Ancel Keys’s 1958 master plan for the Seven Countries Study (SCS)
This unpublished document was recently found in archives of the Laboratory of Physiological Hygiene at the University of Minnesota. We present it as a prime historical source of the early thinking and plans of Ancel Keys and his collaborators. It is, in fact, a more complete and consolidated version of the SCS design, methods, and conduct than is available in published monographs of the study. It clarifies the hypothesis and details the criteria for selection of regional cohorts. It emphasizes the solely ecological analysis of diet-disease relationships among the cohorts, often misconstrued by casual readers as relationships among individuals. It enables reflection on the plans actually carried out and others unrealized during the 60-year course of the study.
The experience and thinking of Keys and principal investigators is particularly evident in this straightforward but carefully qualified hypothesis to be tested in the SCS.
Hypothesis
“The basic hypothesis is that there are important [regional] differences in age- and sex-specific frequency of heart disease, particularly coronary heart disease, and that these differences are related, at least in part, to differences in the mode of life, especially in the diet, and to measurable differences in the personal characteristics of the members of the population who are clinically healthy at the time of examination. The hypothesis emphasizes the factors of diet, physical activity, and personal habits that are both measurable and alterable, and factors of physico-chemical constitution which are at least measurable with available methods. . . .
“Finally, the hypothesis holds that follow-up study of such populations after an initial examination should discover the statistical prognostic significance of characteristics recorded for the individuals at the time of initial examination, particularly if the populations have a relatively high degree of stability in regard to mode of life, including the diet and physical activity.”
Organization
Keys describes the plan for “Central Organization” of the SCS as one directed by a single “Responsible Investigator” who is supervised by an “Overall Committee” and advised by an “Operating Committee.” This Central Organization would coordinate the whole and provide central analysis of survey data from each and all areas.
In practice, formal meetings were not held of a full complement of Principal Investigators in such an Overall Committee, either for SCS planning or operation. No formal Operating Committee functioned in the SCS. There was, however, a Responsible Investigator, Ancel Keys, who communicated regularly with the Principal Investigators for each regional cohort.
From the late l990s, Keys served as adviser to an SCS editorial group of Alessandro Menotti, Daan Kromhout, David Jacobs, and Henry Blackburn, which is still active in 2022.
Cohorts
Cohorts numbered 600 to 1000 men, encompassing all men ages 40-59 in a defined community, recruited in regions purportedly contrasting in traditional diet and/or in coronary disease rates. Keys explicitly included regions thought to differ in coronary disease incidence or death rates as a criterion for selection of cohorts as well as their purportedly contrasting habitual dietary patterns.
This was a fact, but only in this early plan from Keys was it so formally designated. In the 1950s there were no reliable data about regional differences in heart disease incidence and deaths or in foods consumed. There existed, nevertheless, starkly contrasting traditional eating patterns such as the relatively high vegetable fat of Mediterranean countries and high-animal-fat duet in northern Europe and the U.S., versus the very-low-fat, high-carbohydrate pattern in Japan. The SCS deliberately chose–in this first approach to testing the diet-heart hypothesis–the areas of apparent great contrast in eating tradition or incidence of heart disease rather than a random choice of regions.
We learn in this document that Keys anticipated engaging a rural U.S. cohort as a more appropriate comparison to the international worker cohorts. That plan was not realized, presumably because of its great cost. The U.S. railroad-men occupational cohort–which was neither an intact nor representative population but available and already under study–was, for convenience, used to represent a U.S. worker population.
Methods
Keys describes the standardized formularies–developed for recording the medical history and physical examination data and for reading and classifying ECG findings–as providing “the maximum of objective detail and minimum of present interpretation, with a view to future statistical analysis.” He anticipates the possible addition of standardized measures of cerebrovascular, pulmonary, and peripheral circulatory function, and even “emotional stress.” Except for ventilatory function, these tools were never well developed or effectively used in the SCS.
Diet
For diet, Keys recognized from the outset–based on a vast experience testing repeatability and validity of diet questionnaires–that the diet of individuals could not be adequately represented by quetionnaire surveys of foods consumed, short of many repetitions that were infeasible in field studies among whole populations. Keys proposed chemical measures of diet nutrients consumed in “a series of households chosen to be representative of the area.”
Thus, the SCS collected composite samples of actual foods consumed in randomly selected households of each regional cohort. It repeated this collection during several seasons of the year and computed nutrient content from standard food conversion formularies and, uniquely in the SCS, from central laboratory chemical analyses of the food composites for nutrients and particular fatty acids. The product was the most reliable and representative of the nutrient compostion of a whole population’s diet.
SCS diet-disease relationships studied were only ecological, using the food and nutrient composition representing entire regional cohorts, not data for individuals within the cohorts. Systematic measurement and analysis of individual dietary recall data was attempted only in the Finnish and Netherlands cohorts where investigators with particular interests had obtained ancillary study grants. Casual readers of SCS publications have misconstrued this entirely ecologic approach to diet-disease relationships reported in SCS monographs. All diet and diet nutrient-disease analyses for the study were among cohorts, not individuals within the cohorts.
Physical Activity
Keys includes in the plan: “observation of the actual work of members of the population,” which, however, never became a standard survey procedure. Only occupation was used to represent habitual physical activity in the rural cohorts. Work and leisure-time activity questionnaires and some ergonomic measures taken among the occupations, were used in the U.S. and Italian railroad-men studies but not in the rural worker cohorts.
“Perceived work load” was recorded from select subjects during exercise tests in a few areas where that test was standardized. Indeed, at the same workload, perceptions of work intensity differed between Italian and Finnish cohorts.
Conclusions
There is no record that the original SCS Principal Investigators read or participated in composing this early planning document from Ancel Keys. Those of us working then in the field or who later became the principals did not know of it. We might have better described the design and conduct of the SCS, and sooner corrected others misconstructions about the study, if we had earlier read and discussed this original plan.
Henry Blackburn