CLINICAL PERSPECTIVE What Is New?

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

stress. [6][7][8] Given the high prevalence of smoking in childhood, its health consequences, and the difficulty of smoking cessation, studies of childhood smoking are highly relevant to development of strategies for prevention of adult cigarette smoking and related health issues.
In the United States, ≈87% of adult daily smokers started smoking before age 18 and 95% started before age 21. 9 Despite a 2015 Institute of Medicine report finding that much adult smoking could be prevented by postponing smoking exposure until age 21, 10,11 childhood smoking continues to be a major public health problem. Most children/adolescents are not daily smokers, [12][13][14] and previous longitudinal studies have related the lower-intensity cigarette smoking in adolescence to future regular smoking. [15][16][17][18] However, these studies have been short term (2-8 years), and studies with longer-term follow-up have linked smoking during adolescence only to smoking in the early thirties. 19,20 Despite declining US adolescent smoking prevalence from nearly 40% among 12th graders in 1995 to around 10% in 2018, adolescent smoking continues to affect a significant proportion of the adolescent population. 21 Thus, there is an urgency to reinforce the efforts of policies such as Tobacco 21 22 to decrease initiation of smoking in adolescence and also effectively develop and distribute new data on the adverse effects of adolescent smoking.
We report epidemiological findings from the International Childhood Cardiovascular Cohort (i3C) Consortium, a collaborative study among 7 international cohorts recruited in childhood and followed longitudinally into adulthood. The report supports the earlier Institute of Medicine findings that greater involvement with cigarettes in childhood and adolescence leads to greater probability of adult smoking, and that the probability of being an adult smoker is higher the earlier that the child or adolescent smokes. 10,11 But, in addition, we add to the Institute of Medicine conclusions by extending the relation of adolescent smoking to adult smoking into the forties and by comparing data from 3 international settings: the United States, Australia, and Finland.

METHODS
The data that support the findings of this study are available upon reasonable request through the i3C Steering Committee, Chair Terence Dwyer (terence. dwyer@georgeinstitute.ox.ac.uk).

Study Sample
The 7 i3C cohorts are described in detail in an earlier publication 23 and are listed in Table 1. Each cohort recorded smoking behavior in childhood during the 1970s-1990s with follow-up through midadulthood. Smoking information was obtained from 6687 i3C participants in childhood/adolescence (age 6-19 years) and again during their twenties and during their forties. Of the 6687 persons, 2465 (36.9%) were smokers in their twenties and constituted the sample for cessation analyses. To assess potential for selection bias, we also identified the 10 352 excluded i3C participants who were asked about smoking during childhood/adolescence and were at least 40 years old in 2016, but were not successfully followed.

CLINICAL PERSPECTIVE
What Is New?
• This cohort study independently replicates earlier knowledge that adolescent smoking predisposes to adult smoking and newly extends that knowledge to increased risk among adolescent smokers of adult smoking and reduced cessation in their forties, even those adolescents who only smoked a few cigarettes.
What Are the Clinical Implications?
• The medical community should actively support measures to prevent access to tobacco products before age 21. • Pediatricians particularly should be alert to smoking among their patients and should explain the lifetime risks to child patients and their parents.  Table 1 presents an overview of the smoking-related questions by cohort. The specific algorithms used to generate childhood/adolescent smoking information are provided in Table 2, and detailed questionnaires used in childhood/adolescence are provided in Data S1. Smoking questions differed among the cohorts and were harmonized for this study into childhood/adolescent smoking intensity categories: (1) never; (2) trier (self-reported any of the following to survey questions: "tried"; "a few puffs"; "<10 cigarettes lifetime"; "only once"; "once or twice"; "a few times"; and "at least tried" coupled with either "not habitual" or "never smoked"); (3) nondaily (excludes triers and includes adolescents who smoked more than a few times, but not daily); and (4) daily smoking. Children/adolescents who said they had smoked other than daily, but had insufficient information to classify into the trier versus nondaily categories (n=424) were included in the denominator to achieve unbiased prevalence estimates, but they were not included in any other analysis. In the Bogalusa Heart Study, the National Heart, Lung, and Blood Institute Growth and Health Study, and the Young Finns Study, repeated measures during childhood and/or adolescence were often available.

Childhood Data
In that case, the participant was assigned his or her highest mentioned intensity category. An additional category, adolescent quitter, was used when the participant categorized him-or herself during childhood/adolescence as a quitter or in repeated measures when a report of daily smoking was followed only by reports of quitting or not smoking during adolescence.

Adult Data
Adult data were obtained in sporadic visits, the timing of which varied by cohort. In addition, between 2015 and 2018, an attempt was made to recontact all US and Australian i3C participants to complete a Heart Health Survey, which was designed to assess health status in middle age, including smoking. The Cardiovascular Risk in Young Finns Study did not complete the i3C Heart Health Survey, but participated in a clinic examination in 2011-2012, which served as a middle-age assessment. All examinations were approved by the institutional review board of each cohort. Informed consent was obtained from all adult participants, and participants aged <18 years all assented with parental signed consent.

Adulthood Smoking Outcomes
Primary adult smoking outcomes were defined by the i3C Heart Health Survey questions asking about daily smoking status (answered as either yes or no) during ages twenties and forties (these questions were modeled after questions asked in both the 12and 15-year surveys of The Million Women Study 24 ). In the i3C Heart Health Survey, smokers were categorized by their answer to the question, "Have you ever smoked?" Daily smokers in their twenties then answered the question, "In your 20's, ___typical number of cigarettes a day." Daily smokers in their forties also answered the question, "In your 40s, ___typical number of cigarettes a day." An adult quitter was defined as a daily smoker in their twenties who said they were nonsmokers in their forties. In the Young Finns Study and in the sporadic adult visits preceding the i3C Heart Health Survey, smoking in the twenties was determined from smoking status at the last visit during ages 20 to 29, and smoking in the forties was determined from the last visit during ages 40 to 49 (adult smoking asked in the sporadic adult visits used a variety of questions).
Of the 6687 participants in the analysis, 2802 had repeated measures during ages 20 to 29 years and 2029 had repeated measures during ages 40 to 49 years. For most persons, the last measure within the age decade was unambiguous, showing consistency of either smoking or cessation through the decade. However, 53 participants whose last age twenties measure was "nonsmoker" had been smokers at a previous examination earlier in their twenties. Correspondingly, 24 participants whose last age forties measure was "nonsmoker" had been smokers at a previous examination earlier in their forties. These persons were interpreted as being nonsmokers in the early adulthood and midadulthood decades, respectively. A sensitivity analysis was performed in which they were classified as smokers.

Statistical Analysis
Each participant was categorized in 1 childhood/adolescent smoking category and in 1 category of age at which smoking information was last queried. Selection bias was studied by comparing characteristics in the analyzed sample (N=6687) with those in the excluded sample (N=10 352). Our first focus was on the association of adult smoking and cessation with extent of childhood/adolescent smoking. We examined the association of childhood smoking with daily smoking status (yes/no) during the twenties and forties and with smoking cessation by the forties using unadjusted counts. To account for confounding, we back-transformed smoking and smoking cessation probabilities computed in logistic regression after adjustment for sex, race/ethnicity (white, black, or other), and cohort. Socioeconomic status, a critical correlate of smoking behavior, 25 was assessed as a covariate, as represented by parental educational level. Parental education was restricted to 3 categories to harmonize across countries, given international differences in educational systems. The person's own education as reported in adulthood was assessed in 5 categories. Despite significant differences in smoking in adolescence and in adulthood based on education, such adjustment for either parental or own education did not alter findings substantially, and socioeconomic status is not included as a covariate in the tables.
In the several i3C cohorts, the last age at which childhood/adolescent smoking information was queried was often before adolescence ended (ie, 18-19 years old). Because those who attained a given smoking status at an earlier age had the opportunity to become more involved with cigarettes before adolescence ended, people who attain a given childhood/adolescent smoking category at an early age are expected to have a higher adult smoking prevalence than the corresponding subgroup of people who attained that childhood/ adolescent smoking category at a later age. However, the reverse should be true for quitters: The expectation is that those who stated that they quit later in adolescence should be less likely to smoke as adults than those who said they quit at a more-vulnerable, younger age. We categorized participants according to the age at which smoking status was last queried (6-12, 13-14, 15-17, and 18-19 years old) and stratified the analyses of adult smoking by childhood/adolescent smoking age.
P<0.05 was regarded as statistically significant. All analyses were conducted in SAS software (version 9.4; SAS Institute Inc., Cary, NC).

RESULTS
Of the 6687 persons included in this study, 45.6% (N=3078) were from the United States, 34.9% (N=2319) were Australian, and 19.4% (N=1290) were Finnish (Table 3). There was a wide spread of parental and participant's own education. The 6687 analyzed persons included 57.0% female (N=3814), 37.4% who completed at least college (N=2497), and 25.9% whose parents completed at least college (N=1363). These percentages were higher than in the 10 352 who were asked about smoking during childhood/adolescence and were at least 40 years old in 2016, but were not successfully followed (4746 or 45.9%, 817 or 25.3%, and 961 or 18.7%, respectively; P<0.001 for all comparisons).

Childhood Smoking
As shown in Table 4, data on childhood/adolescent smoking intensity were available for analysis from ages 6 to 12 (N=2136 or 31.9% of participants) to ages 18 to 19 (N=1628 or 24.3% of participants). Overall, prevalence of never smoking was 46.9% (3138 of 6687), but 53.1% (3549 of 6687) smoked at some point during childhood/adolescence. Daily smoking occurred in 12.9% (865 of 6687), with only 14 persons aged 6 to 12 and 90 persons aged 13 to 14 years categorized as daily smokers. Over half of the smokers who did not smoke daily were classified as triers. Evolution of smoking intensity was reflected in the decreasing prevalence of the never and trier subgroups with increasing age, whereas prevalence of nondaily (excluding triers) and daily smoking increased with age.

Prediction of Smoking During Age Twenties
Overall, prevalence of daily smoking in age twenties was 36.9% (unadjusted n/N=2465 of 6687), varying across race-country from 29.6% (382 of 1290, Finns) and 33.8% (338 of 999, US blacks) to 39.8% (922 of 2319, Australians) and 39.5% (803 of 2032, US whites; P<0.001). Data on parental education were available on 5256 of the children/adolescents. Prevalence of smoking in age twenties decreased with increasing levels of parental education: 36.7% (969 of 2641) high school and below, 33.5% (419 of 1252) beyond high school but no college degree, and 32.2% (439 of 1363) college degree and above (P trend, <0.001). Prevalence also decreased with increasing levels of the individual's (N=6673) own education: 58.1% (198 of 341) less than high school, 47.3% (641 of 1356) with high school degree, 39.5% (978 of 2479) with education beyond high school but no college degree, 26.5% (407 of 1538) with college degree, and 24.3% (233 of 959) with a degree higher than college (P trend, <0.001).
Prevalence of daily smoking in the twenties was graded by intensity of smoking across never, trier, nondaily (excluding triers), adolescent quitter, and Also included 424 who had tried smoking during adolescence (not daily smokers), but who could not be classified as trier or nondaily based on available information. These persons were included in the denominator in this table to achieve unbiased prevalence estimates, but not included in any other analysis. These were 6.3% of the sample: 1.4% (n=23) for query at ages 18 to 19; 11.6% (n=184) for query at ages 15 to 17; 9.5% (n=127) for query at ages 13 to 14; and 4.2% (n=90) for query at ages 6 to 12. All but 90 of the 2136 persons in the 6 to 12 category were aged 9 to 12. daily subgroups within each childhood/adolescent age group and inversely related to age when each childhood/adolescent smoking intensity was reached (except among childhood/adolescent daily smokers), after adjustment for sex, race, and cohort (P for trend, <0.001 for each age; Tables 5 and 6). For instance, the gradient for smoking prevalence in the twenties across childhood/adolescent smoking intensity categories within the age group 18 to 19 years was 2.6% in the childhood/adolescent never smokers, 7.8% in the trier subgroup, 20.7% in nondaily (excluding triers) smokers, and 76.3% in daily smokers (P<0.05 for all pairwise comparisons). At the same time, among those who were classified as a  Among the 2.0% (n=137 of 6687) who had been daily smokers in adolescence, but who quit before adulthood, most (n=111) reported quitting during ages 18 to 19. Of these 111, 31.5% relapsed (were smokers in their twenties). The few adolescent smokers who quit by ages 15 to 17 (n=26) had relapse rates in the twenties of 50% to 60%, which was significantly higher than those who quit in later adolescence (P<0.001).

Prediction of Smoking During Age Forties and Smoking Cessation by the Forties
Smoking prevalence during the forties followed a pattern across childhood/adolescent smoking intensity categories similar to smoking prevalence during the twenties, but generally at a lower smoking prevalence (20.8%; 1394 of 6687; Tables 5 and 6). Only 2.6% (10 of 386) of participants were smokers in their forties among those who said they never smoked when last asked about smoking at ages 18 to 19 and were nonsmokers in their twenties (Table 7).
Cessation by their forties was generally inversely graded across the 5 childhood/adolescent smoking intensity subgroups. Among the 2465 persons who were smokers in their twenties, adjusted cessation rate in the forties was 56.4% (unadjusted n/N=400 of 760) in the childhood/adolescent never smoked subgroup, 52.3% (313 of 546) in the childhood/adolescent triers, 47.5% (119 of 243) in the childhood/adolescent nondaily (excluding triers) subgroup, and 37.3% (264 of 692) in the childhood adolescent daily smokers (P trend, <0.001), after adjustment for sex, race, cohort, and age. Childhood/adolescent age-stratified analyses of cessation (Tables 5 and 6) were generally consistent with age-adjusted values.

Smoking Prevalence by Country
The association between childhood/adolescent smoking intensity and adult smoking (during age twenties or forties) was not modified by sex, parental education, or own education. The pattern of a positive association of smoking prevalence in the twenties across childhood/ adolescent smoking categories and an inverse association with age at which smoking was asked were observed in each country ( Table 8). The few countryspecific exceptions to this pattern are either slight deviations from the pattern (ie, in Australia age 15 years, 91.3% [21 of 23] of nondaily smokers versus 88.2% [60 of 68] of daily smokers) or based on small sample size (ie, in the US participants smoking in the twenties was less frequent in those aged ≤12 years than in those aged 13-14 years, but among those aged ≤12 years, the 62.5% [5 of 8] smoking in the twenties among nondaily smokers and the 70.0% [7 of 10] smoking in the twenties among daily smokers were based on small sample sizes). Nevertheless, smoking prevalences in the twenties were lowest in Finland and highest in Australia. In each country, smoking cessation in the forties was lower for childhood/adolescent daily smokers than for other childhood/adolescent subgroups (Table 9), but cessation was unrelated to age at which childhood/adolescent smoking was asked.

DISCUSSION
This international study of smoking in children, and its longitudinal effects on adult smoking, confirms the important points made in the 2015 Institute of Medicine report 10,11 that: (1) smoking in children very often leads to regular smoking in adulthood; (2) smoking experimentation continues to occur throughout adolescence and is more likely to be associated with adult smoking the earlier it occurs; and (3) even trying a cigarette during childhood/adolescence is associated with substantial excess risk of adult smoking. In addition, the findings from the present study expand on the Institute of Medicine report. 10,11 First, older adolescent smokers seem less susceptible than the younger adolescent and childhood smokers to becoming adult smokers: Smoking cessation during ages 18 to 19 was much less likely to be followed by relapse than was cessation earlier in adolescence; we speculate that this finding may be attributed to maturation, which would be greater at the older adolescent age. Also, prevalence of smoking in the twenties was much lower for persons who were nondaily smokers (excluding triers) in ages 18 to 19 than it was for the same groups earlier in childhood/adolescence, although it continued  to be important at a rate of around 30%. Second, the present study extends longitudinal evaluation of childhood/adolescent smoking into the forties. Among 386 persons who never smoked throughout childhood, adolescence, and their twenties, only 10 persons (2.6%) smoked in their forties. Third, this study includes both a European and an Australian cohort, showing that the relationship of childhood/adolescence smoking to adult smoking in those countries follows the same pattern as documented in the United States. It is not clear why smokers with an earlier and moreintense smoking history tend to establish daily smoking and have trouble quitting in adulthood. A strong possibility is the early exposure to nicotine. In line with findings of the 2012 Surgeon General's Report, 9 it has been suggested that nicotine addiction is stronger when smoking initiation occurs earlier in childhood (although no age cut point has been identified), [26][27][28][29] and the developing neurological system of younger children is particularly vulnerable to nicotine. 30 Although we showed that smoking in the twenties was more likely in those whose parents had lower education, we did not have other relevant information, such as the extent of exposure to secondhand smoke in the childhood/adolescent home or the smoking behavior of childhood/adolescent peers. The present study had no specific measure related to nicotine addiction, nor did it have any measure of other possible explanators, including deviance proneness, behavioral and emotional dysregulation, and both adult and peer smoking modeling. 31,32 The data reported in Table 5 could be the basis for antismoking counseling. Although few individuals who never smoked at ages 18 to 19 became daily smokers, even triers at that age had an excess risk of taking up smoking by their twenties compared with those who never smoked during childhood/adolescence. In particular, 33% of triers at ages 15 to 17 as opposed to 9% at ages 18 to 19, were smokers in their twenties. Triers during ages 6 to 14 had a 50% chance of smoking in their twenties, and 18% were still smokers in their forties. These data emphasize the health benefits likely to occur by preventing any childhood smoking exposure, particularly beginning in early adolescence.
Like our results predicting smoking in the twenties from childhood/adolescent smoking, previous studies reported that adolescent puffers (analogous to our trier category) and more-intensive nondaily smokers (analogous to our nondaily [excluding triers] category) are at 1.9 to 3.5 times and 3.1 to 10.8 times the risk, respectively, of daily smoking 2 to 8 years later than those who never smoked in adolescence, 15,16,18 and moreintensive experimenters compared with puffers had 1.9 to 2.7 times the likelihood of smoking in 4 years. 17 The Australian Childhood Determinants of Adult Health study, a member of i3C, followed 3559 children aged 9 to 15 years and reported that children who smoked only a few puffs or <10 cigarettes lifetime were at 1.3 to 3.0 times higher risk of becoming daily smokers in early adulthood compared with those who had never smoked by ages 9 to 15 years. 19 A study conducted in the United States reported that smokers aged 12 to 21 years who smoked over 3 packs per month, but did not smoke daily, were as likely as daily smokers to be smoking 14 years later (odds ratio, 0.97; 95% CI, 0.53, 1.80), whereas nondaily smokers of 1 to 5 cigarettes per month and 6 to 60 cigarettes per month were less likely than daily smokers to be smoking 14 years later (odds ratio, 0.21; 95% CI, 0.15-0.29 and odds ratio, 0.22; 95% CI, 0.14-0.34, respectively). 20 Similar to previous studies, our study showed a graded prevalence of smoking during young adulthood across levels of childhood smoking intensity. However, our results extend those findings to show the adverse effect of childhood/ adolescent smoking into the forties in relation to the finding that daily smoking in the twenties was followed by difficulty with smoking cessation by the forties.
Our i3C findings contribute to the ongoing debate for the Tobacco 21 policy, showing the adverse effect of initiating smoking in childhood/adolescence, thus strengthening the Tobacco 21 aims to prevent or delay access to cigarettes by making sale of cigarettes illegal through age 21. Laws prohibiting the sale of tobacco products to anyone under the age of 21 have been implemented in 19 US states and at least 500 localities, but cover only approximately half the US population, according to the Campaign for Tobacco-Free Kids. 33 A US measure starting in 2020 was signed into law on December 20, 2019. 34 Tobacco 21 laws have not been passed in Europe, 35 and limited evidence of the benefits of offsetting preadult smoking has limited progression in Australia. 36 The present study has some strengths. The available i3C Consortium data had a large sample size in 3 countries, a variety of measurements of smoking variables in childhood and adulthood, and extended duration of follow-up relative to the earlier CDAH report. 19 In addition, the study collected childhood smoking data individually in clinics or through research staff independent of schools, which would minimize desirability bias that might otherwise occur in the school classroom environment.
Limitations include that misclassification of childhood smoking status is possible. First, i3C included heterogeneous questionnaires on childhood smoking across i3C cohorts, so not all studies contributed equally to the findings. Second, children may have difficulty in understanding and responding to some smoking questions, and this difficulty is greatest for young children. Third, the natural history of smoking is inevitably affected by policy change, such as the Master Settlement Agreement in 1998 in the United States, and can change over time; US adolescent smoking prevalence has declined since i3C data were collected. 21 Nevertheless, the results of this study remain relevant in terms of the greater the adolescent smoking intensity, the higher the likelihood of becoming an adult smoker. Fourth, our findings may underestimate the prevalence of smoking in both childhood and adulthood, given that those excluded (Table 3) had lower parental and own education than those included. Fifth, we did not consider smoking products other than cigarettes, particularly electronic cigarettes, which have become increasingly popular since 2010. 37 Conflicting data have been reported regarding the association between electronic cigarettes and cigarette smoking cessation. 37,38 Finally, our results may not be generalizable to populations from low-and middle-income countries, where there is currently higher prevalence of established smoking and where societal conditions unique to those countries may influence the probability that an adolescent will smoke as an adult. 1 In summary, these international, long-term follow-up data showed that smoking in children and adolescents, even only a few times, was associated with substantially higher probability of being a daily smoker not only in young adulthood, but also into mid-adulthood. There was also a lower probability of cessation through the forties, graded by childhood/ adolescent smoking intensity, among individuals who smoked in the twenties. Smoking experimentation continued to evolve throughout adolescence. These data reinforce that children who might otherwise be inclined to try smoking should instead completely avoid smoking. Our findings were consistent with concepts reported by the Institute of Medicine that postponing smoking exposure until age 21 remarkably reduced smoking prevalence 10,11 and may be interpreted as supporting policy strategies in the United States, Europe, and Australia, which effectively make nicotine products harder for children to access, such as Tobacco 21 in the United States. Quitting cigarette smoking is well known to be difficult, and the surest approach not to be a smoker in adulthood is never to start smoking at any point in time. However, extra efforts should be made to prevent smoking experimentation and/or initiation earlier in the life course given that it appears to be critically related to moreunbreakable lifelong smoking patterns.