Georg Ivanovas From Autism to Humanism - systems theory in medicine
2.5 The limits of evidence based medicine
Statistics produce data, correlations, but never causal relations. In economy the reciprocal relation between skirt length and economy growth (the hemline index) is so well established that quite a lot of economists believe in it (Schultze 2008).
The standard example in statistic literature is the correlation between the storks and births. Here we see that the decline of storks in the years 1972-1985 in Lower Saxony correlated with the decline of births in this period of time (Schwarze 2001: 18). If we still believe the model of the stork bringing the babies, the meaning of this statistic is obvious to us. (1)
population of storks and births in Lower Saxony
consumption of potatoes (kg per capit) and consumption of electric energy (bill.kWh)
Not so obvious but more scientific, as the parameters are specified in a metric system, is the correlation between the consumption of potatoes (in kg per person) and the consumption of electric energy (in bill. kWh). This correlation (Schwarze 2001: 18) is so accurate that it can’t be accidental. But again we are not inclined to see this correlation as causal. We will use other mechanisms to explain it. Or take this: Turkish children in Germany suffer about half as often from asthma than their German counterparts. However, the more the Turkish families adopt a German life style the more the incidence approximates the German figures. The best correlation is the language spoken. The more the Turkish family speaks German the higher the incidence of asthma (Swaf 2006).
In this category might also fall the correlation between fetal femur size and blood pressure at the age of 6 (Blake et al 2002) or: “Women who change partner between their first two births are at an increased risk of delivering a preterm, low birth weight baby with an increased risk of infant mortality compared with women who have the same partner for both births” (Vatten/Skjærven 2003). When left-handers face a greater cancer risk theories are developed that some hormone-like chemicals during pregnancy provoke both: the status of a left-hander and the risk becoming cancer (Khamsi 2005).
And what are we doing with the Caerphilly study (Ebrahim et al 2003) that showed that men shaving less than once a day had an increased incidence of stroke (70%) and generally a higher mortality (30%). The authors did not succeed in finding other factors to explain this correlation.
As the Caerphilly study was a large study (2.438 men observed over 20 years) we have to take the results seriously. What should be the consequences of this finding? According to the usual procedure shaving machines should be provided by the national health services of each country. Education campaigns should be initiated promoting the benefitial effects of shaving. It should be thought of and tested in small studies whether or not depilatory creams are useful, as it might be assumed that the beard creates health problems.
What sounds at first as satirical is nothing than the real satire of every day medicine. Normally we do not realize that the therapeutic strategies taken on the ground of statistics are on the same level as the strategies presented here for the Caerphilly study. As the subject is mostly diabetes or lupus erythematosous it is not so striking.
Men who experienced the Leningrad siege have higher systolic and diastolic blood pressure and excess mortality from ischaemic heart disease and stroke. This was attributed to starvation (Sparen et al 2004). There were, however, quite a lot of other factors during the siege of Leningrad that might have affected health and one of the authors stated in the following discussion more precisely that the problem was the “trauma of the siege” (Vågerö et al 2004).
Television is associated with obesity in childhood (Stettler et a 2004). Nobody would believe that there is a causal relationship. But when childhood exposure to cigarette smoke is associated with a higher risk of back pain in later life, the authors see a causal effect of the smoke on the developing spine (Eriksen 2004).
These are only some of the dubious causalisations the information services present us every day. But also all other causalisation happens at one’s own discretion. Evidence is always an ‘evidence of obviousness’. It produces what an external frame, a legend (chap. 3.8) or a paradigm (chap. 3.10) allows to. A statistic is just able to make a hypothesis more or less plausible. Nothing more. To believe that statistics produce knowledge is like hunting Easter eggs oneself has hidden (Simon 1993: 31).
Sometimes we have the opposite: After the first meta-analysis for homeopathic treatments showed its effectiveness (Linde et al 1997) there were statements like: “More importantly yet, if the basic sciences gave us very strong reason to believe that a drug would not be effective, then it is appropriate to be very cautious when interpreting apparently positive clinical results” (Sehon/Stanley 2003).
In fact, we should always be cautious when interpreting apparently positive clinical results. And we should be aware that all interpretation of statistics is arbitrary and “an open debate” (McCromack/Greenhalgh, 2000; Coradi/Taylor, 2003).
(1) Illustrations with courtesy by Verlag Neue Wirtschafts-Briefe, Herne, Berlin