The conventional wisdom is that, in the majority of cases, continuous data should be analyzed as continuous because conversion to categoric variables will almost systematically result in loss of power. As such, we analyzed the quantity of emphysema as a risk factor for lung cancer as a continuous variable, with adjunct analyses of emphysema as a discrete variable. We agree, however, that should an all-or-none effect of emphysema on lung cancer risk be confirmed, analyzing our data as a dichotomous variable (emphysema vs no emphysema) could theoretically make sense and could unmask a signal not otherwise evident. Indeed, the crude OR calculated by Dr Smith using a threshold of 5% is 1.7 (1.0-2.9). The use of a density threshold method makes a comparison between emphysema and nonemphysema groups impossible (because having no voxel < −900 Hounsfield units is improbable in any data set), but using an arbitrarily defined threshold of 5%, the adjusted OR is 1.85 (1.05-3.26), P = .034.