The Causal Conundrum: The Diet-Heart Debates and the Management of Uncertainty in American Medicine Free

Abstract

Starting in the 1950s, physicians and researchers began to debate the exact nature of the relationship among blood cholesterol, diet, and cardiovascular risk. Using professional medical, public health, and scientific journals, this article examines the history of a series of intense and sustained debates regarding the credibility of the diet-heart hypothesis, which proposed that diet was causally linked to coronary artery disease. Brought about by intellectual disagreements and illuminated by personal quarrels, these debates created a profound professional rift among researchers who debated whether observational data could be used to prove that dietary intake caused heart disease and who sought to differentiate between “good” and “bad” science. But while the debate persisted into the early 1980s, Americans had begun to adopt the diet-heart hypothesis as public health truth as early as the 1960s, embracing cookbooks promoting “heart healthy” diets that promised to prevent coronary artery disease. Although critics and advocates of diet-heart continued to debate the theory's finer points, the widespread adoption of diet-heart in American homes meant that the debate had become almost moot by the time the National Heart, Lung, and Blood Institute officially endorsed the hypothesis in the 1980s.

On June 16, 1948, President Harry Truman signed the National Heart Act, effectively creating the National Advisory Heart Council and the National Heart Institute (NHI) within the Public Health Service. Six weeks later, Surgeon General Leonard A. Scheele established the NHI as one of the newly named National Institutes of Health to assume responsibility for the administration of all research and training activities related to heart disease. In relatively short fashion, physician-scientists and epidemiologists received an immediate economic boost from the creation of the NHI, and Congressional appropriations to the NHI budget grew steadily during the institute's first decade.¹ Institute funds were dispersed to numerous investigators working at the intersection between laboratory and epidemiological research on heart disease. For physician-scientists and epidemiologists interested in the dietary origins of heart disease, this freeing of monies for research would have significant consequences.

By the 1950s, heart disease researchers could agree on several basic tenets. First, a person's blood level of cholesterol was the result of two sources: ingested fat, which the body then converts into cholesterol, and cholesterol endogenously produced by the liver. Second, atherosclerosis, or the process of plaque formation in arteries, was largely understood to cause coronary artery disease and myocardial infarction, or heart attacks.² Both of these concepts were the result of several decades of experimental and autopsy research on animals and humans.³ Finally, and perhaps most urgently, researchers agreed that the incidence and prevalence of heart disease in the United States was rising.⁴ The general category of heart disease had been the leading cause of death in the United States since 1910, but public apprehensions about the perceived coronary epidemic became increasingly palpable in the years following World War II, particularly after President Dwight Eisenhower's highly publicized 1955 heart attack.⁵ This came several years after arteriosclerotic heart disease had been adopted by the ICD as a separate cause of death and included in vital statistics compiled by the National Center for Health Statistics.⁶ When President Truman signed the National Heart Act into law, he effectively endorsed this perception of a public health crisis and opened the funding floodgates for researchers studying the diagnosis, treatment, and prevention of heart disease.

In large part, this new influx of attention to heart disease and funding for research in the area was directed at the study of atherosclerosis.⁷ Just a few years prior, reflected one prominent cardiovascular disease researcher, atherosclerosis had been “a step-child problem of medical research commanding very limited resources of personnel, equipment, plant, and money.”⁸ But starting in the 1950s, atherosclerosis gained credibility and support, as a multidisciplinary community of clinical investigators pursued new research programs dedicated to unraveling the diagnostic and therapeutic mysteries of coronary artery disease.

The primary relationship these researchers sought to understand was that between a person's dietary intake and their risk of developing coronary artery disease. The idea that there was a causal relationship between the two, which came to be known as the diet-heart hypothesis, dominated over thirty years of research and debate among heart disease researchers.⁹ For them, diet-heart struck at the core of three central issues in public health research. First, could the connection between what someone ate and their risk of heart disease be identified as a causal one even though studies merely showed that the two factors were correlated? Second, what constituted “good” and “bad” science? In the process of these debates, diet-heart became a professional and personal battleground as proponents and dissidents insulted each other's study designs and research techniques, many of which represented novel approaches to evaluating human health. Finally, researchers wondered how they could translate a population-based theory, still shrouded in a veil of uncertainty, into public health guidelines for treating individual patients.

This paper takes a thematic, rather than a chronological, approach. From the 1950s to the late 1970s and early 1980s, the basic components of the debate remained the same. In fact, it is the persistent and unchanging nature of the diet-heart debate that makes the controversy a particularly valuable case study for investigating the nature of proof-making and scientific argumentation in twentieth-century American medicine. In each arena of debate—determining causation from correlation, evaluating epidemiological studies of disease, and changing how individual patients were treated based on population science—the relationship between epidemiology and medicine was called into question.¹⁰ Chronic disease epidemiology offered enticing possibilities for medical research, but simultaneously presented researchers with new doubts. The story of the diet-heart hypothesis can thus help us understand how researchers working on the relationship between diet and heart disease navigated their way through the murky waters of multidisciplinary medical research at midcentury and managed the uncertainty that came along with it.

CONSTRUCTING DIET-HEART

Ancel Keys, a professor of physiology at the University of Minnesota, was one of the first researchers to argue that there existed a relationship between an individual's blood cholesterol levels and his risk of future coronary artery disease.¹¹ Keys' supposition represented the outgrowth of his earlier work on what he called “physiological hygiene.”¹² This research involved correlating an individual's behavior with his or her physiology as measured by laboratory values. Ultimately, Keys hoped to be able to advise individuals on ways to change their behavior to promote healthier living and reduce their risk of disease.¹³ This vision had guided Keys' influential starvation studies during World War II, which had demonstrated the pliability of an individual's physiological state.¹⁴ While subjecting conscientious objectors to a variety of limited diets, Keys observed the effects of these diets on participants' mental and physical states in order to advise the military on optimal feeding strategies for soldiers and victims of wartime famine in Europe.¹⁵

In 1947, Keys launched his next major study, the Minnesota Business and Professional Men's Study, which would prove essential to the development of the diet-heart hypothesis. In it, Keys and his fellow researchers followed five hundred healthy men for fifteen years and sought to understand why some men developed coronary artery disease while others did not. They observed that men with lower cholesterol levels were less likely to develop heart disease.¹⁶ In addition, they were able to determine that the risk of developing coronary artery disease steadily increased as the subjects' blood cholesterol level rose.¹⁷

In 1953, Keys formally introduced what would become the diet-heart hypothesis in the Journal of the Mount Sinai Hospital. Based upon an analysis of laboratory, mortality, and population-based data that he had collected from multiple outside sources, Keys constructed the diet-heart hypothesis based on a set of interrelated statistical associations. First, he cited a high mortality rate among adults in the United States from degenerative heart disease. Second, he reported clinical and laboratory data that indicated that total blood cholesterol levels were associated with dietary fat intake. Third, he presented data from the World Health Organization in multiple countries showing that total blood cholesterol measurements could identify men who were more likely to develop coronary artery disease. Lastly, Keys suggested that Americans' increasingly fatty diets were linked to their increased mortality from degenerative heart disease.

In short, Keys was arguing that more Americans were dying of heart disease, and that this was due to high cholesterol levels caused by fatty diets.¹⁸ Like he had with the Minnesota Business and Professional Men's Study, Keys was arguing for the diet-heart hypothesis based on correlative statistical data. In this case, however, his argument was even more tenuous because it was based on multiple sets of unrelated data he had obtained from outside sources. It was this premise with which critics would take issue in the coming debates about diet-heart.

Four years later, Keys would launch the largest and most influential study of his career. It would also become the most divisive. Devised and launched in 1957 by Keys and an international team of specialists, the Seven Countries Study studied the diet, lifestyle, and incidence of coronary artery disease among 12,763 randomly selected middle-aged men from Finland, Greece, Italy, Japan, the Netherlands, the United States, and Yugoslavia. One of the most distinctive aspects of the Seven Countries Study was the effort to characterize each region's diet. Researchers observed what subjects ate and collected duplicates of their meals, which were analyzed for fat, carbohydrate, and protein content. They then used statistics to analyze the relationships between fat intake, blood cholesterol levels, and heart disease. The study seemed to confirm the diet-heart hypothesis; regional averages of saturated fat intake and blood cholesterol levels were both correlated with a population's rates of coronary artery disease.¹⁹

In essence, the diet-heart hypothesis was the apex of Keys' physiological hygiene. It represented a risk that might be modified through behavioral change, based on observational laboratory data collected from diverse populations. The diet-heart hypothesis also represented an intellectual stepping-stone toward a deeper, more nuanced conception of risk and disease prevention. First employed as a term by Framingham Study researchers Thomas Dawber and William Kannel, the risk factor offered researchers and physicians a new way to conceptualize and manage chronic disease.²⁰ In the case of coronary artery disease, physicians could theoretically translate the risk factor of high blood cholesterol into a discrete preventive strategy: maintaining a low-cholesterol diet.²¹

THE CORRELATION–CAUSATION QUANDARY

By 1955, researchers agreed that elevated blood cholesterol levels were associated with an increased risk of future coronary artery disease.²² But determining whether and how dietary intake caused these elevated blood cholesterol levels was still fraught with debate. In particular, researchers disagreed about whether statistical relationships, even very strong ones, could be used to prove causation.

This debate mirrored the one simultaneously occurring among researchers studying the possible carcinogenic effects of tobacco.²³ Iconic case–control studies conducted by Austin Bradford Hill and Richard Doll and by Ernst Wynder and Evarts Graham in the 1940s had suggested a causal link between smoking and lung cancer using retrospective observational evidence.²⁴ Because retrospective analyses failed to limit experimental biases (and because it was implausible to conduct truly randomized studies of smokers versus nonsmokers), Doll and Hill devised a second form of cohort study that would assess risk prospectively. By following a group of smokers and nonsmokers over an extended period of time, Doll and Hill calculated the risk of developing lung cancer in real time.²⁵ Finally, the debates underlying the 1964 Surgeon General's Report on Smoking and Health revealed the profound discomfort committee members felt with equating epidemiological studies linking smoking and lung cancer to a statement alleging a causal relationship between the two.²⁶

As in the case of cigarettes, some heart disease researchers were eager to draw a causal relationship between a fatty diet and heart disease based on strong statistical correlations, while others considered it fundamentally impossible. As we have already seen, Ancel Keys was one such researcher who placed faith in statistical evidence, and his 1953 ecologic association among the six countries would serve as the foundation for much diet-heart research that followed. As Henry Blackburn and Darwin Labarthe have recently noted, Keys' ecologic association, though a relatively minor component of his overall argument, garnered the most attention from critics who raised questions about causal inference.²⁷ Challenges to Keys' ecologic association ushered in the decades-long debate and also sparked calls for the development of guidelines for causal inference.

Even as diet-heart researchers expanded their understanding of dietary fats, support of the hypothesis was already a foregone conclusion for unwavering advocates like cardiologist William Dock. For example, Dock presented observational evidence from studies of Yemenites who immigrated to Israel. According to these studies, the Yemenites had long subsisted on a grain- and vegetable-based diet but had adopted a fattier one upon immigrating to Israel. These dietary changes, he explained, correlated with higher cholesterol levels and higher rates of coronary artery disease for those who had lived in Israel for many several decades compared with those who had more recently immigrated. For him, this was proof that a fatty diet caused heart disease. He also referred to the diet-heart hypothesis as a foregone conclusion, remarking that it had already found “universal acclaim” among scientific professionals and laymen.²⁸

But several of these professionals begged to differ. One letter to the editor by physician Ernest Geiger insisted that despite the “enormous literature” on diet-heart to which Dock had referred, “no evidence has yet been provided which proves that human atherosclerosis is produced by the consumption of fat or, particularly, butter fat.”²⁹ Rather, Geiger, as a self-proclaimed member of the “more skeptical ‘segment,’” noted that he and like-minded colleagues preferred to wait for more evidence to shine additional light on the subject. In closing, Geiger reached back in history to support his view. As Geiger asked, “did Dr. Dock consider what Clio's judgment may be of physicians who, not so long ago, accepted as truth the theory based on statistical coincidence, that malaria was produced by ‘mala aria,’ i.e., by ‘bad air’ or that excessive sugar consumption is responsible for diabetes?”³⁰ In essence, Geiger was cautioning Dock and other diet-heart supporters against making premature assumptions about a causal relationship between diet and coronary risk.

Others reached even further back in history to make the case for diet-heart. In a series of several 1969 pieces in Lancet, internist Haqvin Malmros and surgeon Thomas Cleave engaged in what we might consider “retrospective epidemiology.” Malmros' review article on dietary prevention of atherosclerosis drew upon evidence from laboratory-based and population-based studies to point out the correlation between diet and heart disease and insist that it was not necessary to wait for “final proof” to endorse diet-heart recommendations.³¹ In a letter to the editor, Cleave argued that diets high in fat could not account for the recent spike in heart disease, for high-fat diets had been enjoyed since biblical times. Employing the Bible for evidentiary support, Cleave remarked, “Indeed we need only refer to Deuteronomy, chapter 32, verse 14, to note that even in that distant time Jehovah gave to his people to eat ‘butter of kine, and milk of sheep with fat of lambs.’”³² Something else, he maintained, must account for the increasing incidence and prevalence of cardiovascular disease. In later years, he would suggest sugar and refined carbohydrate consumption were to blame. In response, Malmros encouraged Cleave to consider the very next verse in Deuteronomy: “but Jeshurun (the folk of Israel) waxed fat and kicked: thou art waxen fat, though art grown thick, though art covered with fatness.”³³ In other words, Malmros was alleging that the people of Israel were obese principally because they had consumed a fatty diet. The logical conclusion, although he did not remark on this specifically, was that Israelites had suffered from coronary artery disease just as Americans did in the twentieth century.

Like Cleave, other physicians worried that declaring the diet-heart relationship to be causal might overlook other important mediators, specifically the mechanism by which plaques accumulated on the interior walls of arteries. One such critic, pathologist John B. Duguid, was concerned that the diet-heart hypothesis oversimplified this poorly understood process. As Duguid put it in a 1954 Lancet article, atherosclerosis was “not a specific disease but merely the common end-result of a number of different pathological processes with different causes.”³⁴ He offered an alternative causative model, maintaining that plaques resulted not from excessive amounts of cholesterol and dietary fat in the blood, but rather from defective repair of the endothelium, the layer of cells lining the inside of coronary arteries.³⁵ To Duguid, this alternative model made the diet-heart hypothesis shortsighted because it assumed that dietary fat simply accumulated on the walls of arteries until heart attacks resulted.³⁶ In other words, Duguid was suggesting that researchers had too quickly concluded that high cholesterol caused coronary artery disease, when in fact the relationship between the two might be only a statistical association obscuring a different causal mechanism.

Duguid's argument was a salient one. In fact, detailing the physical and structural changes that prompted plaque accumulation was not of immediate concern to diet-heart researchers. Keys readily admitted as much in a Letter to the Editor, noting that “the diet theory does not seek to explain much about the mechanisms in atherogenesis.”³⁷ This disinterest in mechanism proved to be both a boon and a burden to diet-heart researchers. In another 1954 article defending the diet-heart hypothesis, Keys noted how prior epidemiological and clinical successes had often come before researchers understood underlying disease mechanisms.³⁸ He acknowledged that a detailed understanding of underlying physiological mechanisms would come in time, but that in the meantime, the statistical association could steer disease prevention efforts. Indeed, several years later, Keys would argue that the diet-heart relationship was based upon physiological processes.³⁹

While individual diet-heart advocates and critics debated the merits of statistical analysis, professional medical organizations also weighed in on the subject. In 1957, the American Heart Association (AHA) published a report by five physicians assessing the body of evidence on diet-heart. In it, they explained that none of the laboratory and epidemiological data collected over the course of the preceding decade had definitively demonstrated an association among dietary fat, cholesterol, and heart disease.⁴⁰ The authors' hesitance to endorse diet-heart was reflected in the report's reserved language of equivocation: that “diet may play an important role in the pathogenesis of atherosclerosis,” that “the fat content and total calories in the diet are probably important factors,” and lastly, that “a wide variety of other factors, dietary and non-dietary, may be of equal or greater importance” [italics mine].⁴¹ As the authors concluded, “perhaps the best that can be said is that there is an association that has statistical value.”⁴² Unwilling to draw a causative relationship, but willing to acknowledge the growing body of evidence in favor of diet-heart, the report reflected the Association's hesitance to draw any definitive conclusions.

Over the course of the next two decades, there remained a palpable disconnect between those who argued that evidence in favor of diet-heart was very strong, while others insisted that the debate was still at a standstill. For proponents, the sheer amount of statistical evidence was enough to prove a causal relationship. As University of Michigan physician Frederick Epstein argued in 1957, “though indirect, [the evidence supporting diet-heart] is so strong that the question whether lowering serum cholesterol reduces the risk of coronary heart disease hinges no longer on the results of preventive trials alone.”⁴³ Rather, Epstein implied, the weight of this statistical evidence was sufficient to prove that diets high in fat caused heart disease, even without additional evidence from clinical trials. Nevertheless, he lamented, there would always be critics who questioned the value of statistical data. As Epstein noted, “even if ‘proof’ in a statistical sense is obtained, those who call for ‘hard data’ may not accept the quality of the evidence upon which this demonstration was based.”⁴⁴ For these critics, he explained, statistics would never count as sufficient proof of causation.

By the 1980s, discord remained. Advocates continued to promote numerous population-based studies that demonstrated a strong association between dietary intake and coronary artery disease, while critics maintained that the debate about statistical proof had not moved forward at all. A 1981 paper written by pathologist Henry McGill and two colleagues focused on “the problem of correlations,” as they put it.⁴⁵ As they explained, “skeptics contend that inconsistencies, gaps, and paradoxes in the evidence do not permit one to infer a causative relationship between these dietary components and atherosclerosis.”⁴⁶ The gaps in knowledge about a causative relationship between diet and heart disease that McGill and his colleagues identified were no different from concerns raised in the preceding decades—but McGill and his colleagues saw opportunity, not standstill, in the still-unresolved debate. In their estimation, as researchers continued to assess new evidence on the matter, they would learn how to improve their ability to interpret epidemiological evidence and formulate dietary guidelines based on that very body of evidence.⁴⁷

LABORATORIES, POPULATIONS, AND THE CONTESTED VALUE OF EPIDEMIOLOGY

As we have seen, physicians and researchers engaged in decades-long debates about the theoretical implications of the diet-heart hypothesis, particularly about how to assess causative relationships. Another issue, however, brought about heated, and sometimes personal, exchanges: what distinguished “good” from “bad” science? Which studies were meaningful, and which could not be trusted? Ancel Keys, perhaps the most prominent diet-heart advocate, and physician-researcher George Mann, perhaps one of the most vocal diet-heart critics, had several illuminating exchanges over the course of their respective careers. Supporters and detractors of both men were also quick to chime in on these debates. Although Keys and Mann were far from the only ones to face criticism from or attempt to undercut their intellectual challengers, an analysis of representative exchanges among Keys, Mann, and other cardiac researchers sheds light on the tenor and scope of diet-heart debates.

One of the earliest debates between Keys and Mann took place in the pages of The American Journal of Clinical Nutrition. In May 1955, Mann, then an assistant professor of nutrition at the Harvard School of Public Health, published an article that suggested weight reduction and caloric limitation lowered blood cholesterol levels. Mann placed two healthy young male subjects on experimental diets that consisted solely of pemmican, a mixture of dried meat and berries rendered in lard, for a period of fifteen days. His data demonstrated a marked increase in beta lipoprotein and total cholesterol levels as the subjects transitioned from their conventional diets to the fatty pemmican diet. However, Mann concluded that these increases were not statistically significant enough to warrant a relationship between dietary fat and blood cholesterol.⁴⁸ In other words, Mann's findings did not support the diet-heart hypothesis.

Keys immediately took Mann to task on methodological grounds. Even though he questioned the reliability of Mann's small sample size, Keys was quick to demonstrate that Mann's cholesterol data did in fact support the diet-heart hypothesis. Keys argued that it was Mann's improper use of statistical analysis that produced errant interpretations.⁴⁹ He denounced Mann's report as “an interesting case study in statistical fallacy applied to inadequate evidence.”⁵⁰ In the study, Mann had recorded baseline lipoprotein and total cholesterol levels for both subjects at the beginning of the study, then collected blood samples at several points during the fifteen-day experimental period. Twenty days after both subjects resumed their regular diets, Mann collected a final blood sample, which served as the final “control” measurement. However, he averaged the two men's blood lipoprotein and cholesterol levels to arrive at his conclusions, rather than comparing each man's levels on the diet to his levels before and after consuming only Pemmican. In his rebuttal, Mann admitted that it would have been more appropriate had he compared the differences between the “control” and “experimental” observations for each man rather than calculating mean blood levels in order to account for intraindividual variability as Keys had argued in his letter. Nevertheless, Mann continued to doubt whether such increases in blood cholesterol and lipoprotein levels were meaningfully related to dietary intake.⁵¹ This particular exchange between Keys and Mann is notable in its timing, as randomized clinical trials were becoming the norm and compared intervention groups with control groups. Keys' criticism was timely, and even though Mann did not have a separate control group, the idea of comparing an intervention to a baseline nonintervention was becoming standard within the biomedical research community.⁵²

Keys himself was not immune to sharp criticism from his peers. His 1953 ecological correlation of diet and heart disease risk in six countries, which would serve as the conceptual underpinning for his Seven Countries Study, would be the subject of methodological debate for decades. Critics challenged Keys in his selection of these six countries, maintaining that the study was inherently flawed due to sampling bias because Keys had selected to study countries whose dietary habits would prove his hypothesis. Notably, Herman Hilleboe and Berkeley statistician Jacob Yerushalmy challenged Keys' statistical calculations. Unimpressed with Keys' conclusion that countries with higher amounts of fat consumption experienced higher rates of coronary artery disease, Yerushalmy and Hilleboe re-examined statistics for all twenty-two nations for which data was publicly available from the World Health Organization and United Nations. Their larger data set demonstrated that the supposed association between dietary fat and coronary artery disease was insignificant, if not nonexistent. They concluded that “clearly this tenuous association cannot serve as much support for the hypothesis which implicates fat as an etiologic factor in arteriosclerotic and degenerative heart disease.”⁵³ Despite their painstaking reanalysis of Keys' data, Yerushalmy and Hilleboe's main concern was his methodology. Blind reliance upon statistics, they suggested, could lead to misguided efforts to control not only dietary fat and cholesterol, but any other physiological variable that could be statistically linked to diet. As they argued, “quotation and repetition of the suggestive association soon creates the impression that the relationship is truly valid, and ultimately it acquires status as a supporting link in a chain of presumed proof.”⁵⁴

Keys challenged Yerushalmy and Hillboe's assertion of sampling bias by arguing that they had engaged in their own form of reverse selection by including all forms of heart disease in their analysis, not just coronary artery disease. In so doing, they examined forms of heart disease such as rheumatic heart disease that Keys felt did not have a dietary component or atherogenic etiology.⁵⁵ Several decades later, Keys defended his selection of the nations, offering brief explanations of how each area was selected in order to “counter unfounded suggestions that the cohorts were chosen on the basis of prior knowledge that promised to support a hypothesis of a causal link between diet and incidence of CHD.”⁵⁶ As Keys explained, he selected the countries in question for methodological reasons—he had close professional relationships with researchers in the countries he selected, and those trusted confidants provided a source of reliable data and local knowledge. Nevertheless, the allegation that Keys had cherry-picked international epidemiological evidence to prove his point would haunt him for several decades, even as detractors performed similar studies. In fact, while Keys' analysis of these six countries and his Minnesota Business and Professional Men Study were some of the first of their kind, regional epidemiological studies were becoming extremely popular.⁵⁷ American and international investigators alike initiated no fewer than 38 similar longitudinal cohort studies on various industrial and rural populations, both in the United States and internationally, between the late 1940s and 1970.⁵⁸

Meanwhile, some critics argued that epidemiology was fundamentally incapable of proving the diet-heart hypothesis. George Mann insisted that “epidemiology is apt not to be a decisive method for resolution of these problems…  .  At best, the method may supply a few clues and enough encouragement to gifted people who  …  will find the answer in the laboratory.”⁵⁹ And as British cardiologist John McMichael opined more than two decades later, “epidemiologists seem to rely on less secure information like death certificates and speculative extraneous influences which would be unacceptable in laboratory science.”⁶⁰ To these critics, epidemiology could not supply the proof necessary to demonstrate a causal link between dietary intake and heart disease.

Mann and McMichael represented an extreme point of view. Others agreed that epidemiology, while fraught with uncertainty, offered potential for understanding coronary artery disease. Herman Hilleboe, despite his concerns with diet-heart, remained optimistic that epidemiological evidence could shed new light on the subject. “Some shrewd detective work in the form of epidemiologic studies should help us to identify these precursors and precipitators of coronary artery disease,” he commented in 1957.⁶¹ Berkeley medical physicist John Gofman, who would become one of diet-heart's strongest proponents, echoed Hilleboe's cautious sense of optimism. While conceding that epidemiological evidence remained inconclusive, Gofman nonetheless urged investigators not to discard the hypothesis. “That it produces valuable clues for  …  investigation is not denied even by those who are most vehemently opposed to acceptance of epidemiologic evidence,” he maintained in his book on the topic two years later.⁶²

While many of these epidemiological studies indicated the predictive value of blood cholesterol levels for an individual's risk of heart disease and others even suggested a correlation between diet and heart disease, Keys noted in 1975 that “many attempts have been made to find populations that might refute the [diet-heart] theory.”⁶³ Despite his comments deriding epidemiological research, George Mann himself conducted field studies in Africa intended to disprove diet-heart. By studying indigenous populations, including the Masai of Tanzania and Kenya, he and his colleagues recorded that Masai nomads living in Kenya who subsisted on high-calorie diets rich in saturated fat had very low blood-cholesterol levels.⁶⁴ They concluded that these findings disproved the diet-heart hypothesis.

Keys, however, dismissed Mann's Masai work in several ways. He pointed to studies of other Kenyan nomadic peoples whose dietary behaviors and blood cholesterol levels fit with the diet-heart hypothesis.⁶⁵ Keys then proposed several explanations for the Masai's apparent peculiarity. He maintained that the Masai were “chronically unfed and frequently half-starved,” which might account for their low rates of heart disease, an observation that reflected insights derived from his own work on semistarvation. Furthermore, Mann had suggested that the Masai's inclusion of sour milk in their diet might prevent their blood cholesterol levels from rising.⁶⁶ This theory, Keys pointed out, supported the notion that dietary intake was somehow related to the pathogenesis of heart disease.⁶⁷ Most interesting, perhaps, was Keys' suggestion that the Masai had developed a “very effective feedback mechanism to suppress endogenous cholesterol synthesis.”⁶⁸ This evolutionary explanation harkened back to George Dock's 1957 generational assessment of Yemenites, who had experienced lower rates of heart disease while eating their traditional diet prior to moving to Israel. Keys himself observed similar changes in his examination of Japanese men and their diets, blood cholesterol levels, and coronary artery disease rates in Japan, Hawaii, and Los Angeles. Keys and his colleagues found that Japanese farmers living in Koga, whose diet consisted of <10 percent fat calories, had particularly low blood cholesterol levels, while Nisei living in Los Angeles had blood cholesterol levels comparable to other American cohorts.⁶⁹

Those who raised challenges to the hypothesis found themselves on the outside looking in. Mann's status as a diet-heart critic was solidified in a People Magazine profile entitled “Dr. George Mann Says Low Cholesterol Diets are Useless, but the ‘Heart Mafia’ Disagrees.”⁷⁰ Mann would be labeled by some of his colleagues as a “heretic.” In a 1977 editorial in the Lancet, nutritionist John Rivers explained “that [Mann's] opposition should justify the term ‘heretic’ is not a reflection of the unchallengeable nature of the evidence supporting the ‘lipid hypothesis,’ but the extent to which official gatherings of experts in different countries have made it national policy.”⁷¹ Mann defended his “heretical” views by insisting that he was only being a good scientist. In his view, a good scientist was one who put his hypothesis to the test and relied on the scientific method to confirm causal claims. Diet-heart advocates, he insisted, had lost sight of or simply ignored this scientific ideal. In his estimation, any heresy was on the part of “diet/heart enthusiasts who, by their noisy propaganda hope to coddle and prolong the diet/heart hypothesis.”⁷² Just as Mann insisted that diet-heart advocates had lost sight of what it meant to conduct good science, John McMichael similarly regarded diet-heart endorsements as “quite an unwarranted extension of hope over experience.”⁷³

In 1977, Mann declared an end to the “diet-heart” era in an obituary of sorts in the New England Journal of Medicine. Contending that evidence of a diet-heart connection was inconclusive at best and wholly incorrect at worst, Mann marshaled evidence from other epidemiological studies that suggested that dietary habits had little or no influence on blood lipid concentrations.⁷⁴ But the debate, at least in the pages of medical journals, was not over. Two years later, three of Keys' colleagues at the University of Minnesota's Laboratory of Physiological Hygiene issued a rigorous defense of diet-heart. They sought to clarify the most stringent methodological critique advanced by the diet-heart opponents: that because cross-sectional analyses had shown no correlation between diet and blood cholesterol levels, there could be no meaningful relationship between the two.⁷⁵ Supporters of the diet-heart hypothesis had long relied upon controlled dietary experiments that demonstrated a clear correlation between diet and blood cholesterol levels in subjects over long periods of time.⁷⁶ The hypothesis as they understood it was built upon a series of indirect, yet interrelated, relationships: if dietary modification lowered the concentration of cholesterol in the blood, and blood cholesterol levels were directly related to the development of atherosclerosis, then proper dietary modification would presumably reduce one's risk for coronary artery disease. As the Minnesota researchers stressed, “the link of diet to coronary heart disease is presumably not direct but through its effect on serum cholesterol.”⁷⁷ They distinguished between the three primary modes of experimental methods for examining the relationship among diet, blood cholesterol concentration, and coronary artery disease: controlled feeding experiments, dietary change experiments (like Mann's Pemmican study), and longitudinal studies, such as Framingham and Seven Countries. As the Minnesota researchers explained, controlled feeding experiments and longitudinal studies showed diet-heart best, while cross-sectional studies did it worst. This reality, they argued, did not diminish the strength of evidence in favor of the hypothesis.

After he retired from medicine, Mann would proclaim diet-heart to be “the greatest health scam of the century.”⁷⁸ Though Mann's language may have been bombastic, it represented his long-held belief that diet-heart advocates were misguided by their blind faith in an unproven hypothesis and their abandonment of proper scientific methodology. To a certain degree, the debate came down to the question of epidemiology's role in clinical medicine. For diet-heart advocates, epidemiology represented a powerful tool that could provide answers to otherwise unanswerable questions. For diet-heart critics, epidemiology represented an imperfect science that could never truly answer questions about the workings of the human body.

FROM RESEARCH TO REALITY

While methodological debates raged on, proponents of the diet-heart hypothesis worked to translate their knowledge into concrete public health recommendations. This proved difficult, as official organizations like the AHA continued to grapple with the cloud of uncertainty surrounding diet-heart. How could public health officials and leaders use population-based data, the veracity and clinical applicability of which was in question, to guide recommendations for the American dinner table? Despite these hurdles, the proponents found that in many cases they were able to ignore much of the debate and make diet-heart a reality for American physicians and their patients. In large part, this was due to the popularity of heart-healthy cookbooks that allowed Americans to take diet-heart into their own hands. In many instances, physicians struggled to keep up.

As we have seen, in its 1957 report the AHA had been hesitant to characterize the relationships among dietary fat, blood cholesterol, and heart disease as definite, let alone causal. As the authors concluded, “the evidence at present does not convey any specific implications for drastic dietary changes, specifically in the quantity or type of fat in the diet of the general population, on the premise that such changes will definitely lessen the incidence of coronary or cerebral artery disease.”⁷⁹ In other words, the authors were unwilling to advocate in favor of any dietary recommendations based on the diet-heart hypothesis.

Just four years later, the organization had changed its tune. In a follow-up report authored by several fervent supporters of diet-heart, including Ancel Keys himself, the AHA recommended that individuals with a high risk of heart attack or stroke substitute polyunsaturated fats for saturated fats in their diets.⁸⁰ By the late 1950s, saturated fats had been pinpointed as particularly dangerous by diet-heart advocates and this represented an important endorsement of diet-heart.⁸¹ Even more, the report included an early call for food manufacturers to declare the levels of dietary fat content in their products.⁸² Despite the forceful tone of the report, the committee exercised restraint by insisting that these recommendations, and indeed all dietary changes, be carried out only under medical supervision.⁸³ While George Mann would later argue that the report's conclusions were foregone, the committee responsible included early skeptics of the hypothesis such as Irvine Page and Frederick Stare.⁸⁴ Indeed, the 1961 AHA report suggested that although scientific consensus about diet-heart did not yet exist, clinical practice would soon adopt it as truth.

The AHA's 1961 report came just as Americans were already beginning to put diet-heart into practice within their own homes. Two years prior, Ancel Keys and his wife Margaret Keys published Eat Well and Stay Well, a cookbook that introduced the Mediterranean diet—notable for its low levels of saturated fat—to a large swath of the American public.⁸⁵ Keys and Keys used straightforward language to explain the diet-heart hypothesis to their readers. Summarizing decades of research on atherosclerosis in animals and humans, they concluded, “the fats we eat have a major effect on the amount of cholesterol in the blood, and it is this cholesterol which is deposited in the walls of the arteries.”⁸⁶ And yet they did not shy away from the debates around diet-heart. “Suspicion is strong that luxurious high fat diets may promote coronary heart disease,” they explained, noting that “many medical scientists would say that suspicion is too mild a term in view of the evidence.”⁸⁷ Still, they argued that the evidence for diet-heart was “impressive” and invited those who were persuaded to follow their dietary recommendations.⁸⁸

Eat Well and Stay Well debuted at number nine on the New York Times bestseller list, sold over 75,000 copies in four months, and exhausted four printings during that span of time.⁸⁹ In celebration of this success, Time magazine made Ancel Keys, whom they deemed “Mr. Cholesterol,” the cover subject for its January 13, 1961, issue.⁹⁰ The cookbook spurred the publication of numerous “heart health” advice books that, more often than not, endorsed a low-cholesterol, low-fat diet.⁹¹ One of these, Dietary Prevention and Treatment of Heart Disease, had been authored by another diet-heart researcher, John Gofman.⁹² Despite uncertainty in the medical literature, diet-heart proponents like Keys and Gofman had been able to bypass the medical establishment by taking their recommendations to the public directly.

For some diet-heart supporters, these changes were not enough. In a 1969 Lancet article, Swedish physician Haqvin Malmros argued that medical professionals should be more involved in advocating for heart-healthy eating, rather than becoming mere bystanders to the cookbook-purchasing public. Specifically, Malmros called for medical authorities and food industry officials to educate laypeople to replace saturated fat with unsaturated fat in their diets. Malmros also raised a critical question that advocates and critics alike would consider for decades to come: whether we must wait for “final proof” before instituting dietary recommendations and other anti-coronary measures.⁹³ Despite the lack of such proof, he concluded, “most experts now believe the evidence to be so strong as to justify a modification of the diet.”⁹⁴

Subsequent versions of the AHA statement on diet and heart disease reflected this transition. While the 1961 recommendations focused on individuals who had a higher risk of heart attack or stroke than the general population, the 1965 recommendations were expanded to all individuals.⁹⁵ By 1968, the AHA made specific recommendations as to how individuals should alter their diets. For example, a diet comprising less than 40% of any kind of fat was highly advised, as was one containing twice as many polyunsaturated fats as saturated fats.⁹⁶ Within professional circles, diet-heart remained a hypothesis in the late 1960s. When it came to public health recommendations, however, it was much closer to a truth.

Despite this trend, detractors like George Mann continued to declare war on diet-heart. Mann's provocative 1977 diet-heart “obituary” produced a series of letters to the editor that were published in two separate issues of the New England Journal of Medicine. One response was a full-length rebuttal by members of the AHA's Nutrition Committee, an advisory group charged with the development of dietary guidelines. In an introductory note, New England Journal editor Arnold Relman emphasized that rebuttal articles were atypical for the journal, but explained that the significance of the debate surrounding the diet-heart hypothesis warranted such an article.⁹⁷

In the rebuttal, the Nutrition Committee acknowledged the existence of uncertainty, but argued that the only path forward was for individual physicians to use their clinical judgment in patient care. “Within a framework of suggestive, but not unequivocal, scientific proof that dietary modification will ameliorate or prevent coronary artery disease,” the authors offered, “physicians must share the burden of uncertainty.”⁹⁸ Because the efficacy of programs of dietary prevention was unclear, the Nutrition Committee suggested that the biomedical community focus on assuring their feasibility and safety, and cautiously endorsed prevention and dietary modification in the meantime. The Committee then concluded the report in a noncommittal fashion, suggesting that “physicians often must make decisions in the absence of absolutely conclusive scientific proof. A reasoned resolution of the controversy is not currently possible.”⁹⁹ In other words, individual clinical judgment would have to supersede scientific “truth”—at least for now.

A reasoned, yet controversial, resolution of the controversy came more than half a decade later in 1984, when the National Institutes of Health sponsored a consensus conference on the subject. Beginning in the late 1970s, the NIH hosted a series of consensus conferences on assorted medical controversies. The NIH consensus development program presented the perfect opportunity for advocates and critics to address the long-standing debate regarding the diet-heart hypothesis. The December 1984 consensus conference followed the conclusion of the National Heart, Lung, and Blood Institute's (NHLBI) Coronary Primary Prevention Trial.¹⁰⁰ Though ostensibly a drug trial, the study demonstrated that the greater the degree of cholesterol reduction, the greater the reduction in cardiovascular risk. The NHLBI and the Office of Medical Applications of Research, which oversaw the NIH consensus program, convened the conference to address the unresolved question of lowering blood cholesterol to prevent heart disease.

After two days of deliberation, the committee agreed that multiple lines of evidence demonstrated a strong relationship between lowering cholesterol levels and reducing coronary artery disease. It recommended that individuals with moderate and high-risk cholesterol levels be treated by dietary means and with drug treatment if dietary modification proved insufficient, and that all Americans above the age of two reduce total caloric and saturated fat intake. Finally, the committee recommended the development of concerted public education campaigns to promote the treatment of elevated cholesterol levels and the expansion of food labeling practices to identify cholesterol, saturated, and unsaturated fat content.¹⁰¹

Some critics argued that the final consensus statement erased any signs of dissent.¹⁰² Edward Ahrens, a professor and lipid researcher at Rockefeller University, and Michael Oliver, a British cardiologist, spoke at the conference but later questioned the consensus process. Oliver argued that the conference was “contrived” and that the conference was organized to ensure that a final statement would be favorable toward diet-heart.¹⁰³ Ahrens focused instead on how the conference panel relied on epidemiological studies as well as dietary and drug trials to fashion its cholesterol-lowering guidelines for the general population. For Ahrens, such evidence was suggestive but not definitive: for him, “correlations, no matter how strong, are never proof.”¹⁰⁴ Although multiple lines of evidence suggested that high-risk individuals benefitted from a cholesterol-lowering diet, Ahrens emphasized that it was unclear if low or moderate-risk individuals would benefit from a cholesterol-lowering diet in the same way. Yet the final consensus statement made that very extrapolation. As Ahrens explained, “it promises benefits without giving the evidence to back up that promise. By failing to emphasise what we do not know, the statement sweeps these weaknesses in our evidence under the-rug, as if they were trivial.”¹⁰⁵

Essentially ignoring these criticisms, the recommendations of the committee became public policy with the November 1985 launch of the National Cholesterol Education Program (NCEP). Modeled after the National High Blood Pressure Education Program, which the NHLBI had established in 1972, the NCEP was to be a cooperative effort among community groups, volunteer organizations, and health agencies. NHLBI leaders envisioned that the NCEP would raise awareness among the public about high blood cholesterol as a cardiovascular risk factor and tout the merits of lowering cholesterol to prevent heart disease. The first set of treatment guidelines published by the NCEP in 1986 was even subtitled “From Controversy to Consensus.”¹⁰⁶ As NCEP coordinator James Cleeman and NHLBI director Claude Lenfant declared, “we have for the first time a detailed set of guidelines giving practical advice for dealing with an individual adult patient's cholesterol problem.”¹⁰⁷ In fact, Cleeman and Lenfant pointed to the consensus conference as a defining moment in the development of the NCEP's Adult Treatment Panel guidelines.¹⁰⁸

ASSESSING UNCERTAINTY

The history of the diet-heart debates stands as an enlightening example of rising tensions regarding epidemiology's relationship to postwar biomedicine, bringing to the fore several vexing questions. What counted as scientific proof? How were researchers and physicians supposed to interpret the substantial body of clinical investigations that produced nondefinitive and sometimes contradictory evidence? Furthermore, how were physicians supposed to apply that data to clinical practice? As one cardiologist lamented in 1990, “what is so puzzling is why we have to work so hard to sell the message given what seems to be an unbeatable amount of hard evidence.”¹⁰⁹

Although cholesterol skeptics have been relegated to fringe status since the 1990s, disagreement about what constitutes “heart healthiness” persists.¹¹⁰ In fact, the debate about the diet-heart hypothesis has gradually shifted toward statin therapy and its associated risks.¹¹¹ Indeed, the addition of cholesterol-lowering drugs to the therapeutic landscape—with Pfizer's Lipitor becoming the best-selling prescription drug in history—has altered the dynamics of what it means to treat high blood cholesterol. For many health professionals and especially for the broader American public, the ability of such drugs to reduce cardiovascular risk is taken as “proof” that cholesterol management is a key to cardiovascular health. The commercial success of these blockbuster drugs further underscores the continuing influential role that cholesterol plays in American life. And yet statin therapy has created a more divided therapeutic landscape than ever, given its ability to lower cholesterol levels independent of dietary modifications and sometimes, seemingly in spite of the lack of such changes.

The November 2013 release of new cholesterol treatment guidelines developed by the American College of Cardiology and the American Heart Association promises to dramatically change this decades-long conversation. Based on a review of randomized, controlled clinical trials, the ACC/AHA guidelines outlined a new calculus by which physicians should prescribe statins. According to the guidelines, not only should the decision to prescribe statins not be based on cholesterol levels alone, but there are no longer target LDL levels nor are physicians encouraged to monitor their patients' cholesterol levels once they start taking cholesterol-lowering medications. Instead, the medical decision-making process is mediated through the use of new risk calculators that recommend treatment if a patient's ten-year risk of a heart attack or stroke is greater than 7.5 percent. According to the committee that created them, the guidelines prioritize the use of statins while still emphasizing the foundational role of lifestyle and dietary modifications.¹¹²

Yet the ACC/AHA guidelines were met by almost immediate condemnation and confusion. Condemnation came from some who contended that the online risk calculator overestimated cardiovascular risk.¹¹³ Others questioned the way in which final guidelines were released without opportunity for public review and comment, as has become the norm for many consensus development programs.¹¹⁴ Confusion stemmed from a pronounced shift away from a decades-long standing emphasis of cholesterol reduction to one of risk reduction more broadly construed.

This seemingly nuanced distinction is significant, particularly when viewed in light of the diet-heart debates. First, it demonstrates how developing scientific consensus on the subject of cholesterol control remains a difficult proposition. The rise of the statins over the past thirty years has indicated tacit acceptance of controlling cholesterol to prevent heart disease, even as the debate regarding the diet-heart hypothesis continued apace. This theory was epitomized by the previous set of guidelines, known as ATP-III, which put forward numerical LDL goals for all patients, suggesting that the more one's cholesterol were reduced, the lower one's cardiovascular risk.¹¹⁵ By eschewing numerical LDL goals, the new guidelines have moved away from a dose-dependent prescriptive approach (that drug doses should be increased until they have an adequate effect on blood cholesterol levels). Instead, physicians are instructed to start patients on “low” or “moderate” doses of statins based on their risk profile without checking to make sure their LDL levels are reduced to any preagreed “acceptable” level.

Second, and equally important, the rise of statin therapy and the new guidelines has coincided with a movement away from the cholesterol theory of coronary artery disease. As David Jones has demonstrated, there has long existed a lack of consensus within the medical community about the cause and management of heart attacks.¹¹⁶ This is particularly true in terms of the relative value and interrelatedness of lifestyle modifications and pharmacological treatments to prevent heart disease. At the center is the checkered past of the plaque rupture hypothesis, or the idea that heart attacks are caused not by the gradual buildup of clots that eventually obstruct coronary arteries but by existing plaques that become vulnerable and suddenly rupture, blocking off the artery. The hypothesis places the very mechanism of statins in question: do they work because they prevent the buildup of plaques in the first place or might they have another, anti-inflammatory effect that reduces the instability of some plaques, reducing the risk of rupture?¹¹⁷ Regardless, statins have remained exceedingly popular, reminding us that understanding the relationship between the disease process and the precise mechanism by which a treatment works is sometimes nominally important.

The diet-heart debates illuminate the ways in which unresolved tension between population-based and clinical-based science persisted due to the different disciplinary backgrounds of the actors involved. Mann's two-person refutation of the diet-heart hypothesis and Keys' response to charges of selection bias offer examples of efforts to consolidate different evidentiary standards. Moreover, while disputes about selection bias and causal inference focused attention on the theoretical underpinnings of the diet-heart hypothesis, discussions about how to apply diet-heart to medical practice were just as prominent.

Perhaps most impressive was the way in which the character and color of diet-heart discourse maintained a sense of continuity over a period of several decades. Even as researchers kept up long-standing epidemiological and laboratory studies or introduced new protocols and investigations, the same basic set of issues—and arguments—were the subject of renewed focus decade after decade. For diet-heart advocates, “selling the message” meant grappling with how to use the data derived from complex population-based studies to formulate “heart healthy” recommendations for individuals. On the one hand, using this data enabled researchers and policymakers to demonstrate how epidemiological investigations could contribute to the formulation of “heart healthy” recommendations. On the other hand, it also brought attention to unresolved methodological and analytical issues, providing diet-heart critics with a ready source of ammunition to challenge the diet-heart hypothesis.

The diet-heart hypothesis was one of the most promising concepts in twentieth-century American medicine. It was simultaneously one of the most divisive, spurring advocates and critics to interpret the same body of evidence in starkly different ways. In the process, it raised questions about the value of epidemiological research, the kinds of conclusions it could provide, and what comprised proper medical research. Despite the static and often vitriolic nature of the debate over approximately thirty years, it ironically became irrelevant as policymakers hurried to endorse a hypothesis that had already become clinical and public health orthodoxy.

Acknowledgments

Many thanks to Henry Blackburn, Deborah Doroshow, Robert Hackey, Joel Howell, David Jones, Chin Jou, Daniel Kevles, Susan Lederer, Gerald Oppenheimer, Naomi Rogers, and Sarah Tracy for taking the time to read and comment on early drafts of this article, and to the two anonymous reviewers for their constructive feedback.