BACKGROUND: Globally, some 45 million children under five years of age are wasted (low weight-for-height). Although 2023 World Health Organisation guidelines on their care did not aim to identify optimal weight gain, they did mention 5-10 g/kg/day as a target, which is a change from prior guidelines that recommended 10-15 g/kg/day, when inpatient-only care was the norm. We aimed to inform future policy/programming on weight gain targets. METHODS: For this systematic review and meta-analysis, we searched Embase, Global Health and Medline. The final search was on 23/02/2024. Papers were included if they reported weight gain of children aged 6-59 months with severe malnutrition during inpatient (facility-based), outpatient (home-based), and hybrid treatment (initially inpatient and progressing to outpatient treatment). Summary data were extracted, and quality was assessed using a NICE Quality Appraisal Checklist. Our primary outcome was mean rate of weight gain (g/kg/day) during treatment. We conducted random-effects meta-analysis to describe pooled mean weight gain by programme type. Meta-regression investigated potential associations of weight gain with length of stay and programme outcomes. We registered the study on PROSPERO (CRD42023266472). FINDINGS: Our search yielded 3173 papers. We reviewed 321 full texts, identifying 126 eligible papers. Of these, 104 papers, including some 240,650 participants, reported weight gain as g/kg/day and were eligible for meta-analysis. Mean rate of weight gain was 8.8 g/kg/day (95% CI: 7.6, 9.9; I2 = 97.8%) across 18 inpatient programmes, 3.4 g/kg/day (95% CI: 2.0, 4.7; I2 = 99.4%) across 12 hybrid programmes, and 3.9 g/kg/day (95% CI: 3.4, 4.4; I2 = 99.7%) across 60 outpatient programmes. We found inconsistent evidence of an association between slower weight gain and higher mortality: there was weak evidence of association after adjusting for programme type (coefficient = -0.4; 95% CI: -0.7, -0.02; p = 0.04; n = 118 programmes). There was high heterogeneity between studies. Details of weight gain calculation methods varied. We found no evidence for publication bias when accounting for programme type (Egger's test p-value = 0.2). INTERPRETATION: Weight gain in outpatient programmes was markedly slower than in inpatient treatment. Clearer reporting of weight gain and a better understanding of the sequelae of faster/slower recovery is important to set future weight gain targets. Our results set an important baseline for current programmes to benchmark against. FUNDING: Medical Research Council/Global Challenges Research Fund, grant number: MR/V000802/1.