We quantitatively synthesized findings of quasi-experimental and randomized control studies of interventions designed to improve the mathematics achievement of secondary school students with mathematics difficulties (MD). We identified 45 studies (49 interventions) published between 1978 and 2020 and classified interventions into five categories: technology-based interventions (TBI), schema-based interventions (SBI), use of visual representations, cognitive-based instruction, and other. We used robust variance estimation (RVE) to address dependency in effect sizes (ESs). The cumulative effect size across 139 outcomes was moderately large (g = 0.52), with a substantial amount of between-study heterogeneity. A multiple meta-regression analysis showed several significant moderators of interventions' effectiveness, such as content domain, intervention length, and dependent measure. We discuss limitations and implications for research and classroom practice.