The nature of these relationships may be influenced by misclassification bias, in that smokers may not be able to accurately recall lifetime brand use patterns. In addition, selection bias may occur if symptoms or diseases cause smokers to switch to lower-yield brands and if persons who switch to lower-yield brands exhibit different lifestyle characteristics that may influence disease risk (Samet, 1996). For example, persons who switch to lower-tar products appear to be more likely to eat more fruits and vegetables (Haddock et al., 1999). Although the magnitude of such biases may eventually prove to be small, research to assess their potential impact is warranted.

In related analyses, Thun and colleagues (1997a) compared the relative risks of lung cancer and coronary heart disease in Cancer Prevention Study I (from 1959 to 1965) with those observed in Cancer Prevention Study II (from 1982 to 1988). The risks did not decrease across studies when the authors stratified for gender, duration of smoking, age, and number of cigarettes smoked daily, even though the average tar yield of cigarettes consumed decreased substantially during that time. The results of this analysis could be influenced at least in part by unmeasured differences in lifetime smoking patterns (e.g., number of cigarettes smoked daily during adolescence) and by lifestyle factors (e.g., fruit and vegetable consumption). The weight of the evidence indicates that lower-tar and nicotine yield cigarettes have not reduced the risk of disease proportional to their FTC yields, in part because smokers compensate to obtain more nicotine (Burns, 2000; Kozlowski and Pillitteri, 1996) and in part because the products themselves contain higher concentrations of selected carcinogens (Hoffman et al., 1996).

Increased prevalence of the use of lower-tar and nicotine cigarettes has been associated with an increase in the percentage of lung cancers that are adenocarcinomas and a decrease in squamous cell carcinomas (Levy et al., 1997; Thun et al., 1997b). This changing histological pattern may be influenced by increased levels, over time, of nitrosamines in cigarette smoke and by increased inhalation of lower-nicotine-yield cigarettes, as smokers attempt to compensate to reduced nicotine yields by inhaling more deeply (Hoffman et al., 1996; Kozlowski and Pillitteri, 1996).

The Gallup Organization has polled Americans about their perceptions of cigarette smoking since 1949 (Moore, 1999). For example, in 1999, 95% of Americans considered smoking to be harmful, up from 60% in 1949. In 1999, 92% of Americans considered cigarette smoking to be one of the causes of lung cancer, compared to only about 40% in 1954. Additionally, about 80% of smokers considered smoking to be one of the causes of