Case Control Studies in CVD Epidemiology
As Oppenheimer has discussed in his historical background of CVD epidemiology, the field emerged with almost exclusive use of prospective designs for observational studies of risk, in contrast to the case-control approach that dominated the origins of cancer epidemiology (Oppenheimer 2006). Among many influences on this simple truth were the greater frequency of coronary than cancer events, along with the unreliability of coronary diagnoses and the easier comprehension of the cohort approach by the predominantly amateur investigators in CVD epidemiology.
Case-control strategies involved more sophisticated knowledge of design, while they lacked the already obvious causal criterion of “temporality”: that the cause be demonstrated to precede the event. Moreover, lower-order associations were also already appreciated for suspect risk factors and CVD events (e.g. 2-to 4-fold) compared to those between cigarette smoking and lung cancer (10 to 20-fold). Then, Felix Moore’s expert advice, as statistician at the National Heart Institute, had set forth the advantages of prospective design at the critical moment around Framingham’s founding.
Historically, nevertheless, the first attempt at epidemiological study in coronary disease had, in fact, a case-control design: Paul White’s comparison of characteristics among controls with those in patients following myocardial infarction at age 40 years or younger (Gertler and White 1954). Unique, and begun in 1946, the study results by 1954 were found wanting in inferences possible. White and Gertler converted it forthwith into a 25-year prospective study among the original cohort and two groups of controls (Gertler and White 1976).
Almost a decade later, due to a casual suggestion from Bill Kannel, and with the enthusiasm of Warren Winkelstein, then editor of the American Journal of Epidemiology, a younger member of the Framingham Study staff, Gary Friedman, was encouraged to explore, in the same population, risk ratios computed for the now-traditional risk characteristics but using case-control design compared to those calculated from the Framingham ten-year incidence data (Friedman et al. 1966).
Case-control versus Cohort: Framingham Risk Ratios
Thus, in 1965, as part of their learning process about the relative findings and merits of case-control and prospective designs for risk factor analysis, the Framingham Study group took its natural opportunity to compare ratios of risk based on ten-year follow-up data with ratios for cases versus controls as found in initial prevalence and later incidence data for the same Framingham population (ibid.).
Their preamble to this comparison provides a classic homily on strengths and limits of three observational study designs: prevalence, “case-series” (case-control), versus prospective (cohort) design (ibid., 366-367). The actual study computes relative risk as follows:
|Disease rate in persons with the factor|
|Disease rate in persons without the factor|
(With the prevalence rate of .0214 from the Baltimore City Survey as the independent base rate.)
They were pleased to find “surprisingly good agreement” among risk ratios by the three different methods for all characteristics except serum cholesterol level, and found likely explanations for that discrepancy (low prevalence, aging of the population, and probable treatment effects). In fact, they were so impressed with the similarity of findings that a decade later Friedman used the rich data base of Kaiser-Permanente Hospitals’ many cases and systematic measurements to carry out a risk analysis of cases and controls for the independent contribution of several “novel” factors including family history and sleep habits (Friedman et al. 1974). (Henry Blackburn)
Oppenheimer, G. 2006. Profiling Risk: The Emergence of Coronary Heart Disease Epidemiology in the U.S. 1947-1970. International Journal of Epidemiology 35: 515-519.
Friedman, G.D., Kannel, W.B., Dawber, T.R., McNamara. P.M. 1966.
Comparison of prevalence, case history and incidence data in assessing the potency of risk factors in coronary heart disease. American Journal of Epidemiology 83: 366-78.
Friedman, G.D., Klatsky, A.L., Siegelaub, A.B., McCarthy, N. 1974. Kaiser-Permanente epidemiologic study of myocardial infarction: study design and results for standard risk factors. American Journal of Epidemiology 99: 101-16.
Gertler, M. M., and P. D. White. 1954. Coronary heart disease in young adults. Commonwealth Fund. Cambridge: Harvard University Press.
Gertler, M. M., and P. D. White. 1976. Coronary heart disease: A 25-year study in retrospect. Oradell, NJ: Medical Economics Co.
Early Cardiovascular Epidemiology:
Case-Control Studies (1947-1972)
|Study Publication Title||Common Name of Study||Principal Investigator||Dates of Study|
|Coronary Heart Disease in Young Adults. A Multidisciplinary Study. Cambridge, MA; Harvard University Press, 1954: 1-218.*||Coronary Heart Disease in Young Adults||White, P.D.||1946- 1954|
|Comparison of Prevalence, Case-history and Incidence Data in Assessing the Potency of Risk Factors in Coronary Heart Disease. 1966. American Journal of Epidemiology 83: 366-378||Case-control versus Cohort Analysis in the Framingham Study||Friedman, G.D.||1964|
|Kaiser-Permanente Epidemiologic Study of Myocardial Infarction. 1974. American Journal of Epidemiology 99:101-116.||Case-control Study of Standard Risk Factors||Friedman, G.D.||1972-73|
*The White study was converted to a prospective study of the cohort with both pair-matched and unmatched controls, in a 25-year follow-up (see Cohort Studies).