Wednesday, April 13, 2005
Does silica in French tap water prevent Alzheimer's?
A French study seems to suggest that higher levels of silica contained in the tap water of some French cities may operate to reduce the incidence of Alzheimer’s disease. This is potentially big news … really big news. But how accurate can this be? Is this another “cold fusion” report? Or a “saccharin causes cancer” study?
As regular readers know, I’m something of a skeptic when it comes to “research” that concludes that statistics “prove” something. Whether scientific research or public opinion polling, most statistical correlations are that and nothing more. Most statistical research tries to make claims way beyond what the data actually shows, and quite frankly, an awful lot of what passes for “research” is sloppy, filled with unproven assumptions, questionable methodology, and conclusions that require a real leap of faith in order to get from the data to the conclusion.
Supposedly scientific research, in particular, is often completely misrepresented, with a mere statistical correlation being pushed as a cause and effect relationship. A philosophy professor who hailed from the Boston area once explained the distinction to me this way: “Every time I lay on my couch and drink Narragansett beer while watching the Red Sox on t.v., they win. But no matter how many times this happens, 100, 1,000, 1,000,000, it’s still nothing but a statistical correlation. It can never establish that my laying on the couch drinking beer causes the Sox to win.”
Much of what passes for “scientific research” falls into just this error, and makes the unsupported leap from statistical relationship to causation. One of the most common ways this error creeps in is the failure to follow subjects over long periods of time, and the failure to account for other possible variables. “Snapshot” research, where a group of people or a particular phenomenon is statistically compared or analyzed at a particular point in time, is inherently unreliable, and the fewer other variables the researchers account for, the more unreliable it becomes. It is like checking the weather this afternoon, and concluding based on that observation that the climate in the area is generally sunny and warm.
This French study, though, on its face seems to be more than the usual slop job pawned off as “research”. First, the study followed the same group of people over a period of several years, apparently ten years or more per participant. It involved about 7,500 subjects, a rather large number for a long term study by today’s standards. And a real effort was made to account for other variables which might effect the data.
And – this is important – the study didn’t start out to find a relationship between silica in drinking water and Alzheimer’s. The principal focus of this long and apparently pretty thorough epidemiological study was apparently osteoporosis – bone loss experienced (most commonly by women) as humans age.
Why is this important? I’m always just a little bit suspicious when researchers “find” what they told the financial sponsors of the study they were looking for!
The Irish Brigade has a discussion of the study, and links to both the Reuters report and the actual published study.
As regular readers know, I’m something of a skeptic when it comes to “research” that concludes that statistics “prove” something. Whether scientific research or public opinion polling, most statistical correlations are that and nothing more. Most statistical research tries to make claims way beyond what the data actually shows, and quite frankly, an awful lot of what passes for “research” is sloppy, filled with unproven assumptions, questionable methodology, and conclusions that require a real leap of faith in order to get from the data to the conclusion.
Supposedly scientific research, in particular, is often completely misrepresented, with a mere statistical correlation being pushed as a cause and effect relationship. A philosophy professor who hailed from the Boston area once explained the distinction to me this way: “Every time I lay on my couch and drink Narragansett beer while watching the Red Sox on t.v., they win. But no matter how many times this happens, 100, 1,000, 1,000,000, it’s still nothing but a statistical correlation. It can never establish that my laying on the couch drinking beer causes the Sox to win.”
Much of what passes for “scientific research” falls into just this error, and makes the unsupported leap from statistical relationship to causation. One of the most common ways this error creeps in is the failure to follow subjects over long periods of time, and the failure to account for other possible variables. “Snapshot” research, where a group of people or a particular phenomenon is statistically compared or analyzed at a particular point in time, is inherently unreliable, and the fewer other variables the researchers account for, the more unreliable it becomes. It is like checking the weather this afternoon, and concluding based on that observation that the climate in the area is generally sunny and warm.
This French study, though, on its face seems to be more than the usual slop job pawned off as “research”. First, the study followed the same group of people over a period of several years, apparently ten years or more per participant. It involved about 7,500 subjects, a rather large number for a long term study by today’s standards. And a real effort was made to account for other variables which might effect the data.
And – this is important – the study didn’t start out to find a relationship between silica in drinking water and Alzheimer’s. The principal focus of this long and apparently pretty thorough epidemiological study was apparently osteoporosis – bone loss experienced (most commonly by women) as humans age.
Why is this important? I’m always just a little bit suspicious when researchers “find” what they told the financial sponsors of the study they were looking for!
The Irish Brigade has a discussion of the study, and links to both the Reuters report and the actual published study.
Subscribe to Posts [Atom]