An article in the August 25 issue of the Journal of the American Medical Association (JAMA), “Sugar-Sweetened Beverages, Weight Gain, and Incidence of Type 2 Diabetes in Young and Middle-Aged Women,” adds yet another chapter to the feeding frenzy that drives our nation’s love affair with epidemiological risk factorology. The article is a textbook case of misusing epidemiological research for the development of public health recommendations.
The JAMA article is strewn with misleading and sometimes-inaccurate statements and enough statistical hocus-pocus to make all but the most adept junk science sleuth dizzy. Perhaps the most glaring problem with the article, however, is the blurring of the distinction between correlation (or association) and causation.
A correlation simply describes the strength (but not the meaning) of a relationship between two factors. It turns out, for example, that there is a correlation between baldness in men and heart disease. This simply means there is some relationship between baldness and heart disease. But such a relationship is not necessarily causal. Certainly no one would claim that giving a bald man a toupee would decrease his risk.
This is because baldness does not have any influence on heart health–it is simply a factor that happens to be found more often in men with heart disease. When we say that any two factors such as baldness and heart disease are positively correlated, we are saying nothing about whether one causes or even affects the other.
Causal Link Not Established
In the JAMA article, the diseases in question are obesity and diabetes and the food involved is the sugar in sweetened beverages. The problem is that the identification of this food as a risk factor in this study may or may not mean there is a causal link with the diseases in question.
In a discussion on the limitations of epidemiological research in the journal Science, leading UCLA epidemiologist Sander Greenland summed up the complexities involved with obscuring the differences between correlation and causation by saying, “There is nothing sinful about going out and getting evidence, … nothing sinful about seeing if that evidence correlates. … The sin comes in believing a causal hypothesis is true because your study came up with a positive result.”
It is entirely possible that subsequent studies will not find an association between this food and these diseases. In fact, previous studies have actually suggested the opposite association between sugar consumption and obesity, and after reviewing the relevant research, the American Diabetes Association concluded in a recent Position Statement that “intake of sucrose and sucrose containing foods does not need to be restricted because of concern about aggravating hyperglycemia.”
JAMA Article Deceptive
Unfortunately, this “sin” is committed numerous times in the JAMA article, as the critical distinction between correlation and causation falls by the wayside. In setting up their argument in the very first paragraph, the authors correctly state, “recent evidence suggests an association between the intake of sugar-sweetened soft drinks and the risk of obesity in children.” In the very next sentence, however, they make the unwarranted jump to causation, saying, “besides contributing to obesity, sugar-sweetened drinks …”–directly claiming a causal link.
In their closing comments the authors state, “because of the observational nature of the study, we cannot prove that the observed associations are causal.” Yet this does not stop them from making the jump to causation in the next paragraph, where they recommend, “Public Health Strategies to prevent diabetes and type 2 diabetes should focus on reducing sugar-sweetened beverage consumption.”
Claimed Association Is Weak
A closer look at the findings shows even the proposed associations between the variables are weak at best. After correcting for confounding factors, the relative risk of developing diabetes in women drinking the greatest vs. the least amount of sugar-sweetened beverages was 1.32. As Gary Taubes noted in a 1995 article in Science magazine, epidemiologists generally agree that relative risks less than 2 should be ignored or at least viewed with extreme skepticism, particularly when there is conflicting research available.
Applying epidemiological research in this fashion is simply bad science. It tends to scare and confuse people, and it greatly oversimplifies the complicated etiology of the types of chronic conditions in question. A number of leading scientists involved in conducting this type of research have acknowledged the significance of this problem. Those who have not should heed the warning of Dimitrios Trichopoulos, head of the epidemiology department at Harvard School of Public Health, as quoted by Taubes:
“We are fast becoming a nuisance to society. … People don’t take us seriously anymore, and when they do take us seriously, we may unintentionally do more harm than good.”
Jonathan Robison, Ph.D. ([email protected]) holds a doctorate in health education/exercise physiology and a master of science in human nutrition from Michigan State University. An earlier version of this article, with references, appeared on TechCentralStation on August 25, 2004. Reprinted with permission.