Monday, March 31, 2008


The post below is an excerpt from Sandy Szwarc. I myself made a derisive comment on the study that Sandy debunks on 11th. Sandy has however taken the trouble of looking at the study closely -- and finds that it is an even bigger fraud than I suspected. So this is yet another instance of the failure of peer-review in the madhouse that is the medical literature.

I see that Sandy has a critique coming up of the study about pot bellies that I dismissed on 28th. as "utter crap". Should be fun to see what she says! H/T to Ecomyths for drawing my attention to the article below. I should read Sandy regularly.

This new study was reported by the media as having found “conclusive evidence” that eating sugary cereals, white bread and other carbohydrates with high glycemic index increase risks of cancer and other modern “lifestyle diseases.”

This paper, published in the American Journal of Clinical Nutrition, was a meta-analysis of 37 published observational (epidemiological) studies that had looked for correlations between glycemic index/load and diseases. No cohort or clinical intervention trials were included.

The studies that the Australian authors compiled had differing designs: different age subjects; different study durations; different techniques to identify diseases (self-reported, chart reviews and only 4 involved actual physical exams); and had adjusted for different confounding factors known to contribute to disease. “Most of the studies used basic models, controlling for age and sex only,” however, the authors admitted.The studies used a food frequency questionnaire at enrollment to calculate GI/GL, but all used different units to report levels, so the authors converted the results to a single scale.

The authors noted that a significant limitation of their meta-analysis was that none of the studies they included had “actually validated the assessment of GI or GL using another dietary method or against an objective standard. The assignment of GI values to foods in a nutrient database is to some extent subjective,” they wrote, “and may be unreliable when extrapolating from one country to another.” They then separated out ten different diseases (type 2 diabetes, heart disease, stroke, breast cancer, colorectal cancer, pancreatic cancer, endometrial cancer, ovarian cancer, gastric cancer and gallbladder disease) and compiled the different risk ratios associated with GI/GL reported in each study.

This paper epitomized the weaknesses of meta-analyses, which lump together diverse and weak or null studies hoping to derive a significant effect. These problems make it imperative that any correlations that emerge from such analyses are tenable and clear; and beyond random chance, statistical modeling errors and confounding factors.

To cut to the chase, try as they might, they were unable to find a viable correlation between GI/GL and any disease. None of the 37 studies, separately or lumped together, could come up with a tenable link. The relative risks for all of the diseases hugged either side of null (relative risk 1/1 = 1):

The researchers, however, arrived at a different conclusion:
"This meta-analysis provides high-level evidence that diets with a high GI, high GL, or both, independently of known confounders, including fiber intake, increase the risk of chronic lifestyle-related diseases. The effect was modest overall... Overall, the GI had a more powerful effect than did the GL... The findings indicate that the judicious choice of low-GI foods offers a similar or higher level of protection as whole-grain foods or high fiber intake in the prevention of chronic lifestyle-related disease"

Are we being sold something? When a study’s conclusions contrast so dramatically with the actual data, it is helpful to ask why.Not mentioned by any news reporter was that, except for the statisticians, the researchers who conducted this study are authors of GI diet books. Not only that, but they are with a GI testing service and a GI-based licensing program in Australia. Through the GI Symbol Program run out of the University of Sydney, food companies purchase license agreements with Glycemic Index Ltd. to put its GI symbol on their products. It’s a marketing gimmick, a “seal of approval” to identify foods for shoppers that are the “healthier choices.” The study’s lead author, Alan Barclay, is the founding director of GI Ltd., the company behind Australia’s GI Symbol Program, which is under the direction of another of the study authors, Jennie Brand-Miller....

With so much money and careers at stake in finding evidence to support low-GI foods, and in worrying people that high-GI foods could cause diseases and health problems, it might help explain why this study’s null findings were interpreted positively. Even the confirmation bias of well-meaning professionals can lead them to see proof of what they believe when the evidence disproves it.It’s not science itself that we can’t trust, it’s the bad portrayals of science.

Seeing for ourselves the actual null and negative findings, as was evident in all 37 studies on nearly 2 million people used in this analysis, offers us the most valuable information. If a significant correlation had been found and had biological plausability, a possible cause might then be tested by conducting randomized, controlled intervention clinical trials. If an epidemiological study can find no significant correlation, no link, between a food or ingredient and a disease, then there is nothing to base a suggestion for a causal link. It's a nonissue.

Null findings testing an hypothesis are important, as they enable science, and us, to move on, rather than continue to beat a dead horse or needlessly worry. This is the source of Albert Einstein’s famous saying: “No amount of experimentation can ever prove me right; a single experiment can prove me wrong.”

Study #2: No support for sugar fears

The second study was of no interest to PR agencies, and hence, never made the news. This epidemiological study, was published in the International Journal of Cancer. Its data was from 120,852 people in the Netherlands Cohort Study, who’d also completed food frequency questionnaires in 1986. After 11.3 years of follow-up, the researchers found no genuine relative risk for colorectal cancers related to dietary glycemic index or glycemic load among the men or women. They also found no clear associations between any cancer subsite and dietary factors. Relative risks for cancers associated with sugary, high-glycemic foods ranged from 0.83 to 1.20. The relative risks all hugged either side of null (RR=1) with none tenable, like the other observational studies.

Unlike the previous study, however, these researchers reported their findings factually: “Overall, our findings do not support the hypothesis that a diet with a high glycemic load or index is associated with a higher risk of colorectal cancer.” And conversely, low-glycemic foods aren’t associated with lower risks. There is no link. As hard as some have tried to prove that their believed way of eating is the only right way and can lead us all to optimum health and wellness, while preventing cancer and chronic diseases of aging; the facts continue to show that well-fed people around the world have hugely varied diets with no consistent correlations with health or longevity. More important, efforts to scare us about foods some don’t believe we should be eating and to prove that ‘bad’ foods can give us cancer or some other frightening disease have no credible evidence, either.

No comments: