Behaviour change techniques in physical activity‐focused interventions for young people at risk of problematic substance use: A systematic review and meta‐analysis

This systematic review investigates behaviour change techniques in interventions promoting physical activity for young people aged 12–25 years at heightened risk of problematic substance use, and the effect of these techniques on physical activity participation and substance use outcomes.


| INTRODUCTION AND REVIEW RATIONALE
The development of substance use problems in early adulthood has been identified as risk a factor for disrupted future life trajectories (Degenhardt et al., 2016;Hall et al., 2016;McGorry et al., 2007).
Problematic substance use has its peak onset during adolescence, a developmental stage characterized by the formation of adult identity and behavioural experimentation (Aggleton et al., 2006;Helzer et al., 1991).Further, adolescence and early adulthood are marked by a decline in adaptive behaviours, such as engagement in physical activity, with young people's physical activity levels consistently falling below international recommendations (Allison et al., 2007;Finne et al., 2011).Consequently, global research has increasingly focused on strengthening and promoting a variety of health-promoting behaviours to improve the life trajectories of young people and mitigate the long-term consequences of problematic behaviours (Linke & Ussher, 2015;Lynch et al., 2013;Smith & Lynch, 2012).
Physical activity behaviour is associated with substance use behaviour (Kwan et al., 2012;Linke & Ussher, 2015;Moore & Werch, 2005), namely an increase in physical activity appears to be associated with lower levels of illicit substance use (Linke & Ussher, 2015).Complex behavioural interventions aim to simultaneously change multiple health behaviours by strategically applying behaviour change techniques targeting explicit and implicit health management (An et al., 2013;Bourke et al., 2022;Brug et al., 2005;Michie et al., 2011).One example is interventions aiming to increase physical activity levels while simultaneously reducing problematic substance use (An et al., 2013).These interventions are based on research indicating that increased physical activity may yield a decrease in reported substance use outcomes in youth (Kirkcaldy et al., 2002;Linke & Ussher, 2015;Moore & Werch, 2005;Simonton et al., 2018) and studies using physical activity prescriptions to both increase physical activity and decrease substance use (Cabrera, 2020;Gustavsson et al., 2018;Kallings, 2016).(Michie et al., 2011;Michie et al., 2013;Teixeira et al., 2020;West et al., 2019).By doing so, Michie et al. aim to increase the possibility of identifying effective components within interventions, enhance their replicability and improve both implementation and evidence synthesis (Michie et al., 2011).

| Behaviour change
The CALO-RE taxonomy, an extension of a previous taxonomy (Abraham & Michie, 2008), identifies 40 different behaviour change techniques (BCTs) extracted from numerous international publications that aim to increase guideline-concordant physical activity behaviour.
For more details about the CALO-RE taxonomy, see table 3 in Michie et al. (2011).
A recent review reported favourable outcomes (e.g.reduction in alcohol, opiate and stimulant use) in adult populations for the application of the following BCTs with physical activity interventions targeting substance use: instruction on how to perform the behaviour, social support (unspecified), behavioural practice/rehearsal, problem-solving, pharmacological support (i.e. for smoking cessation, as detailed by Hartmann-Boyce et al. (2018)), goal setting (behaviour), self-monitoring (behaviour) and biofeedback (Thal et al., 2022).A meta-analysis was not performed in this review, and it remains unclear if the use of these BCTs in interventions has an effect on younger populations.

| METHOD
The review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (http://www.prisma-statement.org)and the APA Quantitative Meta-Analysis Article Reporting Standards (MARS) (Appelbaum et al., 2018).

Publications reporting on interventions for young people at risk
of problematic substance use that promote physical activity and report on substance use and physical activity participation were considered.Doctoral publications, non-English, protocols, prospective publications and abstracts were excluded.
A review protocol was registered before the literature search, and there were no deviations from this protocol.Analysis and extraction of behaviour change techniques were based on the CALO-RE taxonomy described by Michie et al. (2011).
Studies reporting on single BCT interventions were compared to studies reporting on multiple BCT interventions.This approach was consistent with previous research conducted by Michie et al. (2009), investigating interventions that applied very few BCTs to interventions that applied a larger number of BCTs in low-income countries.
Findings suggested that different numbers of BCTs may yield different effects on outcomes (Michie et al., 2009).
A risk of bias assessment was completed in duplicate in Covidence for all included studies according to the Cochrane RoB2-tool (Higgins, Savovi c, et al., 2021) for randomized controlled studies (RCT) and Cochrane ROBINS-I tool (Sterne et al., 2021) for non-randomized studies (NRS), as well as a GRADE rating for determining the certainty of evidence (GRADE Working Group., 2004;Guyatt et al., 2008).
The comprehensive software [("Comprehensive Meta-analysis" (version 3) (www.meta-analysis.com)]was used to calculate outcome effects according to a random effects meta-analysis based on reported numerical data (pre/post means, standard deviations (SD), mean changes, sample sizes).Analyses of publication bias, sensitivity and subgroup effect (a priori planned and unplanned) were undertaken for all studies; manual conversions were performed where necessary.Studies were excluded if reported data could not be converted into numerical data relevant for synthesis and raw data could not be obtained from study authors.
Publication bias was explored using funnel plots.Studies with inadequately reported outcomes and missing data that could not be retrieved through manual calculations or contacting respective authors were excluded from the analysis.

| Openness and transparency
According to MARS, the review authors determined the study selection, all data exclusions (if any), manipulations and measures in the review (Appelbaum et al., 2018).Data and research materials are available upon request.

| RESULTS
From 5427 identified records, 28 studies describing a range of different study designs (randomized controlled trials, non-randomized controlled trials, cluster and crossover designs) were included (Figure 1) The mean age of the included population was 20.7 years.Assessed substances included alcohol, tobacco and illicit substances, such as methamphetamine.For exclusion reasons of studies see supplemental list of excluded studies.

| Behaviour change techniques
Behaviour change techniques for increasing physical activity behaviour were extracted from the included studies.The most frequently used BCT was providing instruction on how to perform behaviour.This strategy was reported in 18 of the included studies.
The strategies providing information on the consequences of behaviour in general and goal setting (referring to the behavioural processes of setting goals, rather than the outcome) were reported by 11 of the included studies.Another frequently reported strategy was facilitating social comparison to change individual participants' behaviour.The eight most reported BCTs according to their frequency can be found in Table 1.A full list of extracted BCTs, participation rate and reported meintance of behavior change is provided in Table 2.

| Risk of bias within the included studies
According to the risk of bias assessment for RCTs, the studies and study domains listed in Tables 3 and 4 were rated as high risk.
Based on the risk of bias assessment, the level of evidence and thus confidence in the range of an effect estimate (Schünemann et al., 2021) was rated high for included RCTs, however, the level of evidence was downgraded to low for all included non-randomized trials (NRS).
T A B L E 1 Most frequently reported behaviour change strategies in included studies according to CALO-RE taxonomy.

| Results of syntheses according to BCTs
Several meta-analyses were computed focusing on BCTs.To ensure comparability, only the most common clinical measures and final time points were used for calculations.Where necessary, multiple comparison groups were combined into single groups, with missing data manually computed or imputed from external sources according to Cochrane recommendations (Higgins, Thomas, et al., 2021) using Matlab, version R2021a, (www.mathworks.com,Natick, USA).

| BCTs and other outcomes
No subgroup analyses based on the number of BCTs could be computed for alcohol use outcomes (all studies in this subgroup comparison were high in reported BCTs), illicit substance use (no indication of heterogeneity) outcomes or physical activity (all studies were high in reported BCTs).

| DISCUSSION
This review explored the effect of BCTs aiming to increase engagement in physical activity on reducing substance use among young people aged 12-25 years at risk of problematic substance use.Note: Confounding: S = Clusters not comparable, potential confounding.Allocation: M = Allocation according to order in which participants presented (Everson et al., 2006); M = Allocation according to predefined characteristics; M = Unclear allocation (Scott & Myers, 1988); S = Students self-selected into study/intervention group (Tesler et al., 2018).Missing outcome data: S = High drop-out rate (Scott & Myers, 1988).Outcome assessment: C = Exercise intensity not measured (Everson et al., 2006); S = Subject to social desirability bias (Tesler et al., 2018).Outcome reporting: S = Only significant measures reported (Scott & Myers, 1988).Other sources of bias: S = Significant baseline differences and crossover effects, prospective study, no clear adjustments; C = Big difference in N between intervention/control group, potential conflict of interest (Tesler et al., 2018).(Michie et al. (2011).Of these, providing instruction on how to perform the behaviour was the most frequently used BCT (reported in 18 studies) to increase physical activity.Other frequently reported BCTs included providing information on consequences of behaviour in general, goal setting (behaviour), facilitation of social comparison and providing information about others' behaviour.
In addition to identifying and extracting BCTs that aim to increase physical activity, the effect of the number of reported BCTs on the post-intervention substance use and physical activity outcomes was reported to investigate the potential impact of using BCTs as part of physical activity interventions.Fourteen studies had sufficient information to be included in the meta-analysis.Analyses revealed a significant effect of BCTs on the observed pre-post behavioural differences in combined substance use outcomes (frequency of use, craving, intent of use).This effect was accounted for by interventions that reported a single BCT and interventions that reported multiple BCTs.The behaviour change technique most frequently used in interventions that were low in reported BCTs was providing instructions on how to perform the behaviour.
Results based on the subgroup analyses, according to reported BCTs need to be interpreted with caution (Deeks et al., 2021).Subgroups may contain different interventions (and different amounts of information) and thus have different abilities to detect effects (i.e. an effect may be detected more readily in one subgroup than another).
Assumptions that a significant subgroup factor explains observed heterogeneity may therefore be premature (Deeks et al., 2021, chapter 10.11.3).tions (Michie et al., 2009, p. 612), indicating that changes in behaviours may be difficult to maintain.

| Subcategories of substance use
Subgroup analysis according to the number of reported BCTs was only possible for tobacco use outcomes, due to significant heterogeneity in the findings for the other outcomes.Interventions reporting a single BCT showed a significant effect on the pre-post observed differences in tobacco use outcomes.In contrast, the category of interventions with multiple BCTs did not.One explanation for this is intervention duration may be a moderating factor, given that interven-  et al., 2021).
We found that among the most frequently reported BCTs is social comparison.In previous research, social comparison has been identified as a predictor of health behavioursuch as nutritional intake and physical activityamong adolescents (Luszczynska et al., 2004;Patel et al., 2016), especially when associated with additional financial incentives.However, Arigo et al. (2020) et al., 2020;Gardner et al., 2010).
Compared to a review investigating the use of BCTs in physical activity interventions for adults with substance use disorder (Thal et al., 2022), our review found differences in commonly reported BCTs in the included studies.While there was an overlap in identified  One limitation is that evidence synthesis was based on reported BCTs, which may not have captured the BCTs that were actually applied within the interventions.This highlights the importance of adequate and thorough reporting of applied BCTs as part of intervention descriptions to allow confident evidence synthesis.
Notably, this review investigates the effect of reported BCTs aiming to promote physical activity on substance use outcomes.An earlier review investigates the overall efficacy of interventions that promote physical activity (Klamert et al., 2023).To facilitate the examination of the effect of reported BCTs, different substance use outcomes were combined (frequency of use, cravings, intent to use).This poses a notable limitation to this review and is not advisable for other research questions.
Further, while analyses based on extracted BCTs explain part of the observed variance in outcomes, the heterogeneity in the established subgroups based on the number of BCTs remains significant, indicating the presence of other confounding or moderating variables which have not been identified.This may partly be explained by the general large heterogeneity in the included studies regarding intervention design, setting, outcome measures and physical activity promotion, as noted previously by Thal et al. (2022).
Last, the quality of included evidence, which had to be downgraded for several domains due to the risk of bias in the included studies, poses a limitation to the existing evidence base.
Michie et al. have developed taxonomies (i.e.classification systems) to identify and organize behaviour change techniques and their associated mechanisms of change Klamert et al. (2023) found evidence for beneficial effects of interventions that promote physical activity on alcohol, tobacco and illicit substance use outcomes in young people aged 12-25 years at heightened risk of problematic substance use.These interventions improved tobacco use, alcohol use and illicit substance use outcomes, including reducing intentions and/or cravings to use substances, and increasing physical activity participation.This current review reports on the behaviour change techniques that were applied in studies included in the Klamert et al. (2023) review.Accordingly, the current study aimed to: 1. Describe behaviour change techniques reported in interventions that aim to promote or increase physical activity in young people at risk for problematic substance use, according to the CALO-RE taxonomy; and 2. investigate the effect of the number and type of reported behaviour change techniques included in interventions on substance use outcomes and physical activity in young people at risk for problematic substance use.
The study search was completed in January 2021 and updated in November 2022, using four databases (PsycINFO, CINAHL, SPORT-Discus and Medline), reference lists of relevant publications and existing systematic reviews.Indexed and free text terms were included via Boolean operators.Abstract and full-text screening, study selection and inclusion were performed in duplicate by researchers experienced in systematic review methods in physical activity and substance use.Studies were considered according to the following inclusion criteria: the studied population was young people (12-25 years) at risk for problematic substance use (ie., by virtue of low socioeconomic status, socioeconomic disadvantage, highrisk behaviour and comorbid mental illness (Klamert et al., 2023)); intervention type was interventions that promote physical activity; studies included a control group (active or inactive); and studies reported on substance use outcome measures.Studies were eligible for inclusion if interventions either solely focused on physical activity promotion or used physical activity promotion among other intervention elements.The screening was conducted using Covidence (Extraction version 2.0, www.covidence.org,Melbourne) Intervention and participant characteristics, reported behaviour change techniques and the outcomes of primary (substance use outcomes) or secondary interest (physical activity engagement levels) were extracted by one reviewer and checked by a second reviewer.All reported time points and measures were extracted, including the different measures used to assess a single outcome within a study.Data were reported as weighted, standardized effect sizes (Hedge's g).
Interventions included trials involving light to vigorous exercise and interventions with either physically active elements or physical activity-promoting messages or education, all of which were within the scope of 'physical activity promotion' interventions.Twenty-eight studies were included in the narrative section of the review, fourteen of which provided sufficient information to be included in the meta-analysis.A total of 27 different BCTs were identified within the interventions according to the CALO-RE taxonomy T A B L E 4 Risk of bias assessment for included NRS (non-RCTs) (Cochrane ROBINS-I).
with a single BCT (single session or short-term interventions) and interventions with multiple BCTs (longer-term interventions) demonstrated a significant effect of physical activity promotion on young people's combined substance use outcomes, with a significantly larger effect reported for single BCT, shortterm interventions.These findings are consistent with previous research suggesting that the effects of behaviour change interventions may dilute over time, which is why large effects are commonly observed in short-term interventions rather than long-term interven- note that social comparison may only be an effective facilitator of change under some circumstances.Underlying behaviour change theories are rarely acknowledged in intervention descriptions, and the application of social comparison as a behaviour change technique to increase physical activity often remains unclear.This is partly due to the absence of information, such as the dimensions or domains relevant for comparison, which creates confusion as to how to best apply this BCT and prevents a clear understanding of the contexts in which this BCT seems to work effectively.Arigo et al. (2020) thus propose that response variability should be investigated according to behaviour change theory and used to increase tailoring and thus the effectiveness of promoting physical activity through social comparison (Arigo BCTs (instruction on how to perform behaviour, goal setting (behaviour), self-monitoring (behaviour)),Thal et al. (2022) additionally highlighted social support (unspecified), behavioural practice/rehearsal, problem-solving and pharmacological support as promising techniques for adults.Our review found social comparison, providing normative information and information about behavioural consequences and modelling the goal behaviour as additional commonly reported BCTs to increase physical activity behaviour.The differences in identified BCTs can be explained by the different target populations (adults versus young people), intervention type and taxonomy used (general BCT taxonomy versus taxonomy of BCTs aiming to specifically increase physical activity).For example, pharmacological support has previously been more likely to be used in adult populations and has only in recent years moved into the research focus for younger populations (Hartmann-Boyce et al., 2018; Mann et al., 2014; Squeglia et al., 2019).Similar to Michie et al. our review further demonstrated that both low and high levels of reported BCTs seem to have a significant effect on substance use outcomes.Dombrowski et al. (2012) have also shown in a previously conducted meta-regression that increased numbers of BCTs are not always associated with better outcomes and Arigo et al. (2020) have pointed out that BCTs may work under some circumstances but not others and thus rely on situational context, indicating that the type of BCT may be more critical than the number.Michie et al. (2009) also point out that using a larger number of BCTs, compared to a smaller number, does not automatically imply a larger effect.The limited body of existing research on physical activity-focused interventions for young people at risk of problematic substance use is characterized by large heterogeneity.Theory-based evidence syntheses are crucial for informing and developing policy and practice.As Gardner et al. (2010) posit, research conceptualization, categorization of interventions and evaluation of intervention components according to explicit behaviour change theory could facilitate theoretical coherence in the large heterogenous body of research evidence, which in turn is critical for responsibly informing policy and practice change.

4. 3 |
Strengths and limitationsA strength of this systematic review is the identification and extraction of BCTs reported in interventions to increase levels of physical activity and improve substance-related outcomes use in young people.The current study is the first, to our knowledge, to report on the effect of BCTs in the context of substance use outcomes in young people aged 12-25 years.This synthesis enables comparison across interventions, adds relevant knowledge to the existing body of evidence and can further provide meaningful directions for policy, practice change and service improvement beyond single intervention studies.F I G U R E 1 Prisma flow diagram of study selection (see also Klamert et al., 2023).F I G U R E 2 Effect of reported number of BCTs in interventions that promote physical activity on substance use outcomes.This review is also the first to extract BCTs in this setting based on the internationally established CALO-RE taxonomy, which provides a suitable framework to compare the findings of this review with other reviews investigating health behaviour change, such as de Bruin et al. (2009) and Jacobs-van der Bruggen et al. (2009).This taxonomy sets a scientific basis for exploring the effect of different combinations of BCTs in varied contexts to improve and increase behavioural responses in this population.

4. 4 |
Implications, recommendations and directions for future research This study provides insight into commonly used BCTs in physical activity and substance use, as well as the importance of considering BCTs as part of health interventions and understanding related mechanisms of change.Nevertheless, the circumstances under which different BCTs influence outcomes vary and remain poorly understood.Consequently, more research is recommended to investigate and apply different BCTs, targeting specific health behaviour, in various contexts.Additionally, guideline or template development for reporting BCTs delivered within behaviour change interventions is recommended to facilitate comparability among interventions, ensure high-quality evidence synthesis and strengthen policy development.Directions for future research and important factors to include in conceptualizing and designing future studies include the long-term behavioural change effects of short-term (1-2 sessions) interventions, considering that behaviour change interventions often experience a time-dilution effect over time (Michie et al., 2009).Other recommendations for research include the investigation of the superiority of distinct combinations of BCTs, and project resources and researcher preferences associated with BCT choice.Additionally, knowledge translation and implementation should be considered within this research area.Namely, effectiveness studies should include a focus on implementation in substance use and/or mental health services and educational settings, should the intervention be effective, to allow knowledge translation and implementation of effective BCTs into practice settings.5 | CONCLUSION Application of BCTs has a significant effect on post-intervention substance use outcomes in young people aged 12-25 years at risk of problematic substance use.While applying several BCTs within an intervention demonstrates a small, significant effect on outcomes, short-term (1-2 sessions) interventions using only single demonstrated a larger effect on post-intervention substance use outcomes.These findings should be interpreted with caution.While certain combinations of BCTs may be superior to others, favourable behavioural responses to reported BCTs are likely to be context-dependent, suggesting more research is needed.The findings support the benefits of applying BCTs in interventions that promote physical activity in young people at risk of problematic substance use.Future research should improve the description of BCTs within interventions to better inform clinical decision-making and public policy.
Behaviour change techniques and implementation strategies in included studies.
Baseline sample N = 104, but 19 drop-outs after pre-testing.Final sample N = 85.15 Students in intervention group excluded as they attended <8 sessions.
Risk of bias assessment for included RCTs (Cochrane RoB2).
T A B L E 3