Passive Smoking: Glossary

Achtung! If you're going to read this section (and, of course, you should), you will find it much easier as a PDF. This is the complete Appendix of Velvet Glove, Iron Fist. View it here.


Passive Smoking: Introduction


The Evidence



Glossary of terms


Case-control study

Epidemiological research which surveys a sample group (the cases) and compares them to a healthy sample group (the controls). In this instance, the cases are lung cancer sufferers. When the subject is no longer alive to answer questions, family members may be asked to provide information. Also known as a retrospective study.


Cohort study

Epidemiological research which monitors a group of people (the cohort) over a period of time, usually several years. Typically, they are surveyed at the outset of the study, at its end and sometimes in between. In the field of secondhand smoke research, the cohort are healthy individuals at the outset and are monitored for lung cancer over time. Because they are healthy to begin with, the possibility of recall bias is eliminated. For this, and other reasons, cohort studies are considered of greater value than case-control studies. Also known as a prospective study.


Confidence interval

The margin of error between the lower and upper end of the relative risk. Shown in brackets after the relative risk.


Confounding factor

Any external factor that may influence the results of a study. In the field of lung cancer research, these particularly include age, occupation, asbestos exposure, radiation, diet and smoking history.


Date-dredging

A term used to describe the practice of asking subjects about a vast number of lifestyle factors in the hope of finding an association with one or more of them.


Dose-response relationship

If A leads to B, any increase in A should lead to an increase in B. In the case of active smoking, there has been proven that heavier smokers have a greater risk than lighter smokers and that risk increases according to the number of years spent smoking. If secondhand smoke poses a risk to nonsmokers, one would therefore also expect to see a linear relationship between lung cancer risk and length and intensity of exposure. If the reverse happens, it is an inverse dose-response relationship.


Null hypothesis

The premise that no relationship exists between A and B


Null study

Any epidemiological result which does not achieve statistical significance and therefore supports the null hypothesis.


Publication bias

There is an accepted tendency of medical and scientific journals to favour publishing epidemiological studies which statistically significant results. Although it is not necessarily the case, null studies are generally of less interest to their readers and the media. If null studies are not

published, evidence for the null hypothesis will be under-reported. Publication bias may also come into play when journals refuse to publish studies that do not support the journal's editorial stance.


Recall bias

In case-control studies, where information is gathered retrospectively through interviews, there is a known tendency for some subjects to exaggerate or downplay their exposure to A as a result of what they have been told about B. In the field of secondhand smoke research, this usually means lung cancer cases overstating their past exposure to tobacco smoke because they see their condition as uniquely related to smoking.


Relative risk/risk ratio

This is the headline estimate - or the best guess - provided by epidemiologists to show the relationship between A and B. The relative risk falls between the low and high end of the confidence interval.


Statistical significance

The degree to which a finding is greater or lesser than would be expected by chance with all other things being equal. Typically, epidemiologists aim for a 95% level of statistical significance ie. that there is only a 5% probability that the result came about by chance and is not indicative of a genuine phenomenon. Occasionally it is lowered to 90% (as by the EPA in 1992) thereby doubling the chance of recording a statistically significant result.


Smoker misclassification

When information about smoking behaviour is self-reported (as is usually the case), there is a known tendency for some smokers to classify themselves as nonsmokers (but rarely, if ever, the other way round). This misclassification leads to a muddying of results. Since around 1990, some researchers have tested their subjects for cotinine - which is present at high levels in users of tobacco. This procedure has helped them to exclude current smokers from their sample groups but because cotinine disappears from the body after several weeks, does not eliminate former smokers. In case-control trials, this can be a significant problem, since many smokers quit their habit when diagnosed with lung cancer.