Effectiveness of peer-led health behaviour interventions on adolescent’s mental health and wellbeing: a systematic review and meta-analysis

Literature search and screening

Database searches identified 5212 records. Following removal of duplicates, and title and abstract screening, 106 full-text articles were retrieved, of which 7 were included (see Fig. 1 for study selection and exclusion reasons).

PRISMA flow diagram showing the study selection process.

Study characteristics

Table 1 provides an overview of included studies. The included studies encompass 7,060 adolescents from 151 schools with study samples ranging from 114 to 2120 participants. Interventions were implemented in the UK^37,38,39, USA⁴⁰, Canada⁴¹ and Norway⁴² between 2010 and 2024. Six studies used a randomised controlled design, and one study used a quasi-experimental design⁴⁰. Three studies cited using an established theory (e.g., Social Cognitive Theory)^38,39,40 and three cited principles that underpinned the program (e.g., goal-directed behaviour, self-efficacy)^37,41,42. All studies aimed to improve the targeted lifestyle behaviour(s), while two studies also specifically aimed to improve mental health^42,43.

Intervention characteristics

Five interventions focused on physical activity^{37,38,39,42,43}, one on nutrition⁴¹, and one on both physical activity and nutrition³⁶. Interventions ranged from daily^38,39, twice per week^37,42, weekly^41,43, to every three weeks⁴⁰, and lasted from six to 29 weeks in duration. See Table 1 for details on intervention session length, frequency and duration.

Peer-leaders

Peer-leaders consisted of year nine students^37,43, pupils aged 15–18 years³⁸, and final year health students⁴⁰. Three studies^39,41,42 did not describe the peer-leaders. Across all studies, except Avitsland et al.⁴², peer-leaders were older than peer-recipients. The intervention by Avitsland et al.⁴² was fundamentally different in that peer-leaders were not selected and trained, but in each lesson, children organically emerged as leaders for the activity.

Selection of peer-leaders

Selection of peer-leaders was generally not well reported, however, studies reported older pupils volunteering³⁹, school staff nominating students^37,38,41, and peer-leaders being selected based on their understanding of the population, ability to provide support to peer-recipients and their role-modelling capabilities⁴⁰.

Training of peer-leaders

Training of peer-leaders was well described in one study³⁸, one study provided no detail³⁷, and the remaining studies provided a vague description of the training process. Collectively, peer-leadership training involved providing information about the research protocol, educational content for lessons, teaching methods and activities, strategies to promote engagement and motivation, and mentoring and social support principles. Time spent training peer-leaders varied from a single lunchtime session³⁹, a 1-hour session³⁷, 2 full days with 30-minute booster sessions each week⁴¹, weekly⁴³, and twice weekly for 10 weeks⁴⁰. It was unclear whether the training occurred during class time at school, during break time at school or outside of school time. Two interventions^38,39 included the use of closed social media groups to provide additional support (i.e., answer questions, give tips, provide information) and a place for discussion for peer-leaders.

Role of peer-leaders

Peer leaders across the included studies primarily facilitated health-related activities, delivered educational content and provided social support. Their responsibilities often included leading specific activities, such as guiding physical activity sessions^37,38,39,43, or running classroom-based lessons focused on health behaviours, such as physical activity or nutrition^40,41. Four studies^38,40,41,43 involved peer leaders delivering structured educational content using pre-prepared materials to promote health behaviours, and providing social support and peer accountability was explicitly mentioned in three studies^37,38,40. Adult supervision was present in five studies^{37,38,39,40,41}, with teachers or facilitators occasionally stepping in to support or oversee activities. The extent of the peer-leaders involvement varied considerably across studies, ranging from partial facilitation of activities^41,43to delivering the majority of the intervention independently^38,40.

Risk of bias and quality assessment

Six studies were considered to be at ‘high risk’ of bias^{37,38,39,40,41,43} due to deviations from the intended intervention (5/7) or not randomly allocating students (1/7). One study was considered to have ‘some risk’ of bias⁴². See Table 2 for the full risk of bias assessment for each study.

Mental health

Meta-analysis

Two studies^42,43 utilised the Strengths and Difficulties questionnaire which reports psychological difficulties, expressed as the total difficulties score, which ranges from 0 to 40 with higher scores signifying worse outcomes. This meta-analysis included a total of 1907 participants, revealing a non-significant effect (MD = 0.6, 95% CI −3.52 to 4.72; p 0.32; Fig. 2). Moderate heterogeneity was present in the analysis (Cochran’s Q = 1.47, p = 0.23; I² = 32%).

Forest plot of effect of peer-led lifestyle intervention on psychological difficulties.

Wellbeing

Meta-analysis

Two studies reported on wellbeing using two different tools – the Warwick-Edinburgh Mental Wellbeing Scale and the KIDSCREEN-27. For the latter, the psychological wellbeing subscale, which aligns with the construct of wellbeing, was used in the analysis. This meta-analysis included a total of 1297 participants. The pooled analysis revealed no effect on peer-recipients’ wellbeing (SMD = 0.00, 95% CI −2.94 to 2.94; p 1.0; Fig. 3). The analysis displayed substantial, significant heterogeneity between studies (Cochran’s Q = 12.6, p < 0.001; I² = 92%).

Forest plot of effect of peer-led lifestyle intervention on wellbeing.

Self-efficacy for physical activity

Meta-analysis

Two studies^39,40 examined the impact of peer-led lifestyle interventions on self-efficacy for physical activity, encompassing a total of 282 participants. The meta-analysis revealed a non-significant small favourable effect (SMD = 0.18, 95% CI −3.08 to 3.44; p 0.0.61, Fig. 4), with considerable heterogeneity between studies (Cochran’s Q = 4.07, p = 0.04; I² = 75%).

Forest plot of effect of peer-led lifestyle intervention on self-efficacy for physical activity.

Self-efficacy

Narrative synthesis

Self-efficacy was assessed more broadly in four studies, though studies varied substantially in both the measurement approach and the specific domain of self-efficacy assessed. Two studies focused on dietary self-efficacy^40,41, one on walking self-efficacy³⁹, and one on self-efficacy for support-seeking for physical activity³⁷. The HART Project⁴⁰ and Goals for Health⁴¹ both examined dietary self-efficacy but employed different measurement approaches, precluding quantitative synthesis. The HART Project reported significant positive changes in self-efficacy while Goals for Health found initial improvements in the intervention group (ES = 0.05), however they were not sustained at final follow-up. Notably, significant gender and racial differences were identified in dietary self-efficacy outcomes in the Goals for Health intervention, such that males reported less confidence in their ability to eat healthier than females and African Americans reported lower self-efficacy to eat healthier than Caucasian Americans⁴¹. For physical activity-related self-efficacy, the WISH study³⁹ reported no significant between-group differences post-intervention. In contrast, GoActive³⁷ focused on self-efficacy for seeking support for physical activity, reporting significant improvements in the intervention group (difference: 0.5, 95% CI: 0.1 to 0.9, p = 0.03).

Emotional Regulation

Narrative synthesis

One study examined emotional regulation outcomes through a peer-led walking intervention, reporting no significant difference³⁸.

Sensitivity analyses

Sensitivity analyses were planned but could not be conducted due to the limited number of studies included in each meta-analysis.

Publication bias

No pooled analyses included ≥ 10 studies, therefore formal evaluation of publication bias was not possible.