difficult to determine the percentage of deaths that can be attributed to smoking.
The most straightforward approach is to follow a large number of people over a number of years and compare the mortality rates of smokers and nonsmokers to determine the additional risk of death caused by smoking. Such prospective cohort studies, as they are called, have the advantage of allowing researchers to collect a great deal of information from participants and then use that information to determine the relationship between various medical and lifestyle factors—smoking history or high blood pressure, for example—and the various causes of death. Nonsmoking factors affecting mortality can be controlled in statistical analyses of the effects of smoking.
The main weakness of such studies in determining the risks of smoking is the difficulty of obtaining a precise measure of smoking behavior. Most prospective cohort studies ask participants about their smoking behavior when they first enter the study and assume that this behavior remains fixed throughout the course of the study (Preston et al., 2010b). In reality, people vary their smoking behavior over time, with some quitting and others starting or restarting and still others increasing or decreasing how much they smoke. Furthermore, 20 to 30 years can elapse between the time a person starts smoking and the time serious health effects appear (Lopez et al., 1994). The resulting errors in recording the smoking behavior of study participants will generally weaken whatever connections the study might find between smoking and various causes of death. Thus such studies will typically underestimate the risk of smoking.
The Cancer Prevention Studies (CPS-I and CPS-II) conducted by the American Cancer Society, which cover the years 1959–1965 and 1982–1986, respectively, are two of the largest such studies to date to examine the effects of smoking on mortality, and they have provided some of the most important information available on the subject. In the CPS-II, investigators asked 1.2 million volunteers to fill out questionnaires asking about their jobs, diet, alcohol and tobacco use, medical history, and family history of cancer as of 1982. Since then the study has kept track of deaths among the 1.2 million participants by monitoring the National Death Index (American Cancer Society, 2009). By 2006, 488,000 of the participants had died.
Working from the data accumulated by the CPS-II and comparing the numbers of smoking-related deaths among smokers with the number of deaths among nonsmokers, Mokdad and colleagues (2004) calculated that smoking had been responsible for 435,000 deaths in 2000. A second group used the CPS-II data to calculate how smoking affects life expectancy. Focusing on the difference in life expectancy between smokers who quit smoking