Background: Education plays a pivotal role in empowering individuals with the knowledge and skills needed for careers, economic progress, and societal engagement. Dropping out of school before achieving a qualification undermines these opportunities and has an impact on individuals and society (Audit Commission, 2010; OECD, 2023). International and national policies prioritize reducing dropout rates. For example, the 2021 target of the European Council Resolution is to reduce the dropout rate below 9% by 2030 (Council Resolution, 2021). School dropout is a complex, multi-faceted issue influenced by interrelated factors at the individual, family, school, and community levels (Bronfenbrenner, 1989; Dupéré et al., 2020). Structured interventions have been developed to prevent school dropout, especially in Anglo-Saxon countries. Assessing the effectiveness of those programs is challenging, as it needs longitudinal evaluations to examine the long-term impact on school completion. Consequently, many studies focus on predictors of dropout, such as student academic achievement, attendance, and behaviors (Battin-Pearson, 2000; EASNIE, 2019; Rumberger, 2011). We conducted a meta-analysis on the effects of school dropout programs on student academic achievement. There are not recent meta-analyses focusing on the impact of school dropout programs on student performance. Wilson et al. (2011) and Tanner-Smith and Wilson (2013) assessed the impact of prevention and intervention programs on school dropout rates and student absenteeism. More recently, Wang et al. (2024) focused on the effects of programs on high school completion. Purpose: This review examined the impact of prevention and intervention programs targeting school dropout on student academic achievement. Since the effects of educational interventions vary across different student populations and contexts, our aim was to explore this variation to understand for whom and under what conditions these programs work. Our research questions were: RQ1. What is the average effect of dropout programs on K-12 student achievement? RQ2. How heterogeneous are the effects of dropout programs? RQ3. What program types are more effective for which students in improving student academic achievement? We identified six key features based on dropout prevention theory and research (Table 1) that are critical to consider (e.g., Dupéré et al., 2015; Rumberger, 2011; Wilson et al., 2011). We then outlined four categories of dropout programs based on combinations of these features in order to explore how they may work and for whom (Figure 1). We expected high-dosage, personalized interventions tailored to the needs of at-risk students to be the most effective program category. We also expected multi-strategy programs involving student families and communities to have a significant positive effect. Method: The review protocol was preregistered on ("blinded"). Search Strategy: A comprehensive search was conducted to retrieve both published and unpublished studies. The search strategies included: (a) database search (e.g., ERIC, Education Source, and ProQuest Dissertations and Theses), (b) searches on websites of research firms and educational associations, (c) backward and forward citation chasing in previous reviews on school dropout, and (d) contacting prominent researchers in the field. Eligibility Criteria: To be included, studies needed to focus on the following: (a) Students in K-12. Eligible programs may be targeting students, their teachers, parents, school leaders, or other relevant stakeholders, but need to evaluate students' outcomes. (b) School-based, school-affiliated, and community-based programs that perform actions with the expectation of having a beneficial impact on staying in school. (c) Business-as-usual or regular school practice as comparison condition. (d) Quantitative outcome measures of student academic achievement. (e) Studies published from 2011 onwards with no geographical restrictions. (c) Group design studies, including Randomized Control Trials (RCTs) or a Quasi-Experimental Designs (QEDs). Data Collection and Analysis: We selected relevant studies using a two-step process (i.e., title and abstract screening and full-text review), with each record reviewed by two independent reviewers. A draft codebook was developed based on the PICOS framework to extract relevant characteristics. Study characteristics were extracted by one coder, and the codes were validated by a second experienced coder. We conducted critical appraisal by coding methodological characteristics showed in Table 2. We used Hedges' g as the effect size measure. For QEDs and RCTs affected by attrition, we computed effect sizes adjusted for pretest scores. We preferred to adjust for covariates also in RCTs, whenever possible. We applied the adjustments described in Hedges (2007) to cluster-assignment studies. Several studies reported multiple effect sizes. To account for the dependence structure of data, we used a Correlated Hierarchical Effects model with Robust Variance Estimation (Hedges et al., 2010; Pustejovsky & Tipton, 2021). We estimated a first model controlling for methodological covariates (i.e., publication status, design, measure type) to obtain the average treatment effect. Heterogeneity was quantified using 95% prediction interval. We used a multiple meta-regression model to explore the potential moderating effect of theory-based categories of programs, controlling for methodological covariates. The analyses were conducted using the R "metafor" package (Viechtbauer, 2010). Results: After the selection process (Figure 2), 46 studies were included. Key characteristics of the studies are displayed in Table 3. We found a weighted average effect size of 0.12, 95% CI [0.07, 0.18]. The 95% prediction interval (95% PI = [-0.32, 0.57]) indicated substantial heterogeneity, thus the need to explore potential moderators of the effect. We are currently working on the heterogeneity analysis by testing the program categories presented in Figure 1 in a multiple meta-regression model. We plan to present the results at the conference, should our abstract be accepted. Conclusions: Preliminary findings suggest that the effect of programs aimed at tackling dropout on student achievement varies substantially. The moderator analysis will play a key role in understanding what program types are effective for which students, providing valuable information for school decision-making. After completing our analysis, we will discuss the results based on the effectiveness of different program categories, the student groups they serve, and their implementation cost. For example, we expect prevention programs at the class/school level to have a smaller effect compared to personalized programs targeting at-risk students. However, prevention programs are much less expensive and may be highly desirable in schools aiming to optimize educational opportunities for all students before the problem manifests.