**Background:** Global aquaculture production reached 82 million tonnes valued at USD 250 billion in 2018, with over 20 million people engaged in the sector. Women account for 19% of the aquaculture workforce. Despite environmental concerns including mangrove deforestation and chemical use, aquaculture is promoted as a pro-poor activity with potential to address poverty, nutrition, and gender inequality. However, empirical evidence on its impacts remains scarce and mixed. This systematic review aims to fill this gap by synthesizing evidence on aquaculture interventions in low- and middle-income countries across four domains: productivity, income, nutrition, and women's empowerment.
**Methods:** The review follows Campbell Collaboration MECCIR standards and the 'effectiveness+' framework. To address research questions 1–3 (impact on outcomes), the authors will include randomized controlled trials (RCTs) and nonrandomized studies using methods that control for confounding: regression discontinuity, instrumental variables, difference-in-differences, fixed effects, interrupted time series, and matching approaches. Studies must be set in low- and middle-income countries, published from 1980 onward, in any language. The intervention definition is broad, covering any project, programme or policy providing new or improved aquaculture activities along the value chain. Primary outcomes are grouped into productivity (e.g., yield, technology adoption), income (e.g., earnings, consumption expenditure), nutrition (e.g., dietary diversity scores, anthropometry), and women's empowerment (e.g., Women's Empowerment in Agriculture Index, decision-making). Secondary outcomes include environmental and social measures. To address research question 4 (barriers and facilitators), the authors will include supplementary evidence from process evaluations, formative assessments, and project documents. For research question 5 (cost-effectiveness), full and partial economic evaluations will be sought. Searches will cover academic databases (3ie Development Evidence Portal, EBSCO, Econlit, Scopus) and grey literature sources (FAO, WorldFish, USAID, IFPRI, ADB, and others). Title and abstract screening will use machine learning priority screening in EPPI-Reviewer 4, with double screening of a random test set and records with ≥20% inclusion probability. Full-text screening will be double-screened independently. Data extraction will be double-coded for quantitative data. Risk of bias will be assessed using the 3ie risk of bias tool, covering baseline confounding, selection bias, attrition, performance bias, measurement bias, and reporting bias. Studies will be rated as 'Low risk of bias', 'Some concerns', or 'High risk of bias'. Effect sizes will be calculated as standardized mean differences (Cohen's d, adjusted to Hedges' g) using appropriate formulae for continuous, binary, and regression-based outcomes. Unit of analysis errors will be corrected using clustering adjustments. Meta-analysis will use inverse-variance weighted random-effects models, conducted in R using metafor and robumeta packages. Dependent effect sizes will be handled via robust variance estimation (RVE) where feasible (≥4 degrees of freedom), or through data selection techniques otherwise. Heterogeneity will be assessed using Q, I², and τ² statistics. Subgroup analyses and meta-regression will explore heterogeneity by extrinsic (funder, publication type), methodological (study design, risk of bias), and substantive characteristics (gender, socioeconomic status, land ownership, intervention type, geography). Sensitivity analyses will examine the influence of single studies, high risk of bias studies, and outliers. Publication bias will be assessed using contour-enhanced funnel plots, Egger's test, and comparison with pre-registration records. Qualitative evidence will be appraised using an adapted CASP checklist. Cost data will be synthesized narratively, with cost-effectiveness analyses conducted only for interventions with statistically significant effects.