Blaming the Food, part 1: Measuring Bad

"Measurement is the first step that leads to control and eventually to improvement.  If you can't measure something, you can't understand it.  If you can't understand it, you can't control it.  If you can't control it, you can't improve it."

                                                  -H. James Harrington  

For the past few hundred years, there have been two schools of thought regarding diet and obesity.  One says that no one food is inherently fattening and so we must endorse universal moderation.   The other says that there are most certainly fattening foods and that they should be avoided, or at least restricted, to stave off or reduce body fat.  It has only been in the past thirty or forty years that dietary fat has been regarded as particularly fattening; starchy and sweet foods have long been regarded as the foods that lead to corpulence.  But the notion that refined carbohydrates are fattening has been just that - a notion.  Is there some way that we can measure or quantify the "fattening-ness" or "obesogenicity" of one food over another?

In the early 1980's, the Glycemic Index (GI) was developed by Dr. David Jenkins to establish the physiological basis for the carbohydrate exchange program that was prescribed to diabetics.  Specifically, the GI quantifies the blood glucose response to different foods based upon a standard amount of carbohydrates in the food (Jenkins American Journal of Clinical Nutrition 1981 24: 257-264).  It is used as an indicator of how fast or slow glucose is absorbed from foods.  At this time, diabetics were being prescribed a high carbohydrate diet (45-55% kcal in the diet) rather than the traditional moderate carbohydrate diet (30-40%) that had been used even after the advent of insulin (The Hartford Courant, 2 January 1983); therefore, it was prudent to establish which carbohydrates would be better tolerated by diabetics in order to maintain sufficiently low blood glucose levels.  The GI would offer a quantitative way to validate this exchange program.

Briefly, the GI is determined by first measuring changes in blood glucose over the course of three hours after eating a single food.  The portion of food is the amount of food that contains 50 grams of carbohydrate.

The glucose response to this test food is then compared to the response of a standard food - a 50 gram glucose solution or bread - by taking the quotient of the area under the curve of each response.  This quotient is multiplied by 100 so that the GI is a whole number.  If pure glucose is used as the test food, then all other foods have a glycemic index of 100 (equal to pure glucose) or lower than 100; if bread is used as the test food, then foods such as glucose and cornflakes will actually be higher than 100.

The results of the GI are not entirely revolutionary.  Meats and dairy have a considerably lower GI than starchy foods, and beans fall somewhere in between, although they are closer to meats than starches.  However, the GI does give some interesting insight.  Despite both being composed of processed wheat, the glycemic index of pasta is 40% lower than that of bread; and boiled white potato is nearly as high as pure glucose while sweet potatoes have a GI of half that.  But it is far from perfect.

Since the inception of the GI, it has taken considerable - and plenty deserved - criticism.  For one, it looks at foods in isolation, which by and large, is not how we eat our food.  Protein can raise insulin, which will lower glucose in the blood, and fat retards digestion and absorption, which may explain why when foods are combined, the predicted GI often diverges from the observed GI.  Furthermore, not every food is eaten in 50 gram portions.  Carrots and white potatoes both have an exceedingly high GI, but a person would have to eat nine medium carrots versus one and one-half white potatoes to achieve that 50 grams of glucose.  So how can this be accounted for?

In 1997, Salmerón and colleagues (Salmerón Diabetes Care 1997 20: 545-550) published a study looking at the influence of the GI and intake of cereal fiber on the risk for developing Type 2 diabetes.  In their paper, they introduced the Glycemic Load (GL).  Because insulin is secreted primarily in response to dietary carbohydrate, the GL was used "as an indicator of a glucose response or insulin demand induced by the total carbohydrate intake."  Or more succinctly, GL estimates the insulin response to a single food, a meal, or an entire diet.  The GL is the product of the carbohydrate content of a food (or meal) times the GI of the food (or meal), which is then divided by 100 in order to return the GI from a whole number back into a ratio.

We now have two estimates for the bodily response to dietary carbohydrate.  Glycemic index is an estimation of the rate of absorption of carbohydrates from different foods.  It is valid if two different foods have similar portion sizes or contain a similar amount of carbohydrate.  The glycemic load estimates the insulin response that will be induced by different foods.  Because it takes into account the carbohydrate content of the food, it remains valid for foods that are not commonly eaten in 50g portions and can be used to differentiate between foods of varied carbohydrate composition that have similar glycemic indexes.  But to be critical scientists, we want to know: Are these estimates complete?  Do they accurately depict how our foods affect us? Can they be used to determine the obesogenicity of a food?  And if they can, is it reasonable to translate it from laboratory to table?