No, It's Not the Sugar – Bittman and MotherJones have overinterpreted another study

Diet seems to be all over the New York Times this week, with an oversell of the benefits of the Mediterranean diet, and now Mark Bittman, everyone’s favorite food scold, declaring sugar is the culprit for rising diabetes. His article is based on this interesting new article in PLoS One and begins with this wildly-inaccurate summary:

Sugar is indeed toxic. It may not be the only problem with the Standard American Diet, but it’s fast becoming clear that it’s the major one.

A study published in the Feb. 27 issue of the journal PLoS One links increased consumption of sugar with increased rates of diabetes by examining the data on sugar availability and the rate of diabetes in 175 countries over the past decade. And after accounting for many other factors, the researchers found that increased sugar in a population’s food supply was linked to higher diabetes rates independent of rates of obesity.

In other words, according to this study, obesity doesn’t cause diabetes: sugar does.

No! Not even close. I hate to repeat his misstatement, because I’d hate to reinforce this as a new myth, but it’s critical to see his full mistake here. This is a wildly inaccurate summary of the authors’ findings, and one they don’t even endorse in their discussion. Bittman has actually just said “obesity doesn’t cause diabetes”, and now has proven himself a deluded fool.

Let’s talk about this paper. This is what is called an “ecological study”, which means it studies populations as a whole, rather than individual patients. Using data from the United Nations Food and Agricultural Organization, the International Diabetes Federation, and various economic indicators from the World Bank, the authors compared populations of whole countries, in particular the prevalence of diabetes correlated to other factors such as GDP, urbanization, age, obesity, and availability of certain varieties of food like sugar, meat, fibers, cereals and oil. Using the rise, or fall, of diabetes prevalence over the last decade in various countries, they correlated this increase with increasing availability of sugar, obesity, urbanization, aging populations etc., and found a few interesting things. For one, increases in GDP, overweight and obesity, tracked significantly with increasing diabetes prevalence. But interestingly, when those factors were controlled for, increasing availability of sugar also tracked linearly with increasing diabetes prevalence, and the longer the duration of the exposure, the worse it got.

However, this does not mean that “obesity doesn’t cause” diabetes, if anything, it’s further support for the exact opposite. While a correlative study can’t be a “smoking gun” for anything, the data in this paper supports increasing modernization/GDP, obesity, and sugar availability are all correlated with higher diabetes prevalence. Even if the sugar relationship is causal, which is no guarantee, the increase in sugar availability could only explain 1/4 of the increase in diabetes prevalence. Obesity is still the main cause of diabetes, which can be demonstrated on an individual level by increases in weight resulting in loss of glycemic control, and subsequent weight loss results in return of euglycemia. In particular, the results of studies of bariatric surgery, in both restrictive and bypass procedures, weight loss is accompanied by improvement in diabetes. The attempts of toxin paranoids like Bittman to reclassify sugar as a diabetes-causing agent, and to dismiss obesity as a cause, are highly premature.

Mother Jones, has a slightly more balanced read, but it still oversells the results.

This is a correlation, of course, and correlation does not always equal causation. On the other hand, it’s an exceptionally strong correlation.

Well, that’s another overstatement. Want to see a picture?

http://www.plosone.org/article/fetchObject.action?uri=info:doi/10.1371/journal.pone.0057873.g002&representation=PNG_M

Article Source: The Relationship of Sugar to Population-Level Diabetes Prevalence: An Econometric Analysis of Repeated Cross-Sectional Data

Basu S, Yoffe P, Hills N, Lustig RH (2013) The Relationship of Sugar to Population-Level Diabetes Prevalence: An Econometric Analysis of Repeated Cross-Sectional Data. PLoS ONE 8(2): e57873. doi:10.1371/journal.pone.0057873

Figure 2. Adjusted association of sugar availability (kcal/person/day) with diabetes prevalence (% adults 20–79 years old).

I wonder what the R-squared is on that line fit. Now, consider a comparison with obesity rates by diabetes prevalence:

Figure 1. Relationship between obesity and diabetes prevalence rates worldwide.

Obesity prevalence is defined as the percentage of the population aged 15 to 100 years old with body mass index greater than or equal to 30 kg/meters squared, from the World Health Organization Global Infobase 2012 edition. Diabetes prevalence is defined as the percentage of the population aged 20 to 79 years old with diabetes, from the International Diabetes Federation Diabetes Atlas 2011 edition. Three-letter codes are ISO standard codes for country names.
doi:10.1371/journal.pone.0057873.g001

Hmm, they didn’t fit a line here, but I can bet the fit would be better. Diabetes strongly correlates with BMI, this has been shown time and again using national survey data like NHANES or SHIELD. And before people start whining about BMI as an imperfect measure of obesity, it is perfectly appropriate for studies at a population level, and other metrics such as waist size, hip/waist ratios etc., all show the same thing. Diabetes risk increases linearly with BMI, with as many as 30% of people with BMI > 40 having diabetes, and further, we know from cohort and interventional studies that weight loss results in decreased diabetes. Much of this data is correlative as well (with the exception of the weight-loss studies), and the study that would prove this for certain – dividing people into diets providing excess fat, vs sugar, vs mixed calories, vs controls, with resultant measurement of diabetes rates, would be unethical. Either way, declaring sugar the enemy is both incomplete, and premature. While this paper provides interesting correlative evidence for increased sugar availability increasing diabetes prevalence, it is still subject to risk of confounding errors, it is correlative, and the link does not explain away other known causes of type II diabetes such as obesity. It is a warning however, and we should dedicate more study towards determining if sugar consumption (rather than mere availability) is an independent risk factor for type II diabetes.

Bittman has wildly overstated the case made by this article. He should retract his claims, and the title and false claims should be corrected by the editors. This is a terrible misrepresentation of what this study shows.