The World Health Organization’s Early Adolescent Skills for Emotions: A Systematic Review on Current Evidence regarding Effectiveness and Implementation

Abstract

Introduction

Adolescents in many low- and middle-income countries are frequently exposed to adversities such as armed conflicts and economic hardship. These stressors occur during a critical developmental period marked by profound physical, social, and psychological changes, increasing vulnerability to mental disorders. To address this mental health-care need, the World Health Organization (WHO) and the United Nations Children's Fund (UNICEF) developed Early Adolescent Skills for Emotions (EASE), a 10-session group intervention for adolescents and their caregivers. This systematic review summarizes current evidence regarding the effectiveness and implementation of EASE.

Methods

We searched five electronic databases up to July 26, 2025, for trials evaluating the effects of EASE on adolescents’ mental distress and caregivers’ parenting behaviors. Random-effects multilevel meta-analyses were conducted using standardized mean differences (SMDs) at post-intervention and 3-month follow-up, complemented by a qualitative synthesis of implementation outcomes. Preregistration ID: CRD42023488490.

Results

No evidence was found for superiority of EASE over a 1-session psychoeducation control on overall mental distress or caregiver outcomes. However, for the primary outcome, internalizing symptoms, there was strong evidence of a very small to small favorable effect at post-intervention, SMD = −0.17, 95% CI: [–0.31, −0.04], which was maintained at follow-up. Implementation studies indicated that EASE is acceptable, feasible, and delivered with good fidelity; yet low caregiver engagement and contextual challenges hindered implementation.

Conclusion

Observed effects were smaller than those reported for a comparable WHO intervention for adults, highlighting substantial scope for optimization. Additional high-quality trials are warranted to improve the effectiveness of EASE and maximize its potential public health impact.

Plain Language Summary

Adolescents in many low- and middle-income countries often face poverty, conflict, and other hardships that increase their risk of mental health problems. To address these challenges, the World Health Organization and the United Nations Children's Fund developed Early Adolescent Skills for Emotions (EASE), a group program with 10 sessions for adolescents and their caregivers. EASE teaches coping and emotional skills to adolescents and parenting skills to caregivers. This review examined all available studies that tested how well EASE works and how it has been implemented. Researchers searched five scientific databases for trials published up to July 2025. They compared the effects of EASE with those of a 1-session educational program. Overall, EASE did not show clear advantages over the shorter program in reducing general emotional distress or improving caregiver behaviors. However, for internalizing symptoms – such as anxiety and depression – EASE showed a small but consistent improvement, which lasted for at least 3 months after the program ended. Studies on implementation found that EASE was generally well accepted and feasible, and facilitators were able to deliver it as intended. Nonetheless, low caregiver participation and contextual challenges – such as limited resources and social barriers – made implementation difficult in some settings. In summary, EASE offers modest benefits for adolescent mental health and is practical to deliver, but its effects are smaller than similar programs of the World Health Organization for adults. More high-quality studies are needed to strengthen its effectiveness and expand its impact.

Introduction

Major crises such as wars, pandemics, and economic crises have resulted in increased stress for many people. There is robust evidence that especially children and adolescents are negatively affected by those crises [1]. In this age group, societal crises fall into a life phase of substantial physical, social, and psychological transformations [2], putting adolescents at risk for the onset of mental disorders [3]. Adolescents in low- and middle-income countries (LMICs) are at particular risk for negative changes in mental health as they are often faced with additional negative socioeconomic and environmental circumstances [4].

Although most adolescents live and need care in LMICs, there is only little evidence on prevalence rates of mental disorders in those countries [5] and even less evidence on effective interventions to reduce mental distress or promote mental health [6, 7]. Systematic reviews [8–10] on psychosocial interventions yielded less favorable effects for children and adolescents compared to adult populations, with a substantial number of studies also pointing to between-outcome differences and worsening of symptoms [8, 11]. Even when interventions showed favorable effects, they were often small. Those findings may arise from problems with the intervention itself but could also result from under-studied challenges with implementation [12, 13]. Thus, so far there is no strong evidence in favor of any psychosocial intervention for adolescents exposed to adversity in LMICs.

To address this gap, the World Health Organization (WHO) and the United Nations Children’s Fund (UNICEF) published the psychosocial intervention Early Adolescent Skills for Emotions (EASE) [14, 15]. EASE is part of the WHO series of low-intensity scalable transdiagnostic programs [16] and was developed for 10- to 14-year-old adolescents with internalizing symptoms such as depression and anxiety in low-resource settings [17]. The program comprises 10 group sessions, with seven 90-min sessions addressing adolescents and three 120-min sessions targeting one of their caregivers. Sessions for adolescents involve psychoeducation, problem solving, stress management, behavioral activation, and relapse prevention. Caregiver sessions consist of psychoeducation, active listening, quality time, praise, self-care, and relapse prevention. Based on feasibility trials [17–19], minor modifications were made to the intervention that involved a larger number of interactive activities in the adolescent sessions, a redesign of storybooks used for adolescents, a shortening of caregiver sessions, and additional training for facilitators [19].

Although several systematic reviews have summarized evidence on psychosocial interventions for children and adolescents affected by humanitarian emergencies [9, 11, 20–23], and initial syntheses exist for WHO interventions targeting adults [24, 25], only one systematic review to date has focused specifically on the effects of EASE in adolescents [26]. That review reported evidence for small to moderate favorable effects on internalizing symptoms but no effects on other outcomes. Importantly, it did not assess caregiver outcomes or implementation aspects, which may help explain the limited effects observed. In addition, an assessment of the certainty of evidence [27] was missing, and the synthesis did not account for statistical interdependencies among outcomes within the same trial, which may have biased the meta-analytical estimates [28].

Given the high need for care, the increasing burden of stressors in LMICs, and the potential large-scale public health impact of WHO-promoted interventions, summarizing preliminary findings on the effects and implementation of EASE is essential for establishing a robust evidence base to inform scaling-up efforts. This aligns with the WHO mandate to provide evidence-based interventions [14, 16], for which high-quality systematic reviews and meta-analyses offer the strongest form of evidence base beyond randomized controlled primary studies. The present review seeks to address this gap by providing a comprehensive, up-to-date synthesis of current evidence on both the effectiveness and implementation of EASE.

Methods

This systematic review adheres to standards of the Cochrane Collaboration [27, 29] and is reported in line with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [30] (PRISMA) (see online suppl. material (SM) 1 for the PRISMA checklist; for all online suppl. material, see https://doi.org/10.1159/000549240). Differences between the prospective preregistration on December 15, 2023 (PROSPERO-ID: CRD42023488490; OSF-ID: 10.17605/OSF.IO/K965C) and the final review are presented as SM2. All materials relevant to this systematic review are available from the OSF project (osf.io/v3md5).

Search Strategy

The search strategy was developed based on a previous review on two WHO interventions for adults [24]. Five databases were searched for primary studies published between 2018 and July 26, 2025, that is, APA PsycNet, Cochrane Central Register of Controlled Trials (CENTRAL), Embase, Scopus, and Web of Science. The search strategy comprised two clusters of search terms related to (a) EASE and (b) study design, with terms within clusters being linked with OR and clusters being connected using AND (see SM3 for the full search strategy). Where applicable, we used high-sensitive search filters for randomized controlled trials (RCTs) [31]. To include the most recent findings, we searched the preprint databases medRxiv, PsyArXiv, and OSF Preprints (last update: July 28, 2025) and created a Google Scholar citation alert for “Early Adolescent Skills for Emotions” (last alert received: August 31, 2025). Moreover, to maximize sensitivity, we screened the reference lists of relevant systematic reviews and eligible primary studies and searched for studies citing the included trials and the original introductory EASE publication [14] using Web of Science.

Selection Criteria

Eligible studies were (cluster) RCTs examining the effects of EASE in stress-exposed populations of all ages. Studies were eligible if EASE was delivered as initially proposed [14] or amended with additional components but ineligible when they examined stepped-care approaches with EASE being the second step. All comparators were eligible, including waitlist, (enhanced) treatment as usual ((E)TAU), and active controls. Eligible studies assessed one of the following outcome types: distress (e.g., anxiety symptoms, attention problems, depressive symptoms, externalizing symptoms, general distress, internalizing symptoms, PTSD symptoms), positive mental health (e.g., wellbeing, quality of life, psychosocial resources), functional status, parenting behavior, adverse events, physical health, family-related outcomes, health service use, and economic outcomes. Moreover, records were eligible if they reported implementation outcomes or described barriers and facilitators of implementation for those trials.

Study Selection

Deduplication was performed in Zotero [32] as well as using Deduplicator [33] for the latest search update. After a trial run on 100 records showing perfect agreement, titles/abstracts and full texts were screened by two reviewers (L.M.S., C.F.B., A.B., J.B., M.C.) independently in Rayyan [34]. Interrater reliability was almost perfect at the title/abstract level (kappa = 0.98) and full-text level (kappa = 0.89). For the search update on July 26, 2025, we utilized Rayyan’s built-in relevance classifier [35]. Initially, 150 records were screened manually to train the artificial intelligence-based relevance scoring system. Two reviewers (C.G.S. and S.K.S.) then proceeded with manual screening, prioritizing records with the highest relevance scores according to the system. Screening was discontinued once 500 consecutive records were rated as “exclude.” The interrater agreement was almost perfect at title/abstract level (kappa = 0.91) and perfect at full-text level (kappa = 1.00). At both stages of screenings, disagreements were resolved through discussion or by consulting a senior team member (K.L., S.K.S.).

Data Extraction

We developed a customized data extraction sheet for this review based on our previous work in the field [24] and implementation domains [36] (i.e., acceptability, feasibility, fidelity, implementation costs, sustainability, appropriateness, penetration, and adoption; see SM4). Additionally, we extracted data on barriers and facilitators of implementation. All data of eligible primary studies were extracted by one reviewer and checked by a second (L.M.S., C.F.B., C.G.S.). Any disagreements were resolved through discussion or consolation with a senior team member (K.L., S.K.S.).

Quality Appraisal

Two team members (L.M.S., C.F.B.) evaluated the risk of bias from primary studies using the Cochrane risk-of-bias tool for randomized trials (RoB 2) [37], assessing bias from the randomization process, deviations from intended intervention, missing outcome data, outcome measurement, and selective reporting. For cRCTs, we additionally assessed the risk of bias from the identification/recruitment of participants. Ratings were performed at single outcome and study levels. Judgments were “low,” “high,” or expressed “some concerns.”

Publication bias was assessed using visual inspection of contour-enhanced funnel plots [38] and statistically using rank correlation tests [39] or approximated rank correlation tests for multilevel models by including sampling error as a moderator to the main analysis [40]. A prediction of effect estimates by sampling error would be indicative of a publication bias.

Data Synthesis and Analysis

Eligible studies were summarized in tabular form. A narrative synthesis based on implementation domains [36] as well as barriers and facilitators of implementation was performed. For intervention effects, pairwise meta-analyses were performed separately for primary outcomes if at least two effect estimates were available per outcome type (e.g., internalizing symptoms) and assessments were sufficiently homogeneous. In cases where data needed for calculation of effect estimates was missing or unclear, primary study authors were contacted. Whenever available, we used data from intended-to-treat analyses for our calculations.

Meta-analyses were performed in R version 4.5.0 using the packages metafor [41], clubSandwich [42], and dplyr [43] for frequentist analyses. Bayesian modeling was performed in JASP version 0.95 [44] based on the RoBMA package [45]. All frequentist analyses used random-effects models and maximum likelihood estimates with an inverse variance method. Standardized mean differences (SMDs, Hedges’ g) at post-intervention and follow-up assessments were used as effect estimates, and their 95% confidence intervals (CIs) as indicators of statistical significance. SMDs were calculated based on means and standard deviations, with positive SMDs indicating unfavorable intervention effects for distress indicators but favorable effects for wellbeing, school engagement, and parenting behavior. In cRCTs, effect estimates were corrected for clustering effects [27]. As the cRCTs did not report on corrected standard errors, we used the formula 1 + (M – 1) • ICC to estimate design effects, with M being the average cluster size and ICC the intra-cluster correlation. As data from other cRCTs was not available, we derived the estimate for the ICC from Hamdani et al. [46], who reported an ICC of 0.047 for their primary outcome, which is in line with mild conservative estimates used in previous research [47, 48].

For adolescents, we calculated separate models for the larger category of distress (including attention problems, depressive symptoms, externalizing symptoms, functional impairment, general distress, internalizing symptoms, and PTSD symptoms), caregiver-reported distress, wellbeing, and school engagement, for both post-intervention and follow-up assessments, resulting in seven models. For distress, we used multivariate multilevel models nesting effect estimates within studies and outcome types [49]. Cluster-robust estimates were used to account for non-independent effect estimates. As no information was available for between-outcome correlations within studies, symmetric covariance matrices were imputed based on a correlation of ρ = 0.60, with other correlations being used for sensitivity analyses. Specifications of multilevel models were examined by calculating profiles of log-likelihood. For caregiver-reported mental distress, wellbeing, and school engagement, we employed standard random-effects meta-analysis as no dependencies emerged among included effect estimates.

For caregivers, we calculated separate models for the larger category of parenting behavior (including the subscales of the Alabama Parenting Questionnaire [50]: parental involvement, positive parenting, poor monitoring and supervision, inconsistent discipline, and corporal punishment) and general distress. For parenting behavior, we employed a multilevel model using the same approach as for distress outcomes in adolescents. For the purpose of these models, all effect estimates were revised in a way that higher values indicated more favorable intervention effects. Caregivers’ general psychological distress was examined by means of standard random-effects meta-analysis.

Statistical heterogeneity was assessed using Cochran’s Q [51], with a significant Q indicating the presence of heterogeneity. To quantify the amount of heterogeneity in our analyses, we used the I² statistic (range: 0–100%) at single outcome level, with values of 50% and higher indicating substantial heterogeneity [27]. Due to the limited between-study heterogeneity with respect to sample characteristics (e.g., participants’ age, gender balance), no moderator analyses were performed for these variables.

Sensitivity analyses were performed for divergent between-outcome correlations (ρ = 0.40, ρ = 0.80), type of control group at follow-up assessment ((E)TAU vs. waitlist), study type (feasibility/pilot vs. fully powered effectiveness trials), and risk of bias (analyses limited to effect estimates rated at “low risk” of bias). For the analyses on the type of control group, we employed moderator analyses using meta-regression to derive subgroup estimates, with a significant Q_M statistic indicating the presence of a moderator effect [41]. Additionally, we repeated our analyses on key outcomes¹ (i.e., overall distress indicators, depressive symptoms, general distress, and internalizing symptoms) for adolescents using a Bayesian approach. Bayesian methods are particularly well-suited for contexts with a limited number of studies and small effect sizes as they allow for the incorporation of prior information to stabilize estimates under data constraints [52, 53]. This approach enabled us to gain deeper insights into the evidence for small but potentially meaningful effects of EASE [27]. For those random-effects models, we chose weakly informative priors as no strong prior knowledge was available [54, 55], reflected in the use of N(0,1) as prior for the intervention effect (µ) and a half-Cauchy prior for the heterogeneity (τ) with a location of zero and a scale of 0.5, i.e., HC(0, 0.5). SMDs at post-intervention and follow-up assessments, adjusted for clustering effects, were used as effect estimates and are reported with 95% credible intervals. These intervals represent the ranges that, given the observed data and prior assumptions, contain the true parameter value with 95% probability. In addition, we present 95% prediction intervals, which indicate the range within which a future observation is expected to fall with 95% probability [56]. For our analyses on overall distress indicators, we employed a multilevel approach nesting effect estimates within studies. We report $\widehat{R}$ values as well as effective sample sizes (ESS) for model diagnostics, with $\widehat{R}$ values below 1.01 and ESS exceeding 400, indicating a sufficient number of iterations [57]. To evaluate the strength of evidence for the effectiveness of EASE, we calculated Bayes factors (BF₁₀), where values greater than 1 indicate stronger support for the intervention relative to the control. Although BFs are continuous measures, conventional guidelines suggest that BF₁₀ values between 1 and 3 constitute anecdotal evidence, values from 3 to 10 indicate moderate evidence, and values exceeding 10 reflect strong evidence [58]. Certainty of evidence was evaluated at a single outcome level at both assessments by two independent raters (L.M.S. and C.F.B.) using the Grading of Recommendations, Assessment, Development, and Evaluations [27] (GRADE).

Results

Study Selection

Our search for primary studies yielded 6,850 eligible records, with 2,065 duplicates being removed. Of 4,785 records screened at the title/abstract level, 20 full texts were assessed, of which 6 studies (15 reports) were eligible. One study and eight additional reports [59–65] were obtained via search updates and citation alerts (see Fig. 1). No additional records were identified by our search in systematic reviews and personal communication. We contacted the corresponding author to request data from a recently published study protocol [66], but did not receive a response to our email inquiry. In total, 7 eligible studies were included, represented by 23 reports.

Fig. 1.

Flowchart according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). APA, American Psychological Association; CENTRAL, Central Register of Controlled Trials, n, number of records.

Characteristics of Included Studies

Characteristics of included primary studies [17–19, 46, 65, 67, 68] published between 2021 and 2024 are presented in Table 1, with studies comprising 1,502 adolescent participants (682 allocated to intervention groups, 820 allocated to control groups) and 1,428 caregivers (intervention groups: 654, control groups: 774). Data on both adolescents and caregivers were available for all trials. Four studies were pilot and feasibility RCTs [17–19, 65], and three studies were designed as fully powered effectiveness RCTs [46, 67, 68], with three trials employing cluster-randomized designs [18, 46, 65]. None of the studies reported on later phases of the research process (e.g., dissemination, scaling up). For five studies [19, 46, 65, 67, 68], 3-month follow-up data were available, while two trials [61, 67] also reported results of a 12-month follow-up. Two studies each were conducted in Jordan [17, 68], Lebanon [19, 67], and Pakistan [46, 65], and one study was carried out in Tanzania [18], that is, all studies were conducted in LMICs according to the World Bank classification [69].

Table 1.

Characteristics of included primary studies

Study ID	Country	Stressor(s)	Adolescent population (type, age (M±SD), %female)	Caregiver population (type, age (M±SD), %female)	Adolescents: nrandomized, nIG, nCG	Type of intervention	Providers of intervention	Type of comparator	Outcomes
					Caregivers: nrandomized, nIG, nCG
Feasibility and pilot trials
Akhtar et al. [17]	Jordan	Syrian civil war	Syrian refugees: IG: 11.52±1.35, CG: 12.00±1.33; IG: 45.5%, CG: 42.3%	Syrian refugees: IG: 36.94±5.84, CG: 37.77±8.27; IG: 93.9%, CG: 96.2%	A: 59, IG: 33, CG: 26	EASE	Lay-guided	(E)TAU	A: Attention problems, caregiver-rated distress, depressive symptoms, externalizing symptoms, functional impairment, general distress, internalizing symptoms, PTSD symptoms, school engagement, wellbeing
					C: 59, IG: 33, CG: 26				C: Corporal punishment, general distress, inconsistent discipline, parental involvement, poor monitoring, positive parenting
									Implementation: Acceptability, feasibility, barriers of implementation
Brown et al. [19, 63]	Lebanon	Humanitarian crises (mixed stressors)	Syrian refugees, adolescents residing in Lebanon: IG: 11.70±1.40, CG: 11.70±1.10; IG: 45.7%, CG: 43.8%	Syrian refugees, caregiver residing in Lebanon: IG: 38.40±7.90, CG: 38.40±7.90; IG: 67.9%, CG: 76.9%	A: 67, IG: 35, CG: 32	EASE	Lay-guided	(E)TAU	A: Caregiver-rated distress, depressive symptoms, functional impairment, general distress, PTSD symptoms, wellbeing
					C: 54, IG: 28, CG: 26				C: Corporal punishment, general distress, inconsistent discipline, parental involvement, poor monitoring, positive parenting
									Implementation: Acceptability, feasibility, fidelity, appropriateness, barriers of implementation
Fine et al. [18] (cRCT)	Tanzania	Political crisis in Burundi	Burundian refugees: IG: 12.70±1.40, CG: 12.00±1.50; IG: 63.2%, CG: 47.7%	Burundian refugees: IG: 40.30±11.70, CG: 38.60±8.02; IG: 77.1%, CG: 79.3%	A: 82, IG: 38, CG: 44	EASE	Lay-guided	(E)TAU	A: Externalizing symptoms, functional impairment, general distress, internalizing symptoms, PTSD symptoms, somatic complaints, wellbeing
					C: 64, IG: 35, CG: 29				C: Corporal punishment, general distress, inconsistent discipline, parental involvement, poor monitoring, positive parenting
									Implementation: Acceptability, feasibility, fidelity, appropriateness, barriers of implementation
Hamdani et al. [65] (cRCT)a	Pakistan	Humanitarian crises (mixed stressors)	Adolescents residing in Pakistan: IG: 13.69±0.85, CG: 13.50±0.68; IG: 55.2%, CG: 46.7%	Caregiver residing in Pakistan: IG: NR, CG: NR; IG: 100%, CG: 100%	A: 59, IG: 29, CG: 30	EASE	Lay-guided	Waitlist	A: Attention problems, depressive symptoms, externalizing symptoms,
					C: 59, IG: 29, CG: 30				general distress, internalizing symptoms, somatic complaints, wellbeing
									C: Corporal punishment, inconsistent discipline, parental involvement, poor monitoring, positive parenting
									Implementation: Acceptability, feasibility, fidelity, facilitators of implementation
Effectiveness trials (fully powered (c)RCTs)
Bryant et al. [61, 62, 68]a	Jordan	Syrian civil war	Syrian refugees: IG: 11.65±NR, CG: 11.63±NR; IG: 49.7%, CG: 49.3%	Syrian refugees: IG: 38.00±7.30, CG: 38.40±7.50; IG: 96.2%, CG: 96.9%	A: 471, IG: 185, CG: 286	EASE	Lay-guided	(E)TAU	A: Attention problems, caregiver-rated distress, depressive symptoms, externalizing symptoms, functional impairment, general distress, internalizing symptoms, PTSD symptoms, school engagement, wellbeing
					C: 471, IG: 185, CG: 286				C: Corporal punishment, general distress, inconsistent discipline, parental involvement, poor monitoring, positive parenting
									Implementation: Feasibility, fidelity
Hamdani et al. [46] (cRCT)a	Pakistan	Humanitarian crises (mixed stressors)	Adolescents residing in Pakistan: IG: 13.89±0.75, CG: 13.85±0.75; IG: 45.7%, CG: 50.4%	Caregiver residing in Pakistan: IG: NR, CG: NR; IG + CG: 95.4%	A: 566, IG: 282, CG: 284	EASE	Lay-guided	Waitlist	A: Attention problems, depressive symptoms, externalizing symptoms, general distress, internalizing symptoms, somatic complaints, wellbeing
					C: 566, IG: 282, CG: 284				C: Corporal punishment, inconsistent discipline, parental involvement, poor monitoring, positive parenting
									Implementation: Feasibility, fidelity
Jordans et al. [59, 63, 64, 67]a	Lebanon	Humanitarian crises (mixed stressors)	Syrian refugees, adolescents residing in Lebanon: IG: 11.70±1.30, CG: 11.90±1.30; IG: 46.3%, CG: 50.9%	Syrian refugees, caregiver residing in Lebanon: IG: 38.00±7.40, CG: 38.80±8.20; IG: 83.9%, CG: 82.8%	A: 198, IG: 80, CG: 118	EASE	Lay-guided	(E)TAU	A: Attention problems, caregiver-rated distress, depressive symptoms, externalizing symptoms, functional impairment, general distress, internalizing symptoms, PTSD symptoms, wellbeing
					C: 155, IG: 62, CG: 93				C: Corporal punishment, general distress, inconsistent discipline, parental involvement, poor monitoring, positive parenting
									Implementation: Acceptability, feasibility, fidelity, barriers of implementation

If more than one report from a single (c)RCT was used for our review, all reports are cited in this table. Lay-guided indicates non-specialist facilitators with different educational backgrounds (e.g., high school education, Bachelor degree holders, graduate degrees in psychology) and/or experience in delivering psychological interventions.

A, adolescents; C, caregivers; cRCT, cluster-randomized trials; (E)TAU, (enhanced) treatment as usual; EASE, Early Adolescent Skills for Emotions.

^aParticipants in these studies were most likely also affected by COVID as an additional stressor during the study period.

On weighted average, adolescents were 12.6 years old (SD 1.1, range: 11.6–13.9 years), and 48.8% (range: 44.1–54.9%) of the adolescent participants self-identified as girls. Caregivers were on average 38.3 years old (SD 8.0, range: 37.3–39.5 years), and 92.8% (range: 72.2–96.6%) self-identified as women. All participants were exposed to significant stressors, including forced displacement after prolonged periods of violence [17–19, 67, 68] and chronic adversities [19, 46, 65, 67], with the COVID-19 pandemic also falling into the study period for some studies [46, 65, 67, 68]. The weighted mean of traumatic events reported by adolescents was 7.5 (SD 4.3, range: 5.0–15.4 events).

All studies employed EASE in community settings, including health-care facilities [18] and educational settings [46, 65]. In line with the EASE manual [15], all studies employed group versions of EASE for both adolescents and their caregivers. In all trials, lays and/or peers delivered the intervention with varying amounts of training and heterogeneous educational backgrounds, with two trials not reporting on the training of intervention providers [46, 65]. In all trials with sufficient information, the duration of the intervention was 7 weeks for adolescents, with planned weekly 90-min sessions. Two trials [19, 67] reported that some sessions took longer (120–135 min). All trials included three sessions for caregivers, with planned 90- to 120-min sessions that were employed weekly [67], biweekly [46, 65, 68], or at given timepoints during intervention delivery for adolescents [17, 19]. Again, three trials [19, 46, 67] reported on variations in session duration, with durations ranging between 90 and 120 min. Five [17–19, 67, 68] out of seven trials used (E)TAU as a comparator, while only two trials [46, 65] employed waitlist designs. (E)TAU included single-session psychoeducation at participants’ homes, with psychoeducation also being a component of EASE, yet delivered in another form and setting.

Quality Appraisal

Risk of Bias

Overall study quality was rated as moderate, but the risk of bias was judged high across all trials, primarily due to concerns regarding outcome measurement, selective reporting, and deviations from intended interventions (see SM5 and SM6). The risk of bias from randomization and missing data were consistently low for adolescents and only partly elevated for caregivers.

Publication Bias

No evidence for publication bias was detected for analyses on adolescent distress, wellbeing, or caregiver-rated distress at post-intervention or follow-up, nor for caregiver general distress. For parenting behavior, rank correlation tests suggested funnel plot asymmetry, but this was attributable to a single outlying effect estimate, and no evidence for a bias emerged at follow-up (see SM7).

Intervention Effects for Adolescents

Six studies² (reporting 38 effect estimates) were included in our analysis on distress indicators at post-intervention assessment for adolescents (see Table 2, Fig. 2). Across all distress indicators, we found no evidence for favorable effects of EASE over (E)TAU and waitlists, SMD = −0.10, 95% CI: [–0.26, 0.06], p = 0.162, with substantial heterogeneity, Q(37) = 54.20, p = 0.034. There was only close-to-significant evidence for between-outcome differences, Q_M(7) = 12.25, p = 0.093. At the single outcome level, effect estimates ranged between SMD = −0.01, 95% CI: [–0.16, 0.14], for PTSD symptoms, and SMD = −0.17, 95% CI: [–0.31, −0.04], for general distress, and SMD = −0.17, 95% CI: [–0.31, −0.04], for internalizing symptoms, with estimates being all very small to small in size and only significant for general distress and internalizing symptoms and close-to-significant for attention problems and somatic complaints. For caregiver-rated distress, wellbeing, and school engagement, we found no evidence for favorable effects (see SM8 for forest plots). The certainty of evidence was very low to low across all outcomes (see SM9). The meta-analysis of five studies comprising 31 effect estimates found no evidence in favor of consistent advantages of EASE over (E)TAU or waitlist conditions regarding distress indicators at 3-month follow-up, except for a small favorable effect on internalizing symptoms, SMD = −0.20, 95% CI: [–0.37, –0.02] (see SM10 and SM11 for details). Long-term data at 12-month follow-up likewise showed no sustained favorable effects [61, 67].

Table 2.

Results of the analyses on adolescents comparing EASE and (E)TAU and waitlists at post-intervention and 3-month follow-up

Analysis	n	k	Effect estimates				Heterogeneity
			SMD	95% CI	95% PI	p value	Q	df	p(Q)	I 2
Post-intervention assessment
Distress indicators	6	38	−0.10	[–0.26, 0.06]	[–0.42, 0.22]	0.162	54.20	37	0.034
Limited to fully powered (c)RCTs	3	21	−0.13	[–0.45, 0.19]	[–0.71, 0.46]	0.228	36.34	20	0.014
Attention problems	4	4	−0.12	[–0.26, 0.02]		0.086				38.5
Depressive symptoms	5	5	−0.05	[–0.18, 0.09]		0.511				38.1
Externalizing symptoms	5	5	−0.07	[–0.21, 0.07]		0.302				38.2
Functional impairment	5	5	−0.11	[–0.26, 0.04]		0.165				33.3
General distress	6	6	−0.17	[–0.31, −0.04]		0.012				37.7
Internalizing symptoms	6	6	−0.17	[–0.31, −0.04]		0.011				42.1
PTSD symptoms	5	5	−0.01	[–0.16, 0.14]		0.903				33.3
Somatic complaints	2	2	−0.16	[–0.33, 0.01]		0.071				37.8
Other outcomes
Caregiver-rated distress	4	4	0.05	[–0.09, 0.20]		0.460	3.53	3	0.317	0.01
School engagement	2	2	−0.34	[–0.95, 0.27]		0.279	4.57	1	0.033	78.1
Wellbeing	6	6	0.11	[–0.08, 0.30]		0.258	12.20	5	0.032	56.3
Follow-up assessment (3 months)
Distress indicators	5	31	−0.13	[–0.36, 0.10]	[–0.58, 0.33]	0.199	46.09	30	0.030
Limited to fully powered (c)RCTs	3	21	−0.13	[–0.61, 0.36]	[–1.03, 0.77]	0.369	35.72	20	0.017
Attention problems	4	4	−0.09	[–0.27, 0.09]		0.315				53.2
Depressive symptoms	5	5	−0.15	[–0.32, 0.03]		0.094				53.4
Externalizing symptoms	4	4	−0.04	[–0.21, 0.14]		0.667				53.3
Functional impairment	3	3	−0.08	[–0.27, 0.10]		0.371				49.0
General distress	5	5	−0.13	[–0.31, 0.04]		0.134				53.3
Internalizing symptoms	5	5	−0.20	[–0.37, −0.02]		0.025				55.8
PTSD symptoms	3	3	−0.13	[–0.32, 0.05]		0.161				49.0
Somatic complaints	2	2	−0.22	[–0.43, −0.01]		0.040				41.7
Other outcomes
Caregiver-rated distress	3	3	−0.03	[–0.18, 0.12]		0.711	0.21	2	0.901	0
Wellbeing	4	4	0.07	[–0.14, 0.28]		0.499	19.05	4	0.001	73.3

The multilevel meta-analysis on distress indicators included attention problems, depressive symptoms, externalizing symptoms, functional impairment, general distress, internalizing symptoms, PTSD symptoms, and somatic complaints. Due to qualitative differences, wellbeing, school engagement, and caregiver-rated distress were analyzed separately. For distress indicators, negative SMDs indicate favorable effects of an intervention [i.e., lower distress in the EASE group compared to (E)TAU/waitlist]. For wellbeing and school engagement, positive SMDs indicate favorable intervention effects [i.e., higher levels of wellbeing and school engagement in the EASE group compared to (E)TAU/waitlist]. All tests and reported statistics use cluster-robust estimates to account for non-independent effect estimates within studies.

df, degrees of freedom; (E)TAU, (enhanced) treatment as usual; I², heterogeneity index in percentage (range: 0–100%); k, number of effect estimates; n, number of studies; PTSD, posttraumatic stress disorder; Q, Cochran’s Q statistic with p value; SMD, standardized mean difference; 95% CI, 95% confidence interval; 95% PI, 95% prediction interval.

Fig. 2.

Forest plot of the meta-analysis on the effects of EASE [14] on mental distress outcomes for adolescents at post-intervention assessment. Negative effect estimates indicate favorable effects for mental distress outcomes. Except for general distress and internalizing symptoms, no favorable effects emerged. CI, confidence interval; F, feasibility/pilot trials; SMD, standardized mean difference.

Intervention Effects for Caregivers

Six studies (reporting 30 effect estimates) were included in the analyses on parenting behavior at post-intervention assessments. Across all parenting dimensions, we found no evidence for a favorable effect of EASE over (E)TAU and waitlists, SMD = 0.15, 95% CI: [–0.12, 0.41], p = 0.218; however, there was evidence for between-study heterogeneity, Q(29) = 1,126.78, p < 0.001 (see Table 3; Fig. 3). There was no evidence for differences between parenting dimensions, Q_M(4) = 7.77, p = 0.100. When we examined single parenting dimensions, effect estimates ranged between SMD = −0.17, 95% CI: [–0.62, 0.28], for corporal punishment, and SMD = 0.66, 95% CI: [0.20, 1.11], for positive parenting. Except for the medium favorable effect of EASE on positive parenting driven by a single study [68], there was no evidence for favorable effects. Also, for mental distress in caregivers, there was no evidence for a favorable effect of EASE over (E)TAU and waitlists, SMD = −0.12, 95% CI: [–0.43, 0.19], p = 0.435 (see SM12 for forest plots). The certainty of evidence was consistently very low (see SM13). The meta-analysis of five studies with 25 effect estimates found no evidence for overall favorable effects of EASE compared to (E)TAU or waitlists on parenting behavior at 3-month follow-up, except for a small significant effect on inconsistent discipline (see SM14 and SM15 for details). Also, at 12-month follow-up, no sustained benefits of EASE on parenting behavior or caregiver distress were observed [61, 67].

Table 3.

Results of the analyses on caregivers comparing EASE and (E)TAU and waitlists at post-intervention and 3-month follow-up

Analysis	n	k	Effect estimates				Heterogeneity
			SMD	95% CI	95% PI	p value	Q	df	p(Q)	I 2
Post-intervention assessment
Parenting behavior	6	30	0.15	[–0.12, 0.41]	[–1.33, 1.62]	0.218	1,126.78	29	<0.001	​
Limited to fully powered (c)RCTs	3	15	0.23	[–0.43, 0.88]	[–2.88, 3.32]	0.276	1,074.76	14	<0.001	​
Corporal punishment	6	6	−0.17	[–0.62, 0.28]	​	0.462	​	​	​	93.3
Inconsistent discipline	6	6	0.19	[–0.26, 0.65]	​	0.403	​	​	​	93.1
Parental involvement	6	6	0.12	[–0.33, 0.57]	​	0.598	​	​	​	93.3
Poor monitoring	6	6	−0.07	[–0.52, 0.38]	​	0.766	​	​	​	93.3
Positive parenting	6	6	0.66	[0.20, 1.11]	​	0.005	​	​	​	91.8
General distress	5	5	−0.12	[–0.43, 0.19]	​	0.435	10.02	4	0.040	67.8
Follow-up assessment (3 months)
Parenting behavior	5	25	0.09	[–0.11, 0.29]	[–0.52, 0.70]	0.292	182.92	24	<0.001	​
Limited to fully powered (c)RCTs	3	15	0.07	[–0.34, 0.47]	[–0.98, 1.12]	0.553	167.14	14	<0.001	​
Corporal punishment	5	5	0.06	[–0.17, 0.28]	​	0.624	​	​	​	75.8
Inconsistent discipline	5	5	0.30	[0.07, 0.52]	​	0.010	​	​	​	75.1
Parental involvement	5	5	0.05	[–0.18, 0.27]	​	0.689	​	​	​	75.7
Poor monitoring	5	5	−0.02	[–0.25, 0.20]	​	0.843	​	​	​	75.8
Positive parenting	5	5	0.07	[–0.15, 0.30]	​	0.538	​	​	​	75.8
General distress	3	3	−0.12	[–0.35, 0.12]	​	0.218	2.34	2	0.311	17.0

The multilevel meta-analysis on parenting indicators included corporal punishment, inconsistent discipline, parental involvement, poor monitoring, and positive parenting. Due to qualitative differences, caregivers’ general distress was analyzed separately. For parenting, positive SMDs indicate favorable intervention effects [i.e., lower corporal punishment, less inconsistent discipline and poor monitoring as well as higher levels of parental involvement and positive parenting in the EASE group compared to (E)TAU/waitlist]. For general distress, negative SMDs indicate favorable effects of an intervention [i.e., lower distress in the EASE group compared to (E)TAU/waitlist]. For parenting, tests and reported statistics use cluster-robust estimates to account for non-independent effect estimates within studies.

df, degrees of freedom; (E)TAU, (enhanced) treatment as usual; I², heterogeneity index in percentage (range: 0–100%); k, number of effect estimates; n, number of studies; Q, Cochran’s Q statistic with p value; SMD, standardized mean difference; 95% CI, 95% confidence interval; 95% PI, 95% prediction interval.

Fig. 3.

Forest plot of the meta-analysis on the effects of EASE [14] on parenting outcomes for caregivers at post-intervention assessment. Single outcomes represent subdomains of the Alabama Parenting Questionnaire [70]. Positive effect estimates indicate favorable effects for parenting outcomes (effect estimates were inverted for three subscales). Except for positive parenting, no favorable effects emerged. CI, confidence interval; F, feasibility/pilot trials; SMD, standardized mean difference.

Adverse Effects of EASE

Three trials found no adverse effects of EASE either for adolescents or caregivers [17, 46, 61, 65]. Four trials [18, 19, 63, 67, 68] reported adverse effects (e.g., physical abuse and violence experienced by adolescents, thoughts of suicide by adolescents or caregivers), which were in most cases unrelated to the intervention. In 1 case [18], the adverse effect was linked to the intervention, with an adolescent being harassed for not receiving EASE. Among those trials reporting adverse events, the number varied between 6 [68] and 24 [18].

Sensitivity Analyses

We examined whether smaller or larger between-outcome correlations (ρ = 0.40, ρ = 0.80) in multilevel models would have changed our results, which was neither the case for adolescent nor caregiver outcomes (see SM16). Only for two studies using waitlist controls [46, 65], analyses on mental distress at follow-up assessment yielded evidence for favorable effects, SMD = −0.33, 95% CI: [–0.48, −0.18], Q_M(2) = 18.03, p = 0.001. The same applied to parenting behavior at follow-up assessments, where effects were favorable for both studies employing waitlist controls, SMD = 0.23, 95% CI: [0.23, 0.23], Q_M(2) = 62.41, p < 0.001. When our analyses were limited to effectiveness studies excluding pilot and feasibility trials [17–19, 65], results remained unchanged. The same applied to analyses that were limited to effect estimates with low risk of bias (see SM17).

To corroborate our frequentist meta-analyses, we re-analyzed overall distress, depressive symptoms, general distress, and internalizing symptoms at post-intervention and 3-month follow-up using a Bayesian random-effects model (see Table 4). Specifically, we found strong evidence for a very small to small favorable effect of EASE compared to (E)TAU and waitlist controls on internalizing symptoms at post-intervention, µ = −0.21, 95% credible interval [–0.31, 0.00], BF₁₀ = 22.38. No evidence emerged for favorable effects on other key outcomes. Effects on internalizing symptoms were sustained at follow-up, µ = −0.18, 95% credible interval [–0.41, 0.00]; however, the strength of evidence at follow-up was only anecdotal, BF₁₀ = 2.04.

Table 4.

Results of the meta-analysis on key outcomes using a Bayesian approach

Parameter	SMD	95% CI	95% PI	BF10	$\widehat{R}$	ESS
Post-intervention assessment
Distress indicators (k = 38 from 6 (c)RCTs)
µ	−0.01	[–0.16, 0.00]	[–0.52, 0.49]	0.15	1.002	1,104
τ	0.16	[0.08, 0.29]	​	∞	1.002	2,649
ρ	0.88	[0.57, 1.00]	​		1.00	5,816
Depressive symptoms (k = 6)
µ	−0.02	[–0.22, 0.00]	[–0.56, 0.51]	0.24	1.00	7,428
τ	0.23	[0.07, 0.48]	​	63.12	1.002	6,827
General distress (k = 6)
µ	−0.05	[–0.28, 0.001]	[–0.58, 0.54]	0.46	1.001	2,804
τ	0.22	[0.00, 0.53]	​	6.52	1.001	3,858
Internalizing symptoms (k = 6)
µ	−0.21	[–0.31, 0.00]	[–0.33, 0.02]	22.38	1.00	2,850
τ	0.02	[0.00, 0.23]	​	0.21	1.001	2,224
Follow-up assessment
Distress indicators (k = 31 from 5 (c)RCTs)
µ	−0.03	[–0.21, 0.00]	[–0.44, 0.38]	0.30	1.002	740
τ	0.18	[0.10, 0.34]	​	∞	1.001	1,675
ρ	0.85	[0.88, 0.50]	​		1.00	4,867
Depressive symptoms (k = 5)
µ	−0.10	[–0.41, 0.00]	[–0.75, 0.59]	0.84	1.001	4,263
τ	0.28	[0.07, 0.60]	​	59.1	1.001	4,567
General distress (k = 5)
µ	−0.02	[–0.28, 0.06]	[–0.71, 0.68]	0.23	1.001	12,874
τ	0.30	[0.13, 0.66]	​	1,427.86	1.003	4,549
Internalizing symptoms (k = 5)
µ	−0.18	[–0.41, 0.00]	[–0.70, 0.53]	2.04	1.001	2,010
τ	0.19	[0.00, 0.60]	​	2.51	1.002	2,028

SMD, standardized mean difference; 95% CI, 95% credible interval (≠ 95% confidence interval); 95% PI, 95% prediction interval (only reported for SMDs); $\widehat{R}$, convergence diagnostic for assessing the convergence of Markov Chain Monte Carlo simulations; ESS, diagnostic for sampling efficiency; BF₁₀, Bayes factor, representing the ratio of the evidence for two competing statistical models and used to quantify support for one model over the other, with BF₁₀ >1 indicating evidence for favorable intervention effects of EASE (µ) or the presence of heterogeneity (τ); (c)RCT, (cluster) randomized controlled trial.

Implementation of EASE

Information on implementation was available from 13 reports [17–19, 46, 59–65, 67, 68]. A detailed summary of all implementation outcomes can be found in the open materials (osf.io/v3md5).

Acceptability

Eight reports [17–19, 59, 60, 62, 63, 65] assessed acceptability, mostly using qualitative interviews, focus groups, and workshops with participants and intervention providers. In four reports [17–19, 62], acceptability was rated as high for adolescents and caregivers, and two reports [63, 65] indicated mixed findings. Participants reported to benefit from sharing their emotions and thoughts [59, 63], and participants as well as providers were positive about the content and structure of the intervention [17–19, 60, 62, 63, 65] as well as the group format [62]. Providers also responded positively to the design of the provider training [62, 63]. Some factors were reported to reduce the acceptability, including repeated diagnostic assessments [17] and unclear a priori information on the structure and content of EASE [17, 19, 63]. Additionally, some providers reported feeling constrained by the manual [62]. Some elements, such as the vicious cycle of inactivity, were reported to be hard to understand and relate, and participants would have enjoyed more interactive components [18, 19, 60, 63]. Caregivers reported concerns with respect to information on suicidality in the adolescent sessions [19, 63], and some participants as well as stakeholders were worried about stigmatization related to mental health problems and illiteracy [63, 65]. Moreover, the organizational and emotional strain for intervention providers was reported to be high [59, 63]. Initial feedback already resulted in modifications for later trials (e.g., earlier start and shortening of caregiver sessions [46, 67, 68]).

Feasibility

Information on feasibility was available from 12 reports [17–19, 46, 59–63, 65, 67, 68], with feasibility being assessed via attendance, retention, and completer rates [18, 19, 46, 60–63, 65, 67, 68] and/or using qualitative interviews and focus groups [17–19, 59, 62, 63, 65]. Adolescents attended 4.37 to 5.39 sessions of the 7-session program, while mean attendance of caregivers was lower, with 1.5 to 1.78 out of 3 caregiver sessions [17, 19, 61–63, 67, 68]. Completer rates were estimated as those adolescents (86%) and caregivers (54–55%) who attended all sessions [19, 65] or a predefined number of sessions (i.e., 5–6 of 7 sessions for adolescents: 63–70%; 2 of 3 caregiver sessions: 48–67%; [19, 46, 63, 67]). Reasons for unattendance were problems with gender-mixed groups, public transport, and competing family or occupational obligations [19, 20, 62, 63, 68]. Some challenges were reported regarding the supervision of intervention providers, including delays in supervision sessions and difficulties attending joint sessions with co-providers [62].

Fidelity

Fidelity was assessed in 9 reports [18, 19, 46, 59, 60, 63, 65, 67, 68]. Most studies employed standardized checklists for fidelity assessment, which were rated by supervisors or self-rated by providers [18, 19, 46, 60, 63, 65, 67, 68]. Results were reported inconsistently, including mean quality ratings per intervention component [46, 68], percentages of completed components [18, 46, 65, 67, 68] or specific ratings [63, 67], and overall qualitative summaries [59, 60, 63]. In all reports, fidelity was consistently reported to be good to very good.

Appropriateness

Five reports [18, 19, 61–63] examined appropriateness through qualitative interviews, focus groups, and workshops. They found that materials were unsuitable for participants with limited literacy skills and insufficiently adapted to age and gender [61–63]. Additional concerns included gender-related restrictions for women [62, 63] and safety issues related to travel [62]. Furthermore, gender-mixed groups, as well as the group size and duration of sessions, were considered inappropriate by some participants [62].

Implementation Costs

One report [64] examined implementation costs relative to (E)TAU using the Client Service Receipt Inventory [71], with some indication of higher costs in the (E)TAU group. However, the findings were inconclusive due to the small sample size.

Other Implementation Outcomes

To date, no evidence is available on the sustainability, penetration, or adaptation of EASE; only recommendations have been reported for those outcomes [62, 63].

Barriers and Facilitators of the Implementation

Nine reports [17–19, 59, 61–64, 67] studied barriers of implementation, while facilitators were only examined in two studies [62, 65]. Most studies used qualitative interviews and focus groups [17–19, 59, 62, 63] to study barriers and facilitators, while a single study derived information from a research team [67]. Main barriers were problems with transport [17, 62] and travel safety [62], competing family or professional obligations [17, 59, 62, 63], economic hardship and generally high levels of adversity [63, 64], deprived basic needs [63], restrictions for girls and women [62, 63], problems with the complex intervention design [59, 63], local conditions (e.g., child labor [19, 59, 63], infrastructure [59, 63]), local attitudes (e.g., mental health stigma [59, 63]), unforeseen critical incidents (e.g., the COVID-19 pandemic [59, 61–63, 67]), problems with recruiting intervention providers [18], and insufficient networking with local institutions (e.g., schools [18]). Facilitators included the involvement of non-specialized providers and caregivers, a safe and comfortable session environment, online sessions for adolescents, online supervision for intervention providers [62], and effective communication with institutions [65]. However, the online format employed during the COVID-19 pandemic was also reported as challenging for some participants [62].

Discussion

This systematic review examined the effectiveness and implementation of EASE [7, 14, 15], a scalable transdiagnostic intervention developed by the WHO and UNICEF, to address internalizing symptoms in stressor-exposed adolescents in low-resource settings. Based on seven studies comprising 1,502 adolescents and 1,428 caregivers, we found no evidence for favorable effects of EASE on overall mental distress outcomes, wellbeing, and school engagement in adolescents. At the single outcome level, we found strong evidence for very small to small favorable effects on internalizing symptoms, which appeared to remain stable at 3-month follow-up and were at least close-to-significant in sensitivity analyses. Effects on parenting behavior were inconsistent and driven by outliers [46, 68], with evidence for small to large favorable effects on positive parenting at post-intervention assessment and very small to medium favorable effects on inconsistent discipline at follow-up. No evidence emerged for favorable effects on caregiver distress.

Our synthesis on implementation outcomes suggested overall good acceptability, feasibility, and fidelity of EASE. Initial problems with acceptability related to intervention content have already been addressed by later modifications of the intervention (e.g., a larger number of interactive elements and modified storybooks [19, 67]). Additional problems resulted from low participation of caregivers, reflected in low attendance and completer rates [46]. For this reason, the sessions for caregivers were shortened as part of the modifications. However, as qualitative data suggested that low caregiver engagement likely results from more fundamental issues (e.g., deprived basic needs, lack of transport, competing obligations, illiteracy), modifications may not be sufficient to solve these problems. By contrast, the (E)TAU conditions [18, 19, 67, 68] might be more suitable to address caregivers’ needs as the 1-session psychoeducation sessions were delivered at home, took less time, likely required fewer literacy skills, and might be easier to combine with competing obligations. The difficulties with caregiver engagement in EASE may also account for the missing evidence for favorable effects on parenting behavior.

In line with a recent meta-analysis [26], intervention effects in adolescents were very small to small and only found consistently at post-intervention and follow-up assessment for internalizing symptoms (SMDs = −0.17 to −0.20), with effects being numerically smaller than for a comparable WHO intervention for adults [24, 25]. In our frequentist analyses, very small to small favorable effects on general distress were significant at post-intervention assessment but non-significant at 3-month follow-up, while effects on somatic complaints were only significant at follow-up. No evidence for a favorable effect of EASE emerged for depressive symptoms as an important facet of internalizing symptoms [72]. Thus, overall favorable effects on internalizing symptoms might be driven by anxiety symptoms, which were only examined in one trial as a single outcome [46]. This study showed significant favorable effects for anxiety, but not for depressive symptoms. These findings were corroborated by Bayesian analyses, which provided strong evidence for short-term favorable effects on internalizing symptoms. At 3-month follow-up, the effects persisted, although the evidence was weaker. Thus, EASE only partly meets its goal of effectively reducing internalizing symptoms in stressor-exposed adolescents. We found no evidence for favorable effects on attention problems, externalizing symptoms, functional impairment, and PTSD symptoms. This is in line with previous evidence on other transdiagnostic psychosocial interventions in children and adolescents [9, 11, 73] and low-intensity transdiagnostic programs in adults [24, 25] and may suggest that those symptom domains need treatments specifically targeting behavioral problems and trauma memory [74]. This may also contribute to the overall, at most small, favorable effects on adolescent mental health outcomes as symptoms of trauma and dissociation are highly prevalent in clinical samples [75] and among adolescents exposed to man-made trauma [76, 77].

Different reasons may account for the small and partly inconsistent effects of EASE. First, in some studies [17–19], distress levels of participants were relatively low at baseline, which may have resulted in floor effects. Second, the comparator used in most studies, a 1-session psychoeducation delivered at participants’ homes (i.e., (E)TAU), might also be an effective intervention for reducing mental distress. This notion is evidenced by a post hoc subgroup analysis showing that (E)TAU outperformed EASE on reducing depressive and internalizing symptoms in those with heightened depressive symptoms, and also in the total sample, EASE was no longer superior to (E)TAU at 12-month follow-up [61, 67]. Those findings might be accounted for by the higher acceptability and feasibility of (E)TAU for caregivers. Moreover, our sensitivity analysis revealed a trend toward more favorable effects for the two studies involving waitlist controls [46, 65]. If this holds true in future trials, EASE is unlikely to be cost-effective as 1-session psychoeducation at home can be delivered at lower costs, even though healthcare costs may be slightly higher following (E)TAU [64]. Third, the inconsistent effects on caregiver outcomes might suggest that EASE fails to produce significant changes in caregiver behavior, which might be a meaningful mechanism of change in this age group [78]. In line with this notion, Bryant et al. [68] found that a larger decrease in inconsistent parenting behavior was associated with reduced internalizing symptoms and attentional problems in adolescents. Additional analyses are needed to link the effects on adolescent and caregiver outcomes. Fourth, problems with implementation may account for small intervention effects. Even though implementation was generally evaluated as good, barriers resulted from the stressors that may also cause the need for intervention (e.g., economic hardship, low-resource local infrastructure, mental health stigma). In the face of highly adverse living conditions, EASE might be at risk for putting too much (additional) strain on participants. The finding that a large number of adolescents had comparably low distress levels [17–19] might also be interpreted as an effect of self-selection, with more distressed adolescents deciding not to participate.

Additionally, our synthesis on implementation outcomes provides insights that could inform further improvements of EASE, potentially enhancing the magnitude of intervention effects. For instance, participants noted that the program lacked adaptation to age and gender, and girls and women reported difficulties attending sessions due to gender-related restrictions [62, 63]. Future adaptations of EASE might, therefore, include tailoring intervention components to age and gender (e.g., incorporating a girl character into intervention materials) and introducing targeted support measures to facilitate participation of girls and women in group sessions. However, such modifications should be carefully considered in collaboration with the expert teams that have developed and implemented EASE [14, 15].

The results of this review should be considered in light of some limitations. First, this evidence synthesis is based on only seven trials, which limits the generalizability of our findings. Although all trials evaluated a single intervention – likely reducing heterogeneity compared with similar reviews [9, 11, 20–23] – substantial between-study differences (e.g., stressor exposure, mental distress levels, competencies of lay providers) remained, which may not have been fully captured in our analyses. Second, our analyses are not suited to conclude that favorable effects of EASE are definitively absent as they likely lack the statistical power to robustly test the null hypothesis [79]. We attempted to address this limitation through post hoc Bayesian meta-analyses; however, our findings indicate only that there is currently no evidence of favorable effects for many outcomes, rather than robust evidence supporting null effects. Third, the quality of included studies was moderate, except for outcome assessment, which caused an overall high risk of bias, and the certainty of evidence was very low to low across all outcomes. Fourth, evidence on implementation outcomes comes from pilot and effectiveness trials; findings from later research phases (i.e., dissemination, scaling-up) are not yet available. Those may provide deeper insights into barriers and facilitators of implementation. Fifth, to our knowledge, at least one trial [66, 80] is still ongoing, and we cannot exclude that findings of future studies may change our results. Thus, our systematic review should be updated when a larger evidence base becomes available. Such studies will allow more robust conclusions as to whether the absence of evidence for favorable effects beyond internalizing symptoms reflects a true lack of effects on other outcomes or is instead due to limited statistical power.

This raises the question of what practical implications can be drawn from our findings. We believe that the current evidence base does not allow for strong claims in either direction, and our results should not be misinterpreted as a call to discontinue EASE. While we found strong evidence for favorable short-term effects of EASE on internalizing symptoms, these effects were only very small to small in magnitude, and the strength of evidence diminished at follow-up. At the same time, implementation findings were encouraging as EASE has been delivered successfully even in highly challenging settings [12].

There is a clear need for further high-quality RCTs evaluating the most recent version of EASE across diverse populations and contexts. Such studies – ideally integrating effectiveness and implementation research through hybrid designs [13] – will provide a stronger basis for concrete policy implications, which may include either discontinuation of EASE or its further adaptation and improvement. Future trials, or pooled analyses across a larger number of studies [81], should also address the crucial aspect of cost-effectiveness using sample sizes large enough to allow valid conclusions [64]. Such analyses are essential given the robust, yet only very small to small, favorable effects of EASE on its primary outcome, which might challenge its cost-effectiveness.

More broadly, there remains an urgent and unmet need for low-intensity interventions that support stressor-exposed youth in maintaining or regaining mental health [7]. EASE may have the potential to address this need, although further adaptations may be required. Moreover, the challenging situation of adolescents living in LMIC may also require the combination of multiple measures and stepped-care approaches [82]. Clinical guideline development teams of the WHO [83] and other institutions can build on this systematic review by integrating current scientific evidence with the expertise of relevant stakeholders to derive practical recommendations.

Conclusion

Based on seven studies evaluating the effectiveness of EASE, we found strong evidence for very small to small favorable effects compared with 1-session at-home psychoeducation and waitlist controls on internalizing symptoms, with effects persisting at short-term follow-up. We found no evidence for consistent favorable effects on other outcomes, including caregiver behavior. Effects were more favorable in trials using waitlist controls. Our findings indicate that while EASE is acceptable and feasible in challenging settings, its incremental benefit over a brief psychoeducation session might be modest, raising questions about its cost-effectiveness. Substantial within-study heterogeneity further suggests that intervention effects likely vary across individuals. Future in-depth individual participant data meta-analyses, as well as high-quality primary studies, are needed to clarify the relationship between caregiver and adolescent outcomes and to identify participant-level factors associated with more favorable effects of EASE.

Acknowledgments

We thank Alizé Bédier, Marie Corneli, and Jan Broll for their assistance with the literature search.

Statement of Ethics

A statement of ethics is not applicable because this study is based exclusively on published literature. A consent to participate statement is not applicable because this study is based exclusively on published literature. All studies included in our systematic review obtained informed written consent from participants in line with the World Medical Association Declaration of Helsinki.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

This systematic review received no external funding.

Author Contributions

L.M.S.: conceptualization (supporting), methodology (equal), investigation (lead), formal analysis (supporting), data curation (lead), writing – original draft (equal), and project administration (lead). C.G.S.: formal analysis, data curation, and writing – review and editing (equal). C.F.B.: investigation, data curation, and writing – review and editing (equal). K.L.: conceptualization (equal), methodology (supporting), resources (lead), writing – review and editing (equal), and supervision (equal). S.K.S.: conceptualization (equal), methodology (equal), resources (supporting), writing – original draft (equal), visualization (lead), and supervision (equal).

Funding Statement

This systematic review received no external funding.

Data Availability Statement

This systematic review was based on data published in previous studies. The data extracted from primary studies for meta-analysis on qualitative synthesis on implementation outcomes as well as all code used for our analyses can be found in the Open Science Framework project (osf.io/v3md5).