Objective To describe the determinants of self-initiated smoking cessation of duration of at least 6 months as identified in longitudinal population-based studies of adolescent and young adult smokers.
Methods A systematic search of the PubMed and EMBASE databases using smoking, tobacco, cessation, quit and stop as keywords was performed. Limits included articles related to humans, in English, published between January 1984 and August 2010, and study population aged 10–29 years. A total of 4502 titles and 871 abstracts were reviewed independently by 2 and 3 reviewers, respectively. Nine articles were retained for data abstraction. Data on study location, timeframe, duration of follow-up, number of data collection points, sample size, age/grade of participants, number of quitters, smoking status at baseline, definition of cessation, covariates and analytic method were abstracted from each article. The number of studies that reported a statistically significant association between each determinant investigated and cessation were tabulated, from among all studies that assessed the determinant.
Results Despite heterogeneity in methods across studies, five factors robustly predicted quitting across studies in which the factor was investigated: not having friends who smoke, not having intentions to smoke in the future, resisting peer pressure to smoke, being older at first use of cigarette and having negative beliefs about smoking.
Conclusions The literature on longitudinal predictors of cessation in adolescent and young adult smokers is not well developed. Cessation interventions for this population will remain less than optimally effective until there is a solid evidence base on which to develop interventions.
- Smoking cessation
- young adult
- prevention and control
- health behaviour
Statistics from Altmetric.com
- Smoking cessation
- young adult
- prevention and control
- health behaviour
In spite of considerable declines in the prevalence of smoking in the past decade, tobacco use remains the leading preventable threat to public health in many Western countries. Further, the steady decline in the prevalence of youth smoking since the mid 1990s seems to have stalled in several countries.1 In 2006–2007, nearly 50 000 Canadian youth in grades 5–9 were current smokers.2 Clearly more research is needed to inform evidence-based tobacco control interventions for young smokers.
One of the cornerstones of tobacco control is cessation. While many young smokers express the desire to quit, most have a great deal of difficulty in doing so.3 Approximately 70% of adolescent smokers try to quit each year.4 Many make multiple cessation attempts, but 90% of those who try to quit relapse within 1 year. The probability of relapse is higher among older adolescents and among those who smoke daily.4 A recent investigation of the natural course of cessation suggests that cessation in youth may be understood as a progression of phases and that novice smokers lack awareness of the difficulty that they will experience as they attempt to quit, at least during the early phases.3 Specifically, the first phase occurs 1–2 months after the first puff, as young smokers increase exposure to nicotine through inhalation and smoking a first whole cigarette. This phase is characterised by naïve certainty about the ability to quit, when young smokers confidently declare that they have stopped smoking completely and forever. This is followed rapidly by expression of a conscious desire to quit and the growing realisation that quitting requires serious effort. Over the next 2 years as cravings, withdrawal symptoms and tolerance set in and as smoking escalates to monthly cigarettes use, novice smokers gradually lose confidence in their ability to quit. Approximately 1 year after reporting loss of confidence and after escalation to daily smoking, young smokers come to realise that they now smoke because it is very hard to quit. Full-blown WHO International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) tobacco dependence manifests approximately 14 months thereafter.
While intensity, frequency and duration of cigarette use along with concomitant nicotine dependence are likely strong determinants of successful cessation in youth as well as in adults, little is known about other determinants of self-initiated cessation in young smokers. There are at least four reviews to date5–8 that attempt to synthesise the literature on the determinants of cessation in young smokers, but the results of these reviews may not be optimal because they did not use systematic methods to identify articles to retain for analysis,5 the definition of cessation varied widely across studies retained for review,5–8 they included studies from one calendar year only,6 7 and/or they focused on cross-sectional,6–8 intervention and clinical studies.6
The objective of the current paper was therefore to systematically summarise the empirical evidence across longitudinal population-based studies on the determinants of longer-term self-initiated cessation in adolescents and young adults. Because longitudinal studies by definition assure that exposure precedes outcome, we opted to restrict our review to longitudinal data to the exclusion of cross-sectional studies, thereby providing relatively stronger evidence for causal associations. In addition, we opted to include only those studies in which participants had quit smoking for 6 months or longer in order to preclude obscuring the determinants of short-term versus longer-term cessation, because smoking cessation of at least 6 months of abstinence corresponds to the recommended definition of successful abstinence among untreated smokers,9 and because relapse most often occurs during the first 6 months of abstinence4 10 especially among youth.11
We carried out a systematic search using the PubMed and EMBASE databases to identify longitudinal studies published in English language journals between January 1984 and August 2010 that described the determinants of self-initiated smoking cessation among young persons aged 10–29 years, in five steps: (1) keyword searches of the PubMed and EMBASE databases; (2) scan of titles to eliminate clearly irrelevant articles; (3) review of abstracts and selection of articles for in-depth analysis and data abstraction; (4) citation search of articles retained for analysis; and (5) in-depth review of articles with data abstraction. Figure 1 describes the results of the search and the following paragraphs describe each step in more detail.
Search terms used in the PubMed database search included: (smoking (MeSH Major Topic) OR tobacco (MeSH Major Topic)) AND (cessation OR quit OR stop) AND (longitudinal OR prospective OR cohort), with the following limitations: ‘humans’, ‘English’, ‘child: 6–12 years’, ‘adolescent: 13–18 years’, ‘young adults: 19–24 years’, ‘adult: 19–44 years’ and ‘publication date from 1 January 1984 to 31 August 2010’. This search, carried out in September 2010, yielded 1468 titles. Search terms used in the EMBASE database search included: (smoking cessation (‘smoking cessation’ as keyword; ‘smoking cessation’ as subject heading) OR tobacco cessation (‘tobacco cessation’ as keyword; ‘smoking cessation’ as subject heading) OR quit smoking (‘quit smoking’ as keyword; ‘smoking cessation’ as subject heading) OR stop smoking (‘stop smoking’ as keyword; ‘smoking cessation’ as subject heading)) AND (longitudinal (‘longitudinal’ as keyword; ‘longitudinal study’ as subject heading) OR prospective cohort (‘prospective cohort’ as keyword; ‘cohort analysis’, ‘prospective study’ and ‘follow-up’ as subject heading)). Limitations included: ‘human’, ‘English language’, ‘publication year: from 1984 to current’, ‘school child: 7 to 12 years’, ‘adolescent: 13 to 17 years’ and ‘adult: 18 to 64 years’. This search, carried out in September 2010, yielded 2239 titles.
Two authors (SC, BL) screened the 3707 titles retained from these database searches to eliminate all articles that were not longitudinal, not population based, or that did not relate to the identification or description of the determinants of self-initiated smoking cessation in adolescents or young adults.
A total of 176 of the 787 abstracts retained for review after title screening were duplicates. Three authors (SC, BL, JOL) reviewed the 611 unique abstracts retained. Articles excluded during this step included sufficient information in the abstract to ascertain that the study:
Was cross-sectional or clearly not prospective (or that the exposure did not precede the outcome).
Pertained to the evaluation of a cessation intervention that was successful (in order to assure that the evidence was not obscured by exposure to a cessation intervention). We did however retain evaluation studies in which the intervention had no impact on cessation, as well as data from comparison or control groups within an intervention study that were not exposed to any intervention.
Pertained to select clinical study populations (ie, pregnant women, alcoholics, HIV positive individuals, trauma patients, drug abusers, patients with psychiatric disorders, pulmonary disease, heart disease, or in remission from a cancer). Our intent was to focus on population-based samples in order to maximise external generalisability of the findings.
Did not include data that pertained specifically to young persons aged 10–29 years.
Did not assess cessation as an outcome (ie, the outcome was prevalence of smoking, relapse, initiation of smoking, cessation attempts, reduction of smoking, maintained cessation, nicotine dependence, resistance to smoking, retraction of smoking status).
Reported qualitative data only. While qualitative data provide rich descriptions in population subgroups, the findings are not necessarily externally generalisable. In addition, qualitative data do not lend themselves easily to reportable quantitative estimates of the strength of an association.
Was based on the analysis of trajectories and/or used the transtheoretical model as a conceptual underpinning. Our intent was to study a discrete outcome (ie, cessation period of a duration of at least 6 months), which was often not explicitly identifiable in the studies that described trajectories or used the transtheoretical model.
Discrepancies between authors in the decision to eliminate an article from further review were discussed among three authors (SC, BL, JOL), and final decisions were reached through consensus.
A total of 176 articles from among the 611 abstracts reviewed were retained for review either because they clearly addressed the study topic or because the abstract did not incorporate sufficient information to allow its exclusion. Two authors (SC, BL) read each article in depth. In addition to the exclusion criteria described above, inclusion criteria applied during this step included that:
The outcome was defined as abstinence from smoking for at least 6 months.
The study assessed all potential determinants prior to cessation.
The data were collected prospectively.
The data were analysed quantitatively.
The study population was population-based.
Discrepancies between the authors in terms of which articles to retain were discussed among the three authors (BL, SC, JOL), and final decisions were reached through consensus. Of the 176 articles reviewed, 169 were excluded after in-depth review. A total of seven articles met the inclusion criteria and were retained for data abstraction.
In January 2011, two authors (SC, BL) searched all related citations from the seven articles retained for data abstraction in the PubMed database using the same limitations mentioned above (‘humans’, ‘English’, ‘child: 6–12 years’, ‘adolescent: 13–18 years’, ‘young adults: 19–24 years’, ‘adult: 19–44 years’ and ‘publication date from 1 January 1984 to 31 August 2010’). We retained 795 titles that were not reviewed during our earlier (ie, in September 2010) PubMed search. The abstracts for all 795 titles were reviewed, and 84 articles were retained for review using the exclusion and inclusion criteria described above. Of the 84 articles 2 were retained, yielding a final total of 9 articles retained for data abstraction.
To examine the evidence for specific predictors of self-initiated smoking cessation in adolescents and young adults, the following data were extracted from each article: study location, timeframe (ie, the years during which the study took place), duration of follow-up (ie, calendar time period used in the analysis), number of data collection points after baseline, sample size, age/grade of participants at baseline, number of quitters, smoking status at baseline, definition of cessation, the list of covariates adjusted for in the multivariate analyses and the analytic method. In order to facilitate comparison of results across studies, potential predictors identified within each study were stratified into five groupings: (1) sociodemographic factors; (2) psychosocial factors; (3) social influences; (4) behavioural factors; and (5) smoking related variables. For the purposes of this review, results that were marginally statistically significant (ie, p=0.057) were considered to be non-statistically significant. To facilitate interpretation, we elected to exclude results on interaction effects from three articles.12–14
Our database search identified nine articles that used longitudinal study designs to identify factors associated with self-initiated cessation in population-based samples of adolescents and young adults. Four articles described data originating from the same cohort: Project ALERT was a drug use prevention programme targeted to grades 7 and 8 students in the USA, which had no impact on cigarette smoking once the intervention was discontinued.15 These four articles identified factors related to cessation at different ages (ie, in grade 10,13 grade 12,12 at age 18,16 and at age 2317). Table 1 overviews the characteristics of the nine articles retained for analysis.
Eight of the nine studies were conducted in the USA; one was conducted in Taiwan. The study populations in all articles retained were school-based. Study populations ranged in age from 1219 to 23 years17 at baseline, and from 1419 to 29 years17 at the final follow-up. For each study, we confirmed that the relevant cessation outcome occurred prior to age 29 years. Length of follow-up ranged between 15 months20 and 14 years14 across studies. Most studies12 13 16 18 21 referred to psychosocial theory as the conceptual underpinning for the investigation and the selection of potential determinants of cessation. Specifically, these included cognitive theory,23 24 social learning theory,25 26 the theory of planned behaviour,27–29 problem behaviour theory,30 the theory of reasoned action,28 31 the social development model,32 and self-efficacy theory.23 33
The definition and method of measuring smoking differed across studies. Specifically, several studies defined a smoker either by the number of times a participant smoked per year12 13 or per month,17 or by frequency of smoking (ie, monthly,19 weekly18 and daily).14 18 Two studies16 21 defined smoking status as cigarette use during the past year. Mittelmark et al19 was the only study that validated self-reports of smoking using biological markers including carbon monoxide and saliva thiocyanate levels. Hansen et al20 did not define smoking status explicitly.
Similarly, smoking cessation was defined differently across studies. Three of the nine studies defined smoking cessation as abstinence of duration ≥6 months, and six studies defined cessation as abstinence of duration ≥12 months. More specifically, Bricker et al,18 Tucker et al,12 17 Ellickson et al13 16 and Chang et al21 specified the duration of cessation (ie, not smoked in the last year, at least 6 months of abstinence or most recent quit attempt lasted ≥6 months), while Mittelmark et al19 and Hansen et al20 reported cessation as being a non-smoker at two consecutive surveys separated by 6 months. Rohde et al14 asked participants older than age 25 if they had not smoked during the 12 months prior to reaching 25 years of age.
There was wide variability across studies in the potential determinants of cessation investigated, as well as in the definitions and methods of measuring the determinants. Most determinants were studied only once or twice across the nine studies. In addition, there were differences in the analytic approaches used to identify predictors of cessation. Most studies used logistic regression; two studies used discriminant analysis; and one study used the generalised estimating equation approach. Six of nine studies provided details about which covariates were adjusted for in the multivariate analyses. Two reports12 13 stratified the analyses by sex. Finally, the threshold between a statistically significant and a marginally statistically significant effect was not the same across studies. Thus for Tucker et al12 peer smoking (controlling for age at first use) was considered ‘marginally statistically significant’ with a p value of 0.05, while another article13 considered that peer smoking (controlling for smoking quantity) was statistically significant with a p value of 0.05.
Online table S1 summarises the univariate and multivariate results across the nine studies in terms of whether or not a specific factor was statistically significantly associated with self-initiated smoking cessation. The table, which is available at the Tobacco Control website (http://tc.bmj.com), describes the results of the nine studies retained in more detail, including ORs and p values (when available) for all variables investigated.
Despite the wide variability across studies in design features and methods, with the single exception of parental education, there were no discrepancies across studies in the direction of associations detected for those determinants that were detected as statistically significantly related to cessation. For parental education, one study13 reported that higher levels of parental education were associated with cessation in adolescents, while a second article16 suggested that lower levels of parental education were positively associated with cessation in young adults. In general, no strong patterns emerged in terms of the five groupings of potential determinants (ie, none of the groupings of sociodemographic, psychosocial, behavioural, social influences, or smoking-related factors appeared to be consistently or more or less strongly associated with smoking cessation across variables within the grouping).
Two of the nine articles12 13 carried out sex-specific analyses, and reported that females were more likely to quit smoking when they did not have friends who smoke, while males who were older at first use of cigarettes were more likely to quit. The results reported according to sex were discordant between studies. For example, parental disapproval of cigarette use was significant for males in one article only, while the reverse was true in the second. One article examined a possible interaction between gender and nicotine dependence, and reported that women were less likely to quit than men when they were nicotine dependent (data not shown).14
Table 2 summarises the number of studies that reported that a variable was statistically significantly associated with cessation, from among all the studies that examined the variable. Peer smoking was the factor most often investigated, followed by age at first cigarette use, beliefs about smoking, perceived prevalence of smoking among peers, parental support, alcohol use, education and ethnicity. Five variables were statistically significantly associated with cessation in at least three studies, including not having friends who smoke, not having intentions to smoke in the future, resisting pressure to smoke, being older at first use of cigarettes and having negative beliefs about smoking. Peer smoking was statistically significant in five of seven studies, and intentions to smoke and cigarette resistance self-efficacy were statistically significant in three of three and in three of four studies, respectively. Age at first use of cigarettes and beliefs about smoking were statistically significant in three of five studies. If marginally statistically significant effects were redefined as statistically significant, having no friends who smoke, age at first use of cigarettes and frequency of smoking became statistically significantly associated with cessation in six of seven studies, four of five studies and three of three studies, respectively.
Several factors (ie, gender, attending fewer different schools, being married, good mental health, being less rebellious or less delinquent, not using drugs, having no parents or siblings who smoke and friend support) were identified as statistically significant predictors of cessation in one or two studies, but were not statistically significant or they were not evaluated in others. Good physical health, low frequency of smoking, living with both parents, no cigarette offers and parental disapproval of cigarette use were statistically significant in two of the three or four studies which examined these factors. However, there were differences between females and males for living with both parents and parental disapproval. Among the two, three, four or five studies that examined socioeconomic status, part time work, pregnancy, perceived prevalence of smoking among peers, having exemplars who modelled cigarette use, having friends who disapproved of cigarette use, thinking that smoking has positive consequences and age, none reported a statistically significant association with cessation. The remaining variables listed in table 2 were only examined only once.
Empirical reviews generally conclude that smoking cessation programmes for young smokers have limited efficacy.34–36 This is perhaps not surprising since the determinants of cessation beyond those linked to smoking and nicotine dependence are not well understood, and the development of effective cessation approaches may be limited by incomplete understanding of the factors that influence cessation in this population. In this current review of nine longitudinal studies on the determinants of cessation in adolescent and young adults, we found evidence across studies that having no friends who smoke, having no intention to smoke in the future, resisting peer pressure to smoke, being older at first cigarette use and having negative beliefs about smoking predict longer-term cessation.
Sussman et al5 published the only similar review on this topic. Specifically his review, published in 2002, incorporated 17 longitudinal studies on self-initiated quitting published between 1972 and 2001, including 6 studies in which youth were exposed to a drug prevention programme. The results reported in the Sussman et al review agree with our results in that both identified having no friends who smoke, having negative beliefs about smoking and having no intention to smoke in the future as predictors of youth cessation. The Sussman et al review also identified lower pretest smoking and having less experience with smoking, believing that society should step in to place controls on smoking, feeling relatively hopeful about life and having parents and/or siblings who smoked, as predictors of successful cessation. Also inconsistent with our review, Sussman et al did not identify resisting peer pressure to smoke or being older at first cigarette use as significant predictors of youth cessation.
Differences in findings between Sussman et al and our review may relate to timing (ie, the Sussman et al review covered 1972 to 2001, while ours covered 1984 to 2010) and to the fact that Sussman et al did not define a minimal duration of abstinence as an inclusion criterion. Further these authors did not specify any inclusion or exclusion criteria for articles to be retained; nor did they clearly describe the method(s) or search criteria used to identify studies. Three other reviews published in 2000,7 2004,6 and 2004,8 respectively included studies from one calendar year only,6 7 and they incorporated results across cross-sectional,6–8 interventional and clinical studies.6
The predictors identified in our review seem credible as possibly causally related to cessation. Being older at first cigarette use likely relates to lower lifetime exposure to cigarettes, which may link with lower levels of nicotine dependence and in turn, with a higher likelihood of quitting in adolescence or young adulthood. Psychosocial indicators including intention to smoke in the future, resisting peer pressure to smoke and having negative beliefs about smoking align with social learning theory and may be key underpinnings of behaviour choices and useful targets for intervention. Finally friends' smoking is a strong and well established predictor of youth smoking as well as cessation, and highlights the importance of social norms in this domain.
It is notable that few studies examine nicotine dependence as a possible physiological determinant of cessation. In one of the few studies that did so, Rhode et al14 reported that young adults who were nicotine dependent were less likely to quit smoking. An analysis of possible interaction effects within the same study revealed that women who were nicotine dependent were six times less likely to quit smoking than men. Given accumulating evidence that many young smokers begin to experience symptoms of nicotine dependence such as cravings early on in the natural course of smoking onset,37 38 increased understanding of the extent to which nicotine dependence relates to cessation in novice smokers may be critical in the development of effective cessation strategies for youth.
In addition to difficulty synthesising across studies that span 1981 to 2004 and that differ in design and methods, limitations of this review include that most studies retained were conducted in the US, so that external generalisability of the findings to other countries, and in particular, to low and middle income countries, may be limited. Second, we restricted the review to population-based studies so that the results are likely not generalisable to select study populations (ie, young people who are homeless, street-involved, mentally ill, substance using, traumatised, school leavers, youth in transition, Aboriginal, low income). Four of the nine articles were based on Project ALERT. Despite possibly further limiting external generalisability, we opted to include all four ALERT articles because there were so few eligible studies available. Our review was limited to studies that report on cessation of a duration of at least 6 months rather than permanent lifetime cessation. Only two of the nine studies addressed differences in the predictors of cessation by sex, so that whether or not programmes and policy need to take sex and/or gender into account cannot be established. Finally, the age ranges covered in this review are wide. Individuals in these age ranges differ markedly in developmental stage and have widely variable life experience. Longitudinal studies may be needed to examine effect modification in the predictors of cessation by age, if the predictors of cessation are dependent on age.
The literature on the longitudinal determinants of cessation in adolescents and young adults is relatively undeveloped and subject to numerous limitations. It is likely that cessation interventions for young smokers will remain less than optimally effective until there is a solid evidence base to inform the development of effective cessation interventions for youth. Development of the evidence will comprise, among other concerns, more complete understanding of the determinants of long-term cessation in this population. In the meantime, given that many young smokers want to quit, practitioners may want to consider incorporating the results of this review on the determinants of cessation (ie, that not having friends who smoke, not having intentions to smoke in the future, resisting pressure to smoke, being older at first use of cigarettes and having negative beliefs about smoking relate to youth cessation) into cessation programmes and policy targeted to adolescents and young adults.
Funding This study was funded by the Canadian Cancer Society. JO'L holds a Canada Research Chair in the Early Determinants of Adult Chronic Disease.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.