Information sources and search strategy

We searched MEDLINE, PubMed, PsycINFO, CINAHL Plus and Scopus as sources of academic databases and Google Scholar as a source of grey literature. We conducted our initial literature search in September 26, 2022, but later updated our search in October 20, 2023 by using various combinations of medical subject headings (MeSH terms) and keywords (e.g., “smoking cessation”, “cigarette*”, “quit”, “abstinence”, “predict*”, “risk factors”, “correlate*”, “associate*”, “emerging adulthood”, “youth*”, “adolescent*”, and “young adult*”) (see S1 Appendix). From the initial Google Scholar search, first 200 results were considered for title and abstract screening, because we did not get any relevant article following this threshold. Later, we additionally screened first 100 results of the most recently published articles from the updated Google Scholar search. All search results were further limited to review articles on humans and English language papers. We also searched the reference lists of the included articles. One reviewer (AK) conducted both academic and grey literature search and imported the articles to the Covidence workflow platform where duplicate articles were automatically removed. Another reviewer (NS) searched the reference lists of the finally included articles looking for any potential paper which matched the eligibility criteria.

Eligibility criteria

We included any systematic reviews which summarized factors possibly influencing smoking cessation related behaviors including intention to quit smoking, quit attempts and smoking abstinence among young people between age 10 and 35 years. Hence, other types of literature reviews including scoping and narrative reviews were excluded. Although World Health Organization (WHO) has defined people between age 10 and 24 years as adolescents and young adults [17], several studies identified people up to age 35 years as young adults [18–20]. Moreover, previous research has shown that quitting smoking by 34 years significantly lowers smoking mortality and morbidities [4]. Hence, we decided to expand our target population to people between the ages of 10 and 35 years, so that the findings of this study could be applicable for a broader group of young population. We also included reviews where the target population included people between 10 and 35 years as sub-population, or the review assessed predictors of smoking cessation as one of the various other outcomes. We excluded systematic reviews on 1) efficacy of smoking cessation interventions; 2) predictors of adverse health effects of smoking; 3) predictors of smoking prevention and other smoking behaviors such as smoking initiation, escalation, and smoking prevalence; 4) predictors of other tobacco products (OTPs) cessation; and 5) predictors of cessation of dual use of cigarettes and e-cigarettes or polysubstance use. We excluded reviews that assessed the efficacy of smoking cessation interventions because most of the studies included in these types of reviews were randomized controlled trials [21,22], which are most suitable for examining cause and effect relationship. However, rather than identifying which intervention will be more effective for causing smoking cessation, we intended to summarize the individual and environmental risk factors that influence young peoples’ smoking cessation related behaviors.

Selection process

Two reviewers (AK and NS) independently screened each title and abstract in accordance with the inclusion and exclusion criteria. Next, each full text of the remaining articles was reviewed by the two reviewers (AK and NS) and any disagreements were resolved by further discussions among them.

Data collection process, data items, and effect measures

We used a custom-made data extraction form to extract relevant data on general characteristics of the included reviews (author and year, number of databases searched, search period, number of included studies, study design of included studies, age group, primary outcome, number of predictors identified, key findings, special features and critical appraisal score). For any given review, we also extracted data on all predictors (both statistically significant and insignificant) of smoking cessation related behaviors reported in the review. The included reviews reported statistical significance of the predictors using either a narrative data synthesis approach, or by considering p-values < 0.05 or 95% confidence intervals reported in their included studies. We also extracted data on the direction of association of the predictors as reported in the reviews, which were either positive or negative or inconsistent. One particular review [12] did not report the direction of association of their identified predictors clearly. We contacted the corresponding author enquiring about details of the findings but did not receive any responses. Hence, we documented the direction of association for these predictors as unclear. For the reviews which assessed predictors of different cigarette or tobacco use related behaviors, we limited our data extraction to the studies included in their review that evaluated only smoking cessation behaviors. While one reviewer (AK) extracted the data from all final included reviews, another reviewer (NS) checked for accuracy of the extracted data. Disagreements were resolved by discussion between the reviewers.

Synthesis methods

We grouped outcomes in the included reviews into two categories: intention to quit smoking and smoking abstinence/quit attempts. While intention to quit smoking was indicated by future intention to quit among current smokers, smoking abstinence was usually defined by the reviews as successfully quitting smoking for a defined time-period such as 6 months. Some reviews used the term ‘smoking cessation’ as their outcomes which was non-specific and could be interpreted as either smoking abstinence or previous quit attempts. So, we combined these outcomes with smoking abstinence and made a category named smoking abstinence/quit attempts. Most of the included reviews used broader age groups as target population. Hence, we made three categories of overlapping age groups and summarized the predictors of smoking abstinence/quit attempts for individual age groups.

By following the categorization method used in a previous systematic review [14], each identified predictor was further categorized into the following groups based on the frequency of their reported significant association in the reviews: 1) probable predictor (the predictor was reported by at least 2 reviews and found statistically significant in at least 50% of the reviews); 2) possible predictor (statistical significance was observed in only one review or <50% of the total reviews); 3) insufficient evidence (when the predictor was reported by only one review and no statistically significant association was found); 4) probably unrelated (no statistically significant effect identified in at least 2 reviews). We determined the direction of effect for each individual predictor as positive or negative based on the reported direction of association observed in majority of the reviews. For any given predictor, if an unclear association was identified for one particular review [12], we ignored it and considered the direction of association found in other reviews while determining the final direction of effect. If a predictor with unclear direction of association was examined by a single review, we grouped it as ‘insufficient evidence’ factor. An additional category- ‘inconsistent direction’ was added to the previous 4 categories to represent the predictors for which 1) equal number of reviews reported positive and negative associations; or 2) equal number of reviews reported positive/negative and inconsistent associations. Data synthesis was conducted by one reviewer (AK) and checked for accuracy by another reviewer (NS).

Quality assessment

For quality assessment of the included reviews, we used the Joanna Briggs Institute (JBI) critical appraisal checklist for systematic reviews [23]. We choose the JBI critical appraisal checklist instead of other critical appraisal tools (e.g., A measurement tool to assess systematic reviews (AMSTAR) [24] or Risk of bias in systematic review (ROBIS) [25]), because the applicability of this tool is wide and more appropriate for systematic review of non-experimental studies [26,27]. It contains an 11-item checklist, where for each item appraised, we assigned a score of 1, if the criterion was met, and 0, if the criterion was not met or was unclear. We modified the 11-item checklist to a 10-item checklist for selective reviews of qualitative studies, because item no. 9 was not applicable for these type of reviews [23]. Finally, all critical appraisal scores were reported as a percentage of the assigned total numerical scores instead of individual points. We followed a recent overview [28] to present our results for the JBI critical appraisal checklist and determined an overall score of at least 70% indicating satisfactory methodological quality. Two reviewers (AK and NS) independently scored all reviews, and any disagreement was resolved by discussion.

Certainty assessment and overlapping between the reviews

As the included reviews did not conduct any meta-analysis and mostly presented their findings by narrative data synthesis, we used the Confidence in Evidence from Reviews of Qualitative Research (GRADE-CERQual) approach [29] for assessing certainty or confidence in the body of evidence. Although the GRADE-CERQual [29] is mainly built for conducting systematic review of primary studies, we followed similar approach applied in previous research [30] and recommendations on adapting existing tools for assessing certainty of the evidence in an overview [31]. The certainty assessment in GRADE-CERQual is based on four components- methodological limitations, coherence, adequacy of data, and relevance [29]. We evaluated the confidence in evidence on categories of the predictors for each individual outcome by assessing whether the involved reviews had any concerns regarding these four components. Based on the judgements made for each of the four components, an overall assessment of certainty or confidence was made as high, moderate, low, or very low confidence [29]. Two reviewers (AK and NS) independently provided their judgements, and any disagreement was resolved by discussion.

As the finally included reviews addressed the same research question (e.g., population, exposure, and outcomes) and search period of these reviews overlapped with each other, we anticipated that some primary studies were included in multiple reviews, which might introduce bias in the results [32]. Hence, we measured the degree of overlapping between the reviews by using the corrected covered area (CCA) method as directed in the Pieper et al., 2014 [32]. Covered area (CA) and CCA are calculated by:

Here, N is the number of total studies across all reviews (including multiple counting of the same study), r is the number of primary studies and c is the number of reviews. The CCA value 0–5 indicates slight overlap, 6–10 moderate overlap, 11–15 high overlap and >15 means very high overlap between the reviews [32].

Although one of our included reviews [12] reported including total 51 studies in the final review, we could only retrieve 18 studies from their reference list. Our efforts to get the full reference list failed to get any responses from the authors of the review [12]. Hence, we counted only the retrieved 18 original studies for measuring the final CA and CCA measures.

Study selection

The search of academic electronic databases yielded 999 articles. Following removal of 307 duplicates, a total of 692 articles were reviewed by title and abstract screening. An additional 300 articles were screened through Google Scholar search and 1 article was added from searching reference lists of the final included articles. Finally, a total of 24 articles were selected for full-text screening. After removing 13 articles for various reasons (Fig 1), we included total 11 articles [11–14,33–39] in the final review. The detailed selection process of the articles is presented in the PRISMA flow diagram (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow diagram demonstrating selection of studies.

https://doi.org/10.1371/journal.pone.0299728.g001

Study characteristics and results of individual studies

The summary statistics of the included studies and general characteristics of individual reviews are presented in Tables 1 and 2 respectively. Out of 11 final systematic reviews [11–14,33–39], only 2 [12,33] were published before 2010, the remaining 9 were published between 2011 and 2022. Only 2 reviews [35,37] assessed predictors of intention to quit smoking, while all reviews [11–14,33–39] identified predictors of smoking abstinence/quit attempts. Four [13,34,36,37] out of 11 reviews included only 1–5 studies on smoking cessation in their review, 2 reviews [35,39] included 6–10 studies, 2 reviews [11,33] included 11–20 studies, and 3 reviews [12,14,38] included >20 studies. The highest number of studies included in one review [12] was 51 studies, however, only 18 of them were retrievable. The second highest number of studies included in one review [38] was 39. Five reviews [13,34–37] detected <20 predictors, while 4 reviews [11,12,33,38] detected 21–50 predictors and 2 reviews [14,39] detected >51 predictors. The highest number of predictors identified by a single review [14] was 67 (Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary statistics of included reviews (N = 11).

https://doi.org/10.1371/journal.pone.0299728.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. General characteristics of all included reviews (N = 11).

https://doi.org/10.1371/journal.pone.0299728.t002

Synthesis of results on predictors of smoking cessation

We identified a total of 123 predictors, of which 4 were predictors of intention to quit smoking and 119 were predictors of smoking abstinence/quit attempts. There was no ‘probable’ predictor for intention to quit smoking, because the low number of studies (n = 2) examining this outcome focused on completely different factors. Hence, it did not meet our criteria of ‘probable’ predictors and we categorized these 4 factors as ‘possible’ predictors. Among them, self-recognizing smoker identity, and interactions between smoker identity and other identity like being a mother increased the likelihood, while positive risk acceptance with smoker identity decreased the likelihood of having intention to quit smoking among 16–34 years age people [35]. Additionally, using flavored tobacco was found to lower the probability of future intention to quit smoking among 12–18 years age group [37].

Tables 3–5 and S2 Appendix presents the distribution of 119 predictors of smoking abstinence/quit attempts by 3 overlapping age groups- ~12–21 years, ~16–34 years, and ~10–29 years. While the predictors for the first two age groups were very small in number, the four reviews [14,33,38,39] conducted on the ~10–29 years age group identified on an average of 46.5 predictors. No ‘probable’ and ‘probably unrelated’ predictors were found for ~12–21 years age group, while 22 ‘possible’ predictors, 3 predictors of ‘insufficient evidence’, and 1 predictor of ‘inconsistent direction’ were identified. Among the predictors identified for the ~16–34 years age group, 2 were ‘probable’, 6 were ‘possible’, 17 had ‘insufficient evidence’, and 2 were ‘inconsistent direction’ predictors. The 2 ‘probable” predictors were self-efficacy/confidence in quitting and tobacco price increase, both were found to have positive direction of association with smoking abstinence/quit attempts (Table 3). No ‘probably unrelated’ predictors were found for this age group.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. ‘Probable’ predictors of smoking abstinence/quit attempts by different age groups.

https://doi.org/10.1371/journal.pone.0299728.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. ‘Possible’ predictors of smoking abstinence/quit attempts by different age groups.

https://doi.org/10.1371/journal.pone.0299728.t004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. ‘Insufficient evidence’, ‘probably unrelated’ and ‘inconsistent direction’ predictors of smoking abstinence/quit attempts by different age groups^*.

https://doi.org/10.1371/journal.pone.0299728.t005

For the ~10–29 years age group, we detected 29 ‘probable’, 39 ‘possible’, 26 ‘insufficient evidence’, 2 ‘probably unrelated’, and 14 ‘inconsistent direction’ predictors of smoking abstinence/quit attempts. The ‘probable’ predictors included a wide variety of factors including socio-demographic factors, nicotine dependence, attitudes regarding smoking, mental health and psychological factors, behavioral factors, peer and family related factors, and policy related factors. (Table 3). The socio-demographic factors- older age at smoking initiation, higher education, married/living with partners, and pregnancy/becoming parent had positive association with smoking abstinence/quit attempts, while factors related to dependence- level of nicotine dependence, cravings, frequency, and intensity of smoking had negative association. Higher self-efficacy/self-confidence in quitting increased the likelihood, but having pro-smoking attitudes, and susceptibility to smoking reduced the likelihood of smoking abstinence/quit attempts. Among the mental health and psychological factors, good self-management skills had a positive association, and stress level and depression had negative association. Additionally, having good self-perceived general health increased the probabilities of smoking abstinence/quit attempts, while previous history of cigarette use, alcohol or other substance use lowered the likelihood of smoking abstinence/quit attempts. Among the peer and family related factors, parental monitoring, parental or peer support, and parents/family members quitting, friends/peers quitting were positive predictors, and family/household smoking, peer smoking, and social acceptability of smoking were negative predictors. Additionally, increasing tobacco prices, restricting cigarette availability, and ban on cigarette coupons were found to have positive associations with smoking abstinence/quit attempts (Table 3).

Quality assessment findings

The average critical appraisal score of the 11 reviews was 71.4% with a standard deviation of 156%, and 95% confidence interval of 60.6% to 82.1% (Table 1). Seven reviews [11,13,34–38] had a total critical appraisal score of at least 70%, indicating satisfactory methodological quality, while two [13,38] reviews were considered to have a score of at least 90% (Table 2). Three reviews [14,33,39] had a critical appraisal score between 50 and 69% and one review [12] had a score of less than 50%. None of the reviews assessed the likelihood of publication bias in their review. Thirty percent (n = 3) of the reviews used adequate sources for searching studies and 45% (n = 5) of the reviews had an appropriate search strategy. Seventy-three percent of the reviews had conducted appropriate critical appraisals by two or more reviewers. Approximately 80% to 90% of the reviews had clear and explicit research questions, appropriate inclusion criteria, adequate methods of data extraction and data synthesis. All reviews provided recommendations for policy and/or practice based on their findings and direction for future research (see S1 Table).

Certainty assessment and overlapping between the reviews

Tables 6 and S2 of the Supporting Information summarizes the findings and evidence profile on individual components of the GRADE-CERQual certainty assessment. Overall, we had moderate confidence on the ‘probable’, ‘possible’, ‘insufficient evidence’, and ‘inconsistent direction’ predictors of smoking abstinence/quit attempts. However, our confidence on the ‘probably unrelated’ factors of smoking abstinence/quit attempts and 4 ‘possible’ predictors of intention to quit smoking were low.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. GRADE-CERQual summary of findings on certainty of the evidence.

https://doi.org/10.1371/journal.pone.0299728.t006

We identified a total of 138 primary or index publications out of the 11 reviews [11–14,33–39] (see S3 Table). We found only 7 primary studies which were included in multiple reviews. The percentage of overlap was 0.05%, CA was 0.10 and CCA value was 0.01. Hence, the degree of overlap between the reviews was slight or minimal.