behavioral patterns, age, and the presence of co-existing disease (Evans 1978; Gordis 2000; Lilienfeld and Lilienfeld 1980; MacMahon et al. 1960; Susser 1985).
(19.) I do not mean to suggest that the surveillance of individuals and communities was a new phenomenon emerging in the mid–twentieth century. In fact, governmental apparatuses to measure, survey, record, and regulate patterns of health and disease have been in existence since at least the eighteenth century (e.g., Armstrong 1983; La Berge 1992; Rose 1990). The nineteenth and early twentieth centuries saw a proliferation of government agencies, professional specializations, and philanthropic and volunteer organizations set up to deal with public health issues, including surveillance and intervention activities into practices of health, lifestyle, self- and (in the case of women) other-care, and hygiene (e.g., Fee and Porter 1992; Rogers 1992). However, in the case of studies undertaken to address etiologic questions of cardiovascular disease, the mid–twentieth century does represent a watershed in the mass surveillance and measurement of populations via epidemiologic techniques.
(21.) New organizational guidelines allocated research into methods of disease prevention and control to the National Institutes of Health, while projects concerned with practical control measures belonged elsewhere in the Public Health Service (Dawber 1980: 17).
(30.) Nieto (1999) and Aronowitz (1998: 223–4, note 2) cite the first published use of the term “risk factor” to a paper by Kannel and colleagues (1961). Aronowitz (1998: 97–8) notes, however, that the notion of “disease predisposition,” prominent in the first half of the twentieth century, anticipated the conceptualization and identification of “risk factors” with respect to heart disease.
(32.) Nor, as important, did it force physicians to engage further in preventive practices so much as add to the list of conditions (e.g., hypertension, hypercholesterolemia) they could treat medically (Aronowitz 1998).
(37.) Moreover, as Pollock (2007: 99–100) argues, the location for the study was legitimated in part through claims that Framingham was a “normal,” typical, average, and therefore representative reflection of a pan-white, Euro-American town. Two additional findings from Pollock’s work are worth noting here. First, this notion of a homogenous, “white” population is one that had to be actively constructed in the process of collecting, recording, and storing data on categories variously termed as “race,” “national origin,” and even nutrition for the Framingham Study (Pollock 2007: 111–8). And second, the language shifted fairly quickly from claims of representation to those of extrapolation, that the Framingham sample was not so “grossly atypical” that the results could not be generalized to other (white) Americans (Pollock 2007: 103–10).
(38.) Moreover, what we take to be “heart disease” has changed since the Framingham days. Pollock explains this through her comparison of Framingham to the Jackson Heart Study, begun in 2000 and often referred to as the “Black Framingham.” Similar to Framingham, the Jackson Heart Study selected participants from the residents of a particular town, this time African Americans living in Jackson, Mississippi, to study atherosclerosis (the accumulation of fatty plaques along the walls of the coronary arteries) and its risk factors among this high-risk racial group. As Pollock argues: “Framingham has attended first and foremost to the coronary artery disease from combined risk factors of the white middle class. Jackson attends to the burden of morbidity and mortality on people of color who were left behind by America’s post-war medical advances. Medical knowledge about the diseases of the heart comes to include an expanding notion of who counts as American: from WASP, to pan-white, to racially fractured, and in parallel from high SES to middle class and then also low SES” (Pollock 2007: 139).
(p.233) (41.) Numerous epidemiologic studies examining the effects of preventive programs at the community level have also been initiated. However, as these studies investigate the efficacy of heart disease interventions, rather than the determinants of heart disease incidence and distribution, they are outside of the scope of this project and are not discussed here.
(44.) Approximate total sample sizes are provided to give a sense of the large numbers of individuals being enrolled in these studies.
(55.) In the Framingham Study, Type A behavior pattern was measured through questionnaires that included the following questions and items: (1) “traits and qualities which describe you: being hard-driving and competitive, usually pressed for time, being bossy or dominating, having a strong need to excel in most things, eating too quickly”; (2) “feeling at the end of an average day of work: often felt very pressed for time, work stayed with you so you were thinking about it after working hours, work often stretched you to the very limits of your energy and capacity, often felt uncertain, uncomfortable, or dissatisfied with how well you were doing”; and (3) “do you get upset when you have to wait for anything?” (Haynes et al. 1978: 382).
(58.) Lipid fractions include low-density and high-density lipoproteins—the now fairly well-known LDL (or “bad”) and HDL (or “good”) cholesterol—as well as triglycerides.
(69.) See, for example, Clarke and Olesen (1999), Morgen (2002), Nelson (2011), Ruzek (1978), Smith (1995), and White (1994). It is important to note that many of the struggles chronicled in these works articulated racial, class, and gender oppressions as of a piece, woven together, and whose intersecting, cumulative effects impoverished the well-being of multiple vulnerable groups.
(75.) For example, Irwin and Wynne (1996b). Irwin and Wynne (1996a: 218) conclude that “what scientists interpret as a naïve and impracticable public expectation of a zero-risk environment can … be seen instead as an expression of zero trust in institutions which claim to be able to manage large-scale risks throughout society.” The analyses described in the following chapters seek in part to explore this proposition in the context of accounts of cardiovascular risk and causation.
(77.) Observational data come from cohort and similar kinds of studies, where interventions or treatment are not being attempted.
(78.) Interestingly, a 1985 publication of results from the Framingham Heart Study (Wilson et al. 1985) indicated that HRT increased women’s risk of stroke, but this finding was dismissed until the Women’s Health Initiative publication in 2002 (Levy and Brink 2005).
(79.) In fact, because of the etiologic complexity of chronic disease, the term “risk factor” rather than “cause” is usually applied (Evans 1978: 167). Some (e.g., Dawber 1980) have gone so far as to state that “cause” is a word that epidemiologists prefer to avoid, as it connotes that a factor labeled as such would lead to disease in every case of exposure, when disease-related factors that still have a causal effect rarely if ever work in such a fashion. While the judgment of whether a particular risk factor operates in an associative or in a cause–effect manner is traditionally represented as a separate stage in the epidemiologic method, in everyday (p.235) epidemiologic parlance the terms “risks” and “risk factors” often refer to suspected direct or indirect “causes” or “causal mechanisms” of heart disease.
(85.) External validity also depends on the response rate and retention of participants throughout the course of a longitudinal study, in addition to the initial eligibility criteria for inclusion that are discussed here.
(86.) How cardiovascular epidemiologists manage the measurement and conceptualization of such social variables is the focus of later chapters.
(100.) Many commentators (e.g., McMichael 1995; Pearce 1996; Vandenbroucke 1994) have observed that the concept of an epidemiology that takes seriously, in theoretical and methodological terms, the influence of social relations and structural environments is not new. They argue that these kinds of approaches are in fact a return to the ecological, environmental, community, and social concerns of epidemiology’s early pioneers, such as Edwin Chadwick, Friedrich Engels, William Farr, Rudolf Virchow, and John Snow. Susser (1985: 153) notes that more recently, during the mid-twentieth-century epidemiologic transition from infectious to chronic diseases as leading causes of death in the United States, it became increasingly clear that social and environmental factors were significant to the incidence of illness. Key innovators in this tradition include Cassel (1976), Graham (1963), Susser (1964), and Syme et al. (1965).
(p.236) (102.) An often-used example of an ecological fallacy is the assumption, based on observations that societies with high-fat diets have high rates of heart disease, that high-fat diets lead to heart disease. Such an assumption is impossible to make from this aggregate level of data as it is not possible to tell whether those individuals developing heart disease in a particular country are in fact those with the high-fat diets; exposure cannot be linked to the incidence of disease. However, this focus on the individual level to the virtual exclusion of group-level factors is changing somewhat; these moves are taken up in the Conclusion.
(110.) Poole and Rothman (1998). See, for example, Coughlin (1998), Diez-Roux (1998a), Kogevinas (1998), Koopman (1996), Krieger (1994), Krieger and Zierler (1996; 1997), Mackenbach (1998), McMichael (1995), McPherson (1998), Pearce (1996), Savitz (1997), Savitz et al. (1999), Shy (1997), Susser (1989; 1998), Susser and Susser (1996b), and Taubes (1995).
(118.) Bhopal (1997; 1999). Kuhn (1962/1996) argues that in “normal science,” research is firmly based upon past scientific achievements that a particular scientific community accepts as the foundation for its practice. A reigning paradigm provides the conceptual means to guide current research and by which certain problems are taken to be scientific and the proper domain of scientists. Normal science continues until anomalies accumulate, producing a scientific crisis and engendering a paradigm shift. Kuhn’s emphasis on the accumulation of anomalies as the driving force behind scientific crises and paradigm shifts has been critiqued by more recent works in science studies (Gieryn 1999; Hess 1997; Star 1986, 1989) as overly intellectualist and neglectful of the influence of broader social and cultural transformations within which the practice of science is always cradled.
(121.) As Gieryn (1999: 23) argues, “[I]f the stakes are autonomy over scientists’ ability to define problems and select procedures for investigating them, then science gets ‘purified,’ carefully demarcated from all political and market concerns, which are said to pollute truth; but if the stakes are material resources for scientific instruments, research materials, or personnel, science gets ‘impurified,’ erasing the borders or spaces between truth and policy relevance or technological panaceas. The sociological question is not whether science is really pure or impure or both, but rather how its borders and territories are flexibly and discursively mapped out in pursuit of some observed or inferred ambition—and with what consequences, and for whom?”