Treatment of severe and moderate acute malnutrition in low- and middle-income settings: a systematic review, meta-analysis and Delphi process

We developed comprehensive search strategies for the following databases: Medline, Embase, Web of Science, WHO regional databases and the Cochrane library (see additional file 1). We conducted hand searches for sources of grey literature, including the Emergency Nutrition Network and Epicentre websites, Grey Literature Review and the World Bank website. We also issued a call to key non-governmental organizations requesting reports from their programs.

Literature published after 1970 was included and we did not restrict by language. Searches were conducted between October 9, 2012 and November 3, 2012. We did not limit by study design type or by publication type when selecting studies for inclusion. However, we excluded before-and-after studies in the meta-analysis, as we were not confident in the abilities of these studies to isolate the intervention effect separately from the confounding variables.

We defined MAM as weight-for-height z-score (WHZ) between -2 and -3 standard deviations (SD), weight-for-height (WFH) 70-80% of the NCHS or WHO reference median or mid-upper arm circumference (MUAC) of 115-125mm. We defined SAM as weight-for-height z-score (WHZ) <-3 SD, weight-for-height (WFH) <70% of the median NCHS or WHO reference or mid-upper arm circumference (MUAC) <115mm or oedema [4].

We contacted authors who used alternative definitions of acute malnutrition or in cases where there was insufficient information available in the publications to request additional information or disaggregated data. An asterisk next to the authors’ names in the forest plots indicates the use of unpublished data.

We coded and categorized the types of interventions in each article. For moderate acute malnutrition, we conducted a meta-analysis only on ready-to-use-supplementary food (RUSF) compared with CSB, as this was the only comparison with multiple studies that could be pooled. Likewise, for severe acute malnutrition, we conducted a meta-analysis on RUTF compared with standard therapy. No study has included true control groups using placebo or no intervention for ethical reasons. We also conducted a meta-analysis on two studies comparing inpatient treatment to ambulatory treatment for children with SAM and MAM, as well as a meta-analysis comparing locally-produced RUTF to imported RUTF for rehabilitation of children with SAM.

We included outcomes needed for LiST and those routinely used for program monitoring. These include: mortality, recovery rate (as defined by authors), relapse rate (as defined by authors), default rate, time to recovery, and change in anthropometric measures such as weight, height, MUAC and WHZ. Outcomes with more than one data point were included in the final analysis.

We used a standardized data abstraction form to collect information regarding study characteristics, descriptions of the interventions and comparisons, outcomes of interest and effects as well as quality of the studies. We assessed quality based on the CHERG adaptation of the GRADE technique at the individual study level and at the outcome level [20]. Studies were classified as high, moderate, low or very low quality. The quality of each study was assessed based on study design, methods and generalizability.

Quality assessment at the outcome level was graded based on volume and consistency of the evidence, strength of the pooled effect and strength of statistical evidence based on the p-value. Levels of heterogeneity were assessed by visual inspection, looking for overlapping confidence intervals, and by the I² value. An I² value of >50% was considered to be evidence of substantial heterogeneity.

The meta-analysis was conducted using RevMan 5.2®. We applied generic inverse variance methods to all analyses and used a random effects model in all cases; summary estimates are presented as relative risk (RR) or mean difference (MD) and 95% confidence intervals (CI).

Three articles, representing two unique trials, were located that compared community-based management with RUTF versus standard therapy in children with severe acute malnutrition [21‐23]. Standard therapy entailed treatment in an inpatient facility until complications resolved, with the subsequent provision of Corn-Soy blend (CSB) for feeding the child at home. Both were quasi-experimental trials set in Malawi. In one trial, all children were treated for infectious and metabolic complications in a nutritional rehabilitation unit; they were enrolled into the trial upon discharge and given either RUTF or CSB for home-treatment [21, 23]. The other trial enrolled children upon discharge from a nutritional rehabilitation unit as well as directly from the community. All of the children in the standard therapy group and about half of the children in the RUTF group had received inpatient treatment [22]. The first did not test children for HIV, but presumably included a mix of children who were HIV-infected and HIV-uninfected and took place from 2002 to 2003 [22]. The other two articles reported data from the same trial, which took place in 2001. One reported data on the HIV-infected cohort [23], while the other reported data for children who were HIV-uninfected [21]. We assessed the quality of the studies as moderate/low [22] and moderate/high [21, 23].

There were no significant differences in mortality (figure 2). Children who received RUTF were 1.51 times more likely to recover (defined as attaining WHZ ≥ -2) than those receiving standard therapy (RR: 1.51, 95% 1.04 to 2.20) (figure 3). There was substantial heterogeneity (I² = 92%), the effect was only marginally statistically significant, and this outcome was graded as low quality (see table 1 for quality assessment). Children who received RUTF had a greater average height gain (MD: 0.14, 95% CI 0.05 to 0.22) and MUAC gain (MD: 0.11, 95% CI 0.05 to 0.18); both outcomes were graded as moderate/low quality. Average weight gain in the RUTF group was also greater (MD: 1.27, 95% CI 0.16 to 2.38) and this outcome was graded as moderate quality (figure 4).

A literature review by Schofield and Ashworth [24] indicates that between the 1950s and 1990s, case fatality rates (CFR) were typically 20-30% among children treated for SAM in hospitals or rehabilitation units. Average CFR was higher (50-60%) among children with oedematous malnutrition. The persistence of high CFR was attributed to faulty case management, and the authors called for prescriptive treatment guidelines as part of a comprehensive training program. In the 2008 Lancet Maternal and Child Undernutrition Series, the preliminary review estimated that treating children according to the WHO Protocol compared to standard care would result in a 48% reduction in deaths (RR 0.52, 95% CI 0.43, 0.64) [13]. This was used in the model to determine impact on mortality for SAM but requires refinement for the reasons mentioned in the introduction.

In terms of recent studies, we found one before/after study [25], two retrospective chart reviews [26, 27], one quasi-experimental study [28] and four prospective cohorts [29‐33] that examined the case fatality rates and recovery rates of children with SAM treated according to the WHO protocol or an adapted WHO protocol. There was also one cluster RCT that compared inpatient treatment to home-care and day-care treatment [34, 35]; this study contained methodological issues and did not adequately describe the intervention (see additional file 3 for study assessment).

None of the studies provided sufficient information to ensure that each step of the WHO protocol was followed and many noted variations from the protocol. One study [31, 32] excluded children with severe complications and thus may not be generalizable. Case fatality rates ranged from 3.4% to 35% (see table 2). The highest CFR stemmed from a cohort of HIV-infected children [31, 32], while the lowest rate was achieved in a study that provided few details on the protocol followed [34]. Only two studies provided information on recovery rates, which were 79.7% and 83.3%, respectively [28, 31, 32].

Ashworth noted that issues of training were paramount to improving outcomes, as mortality rates increased with the influx of new, untrained doctors into a hospital [29]. Two additional observational studies documented that implementing changes to dietary and clinical management did not seem to be sufficient to promote substantial reductions in case fatality rates. Key factors associated with improved outcomes were related to quality of care and institutional culture, including staff morale, attentiveness of nurses and support structures at the managerial level [36, 37].

Our review identified five studies investigating the effect of Ready-to-Use Supplementary Food (RUSF) compared to Corn Soy Blend (CSB) in moderately malnourished children under five years of age [38‐42]. Two of the studies were cluster randomized controlled trials (cRCTs), one set in 10 health centres and health posts in the Sidama zone of Ethiopia [39] and the other in the Dioila health district in Mali [38]. Three of the studies were randomized controlled trials (RCTs). Two were located in southern Malawi [40, 41], and one in the Zinder region of southern Niger [42]. Two studies took place from 2007 to 2008 [38, 42]; the remaining three studies took place during 2009 and 2010 [39‐41]. We assessed the quality of the studies to be low [42], moderate [38], moderate/high [39, 40] and high [41] quality (see additional file 3).

There were no significant differences in mortality for children given RUSF compared to those who received CSB (figure 5). However, the non-response rate was significantly lower in the RUSF group (RR: 0.65, 95% CI 0.47 to 0.90). This outcome has considerable heterogeneity (I² = 57%) and was graded as moderate quality (see table 3). Children in the RUSF group were also significantly more likely to recover (RR: 1.11, 95% CI 1.04 to 1.18), although this estimate contained substantial heterogeneity and was graded as moderate/low quality (figure 6). The rate of height gain did not differ between the intervention and comparison groups. Children who received RUSF had an average MUAC gain of 0.04 mm per day (MD: 0.04, 95% CI 0.01 to 0.07) and had an average weight gain of 0.61 g/kg/d higher (MD: 0.61, 95% CI 0.24 to 0.99) than those who received CSB (figure 7). Upon discharge or completion of the study, children who received RUSF had an average weight-for-height z-score that was 0.11 z-scores greater than those who received CSB (MD: 0.11, 95% CI 0.04 to 0.17). While this is statistically significant, it may not be a sufficient difference to be clinically important. Nackers et al. [42] followed-up with children 6 months post-intervention. There were no significant differences in sustained recovery (defined as maintaining WFH ≥ 80% of the NCHS median post-treatment) or height gain.

The comparison groups in two of the studies used standard CSB [39, 41]. Two of the studies used “CSB++”, which contains a revised micronutrient profile and contains higher quality protein through the addition of milk powder [38, 40] and the last study used “CSB-based pre-mix”, which contains additional oil and sugar [42]. We performed a sensitivity analysis, separating out studies using improved CSB (CSB++ and CSB-based pre-mix) from those using standard CSB. For mortality, the two studies using improved-CSB slightly favoured the comparison group, while the study using standard CSB favoured the intervention group. For the rate of height gain, there is a very slight difference in the direction of effect, but again there was no significant difference between the subgroups. The remaining outcomes showed consistent directions of effect. There were no significant differences between the subgroups for any outcomes.

Two trials, one in Senegal, graded as low quality [43] and the other in Malawi, graded as moderate quality [44], compared imported RUTF to locally produced RUTF used in community-based management of SAM. There was no significant difference in weight gain between intervention groups (figure 8). This effect was consistent (I² = 0%) and the overall outcome was graded as moderate/low quality.

Two studies compared home-based treatment to inpatient treatment in children without severe complications. One moderate quality study in Niamey, Niger, enrolled children with MAM and SAM who were about to be discharged from the hospital [45]. The other study, graded as low quality, allocated children presenting to the nutrition unit in Dhaka, Bangladesh, to receive either home-based or inpatient care [35]. A third arm of this trial, day care, was not analyzed because it could not be pooled.

There was no significant difference in mortality between home-based or inpatient care (figure 9). However, the studies produced opposite directions of effect and the overall quality of this outcome was rated as very low due to issues with study design, analysis and availability of key study details (table 4).

We identified several interesting singular studies that could not be pooled in the meta-analysis. Oakley et al. [46] studied the effect of an RUTF consisting of 25% milk versus another with 10% milk in treating children with SAM, and found that the RUTF with the higher milk content was associated with a statistically significant higher recovery rate (p<0.05). A trial in urban Senegal randomized children with SAM and MAM to receive RUTF or F100 (a therapeutic milk-based product used for nutritional rehabilitation in inpatient facilities) [47]. The study found a statistically significant difference in the rate of weight gain: children who received RUTF gained an average of 5.50 g/kg/d more than those receiving F100 (MD: 5.50, 95% CI 3.00 to 8.00). Branger et al. [48] investigated the effect of adding spirulina, a microscopic algae to standard treatment or standard treatment plus fish for children with moderate and severe malnutrition in Burkina Faso. The authors found no significant differences between treatment groups.

Navarro-Colorado [49] found no significant differences in duration of rehabilitation or weight gain in children with severe acute malnutrition randomized to receive F100 or BP100, a ready-to-use food in biscuit form, although children receiving BP100 had a significantly higher energy intake. Finally, a double-blind, randomized, placebo controlled efficacy trial investigated the effect of adding probiotics and prebiotics to RUTF compared to standard RUTF. The study found no significant difference in the primary outcome, nutritional recovery, or any of the secondary outcomes, including mortality. HIV-infected children were at a higher risk of death in both groups, but HIV did not confound or modify the non-statistically significant effect of the intervention [50].

We found very few rigorous trials that compared the provision of therapeutic or supplementary foods to other types of interventions aimed at modifying the upstream factors that contribute to the development of wasting. Fauveau [51] compared education on appropriate complimentary feeding to education plus supplementary feeding in children aged 6-12 months. The supplementary food package contained rice, wheat, lentil power and cooking oil. While the group receiving the supplemental feeding package had a significantly higher monthly weight gain at three months, this result was not sustained at six months of follow-up.

A new randomized controlled trial comparing RUTF to RUTF plus antibiotics (either amoxicillin or cefdinir) in children with uncomplicated SAM in an outpatient setting found that the mortality rate was significantly higher in children receiving placebo than either antibiotic arm (amoxicillin RR: 1.55, 95% CI 1.07-2.24; cefdinir RR: 1.80, 95% CI 1.22-2.64) [52]. The proportion of children who recovered was significantly lower among the placebo arm compared to either antibiotic arm (amoxicillin: 3.6%, 95% CI 0.6-6.7; cefdinir: 5.8% lower, 95% CI 2.8-8.7), with no significant differences in death or recovery between the two antibiotic arms. Rates of weight gain among children who recovered were higher in the antibiotic arms compared to the placebo arm. HIV status was not known for over half of the children in the study. Additional studies are needed to strengthen the evidence base on whether children with uncomplicated SAM should be provided routine antibiotics.

We received replies from 15 participants in round 1 (83%) and 13 participants responded to round 2 (72% of total, 86% of round 1 participants). All participants provided input on what constitutes ‘optimal care’ of MAM; 13 participants contributed to the mortality and recovery rates in each round.

For inpatient treatment of complicated SAM according to the WHO protocol, the panel of experts estimate a CFR of 14% (range: 5-30%) and recovery rate of 71% (25-95%). The lower bound of the recovery rate is 25% results from one expert who expressed that a large proportion of admissions would default before recovery is reached. For community-based treatment of SAM, the CFR was estimated at 4% (range: 2-7%) and a recovery rate was estimated at 80% (range: 50-93%). The mortality rate for MAM based on optimal treatment proposed by the experts (see additional file 4) was estimated at 2% (range: 0-4%) and recovery rate was 84% (50-100%).

It should be noted that true consensus in estimating CFRs and morality rates was not achieved through this process, as a few participants did not provide a single estimate for each outcome, stating that the intervention effects varied considerably by context. There was a convergence of ideas around the general approach to managing MAM, as illustrated by the major themes described in additional file 4. Consensus was not achieved regarding whether all children with MAM (in areas of high HIV prevalence) should be screened for HIV, the relative importance of each component of the management approach, or the ideal form of food to provide (whether there are other foods that are as effective as RUSF).