Junkfood Science: <i>Special science exclusive:</i> Have your steak and enjoy it too!

November 22, 2006

Special science exclusive: Have your steak and enjoy it too!

When it comes to food and health news, the saturation of media coverage is usually inversely proportional to the soundness of the research. Sensational claims sell a lot of newspapers, lure viewers and listeners, and create good buzz. The trouble is, most of us don’t realize we’re getting Hitchcock-like fiction when we turn on the news. With last week’s scare du juor — “red meat increases risk for breast cancer” — no mainstream news reporter took a critical look at the study they were reporting, preferring to simply pass along the press release. No one except Steve Milloy, of course, who wrote an excellent review here.

Our first clue that there was something more afoot in this media blitz than brilliant science, is that this one study — out of the thousands of new studies released every single day — was reported in every media outlet, on the exact same day, all saying the exact same thing. This is evidence of brilliant marketing, but not much more. Like everything in media today, it deserves viewing with the same skepticism as you would any other commercial. The study wasn’t nearly as well done as its marketing...


This was another study by researchers from Brigham and Women's Hospital and Harvard Medical School mining a database, called the Nurses Health Study, compiled from answers to lifestyle questionnaires that have been collected from around 120,000 nurses since 1976. This database has been the source for more than 500 “studies” looking for correlations on everything under the sun. The findings often disagree, depending on what the researchers want to find. It’s the source for many of those media reports of scientists finding something one week and something entirely different the next.

Food frequency questionnaires sent to 90,659 premenopausal women in 1991, 1995 and 1999 were used for this study. These asked the nurses how many times they had eaten 130 different food items during the past year, and each food item had 9 possible responses, ranging from never to 6 or more times a day. It’s unimaginable that anyone has the memory, time or patience to complete such a survey, which makes the accuracy of the information going in precarious at best. Even more telling of the dubious quality of the data, the women didn’t give any information about how much they ate because they weren’t asked amounts or portion sizes. So the researchers guestimated.

The researchers then obtained self-reported cases of breast cancer, a total of 1021 cases, which were confirmed using pathology reports.

Here’s where the fun begins. The researchers put all this into their computer, created a computer model that would pull select information and put it together how they chose, and then waited....

They were unable to find any measurable correlation between red meat intake and premenopausal breast cancers. That should have ended it right there. But didn’t.

What the raw data showed was that only about 9% of the young women ate more than 11/2 servings of meat a day. Most ate less than 5 servings a week. But the highest meat eaters were also more likely to smoke, nearly two times more likely to have more than three children, and slightly less likely to be taking oral contraceptives or have a history of benign breast disease. So, the researchers made statistical adjustments to their computer model to attempt to control for these and other possible confounding factors. No one knows if they adjusted correctly or if they included all of the other possible things that might have a role in these cancers.

Despite the nonfinding, they decided to dig deeper looking for correlations among hormone receptor status of the cancers. Except they didn’t have any information on the hormone status for nearly 25% of the cancers — with more of the data missing among the women eating less than 5 servings a week than among women eating the most. Meaning, they weren’t looking at data representative of most premenopausal women.

After all that, they arrived at a finding that women eating the most red meat had a 97% greater chance of getting hormone receptor positive breast cancers than those eating the least.... Except not all red meat, only if it wasn’t eaten in a sandwich or as hotdogs and bacon — they didn’t even try to give a biological explanation for that!

Now a 97% increase sounds like a horrific amount, doesn’t it? But that refers to “relative risk,” not actual risk. Only 0.6% of the women eating more than 11/2 servings a day developed hormone receptor positive cancer as compared to a 0.45% incidence among the women eating meat 3 or less times a week.That is a difference in actual risk of only 0.15% over twelve years. This is such a small number as to be unreplicable. It could just as easily have been chance or a tiny mathematical problem in their computer model.

[And the incidence among most of the women — including those eating more than a serving of meat a day! — was only 0.5%, making any real difference even less relevant.]


Reporting the actual numbers, of course, wouldn’t have created nearly as dramatic of headlines.

Accompanying this story were scares of rising rates of hormone receptive breast cancers. The truth, according to the American Cancer Society and National Cancer Institute, however, is that incidences of breast cancer in women have been stable since around 2000 and any perceived increases have been because of increased use of mammography and hence increased diagnosis.

But let’s stop right here. You’re probably realizing that playing with statistics is a mighty problematic venture. And you would be right. These types of studies are the most susceptible to errors and bias. That’s why in these data dredges to find correlations, relative risks have to be particularly significant to even be deemed tenable. They also have to be well above what one would find by chance — more than the toss of the dice.

So scientists do not take as credible anything which isn’t at least 2 to 3 times over baseline, or a 100% to 200% difference. Ernst Wynder, MD, founder and director of the American Health Foundation and editor of Preventative Medicine prior to his death, said anything less than 200% is suspect. And Marcia Angell, MD, former editor-in-chief of the New England Journal of Medicine said they looked for 200% or more before accepting a study for publication. Yet, we frequently hear about studies finding risks of 30% or 50% reported as if that means anything. So, 97% might be considered statistically significant to these researchers, but they wouldn’t be taken as worthy of note by prudent scientists or meaningful in real life at all.

These types of studies, looking for links among large groups of people, were never meant to be used as they are today. They don’t apply to individuals and most importantly, they can never show a cause for anything because they don’t actually test anything. These studies are called epidemiological studies because they were originally meant to be used as the first step in narrowing down potential factors in infectious diseases, which would then be tested in a series of clinical studies to confirm the origin of the disease.

What that means is, if a well designed population study can’t even find a correlation, then good scientists move on and look somewhere else for the cause. So those studies finding no evidence of a significant correlation are especially valuable in today’s scare-driven climate. Yet the media failed to give us the full story and reveal that multiple population studies — even using data on far larger groups of women than this and even studies done by these same researchers — have found no evidence for a link between breast cancer and red meat. For example, a study of 351,041 women published in the International Journal of Epidemiology in February 2002 and a study of 85,000 women ages 30 - 55 published in the International Journal of Cancer in January 2003. Nor is there a body of clinical evidence to support causation.

It is the most basic fundamental principle of research that correlations do not make for a cause, no matter how strong a correlation might be. Wearing a bra has been associated with a 12,500 times greater risk for breast cancer, not because bras cause cancer, but because most breast cancer is found in women.

So while Dr. Nancy Snyderman told NBC’s Today Show viewers on November 14th that we now know that there is a cause and effect between eating red meat and breast cancer....we know nothing of the sort.

© Szwarc 2006

Bookmark and Share