Eye View 

by David Charbonneau


Science is easier to believe when it coincides with our opinions

 

August 14, 2013



Science is easier to believe when it coincides with our opinions. Since I would like to eat whatever I want without worrying about my weight, I was pleased with recent news. CTV reported: “Study finds moderately overweight people might live longer.”

It seemed like the real thing. In January, the reputable Journal of the American Medical Association (JAMA) published an analysis that showed that fatter people live longer.

But it turned out to be wishful thinking. Walter Willett, chair of the nutrition department at Harvard University’s School of Public Health in Boston was quick to throw cold water on the report. “This study is really a pile of rubbish,” he said unequivocally, “Anyone who believes that is just completely nuts.”

How could JAMA have got it so wrong? As is often the case, the devil is in the details as Bonnie Liebman from the Centre for Science in the Public Interest found out.

The problem was with what JAMA had included, not what they left out. It was a meta-analysis; a review of 97 different studies. Smokers were included but shouldn't have been because they tend to be thin and die younger. Their inclusion, about 20 per cent of the study, makes overweight people look relatively healthy.

JAMA also included sick people who, at the end of their lives, loose a lot of weight. Again, the inclusion of thin, unhealthy people makes fat people appear relatively healthy.






JAMA chose to ignore cause and effect by including such distortions. Thinness, in the case of smokers and the sick, is an effect of illness not a cause of reduced life expectancy. Being thin doesn't cause them to die - - being sick causes them to be thin. And being fat does not result in increased life span.

Despite wishful thinking, excess weight remains risky. Expanding waistlines contribute to diabetes, heart disease, cancer, stroke and death. Cancer as a result of being overweight kills about 9,000 Canadians annually.

Still worse, sensational reports get more attention than reliable ones that report the cold, hard truth. One underreported study by the U.S. National Cancer Institute tracked 1.46 million people over 10 years and clearly linked excess weight and cancer. JAMA was invited to join with their relatively small number of 37,000 participants but they declined.

Interpretation can be tricky. JAMA wasn't factually wrong; it was misleading in what it claimed. The results would have been useful to an insurance company who wanted a statistical connection to weight and death for whole populations.

So, how can you tell if a scientific study is reliable? In meta-analysis studies, make sure the what's included or excluded doesn't distort claims. In experiments, the double-blind model with a randomized control group is the gold standard. The sample size should be large enough to represent whole populations. The experiment should yield the same results each time it is run regardless of who conducts it.

Studies that confirm what we wish were true, without investigation, are self-deception.


David Charbonneau is the owner of Thompson Studio
He can be reached at dcharbonneau13@shaw.ca

 





go back to my Columns in the