and those who never smoked, the researchers estimated that quitting smoking at age 35 added 6.9 to 8.5 years of life expectancy for men and 6.1 to 7.7 years of life expectancy for women compared with those who kept smoking (Taylor et al., 2002).
The reason why the number of deaths attributable to cigarette smoking has stayed as high as it has despite large reductions in the prevalence of smoking is that heavy smoking in the past has left a clear imprint on current mortality levels. One feature of that imprint is that the relative mortality levels of smokers have risen relative to nonsmokers, reflecting the heavy smoking histories of smokers today. Comparisons of the CPS-I and CPS-II data indicate that the age-adjusted death rate from lung cancer per 100,000 person-years increased among current smokers between the period covering the CPS-I (1959–1965) and that covering the CPS-II (1982–1986), from 187.1 to 341.3 in men and from 26.1 to 154.6 in women (Thun et al., 1997). The increased mortality risk appears to be partially attributable to differences between the two studies in smokers’ average age of initiation, number of cigarettes smoked per day by smokers, and duration of smoking. Among those who had never smoked, the age-adjusted death rate did not change substantially between studies for either males or females.
Prospective cohort studies make it possible to estimate directly the number of deaths caused by smoking or other factors, but they are expensive to carry out, require decades of commitment, and often are not available for populations one wishes to study. In particular, while the CPS-II offers reliable data on a large (but nonrepresentative) subgroup of the U.S. population, few other countries have comparable studies. Therefore, researchers have developed a variety of indirect ways to determine the number of deaths caused by smoking.
The major challenge researchers face when attempting to estimate smoking-related mortality in a population is obtaining reliable information about the smoking habits of the population. Peto and colleagues (1992) devised an innovative way of getting around the fact that good data on the smoking habits and histories of most populations one wishes to study are relatively scarce. The basic idea is that one can use the rate of lung cancer in a population to obtain a reasonably good estimate of the total smoking burden in that population. One starts with the amount of lung cancer mortality in the population, which is available wherever careful cause-of-death statistics are kept. The underlying assumption is that even if no one in a particular population of people smoked, there would still be a certain small number of lung cancer deaths, and it is possible to know what that small number would be by, for example, looking at the lung cancer