eases other than lung cancer in this mixed population are then assumed to be proportionately reduced according to the measure of the maturity of the epidemic based on lung cancer rates. These adjusted or scaled relative risks thus incorporate the cumulative and simultaneous effects of exposure (P) and calculated relative risks (RR) on mortality, and in principle can be applied to calculate mortality attributable to smoking in the specific population under study. (See Peto et al., 1992, for more details about the methodology discussed here.)
To describe the methodology, it is first necessary to distinguish between smoking deaths from lung cancer (since virtually all the excess risk of lung cancer among smokers in the ACS cohort is actually attributable to smoking) and other causes (for which the excess mortality of smokers may be due to other differences, or confounding, between U.S. smokers and nonsmokers). Consider first the estimation of smoking-attributable deaths from lung cancer. In this case, we assume that all of the excess mortality from lung cancer observed among smokers is due to the habit, and estimate the national smoking-attributable lung cancer mortality by subtracting the smoothed U.S. nonsmoker rates from the national rate and multiplying by the population at risk in the given country (see Peto et al., 1992:1278, for the age-sex-specific smoothed rates). This is done for each age group 35-79 years. Below age 35, lung cancer is extremely rare (as indeed are most smoking-induced illnesses). Above about age 80, lung cancer death rates may become unstable or unreliable, and hence the attributable fraction for ages 80+ is assumed to be the same as for ages 75-79. Since the procedure is based on the assumption that the smoothed U.S. nonsmoker lung cancer rates adequately describe the levels for nonsmokers in other countries, if the observed national rate at any age is less than the U.S. nonsmoker rate, smoking-attributable deaths in that and all higher age groups are conservatively set to zero. That is, if the effects of smoking on lung cancer are not yet evident at younger ages, the epidemic is assumed not to be present at older ages.
To estimate smoking-attributable mortality from diseases other than lung cancer, a more complex procedure is required, since it cannot be assumed that the absolute rates among nonsmokers will be comparable in different populations as was done for lung cancer. There may well be important differences in other major risk factors for vascular disease (e.g., hypertension, blood lipid levels) or upper aerodigestive cancers (alcohol) among nonsmokers in different countries. For the same reason, it cannot be assumed that all of the excess mortality from these diseases among smokers in the ACS cohort, compared with nonsmokers, is due to tobacco. Smokers can be expected to be generally less health conscious than nonsmokers, and hence are more likely to adopt other deleterious health habits (e.g., poor diet, excessive alcohol consumption) that either independently or synergistically interact with smoking to increase the risk of death. Thus in the ACS cohort, part of the excess mortality of smokers from diseases other than lung cancer may well be attributable to factors other than smoking. In an attempt to control for this confounding and thus to ensure that the risks of tobacco are not