Sunday, April 08, 2007

Improving Medical Research

I recently read an article called Office workers risk blood clots in BBC News, and the thought crossed my mind that this is an excellent reason to be a smoker: you get to get up every hour or so and take a little walk, stretch your legs, reduce the risk of deep vein thrombosis. You also rest your eyes, have a mental health break, get to know your coworkers... it seems likely that smokers are happier, more sociable, more laid-back workers with lower rates of certain health conditions. There are people who smoke moderately (4 or 5 cigarettes a day) who might not even be hurting their health. But of course medical stories will never mention such things because they don't want to promote smoking.

But if medical research isn't objective and scientific, then what is it? Just a mouthpiece for things we already believe? It seems too often to be just that. At the worst extreme, organizations fund research that confirms their viewpoint. Hence, all those stories that red wine is good for you are funded by the French government; the Dutch are behind the pro-chocolate studies; and we all know about drug research that doesn't turn up problems with new drugs.

Even at its best, medical research is skewed too much by the beliefs of researchers. Here's another story from recent press: Soft drinks clearly associated with diabetes - report (from the New Zealand Herald). This article starts, "A review of published studies shows a clear and consistent relationship between drinking sugary (non-diet) soft drinks and poor nutrition, increased risk for obesity -- and increased risk for diabetes. There is no denying that sugar-loaded soft drinks are having "a negative impact on health," Dr Kelly Brownell, director of the Rudd Centre for Food Policy and Obesity at Yale University in New Haven, Connecticut, said in a telephone interview."

Actually, there is a lot of denying it. A study that finds that people who drink soda pop are more likely to have health problems does not tell us at all that soda pop causes health problems. In fact, the likely culprit is that people who tend to drink soda also tend to have other habits that lead to health problems. Like they're more likely to eat meals at fast food restaurants, or they drink less milk, or they're less likely to be interested in diet. Or maybe it isn't the sugar intake that causes the diabetes but diabetes that is causing the sugar craving. Or another condition that causes both - perhaps even a mental condition like depression or plain old teenage angst (overeating as a slow form of suicide).

But okay, okay, it's pretty likely that drinking large quantities of sugary pop (which is high in sugar) is going to have an effect on your blood sugar, and because it's high calorie it's probably connected to weight gain, but did the study prove these connections? Not at all. This study didn't even bother to measure how much sugary pop was drunk, so there's no indication that it was consumed in harmful quantities. How are we going to learn anything new if we just keep "proving" what we already believe to be true?

(By the way, I neither smoke nor drink sugary soda pop, so please don't take this post as an argument for either of them.)

Along with the problems of the way the data is analysed, there are problems with the way it's collected. Why do we have thousands upon thousands of short-term, small studies? Why doesn't our government take most of its funding for medical research and put it into one huge long-term study? It could then make the information available to everyone in a huge database - as it does with economic data.

Broader data collection might reduce some of the biases that occur in research design. Factors that were considered irrelevant might emerge as worth looking into. Where possible, throw in DNA, family history, information about every place the individual lived, and a lifetime of medical data.

For specific issues, researchers could prepare add-on studies. The goal would always be to have as large a sample size, and long a time frame, as possible, so data collection would often go on for longer than the researcher's own time frame.

Allowing many people access to the data would lead to much better, more objective analysis. It would also allow for much better reviews.

* The Scandal of Poor Medical Research
* 500,000 people, a span of decades - and a waste of time and money?


1 comment:

Graeme said...

Statistics are fun. You can use them creatively to prove or disprove whatever you want, and the vast majority of the population doesn't know enough about statistical analysis (myself include, for the most part) to be able to argue with you.

I've read that more black people are executed in the USA than white people. What does this tell you? Is it because black people are criminals (black people commit more crimes for which the death penalty applies), or is it because white people are racist (juries are more likely to execute a black person than a white person convicted of the same crime, or black people are more likely to be arrested for or convicted of crimes involving the death penalty)? More than likely, some combination of the two, but in what proportion? The statistics themselves tell you nothing.

As you said, proving a connection is one thing, but proving causality is something entirely different.

It never occurred to me that there could be any advantages to being a smoker, but you've mentioned a couple of interesting points. I still think they pale in comparison to the damage you're doing to your body, and I always laugh when I see smokers huddling outside when it's -20 degrees. I wonder how many people are "summer smokers" because they refuse to stand outside when it's freezing.