Lies, damned lies and statistics

[JonStamford]

In the field of Parkinson's, much is made of epidemiological studies linking pesticides, well water, and so forth with a greater likelihood of developing Parkinson's. As with all epidemiological studies, the link between recorded phenomena (a higher prevalence of Parkinson's in areas where there is greater exposure to pesticides) may seem blindingly obvious. But epidemiology is coy on the subject. All epidemiology postulates is that there is a link. It does not suggest in which direction that link may manifest itself.

Let me play devil's advocate for a moment. Although we would extrapolate the data to indicate greater pesticide exposure leads to a greater likelihood of Parkinson's, this cannot be strictly drawn from the data. The data could equally be taken to mean that some aspect of pesticide exposure perhaps attracts people with Parkinson's. Greater pesticide exposure suggests a rural community. Such a rural community might have a skewed age distribution. Young people might find such a community unattractive while older individuals might be drawn to the rural life. As Parkinson's is generally a condition of older people, a higher number of people with Parkinson's might be expected.

You see what I mean? Things are rarely as simple as they seem. Of course the best epidemiological studies take things like age distribution into account. But even these are subject to challenge. Richard Doll, the famous epidemiologist, knew more about life's risks and perils than most and was among the first to report on the association between lung cancer and smoking, reporting a higher proportion of smokers among lung cancer victims than among the unaffected. Still, for many years, even these findings were challenged, often using similar lines of reasoning.

A recent study from British Columbia via Denmark (sjweh.fi/show_abstract.php?... suggests that an agricultural lifestyle rather than specific exposure to pesticides may be causal. Whether or not that proves to be the case, it highlights the pitfalls in epidemiology. Statistics are wonderful servants but poor masters.

Doll once ended a lecture on life's risks by saying that the first 10 minutes of life were the most dangerous. A voice from the back said "what about the last 10 minutes?"