Junkfood Science: JFS Exclusive: Part Two of the country’s largest clinical trial on healthy eating

October 16, 2007

JFS Exclusive: Part Two of the country’s largest clinical trial on healthy eating

Last year, the primary outcomes of the Women’s Health Initiative Dietary Modification Trial were reported in the Journal of the American Medical Association. Even though this was one of the largest, longest and most expensive randomized controlled diet clinical trials in the history of our country — and finally tested the claims being made about “healthy” eating that are the foundation of the government’s nutrition and obesity prevention initiatives — few of us heard the results.

In fact, nearly two years later, few people still know what this enormous undertaking found, as revealed in yesterday’s Exclusive Report. Briefly, more than 19,000 women in the intervention group endured intense behavioral interventions to encourage “healthy eating,” and they watched what they ate for more than eight years (reducing dietary total fats and saturated fats, eating 5 or more servings of fruit and vegetables and 6 or more servings of whole grains per day). In the end, none of the expected benefits of “healthy” eating were realized.

There were no significant differences in the incidences of breast cancer, colon cancer, heart attacks or strokes, or weight changes, among those who ate a restrictive “healthy” diet and the control group (29,294 women) who ate whatever they pleased.


The latest findings on 30 cancers

Since then, secondary analyses have been combing through the study data trying to find some correlation between diet and cancers. This week, the first large such report was released in the online issue of Journal of the National Cancer Institute, in advance of print publication. The researchers looked at the incidences of about thirty different invasive cancers, verified by pathology reports:

· Gynecologic (Cervix, Genital organs, Uterus, others)

· Renal/urinary (Kidney, Bladder, Urinary organs)

· Digestive (Oropharynx/esophagus, Stomach, Biliary tract, Pancreas, Liver, Other)

· Respiratory (Lung, Other)

· Brain/nervous system (Brain, Nervous system)

· Bones/connective tissue (Bones, joints, cartilage, Connective/soft tissue)

· Multiple myeloma (Blood/lymphatic, Hodgkin disease, Non-Hodgkin lymphoma, Leukemia, Lymph nodes)

· Endocrine (Thyroid, Other)

· Other sites (Melanoma of skin, Other/unknown sites)

They statistically compared the number of cases among the group that ate the “healthy” diet to the number in the control group. As you’ll remember, a relative risk of 1.0 means there is the same likelihood of something occurring in one group as another (fractions: 2/2 =1) and it’s a null finding. For relative risks to be greater than what would happen by chance and random fluke, they need to deviate significantly from 1.0 — at least 3 (200%). These authors then reported the correlations as hazard ratios (odds ratios) which can exaggerate relative risks, meaning they need to be viewed especially cautiously.

What did the data show?

There were no statistically significant differences for any of the cancers. Eating “healthy” versus eating whatever they chose made no tenable difference in any of the cancers.


Splitting hairs 101

We can go down the list of the stats for all of the cancers and see that the hazard ratios are all hugging 1.0 — with insignificant variations above and below 1.0.

Of course, if someone wanted to scare people, they’d point out the tiny deviations above 1.0 and if they wanted to convince people to eat low fat, they’d point out the tiny deviations below 1.0. While such hair-splitting pursuits are popular (the media loves them), they are not credible. And they shouldn’t worry us because no matter how serious relative risks may sound, that still doesn’t necessarily make them clinically significant and above random chance. Don't let numbers be used to spook you.

So, for example, the low-fat “healthy” diet was associated with a 4% higher risk for melanoma, 8% higher risk for uterine cancer, 10% higher risk for stomach cancer, 20% higher risk for genital organ cancers and brain cancers, 49% increase in lymph node cancers, 96% increase risk for biliary tract cancers and 130% increase in liver cancer. In contrast, it was associated with an 7% lower risk for esophagus cancer, 8% lower risk for lung and lymphatic cancers, 14% lower risk for bone cancer, 22% lower risk for kidney cancer, 25% lower risk of pancreas cancer, and 54% lower risk for cervical cancer. But don’t be fooled: All are ascribed to chance.

As the authors explained about even the greatest hazard ratio deviations in this study:

The results... for Hodgkin disease (P=.05, HR [hazard ratio] = 0.19), biliary tract cancer (P=.20, HR = 1.96), and liver cancer (P=.31, HR = 2.30) can readily be attributed to chance. They arise in the context of approximately 25 comparisons, each based on a small number of disease events.

It is important to remember that relative risks or odds are not the same as actual incidences of disease; so, that 130% higher risk of liver cancer obviously does not mean that 130% of women got that cancer. In actuality, the total incidences of liver cancer was 0.0368%, with only 3 actual cases difference between the 48,835 women. As the researchers noted, it's not statistically greater than a random toss of the dice.

Diet simply was not shown to be a valid factor in causing or preventing any of the invasive cancers.


Reassuring evidence

These findings could, and hopefully will, relieve guilt and anxiety for hundreds of millions of women worried that what they eat could be to blame for their cancers or their chances for survival. Similarly, a recent clinical trial showed that eating gobs of fruits and vegetables doesn’t affect cancer recurrence. All foods can be enjoyed and shared as pleasureable nourishment of body and soul, rather than toxins or cures.

The WHI trial data showed a few other points that may help women feel less worried. Other things that made no difference in risks for the invasive cancers between the dieters and control group included: diabetes; dietary fats, grains, or fruits and vegetables intake (in separate analyses); physical activity (from less than 1.5 to more than 14.8 METS/week); and BMI. Yes, despite all of the fears about body weights, incidences of invasive cancers were not significantly different among women who were fat (in the “overweight,” “obese” and even “morbidly obese” medical categories).


In the news

So, how did the media report these findings?

These were typical of the news stories this week: Reuters Health headlined with “Low-fat diet may lower ovarian cancer risk” and CNN headlined with “Low-fat diet may cut ovarian cancer risk.” CNN reported:

T]his week, researchers unveiled the first hard evidence that switching to a low-fat diet late in life can lower the odds of ovarian cancer... It's arguably the most promising finding of the mammoth Women's Health Initiative dietary study.... “It's quite noteworthy,” [co-author Dr. Ross] Prentice said of the ovarian protection. “We're really pleased to have something positive to say to American women — that undertaking a low-fat diet likely reduces your risk of ovarian cancer and perhaps other cancers as well.”

That doesn’t even remotely sound like the same study, does it? They sure left out a lot.

Where did the media come up with this take? Regular readers don’t even need to ask. In advance of the print publication, a memo to the media went out from the publisher titled “Low-Fat Diet Possibly Linked to Lower Risk of Ovarian Cancer.” And a similar press release was issued from the NIH/National Heart, Lung and Blood Institute (which funded the study) titled: “Low-fat Dietary Pattern May Lower Risk Of Ovarian Cancer.” The media reported just that.

Let’s look at what the study actually found about ovarian cancer. The ovarian cancer hazard ratio associated with the dieters was just to one side of 1.0 — 17% lower — but, as the study authors reported in their published paper, this “was not statistically significantly less than 1.0.”

To put this calculated odds ratio into context, that is 17% of the actual incidence of ovarian cancer among women. And the actual incidence is 0.0129% according to the Centers for Disease Control and Prevention, U.S. Cancer Statistics.

Where the spin went from there is an example of fishing, where researchers dredge data and perform multiple analyses and permutations trying to find some relationship to hang their hat on. “The more analyses conducted, the more likely that some findings will be detected by chance and do not actually exist,” wrote Dominic Carone, Ph.D., neuropsychologist and assistant professor at SUNY Upstate Medical University, New York. In this case, they divided hazard ratios by years and various groupings and were able to find that the risks for ovarian cancer were 16% higher among the dieters in the first four years of the trial, but 40% lower in the last 4 years, leading to their statement that “the hazard ratio decreased with increasing intervention duration...[and] a low-fat dietary pattern may reduce the incidence of ovarian cancer.”

In reality, the difference in incidences of ovarian cancer was a mere 0.07 case per 1,000 women. In fact, the difference between the “healthy dieters” and ad lib group for all of the invasive cancers amounted to only 0.01 case per 1,000 women. As the authors had noted, “not statistically significant.” Not even close.


© 2007 Sandy Szwarc

Bookmark and Share