Genetic Study Supports Carbohydrate-Insulin Model of Obesity

Dr. David Ludwig
4 min readJan 2, 2018

According to the Carbohydrate-Insulin Model of Obesity (CIM), the processed carbohydrates that flooded our diet during the low-fat diet craze undermine our metabolism and drive weight gain. Put simply:

  1. Processed carbohydrates — think white bread, white rice, potato products, low -fat snacks — raise insulin more than any other food, calorie for calorie. This is just Nutrition, 101.
  2. Insulin is the Miracle-Gro for your fat cells. A child with new onset type 1 diabetes — unable to make enough insulin — will invariable lose weight until receiving treatment, no matter how many calories she consumes. Give that child the right amount of insulin, and weight trajectory returns to normal. Give that child (or an adult with type 2 diabetes) too much insulin and excessive weight gain will predictably result. This is just Endocrinology 101.
  3. When too many of the calories we eat get locked away in fat cells, there aren’t enough calories to supply the needs of the brain and other organs. So we get hungry and “overeat.” And to make matters worse, metabolism slows down, further fueling weight gain. This is just Obesity 101.

(For more, see this NY Times op-ed and these two articles in JAMA.)

While each of these steps is fairly non-controversial, the three together — that is, the CIM — make for one of the most hotly contested topics in nutrition today. No less than 6 recent academic reviews have aimed to dismiss the CIM (see here, here, here, here, here and here), primarily citing the work of one investigator.

Elsewhere, I’ve argued that these - categorical - rejections are based on a misinterpretation of very short-term feedings studies and disregard for extensive supportive data. Moreover, the conventional view of obesity, focused on “calorie balance” — has utterly failed to explain the obesity epidemic, beyond people’s inability to control themselves in the modern food environment. We’re in desperate need of new thinking.

Admittedly, the CIM remains unproven. The definitive clinical trials are exceeding expensive and, even if they do get funded, will take years to perform. Unfortunately, conventional observational (“associational”) research can’t come to the rescue, because of the difficulty of determining which comes first:

overeating > increasing body fat > high insulin secretion (Conventional View)


high insulin secretion> increasing body fat > overeating (The CIM)

This difference is of much more than just theoretical interest, with direct implications for how best to prevent and treat obesity. If the Conventional View is right, we need to focus even more intensively on cutting back calories, for example with a 1600 calorie diet. If the CIM is right, the emphasis should instead be placed on lowering insulin secretion with a lower-carbohydrate/higher-fat diet and other supportive dietary and lifestyle measures. Calorie balance will then adjust naturally due to reduced hunger, greater satiety and faster metabolism.

Fortunately, a new scientific method, called Mendelian Randomization (MR), can help disentangle cause and effect. MR is based on the fact that a child receives genes from both parents at conception in a random fashion (for all practical purposes).

Let’s say you are interested in studying the effects of an exposure (for example, HDL-cholesterol) on a health outcome (heart attack). First you identify genetic determinants of that exposure within a population. Then you see whether those genes predict likelihood of the outcome. In the case of HDL-cholesterol and heart disease, the results from MR provide evidence against a long-standing belief based on less sophisticated methods.

In an article published today in Clinical Chemistry’s Special Issue on Obesity, my colleagues in Boston and I used MR to test whether higher insulin secretion throughout one’s life (the exposure) is related to higher body weight (the outcome)— a key assumption of the CIM. Using several large international databases, we created a genetic score that predicts how much insulin a person releases after consuming carbohydrate (technically, insulin 30 minutes into an oral glucose tolerance test). We found that genetically-determined insulin secretion predicted body mass index with extremely high confidence (p<0.00000000000000000001) and a potentially large effect across the population. Body weight of high insulin secretors was about 10 pounds more, on average, than low insulin secretors (comparing 90th to 10th percentiles) — an effect that might be even greater among those consuming the most carbohydrate. Of particular importance, the “reverse” relationship was null. That is, genetically determined body mass index did not predict insulin secretion to any degree (p=0.43).

These results are, as highlighted in an accompanying editorial, “less susceptible to confounding and reverse causation, the Achilles heel of other association studies … information that must be incorporated into our thinking and not rejected on the grounds that it conflicts with a rigid intellectual position”. (The editorial also provides a concise history of CIM, whose origins date back a century.)

Pending further research, all sides of this debate would do well to avoid proclamations of having either proven or disproven the CIM. Ultimately, we will likely find that the truth is more complicated than expected, with important variations (based in part on biological differences) that determine an individual’s optimal diet for obesity treatment, chronic disease prevention and longevity.

The full text articles are freely available for 2 weeks: study and editorial

For all articles in the special Obesity Issue of Clinical Chemistry, click here



Dr. David Ludwig

Physician, Nutrition Researcher, and Public Health Advocate. #1 NY Times bestselling author ofALWAYS HUNGRY? and ALWAYS DELICIOUS