Background: Current (1999) World Health Organization guidelines recommend giving routine antibiotics (AB) for all children with severe acute malnutrition (SAM), even if they have uncomplicated disease with no clinically obvious infections. We examined the evidence behind this recommendation.
Methods and findings: OVID-MEDLINE, EMBASE, COCHRANE, GLOBAL-HEALTH, CINAHL, POPLINE, AFRICA-WIDE-NiPAD, and LILACS were searched for AB efficacy, bacterial resistance, and infection rates in SAM. Following PRISMA guidelines, a systematic review and meta-analysis were performed. Three randomised controlled trials (RCT), five Cochrane reviews, and 37 observational studies were identified. One cohort-study showed no increase in nutritional-cure and mortality in uncomplicated SAM where no AB were used. (p>0.05). However, an unpublished RCT in this setting did show mortality benefits. Another RCT did not show superiority of ceftriaxone over amoxicilllin for these same outcomes, but adressed SAM children with and without complications (p = 0.27). Another RCT showed no difference between amoxicillin and cotrimoxazole efficacies for pneumonia in underweight, but not SAM. Our meta-analysis of 12 pooled susceptibility-studies for all types of bacterial isolates, including 2767 stricly SAM children, favoured amoxicillin over cotrimoxazole for susceptibility medians: 42% (IQR 27-55%) vs 22% (IQR 17-23%) and population-weighted-means 52.9% (range 23-57%) vs 35.4% (range 6.7-42%). Susceptibilities to second-line AB were better, above 80%. Prevalence of serious infections in SAM, pooled from 24 studies, ranged from 17% to 35.2%. No study infered any association of infection prevalence with AB regimens in SAM.
Conclusions: The evidence underlying current antibiotic recommendations for uncomplicated SAM is weak. Susceptibility-studies favour amoxicillin over cotrimoxazole. However, given that these antibiotics have side-effects, costs, and risks as well as benefits, their routine use needs urgent testing. With reliable monitoring, we believe that there is sufficient equipoise for placebo controlled RCTs, the only robust way to demonstrate true efficacy.