Re: sex, smoking, and cancer: a reappraisal.

In a recent issue of the Journal (1), Perneger criticizes us (2), as well as a number of other authors, for stating that case–control study results dealing with relative risks of cancer also apply to absolute risks of cancer. The studies in question concerned male–female differences in risks of lung and bladder cancer according to cigarette smoking history. Our results did indeed show that female smokers had a higher relative risk of lung cancer compared with male smokers, at the same level of smoking (2), and we further concluded that female smokers had greater absolute risk of lung cancer than male smokers. Zang and Wynder (3) concluded similarly. Perneger is theoretically correct that the latter conclusion concerning absolute risks does not follow from the former one on relative risks. However, his criticism is wrong because its validity depends upon the male and female absolute risks for never smokers, i.e., the baseline category for the relative risks. That is, a high enough baseline lungcancer risk among male never smokers compared with female never smokers could, in theory, account for our findings of twofold to threefold higher smoking-related relative risks for females (2). As we pointed out 7 years ago (4,5), data on these baseline absolute risks had been available for some time. At that time, nine studies had demonstrated that the difference in baseline risk between male and female never smokers is small and, therefore, cannot explain the twofold to threefold higher relative risks seen for females. This result has been confirmed by the Nurses Health Study/Health Professionals’ Follow-up Study, with incidence rates among male and female nonsmokers of 10.1 and 12.2 per 100 000 per year, respectively (ratio 1.21) (6). In total, these 10 studies identified 1022 cases of lung cancer among nonsmoking males and 1832 among nonsmoking females. A combined precision-weighted estimate of the common female/male ratio of absolute age-adjusted rates from these 10 studies is 0.80 (95% confidence interval 0.73 to 0.87). Even if we use a conservative value for this ratio equal to the lower 95% limit, the highest relative risk solely attributable to differing baseline risks is 0.73 1.37. Thus, twofold to threefold higher relative risks for female smokers compared with male smokers require that female smokers be at higher absolute risk than male smokers at the levels of cigarette smoking where these twofold to threefold relative risks occur. These levels are not very great, beginning at less than 5 pack-years (2) or 5 kg of tar (7). The above evidence strongly suggests that women who have been regular smokers since young adulthood are at greater absolute risk of lung cancer than are men with the same smoking history, and they certainly are at greater risk than men relative to the risk levels they could achieve by not smoking. Contrary to Perneger’s discussion (1), this conclusion does not depend on whether multiplicative, additive, or other regression models are used for analysis.