More epidemiology shenanigans

By Sandy Szwarc, BSN, RN, CCP

The media has been quick to report, verbatim from the press release, of a new study finding that high fructose corn syrup (HFCS) could be contributing to a global epidemic of type two diabetes. The study reportedly found that countries with HFCS in their food supply were associated with a “significantly” 20% higher risk for diabetes compared to countries that don’t.

‘High fructose corn syrup appears to pose a serious public health problem on a global scale’

“HFCS appears to pose a serious public health problem on a global scale,” said principal study author Michael I. Goran, professor of preventive medicine, director of the Childhood Obesity Research Center and co-director of the Diabetes and Obesity Research Institute at the Keck School of Medicine at USC. “The study adds to a growing body of scientific literature that indicates HFCS consumption may result in negative health consequences distinct from and more deleterious than natural sugar.”…

Even without knowing the details of how the study was conducted or what the body of the scientific evidence on HFCS and diabetes has proven, the public had a quick and easy way to know this study was junk science: the 20% increased risk was nowhere near the 200 to 300 percent that scientists consider necessary to be even considered tenable. In other words, the results were no greater than might have been found by chance and random coincidence or reflect statistical or modeling errors and biases, as well as countless confounding factors.

This epidemiological paper was a classic example of the very worst problems rife in these studies. It’s worth reminding ourselves just how uncredible these studies can be  and that the associations they find should never be taken as indicative of a cause. In fact, after taking a look at a few illustrations of this study’s methodology, it is unimaginable this paper even passed peer review to be published at all.

As we know, epidemiological studies are done in a computer, using software that dredges through vast databases and performs regression computer modeling and statistical manipulations to find correlations with some health outcome. These computer programs can—and do—pull out links in almost unlimited combinations, many of which are spurious and contradictory, regardless of whether there is any real connection between the variables. These types of studies are also the most easily manipulated to arrive at whatever conclusions the researcher set out to find.

Premises

In this study, published in Global Public Health, the authors were from the preventive medicine department at the University of Southern California, and the social and cultural anthropology institute at the University of Oxford. They began by stating the premise for their research, repeating two popular myths about type 2 diabetes and about HFCS:

  • Sugars and refined carbohydrates are “directly linked to risk for obesity and type 2 diabetes;” and
  • “HFCS is especially detrimental to metabolic health and risk for type 2 diabetes.”

Not only are both premises scientifically implausible, the research they cited in support of their beliefs has been debunked by the scientific community and even retracted by their authors.

First, sugar does not cause diabetes. High blood sugars are a symptom of the disease, not the cause. Diabetes is a metabolic disorder that is primarily genetic, with type 2 considerably more inheritable than type 1 diabetes. Concordance rates among identical twins for abnormal glucose metabolism, for example, are nearly 100% and have been estimated at 70 to 90% for type 2 diabetes. (That’s greater than the 60 to 80% attributed to our height.) While there are various forms of type 2 diabetes, it is also a disease of aging, with cases rising exponentially with age. Weight gain can also be a symptom of diabetes and insulin resistance (weight also rises with aging), but obesity’s association with diabetes doesn’t make it the cause, either. Sugars have been studied up and down, inside out and sideways for decades and repeatedly found to not cause diabetes, obesity, heart disease, cancer or any other chronic disease. These study authors, however, wrote:

The global trends in the rising prevalence of obesity and type 2 diabetes parallel shifts in dietary patterns that have resulted from an increase in processed, ‘Western’ style foods with high energy density, which have now become popular in many countries. A feature of most Western diets is the consumption of high levels of refined carbohydrate, and in particular, sugar….

Second, HFCS is metabolically and nutritionally equivalent to ordinary table sugar (sucrose), composed of about equal parts of fructose and glucose. The source of sucrose in our diet — beets, cane, dates, corn, etc. — is irrelevant to the human body, which breaks down all sugars and metabolizes them the same. Nor can our bodies distinguish between natural or added sugars, as they are chemically identical. HFCS is made in two formulations, 42% fructose and 55% fructose, with the rest glucose. The 55% version, used in many sodas, is exactly the same sweetness as sucrose that’s table sugar.

The availability of HFCS, or any other food in our food supply, doesn’t make that ingredient the cause of an illness. While HFCS has been an increasing source of sugar in our diet, it has simply replaced earlier sugar sources, such as beets or cane, while the total amounts of fructose and sugar in our diets have remained relatively constant for the past thirty years, according to the FDA. Total global sugar consumption has increased roughly in line with the increase in the world’s population, but per capita  consumption has remained stable for more than 40 years and has even declined in industrialized countries, such as the United States, according to the FAOSTAT (Food and Agricultural Organization of the United Nations), the World Sugar Research Organisation, and the 2012 USDA Economic Research Service Sugar and Sweeteners Outlook.

Popular scares about HFCS have been based on the misconception that HFCS is the same as fructose. Advocates of this popular misconception have inappropriately used studies looking at pure fructose to point to malabsorption problems. But fructose is never eaten that way; it always accompanies glucose in roughly equal proportions, where malabsorption problems disappear. No clinical studies of HFCS as it’s normally consumed have ever shown it to be harmful. These study authors, however, wrote:

A growing body of evidence supports the hypothesis that in addition to overall sugar intake, fructose is especially detrimental to metabolic health and risk for type 2 diabetes. This is of particular concern given the global changes that are occurring in the use of high fructose corn syrup (HFCS) in food and beverage production…

The authors not only incorrectly used studies looking at unrealistically high percentages of pure fructose to suggest a health danger, they went on to repeatedly imply that manufacturers are secretly putting more fructose in sodas, citing a prior study they had conducted where they had reportedly found an undisclosed higher fructose content. That study, however, had been resoundly criticized by the scientific community and found to have used invalid laboratory testing methods and the authors later admitted there were errors cause by the methods used.

So, with two false premises, the authors went on to examine the relationship between HFCS availability in food supplies around the world and obesity and type 2 diabetes.

Data

In creating their database, as many epidemiological studies do, they compiled countless estimates, assumptions, extrapolations and deductions on HFCS availability and type 2 diabetes prevalence from published and unpublished sources, all of which used different methods and found widely varying results. Such estimates have been shown to have very limited relevance to actual dietary intakes or actual disease rates and are especially inaccurate when comparing different countries with widely divergent health surveillance methods, quality and completeness of health surveys and records, economic status, population genetic homogeneity or diversities, and cultural and social differences. The authors then used population estimates to calculate rates of type 2 diabetes in different countries, and then made a number of arbitrary decisions in setting up their comparison groups.

They didn’t conduct any actual dietary analyses or clinical diagnostic tests.

Diabetes. For their data on the prevalence of type two diabetes, they used two different sources:

  • The International Diabetes Federation (IDF) Diabetes Atlas, fourth edition (2009) which based its estimates for the prevalence of diabetes, according to the authors, on literature searches of published information and contact with IDF member countries. The IDF 2012 edition  wasn’t used. The IDF estimates include the total prevalence of all types of diabetes from age 20, as defined by fasting blood sugar level using the new WHO diagnostic criteria, among 229 countries.
  • The GBMRF Collaborating Group’s global estimates, which drew on health surveys and epidemiological studies from 370 countries and used an innovative Bayesian hierarchical computer modeling to create its estimates.

 From the diabetes prevalence estimates from these two sources, the authors then took the mean values for years 2000, 2004 and 2007; and population estimates for 2010; to estimate the rates of diabetes in each country.

In lumping together country-wide estimates of prevalence of diabetes, they failed to adjust for the age of the population, even though type 2 diabetes significantly increases with aging. Longer lived populations with more seniors would be expected to have more cases of type 2 diabetes. Similarly, they didn’t adjust for the countries’ economic status, population genetic homogeneity or diversities, cultural and social differences, and myriad of other confounding factors.

To add to the flaws in their data, out of the hundreds of countries with diabetes data, they cherry picked only 43 countries to use in this study — disregarding about 90% of the evidence. The IDF, for instance, ranked 28 countries with higher diabetes rates than the United States in 2010, but only three of those countries were used in this study.

Finally, diabetes estimates from the two sources were wildly different, varying as much as 83%, and were so dissimilar the authors weren’t able to combine them. Instead, they used each estimate in a separate analysis.

HFCS. The authors used data on food availability per country from the Food and Agricultural Organization (FAO) of the United Nations database from 200 countries. The FAO lumps all sweeteners together, however, and doesn’t isolate HFCS as a unique commodity, the authors noted. They also wrote that data on world HFCS production was limited, so they used two additional sources to estimate the availability of HFCS in each country’s food supply: F.O. Licht’s  International Sugar and Sweetener Report and data on HFCS trade quotas for the EU countries.

Using industry production, food supply data, export data and market/price reports are not measures of actual dietary intakes among people or populations and have repeatedly been shown to be unreliable and overestimate and inflate people’s dietary consumptions. Such industry data also fails to adequately account for waste, losses (during storage, preparation, cooking, plate losses, etc.) and nonfood uses, such as HFCS used in animal feed, leather tanning, gum, cosmetics and medications. FAO supply data also doesn’t include foods produced at home.

Until as recently as the late 1980s, for example, Cuba was the world’s largest sugar exporter and third largest sugar producer, but Cuba wasn’t included in this study. Currently, according to the FO Licht’s report, Brazil dominates the sugar industry, exporting 25 million tons in 2010, over five times more than any other country, and produces most of the world’s sugar. But Brazil wasn’t included in this study, either.

With this many problems in the database they used, they would need to show a mighty large correlation between HFCS and diabetes to rise above random chance, a plethora of statistical or modeling errors and biases, and countless confounding factors. So what did they find?

Results

There was no relationship between the availability of HFCS in the countries and their prevalence of diabetes, regardless of the source of diabetes data. Countries such as India with no HFCS consumption had one of the highest GBMRF diabetes estimates — 26% higher than the United States, which had the most HFCS consumption used in their report. In fact, 17 countries — all with none or a fraction of the HFCS available in the U.S. — all had diabetes rates much higher than the United States (averaging 16% higher) There were also no differences between the countries in total sugars and other sweeteners, cereals, total calories or BMI.

The authors were not able to show that higher levels of HFCS were associated with increased risks for diabetes or that people with diabetes consumed more HFCS.

They decided to lump countries together and compare those with no HFCS availability to those with HFCS in the food supply. Except, they didn’t actually do that, although that’s what was reported. Instead, they drew an arbitrary dividing line between countries with HFCS availability of 0.46 kg/year/capita and those with 0.54 kg/year/capita. Only then could they conclude:

Our analysis revealed that countries electing to use HFCS in their food supply have a diabetes prevalence that is about 20% higher than that in countries that do not use HFCS.

That correlation was not acknowledged as being untenable. Instead, the authors went on for pages discussing the possible negative health effects of HFCS in the Twentieth Century; speculating that HFCS was responsible for raising  healthcare costs the equivalent of $95 million a day in the U.S., and suggesting a “tax” of 10 cents per food or beverage item containing HFCS to offset healthcare costs; and writing of the need to address public health strategies, rewrite national dietary guidelines, mandate food warning labels, and adopt agricultural and trade policies aimed at HFCS “as a means to tackle the increasing global prevalence of diabetes.”

You can decide for yourself if there was a political agenda in their conclusions. But, there was certainly no science.

10 responses to “More epidemiology shenanigans

  1. Feel free to visit my blog where I discuss that sugar tariffs and government intervention got us into this whole mess to begin with. Just assume that HFCS really is the poison everyone is saying it is (which I state in my blog is a pile of horse manure). http://caswansong1.blogspot.com/2012/12/the-path-to-fascism-from-controlling.html “The path to Fascism: from controlling surgar to Socialist Healthcare.”

  2. “media quick to report” – and slow to retract (on p. 32.) There is a recent post by Robert Brown over with Watts “Truth is the Daughter of Time.” Seems generally relevant to the skeptic of maniupulative media pseudo-science.

  3. I could make a far more convincing case that Type 2 Diabetes is caused by a synergism between the use of spoons and the use of chairs.

  4. Well done, Sandy. Pity that the issue is more about control-politics, pricking those evil soft-drink manufacturers (the new oil & tobacco piñatas), and getting one’s next round of federal funding, than about science/truth. This kind of thing could not happen with a properly educated and attentive public.

  5. Snorbert Zangox

    Welcome back Sandy

  6. Thanks Sandy for the comprehensive de-bunking! At least on our breakfast TV news this morning they had a local doctor who said the study was rubbish, although if you only caught the headline you would not have heard this. What has happened to the journals that they publish this stuff?

  7. With trepidation I disagree with the author. First, 20% seems material to me and I don’t understand why 200-300% is the bar that has to be passed to be epidemiologically significant.

    While I have no disagreement that sugar doesn’t cause diabetes, sugar does contribute to obesity and there seems to be a link between obesity and the onset of diabetes.

    Fructose and sucrose may been the same, but fructose is associated with the formation of advanced glycation end products which can result in numerous health issues, including perhaps diabetes.

  8. David, two short books may be of help in understanding tenability versus statistical significance, and correlations versus causation, in epidemiological research. Steve Milloy’s book, Junkscience Judo, is a great resource. Another is Dr. John Brignell’s, The Epidemiologists – Have they got scares for you!

  9. @David — John Brignell’s “Sorry, Wrong Number” is a better book to start with. “The Epidemiologist’s” is a good book, but “Sorry” is a better starting point. The fundamental problem with with epidemiology though is that it is a focus on the dead and not the living. We try desperately to prevent death and forget how many people survive the “causes” the epidemiologist discover. Even the deadliest of discoveries in the epidemiological field, smoking causing lung cancer, is avoided by 92% of smokers.

    The headline though is “Smokers 40 times more likely to get lung cancer than non-smokers”. 40x is 4000%. This is a good reason not to smoke. The 92% of smokers who don’t get lung cancer though should cause all of us to pause and reflect on the danger, especially when it comes to effect in the < 200% category. Often the < 200% cases involve < 10 actual people with the "effect". One study I looked into said "200% increase in risk!". I dug into the report. There were 2 cases in causal side and 1 case on the control.

    The lesson here is that even 4000% increase in risk is not always as clear cut as it would seem. If you start measuring the individuals happiness with their life somehow, that metric becomes completely obscure.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s