Six-month follow-up of the ‘Mi Mejor Plan’ school-based prevention program: a pilot cluster randomized controlled trial among early adolescents in Chile

Abstract

Background

School-based preventive interventions for adolescent substance use often show limited short-term behavioral effects, with evidence suggesting that sustained impacts may depend on continued exposure and reinforcement strategies such as booster sessions. However, few studies have examined the longitudinal effects of booster implementation on substance use and related psychosocial outcomes in culturally adapted interventions.

Methods

We conducted a cluster randomized controlled trial to evaluate the longitudinal impact of a culturally adapted school-based prevention program among sixth-grade adolescents (mean age 11.3 years) in Chile. Schools assigned to the intervention received a 16-session curriculum and three booster sessions, whereas control schools continued with treatment-as-usual prevention activities. Assessments were conducted at baseline, immediately post-intervention, and approximately six months later following completion of the booster sessions. Primary outcomes were substance use behaviors (30-day prevalence of tobacco, alcohol, and cannabis), while secondary outcomes included cognitive and social determinants. Mixed-effects models were used to assess changes over time, accounting for clustering at the school level.

Results

A total of 765 students from 11 schools were enrolled, with 608 assessed at baseline and 426 at follow-up. While no significant reductions in substance use were observed immediately post-intervention, longitudinal analyses indicated improvements in selected secondary outcomes, including more negative beliefs about tobacco and alcohol and enhanced decision-making processes. Several cognitive and social determinants remained improved at the assessment conducted after completion of the booster sessions. An increase in positive beliefs about marijuana was also observed at follow-up. However, effects on behavioral outcomes remained limited across time points.

Conclusions

Several cognitive and social determinants of substance use remained improved at follow-up despite the absence of significant behavioural effects. Longer-term studies are needed to determine whether these changes translate into reductions in substance use.

Trial registration

ClinicalTrials.gov, number NCT04236947; registration date: 17/01/2020. Available at: https://clinicaltrials.gov/ct2/show/NCT04236947

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12889-026-28548-x.

Introduction

Substance use during early adolescence remains a major public health concern due to its short- and long-term consequences for neurodevelopment, mental health, and educational and social trajectories. Early initiation of alcohol, tobacco, and cannabis use has been associated with an increased risk of depression, anxiety, cognitive impairment, psychotic symptoms, suicidality, and the later development of substance use disorders [1–3]. In Latin America, and particularly in Chile, substance use prevalence among school-aged adolescents remains high compared to other regions, underscoring the need for effective, scalable, and culturally appropriate prevention strategies [4, 5].

Schools constitute a key setting for universal substance use prevention during early adolescence, given their broad population reach and their central role in social, emotional, and cognitive development. Over the past decades, numerous school-based prevention programs have been developed and evaluated internationally, aiming to delay initiation and reduce escalation of substance use. Systematic reviews and meta-analyses consistently show that universal school-based programs produce small to moderate preventive effects, particularly for alcohol use, with more modest effects for tobacco and cannabis [6, 7]. Importantly, programs grounded in cognitive-behavioral principles, social competence development, and skills training—such as self-regulation, problem-solving, and resistance to peer pressure—appear to be especially effective in early adolescence [6, 8].

Despite this evidence, several limitations persist in the prevention literature. First, average effects tend to attenuate over time, and many programs fail to demonstrate sustained impacts without continued reinforcement [7, 9]. Second, relatively few studies include follow-up periods beyond 12–24 months, and robust evidence extending into later adolescence or early adulthood is largely restricted to a small number of trials conducted in high-income countries, such as Climate, Preventure, and related combined interventions [10–12]. Third, there is a notable scarcity of rigorous longitudinal evaluations in Latin America, where most available studies remain pilot trials with short-term follow-up [13, 14]. Nevertheless, important contributions have emerged from Brazil through evaluations of the #Tamojunto program, a culturally adapted version of Unplugged [15, 16]. Randomized studies have reported follow-up assessments extending to 9 and 21 months, providing valuable evidence regarding the implementation and effectiveness of school-based prevention programs in the region. However, findings have been mixed, highlighting the challenges of translating evidence-based prevention programs across cultural and educational contexts and underscoring the need for additional longitudinal evaluations in Latin America.

Programs focused on social competence and life skills represent a particularly relevant approach for universal prevention in early adolescence. Interventions such as Life Skills Training, Unplugged, and the Social Competence Promotion Program for Young Adolescents (SCPP-YA) aim to reduce substance use indirectly by modifying key mediators, including beliefs about substances, perceived norms, refusal skills, emotional regulation, and problem-solving abilities [17–19]. Evidence suggests that changes in these mediators often precede observable behavioral effects and may require time and repeated practice to translate into reductions in substance use [20, 21].

In Chile, Mi Mejor Plan (MMP) is a culturally adapted version of the SCPP-YA designed for implementation in the school context during early adolescence. A recently published cluster randomized pilot trial evaluated the acceptability, feasibility, fidelity, and immediate post-intervention outcomes of a 16-session adapted version of MMP delivered to sixth-grade students (approximately 11–12 years old) [14]. The study demonstrated high levels of implementation quality and student acceptability, as well as significant improvements in specific cognitive and social determinants, particularly increased negative beliefs about tobacco and alcohol use. However, no significant effects were observed on substance use prevalence immediately after the intervention, a finding consistent with international evidence indicating that behavioral effects may emerge later and depend on sustained exposure and reinforcement [6, 9].

Recent prevention research highlights the potential importance of booster sessions in maintaining and strengthening intervention effects over time. Booster sessions provide opportunities to rehearse previously learned skills, reinforce cognitive and behavioral strategies, and adapt prevention messages to adolescents’ evolving developmental and social contexts [22]. Studies with longer follow-up periods suggest that interventions incorporating reinforcement components are more likely to produce durable effects, particularly when they continue to target theoretically grounded mediators of substance use risk [10, 12]. Nonetheless, empirical evidence examining the added value of booster sessions within universal, skills-based programs—especially in low- and middle-income countries—remains limited.

This article extends the previously published MMP pilot trial by examining substance use outcomes and theoretically relevant mediating factors following completion of the full intervention, including the original 16-session adapted curriculum and three booster sessions delivered during the subsequent academic year. By incorporating a longer follow-up period, we evaluated whether the improvements in cognitive and social determinants observed immediately after the core intervention were maintained over time and whether additional changes emerged after the booster sessions. Although our design does not allow the specific contribution of the booster component to be isolated, assessing outcomes after its implementation provides valuable information on the durability of intervention effects and may inform future research on reinforcement strategies in school-based prevention programs. In doing so, this study addresses an important gap in the Latin American prevention literature and contributes to a better understanding of the longer-term effects of culturally adapted, school-based substance use prevention programs.

Objectives

General objective

To evaluate the effects of the Mi Mejor Plan (MMP) school-based prevention program on substance use outcomes and theoretically relevant mediating factors among early adolescents, following the implementation of the adapted 16-session curriculum and three booster sessions delivered during the subsequent academic year.

Specific objectives

1. To examine changes in the prevalence and frequency of alcohol, tobacco, and cannabis use from baseline to follow-up after the completion of the full MMP intervention, including booster sessions.

2. To assess whether previously observed improvements in key cognitive and social determinants—such as negative beliefs about substance use, risk perception, refusal skills, problem-solving abilities, emotion regulation, and sense of school membership—are sustained or enhanced at follow-up.

3. To evaluate changes over time in cognitive and social determinants associated with substance use, including positive beliefs about substances, normative beliefs, and intentions for future use.

Methods

Study design

This study is a longitudinal follow-up of a cluster randomized controlled pilot trial evaluating the culturally adapted version of the Social Competence Promotion Program for Young Adolescents (Mi Mejor Plan, MMP). The original trial assessed acceptability, feasibility, fidelity, and immediate post-intervention outcomes following the implementation of a 16-session school-based prevention program [14].

We extended the original evaluation by examining substance use outcomes and theoretically relevant mediating factors after the full implementation of the intervention, including three booster sessions delivered during the academic year following the initial intervention. Schools remained assigned to their original allocation (intervention or control), and outcomes were assessed using a repeated-measures design.

Participants and setting

The inclusion and exclusion criteria for schools are the following.

Inclusion criteria of schools:

1. Schools located in Region Metropolitana (RM), Chile.

2. Schools of Primary Education (1st grade to 8th grade). The MMP curriculum was designed to be implemented in 6th grade.

3. Mixed-sex schools.

4. Schools with at least two classes in the 6th grade.

5. Schools with at least 30 students per class

6. Schools with vulnerability (≥ 50%). We used the School Vulnerability Index – National System of Equality Allocation (IVE-SINAE) to state the degree of vulnerability. This index is the proportion of students with high vulnerability in a given school. This index considers the following socioeconomic variables to group the schools: mother’s educational level, father’s educational level, and total monthly household income, among others [23].

Exclusion criteria of schools:

1. Four or more classes in 6th grade. This criterion was considered for economic and practical reasons because the larger the school, the more expensive the implementation and supervision.

2. Implementing other manualized substance use prevention programs. However, schools may implement preventive lessons in the Orientation curriculum.

Participants were students enrolled in sixth grade (typically 11–12 years old) at the time of the original baseline assessment, attending primary schools with a high level of socioeconomic vulnerability, as defined by the School Vulnerability Index (IVE-SINAE). The IVE-SINAE is a national indicator developed by the National Board of School Aid and Scholarships (JUNAEB) that estimates the proportion of students experiencing socioeconomic disadvantage within a school. The index incorporates information on household socioeconomic conditions, educational vulnerability, and social risk indicators. Higher scores indicate greater levels of student vulnerability. Schools with elevated IVE-SINAE scores were prioritised because adolescents exposed to greater socioeconomic disadvantage are at increased risk for substance use and related adverse outcomes [24, 25], and because these settings are often underserved by evidence-based prevention initiatives [26, 27]. Consistent with the original trial protocol, only schools with an IVE-SINAE score of 50% or higher were eligible for participation.

Schools were identified from the official registry of primary schools in the Metropolitan Region of Chile maintained by the Ministry of Education. Eligible schools were invited sequentially until the target number of participating schools was achieved.

All students who participated in the baseline and post-intervention assessments of the original trial were eligible for inclusion in the follow-up assessment. Written informed consent was obtained from parents or legal guardians, and written assent was obtained from students at each wave of data collection.

Further information about recruitment can be found in the previous report of this trial [14].

Sample size

This follow-up study used the same sample derived from the pilot cluster randomized controlled trial previously reported. As described in the main trial publication [14], the sample size was determined based on feasibility considerations rather than statistical power for effectiveness testing. No additional sample size calculations were performed for the 6-month follow-up. No interim analyses or stopping guidelines were prespecified.

Randomisation

Details of the randomization procedure have been reported previously. In brief, schools were randomized at the cluster level (1:1 allocation to intervention or control) using a computer-generated sequence by an independent statistician. No additional randomization procedures were conducted for the follow-up phase.

Blinding

As described in the primary trial report, blinding of participants, facilitators, and school personnel was not feasible due to the nature of the intervention. No additional blinding procedures were introduced for the follow-up assessments.

Intervention

Mi Mejor Plan (MMP)

The intervention consisted of a culturally adapted version of the Social Competence Promotion Program for Young Adolescents (SCPP-YA), implemented within regular school hours. The cultural adaptation followed a structured adaptation process designed to preserve the core theoretical principles and active components of the original SCPP-YA while improving its acceptability and relevance for Chilean schools. The adaptation included revisions to language, culturally appropriate examples and role-play scenarios, modifications to classroom activities, visual materials, and examples reflecting the Chilean educational system and adolescents' everyday experiences. No changes were made to the program's theoretical framework or core problem-solving model. A detailed description of the adaptation process has been published previously [14]. As described in the original study, the adapted curriculum included 16 classroom-based sessions delivered over one academic year: 11 sessions focused on social problem-solving skills and 5 sessions focused on substance use prevention.

The core component of the program was a structured six-step social problem-solving model designed to enhance self-regulation, goal setting, cognitive flexibility, and decision-making. The substance use prevention module emphasized the application of these skills to high-risk situations, including peer pressure and normative influences related to alcohol, tobacco, and cannabis use. For more details, see the previous publication [14].

The theoretical framework underlying Mi Mejor Plan combines the social problem-solving model originally developed within the Social Competence Promotion Program for Young Adolescents (SCPP-YA) [28, 29] with social-cognitive theories of adolescent substance use prevention [30, 31]. The intervention assumes that substance use behaviours are influenced by cognitive expectations regarding substance use, perceived social norms, behavioural intentions, and adolescents’ ability to generate adaptive responses to social and emotional challenges. This approach is also consistent with evidence indicating that social competence, decision-making skills, and resistance to peer influence are important theoretically relevant mediators against adolescent substance use [32, 33]. Through the six-step problem-solving model, students learn to identify problems, evaluate alternative solutions, anticipate consequences, and make goal-directed decisions. Substance use prevention sessions explicitly apply these skills to situations involving peer pressure, social influence, and substance-related decision-making. Accordingly, the intervention is expected to reduce favourable beliefs about substance use, strengthen perceptions of risk, modify normative beliefs, reduce intentions for future use, and increase refusal skills, which are considered proximal mediators of behavioural outcomes and established predictors of substance use behaviour in adolescence [34]. The selection of mediators assessed here was directly informed by this theoretical framework, as these constructs represent the primary cognitive and social mechanisms through which MMP is expected to influence subsequent substance use behaviours.

Booster sessions

In addition to the original 16 sessions, three booster sessions were delivered during the subsequent academic year. The booster sessions were designed to reinforce previously learned skills, provide additional opportunities for practice, and adapt prevention messages to students’ developmental progression and emerging social challenges. Content focused on revisiting the problem-solving steps, strengthening refusal skills, and reinforcing negative beliefs and risk perception related to substance use. See Suppl. 1 for content details of booster sessions.

Booster sessions were delivered by trained facilitators, following a standardized manual developed by the research team, and were implemented during regular school hours.

The booster sessions were designed to reinforce these proximal cognitive and behavioural mechanisms by revisiting core problem-solving concepts, practising refusal skills, and reviewing substance-related beliefs and consequences. Consistent with prevention research, the rationale for booster sessions is that repeated exposure, rehearsal, and reinforcement may help consolidate previously acquired skills, strengthen retention of intervention content, and sustain changes in substance use–related cognitions over time [32, 35]. Previous studies have suggested that booster sessions can enhance the durability of prevention effects by providing opportunities to revisit key concepts, reinforce behavioural competencies, and address developmental challenges that emerge as adolescents grow older [36, 37]. Accordingly, the booster component of Mi Mejor Plan was intended to support the maintenance of gains achieved during the core intervention and to promote longer-term reductions in substance use risk factors.

Control condition

Schools assigned to the control group continued to receive treatment-as-usual (TAU), consisting of the standard non-manualized prevention content delivered during the “Orientation” class, as defined by national educational guidelines. The research team did not interfere with or standardize the content delivered in control schools.

Measures

Substance use outcomes

Substance use outcomes were assessed using the European Drug Addiction Prevention Trial Questionnaire (EU-Dap), a self-report instrument validated for use in Chilean adolescents [38]. Outcomes included 30-day prevalence of tobacco, alcohol, and cannabis use, and alcohol intoxication. Binary indicators were created for each substance based on standard EU-Dap scoring procedures.

Substance use–related cognitive and psychosocial factors

The following theoretically relevant cognitive and psychosocial factors associated with adolescent substance use were assessed, consistent with the theory of change of the MMP program:

1. Beliefs about substance use. Beliefs about substance use were assessed using subscales from the European Drug Addiction Prevention Trial Questionnaire (EU-Dap) [38]. These scales assess adolescents’ cognitive expectations regarding the potential benefits and harms associated with the use of specific substances, including tobacco, alcohol, and cannabis. For each substance, the instrument distinguishes between positive beliefs, reflecting perceived benefits or desirable effects of use (e.g., social acceptance or emotional relief), and negative beliefs, reflecting perceived risks or adverse consequences associated with consumption. Items are rated on Likert-type scales, with higher scores indicating stronger endorsement of the respective belief dimension. These belief constructs are theoretically grounded in social-cognitive models of substance use [20], where outcome expectations play a central role in shaping adolescents’ decisions to initiate or avoid substance use. The EU-Dap questionnaire has been culturally adapted and validated for use among Chilean early adolescents, demonstrating adequate factorial validity and internal consistency for the belief subscales evaluating tobacco, alcohol, and marijuana use [38].

2. Normative beliefs and future substance use intentions. Perceived social norms regarding substance use and intentions for future use were also assessed using EU-Dap items. Normative beliefs capture adolescents’ perceptions of how common substance use is among their peers, including both close friends and adolescents of similar age. These perceptions reflect the degree to which substance use is viewed as socially normative within the adolescent’s peer context. The future substance use intentions scale assesses the likelihood that adolescents believe they may use substances in the future. This construct captures behavioural expectations that may precede the initiation of substance use. Both constructs are widely recognised predictors of substance use initiation during adolescence. Research has shown that misperceptions of peer norms are strongly associated with increased substance use, as adolescents often overestimate the prevalence of use among peers [39]. In the Chilean validation study of the EU-Dap questionnaire, the normative beliefs and future substance use subscales demonstrated acceptable psychometric properties and meaningful associations with recent substance use behaviours, supporting their relevance as risk indicators and potential mediators of preventive interventions [38].

3. Risk perception and refusal skills. Risk perception and refusal skills related to substance use were measured using EU-Dap subscales [38]. The risk perception scale assesses adolescents’ perceptions of the potential negative consequences of substance use, including health, psychological, and social harms. Higher scores indicate greater perceived risk associated with engaging in behaviours such as smoking, drinking alcohol, or using marijuana. The refusal skills scale assesses adolescents’ perceived ability to resist peer pressure and decline offers to use substances. These items reflect behavioural competence and assertiveness in situations where substances may be offered by peers. Both constructs represent key cognitive and social determinants factors targeted by many school-based prevention programs. Evidence from the Chilean validation of the EU-Dap questionnaire indicates that higher perceived risk and stronger refusal skills have been linked to lower odds of recent substance use, supporting their role as important intermediate outcomes in prevention research [38].

4. Social problem-solving skills. Adolescents’ problem-solving abilities were measured using the Social Problem-Solving Inventory – Revised Short Form (SPSI-R:S), a widely used instrument designed to assess cognitive-behavioural processes involved in identifying, evaluating, and responding to everyday problems [40]. The SPSI-R:S has 25 items and conceptualizes social problem solving as a multidimensional construct that includes both individuals’ orientation toward problems and the strategies they use to address them. The instrument captures key aspects of problem-solving competence, including constructive approaches to problem resolution as well as more maladaptive or impulsive response patterns. Participants respond to items using Likert-type scales reflecting how typical each statement is of their usual behaviour. Higher scores indicate more adaptive problem-solving abilities. Social problem-solving skills are considered an important psychosocial competence during adolescence and have been associated with better coping, lower emotional distress, and reduced engagement in health-risk behaviours [40].

5. Emotion regulation. Emotion regulation strategies were assessed using the Emotion Regulation Questionnaire for Children and Adolescents (ERQ-CA), a developmentally adapted version of the original Emotion Regulation Questionnaire designed to assess how young people manage their emotional responses [41, 42]. The instrument has 10 items and assesses two widely studied regulatory strategies: cognitive reappraisal (6 items) and expressive suppression (4 items). Cognitive reappraisal refers to the tendency to reinterpret emotionally evocative situations in order to modify their emotional impact, whereas expressive suppression reflects the inhibition of emotional expression after an emotional response has been generated. Participants rate items on a Likert-type scale indicating how frequently they use each strategy in everyday situations. Higher scores in the cognitive reappraisal subscale indicate greater emotion regulation. On the contrary, a higher score in expressive suppression indicated lower emotion regulation. This questionnaire has been validated among adolescents in Chile, showing good psychometric properties [43].

6. Psychological sense of school membership. Students’ sense of belonging within the school environment was assessed using the Psychological Sense of School Membership (PSSM) scale, a 18-item questionnaire developed to measure adolescents’ perceptions of acceptance, inclusion, and support within their school community [44]. The PSSM assesses the extent to which students feel respected and valued by teachers and peers, as well as their overall sense of belonging and emotional connection to the school. Participants indicate their level of agreement with statements reflecting feelings of inclusion, support, and recognition within the school context. Higher scores indicate a stronger sense of school membership. A robust body of research has shown that school belonging is a key protective factor during adolescence and is associated with improved academic engagement, psychological wellbeing, and lower levels of risk behaviours, including substance use [44]. The validation in Chile includes 13 of the 18 original items, demonstrating unidimensionality and good psychometric properties [45].

A summary of the number of items, response formats, score ranges, and interpretation of all outcome and mediator measures is provided in Supplementary Table S4.

Data collection procedures

Assessments were conducted at three time points: 1) baseline (prior to intervention), 2) immediate post-intervention (after the 16-session curriculum), and 3) follow-up assessment after completion of the three booster sessions. Questionnaires were administered in classroom settings by trained research assistants using standardized procedures. Confidentiality was ensured by using anonymized participant codes.

The baseline assessment was conducted at the beginning of sixth grade, prior to delivery of the 16-session intervention. Post-intervention assessments were conducted immediately after completion of the core curriculum at the end of the academic year. The three booster sessions were delivered during the subsequent academic year, when students were enrolled in seventh grade. The follow-up assessment was conducted approximately six months after the post-intervention assessment and after completion of all booster sessions. Thus, the follow-up evaluation reflects outcomes observed after exposure to both the core intervention and the booster component. See Supplementary Material 3.

Statistical analysis

The analytical approach for the follow-up outcomes was consistent with that used in the primary analysis. Analyses followed a modified intention-to-treat approach. All participants who completed the baseline assessment were included in the analytical sample and were analysed according to the group to which their school was originally randomised, regardless of the number of intervention sessions attended. Longitudinal mixed-effects models incorporated all available repeated measurements from these participants. Participants who did not complete the 6-month assessment were considered lost to follow-up for the purposes of reporting retention rates and the attrition analysis but contributed all available data collected before dropout to the mixed-effects models under the missing-at-random assumption.

Mixed-effects models were used to examine changes over time and between-group differences at follow-up, accounting for the clustered structure of the data (students nested within schools) and repeated measures within participants. Models included fixed effects for group, time, and baseline values of the outcome, with additional adjustment for age and sex in fully adjusted models. Random effects for school and participant were included. The reported intervention effects represent adjusted between-group differences at each follow-up assessment while controlling for baseline values of the corresponding outcome.

Binary outcomes were analyzed using logistic mixed-effects models and are reported as odds ratios with 95% confidence intervals. Continuous outcomes were analyzed using linear mixed-effects models and are reported as adjusted mean differences with 95% confidence intervals and standardized effect sizes.

Analyses were conducted using all available data within the mixed-effects framework. No additional procedures for handling missing data (e.g., multiple imputation) were implemented. Attrition and response rates at follow-up are reported to characterize data completeness. To evaluate the potential impact of attrition, baseline characteristics were compared between participants retained at follow-up and those lost to follow-up. Results of these analyses are presented in Supplementary Table S2.

No subgroup or sensitivity analyses were conducted. In addition, given the pilot nature of the study and the relatively small number of clusters, no small-sample corrections (e.g., Kenward–Roger or Satterthwaite adjustments) were applied. Model estimates were based on standard maximum likelihood procedures implemented in the mixed-effects framework.

The trial protocol is available at ClinicalTrials.gov (NCT04236947). Additional methodological details are reported in the original trial publications [14, 46].

Ethical considerations

The study was conducted in accordance with the Declaration of Helsinki and received ethical approval from the Ethics Committee of Universidad de los Andes (CEC201983). Participation was voluntary, and students and caregivers were informed of their right to withdraw at any time without consequences.

Harms

As in the primary trial report, harms were not specified as a formal outcome, and no systematic adverse event monitoring protocol was implemented. During the follow-up period, no intervention-related harms were reported.

Results

Participant flow and follow-up rates

A total of 575 schools were assessed for eligibility; 95 were excluded and 315 were not invited after reaching the required sample size. Of the 165 schools invited, 12 were randomized (6 intervention, 6 control). After randomization, but before the baseline assessment, one school in the intervention group withdrew from the study, saying they no longer had the time to implement the sessions. Consequently, 11 schools (765 students) were included in the study: 5 schools (n = 344 students) in the intervention arm and 6 schools (n = 421 students) in the control arm. At baseline, 608 students were assessed (intervention: n = 263; control: n = 345). At post-intervention, 538 students were assessed (intervention: n = 236; control: n = 302). At follow-up, 426 students were assessed (intervention: n = 198; control: n = 228), corresponding to an overall retention rate of 70.1% from baseline. Attrition across assessment points was primarily due to student absence, lack of consent or assent, and non-response. See Fig. 1.

Fig. 1.

CONSORT 2010 flow diagram

Among the 608 participants assessed at baseline, 601 had complete baseline information for the variables included in the attrition analysis. Of these, 419 were retained at follow-up and 182 were lost to follow-up. Attrition analyses are presented in Supplementary Table S2. Retention was higher in the intervention group than in the control group (47.3% vs. 35.7%; p = 0.011). Apart from this difference, participants retained and lost to follow-up were broadly comparable across baseline characteristics. Students lost to follow-up reported slightly higher normative beliefs regarding friends’ substance use at baseline, whereas no statistically significant differences were observed for age, sex, baseline alcohol, tobacco or marijuana use, future substance use intentions, normative beliefs regarding other adolescents, negative beliefs about tobacco, or positive beliefs about marijuana.

Baseline characteristics

Baseline characteristics of participants are presented in Table 1. A total of 608 students were included at baseline (intervention: n = 263; control: n = 345). The mean age was 11.3 years in both groups (SD = 0.6 and 0.5, respectively), and the proportion of females was identical across groups (47.5%).

Table 1.

Baseline characteristics

Variables	MMP Intervention			Control
	n	% or mean	SD	n	% or mean	SD
Sex
Female	125	47.5		164	47.5
Male	138	52.5		181	52.5
Age	263	11.3	(0.6)	345	11.3	(0.5)
Type of School dependency
Subsidized	181	68.8		345	100	-
Public	82	31.2		-	-	-
IVE-SINAE	263	77.8	(12.4)	345	78.4	(7.5)
EU-Dap Substance use
Tobacco in the last month	5	1.9		4	1.2
Alcohol in the last month	18	6.8		18	5.2
Alcohol intoxication in the last month	3	1.1		2	0.6
Marijuana in the last month	3	1.1		-	-	-
EU-Dap risk and protective factors
Positive beliefs about tobacco	242	7.6	(3.2)	313	7.3	(3.2)
Negative beliefs about tobacco	249	9.8	(3.4)	326	10.1	(3.3)
Positive beliefs about alcohol	235	9.7	(4.3)	308	9.3	(4.3)
Negative beliefs about alcohol	247	13.2	(4.6)	323	13.4	(4.7)
Positive beliefs about marijuana	242	7.5	(3.7)	324	7.3	(3.6)
Negative beliefs about marijuana	242	17.0	(6.1)	322	17.4	(5.7)
Future substance use	251	7.9	(3.9)	330	7.1	(3.5)
Normative beliefs friends	253	4.6	(1.4)	336	4.2	(0.9)
Normative beliefs other adolescents	254	6.7	(3.4)	323	6.5	(3.3)
Risk perception	238	19.6	(4.3)	319	19.9	(4.1)
Refusal skills	253	10.7	(2.2)	330	10.9	(1.7)
ERQ-CA
Cognitive Reappraisal	234	19.3	(6.3)	306	19.8	(6.4)
Expressive Suppression	238	13.0	(4.4)	309	12.7	(4.3)
SPSI-R:S	206	49.4	(11.4)	262	51.9	(11.9)
PSSM	221	44.1	(12.7)	283	45.6	(12.0)

MMP “Mi Mejor Plan”, SD Standard Deviation, IVE Indíce de Vulnerabiliad Escolar, SINAE del Sistema Nacional de Asignación con Equidad (in english “School Vulnerability Index (IVE) of the National System of Equitable Allocation (SINAE)”), EU-Dap European Drug Addiction Prevention Trial Questionnaire, ERQ-CA Emotion Regulation Questionnaire for Children and Adolescents, SPSI-R:S Revised Social Problem-Solving Inventory-Revised Scale, PSSM Psychological Sense of School Membership questionnaire

In the intervention group, 68.8% of students attended subsidized schools and 31.2% attended public schools, whereas all students in the control group were from subsidized schools. The mean School Vulnerability Index (IVE-SINAE) was similar between groups (intervention: 77.8 [SD = 12.4]; control: 78.4 [SD = 7.5]).

The prevalence of substance use in the past 30 days was low in both groups, including tobacco (1.9% vs. 1.2%), alcohol (6.8% vs. 5.2%), alcohol intoxication (1.1% vs. 0.6%), and marijuana (1.1% vs. 0.0%) in the intervention and control groups, respectively.

Mean scores for EU-Dap cognitive and social determinants, emotion regulation (ERQ-CA), social problem-solving (SPSI-RS), and school belonging (PSSM) were comparable between groups at baseline.

Substance use outcomes at follow-up

Thirty-day prevalence of substance use and intervention effects at post-intervention and 6-month follow-up are presented in Table 2.

Table 2.

Thirty-day prevalence of tobacco, alcohol, drunkenness, and marijuana, and the impact of the intervention post-intervention and at 6-month the follow-up

	Post intervention			6-month follow-up
	Control	Intervention		Control	Intervention
30-day prevalence	n (%)	n (%)	OR (95% CI)	n (%)	n (%)	OR (95% CI)
Tobacco	4 (1.3)	5 (2.2)	1.04 (0.14 to 7.44)	7 (3.2)	3(1.5)	0.26 (0.03 to 2.02)
Alcohol	18 (6.0)	23 (10.1)	1.51 (0.50 to 4.58)	13 (6.0)	10(5.2)	0.66 (0.18 to 2.41)
Alcohol intoxication	3 (1.0)	4 (1.7)	1.01(0.8 to 12.59)	2 (0.9)	2(1.0)	0.59 (0.03 to 10.27)
Marijuana	2 (0.6)	2 (0.8)	2.38 (0.17 to 32.51)	4 (1.8)	2(1.0)	-a

CI Confidence Interval, OR Odds Ratio

^aThere was not a sufficient number of students who used marijuana to compute the mixed regression models

At post-intervention and 6-month follow-up, the prevalence of tobacco, alcohol, alcohol intoxication, and marijuana use remained low in both groups. No statistically significant intervention effects were observed for any substance use outcome at either post-intervention or follow-up (Table 2).

For marijuana use at follow-up, the mixed-effects logistic regression model could not be estimated due to the insufficient number of events. This likely reflects the very low prevalence of marijuana use in the sample, which can lead to sparse data and quasi-complete separation, limiting the stability and convergence of logistic mixed-effects models.

Changes in cognitive and psychosocial factors

Changes in cognitive and social determinants at post-intervention and 6-month follow-up are presented in Table 3.

Table 3.

Program impact on cognitive and social determinants

	Post-intervention			6-month follow-up
	Control	Intervention	Adjusted Modela	Control	Intervention	Adjusted Modela
Variables	Mean (SD)	Mean (SD)	aMD (95% CI)	Mean (SD)	Mean (SD)	aMD (95% CI)
EU-Dap
Cognitive and Social determinants
Positive beliefs tobacco	7.41 (3.14)	8.22(3.23)	0.58 (−0.04–1.22)	7.36(3.20)	8.08(3.26)	0.36 (−0.32–1.05)
Positive beliefs alcohol	9.12 (4.05)	10.08(4.39)	0.67 (−0.20–1.55)	8.83(4.2)	10.04(4.63)	0.84 (−0.08–1.78)
Positive beliefs marijuana	7.17 (3.58)	8.11(3.71)	0.66 (−0.05–1.38)	7.26 (3.62)	8.36 (3.98)	0.86 (0.09–1.64)
Normative beliefs friends	4.60(1.73)	4.67(1.75)	−0.23 (−0.50 −0.42)	4.65(1.55)	4.46(1.06)	−0.50 (−0.79—−0.20)
Normative beliefs other adolescents	7.21(3.62)	7.35(3.73)	−0.19 (−0.85—0.45)	8.24 (3.96)	7.70(3.81)	−0.87 (−1.58—−0.16)
Future substance use	7.26 (2.97)	7.82 (3.60)	−0.23 (−0.84–0.36)	7.54 (3.19)	7.22 (2.67)	−0.92 (−1.57—−0.27)
Negative beliefs tobacco	10.61(3.12)	11.43(3.39)	1.11 (0.47–1.75)	10.88(3.11)	11.30(3.57)	0.75 (0.01–1.39)
Negative beliefs alcohol	13.54(4.25)	14.79(4.53)	1.33 (0.46–2.21)	13.70(4.12)	14.42(4.95)	0.84 (−0.09–1.78)
Negative beliefs marijuana	17.52(5.36)	18.67(5.55)	1.36 (0.32–2.39)	18.04(5.07)	18.55(5.58)	0.85(−0.26–1.96)
Risk perception	19.73(3.92)	20.05(3.86)	0.61 (−0.12 −1.34)	20.08(3.43)	20.19(3.71)	0.22 (−0.57- 1.0)
Refusal skills	10.80(1.91)	10.44(2.21)	−0.20 (−0.57 −0.16)	10.50(2.21)	10.64(1.88)	0.29(−0.10–0.69)
ERQ-CA
Cognitive Reappraisal	19.38(5.92)	19.35(6.74)	0.44 (−0.83 −1.72)	18.65(5.98)	18.63(6.41)	0.34 (−1.04—1.72)
Expressive Suppression	12.77(4.33)	13.59(4.44)	0.47 (−0.35—1.30)	12.67(3.99)	12.23(4.29)	−0.76 (−1.66 −0.13)
SPSI-RS	52.07(13.05)	50.29(11.87)	0.24 (−2.11 −2.60)	53.00(13.88)	51.72(12.62)	0.96 (−1.52—3.45)
PSSM	44.97(11.36)	44.77(12.17)	0.49 (−1.68 −2.67)	43.15(12.53)	43.30(11.77)	1.97 (−0.36 −4.31)

aMD adjusted mean difference, CI Confidence Interval, EU-Dap European Drug Addiction Prevention, ERQ-CA Trial Questionnaire; Emotion Regulation Questionnaire for Children and Adolescents, SPSI-RS Social Problem-Solving Inventory-Revised Scale, PSSM Psychological Sense of School Membership Scale

^aMixed regression model adjusted for sex, age, and baseline outcome measures

Bold values indicate statistical significance at p < .05

At post-intervention, statistically significant differences were observed in several cognitive and social determinants. The intervention group showed higher scores in negative beliefs about tobacco (aMD = 1.11, 95% CI 0.47–1.75), alcohol (aMD = 1.33, 95% CI 0.46–2.21), and marijuana (aMD = 1.36, 95% CI 0.32–2.39). No significant differences were found for risk perception, refusal skills, normative beliefs, or future substance use.

At 6-month follow-up, significant differences were observed in cognitive and social determinants. The intervention group showed lower normative beliefs regarding friends (aMD = − 0.50, 95% CI − 0.79 to − 0.20) and other adolescents (aMD = − 0.87, 95% CI − 1.58 to − 0.16), as well as lower future substance use intentions (aMD = − 0.92, 95% CI − 1.57 to − 0.27). In addition, higher scores were observed in negative beliefs about tobacco (aMD = 0.75, 95% CI 0.01–1.39).

An increase in positive beliefs about marijuana was also observed at follow-up (aMD = 0.86, 95% CI 0.09–1.64). No statistically significant differences were found for emotion regulation (ERQ-CA), social problem-solving skills (SPSI-RS), or school belonging (PSSM) at either time point.

Discussion

We evaluated the 6-month follow-up effects of the Mi Mejor Plan school-based intervention on substance use and associated cognitive and social determinants among early adolescents. Overall, no statistically significant effects were observed on 30-day substance use outcomes. However, the intervention was associated with meaningful changes in several proximal cognitive and social determinants of substance use. At follow-up, students in the intervention group reported lower normative beliefs and future substance use intentions, as well as higher negative beliefs about tobacco. At the same time, an increase in positive beliefs about marijuana was observed. No effects were found for emotion regulation, social problem-solving skills, or school belonging.

The absence of significant effects on substance use outcomes is consistent with previous research on universal school-based prevention programs delivered in early adolescence, where behavioural effects are often modest or delayed [47]. Systematic reviews and meta-analyses have shown that prevention effects on actual substance use tend to emerge over longer follow-up periods, particularly as exposure and opportunity to use substances increase with age [47].

Importantly, the observed changes in normative beliefs and future substance use intentions are aligned with the theoretical mechanisms targeted by evidence-based prevention programs such as Unplugged and other EU-Dap–based interventions. Normative beliefs, in particular, have been identified as one of the most robust predictors of adolescent substance use across substances [48], and prior research has demonstrated that modifying perceived peer norms is a key pathway through which school-based interventions exert their effects [13, 49].

These findings are also broadly consistent with the post-intervention results reported in the first phase of this study, where early changes were observed in several cognitive determinants but not in behavioural outcomes [14]. The persistence—and in some cases strengthening—of effects on normative beliefs and future intentions at follow-up suggests that these cognitive shifts may be sustained over time, even in the absence of immediate behavioural changes.

More broadly, these results are in line with the prevention literature indicating that universal school-based interventions tend to have stronger effects on intermediate outcomes (e.g., beliefs, intentions, skills) than on substance use behaviour itself, particularly in younger samples with low baseline prevalence [6, 7].

Several factors may explain the pattern of results observed in this study. First, the low prevalence of substance use at baseline and follow-up likely limited the ability to detect behavioural effects. Early adolescence represents a developmental stage where substance use initiation is still relatively infrequent, and prevention efforts may primarily operate by shaping attitudes, norms, and intentions before behavioural patterns are established [8]. Second, the observed reductions in normative beliefs and future substance use intentions suggest that the intervention may be influencing key upstream mechanisms of risk. These constructs are central components of social influence models, and their modification has been linked to later reductions in substance use in longer-term follow-up studies [50]. Third, the increase in positive beliefs about marijuana represents an unexpected finding and should be interpreted cautiously. One possible explanation is the occurrence of unintended intervention effects, whereby increased discussion and salience of cannabis-related topics may have inadvertently strengthened some perceived benefits of use among participants. Although speculative, this interpretation is consistent with previous literature describing potential iatrogenic effects of prevention programs, particularly when increased awareness of substances is not accompanied by equivalent changes in attitudes toward their perceived benefits [51]. An alternative explanation is that broader sociocultural trends toward the normalization of cannabis use may have contributed to increasingly favourable perceptions among adolescents. However, because such societal influences would be expected to affect both groups, they are unlikely to fully explain the between-group difference observed. Further research is needed to better understand the mechanisms underlying this finding and its implications for prevention program design. Fourth, the absence of effects on emotion regulation, problem-solving, and school belonging may reflect the fact that these domains were not the primary targets of the intervention, or that changes in these broader psychosocial constructs require more intensive or longer-term intervention components to be detectable [52, 53].

The interpretation of these findings in relation to the booster sessions warrants caution. Because all students in the intervention group received the booster component and no comparison group without booster sessions was available, the present study cannot isolate the specific contribution of the boosters from the effects of the original intervention. Consequently, although several cognitive outcomes remained significant or appeared stronger at follow-up, these findings should not be interpreted as evidence that the booster sessions themselves caused these changes. Rather, they indicate that intervention effects were still observable at the assessment conducted after completion of the booster sessions.

Several limitations should be considered when interpreting these findings. First, this study was originally designed as a pilot cluster randomized trial and was not powered to detect small intervention effects on substance use outcomes. The combination of a limited number of clusters, low baseline prevalence of substance use, and a relatively small number of outcome events reduced statistical precision and resulted in wide confidence intervals for several behavioural outcomes. In addition, substance use prevalence remained relatively stable across follow-up assessments, particularly in the control group. This limited developmental change reduced the opportunity to observe divergence between study arms and may have further constrained the detection of preventive effects during the study period. Consequently, the absence of statistically significant effects on substance use should not be interpreted as evidence of no intervention effect, but rather as reflecting limited power to detect modest behavioural differences in this early adolescent sample. In addition, the relatively small number of clusters may have increased the sensitivity of estimates to cluster-level variability and reduced the precision of standard error estimation.

Second, retention was significantly higher in the intervention group than in the control group. Although the reasons for this difference cannot be determined, differential attrition between study arms may have introduced selection bias. Although participants retained and lost to follow-up were otherwise broadly comparable and mixed-effects models incorporated all available repeated measurements under the missing-at-random assumption, differential attrition between study arms may have introduced selection bias and should be considered when interpreting the findings. Nevertheless, retention rates observed in the present study are comparable to those reported in several longitudinal school-based prevention trials involving adolescents, including evaluations of the EU-Dap/Unplugged, Climate Schools, and Preventure programs, where participant mobility, school absenteeism, and administrative factors commonly contribute to follow-up losses over time [9, 12, 48]. Third, the study relied on self-reported measures, which may be subject to reporting bias, particularly for sensitive behaviours such as substance use. Fourth, although schools were randomized, some differences in school characteristics remained between study arms, particularly regarding school type. However, socioeconomic vulnerability, as measured by the School Vulnerability Index (IVE-SINAE), was highly comparable between groups. Given the small number of clusters, school-level covariates were not included in the primary models to avoid model overparameterization. Residual confounding at the school level nevertheless cannot be completely excluded.

Fifth, the study was conducted in a relatively small number of schools located within a single region of Chile and primarily serving socioeconomically vulnerable populations. Although this focus was consistent with the objectives of the original trial, the findings may not generalize to schools with different socioeconomic characteristics, educational contexts, or cultural environments. Replication in larger and more diverse school samples is needed to establish the broader applicability of the intervention.

Future research should examine the longer-term effects of the Mi Mejor Plan intervention, particularly as participants transition into later adolescence, when substance use becomes more prevalent. Extending follow-up periods may help determine whether early changes in cognitive and social determinants translate into meaningful reductions in substance use over time.

Additionally, intervention refinements may be warranted to more directly address substance-specific beliefs, particularly regarding marijuana, given the observed increase in positive beliefs. Strengthening components targeting social norms and perceived benefits of use may enhance program effectiveness in this domain.

Further research could also explore the integration of complementary components targeting emotional regulation and coping skills, as combined prevention approaches addressing both substance use and mental health have shown promise in recent studies [9, 11]. Finally, future studies should consider larger samples or higher-risk populations to improve the ability to detect behavioural effects and to better understand for whom and under what conditions the intervention is most effective.

Authors’ contributions

JG, CRB and RA conceptualised the study and secured funding. NR and SR contributed to study design and implementation. JG and GS conducted the statistical analyses. SR and NR contributed to data collection and management. JG drafted the manuscript, and all authors critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript.

Funding

This research was funded by the National Research and Development Agency [ANID]. Unique ID: FONIS SA19I0115. The funding body had no role in the design of the study; collection, analysis, and interpretation of data; or in writing the manuscript. JG received additional support from ANID Center for the Well-Being and Development of Adolescents and Children in the Digital Age (BAND) (CIN250046) and ANID/FONDECYT REGULAR Nº 1261819.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to ethical and data protection restrictions but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

All procedures performed in this study were reviewed and approved by the Ethics Committee of Universidad de los Andes, Chile (protocol number: (CEC 201983, November 5 th, 2019)). The study was conducted in accordance with the Declaration of Helsinki and local regulations. Written informed consent was obtained from parents or legal guardians, and assent was obtained from all participating students prior to enrolment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.