LAUSR

In  Thorax , Bhatt and colleagues’ interesting cross-sectional analysis of data from the COPDGene study cohort demonstrates, somewhat unsurprisingly, that the compound (multiplicative) measure of cumulative exposure to tobacco smoking, ‘pack-years’, is a suboptimal index of exposure with respect to effects on indices of airflow obstruction, spirometric lung volume and emphysema.1 This is unsurprising because there is no biological basis for assuming that the effects should be manifest by a simple multiplicative relationship between duration of smoking and estimated average number of cigarettes smoked per day. Compound measures are convenient, but usually not robust, measures for epidemiological investigation. This study confirms that it is almost always more appropriate (and powerful) to analyse the individual components of the exposure separately and to model the interaction between them. Perhaps more surprising is the near absence of any effect of the estimated …