Longitudinal trajectories of smoking status using 25 year annually-updated data and all-cause mortality followed over 30 years: A community-based prospective cohort study

Abstract

We aimed to evaluate the long-term risk of smoking for all-cause mortality according to smoking status trajectories using 25-year annually-repeated input, traced by group-based trajectory modeling with an extension to account for non-random participant attrition or truncation due to death. We examined 2682 men and 4317 women aged 40 to 59 years who participated in annual health checks for the community-based prospective cohort study, 1975–1984 enrollment in Japan. The main outcome measure was all-cause mortality (follow-up period: median 30.2 years in men and 32.2 years in women). We traced annual smoking trajectories, stratified by sex and smoking status at baseline. For smokers at baseline, we identified five trajectories in both sexes, with different patterns of smoking cessation (e.g., early quitters and lifelong smokers). We calculated HRs and 95% CI of all-cause mortality using Cox proportional hazards regression modeling adjusted for age, body mass index, alcohol intake, blood pressure category, dyslipidemia and glucose category. Compared with one-time-point-based smokers, trajectory-based lifelong smokers had an increased risk of all-cause mortality; HRs were 1.31 (95% CI, 1.18–1.46) in men and 1.26 (95% CI, 0.91–1.73) in women. Among community residents aged 40 to 59 years, 25-year-trajectory-based lifelong smokers had an approximately 30% increased risk for all-cause mortality compared to one-time-point-based smokers. Risk of all-cause mortality among smokers with earlier cessation varied materially. It is necessary to consider the trajectories of smoking status to clarify the long-term excess risk of smoking.