Nog onlangs werd in het Gezondheidsraadrapport het EPA-onderzoek uit 1992 aangehaald als één van de belangrijkste onderzoeken die hun conclusie (dat meeroken kanker veroorzaakt) ondersteunt.
Dit onderzoek wordt op onze site al eerder sterk gerelativeerd, maar recent kwam er weer, in een forum, een discussie op gang over dat onderzoek. Wij publiceren de zwakheden van dat onderzoek, zoals in dit forum opgesomd, hier nog maar eens omdat het een prachtig overzicht geeft van het statistische gesjoemel dat bij dit onderzoek werd toegepast.
Overigens verwees ook een Amerikaanse rechter in 1998 dit onderzoek al voor een groot deel naar de prullenbak…
Other views: Who really is fooling the public?
By John Rowell, The Forum
Published Sunday, March 21, 2004
What happens when bad science is used to promote what many think is a good policy? What happens when scientists stop being skeptics, stop demanding rigorous adherence to the scientific method, stop thinking for themselves? For one thing, we see (Forum, March 13) an attack on me for “deceiving the public” about the risks of secondhand smoke.
Dr. Donald Miller writes, correctly, that the 1992 EPA report held that exposure to secondhand smoke causes lung cancer in non-smokers — by the slimmest of statistical margins. But neither he, nor The Forum’s editorial writers, nor (I daresay) the general public is at all curious about how the EPA reached its finding.
First, the EPA committed itself publicly to a conclusion before its research had even begun. Second, the EPA selected 11 studies which form the basis of the 1992 report, including one (Fontham) for which interim results were then available. However, it excluded two others (Stockwell and Brownson) for which such results were likewise available.
Now, why would that be? Perhaps because the latter two studies showed no statistically significant risk due to workplace exposure to smoke? Had the two studies been included in the meta-analysis, the EPA’s conclusion could not have been the conclusion announced before the work began.
Third, the EPA changed its normal methodology in order to find a statistically significant association. Normally, EPA would use a 95 percent confidence level, but since that would not have sustained its controversial hypothesis, it lowered the confidence level to 90 percent. Why? The EPA does not explain why.
Fourth, the EPA corrected upward the risk of exposure to secondhand smoke “based on calculations in which unknown parameters are replaced by numerical estimates that are subject to uncertainty.” Does this sound even remotely like good science?
An article published in Science on July 31, 1992, perhaps elucidates what’s going on here. An EPA scientist who wrote part of the draft report admitted that “she and her colleagues engaged in some fancy statistical footwork.”
Flawed as it is, the EPA report states a relative risk of 1.19 for exposure to secondhand smoke — a 19 percent increased risk. But Dr. Miller devotes an entire paragraph to discussing a relative risk of 2.0 — a 100 percent increased risk — for such exposure. Now why would that be? Do his words aid our understanding of the issue, or do they mislead us?
Dr. Miller directs my attention to “the new evidence published in scientific journals since 1992.” Naturally, he won’t get specific on what I’m to look at. The studies since 1992 which are familiar to me (Steenland, Zaridze, Blettner, Boffetta, Enstrom and the World Health Organization) don’t support the assertion that exposure to secondhand smoke causes disease in nonsmokers.
In spite of all this, I’m the one charged with “deceiving the public.” Now why would that be?
Rowell is a member of the Moorhead City Council.