Abstract Children with moderate acute malnutrition (MAM) are treated based on low weight-for-length z-score (WLZ), low mid-upper arm circumference (MUAC) or both. This study aimed to assess associations of admission criteria and body composition (BC), to improve treatment of MAM. We undertook a cross-sectional study among 6–23 months old Burkinabe children with MAM. Fat-free (FFM) and fat mass (FM) were determined by deuterium dilution and expressed as FFM (FFMI) and FM index (FMI). Of 1,489 children, 439 (29.5%) were recruited by low MUAC only (MUAC-O), 734 (49.3%) by low WLZ and low MUAC (WLZ-MUAC) and 316 (21.2%) by low WLZ only (WLZ-O). Thus, 1,173 (78.8%) were recruited by low MUAC, with or without low WLZ (ALL-MUAC). After adjustments, WLZ-O had 89 g (95% confidence interval (CI) 5; 172) lower FFM compared to MUAC-O. Similarly, WLZ-O had 0.89 kg/m2 (95% CI 0.77; 1.01) lower FFMI compared to MUAC-O, whereas there was no difference for FMI. However, boys included by WLZ-O compared to MUAC-O had 0.21 kg/m2 (95% CI 0.05; 0.38) higher FMI. In contrast, girls included by WLZ-O had 0.17 (95% CI 0.01; 0.33) kg/m2 lower FMI compared to MUAC-O (interaction, p = 0.002). We found that different criteria for admission into MAM treatment programmes select children with differences in BC, especially FFMI. Trial registration: ISRCTN42569496.