Comparative Analyses

Quantifying the impact of tobacco control policies on lung cancer mortality in the United States

Using five CISNET lung cancer models and detailed information on cigarette smoking, the CISNET lung group is quantifying the impact of tobacco control policies and increased awareness of the health effects of smoking following the seminal Surgeon General's report on past and future lung cancer mortality in the United States. The CISNET group is quantifying the levels of smoking and associated lung cancer deaths that would have occurred if the tobacco control efforts starting mid-century had never been initiated, as well as quantifying the number of lung cancer deaths that would have been avoided had tobacco control been perfect (i.e., if, following the Surgeon General's report, all current smokers quit and there was no new initiation).

Initial projections of the impact of tobacco control on lung cancer mortality from 1975-2000 highlighted the number of lung cancer deaths that were avoided due to the tobacco control efforts that did occur, and an upper bound on how many more deaths could have been avoided if the efforts had been perfect. The results of this comparative modeling project were published in the Journal of the National Cancer Institute (Moolgavkar, Holford et al. 2012), and also as a special monograph in the journal Risk Analysis (2012). In addition to the population-based models, this project resulted in the development of other sophisticated tools for the investigation of the association between smoking habits and human health. These included life tables for mortality associated with specific levels of smoking and a smoking history generator (SHG) (Jeon, Meza et al. 2012) that allows the stochastic simulation of individual smoking histories for the US birth cohorts from 1864-2012, while adjusting for smoking-related mortality.

We are extending this work to project the impact of tobacco control and lung cancer screening policies on current and future lung cancer rates in the US for the overall population, and also for different socio-demographic groups. We will also be extending our analyses to assess the effectiveness of tobacco control policies in middle-income countries (specifically Mexico, Thailand, and China).

Hypothesized Impact of Tobacco Control Policies on U.S. Smoking Prevalence

Hypothesized impact of tobacco control policies on US smoking prevalence

Hypothesized impact of tobacco control policies on U.S. smoking prevalence. Percentage of smokers among the white U.S. male population in a given year (19002000) by given year at birth (different colors). Solid lines represent the percentage of smokers estimated based on National Health Interview Survey data. Dashed lines represent the percentages (hypothesized) that would have been observed if tobacco control efforts had never been initiated.(Adapted with permission from Moolgavkar, Holford et al. 2012)

Impact of Tobacco Control on Lung cancer death rates and counts for men and women aged 30-84 years as observed and for modeled tobacco control scenarios.

Impact of Tobacco Control on Lung cancer death rates and counts for men and women aged 30-84 years as observed and for modeled tobacco control scenarios.

Click the graph for larger view

Impact of Tobacco Control on Lung cancer death rates and counts for men and women aged 30-84 years as observed and for modeled tobacco control scenarios.

Impact of Tobacco Control on Lung cancer death rates and counts for men and women aged 30-84 years as observed and for modeled tobacco control scenarios. ATC = Actual Tobacco Control; CTC = Complete Tobacco Control; NTC = No Tobacco Control. (Reprinted with permission from Moolgavkar, Holford et al. 2012)

Quantifying the potential benefits and harms of CT screening for lung cancer

Five CISNET groups have been engaged in a comparative modeling analysis that aims to predict the impact of low-dose computed tomography (CT) screening on US lung cancer incidence, mortality, and overdiagnosis. The National Lung Screening Trial (NLST) found a significant lung cancer mortality reduction in its CT screening arm in comparison with its chest-radiography (CXR) screening arm, demonstrating that screening heavy smokers with low-dose CT can be effective in the early detection of lung cancer. Meanwhile, the Prostate Lung Colorectal and Ovarian Cancer Screening Trial (PLCO) found no statistically significant difference in lung cancer mortality between a no-screen control arm and a CXR screening arm.

We collaborated with the NLST investigators and with the US Preventive Services Task Force (USPSTF) to calibrate our models to the NLST and PLCO trials (Meza, ten Haaf et al. 2014) and then extrapolate the trial results to the US population (McMahon, Meza et al. 2014; de Koning, Meza et al. 2014). Using our NLST and PLCO-calibrated models, we estimated the population benefits and harms of lung cancer screening in the US, evaluating the relative effectiveness of 576 alternative screening scenarios. In particular, we explored varying the screening age eligibility range (starting at ages 45, 50 or 55 and stopping at ages 75, 80 or 85), the smoking exposure criteria (a minimum of 10, 20, 30 or 40 pack-years) and the time since quit for smokers (a maximum of 10, 15, 20 or 25 years). Clinical concerns about the potential for increased operative mortality in older individuals with a history of heavy smoking, as well as increased comorbidity and reduced eligibility for surgery with curative intent at higher age limits, led us to focus on scenarios with stopping ages of 80 years or younger.

We found that more lung cancer deaths may be averted with more CT screening examinations, but there are diminishing returns (although not a single distinct point). The scenario of annual screening starting at age 60 and stopping at age 80 with 20 pack-years and 25 years since quitting (denoted A-60-80-20), which extends eligibility to individuals with fewer pack-years but starting at a later age, was still efficient with respect to number of screening examinations and lung cancer deaths averted, but it provided relatively fewer life-years gained than did A-55-80-30-15. For the three scenarios of A-55-80-30-15, A-60-80-20-25, and A-55-80-30-25, the number of screening examinations per lung cancer death averted increased progressively (550, 570, and 583 examinations), whereas the number of screening examinations per life-year gained was the highest (that is, the worst) for A-60-80-20-25 (52, 57, and 54 examinations). The A-60-80-20-15 scenario also resulted in the highest number and percentage of overdiagnosed cases. Our results supported the development of new lung cancer screening recommendations by the USPSTF (de Koning, Meza et al. 2014; Moyer 2014).

We are extending this work to assess the potential of risk-based screening strategies (i.e., screening eligibility based on the probability of developing lung cancer within a few years rather than based only on past smoking history), and the cost-effectiveness of different screening strategies in the US. We will also be extending our models and analyses to assess the effectiveness of screening in middle-income countries (specifically Mexico, Thailand, and China).