‘Circular Epidemiology’: Re-Analyzing Ourselves to Death? L. Kuller
There is a tendency in academic epidemiology toward the repetitious re-examination of traditional cardiovascular risk factors, dwelling inordinately on finer points, and “gilding the lily” with minimally enhanced risk discrimination—a tendency that inspired Lew Kuller to coin the term “circular epidemiology,” reflecting, at best, a process of solidifying established ideas and well-tested concepts, and, at worst, an absence of new ideas. (Kuller 1999) Indeed, a certain ennui in the discipline was acknowledged in the 1993 report of the NHLBI Task Force on Research in Epidemiology and Prevention (Task Force 1994).
But what is circular and what is new in CVD epidemiology? What useful knowledge has come from re-examining the traditional risk factors and understanding their effects over the long term and that of their secular change, and of their force among population subgroups by age, gender, and ethnicity, and over the course of time? It is evident that CVD risk factors measured in the 1950s and ‘60s in Framingham have different meaning and impact measured in our century and in new cohorts. Assessment, furthermore, of the independent contribution of so-called novel risk factors requires analysis both in conjunction with and as add-on estimates to the traditional ones. But the results are seldom earth-shaking.
Then there are needed details about quantitative effects; we still argue, for example, about the optimal amount, duration, and frequency of exercise associated with lower CVD risk, best studied in the context of an individual’s past and present activity measurements. We must also measure anew the role of major nutrients as we explore new ideas about micro-nutrients, essential fatty acids, and antioxidants. These important new directions derive much from traditional risk-factor epidemiology and justify at least a part of the decried circularity.
And, in the enlightened era of molecular biology, we need to determine the population frequency and the public health relevance, if any, of genes and alleles related to hypertension and atherosclerosis and to their traditional risk factors. Moreover, re-examination of the role of inflammation in atherosclerosis has grown out of studies and new laboratory findings complementary to traditional epidemiology.
On the other hand, the downside of circular epidemiology is much in evidence in a flurry of studies of the purported new ‘metabolic syndrome,’ ‘Syndrome X,’ which David Goff at Wake Forrest University, characterizes as follows:
“It seems as though when you take a group of five perfectly good, continuously measured [traditional] CVD risk factors, categorize them into five dichotomous variables, sum them, then dichotomize the summed variable, you get a ‘syndrome’ that is related to the risk of cardiovascular disease. If that were not remarkable enough, a truly fascinating feature of the ‘syndrome’ is that it doesn’t predict CVD any better than the five original continuously measured [traditional] risk factors! Nevertheless, we have tremendous effort going into the study of the epidemiology and genetic determinants of the [metabolic] syndrome” (Goff 2006).
Goff adds that “a generation of new trials and evidence,” including the DASH Diet, the DASH Sodium Diet, and the Diabetes Prevention Program, has been “built largely on contributions of past epidemiology” and that new legislation and policies designed to change the environment in order to promote cardiovascular health “point toward the epidemiology of the future” (Goff 2006).
Aaron Folsom at Minnesota responded specifically about the issue:
“The circularity criticism holds. Many lifestyle and individual characteristics have been studied quite completely. Replication is valuable in observational epidemiology but we need no more replication for many factors. But new questions arise for which we need observational epi. For example, we still argue about the optimal amount of physical activity for good health. Subtle aspects of diet have proven important–diet patterns vs diet fat composition. [And I would add, one always has to determine whether and how much a ‘novel’ risk factor adds to prediction over and above the traditional, which requires replication!]
The “inflammation” paradigm arose to a great deal from observational epi and certainly was important. We certainly need cohorts to determine the population frequency and public health importance of the multitude of genetic variants identified by our lab colleagues. I’d be reluctant to identify studies that were a waste of money. The case of [the ‘toxicity’ of] Vioxx is an example of how epidemiology can supplement inadequate information from short term clinical trials.”
David Jacobs, at the University of Minnesota, takes a more philosophical approach to the “circular” charge:
“Epidemiology is imperfect, a very blunt tool, but it is the only way to examine people and populations over the whole life span. Epidemiology is the ultimate arbiter of success or failure of any treatment. . . Thus, epidemiology, including ‘circular epidemiology,’ can be highly relevant. Epidemiology deals in probabilities, not certainties. It applies to everyone without applying to anyone [in particular].
“We deal in probabilities, and there are always individuals who present an exception to any epidemiologic rule. . . Many of the critics of epidemiology are reductionists who really don’t believe in probability. They are looking at deterministic pathways. These are the people . . . who consider observational studies and surveillance of cardiovascular risk and trends [to be] uninteresting and too costly.
“In fact, short of careful design in monitoring trends of morbidity and death in populations, along with remeasuring risk characteristics in successive surveys, no one can really tell what is happening over time in whole populations. Those findings are difficult . . . and costly to obtain. But when you have them, you notice some very basic facts that. . . guide public health action, though they are not specific in terms of mechanisms” (Jacobs 2006).
To better understand the circularity criticism we asked colleagues to suggest prevention studies of the last decades that should have been avoided on this basis. We received thoughtful replies, most of which largely complemented the analysis of the originator of the ‘circular’ term, Lew Kuller. Some felt that things we still don’t know involve perforce, the re-examination of new ideas in old data and old ideas with current data.
Thus, self-criticism in CVD epidemiology pays off. Acknowledgment of the criticism of circularity allows one to move on with, and not abandon, a reliable tool, a complementary discipline, and a broader life view than that of individual or “Personalized Medicine.” (Henry Blackburn)
Folsom, A. Personal interview 2006
Goff, D. Personal communication, 2006.
Jacobs, D. Personal communication, 2006.
Kuller, L. 1999. Circular epidemiology. American Journal of Epidemiology 50:897-903.
National Heart, Lung and Blood Institute, National Institutes of Health. 1994. Report of the Task Force on Research in Epidemiology and Prevention of Cardiovascular Diseases. Bethesda, MD: National Heart, Lung and Blood Institute.