Childhood Smoking, Adult Cessation, and Cardiovascular Mortality: Prospective Study of 390 000 US Adults

Smoking still causes about 100 000 US deaths per year from cardiovascular disease.1 There were an estimated 25 million daily smokers in the United States in 2018, including 5 million who began smoking regularly in childhood (defined here as before age 15 years); of these, an estimated 0.5 million began before age 10 years.2 Previous research has shown that, among daily smokers, those who began youngest have the greatest relative increase in overall adult mortality,2 but better evidence is needed on the relevance of the age when smoking began to cardiovascular adult mortality. We investigated this using the annual US National Health Interview Surveys (a cross-sectional household interview survey of the noninstitutionalized US population) linked to the National Death Index.3 Further details about the National Health Interview Surveys, and the anonymized data that support the findings of this study, are publicly available online (https://www.cdc. gov/nchs/nhis/data-quest ionna ires-docum entat ion. htm). In 1997 to 2014, 424 793 US adults aged 25 to 74 years were surveyed and have been followed for mortality to December 31, 2015. Participants were asked about their health, habits (including smoking, drinking, and physical activity), medical history, and demographic details. Current smokers were categorized by the age they began smoking regularly (as reported at recruitment): <10, 10 to 14, 15 to 17, 18 to 20, or >20 years. Occasional smokers were excluded. Ex-smokers were categorized by the age they last smoked regularly: 15 to 34, 35 to 44, 45 to 54, or 55 to 64 years (excluding the few who had quit at older ages). Ex-smokers who quit within 5 years of death were excluded to reduce reverse causality (by which smokers quit because of the onset of serious illness). Causes of death were recorded according to the International Classification of Diseases, Tenth Revision (ICD-10); our analyses of cardiovascular mortality included cardiac (I00–I09, I11, I13, I20–I51) and cerebrovascular (I60–I69) causes. Prospective analyses were restricted to “premature” deaths (here defined as ages 25–74 years). Cox regression was used to estimate mortality rate ratios (RRs) comparing different smoking histories. For each baseline-defined smoking group (including the reference group with RR = 1.0), the RR is shown with a group-specific 95% CI that reflects the variance of the log risk in just that 1 group.4 (The variance of the log RR comparing 2 different groups is the sum of the variance of the log risk in 1 group and the variance of the log risk in the other group.) The RRs were adjusted for age-at-risk (in 5-year groups; 10 categories), sex, race (5 categories), education (4 categories), region (4 categories), and alcohol consumption (5 categories). All participants provided informed consent. Ethics approval for analyses of this publicly available anonymized dataset was not needed. Analyses used SAS v.9.4 and R v.3.1.1. After excluding participants with missing information on smoking, covariates of interest, or mortality linkage, there were 390 929 participants aged 25 to 74 years (mean age 47, 56% female) at recruitment: 228 165 (58%) never smokers, 88 717 (23%) ex-smokers, and 74 047 (19%) current smokers. Among current smokers, 1403 (2%) had begun before age 10 and 14 421 (19%) at ages 10 to 14 years. During 3.5 million