摘要：Objectives. To develop a smoking indicator that combines the popularity and duration of smoking and the quantity and quality of consumed cigarettes, factors that vary dramatically over time and across generations. Methods. We used retrospective reports on smoking behavior and a time series of cigarette tar yields to standardize nationally representative life-course smoking prevalence rates of 11 generations of US men and women, spanning 120 years. For each generation and gender, we related the standardized data with the corresponding rates of smoking-attributable mortality. Results. Our indicator suggests that US cigarette consumption spread, peaked, and contracted faster than commonly perceived; predicts a significantly stronger smoking–mortality correlation than unadjusted smoking prevalence; and reveals the emergence of a delay (by up to 8 years) in premature death from smoking that is consistent with increasing population access to effective treatments. In fact, we show that, among recent cohorts, smoking health-risk exposure is at a historic low and will account for less than 5% of deaths. Conclusions. Relative to unstandardized measures, our novel, standardized indicator of smoking prevalence describes a different history of smoking diffusion in the United States, and more strongly predicts later-life mortality. The past 2 decades have seen a global tobacco control treaty and enormous sums spent on antismoking campaigns worldwide. Despite these efforts, the estimates of smoking-attributable mortality increased from 3.5 million 1 to 5 million deaths annually, 2 and the latest forecasts suggest that this toll will exceed 8 million by 2030. Progress in reducing smoking prevalence has been generally slow. For example, since the United States launched its high-profile Healthy People initiative in 1990, it has frequently set ambitious targets to limit tobacco consumption, but has consistently failed to fully meet those targets. Public health officials are especially concerned because per capita consumption is now greater in developing than developed countries and continues to increase further, creating expectations of waves of smoking-caused deaths in the decades to come. 3 To frame policy discussions about the relationship between current smoking and future health, researchers commonly rely on the “cigarette epidemic” model. That model describes stylized patterns of aggregate smoking prevalence and its coevolution with smoking-related mortality. 4,5 It suggests that cigarette demand and supply initially increase with economic development but, as the population becomes more educated, health information spreads, and governments adopt antismoking policies, cigarette consumption eventually declines. Importantly, the hump-shaped pattern in overall smoking prevalence is followed by a similar pattern in smoking-attributable mortality a few decades later. This observation suggests that policymakers can minimize the long-run costs of smoking if they limit its uptake before it becomes a widespread social habit. However, the model yields no detailed policy prescriptions. On a more practical level, researchers study the smoking–mortality relationship by generating smoking histories of different demographic groups and using them in microsimulation models to estimate health costs. Among the most state-of-the-art models are those developed by the Cancer Intervention and Surveillance Modeling Network of the National Cancer Institute (e.g., to test the impact of tobacco control policies on smoking-related deaths). 6–8 Albeit more useful for policy planning, these models suffer 3 limitations. First, they rely on retrospectively reported data to generate smoking histories but use an outdated method to account for the differential mortality bias inherent in this type of data. Second, none of the models take into account temporal changes in cigarette quality. Third, no model combines all available information on smoking into 1 comprehensive measure that can be used as a macro-level monitoring device of smoking diffusion. Although some research describes concepts of smoking intensity, 9,10 no available indicator accounts for cigarette quality or has extensive sample coverage. In this paper we addressed all 3 limitations and built the first comprehensive indicator that can be used for surveillance of national smoking patterns, as a more precise alternative to smoking prevalence rates.
