Abstract 2157: Causal effects of delaying smoking initiation on subsequent lung cancer risk

Abstract

Abstract Background: Methods that report relative risks when describing the association between age at smoking initiation and lung cancer adjust for variables unknown or unpredictable at the time of initiation (smoking duration, pack-years etc.). Such methods do not provide valid estimates of the causal effect of smoking initiation. Also, relative risks are not informative for assessing the effects of youth smoking prevention programs or describing the impact of individuals' decisions to start smoking. Applying a causal modelling framework, we estimate absolute lung cancer risk - a metric more relevant to public health than relative risk - given variables fixed at the age of smoking initiation and averaging over subsequent smoking behaviors. Methods: We used the Environment And Genetics in Lung cancer Etiology (EAGLE) study (1942 lung cancer cases, 2116 controls), a population-based case-control study in Lombardy, Italy. We classified risk factors as fixed at smoking initiation (gender, childhood exposure to tobacco smoke [CETS], lung cancer family history [FH], birth cohort) and as causally related to the initiation age (pack-years, cigarettes per day, quitting, nicotine dependency). Using the sampling fractions for controls selection, we reconstructed the original study cohort and estimated absolute lung cancer risks overall and conditionally on fixed factors. We predicted lung cancer rates for different ages at smoking initiation as number of cases per 100,000 person-years (PY) using a binomial linear model and estimated corresponding risk differences (RD) and 95% confidence intervals (CI). Instead of adjusting for the effects of smoking behaviors that happen after initiation, we averaged over them. We show how assumptions about factors unknown at smoking initiation affect the estimates. Results: Delaying smoking initiation from age 16 to age 20 reduces estimated lung cancer rate in men from 149 to 78 cases per 100,000 PY with a RD=71 (95% CI 25 to 120), but there is apparently no corresponding effect in women (RD= -1 [95% CI -50 to 48] from 26 to 27 cases per 100,000 PY). The effect of delaying initiation is larger for those without CETS (RD=113 [95% CI 52 to 174] cases per 100,000 PY from 156 to 43) than those with CETS (RD=45 [95% CI -2 to 92] cases per 100,000 PY from 92 to 47). For those with FH RD was 214 (95% CI -106 to 534) cases per 100,000 PY from 282 to 68, and for those without FH it was 44 (95% CI 3 to 85) cases per 100,000 PY from 85 to 41). RDs were greater for early birth cohorts. There was heterogeneity among individuals with same initiation age but different subsequent smoking behaviors. Conclusion: We estimated absolute lung cancer risk from a case-control study using information fixed at the initiation age and averaging over subsequent smoking behaviors. Our estimates show substantial impact of delaying initiation in men, even if the person eventually smokes. Extrapolation to current youths and young adults requires integration of data on contemporary smoking patterns. Citation Format: Orestis A. Panagiotou, Fangyi Gu, Carolyn Reyes-Guzman, Maria-Teresa Landi, Michele Bloch, Neil E. Caporaso, Sholom Wacholder. Causal effects of delaying smoking initiation on subsequent lung cancer risk. [abstract]. In: Proceedings of the 105th Annual Meeting of the American Association for Cancer Research; 2014 Apr 5-9; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2014;74(19 Suppl):Abstract nr 2157. doi:10.1158/1538-7445.AM2014-2157