As an engineer, I am used to browsing abstracts. PubMed is awfully weak. I searched for secondhand smoke and got 98 hits (seems to be a little light. The most salient articles do not link to anything, even an abstract, though presumably if you owned the journals in question, or had access to a medical library, you could read it.
But the first one I ran into that had an abstract read
Quote:
In 39 children with Legg-Perthes disease who were nonsmokers, the specific aim was to assess relationships among parental cigarette smoking during pregnancy, household smoking before diagnosis of Legg-Perthes disease, hypofibrinolysis, and thrombophilia. Fifteen (38%) children had no secondhand smoke exposure; 24 (62%) had secondhand smoke exposure before their diagnosis. Seventeen (71%) of these 24 children were exposed while in utero to smoking by a parent or live in relative and also had exposure to household smoke during childhood; seven (29%) had only household smoke exposure in childhood. In the full cohort of 39 children, secondhand smoke exposure correlated inversely with the major stimulator of fibrinolysis, stimulated tissue plasminogen activator activity. Of the children exposed to smoking, 48% had low stimulated tissue plasminogen activator activity (< 2.19 IU/ml) compared with 7% of the children without secondhand smoke exposure and 14% of 22 healthy control children. Secondhand smoke exposure had no significant effects on other measures of coagulation. Secondhand smoke exposure while in utero and during childhood appears to lower stimulated tissue plasminogen activator activity and additionally may depress heritable low stimulated tissue plasminogen activator activity, leading to hypofibrinolysis. Hypofibrinolysis may facilitate thrombotic venous occlusion in the head of the femur, leading to venous hypertension and hypoxic bone death, Legg-Perthes disease.
PMID: 9678044 [PubMed - indexed for MEDLINE]
|
OK, now this is certainly not the cream of the crop of studies, but 'c'mon. 39 kids with a given condition are reviewed for exposure to household smoke, and no mention is made of attempts to control for other factors. It is also an incredibly small group, so random fluctuations will be much harder to isolate.
"Of the (24) children exposed to smoking, 48% had low stimulated tissue plasminogen..." As I read that, 11 in a smoking environment had low levels, 13 did not. Harly a clear-cut cause, though may be a contributory factor. The sample size is much to small to tell.
Furthermore, this is just another example of a backwards study -- pick a group with some disorder, then try to force some correlation onto it, with no a priori reason for suspecting any given factor. In my first college statistics class, we were shown a chart correlating the birth rate in post-war London with pig iron production in the US. Surely no one would claim that though it was a correlation of .99, that the US could decrease London's birth rate by decreasing our steel production. Yet "studies" like this get little to no criticism when the results kind of agree with what the chosen would like to see.
I scanned a few other articles, and lo and behold, saw at least one major flaw in the application of statistics in each one.
What I am saying is that very few people have much in the way of statistical background. I have several years of post-doc training in statistics, so they tend to stand out like a sore thumb. But for anyone without a firm statistical background, including most professionals, those errors need to be pointed out by someone who does understand the limits of statistics. Which is why books criticizing papers are generally more revealing than the papers themselves...