An iterative technique for identifying smoking deceivers with application to the Scottish Heart Health Study.

BACKGROUND: The study population consists of 3,977 self-declared nonsmokers for whom complete data on smoking biochemistry were available. Data were obtained from the Scottish Heart Health Study, a random cross-sectional population sample of 40- to 59-year-old men and women.
METHODS: Three biochemical markers of smoking (expired-air carbon monoxide, serum thiocyanate, and serum cotinine) are used together to derive optimum cut-points for distinguishing true nonsmokers from self-declared nonsmokers who are smoking deceivers, using an iterative procedure via a computer program.
RESULTS: The cut-points derived are, for carbon monoxide, 6 ppm (sensitivity, 0.81, and specificity, 0.94, when compared with the other two biochemical markers combined), for thiocyanate, 63.4 mumol/liter (sensitivity, 0.75; specificity, 0.92), and, for cotinine, 17.5 ng/ml (sensitivity, 0.77; specificity, 0.98). These cut-points are generally lower than those reported previously, primarily because other studies have taken self-reported smoking status to be the truth. The presence of deceivers among declared nonsmokers would tend to raise the biochemical levels of this group and hence of the cut-points.
CONCLUSION: The prevalence of smoking deception in the Scottish population is low: the frequency of those at or above all three cut-points is 1.2%, at or above two or more (the preferred definition of a smoking deceiver) is 2.2%, and at or above at least one is 16.4%.