Monday, October 13, 2008

Another example of making mountains out of pimples

I regularly characterize medical research as making mountains out of pimples so another example of it might help reinforce what I mean by that. I mentioned this study on the 10th so I will not reproduce it again. Instead I want to point out that its conclusions, although correct, were not only trivial but rather misleading.

What was found was that autism was 7 times more common among mathematics students at Cambridge university than it was among students in other fields at that university. But how was that finding arrived at? It was arrived at by taking 378 mathematics students and comparing them with a group of 414 non-mathematics students. And there were 7 autistic people in the mathematics group but only one autistic person in the other group. So the "7 times" conclusion was totally correct. But was it important?

It was in fact totally trivial. What was really found was that the incidence of autism among students at Cambridge was very rare -- even among mathematics students. Putting it another way, it is very rare for mathematics students to be autistic. Only 1.8% of them were autistic in fact. And putting it that way gives roughly the opposite impression to the impression given by the original report. What the original report presented as interesting is in fact of no importance whatsoever.

And that sort of finding is absolutely typical of what is reported in epidemiological research. In fact it is a much stronger finding than is generally reported. Where a 700% difference was reported above, a 50% difference would be about the average in epidemiological research (Note the 60% mentioned in the article immediately below). Most epidemiological research is the modern-day equivalent of the old mediaeval debate about how many angels can dance on the head of a pin: An argument about trivia.

Red wine 'reduces lung cancer risk in smokers'

Interesting that men only were looked at. Does the same hold for women? It should if the causal chain assumed below is accurate. White wine is to some degree a "ladies drink" so maybe it is testosterone differences that caused the small effect described below. Perhaps the red-wine/white-wine difference arises because white-wine consumption is greatest among men who are more limp-wristed. I cheerfully admit, however, that my favourite wine at the moment is a Sauvignon blanc

Drinking red wine could reduce the risk of lung cancer among smokers and ex-smokers, research suggested yesterday. Smokers who drank at least a glass of red wine daily were 60 per cent less likely to develop lung cancer than non-drinkers, a study found. However, white wine did not reduce the risk in the same way.

Studies examining the relationship between lung cancer and alcohol consumption have had mixed results, Dr Chun Chao of Kaiser Permanente Southern California in Pasadena and her colleagues told the journal Cancer Epidemiology, Biomarkers and Prevention. Much of this research has failed to adjust for factors like social status that can influence alcohol use and lung cancer risk.

In the current study, Chao and her colleagues looked at 84,170 men aged 45-69 years old. Between 2000 and 2006, 210 of them developed lung cancer. After accounting for age, education, income, exposure to second-hand smoke, body weight, and other factors, the researchers found that their lung cancer risk steadily decreased with red wine drinking, with a two per cent drop seen with each additional glass of red wine a man drank per month. No other type of alcoholic beverage, including white wine, was associated with lung cancer risk.

For men who were heavy smokers, the reduction in risk was greater, with a 4 per cent lower likelihood of developing lung cancer seen for each glass of red wine consumed per month.

Research has shown that wine drinkers may have healthier lifestyles and tend to have more education and higher income than non-wine drinkers, the researchers said. But the fact that reduced lung cancer risk was seen only with red wine, not white, 'lends support to a causal association for red wine and suggests that compounds that are present at high concentrations in red wine but not in white wine, beer or liquors may be protective against lung carcinogenesis,' Chao and her team say.


No comments: