Diet Soda is Clearly an Evil Plot to Kill Us All

One common theme of this blog recognizable to regular readers is that medical studies based on giant data sets, especially those including self-reporting data, are quite limited in their implications for how an individual should live his or her life to promote optimum health. A recent article publicized as linking diet soda consumption to greater risk of diabetes and heart disease, and mortality in general, is a good example of this.

The problem with the study (Lutsey, P.L., L.M. Steffen and J. Stevens, 2008. Dietary intake and the development of the metabolic syndrome: The atherosclerosis risk in communities study. Circulation 117:754-761) is not that it is necessarily wrong. It is that it is too hard to tell what importance the findings have in the context of all the other health information with which we are bombarded daily.

Aside from the problems inherent in self-reporting diet data — which the authors acknowledge in the discussion but which obviously had no effect on likelihood of publication or promotion of the press release — the authors reveal a troubling bias in their assumptions and use of terminology. They conducted a factor analysis on dietary components in an attempt to see which parts of a diet are more highly correlated with a condition called metabolic syndrome (fat, high blood pressure, high cholesterol, etc.) or “MetSyn”. So far so good. But the next problem here is that human beings had to classify the types of food people ate, and these classifications are based on assumptions already about what foods are good and bad for you. For example, “red meat” of course does not take into account what species of mammal was eaten, or the conditions under which it was raised, which surely affect its nutritional and fat content. They use a category “low-fat diary” because they believe there is a reason to distinguish it from non-low-fat dairy.

The use of categories is necessary for their methodology, but it illustrates the problem with nutritional data collection which to this point is always colored by currently held biases about good and bad food. To make matters worse, when their factor analysis revealed two broad dietary patterns (based on their categories) among the people studied, they chose to label these “Western” and “Prudent” dietary patterns. Guess which dietary pattern they have already decided is bad for you, and likely to cause MetSyn?

Although their results showing more people on the “Western” diet to acquire “MetSyn” show a correlation, they are quick to label certain food groups (e.g. dairy) as “protective”. But of course this is based on the studies in the past that have shown certain types of food to have negative effects on human health (usually, though, only when consumed in high quantities). Of course then none of the results were too surprising, except for the finding of diet soda, consumption of which in their model increased risk for MetSyn even more than consumption of sweetened drinks. This was the splashy result that got the newspaper headline.

I have no problem with their explanation, it makes perfect sense. They first admit that the data might be confounded, because diabetics are more likely to drink diet soda than nondiabetics, so which came first? But they also cite a rat study which suggested that artificial sweetener screws up our body’s ability to determine the caloric content of what we are consuming - our mouth says “caloric” and our body says “not” and thus our body may simply stop trusting our mouth. It’s a much more interesting explanation, and makes some intuitive sense, so that makes it easy to ignore the explanation that the data are confounded. The study gets published anyway, and promoters at the journal find the line in the results and discussion that will attract media attention, and bingo.

Those of us who think artificial (= man made) sweetners are potentially nasty unknowns to be avoided if possible, love a result like this. But that still doesn’t mean that if you drink a lot of diet soda you are doomed to be fat and get diabetes. Even when we really know very little about what we are studying, even when our methods are poor, and the results questionable (in some cases contradicted by other studies, as the authors cite), and based on broad assumptions, these studies get published, because they get headlines. Bad methods are considered acceptable to medical journals because there are not necessarily feasible methods that are valid. Assumptions may be based in part on established medical knowledge, but they are mostly based on previous, poorly conducted studies such as this one, not to mention constant propaganda from our media and government about what is good for us, and which is so obviously correct that they completely revamp the propaganda every couple of decades or so. We think we are learning more and more and more when we “confirm” these same assumptions, yet in truth we haven’t even begun to understand the complexity and variation in the human body. Just ask a scientist researching disease cures, who actually has to get it right for anyone to care about her research. Eating dairy products is “protective”? Check that with someone from a genetic heritage of lactose intolerance.

Never forget that these studies are blunt, blunt instruments, that tell us nothing about what works for an individual. A good rule of thumb for most diets is, simply, variety, and not too much of any one thing. That, not forcing down a gallon of skim milk every day, should be the first step for anyone trying to feel better through a diet better suited to his body.

Tags: health

This entry was posted on Friday, February 22nd, 2008 at 12:51 am and is filed under health, humans, statistics. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.