NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Picot J, Hartwell D, Harris P, et al. The Effectiveness of Interventions to Treat Severe Acute Malnutrition in Young Children: A Systematic Review. Southampton (UK): NIHR Journals Library; 2012 Apr. (Health Technology Assessment, No. 16.19.)

Cover of The Effectiveness of Interventions to Treat Severe Acute Malnutrition in Young Children: A Systematic Review

The Effectiveness of Interventions to Treat Severe Acute Malnutrition in Young Children: A Systematic Review.

Show details

4Assessment of clinical effectiveness

Titles and, where available, abstracts of a total of 8954 records were screened and full copies of the 224 references were retrieved (because of resource limitations only references in English were selected for retrieval). After inspection of the retrieved references, 150 were excluded (see Appendix 6): 81 because they did not focus on the patient group of interest, two because the intervention was not relevant, 12 because they did not report the necessary outcomes, 62 were because of their design and four because they were abstracts containing insufficient information to judge study quality, methodology and results (references could be excluded for more than one reason). Seventy-four retrieved references/full papers describing 68 studies met the inclusion criteria of the review. The total number of records assessed at each stage of the systematic review screening process is shown in the flow chart of Figure 3.

FIGURE 3. Reference retrieval flow chart.

FIGURE 3

Reference retrieval flow chart. a, For example, bibliographies of included studies and grey literature identified by the advisory group.

As set out in the protocol for this review, the prioritised list of research questions that resulted from the Delphi process formed the basis for this systematic review. Each of the 68 studies that met the general review inclusion criteria was therefore mapped against the list of prioritised questions to provide an overview of the extent of the available evidence (Table 8).

TABLE 8. Evidence available for each of the research questions prioritised in the Delphi process.

TABLE 8

Evidence available for each of the research questions prioritised in the Delphi process.

The available evidence mapped against 9 of the 15 prioritised questions. For one other question (Q19), no studies focused on the topic of interest; however, very limited evidence was available in two other studies. These 10 questions in which evidence was included were as follows:

  • What methods are effective for treating SAM among infants < 6 months old? (Q19, limited information only)
  • Which form of intravenous (i.v.) fluid administration is the most effective for treating shock? (Q21)
  • What are the best treatments for children with SAM who have diarrhoea? (Q22)
  • What methods are effective in treating infection? (Q7)
  • What is the clinical effectiveness of interventions in different settings (e.g. hospital, community, emergency)? (Q14)
  • Which methods for correcting micronutrient deficiencies are effective? (Q8)
  • What is the overall effectiveness of current programmes/guidance (e.g. the WHO 10-step plan)? (Q1)
  • What methods for treating dehydration are effective? (Q5)
  • What are the most effective methods for feeding during the initial stages of treatment? (Q9)
  • Which methods are effective in the rehabilitation phase? (Q10)

No evidence was found to inform the remaining five questions:

  • How should management of HIV-infected children with SAM differ from those who are severely malnourished but HIV–ve? (Q20)
  • What factors affect sustainability of programmes, long-term survival and readmission rates? (Q18)
  • What is the clinical effectiveness of management strategies for treating children with comorbidities such as tuberculosis (TB) and Helicobacter pylori? (other than HIV infection and diarrhoea, which are considered in Q20 and Q22) (Q15)
  • What factors limit full implementation of treatment programmes (e.g. insufficient training, cultural difficulties and funding limitations)? (Q17)
  • What is the effectiveness of different methods for increasing appetite and food intake to recover lost weight and aid catch-up growth? (Q11)

After the available evidence had been mapped against each research question, the final decision on how many questions would be addressed was taken, based on the extent of the evidence and the resources available for the research. It was decided that project resources were available to review the evidence for the first six questions for which any evidence was available.

  • What methods are effective for treating SAM among infants < 6 months old? (limited information only)
  • Which form of i.v. fluid administration is most effective for treating shock?
  • What are the best treatments for children with SAM who have diarrhoea?
  • What methods are effective in treating infection?
  • What is the clinical effectiveness of interventions in different settings (e.g. hospital, community, emergency)?
  • Which methods for correcting micronutrient deficiencies are effective?

For each question, evidence was included from studies with the most rigorous designs based on the hierarchy of evidence. For all but one question, this meant that only RCTs and CCTs were included. The exception was Q7 (What methods are effective in treating infection?), where a RCT and a retrospective cohort study with control were the only two studies that addressed this question, but each one focused on a different aspect of this topic.

The evidence is presented in the remainder of this chapter with each of the six questions reviewed being considered in a separate section. The evidence is presented in the remainder of the chapter, with each of the six questions reviewed.

What methods are effective for treating severe acute malnutrition among infants < 6 months old? (Q19, rank 1 =)

No research focusing on treating SAM of infants < 6 months old was identified. The majority of studies excluded this age group, and most of those which allowed for the inclusion of this age group did not report on outcomes for this subgroup. Two studies47,48 were identified that did include infants < 6 months of age within their study populations and provided some outcome information for this subgroup; however, the information available was very limited (see Appendix 7). Although data were extracted, no formal quality assessment was undertaken. The findings are presented to illustrate the nature of the studies and should be interpreted with caution.

Nu Shwe's retrospective cohort study with control47 described outcomes at a children's hospital in Myanmar before and after the introduction of the WHO's guidelines for SAM. In the year before the introduction of the WHO guidelines (1999), 11.4% of children were < 6 months of age, but this proportion fell in subsequent years to 10.7% in 2000 and to 6.4% in 2001. No baseline data were presented for the group of children < 6 months of age; thus, the comparability of the cohorts in each year is unknown. The only outcome reported for the group of interest is proportional mortality (the number of deaths for each age group expressed as a percentage of all the deaths), but a statistical comparison between the control year, 1999, and the years 2000 and 2001, when the WHO guidelines were in use, is not reported (control year 1999: cases 11.4%, proportional mortality 12%; WHO year 2000: cases 10.7%, proportional mortality 9.1%; WHO year 2001: cases 6.5%, proportional mortality 12.5%). The author comments that the introduction of exclusive breast feeding programmes may have reduced SAM in children < 6 months of age and may also have contributed to the lower proportional mortality in the < 12 months age group in comparison with other age groups. Nu Shwe47 states that, comparatively, the proportional mortality in the age groups < 6 months and 6–12 months (9–24%) was lower than in the 13–24 months and > 24 months age groups (20–50%).

Hossain and colleagues48 described a prospective cohort study with concurrent control in Bangladesh, which compared a locally adapted protocol for treatment of SAM with the WHO protocol. They included children in the age range 2–59 months, but the number of children enrolled who were aged < 6 months is not reported and no baseline characteristics are provided for this subgroup of children; therefore, the comparability of the groups with regard to children aged < 6 months is unknown. The only outcome reported for the group of interest is weight gain. There was no statistically significant difference in weight gain for the < 6 months age group between the treatment arms [mean ± SD weight gain: Institute of Child and Mother Health (ICMH) protocol 17.5 ± 7.5 g/kg/day vs the WHO protocol 11.6 ± 6.8 g/kg/day; p = 0.21]. The mortality rate overall in each group was 6.7%, but mortality was not reported on separately for children aged < 6 months.

Which form of intravenous fluid administration is most effective for treating shock? (Q21, rank 1 =)

Quantity and quality of research available: shock

One RCT was included that investigated the efficacy of fluid resuscitation solutions for treating hypovolaemic shock in children with SAM.49 The key characteristics of the trial can be seen in Table 9, with further details in Appendix 8. The trial was a phase II safety and efficacy RCT conducted in a district hospital in Kenya, and funded by a global charity.

TABLE 9. Characteristics of the included RCT of children with shock.

TABLE 9

Characteristics of the included RCT of children with shock.

Severe acute malnutrition was defined in this RCT as any of W/H z-score < −3 or W/H < 70% of reference median, a MUAC measurement of < 11.0 cm, or oedema involving at least both feet (kwashiorkor). Participants were also required to have evidence of shock and were categorised as having either severe dehydration/shock (shock and severe dehydrating diarrhoea defined as ≥ 6 watery stools/day) or presumptive septic shock (non-diarrhoeal shock). The trial predominantly evaluated Ringer's lactate isotonic fluid (RL) compared with a standard WHO hypotonic fluid solution [half-strength Darrow's in 5% dextrose (HSD/5D)]. Children with severe dehydrating diarrhoea/shock randomly received RL or HSD/5D, whereas those with presumptive septic shock were randomised to RL, HSD/5D or 4.5% human albumin solution (HAS); although limited data were subsequently reported for the HAS group, owing to small study numbers (n = 6). HSD/5D was given according to the WHO recommendation in a maximum of two boluses of 15 ml/kg over 2 hours, whereas the RL group received 10 ml/kg over 30 minutes (up to a maximum of 40 ml/kg where necessary). HAS was administered in the same dosage as for RL. Follow-up was at 8 and 24 hours for the primary outcome, although the children were followed up intensively for up to 48 hours and thereafter for in-hospital mortality.

Other interventions that all participants received included standard WHO management of SAM comprising treatment of hypoglycaemia, antibiotics and oral rehydration solution (ORS) [rehydration solution for malnutrition (ReSoMal)] for those with dehydrating diarrhoea, and maintenance i.v. dextrose fluids up until tolerance of oral feeds was established.

The trial49 was relatively small with 61 participants, although with few data reported on the six children receiving HAS, this number was reduced to 55 for reported baseline characteristics and most outcomes. Children allocated to the RL and HSD/5D treatment groups were around 15 months of age (though it is not clear from the publication whether this is the mean or median), with a slightly higher proportion being boys (58–59%). The mean W/H z-score at baseline ranged from −3.4 to −3.9 and the mean MUAC was approximately 10 cm. Approximately two-thirds of participants had severe wasting, about 40% were HIV sero-positive (HIV+ve) and around 75% fulfilled the strict WHO definition of advanced shock for severely malnourished children. Of the total included population, approximately twice as many children had severe dehydration/shock as had presumptive septic shock, although within the RL and HSD/5D treatment groups there were approximately an even number of children with each type of shock.

The study was limited to children > 6 months of age with SAM and evidence of shock. The clinical shock criteria were defined and included measures such as a capillary refill time (CRT) > 2 seconds, weak pulse volume and deep ‘acidotic’ breathing, among others (see Appendix 8). Children were excluded if they had known congenital heart disease, severe anaemia, clinical features of pulmonary oedema or raised intracranial pressure. The primary outcome was stipulated as resolution of features of shock, defined as the absence of all of severe tachycardia (heart rate > 160 beats/minute), CRT > 2 seconds or oliguria (urine output < 1 ml/kg/hour) at 8 and 24 hours post treatment. Secondary outcomes included the incidence of adverse events and mortality. Improvements in the W/H z-score or other measures of weight gain were not reported outcomes.

Summary of quality assessment

The methodological rigour of the trial by Akech and colleagues49 was rated moderate overall (Table 10). The trial was potentially at risk of selection bias, because not all of the eligible children who were selected actually participated in the trial either for clinical reasons or because consent was declined. The study was a RCT and an adequate method (use of sealed envelopes) was used for randomisation to treatment groups, resulting in a strong rating for study design. Baseline characteristics and disease severity indices were reported to be balanced across the three fluid intervention arms (although data were not presented for the HAS arm because of small numbers), also leading to a strong rating. However, neither the participants nor the care providers were blinded to treatment and no details were reported regarding the outcome assessors, leading to a higher risk of detection bias and thus a weak rating. For data collection methods, the trial was rated as moderate as it used valid criteria for measuring shock, but it was not possible to judge whether or not these criteria were reliable. There were no dropouts or withdrawals from the trial, only losses because of deaths, and all surviving children completed the study, indicating a low risk of attrition bias. The intervention integrity of the trial was strong as all the participants were deemed likely to have received his or her allocated intervention without any cross-contamination. Appropriate statistical methods were employed in the data analysis and the authors report that all analyses were performed using the intention-to-treat (ITT) principle, although outcomes were presented for all survivors (those who died were not included), rather than for all those randomised. However, the area under curves (AUCs) were calculated in order to compensate for the confounding effect of mortality and, hence, missing observations, leading to biases in the highest risk group and resulting imbalance within the survivors. It should also be noted that the trial was prematurely terminated because of the high overall mortality and inadequate correction of shock in all study arms after an interim review of safety data and consultation with the external safety monitors. As a result, the study did not recruit the required sample size and was therefore underpowered.

TABLE 10. Summary of methodological quality: RCT of children with shock.

TABLE 10

Summary of methodological quality: RCT of children with shock.

Assessment of effectiveness: shock

Mortality

Overall mortality was high, with 51% (31/61) of children not surviving. Of these deaths, 39% (12/31) occurred within 24 hours of recruitment,49 whereas 52% (16/31) of fatalities occurred within 48 hours of enrolment (Professor Kathryn Maitland, Imperial College London, 2011, personal communication). There was no statistically significant difference in mortality rates between the three treatment groups (p = 0.62), nor between children who received RL versus HSD/5D (p = 0.34) (Table 11). On Kaplan–Meier survival analysis, there was no significant difference in time to death when any of the intervention fluids were used for resuscitation (log-rank test combined, p = 0.42).

TABLE 11. Mortality in children with shock.

TABLE 11

Mortality in children with shock.

Mortality rates within a number of subgroups were also reported by Akech and colleagues,49 although not all were presented as comparisons between fluid resuscitation treatment groups. In those with severe diarrhoeal shock, mortality was higher in the standard HSD/5D group than in the RL group {13/19 (68%) vs 9/22 (43%), respectively; p = 0.11 [note: there is a possible error reported in the publication, RL should be 9/21 (43%)]}, although the opposite trend was observed for those with presumptive (non-diarrhoeal) shock [2/7 (29%) vs 4/8 (50%), respectively; p = 0.61 (note: there is a possible error reported in the publication for presumptive shock for HSD/5D)], but neither difference reached statistical significance. Children who fulfilled the WHO malnutrition shock definition at admission were at a statistically significant increased risk of death [risk ratio (RR) 2.0, 95% confidence interval (CI) 0.92 to 4.36; p = 0.05] compared with those who did not fulfil this definition, irrespective of allocated intervention. Similarly, kwashiorkor was associated with an increased risk of death irrespective of treatment arm [odds ratio (OR) 2.2, 95% CI 0.7 to 10.1; p = 0.14], though this was not statistically significant. Mortality in children who were HIV+ve was similar to those that among who were HIV−ve (42% vs 45%, respectively; p-value not reported) and infection with HIV did not significantly increase the risk of death (OR 1.18, 95% CI 0.38 to 3.72; p = 0.76).

Weight gain and anthropometry

Weight gain and anthropometry outcomes were not reported by the Akech and colleagues' trial49 because of the focus of the study (i.e. the trial was designed to look at emergency management of shock rather than nutritional rehabilitation).

Resolution of shock

The proportion of children in whom shock persisted after fluid resuscitation treatment was considerable, but was not significantly different between RL and HSD/5D at either 8 or 24 hours (Table 12). The authors report that a larger decline in the proportion with shock was observed in children who received RL than in those who received HSD/5D, particularly in the diarrhoeal group, but the differences were not significant at any time point (data not shown).

TABLE 12. Persistence of shock in children with shock.

TABLE 12

Persistence of shock in children with shock.

Oliguria

Adequate urinary output was used as a gold standard for successful fluid resuscitation, with oliguria (the production of an abnormally small volume of urine) being a marker of persistent, severe shock. The incidence of oliguria was significantly higher in children receiving the standard WHO HSD/5D solution than in those receiving RL at 8 hours (reported by the authors as p = 0.02 in the table, but p = 0.05 in the text). This trend was also evident at 24 hours, but was no longer statistically significant (p = 0.16) (Table 13).

TABLE 13. Oliguria in children with shock.

TABLE 13

Oliguria in children with shock.

In an additional analysis, the median AUC for the hourly urine output was significantly lower in HSD/5D participants (51 ml/kg/hour, IQR 36–116) than in RL participants (101 ml/kg/hour, IQR 63–141; Kruskal–Wallis chi-squared = 4.6; p = 0.03) (data not shown).

Tachycardia

Persistent tachycardia is an index of unresolved shock and was defined as a heart rate of > 160 beats/minute. Children who received the standard WHO HSD/5D solution had a higher incidence of tachycardia (and hence unresolved shock) compared with those who received the RL solution, becoming statistically significant at 24 hours (p = 0.04) (Table 14).

TABLE 14. Tachycardia in children with shock.

TABLE 14

Tachycardia in children with shock.

In the additional analysis, median AUC of heart rates were similar for both treatments (Kruskal–Wallis chi-squared = 0.3; p = 0.59).

Adverse events

Although the incidence of adverse events was not presented, Akech and colleagues49 did report that no child developed clinical features of pulmonary oedema or allergic reaction (to HAS) during the course of study observation. In addition, no diuretics were required or prescribed during the trial and there were no differences in the mean sodium concentration at admission (133 ± 11 mmol/l vs 134 ± 10 mmol/l, respectively; p = 0.81), 8 hours (134 ± 10 mmol/l vs 139 ± 10 mmol/l, respectively; p = 0.09) or 24 hours (138 ± 9 mmol/l vs 140 ± 9 mmol/l, respectively; p = 0.47) between those who received HSD/5D and RL implying that children did not exhibit the problem of either water or sodium retention.

Other outcomes

Additional outcomes such as severe tachypnoea (rapid breathing of > 60 breaths/minute), creatinine levels and resolution of base deficit (acidosis) were also reported in the trial publication, but have not been presented here. Further details are available in the data extraction forms in Appendix 8.

Summary

  • Only one trial49 was identified that evaluated the efficacy of fluid resuscitation solutions for the treatment of children with SAM and hypovolaemic shock. The trial was relatively small and was rated as having a moderate methodological quality overall. It should be noted that the study was underpowered because of premature termination of the trial because of safety issues (i.e. high overall mortality and inadequate correction of shock in both arms) and the results should therefore be interpreted with caution.
  • The overall mortality rate in the trial was high (> 45%), with no statistically significant differences between treatment groups nor any difference in the time to death between treatment arms. There was an inadequate correction of shock that persisted after fluid resuscitation treatment in both the standard WHO HSD/5D hypotonic solution and the isotonic RL solution groups (> 50%).
  • The incidence of oliguria (used as a marker of persistent, severe shock) was higher in children receiving HSD/5D hypotonic solution than in those receiving RL, being significant at 8 hours, but not at 24 hours. Similarly, children who received the HSD/5D solution had a higher incidence of tachycardia (denoting unresolved shock) than those in the RL group, becoming statistically significant at 24 hours.
  • The isotonic RL solution was found to be as safe as the currently recommended WHO HSD/5D hypotonic solution with no adverse events reported. However, it should be noted that all the fluid solutions were deemed inadequate by the authors in the correction of shock.

What are the best treatments for children with severe acute malnutrition who have diarrhoea? (Q22, rank 1 =)

Eight trials5057 were included that investigated the efficacy of treatments for children with SAM who also had diarrhoea. Within this section, similar trials have been grouped together for ease of comparison between studies. The groupings consist of those with acute diarrhoea and treated with ORS (n = 5,50,51,54,55,57 see Quantity and quality of research available: acute diarrhoea and Assessment of effectiveness: acute diarrhoea) and those with persistent diarrhoea and treated with formula and/or solid diets (n = 3,52,53,56 see Quantity and quality of research available: persistent diarrhoea and Assessment of effectiveness: persistent diarrhoea).

Quantity and quality of research available: acute diarrhoea

Five trials50,51,54,55,57 were included that investigated children with acute diarrhoea, defined as diarrhoea lasting < 2, < 3, < 4 or ≤ 10 days. The key characteristics of these RCTs can be seen in Table 15, with further details of the trials in Appendix 9. All the trials were single-centre RCTs carried out in India51,54,55 or Bangladesh.50,57 One study50 received funding from WHO and one57 was funded jointly by a commercial organisation and an international health research institution. For three studies51,54,55 the primary source of financial support was not stated, although Alam and colleagues51 received funding for materials from a local medical college/university.

TABLE 15. Characteristics of the included studies of children with acute diarrhoea.

TABLE 15

Characteristics of the included studies of children with acute diarrhoea.

Severe acute malnutrition was defined similarly in three trials, being W/H < 70% of the NCHS median,51 and either W/L < 70% of the NCHS median or with bilateral pedal oedema.50,57 Dutta and colleagues54 defined SAM as being < 60% of the Harvard standard W/A without oedema. In the fifth trial, Dutta and colleagues55 included different grades of malnourished children and used the IAP 1972 classification system.45 They did not specifically define SAM.

Two trials51,54 evaluated a hypo-osmolar oral rehydration solution (H-ORS) (containing lower concentrations of sodium, chlorine and glucose), and one trial50 evaluated a modified ORS, termed ReSoMal (containing lower concentrations of sodium, chlorine and citrate, and higher concentrations of potassium and glucose, as well as including other selected minerals). The comparator in these three studies was a standard WHO-ORS. In the fourth trial,55 all participants received a standard WHO-ORS initially with either a zinc-supplemented syrup or a placebo syrup [this study is included in this section rather than in Which methods for correcting micronutrient deficiencies are effective? (Q8, rank 10) because the focus of the study was on treatment of diarrhoea]. The fifth trial57 evaluated three types of ORS, which differed only by the addition of glucose, glucose plus amylase-resistant starch (ARS) or rice powder. In four trials,51,54,55,57 the ORS was given over a period of 4–6 hours, whereas in the ReSoMal trial50 the ORS was given more slowly, over a period of 12–14 hours, with all continuing to receive the ORS thereafter, if necessary, until diarrhoea stopped. None of the studies specifically stated the intended total duration of ORS treatment, although it appeared to be until diarrhoea ceased,50,51,54,55,57 with two studies suggesting that if diarrhoea had not ceased treatment continued for up to 5 days.54,55 The additional treatment with zinc or placebo syrup in the Dutta and colleagues trial55 was continued after discharge until the bottle was finished. None of the studies reported any follow-up beyond the treatment period with the exception of Dutta and colleagues55 who reported outcome data at a follow-up of 30 days.

The trials varied in the other interventions that were offered to participants. Three trials51,55,57 gave i.v. rehydration to participants, where needed, in addition to the ORS. Three trials50,51,57 treated infections with antibiotics, one study54 specifically stated that no drug therapy was given and one55 did not report either way. Most of the studies51,54,55,57 permitted breastfeeding and children were also given solid food where appropriate, with children in the Dutta and colleagues trial54 also given water ad libitum and formula or animal milk. Two trials reported that all children received the standardised treatment for SAM according to either WHO50 or International Centre for Diarrhoeal Disease Research (ICDDR)57 guidelines. It is not clear whether or not this is the case in the other trials, although it is possible that the two H-ORS studies51,54 used the WHO-ORS as their control intervention.

All the trials took place in an inpatient setting, recruiting children from diarrhoea treatment centres50,51,57 or hospital.54,55 The studies were relatively small, ranging from 64 participants in the Dutta and colleagues trial54 to 175 participants in the Alam and colleagues trial.57 Although the Alam and colleagues trial51 included 170 children in total, only 81 of these had SAM, with results reported separately for this group. This trial reported baseline characteristics for the whole study population, with only age and W/A reported in the subgroup with SAM. The five trials included children aged from 3 months to 5 years, although most were toddlers, with the average age being around 1–2 years. In one study,57 around half the participants were boys, in another50 approximately two-thirds were boys, whereas in both trials by Dutta and colleagues54,55 all the included children were boys (for the purposes of ease of collection of urine and stools separately). The last study51 did not report the proportions of males and females.

For three trials,50,51,57 the mean W/A (as a percentage of the NCHS median) at admission ranged from 50% to 59%, for one trial54 about 95% of children were < 60% Harvard standard W/A and for the fifth trial55 around 85% of participants were < 70% Harvard standard W/A. Two studies50,57 reported baseline z-scores, with a mean W/A z-score ranging from −4.3 to −4.7 and a mean W/L z-score ranging from −2.8 to −3.6. The mean duration of diarrhoea before admission was very different in the four trials that reported it, ranging from a mean of around 13 hours57 to 75 hours.50 Dutta and colleagues54 reported a mean of 22 days despite an inclusion criterion of acute diarrhoea for ≤ 72 hours. In three trials, some50,51 or all57 of the children had diarrhoea with cholera.

All five studies had similar inclusion criteria with children required to have SAM, acute, watery diarrhoea for < 48 hours,57 ≤ 72 hours,54,55 < 4 days51 or ≤ 10 days,50 and be within the age range > 3 months and < 5 years. Four trials either required children to have some degree of dehydration51,54,55 or such children were eligible for inclusion.57 Alam and colleagues51 stipulated that children should be included if aged between 3 months and 5 years with non-cholera diarrhoea or if aged > 3 months with a clinical suspicion of cholera. The two trials by Dutta and colleagues54,55 included only males (for the reasons reported above). Children with severe infections were excluded from all five trials. In addition, some trials also excluded those with invasive,51 bloody50,57or a previous episode54 of diarrhoea. Other reasons for exclusion included having chronic underlying disease,55 receipt of i.v. fluids50 or antibiotics,54,55 convulsions,51 being exclusively breastfed54,55 or having signs of kwashiorkor.54

Only two trials specified their primary outcomes. Alam and colleagues57 specified stool output, whereas Alam and colleagues50 specified the proportion of children developing overhydration and with correction of basal hypokalaemia. The other three trials did not specifically report what their primary outcomes were, but the main outcomes presented were similar and included weight gain, duration and volume of diarrhoea, ORS intake and electrolyte concentrations in addition to fluid54,57 or energy intake,51 time to recovery,51,54,55,57 urine output51,57 and requirement for i.v. fluids.51,57 None of the trials reported W/H or W/A z-scores. Further details on all the outcomes reported in the trials can be seen in the data extractions in Appendix 9.

Summary of quality assessment

The methodological quality and the quality of reporting of the five included trials did not vary greatly. Two trials50,51 were rated strong overall, with the other three trials being rated moderate54,55,57 (Table 16).

TABLE 16. Summary of methodological quality: studies of children with acute diarrhoea.

TABLE 16

Summary of methodological quality: studies of children with acute diarrhoea.

Selection bias varied between the studies, with three trials50,54,55 being at potential risk of selection bias. For all of these trials, it was unclear what proportion of selected individuals agreed to participate in the trials before they were randomised. In addition, the included children in both trials by Dutta and colleagues54,55 were considered to be only somewhat likely to be representative of the target population, leading to a higher risk of selection bias. Conversely, the study design of all five trials was strong, with all being RCTs and using an adequate method to generate random allocations. Hence, trial arms within all the studies were balanced with respect to baseline characteristics and confounders, leading to a strong rating. All but one57 trial employed a double-blind method, reporting that the interventions looked identical to participants. Alam and colleagues57 reported that treatments could not be blinded to those involved in the study because of visible differences in the three ORS solutions. Furthermore, neither study by Alam and colleagues51,57 reported sufficient details on the blinding of outcome assessors and they were therefore rated as moderate51 and weak57 as this could lead to detection bias. For data collection methods, all five trials were rated as moderate as they included valid data collection tools, but it was not possible to judge if these tools were reliable.

Sources of attrition bias in clinical trials include losses of participants to follow-up, unequal dropout rates between interventions, selective reporting of outcomes (missing outcomes) and failure to explain why participants are missing (e.g. whether or not they are missing at random). All five trials were rated as strong for withdrawals and dropouts, though they varied in their level of reporting. One trial50 provided both the number and reasons for any losses and had 80–100% of participants completing the study, indicating a low risk of attrition bias. Three trials51,54,57 either did not report any information on dropouts or only reported numbers (without reasons), but had most or all the participants completing the study. Consequently, these were rated as strong as the outcomes can be considered to be reasonably reliable and reflect the study population. In the trial by Dutta and colleagues,55 two contrasting ratings were allocated because all participants completed the acute phase of the study up to the point of recovery (rated strong), but over half the participants were not included in the 30-day follow-up assessments and neither the number nor reasons for the dropouts were reported by the authors (rated weak). The intervention integrity of all five trials was strong, as all the participants were deemed likely to have received their allocated intervention without any cross-contamination. All five trials used appropriate statistical methods in their analysis, although two51,57 did not perform an ITT analysis. For Alam and colleagues,51 this was presumably because the children with SAM were only a subgroup of the total study population. It should be pointed out that all five studies excluded children with severe infections, and as this is not uncommon in hospitalised children with SAM (Professor Kathryn Maitland, Imperial College, London, 2011, personal communication), the results of the studies may not be generalisable to most children with SAM and acute diarrhoea.

Assessment of effectiveness: acute diarrhoea

Mortality

The two studies by Alam and colleagues50,57 were the only trials to report mortality, with no deaths in any treatment group (Table 17). The other trials did not report this outcome, although in the third Alam and colleagues trial51 it is assumed there were no deaths as the children who were not discharged (after having recovered) were accounted for as dropouts. In both trials by Dutta and colleagues,54,55 it remains unclear whether the few children who did not recover within the 5 days of hospitalisation were lost to follow-up or died as no details were reported.

TABLE 17. Mortality in children with acute diarrhoea.

TABLE 17

Mortality in children with acute diarrhoea.

Weight gain

Most of the trials51,54,55,57 reported weight gain as an outcome measure, with two finding significant differences between treatment groups (Table 18). Dutta and colleagues54 found that children receiving the standard WHO/UNICEF-ORS had at discharge gained significantly more weight (p = 0.001) than those receiving the H-ORS (or on day 5 if they did not recover during this period). However, in the Alam and colleagues trial51 weight gain was similar in the H-ORS and WHO-ORS treatment groups. Alam and colleagues57 reported that children receiving the rice-ORS had significantly greater weight gain at 72 hours than those receiving either of the glucose-ORS treatments (p = 0.05). There was no statistically significant benefit on weight gain from a zinc supplement compared with placebo either at the time of recovery or at 30 days follow-up in the trial by Dutta and colleagues.55 Alam and colleagues50 did not provide any numerical data on weight gain, but stated that weight gain before discharge was similar between the groups.

TABLE 18. Weight gain in children with acute diarrhoea.

TABLE 18

Weight gain in children with acute diarrhoea.

Duration of diarrhoea

The length of time that diarrhoea persisted in treated children was reported in four51,54,55,57 of the five trials and can be seen in Table 19. The two trials51,54 evaluating a hypo-osmolar-ORS found similar results. The duration of diarrhoea was statistically significantly shorter in children who received the H-ORS than in those who received the standard WHO/UNICEF-ORS (41.5 vs 66.4 hours, respectively; p = 0.001).54 In the Alam and colleagues trial,51 the duration of diarrhoea was reported separately for a rehydration phase and maintenance phase (as well as overall duration), though the timescale for these phases was not defined. The difference between treatment groups followed the same pattern and was statistically significant during the maintenance phase in favour of H-ORS (95% CI 0.46 to 0.88; p = 0.007), but was no longer significant when the phases were combined as overall duration. Supplementation with zinc was favourable compared with placebo with a mean difference in duration of diarrhoea of approximately 30 hours (p = 0.0001),55 whereas in another study,57 although the median duration of diarrhoea was lower in the rice-ORS group than in the glucose-ORS or glucose-ORS + ARS groups, this did not reach statistical significance.

TABLE 19. Duration of diarrhoea in children with acute diarrhoea.

TABLE 19

Duration of diarrhoea in children with acute diarrhoea.

Frequency of diarrhoea

The frequency of diarrhoea was reported by three trials,51,54,57 although differences in the way this outcome was reported make direct comparisons between trials difficult. Alam and colleagues51 reported the number of stools in a 4-hour period, whereas the other two trials54,57 reported stool output (g/kg and ml/kg, respectively) in several 24-hour periods and also at recovery54 (Table 20). Despite differences in the reporting, for both studies evaluating H-ORS,51,54 the mean frequency of stool output was significantly less in the children receiving H-ORS than in those receiving standard WHO-ORS at all time points. For the third trial, by Alam and colleagues,57 the cumulative mean stool output of children receiving rice-ORS was statistically significantly lower than among children receiving glucose-ORS at 24 hours (32% mean reduction, 95% CI 44% to 174%; p = 0.004), and this statistical difference was maintained at 48 and 72 hours. Compared with the study by Dutta and colleagues,54 data for stool output per kg of body weight were markedly higher in the trial by Alam and colleagues,57 but the reason for this is unclear.

TABLE 20. Frequency of diarrhoea in children with acute diarrhoea.

TABLE 20

Frequency of diarrhoea in children with acute diarrhoea.

Recovery

Two54,55 of the five trials specifically reported recovery (proportion of children who recovered within 5 days) as an outcome (Table 21), with one of these54 also reporting median survival time to recovery. Recovery was defined as the passage of a normal stool or no stool for the last 18 hours,55 or was assumed to be when diarrhoea had ceased (two formed stools passed or no stool for 12 hours).54 A further two trials50,57 reported outcomes that inferred recovery in the children. Alam and colleagues57 reported the time taken to attain an oedema-free W/L of 80% of the NCHS median, whereas Alam and colleagues50 reported the number of children who were adequately rehydrated at 12 hours.

TABLE 21. Recovery in children with acute diarrhoea.

TABLE 21

Recovery in children with acute diarrhoea.

Dutta and colleagues55 found a small but significant (p = 0.04) difference between treatment groups with all children supplemented with zinc recovering within 5 days of hospitalisation, compared with 89% of children receiving placebo. Dutta and colleagues54 also reported a high recovery rate, with all but three children (all in WHO-ORS group) having recovered within 5 days of treatment, but the difference between treatment groups was not significant. However, children treated with H-ORS recovered significantly quicker than those treated with the WHO-ORS (36 vs 53 hours, respectively; p = 0.001).

In the Alam and colleagues57 trial, it took around 7 days for children to attain an oedema-free W/L of 80%, being similar regardless of the type of ORS (p = 0.99).

In the Alam and colleagues trial,50 most of the children in both treatment arms were adequately rehydrated at 12 hours, with no statistically significant differences between groups.

Consumption of oral rehydration solution

Most of the trials51,54,55,57 measured how much ORS was consumed by the children, either as the total amount consumed (litres)51,55 or as ml/kg of body weight54,57 (Table 22). In two trials,51,54 children receiving H-ORS needed to consume less rehydration solution than those receiving the standard WHO-ORS, although this reached statistical significance in only one of the trials (p = 0.0001).54 The other two studies also found significant differences in favour of the intervention groups. Dutta and colleagues55 reported a lower ORS consumption in children supplemented with zinc than in those supplemented with placebo (p = 0.0001). Alam and colleagues57 found that children receiving rice-ORS had a significantly lower ORS intake at 18 hours compared with those receiving glucose-ORS (see Appendix 9). This difference was maintained at each 6-hourly interval thereafter until 72 hours, when there was a 38% reduction in intake (p = 0.012).

TABLE 22. Consumption of ORS in children with acute diarrhoea.

TABLE 22

Consumption of ORS in children with acute diarrhoea.

Adverse effects

Adverse effects were not reported in any detail by the included studies. Two trials51,55 did not report any safety issues, whereas two trials54,57 reported that no children developed symptoms of overhydration. Alam and colleagues50 report that prevention of overhydration is the primary theoretical advantage of ReSoMal. Overhydration was defined as a weight gain > 5% after correction of dehydration at any time during the study period with any of the following signs: periorbital oedema/puffy face, increased heart rate (> 160/minute) or increased respiration (> 60/minute). Although there appeared to be a lower occurrence of over-rehydration in those children who received ReSoMal than in those receiving WHO-ORS, numbers were small and this was not supported statistically (Table 23). Alam and colleagues50 also looked in detail at serum electrolytes and, thus, the incidence of hypo- and hyperkalaemia and hypo- and hypernatraemia (these outcomes have not been reported here as they are not main outcomes of interest to this review, but data are available in Appendix 9). However, it is worth noting that three children in the ReSoMal group developed severe hyponatraemia (low serum sodium) by 24 hours, with one child having a resulting convulsion, which the authors highlight as a safety concern that may limit the use of ReSoMal in its current formulation.

TABLE 23. Adverse effects in children with acute diarrhoea.

TABLE 23

Adverse effects in children with acute diarrhoea.

Other outcomes

Additional outcomes, such as caloric or fluid (water, milk) intake, other fluid losses (e.g. urine, vomit) and correction of hypokalaemia, were also reported by some studies, but have not been presented here. Further details are available in the data extraction forms in Appendix 9.

Summary

  • Five trials evaluated the treatment of children with acute diarrhoea with various types of ORS, including a H-ORS,51,54 a modified WHO-ORS (ReSoMal),50 an ORS containing either glucose, glucose plus ARS or rice powder,57 and supplementation with zinc.55 The trials were all of strong or moderate methodological quality.
  • There were no deaths in the two trials that reported mortality,50,57 and it is assumed that there were no deaths in a third trial,51 as all children who did not recover were accounted for as dropouts.
  • Compared with the standard WHO-ORS, children receiving the H-ORS had a significantly shorter duration and lower frequency of diarrhoea, consumed less ORS and had a quicker time to recovery (one trial54).
  • There appeared to be no benefit from H-ORS with respect to weight gain compared with WHO-ORS.
  • Supplementation with 40 mg elemental zinc (as zinc syrup) in addition to a standard WHO-ORS resulted in a significantly shorter duration of diarrhoea, a better recovery rate and a lower ORS intake, but no difference compared with placebo in terms of weight gain.55
  • Rice-ORS appeared to be more favourable than glucose-ORS in treating children with cholera diarrhoea. The rice-ORS groups had significantly better weight gain, a lower frequency and duration of diarrhoea and consumed less ORS.57
  • ReSoMal did not appear to show any advantage over a standard WHO-ORS in rehydrating severely malnourished children with acute diarrhoea, although it was beneficial in correcting potassium depletion.
  • Adverse effects were not generally reported by the trials, although ReSoMal50 may result in symptomatic severe hyponatraemia and seizures in some patients.

Quantity and quality of research available: persistent diarrhoea

Three included trials, reported in four publications,52,53,56,58 pertained to children with persistent diarrhoea. Persistent diarrhoea was defined as diarrhoea lasting ≥ 14 days (Table 24).

TABLE 24. Characteristics of the included studies of children with persistent diarrhoea.

TABLE 24

Characteristics of the included studies of children with persistent diarrhoea.

All three trials were single-centre RCTs, of which one was a single-blind52,58 and one a double-blind trial.56 The third trial provided no details about blinding.53 The trials were set in Mexico,56 Pakistan53 and Zambia,52,58 and all received external funding. One trial was funded by a grant from a commercial organisation, with one of the authors also receiving support from a global charity.52,58 Of the two remaining trials, one was part-funded56 and one fully-funded53 by a US academic institution, by means of a cooperative agreement with a US Government department. The part-funded trial received a further grant from another US Government department.56

All three trials evaluated varying diets, including soy as either an intervention53,56 or as a comparator (milk followed by a soy porridge).52,58 Bhutta and colleagues53 evaluated a full-strength soy diet against a half-strength buffalo milk diet with khitchri (rice-lentils) and yoghurt (KY), with diets given in gradually increasing amounts over 14 days. The trial by Nurko and colleagues56 compared three intervention strategies – a local chicken-based diet, a soy-based (Nursoy®; Wyeth Laboratories, Philadelphia, PA, USA) diet and an elemental diet (Vivonex® standard; Norwich Eaton Ltd, Surrey, UK) – all provided at gradually increasing concentrations by nasogastric tube for around 16 days if the diet was tolerated. The third trial by Amadi and colleagues,52,58 compared an amino acid-based infant formula (Neocate®, SHS International Ltd, Liverpool, UK), without cow's milk, soy and cereal antigen, with a standard skimmed milk diet (followed by soy-based porridge from week 2) for 4 weeks. One of the trials followed the WHO guidelines for the treatment of persistent diarrhoea and malnutrition52,58 and one the WHO/UNICEF guidelines for hydration (standard glucose–electrolyte i.v. solution).56 In addition, two of the trials provided antibiotic treatment as needed52,56,58 and one provided micronutrient supplements.52,58

All trials took place in the hospital inpatient setting. Sample sizes were small for two of the trials, 5153 and 5656 children, whereas the third RCT included 200 children.52,58 The age of children included ranged from 3 to 36 months. Two of the trials had fairly similar ratios of boys and girls in their trial arms,52,56,58 whereas the remaining trial consisted of boys only (to facilitate separate quantitative collections of urine and faeces).53

Definitions for SAM varied, with Amadi and colleagues52,58 using the Wellcome classification for severe malnutrition (W/A and H/A). The remaining two trials used the NCHS growth reference, with W/A ≤ 80th centile of the median NCHS standard (i.e. Gómez grades II and III malnutrition) described as severe PEM,53 and W/A < 60% of the NCHS 50th percentile for W/A described as third-degree malnutrition of the marasmic type by the Gómez criteria.56 One trial reported W/Ls at baseline.53 W/A z-scores were similar across the three trials, ranging from −3.953,56 to −4.41.53 All three trials excluded exclusively breastfed children.52,53,56,58 Other exclusion criteria were chronic illnesses,56 neurological or serious systemic disorders52,58 and children with kwashiorkor and the presence of intercurrent infections.53 The children in the Amadi and colleagues52,58 trial had a high prevalence of intestinal infection, and around half were HIV+ve. In the trial by Nurko and colleagues,56 64% of the sample had associated conditions (e.g. pneumonia, sepsis, infections) on admission.

There were large differences in the baseline duration of diarrhoea between the trials, reported as around 36.6–48.7 days in one trial,56 an average of 75–150 days in another trial,53 but as ≥ 14 days in the remaining trial.52,58

Trials assessed outcomes of weight gain and some measures of diarrhoea, but only one trial specified these as primary outcomes in addition to mortality.52,58 Other outcomes included treatment success/failure, nutritional recovery and nitrogen balance,56 as well as developmental milestones achieved, activity and play, and laboratory indicators of severity of illness.52,58 For further details on reported outcome measures see Appendix 9.

Summary of quality assessment

Two of the included trials were rated overall as ‘weak’ for their methodological quality and quality of reporting (Table 25),52,53,58 with the third being rated overall as ‘strong’.56

TABLE 25. Summary of methodological quality: studies of children with persistent diarrhoea.

TABLE 25

Summary of methodological quality: studies of children with persistent diarrhoea.

Trials were rated as moderate,52,58 weak53 or strong56 for selection bias. A moderate rating indicates that the selected individuals are at least somewhat likely to be representative of the target population and at least 60% of selected individuals participated in the trial. A weak rating indicates that participants may not be representative of the target population, or that the selection method and/or levels of participation were not described. The two trials with moderate and weak ratings were at potential risk of selection bias.52,53,58 All three trials were rated as strong for their study design (RCTs).

There were no important differences in baseline characteristics between the trial arms, and without potentially confounding variables all three trials were rated as strong. For blinding, only one trial employed a double-blind method and was therefore rated as strong.56 Of the other two trials, both were rated as weak, with one employing a single-blind method52,58 and the other reporting no details.53 It is recognised that blinding of children is not always possible because of the nature of the intervention. This could lead to bias in either the care provided (performance bias) or how the outcomes were assessed (measurement or detection bias), or both. Not blinding children/parents to the research question could lead to reporting bias. Although it may be problematic in some circumstances to blind children/parents to the intervention, the potential bias it can introduce needs to be kept in mind when interpreting the results.

For data collection methods, two trials were rated as moderate.53,56 Although both trials included valid data collection tools, it was not possible to judge if these tools were reliably employed. The remaining trial was rated as weak, as it was not possible to assess if the data collection tools were either valid or shown to be reliable.52,58 One trial52,58 provided both numbers and reasons for withdrawals and dropouts, and with a follow-up rate of ≥ 80% received a strong rating. Of the remaining two trials,53,56 both had lower follow-up rates (60–79%) and one provided inadequate information by giving reasons for withdrawal, but not numbers for each group.53 There was a possible risk of attrition bias in both these trials and they both received an overall rating of moderate for withdrawals and dropouts. For the section of the tool capturing intervention integrity, two trials52,56,58 reported that > 80% of the participants received the intervention, and in the third53 60–79% received the intervention. The consistency of the intervention was measured by all three trials, using weight gain as the measure, and there appeared to be no contamination of the interventions (i.e. all children received the allocated diet only). All trials used patients as the unit of allocation and analysis for statistical analysis of the results, and were judged to use appropriate methods of statistical analysis for the research question. Two of the trials did not perform an ITT analysis,52,56,58 and it was not possible to determine how missing data were dealt with in the analysis in the third trial.53

Assessment of effectiveness: persistent diarrhoea

Mortality

Only Amadi and colleagues52,58 reported mortality as an outcome (Table 26). Although mortality was highest in the Neocate group (22/100), the difference was not statistically significant (see Table 26). The highest number of deaths for the combined treatment groups (43%) occurred in the second treatment week (week 1 = 31%, week 3 = 26% and week 4 = 10%). Irrespective of treatment arm, death was more likely to occur in children with marasmic kwashiorkor (34.9%; p = 0.004), or cryptosporidiosis (no data reported) and in children identified as HIV+ve (24% compared with 11% of HIV−ve children; p = 0.04). Although mortality was not formally identified as an outcome in the trial by Nurko and colleagues,56 the authors reported that five children died during the trial and how many deaths occurred in each group (see Table 26). However, the causes of death, intestinal pneumatosis (n = 2), central line-associated sepsis (n = 2) or bacterial sepsis (n = 1), were reported only for the whole trial population and not by group. Bhutta and colleagues53 also did not specify mortality as an outcome; however, it can be assumed that there were no deaths, because all children either completed the treatment or were accounted for as dropouts.

TABLE 26. Mortality in children with persistent diarrhoea.

TABLE 26

Mortality in children with persistent diarrhoea.

Weight gain

Measures of weight gain were employed by all three trials; however, only one trial reported weight gain relative to initial weight (the benefit of relative weight gain measures is that any effects because of starting differences in body weight are removed). In the trial by Amadi and colleagues,52,58 feeding with Neocate was associated with a 41% better gain in weight from nadir compared with the skimmed milk/soy-based diet, and the difference was statistically significant (p = 0.002) (Table 27). In the trial by Bhutta and colleagues,53 weight gain was higher for the intervention diet of soy than for KY milk, but reached statistical significance only at the end of treatment (i.e. week 2; p < 0.02). Conversely, mean daily weight gain was higher in the KY milk group, but the difference between groups was not statistically significant. It should be noted that there was also a reported weight loss in two children in the soy group (10%) and seven (37%) in the KY milk group (p = not statistically significant). Nurko and colleagues56 reported statistically significant weight gains for all three diets used in their trial for their comparison of weight change from admission versus at the end of the protocol and from admission versus discharge. However, no statistically significant differences between the three treatment arms were reported.

TABLE 27. Weight gain in children with persistent diarrhoea.

TABLE 27

Weight gain in children with persistent diarrhoea.

Anthropometry

Two of the studies reported anthropometry outcomes as well as weight gain. Amadi and colleagues52,58 reported that increases in z-scores of W/A and W/H were statistically significantly higher from admission (W/A, p = 0.018; W/H, p < 0.001) and from nadir (W/A, p = 0.002; W/H, p < 0.001) for the Neocate group, with results mirrored in HIV+ve (W/A, p = 0.007; W/H, p < 0.001) and HIV−ve (W/A, p = 0.01; W/H, p = 0.009) subgroups (Table 28). In the trial by Bhutta and colleagues,53 increases in W/A z-score during the study were significantly greater in the soy group (p < 0.001) than in the KY milk group (p = not statistically significant), but no statistical comparison between the groups was reported. Bhutta and colleagues53 did report a statistical comparison between the groups for improvement in mid-arm circumference, which was significantly higher in the soy intervention group (1.0 cm vs 0.1 cm; p < 0.001).

TABLE 28. Anthropometric outcomes in children with persistent diarrhoea.

TABLE 28

Anthropometric outcomes in children with persistent diarrhoea.

Diarrhoea

Diarrhoea output was quantified either by collecting urine separately from stools (using adhesive urine bags and pre-weighed nappies/diapers)53 or by the use of metabolic beds/cots for separation of stool from urine.56 There were no statistically significant differences in any of the measures of diarrhoea between treatment arms in the two trials53,56 that reported these outcomes (Table 29). In addition, Amadi and colleagues,52,58 who presented no numerical data, stated that there were no differences in either stool number or frequency.

TABLE 29. Diarrhoea volume and frequency in children with persistent diarrhoea.

TABLE 29

Diarrhoea volume and frequency in children with persistent diarrhoea.

Oral rehydration solution intake

Only Bhutta and colleagues53 reported ORS intake, which was significantly reduced by week 2 in the soy intervention arm compared with the KY milk diet (p < 0.05); however, differences in time to recovery were not statistically significant between the two diets (Table 30).

TABLE 30. Oral rehydration solution intake in children with persistent diarrhoea.

TABLE 30

Oral rehydration solution intake in children with persistent diarrhoea.

Calorie intake

Surprisingly, Amadi and colleagues52,58 reported that intake of calories (per kg per day) as liquid feeds, was statistically significantly higher at all time points for the control group (p < 0.0001). However, it should be noted that in addition to the liquid feed based on skimmed milk, the control group also received soy-based porridge from the beginning of the second week. In contrast, Bhutta and colleagues53 found caloric intake (per kg per day) to be only significantly higher for the soy-based intervention arm than for the KY milk arm at the end of week 1 (p < 0.02), and although this remained higher, it was no longer statistically significant at the end of week 2. Caloric intake in the trial by Nurko and colleagues56 was similar in all three diet groups (Table 31).

TABLE 31. Caloric intake in children with persistent diarrhoea.

TABLE 31

Caloric intake in children with persistent diarrhoea.

Treatment success/failure

Although clinical failure appeared to be lower in the soy-based diet arm than in the KY milk arm (no p-value reported), Bhutta and colleagues53 reported no statistical difference between treatment arms. Nurko and colleagues56 reported that there was no statistically significant differences between the three diets in terms of successful outcome (Table 32), with nutritional recovery and treatment failures appearing similar between the groups (p-values not reported). However, across the whole trial population (i.e. analysis not per treatment group), significant differences between treatment success and failure (p < 0.05) were associated with albumin and sodium concentration at admission, as well as the incidence of associated infections. Treatment failures were associated with formula intolerance (Table 33). Of the 15 treatment failures that occurred (see Table 32), 10 were successfully managed. One of the failures in the Nursoy group was because of allergy to the formula. The other five children who failed treatment died (see Table 26).

TABLE 32. Treatment success/failure in children with persistent diarrhoea.

TABLE 32

Treatment success/failure in children with persistent diarrhoea.

TABLE 33. Safety outcomes in children with persistent diarrhoea.

TABLE 33

Safety outcomes in children with persistent diarrhoea.

Safety outcomes

Only the study by Nurko and colleagues56 reported on safety, but there was no statistically significant difference in formula intolerance between the treatment arms (see Table 33). Of those children with formula intolerance, 15 were treatment failures (see Table 32) and four had intestinal pneumatosis (two of those with intestinal pneumatosis died; see Table 26).

Additional outcomes

Additional reported outcomes, such as protein ingested after full diet tolerance or time from diet start to failure, were reported in some studies, but have not been presented here. Further details can be seen in the data extraction forms in Appendix 9.

Summary

  • Three trials52,53,56,58 evaluated the treatment of children with persistent diarrhoea, with each trial comparing different diets. The overall methodological quality was rated as weak for two trials52,53,58 and strong for one trial.56
  • Although all three trials employed a hospital inpatient setting, making diet intake easier to control and regulate, all three trials were judged to be open to a potential risk of bias in a number of areas and results should therefore be treated with caution.
  • There were no significant differences in mortality rates between the diets employed in the two trials reporting mortality.52,56,58
  • None of the diets in the three included trials52,53,56,58 pertaining to children with persistent diarrhoea produced statistically significant improvements in measures of diarrhoea.
  • The majority of diets appeared to be effective in increasing weight, with two out of three trials reporting better results for the diet used in the intervention arm. In the trial by Amadi and colleagues,52,58 Neocate produced greater weight gain over a 4-week period than the standard skimmed milk/soy-based diet, which was reflected by increases in W/A and W/H z-scores, as well as weight increases in both HIV+ve and HIV−ve subgroups. The full-strength soy diet in the trial by Bhutta and colleagues53 also produced better weight gain over a 2-week period than the half-strength buffalo milk diet with rice-lentils and yoghurt given to the control group. This was again reflected by increases in W/A z-scores. In contrast, the three diets of chicken, Nursoy and Vivonex (control) employed by Nurko and colleagues56 were found to be equally effective for weight gain.

What methods are effective in treating infection? (Q7, rank 5 =)

The overarching question for this section included within it broader issues regarding antibiotic therapy (examples of these are available in Appendix 5). No study addressed the overarching question directly, but two studies59,60 were included that investigated different aspects of antibiotic therapy in children with severe malnutrition. As these addressed different questions they are presented in separate sections. Dubray and colleagues59 studied the relative effectiveness of two broad-spectrum antibiotics prescribed systematically to all participants (regardless of confirmation of infection) (see Quantity and quality of research available: different antibiotics in the inpatient setting), whereas Trehan and colleagues60 sought to determine whether or not including amoxicillin in the home-based treatment of uncomplicated severe malnutrition with RUTF led to better recovery rates than treatment with RUTF alone (see Quantity and quality of research available: antibiotic use in the outpatient setting).

Quantity and quality of research available: different antibiotics in the inpatient setting

This question was addressed by one RCT59 that met the inclusion criteria for this review. The key characteristics of this RCT are presented in Table 34, and the full data extraction form in Appendix 10 provides further details.

TABLE 34. Characteristics of the included RCT of different antibiotics in the inpatient setting.

TABLE 34

Characteristics of the included RCT of different antibiotics in the inpatient setting.

Dubray and colleagues59 conducted a randomised, unblinded superiority controlled trial to compare two antibiotic regimens in a therapeutic feeding centre (TFC) in Sudan. This was a single-centre trial funded by an international humanitarian medical aid organisation. Systemic broad-spectrum antibiotic therapy was provided on admission to all participants (with or without any signs of clinical infection), with the aim of improving the outcomes of SAM (reduce mortality and improve nutritional response to feeding). Four hundred and sixty children with SAM were randomly allocated to either ceftriaxone (the intervention, n = 230) or amoxicillin (the comparator, n = 230). Children were eligible to participate if they presented with a W/H < 70% of the reference median [NCHS/Center for Disease Control (CDC) 1977 growth reference curves18] and/or bilateral oedema and/or MUAC < 110 mm. In addition, eligible children had to weigh at least 5 kg and have a height within the range of > 65 cm to ≤ 109.9 cm. Children whose parents refused permission to participate were excluded from the study, as were children who had undertaken treatment with any of the study drugs or had been admitted to any health facility for SAM in the 7 days before admission, children with known hypersensitivity to amoxicillin or ceftriaxone, children whom the physician decided to treat using a different antimicrobial drug on admission and children with acute otitis media (AOM) or severe complications diagnosed on admission.

All participants received the same nutritional rehabilitation and care (further details in Appendix 10). The intervention group received a once-daily intramuscular (i.m.) injection of ceftriaxone at a dose of 75 mg/kg/day for 2 days, whereas the comparator group was given oral amoxicillin (80 mg/kg/day) twice daily over 5 days. When necessary, a second antibiotic (ceftriaxone, chloramphenicol, cotrimoxazole, amoxicillin or metronidazole) was administered as per the TFC protocol.

Dubray and colleagues59 reported that baseline characteristics did not differ significantly between groups. The mean age of the participants was approximately 17 months and just over half of the trial participants were male. More than 70% of the participants had W/H < 70% of the median, 15% had a MUAC measurement of < 110 mm and at least 10% had bilateral oedema. Though there was no diagnostic confirmation of infection, approximately 30% of the participants had fever (≥ 37.5 °C), 1–2% tested positive for malaria, more than 17% presented with an abnormal respiratory rate and at least 10% were moderately dehydrated.

The reported primary outcome was the proportion of children with a weight gain increase of at least 10 g/kg/day calculated over a 14-day period starting on the first day of weight gain after admission. Additionally, the authors considered secondary outcomes such as the recovery rate of discharged children, overall case fatality ratios (CFRs), defaulter rate, referral (to another medical facility) rate and the occurrence of adverse events.

Summary of quality assessment

Although 230 participants were randomly allocated to the ceftriaxone group, two of these were secondarily excluded; thus, only 228 were included in the analyses. The authors state that an ITT analysis was conducted, given that all children who had received at least one dose of the study drug were included. However, because of the post-randomisation exclusion of two participants, this was judged not to be a full ITT analysis during quality assessment.

Dubray and colleagues' study59 was rated moderate in terms of its overall methodological quality, as shown in Table 35. More than 80% of the selected individuals, who are very likely to be representative of the target population, participated in the RCT. The use of a computer-generated block randomisation method and sealed envelopes for allocation was appropriate. Additionally, there were no important baseline differences between groups and the number and reasons for withdrawals and dropouts were reported per group. Therefore, this study was considered strong regarding the selection bias, study design, confounders and the withdrawals and dropouts components of quality assessment. Despite using valid data collection tools, the reliability of the tools is not reported, and, hence, the study strength on data collection methods was rated moderate. For the blinding component, the study was judged to have weak methodological strength because neither outcome assessors nor participants were blinded. Considering that the consistency of the intervention was measured, that 60–79% of the participants received the allocated intervention and that they are not likely to have received an unintended intervention, the intervention integrity is considered to have been ensured. Furthermore, the analysis performed was found to be appropriate for the study design, despite the shortcomings of the ITT analysis.

TABLE 35. Summary of methodological quality: RCT of different antibiotics in the inpatient setting.

TABLE 35

Summary of methodological quality: RCT of different antibiotics in the inpatient setting.

Assessment of effectiveness: different antibiotics in the inpatient setting

Mortality

Dubray and colleagues59 reported several mortality-related secondary outcomes, based on an analysis that excluded two participants who had been randomised, but who did not receive any treatment. As can be seen in Table 36, fewer deaths occurred in the ceftriaxone group, not only within 14 days after admission but also during the whole follow-up period to discharge from the TFC. However, the difference in total deaths during follow-up was not statistically significant (p = 0.62) and no p-value was reported for the former. The 13 deaths that occurred during the first 14 days were because of septic shock (n = 5), lower respiratory tract infections (n = 3), fluid overload (n = 4) and severe dehydration (n = 1).

TABLE 36. Mortality in children receiving different antibiotics in the inpatient setting.

TABLE 36

Mortality in children receiving different antibiotics in the inpatient setting.

Weight gain

Table 37 presents the primary outcomes on weight gain from the Dubray and colleagues59 study, which were success rate and mean overall weight gain, as well as a secondary outcome of weight gain at exit from TFC. The reported success rate is defined as a weight gain ≥ 10 g/kg/day by day 14 or discharge before 14 days of weight gain because the TFC exit criteria were met (maintained W/H ≥ 85% for 7 consecutive days). Mean overall weight gain was calculated 14 days after the first weight gain. The groups showed similar results and no statistically significant differences were found between groups for any of the outcomes.

TABLE 37. Weight gain in children receiving different antibiotics in the inpatient setting.

TABLE 37

Weight gain in children receiving different antibiotics in the inpatient setting.

Length of stay and reasons for exit from the therapeutic feeding centre

As shown in Table 38, the authors reported a slightly shorter length of stay for the ceftriaxone group, but the difference from the amoxicillin control group was not statistically significant. No statistically significant differences were found on the reasons for exit either.

TABLE 38. Length of stay and reasons for exit from the TFC in children receiving different antibiotics in the inpatient setting.

TABLE 38

Length of stay and reasons for exit from the TFC in children receiving different antibiotics in the inpatient setting.

Infection-related deaths and adverse events

Dubray and colleagues59 reported the number of infection-related deaths per type of infection and adverse effects attributed to antibiotics (Table 39). A statistically significant lower rate of adverse events was found in the ceftriaxone group (p = 0.05).

TABLE 39. Infection-related deaths and adverse events in children receiving different antibiotics in the inpatient setting.

TABLE 39

Infection-related deaths and adverse events in children receiving different antibiotics in the inpatient setting.

Summary

  • One RCT59 that compared ceftriaxone (i.m.) with oral amoxicillin met the inclusion criteria of the review for this question. The RCT's methodological quality was summarised as moderate, mainly owing to the fact that blinding of the outcome assessors or the participants was not carried out.
  • Mortality was a secondary outcome of the RCT. Dubray and colleagues59 did not find a statistically significant difference in mortality between the ceftriaxone and amoxicillin groups. Similarly, no statistically significant differences in the number of recovered patients, weight gain, length of stay or reasons for exit from the TFC were found either.
  • A statistically significant lower rate of adverse events was found in participants receiving ceftriaxone (p = 0.05) than in those receiving amoxicillin.
  • No data on resolution of existing infections, development of new infections, relapse or development of antibiotic resistance outcomes were reported.
  • The criteria used to define SAM were broadly in line with current WHO criteria, hence, results are likely to be generalisable to the SAM populations identified by WHO criteria. However, the generalisability to settings where HIV prevalence is high and where children may be receiving long-term cotrimoxazole prophylaxis is uncertain.
  • More than 25% of children in each group received a second antimicrobial treatment (ceftriaxone, chloramphenicol, cotrimoxazole, amoxicillin or metronidazole, in accordance with TFC treatment protocols), which may have reduced the evidence of difference between groups.
  • The study site was chosen because the working conditions were satisfactory, the centre adhered to international standards of nutritional rehabilitation programmes and the political situation was stable. Centres with poorer operational conditions might not be able to reach the same level of care, which might adversely affect outcomes.

Quantity and quality of research available: antibiotic use in the outpatient setting

The key characteristics of the single retrospective cohort study investigating this question are presented in Table 40, and the full data extraction form in Appendix 10 provides further details.

TABLE 40. Characteristics of the included cohort study of antibiotic use in the outpatient setting.

TABLE 40

Characteristics of the included cohort study of antibiotic use in the outpatient setting.

Trehan and colleagues60 conducted a retrospective analysis of outcomes from two cohorts of children in Malawi to determine whether or not including amoxicillin in the home-based treatment of uncomplicated SAM with RUTF led to better recovery rates than treatment with RUTF alone. The study was funded by a US government department. The data were obtained for the same time period from two different feeding projects, one operating in one district of Malawi, the other operating in two other districts of Malawi (the number of feeding centres in each district was not reported). Data from anonymised records of 2453 children who had qualified for outpatient treatment of SAM were included: 1955 children in one cohort had received RUTF alone and 498 children in the second cohort received amoxicillin in addition to RUTF. SAM was defined as W/H z-score ≤ −3 and/or the presence of bilateral pitting oedema. To be eligible for outpatient treatment, children in both cohorts needed to have uncomplicated SAM and a good appetite. Children with poor appetite, altered mental status, compromised perfusion or respiratory distress or who were being transferred from inpatient to outpatient therapy were excluded.

The intervention cohort received a 7-day supply of amoxicillin, equivalent to approximately 60 mg/kg/day, and RUTF to provide 175 kcal/kg/day. Children in the comparison cohort (who met the same criteria for outpatient treatment described above), received the same RUTF provision, but did not receive any antibiotics. In both cohorts RUTF was given until children reached a W/H z-score ≥ −2 with no peripheral oedema for a minimum of 4 weeks and a maximum of 12 weeks. Caregivers of the children in both cohorts were educated about the child's illness and instructed on optimal feeding practices. They were also referred to local health providers with any concerns about other acute illnesses.

The primary outcome was the nutritional recovery rate, with recovery defined as W/H z-score ≥ −2 and no peripheral oedema. Secondary outcomes were survival, W/H z-scores, W/A z-scores, H/A z-scores and presence of oedema.

Summary of quality assessment

Trehan and colleagues' retrospective analysis of two cohorts60 was rated moderate in terms of its overall methodological quality (Table 41). The study was judged to be at a low risk of selection bias (rated strong for selection bias), but because this was a cohort with control study a moderate rating was applied for study design. Although there were some important differences between the cohorts prior to the intervention, the study authors indicated that these were taken into account in the analysis, which enabled the confounders section of the quality assessment to be judged strong. Although study participants were not aware of the research question, the outcome assessors knew what treatment participants had received, which led to a moderate rating for the blinding section. The data collection methods were rated weak because tools were not reported to be either valid or reliable. The final section contributing to the global rating, withdrawals and dropouts, was rated strong because 80–100% of participants completed the study, so the risk of bias owing to missing data was considered to be low. The intervention integrity is difficult to determine because the consistency of the intervention did not appear to have been measured, and it was not possible to tell whether or not any unintended intervention could have occurred in either cohort. There was also some uncertainty regarding the analysis of the data. The study was powered to detect a difference of at least 5% in the recovery rate.

TABLE 41. Summary of methodological quality: cohort study of antibiotic use in the outpatient setting.

TABLE 41

Summary of methodological quality: cohort study of antibiotic use in the outpatient setting.

Assessment of effectiveness: antibiotic use in the outpatient setting

Mortality

Mortality was one of the secondary outcomes of the Trehan and colleagues study.60 The total number of deaths was reported at 4 weeks and at 12 weeks for the overall number of participants in each cohort, but also reported separately for those with and without oedema at baseline (Table 42). The rates of death at both time points were described as similar for each group.

TABLE 42. Mortality in the cohort study of antibiotic use in the outpatient setting.

TABLE 42

Mortality in the cohort study of antibiotic use in the outpatient setting.

Recovery

Recovery was defined as a W/H z-score ≥ −2 and no peripheral oedema (Table 43). Those who remained alive but did not meet the criteria for recovery were classed as remaining malnourished, and those who missed two follow-up visits were categorised as defaulters. At the 4-week follow-up, a greater proportion of children in the RUTF-only cohort had recovered in comparison with the cohort receiving amoxicillin and RUTF (70.8% vs 39.8%; no p-value reported). A statistically significant difference (p < 0.001) in favour of the RUTF-only cohort was reported for the subgroups of children with and without oedema at baseline. In the subgroup of children who recovered after 4 weeks, the W/H z-score was significantly higher in the RUTF cohort than in the RUTF plus amoxicillin cohort (−0.37 vs −0.75; p < 0.0001).

TABLE 43. Recovery in the cohort study of antibiotic use in the outpatient setting.

TABLE 43

Recovery in the cohort study of antibiotic use in the outpatient setting.

At the 12-week follow-up, the overall proportion who had recovered in each cohort was described as similar (no p-value reported). Rates of defaulting were described as similar in the two cohorts at 4 and 12 weeks (no p-values reported). Therefore, the proportions of children classed as remaining malnourished were as expected, with a greater proportion remaining malnourished at the 4-week follow-up in the intervention cohort receiving amoxicillin, but more similar proportions from each cohort were malnourished by the 12-week follow-up (no p-values reported for the between group comparison at either time point).

Other outcomes

A regression analysis was conducted to compare recovery rates, while controlling for differences in baseline characteristics between the cohorts. The model based on outcomes at 4 weeks showed that age (older children) and W/H z-score (higher W/H z-score) at baseline were predictive of recovery (p < 0.001 for both), whereas receipt of amoxicillin was correlated with failure to recover at 4 weeks (OR 0.22; p < 0.001). However, the 12-week follow-up regression analysis demonstrated that none of the baseline factors considered were predictive of recovery. The W/A z-score, H/A z-score and presence of oedema were not correlated with recovery in either the 4- or 12-week analysis. Full results are available in Appendix 10.

Summary

  • One cohort study60 of moderate methodological quality compared amoxicillin plus RUTF with RUTF only for the treatment of uncomplicated SAM in children.
  • Mortality rates were < 5% and similar in both cohorts.
  • The primary outcome of the study was recovery rate (W/H z-score ≥ −2 and no peripheral oedema), which appeared substantially greater at 4 weeks in the cohort of children who did not receive amoxicillin. However, by 12 weeks the proportion of children in each cohort who had recovered was described as similar.
  • The provision of a 7-day course of amoxicillin did not improve recovery rates from uncomplicated SAM in children in Malawi in this cohort when compared with the outcomes of a cohort who did not receive amoxicillin.

What is the clinical effectiveness of interventions in different settings (e.g. hospital, community, emergency)? (Q14, rank 9)

Quantity and quality of research available: settings

Four trials6265 that investigated the clinical effectiveness of interventions in different settings were included. All the studies were CCTs. The key characteristics of these CCTs are presented in Table 44 and Appendix 11 provides further details. These trials were conducted in Niger62 (100 participants), Malawi63 (1178 participants overall, 645 as subgroup with SAM), Jamaica64 (81 participants) and Bangladesh65 (573 participants). Two trials were single-centre64,65 and two were multicentre trials.62,63 Heikens and colleagues64 received funding from the government of the Netherlands, Chapko and colleagues62 were partially funded by a US governmental education fellowship, Khanum and colleagues65 were supported by a UK charity and the UK government and Ciliberto and colleagues63 received funding from the United Nations, a US charity and hospital foundation, a UK humanitarian organisation and the US government.

TABLE 44. Characteristics of the included studies of interventions in different settings.

TABLE 44

Characteristics of the included studies of interventions in different settings.

Chapko and colleagues62 compared inpatient with daily ambulatory rehabilitation, and two of the trials63,64 investigated hospital- and home-based rehabilitation (differing, however, in the level of support provided). These three trials6264 evaluated alternative settings for the rehabilitation phase of treatment for malnourished children, after an initial phase of hospital care common to both treatment arms. In contrast, the fourth trial, by Khanum and colleagues,65 had three trial arms to compare inpatient care with daily ambulatory care for both the initial and the rehabilitation phases of treatment for children with SAM, and with home rehabilitation (after daily ambulatory care during the initial phase of treatment).

Although hospital care was one of the investigated settings in all of the included trials, the inpatient care provided differed among the trials, for instance not only were different diet formulas and number of meals administered, but staff teams also varied in their composition. Similarly, the home-based care involved in the three trials6365 that investigated home-based rehabilitation also differed. Ciliberto and colleagues63 studied home-based care provided by caretakers and involved two weekly clinic visits at which RUTF supplies were given, whereas Heikens and colleagues'64 home-based treatment was supported by community health aides (CHAs) who were trained to offer standard health-service care. The frequency of care provided by CHAs, reported in an earlier publication that did not meet the inclusion criteria for this review,66 was unclear. Khanum and colleagues65 studied home-based care with no food supplements, and trained home visitors made home visits weekly for 1 month, then fortnightly. Trials also differed in the duration of the interventions and the length of follow-up.

Two studies62,63 included both moderately and severely malnourished children using similar criteria. Chapko and colleagues62 included children with a W/H z-score < −2 SD or kwashiorkor (not further defined) and Ciliberto and colleagues63 included children with a W/H z-score < −2 SD or mild oedema (< 0.5 cm of pitting oedema on the dorsum of the foot). However, both these studies were eligible for inclusion in this review. In the study by Chapko and colleagues,62 this was because the median W/H z-score was −3.38, with 70% of children having a z-score < −3, whereas in the study by Ciliberto and colleagues63 separate outcome data were presented for a subgroup of children with SAM (W/H z-score < −3 or presence of oedema). Heikens and colleagues64 also included both moderately and severely malnourished children judged to require hospital admission using the admission criteria of W/A < 80% of the NCHS median, oedema, anorexia, dermatosis or hair condition symptomatic of kwashiorkor and the need for treatment with parenteral antibiotics. Baseline status was described according to the Gómez,26 Wellcome67 and Waterlow classifications,29 which enabled the study to be included because the mean baseline W/A was ≤ 60% of the NCHS median. Khanum and colleagues65 included only children with SAM and used W/H < 60% of NCHS median and/or oedema as their admission criterion.

Reporting of exclusion criteria varied, with Chapko and colleagues62 not reporting exclusion criteria at all. Ciliberto and colleagues63 excluded children < 10 months of age and/or with severe oedema, systemic infection or anorexia. Heikens and colleagues64 excluded children with congenital abnormalities and/or siblings in the present study or in the authors' community study. Khanum and colleagues'65 reasons for excluding children from entry to the study were conditions that might require > 7 days of medical supervision (see Appendix 11), age < 12 months or > 60 months and children having TB or a congenital or metabolic disorder, children whose homes were > 10 km from the unit were also excluded.

Chapko and colleagues,62 Khanum and colleagues65 and Heikens and colleagues64 did not specifically identify their primary outcomes, but their main outcomes included mortality,62,65 utilisation62 (in terms of hospital and ambulatory days), days to reach oedema-free 80% W/H65 and time to discharge,64 W/H,62,64 W/A,62,64 H/A64 treatment completion65 and weight gain.65 Ciliberto and colleagues63 stated that their primary outcomes were successful recovery (W/H > −2 SD while remaining free of oedema), relapse or death.

Summary of quality assessment

As summarised in Table 45, the overall methodological quality for two of the trials62,63 was found to be moderate, while for the remaining two trials it was found to be weak.64,65

TABLE 45. Summary of methodological quality: studies of interventions in different settings.

TABLE 45

Summary of methodological quality: studies of interventions in different settings.

All four trials were judged to be at moderate risk of selection bias because, although 80–100% of the selected individuals participated in each of the four trials, their selected participants were judged as only somewhat likely to be representative of the target population. Two of the trials62,64 stated that children were randomly allocated to groups; however, during quality assessment they were judged to be CCTs (in accordance with the instructions on the use of the quality assessment tool, see Appendix 4) because the method of randomisation was not described. Nevertheless, the quality assessment tool still led to the trials being rated strong in terms of study design. Ciliberto and colleagues' study,63 was the only trial with important differences between groups at baseline (including differences in W/H, details in Appendix 11), but as 80–100% of the relevant confounders were controlled for in the analysis, all trials were rated strong with respect to confounders.

All trials showed weak methodological quality on blinding, as the outcome assessor was not blinded in three of the studies62,64,65 (Ciliberto and colleagues63 is unclear on this matter) nor were the participants blinded in any of them. The studies by Chapko and colleagues62 and Ciliberto and colleagues63 were found to be moderate regarding data collection methods, as their tools were valid, but their reliability was not reported. The Heikens and colleagues64 study was rated weak for data collection methods because information on the validity and reliability of the methods used was not reported. Khanum and colleagues65 used valid and reliable methods for the second follow-up after an additional 12 months, and hence, their study was rated as strong; however, the validity of the tools of the initial study could not be determined and their reliability was not reported, so this initial study was classified as weak. Three of the studies had 80–100% of their participants completing the study; hence, they were rated strong on withdrawals and dropouts, in spite of the fact that only Heikens and colleagues64 reported on both the numbers and reasons for missing data.

Studies vary widely in terms of the integrity of intervention. Chapko and colleagues62 reported that some children did not receive the assigned care; in particular, 11% of those assigned to ambulatory treatment received hospital rehabilitation at the insistence of their mothers, but it was not clear whether or not any of the children assigned to hospital care did not attend. Khanum and colleagues65 reported that 60–79% of their participants received their allocated intervention, and the other two trials63,64 reported 80–100%. Consistency was reported to have been measured by one trial65 and not measured by two trials;62,63 one trial was not clear on the matter.64 Contamination or co-intervention was likely to have occurred in Chapko and colleagues62 trial, whereas the other three studies6365 are not clear on this aspect.

The infant/child was the unit of allocation in three trials,62,64,65 whereas allocation was established per rehabilitation centre by Ciliberto and colleagues,63 whose trial had a stepped-wedge design. The unit of analysis in all four included trials was the infant/child. Overall, statistical methods were found to be appropriate for the design of two of the included studies,62,64 but it was unclear if they were appropriate for the other two studies.63,65 Out of the four studies, only Ciliberto and colleagues63 conducted an ITT analysis.

Assessment of effectiveness: settings

Mortality

Although mortality was a primary outcome of the Chapko and colleagues62 study and the Ciliberto and colleagues63 study, Khanum and colleagues65 reported it as a secondary outcome. None of the studies reported statistically significant differences in mortality between the different settings (Table 46). Chapko and colleagues62 reported a higher proportion of deaths in children in hospital than in the ambulatory setting. Ciliberto and colleagues63 reported only a 2.5% (95% CI −0.8% to 6.8%) difference in mortality between the groups, whereas Khanum and colleagues65 reported that mortality was low and did not differ between the groups (no p-value reported). Heikens and colleagues64 did not specify mortality as an outcome, but reported deaths among data on children lost to follow-up, and consequently no p-value was reported.

TABLE 46. Mortality in studies of interventions in different settings.

TABLE 46

Mortality in studies of interventions in different settings.

The authors of the included studies6365 accounted for the number of deaths that occurred in both the initial and the rehabilitation phases, apart from Chapko and colleagues,62 who reported deaths in the rehabilitation period only. This study had the highest proportion of deaths.

Weight gain

One study, that by Chapko and colleagues,62 did not include weight gain as an outcome. Khanum and colleagues65 reported the mean weight gain from admission to 80% W/H as a primary outcome and weight gain after an additional 12 months' follow-up as a secondary outcome (Table 47). A statistically significant difference in the primary outcome was found (p < 0.001), with inpatient care resulting in a greater daily mean weight gain from admission to the point at which participants reached 80% W/H than either home care or day care. However, after an additional 12 months of follow-up for all participants who reached 80% W/H, no significant differences in weight gain were apparent. Heikens and colleagues64 stated either that the rates of weight gain were similar or that there was no difference between the groups at the different treatment stages, but it is not clear whether or not any formal statistical testing was conducted and no p-values were reported. The exception was the average final rate of weight gain before discharge of 6–7 g/kg/day. The authors stated that this was maintained over a longer period for the long-stay group, but presented no data. Ciliberto and colleagues63 reported a non-statistically significant difference in the rate of weight gain for children at home compared with hospital during the first 4 weeks of the study. However, Ciliberto and colleagues63 performed a multivariate regression analysis, which showed that the overall rate of weight gain was 1.4 (95% CI 1.1 to 1.7) times as great among the severely malnourished children in those subject to home-based therapy than in those who received standard therapy at the rehabilitation unit.

TABLE 47. Weight gain in studies of interventions in different settings.

TABLE 47

Weight gain in studies of interventions in different settings.

Anthropometry

Chapko and colleagues62 reported W/H data in line graphs separately for those who died and for those who survived. Within both groups, there was no significant difference between the ambulatory-based and hospital-based groups (no p-value reported). Heikens and colleagues64 reported measures of W/H, W/A and H/A at discharge, after 6 months' home care (end of intervention) and then at 6-month intervals to 36 months post admission (groups altered in size at later time points; see Appendix 11). At discharge, the hospital (long-stay) group had a better W/H z-score than the home (short-stay) group and the difference between the groups was statistically significant [mean z-score ± standard error (SE): long-stay group −0.49 ± 0.11 vs −1.17 ± 0.16 in the short-stay group; p = 0.001]64 (Table 48). However, 6 and 12 months later, the difference between the groups was no longer statistically significant (p-values 0.105 and < 0.1, respectively), and the difference between the groups narrowed further at 18 months and at later time points (no p-values reported). Ciliberto and colleagues63 found a statistically significantly higher proportion of patients with W/H > −2 SD after 8 weeks of treatment in the group under home-based therapy than in the inpatient group.

TABLE 48. Anthropometric outcomes in studies of interventions in different settings.

TABLE 48

Anthropometric outcomes in studies of interventions in different settings.

Weight-for-age and H/A outcomes were reported only by Heikens and colleagues.64 Statistically significantly higher z-scores were reported for the long-stay group up to 24 months (W/A) or 30 months (H/A), but not thereafter (see Table 48 and Appendix 11).

Only Ciliberto and colleagues63 reported MUAC gain, finding a statistically significantly higher rate of MUAC gain during the first 4 weeks in the group under home-based therapy compared with the inpatient group.

Completion of treatment

Khanum and colleagues65 reported a significantly longer period of time to achieve 80% W/H in the group treated at home than in the group receiving hospital or ambulatory care (Table 49). The hospital (long-stay) group in the Heikens and colleagues64 study was considered to have completed treatment and was discharged when 95–100% W/H was reached (a mean ± SE of 39.45 ± 2.35 days post admission), but these data were not presented for the group that received care at home.

TABLE 49. Completion of treatment in studies of interventions in different settings.

TABLE 49

Completion of treatment in studies of interventions in different settings.

Relapse

Ciliberto and colleagues63 presented the composite outcome of children who relapsed or died (Table 50). It is presumed (although not explicitly stated) that this outcome incorporates the deaths already reported in Table 46, indicating, therefore, that 12 children (10.6%) in the hospital group relapsed, in comparison with 33 (6.2%) in the home group. A multivariate regression analysis was conducted to control for a range of covariates, and this indicated a statistically significantly lower probability of relapse or death [0.5 times (95% CI 0.3 times to 0.7 times)] in the subgroup of SAM children who received home-based therapy with RUTF compared with those receiving standard care at hospital. Khanum and colleagues65 reported on those who were readmitted during the 12-month follow-up. Children were readmitted if they relapsed (became oedematous or were < 60% W/H) or because of medical emergencies. Overall, there were eight readmissions (1.8% of the 437 children followed up), of which 0.6% occurred because of relapse. Data were not presented separately for relapse in each group, but Khanum and colleagues65 stated that relapse did not differ among the groups.

TABLE 50. Relapse after treatment in studies of interventions in different settings.

TABLE 50

Relapse after treatment in studies of interventions in different settings.

Additional outcomes

Other outcomes such as height gain, oedema loss and prevalence of fever, cough or diarrhoea were reported by some studies, but details have not been presented here. Full details are available in the data extraction forms in Appendix 11.

Summary

  • One moderate-quality CCT was found comparing ambulatory care with hospital care.62 Two other included CCTs involved home- and hospital-based therapy (one of them was graded moderate,63 whereas the other CCT's methodological quality was considered weak64), and another methodologically weak CCT65 compared the three settings.
  • Only one trial63 undertook an ITT analysis. There is the possibility, therefore, that the results of the remaining trials62,64,65 are at a higher risk of bias and, consequently, the intervention effect may not have been accurately captured.
  • None of the included studies reported a significant difference in mortality between groups.
  • Conflicting results were obtained for weight gain. No significant differences in weight gain during the first 4 weeks,63 at 12 months of follow-up65 or in the different stages up to 6 months after discharge64 were found between settings. However, a separate multivariate regression analysis in one trial found that overall rate of weight gain was greater among children in the home-based group than in those receiving standard therapy at the rehabilitation unit.63 In contrast, two studies of weaker quality64,65 reported that inpatient care presented a statistically significantly higher mean weight gain to 80% W/H than home or ambulatory care,65 and the final average rate of weight gain was maintained over a longer period in the inpatient group than in the home-care group.64
  • There was no significant difference during a 6-month follow-up period in W/H between the ambulatory and hospital-based groups in one study.62 However, two studies63,64 showed conflicting results for the comparison of hospital and home care. According to Heikens and colleagues,64 the inpatient group showed statistically significant improvement in W/H at discharge compared with the home-based group (supported by CHAs). In contrast, Ciliberto and colleagues'63 home-based group (visiting the rehabilitation centre fortnightly) included a statistically significantly higher proportion of patients with W/H > −2 SD after 8 weeks of treatment than in the inpatient group.
  • Statistically significantly higher W/A and H/A z-scores were found after hospital-based treatment than after home care with support of CHAs.64 In contrast, another trial reported a statistically significantly higher rate of MUAC gain during the first 4 weeks in the group under home-based therapy than in the inpatient group.63
  • Most studies defined SAM with criteria similar to those currently used by the WHO or analysed a subgroup that met them.62,63,65 It is not clear, however, whether or not the participants in Heikens and colleagues' study64 would meet current WHO criteria.
  • Studies varied in the care provided, even if the same setting is considered. For instance, studies on home rehabilitation6365 involved different diets, time and frequency of contact with nutritional rehabilitation centres/staff. Additionally, Chapko and colleagues62 point out that nutritional rehabilitation differed between ambulatory centres, and between the hospital and the ambulatory centres. Ciliberto and colleagues63 provide no indication regarding similarities or differences between centres.

Which methods for correcting micronutrient deficiencies are effective? (Q8, rank 10)

Thirteen trials6882 were included that investigated the efficacy of treatments for correcting micronutrient deficiencies in children with SAM. Any supplements or combinations of supplements were eligible for inclusion, providing other review inclusion criteria (e.g. reported outcomes) were met. Within this section, 10 trials6879 (12 publications) investigating zinc supplements have been grouped together for ease of comparison between studies (see Quantity and quality of research available: zinc and Assessment of effectiveness: zinc). The remaining three trials,8082 each focus on different interventions: potassium,80 nicotinic acid81 and nucleotides (NTs)82 (see Quantity and quality of research available: other supplements and Assessment of effectiveness: other supplements).

Quantity and quality of research available: zinc

A summary of the key characteristics of the 10 trials6879 can be seen in Table 51, with further details of the trials in Appendix 12. Three trials took place in Bangladesh,68,69,73,77 two in India,72,78 and one trial each in Pakistan,79 Kenya,70 Jamaica,71 South Africa74,75 and Chile.76 Two studies were RCTs,68,69,79 the remaining eight were CCTs7078 and all were conducted at a single centre. Three studies did not report on how the study was funded,72,73,78 one study was funded by a commercial research grant programme76 and another study funded from a US academic source.79 The remaining five studies received funding from a combination of a commercial and an academic source (two studies68,69,74,75 with academic sources either in the UK68,69 or in South Africa74,75), or an academic source and a government department (both in Kenya),70 or an academic source and a charity (both based in the UK).71,77

TABLE 51. Characteristics of the included studies providing zinc as a supplement.

TABLE 51

Characteristics of the included studies providing zinc as a supplement.

The age range of the children enrolled in each study varied. In three studies,70,72,77 children ranged in age from a minimum of 1 year to a maximum of either 3 years,70 5 years72 or 7 years77 (this study could be included because the mean age of participants was < 5 years of age). Six studies allowed for the inclusion of children under 1 year in age, with ages ranging from 6 months to about 2.5 years in one study,71 6 months to 3 years in two studies,68,69,79 and in three studies from either 6 months to 5 years,74,75 5 months to 5 years73 or 8 months to 2 years.78 Only one study focused on children aged < 1 year.76 The total number of children enrolled and reported on in the trials ranged from 1171 to 30074,75 with most studies reporting on fewer than 100 participants.

Four studies enrolled approximately equal numbers of male and female children,70,7376 in one study the population contained more male than female children,79 and one study enrolled only male children.71 In the remaining four studies, the ratio of male to female children was not reported.68,69,72,77,78

The criteria for identifying SAM included a W/A assessment in 8 of the 10 studies,6872,74,75,7779 although only three68,69,79 reported baseline W/A for each intervention arm. In three studies,72,78,79 this was the only criterion used. Hemalatha and colleagues72 included children with a W/A < 60% of that expected in comparison with the NCHS reference median using the Gómez criteria,26 whereas Vasudevan and colleagues78 included children with 51–60% expected W/A (PEM grade III) and ≤ 50% W/A (PEM grade IV), based on the Harvard standard according to the IAP 1972 classification of PEM.45 The third study, by Bhutta and colleagues,79 was the only one to focus on children with persistent diarrhoea in combination with evidence of malnutrition, defined as a W/A z-score ≤ 2. This study was included in this section, rather than in Quantity and quality of research available: persistent diarrhoea and Assessment of effectiveness: persistent diarrhoea, because the primary outcome was weight gain (whereas diarrhoea-related outcomes were secondary outcomes), and plasma zinc levels were checked before and after supplementation. The Wellcome classification67 was employed by three studies.70,71,74,75 This categorises W/A as either < 60% or 60–80% of that expected based on the Harvard standard, and combines this with an assessment of whether oedema is present or absent to identify children with kwashiorkor, marasmic kwashiorkor or marasmus (children in the fourth category, undernourished, were not included). One of the studies, that by Makonnen and colleagues,74,75 also included children with W/A > 80% if they had signs and symptoms of kwashiorkor. Two other studies also included an assessment of nutritional oedema in their criteria for assessing SAM. Doherty and colleagues68,69 included children with nutritional oedema, or with W/A < 60% of the NCHS median, or both. Simmer and colleagues77 included children with nutritional oedema, or W/A < 60% and W/H < 70% of local standards (< 42% and < 63% respectively of Western standards, not further defined).

Two studies did not use W/A in their assessment of SAM. Khanum and colleagues73 included children with oedema and/or with ≤ 60% of the W/H expected in comparison with the Harvard reference. Schlesinger and colleagues76 described all the infants in their study as marasmic.

Although the 10 trials6879 investigating zinc supplements have been grouped together for ease of comparison, the interventions varied in many aspects: when zinc supplementation began, the daily dose, the mode used to administer this and the duration it was provided for. In addition to the summary information presented in Table 51, further details of zinc supplementation can be found in Table 52.

TABLE 52. Details of zinc supplementation provided.

TABLE 52

Details of zinc supplementation provided.

The comparator in four trials was a placebo;72,74,75,78,79 three trials did not provide a placebo to the comparator group,70,73,77 and three trials varied dose and/or duration of treatment between the groups.68,69,71,76

Only 2 of the 10 studies specified what their primary outcome measures were.74,75,79 Mortality was reported by two studies,68,69,74,75 weight or weight gain was an outcome in seven studies7073,7779 and anthropometric measures were reported by five studies68,69,7376,79 (including the three that did not report on weight). Eight studies reported on zinc levels,70,7279 four reported on symptoms70,72,74,75,79 (e.g. diarrhoea, oedema) and seven reported one or more other outcomes (e.g. results of biochemical tests, length of hospital stay).68,69,71,72,7477,79

Summary of quality assessment

The trials were assessed against a number of criteria to obtain an overview of their methodological quality. The global rating for methodological quality varied: two trials68,69,79 were rated strong overall, two trials7476 were rated moderate and six trials were rated weak7073,77,78 (Table 53).

TABLE 53. Summary of methodological quality: studies providing zinc as a supplement.

TABLE 53

Summary of methodological quality: studies providing zinc as a supplement.

The assessment of selection bias required a judgement about how likely it was that participants selected to take part in the study were representative of the target population and information about the percentage of selected individuals who did participate. In general, this information was not well reported. Consequently, only two studies74,75,79 were rated ‘strong’ (at low risk of selection bias). Of the remaining studies, four had a moderate rating68,69,73,77,78 and four had a weak rating7072,76 (the latter judged to be at a high risk of selection bias).

The study design of all 10 trials was strong, with two being RCTs68,69,79 and the remaining eight being CCTs7078 (in two of these,72,74,75 there was a suggestion of randomisation but no information was provided about the method, hence they were judged to be CCTs). Both of the RCTs68,69,79 and four of the CCTs71,73,76,77 were judged to have trial arms that were balanced with respect to baseline characteristics and confounders, leading to a strong rating. Reporting in the remaining four CCTs70,72,74,75,78 was either insufficient or unclear, so it was not possible to be certain whether or not the trial arms were balanced, which led to a weak rating. Six trials68,69,72,7476,78,79 were described as double-blind; of these, five used a placebo and were judged strong with respect to blinding.68,69,72,74,75,78,79 One trial76 was judged moderate because it was not clear whether or not the outcome assessor was blinded to the intervention status of the participants. Four trials,70,71,73,77 that did not use a placebo were all judged weak with respect to blinding.

When judging the methodological quality of data collection methods, the judgement could differ depending on the outcome measure. Therefore, the data collection method judgements reported here are for the primary outcomes of this systematic review (see Appendix 12). In five trials,7173,77,78 the data collection tool was not described and, therefore, could not be judged as valid or reliable. The remaining trials all used valid data collection tools, but only two trials68,69,76 provided information indicating that the tools were reliable, which allowed data collection to be judged strong. The other three trials,70,74,75,79 were rated as moderate because no information about the reliability of the tools was reported. Most of the studies were rated as methodologically strong for the item on reporting of withdrawals and dropouts, even though six studies7073,76,78 did not report the number and reasons for withdrawals and dropouts. The strong rating could be applied because the proportion of participants completing these studies lay between 80% and 100%. Only two studies did not gain a strong rating: one70 was judged moderate and one72 was judged weak because < 60% of participants completed the study.

The final two elements of the quality assessment tool, ‘intervention integrity’ and ‘analysis appropriate to question’, did not contribute to the global rating, but nevertheless provided important information about each study. In all studies, 80–100% of participants received their allocated intervention, in six trials the consistency of the intervention was measured,71,73,7679 and in eight trials7073,7679 it was judged unlikely that any unintended intervention had been implemented. All studies allocated individual infants/children to trial arms and in only one study72 was there insufficient detail to determine whether or not appropriate statistical methods had been used. Three studies71,76,79 performed an ITT analysis.

Assessment of effectiveness: zinc

Mortality

Mortality was reported by only two of the studies68,69,74,75 investigating zinc as a supplement (Table 54). Doherty and colleagues68,69 conducted a planned interim analysis of the first 100 participants. Data from the two groups receiving 6 mg/kg zinc were combined and, when compared with the group receiving 1.5 mg/kg zinc, the risk of death in the 90-day study period was significantly greater for those receiving 6 mg/kg (Yates'-corrected chi-squared value of risk of death 4.52, 95% CI for relative risk of 1.09 to 18.8; p = 0.03). This led to the suspension of enrolment to the trial, by which point 141 participants had been recruited. Of the 19 deaths that occurred overall (all three groups combined), 13 occurred during the inpatient phase and 11 of these occurred in one of the two 6 mg/kg zinc groups. The six deaths in the outpatient phase all occurred in children who had received 6 mg/kg zinc as inpatients (intervention group two who subsequently received placebo in the outpatient phase) or were still receiving 6 mg/kg zinc (intervention group 3). The clinician's impression was that sepsis was the cause of death in most cases. Doherty and colleagues68,69 conducted an analysis of a range of predictive/prognostic factors, but found that none of these factors in conjunction with the higher dose of zinc was predictive for death.

TABLE 54. Mortality in children receiving zinc supplements.

TABLE 54

Mortality in children receiving zinc supplements.

Makonnen and colleagues74,75 also recorded the majority of deaths during the initial period of hospitalisation. However, in this trial the authors state that there were significantly more deaths in the control group, who did not receive a zinc supplement, than in the group that did receive a zinc supplement [17.3% mortality in the control vs 4.7% in the zinc group; no p-value given, but a 95% CI of 5.5% to 19.5% was reported (although not clear in the paper, it appears that this is likely to be the 95% CI for the difference between the groups)]. In each group, some participants had to be readmitted after discharge. In the zinc group, two participants were readmitted, one 5 days after discharge, who was subsequently discharged, and a second identified at the 30-day follow-up who subsequently died. In the control group, four participants were readmitted, three after initial discharge, of whom one subsequently died; the fourth was identified at the 30-day follow-up and also subsequently died.

Weight gain

Seven studies reported weight gain as an outcome, although the reported measures varied: overall weight gain, daily or weekly weight gain, grams of weight gained per kg body weight per day and proportion meeting a threshold value of weight gain (Table 55). All the studies except that by Gatheru and colleagues70 reported weight gain relative to initial weight or reported the proportion of participants meeting a threshold that was a measure of relative weight gain. These measures help remove differences due to starting differences in body weight. Three studies,70,73,77 reported at least one statistically significant difference in a weight gain outcome between the groups in favour of zinc supplementation, but four studies71,72,78,79 found no evidence for a statistically significant difference between groups. A meta-analysis was not carried out because it was considered to be inappropriate because of heterogeneity in the dose(s) of zinc provided, differences in the reported outcome measures [units and time point(s) of measurement] and other limitations of the data (missing measure of variance).

TABLE 55. Weight gain in children receiving zinc supplements.

TABLE 55

Weight gain in children receiving zinc supplements.

Gatheru and colleagues70 reported a significantly greater total weight gain in the zincsupplemented group than in the control group (mean ± SD, 531 ± 277 g vs 338 ± 235 g; p < 0.05). However as noted above, the measures used in this study do not take into account starting differences in body weight. Approximately one-quarter of each group did not complete the study, but the reasons for this are not provided and it is not clear whether or not this had any impact on the outcome. Mean daily weight gain was greater in the zinc-supplemented group (67 g/day) than in the control group (47.3 g/day), but no p-value for a statistical comparison of these values was reported.

Khanum and colleagues73 found that during the first 2 weeks of rehabilitation (which was before zinc supplementation began), the rate of weight gain was not significantly different between the groups (see Appendix 12). However, at the start of the third week the intervention group started to receive a zinc supplement and Khanum and colleagues73 found that by the end of that week the group receiving the zinc supplement had gained significantly more weight than the non-supplemented group (mean ± SE: zinc group 580 ± 67.6 g/week, control group 342 ± 86.5 g/week; p < 0.05). This statistically significant effect was maintained in the following week, but by the fifth week of the study (after 3 weeks of zinc supplementation), the difference in weekly weight gain was no longer statistically significant. Khanum and colleagues73 also report that the percentage of participants in the zinc-supplemented group who achieved a mean daily weight gain of > 10 g/kg was statistically significantly greater than in the control group (66% vs 33%; p < 0.02). It is not clear for what time period the mean daily weight gain was calculated.

Simmer and colleagues77 also reported that the percentage of participants in the zinc-supplemented group who achieved a mean daily weight gain of > 10 g/kg was statistically significantly greater than in the control group (42% vs 9%; p < 0.001). However, in this study, although the zinc group gained more weight each week than the control group, statistically significant differences between the groups for mean daily weight gain and mean daily weight gain per kg body weight could not be demonstrated in either week 1 (see Appendix 12) or week 2 (see Table 55).

All four of the studies71,72,78,79 that did not find evidence for a statistically significant difference between groups reported on the rate of weight gain in terms of g/kg/day (the benefit of a relative measure such as g/kg/day is that any effects due to starting differences in body weight are removed). Golden and colleagues71 reported mean weight gain during the first 6 weeks of recovery. Although children in the moderate- and high-zinc groups gained weight more rapidly, the difference between the groups was not significant. Hemalatha and colleagues72 reported on this outcome separately for weeks 1, 2 and 3 (see Appendix 12) where gains appeared similar between groups, and for week 4 (see Table 55) when the difference in outcome was described as not statistically significant (mean ± SE: 22.6 ± 5.100 g/kg/day in the zinc group vs 24.5 ± 5.035 g/kg/day in the control group; no p-value reported). Bhutta and colleagues79 reported g/kg/day weight gained separately for the 14 days of inpatient and for the following 14 days of home-based supplementation. In both cases, although the zinc-supplemented group gained a little more weight than the control group and the differences were not found to be statistically significant (see Table 55; no p-value reported). Bhutta and colleagues also reported the total weight of children at baseline, on day 7 (see Appendix 12) and on day 14 of inpatient therapy, but again there was no significant difference between the groups (mean weight on day 14 in the supplemented group 6.67 ± 1.43 kg vs 7.13 ± 1.42 kg in the placebo group; p = 0.27). Finally, Vasudevan and colleagues78 reported a rate of weight gain in terms of g/kg/day, which was much lower than that of the other studies (zinc-supplemented group 1.4 g/kg/day vs 0.98 g/ kg/day in the control group; p > 0.1). This lower value may be because the authors appear to have based their calculation on the initial weight and a single further weight measurement taken after 3 months.

Anthropometry

Five studies reported anthropometry outcomes (Table 56).68,69,7376,79 Two of these studies also reported weight gain as a separate measure,73,79 but anthropometry outcomes were the only measures capturing weight gain for the other three studies.68,69,7476 One study73 reported a statistically significant difference in anthropometry outcomes between the groups in favour of zinc supplementation, two studies reported mixed results7476 and two studies found no evidence for a statistically significant difference between groups.68,69,79 It was considered inappropriate to carry out a meta-analysis of these outcomes because of heterogeneity in the dose(s) of zinc provided and differences in the reported outcome measures.

TABLE 56. Anthropometric outcomes in children receiving zinc supplements.

TABLE 56

Anthropometric outcomes in children receiving zinc supplements.

Khanum and colleagues73 reported statistically significant differences in favour of zinc supplementation in measures of W/H, W/A and the percentage of patients reaching or exceeding 90% W/H by discharge. There were no significant differences in W/H and W/A at admission (see Appendix 12), but a significant difference in favour of the zinc group was present in W/H by the beginning of the first week (eighth day) and in W/A by the beginning of the second week of nutritional therapy (15th day). As zinc supplementation only began on day 15, it is not clear what led to the differences emerging at this stage. The differences were then maintained for each of the following weeks during zinc supplementation (see Appendix 12) until discharge on day 36 (at discharge W/H in the zinc group 95 ± 1.2, control 86 ± 1.2, p < 0.001; W/A in the zinc group 68.1 ± 1.58, control 59.7 ± 1.77, p < 0.001). Over 75% of participants were discharged with a W/H of ≥ 90% of that expected in the zinc-supplemented group, whereas < 25% of the control group reached W/H of ≥ 90% of that expected at discharge.

Makonnen and colleagues74,75 found that at discharge from inpatient care (after a mean length of stay of approximately 11 days), there were no significant differences in anthropometry between the groups. However, the zinc group continued to receive supplements until 90 days after discharge, and at the 90-day follow-up the proportion of children with W/A < 60% was described as significantly lower in the zinc group than that of the control group (3.6% vs 13.8%, 95% CI for difference −17.2% to −3.1%; no p-value reported), and the proportion with W/A > 80% and no oedema was greater (58.7% vs 28.4%; no 95% CI for difference or p-value reported). The proportion of zinc-supplemented participants whose MUAC remained below the fifth percentile was also lower than those children in the control group (54.1% vs 77.9%, 95% CI for difference −35.2% to −11.5%; no p-value reported).

Schlesinger and colleagues76 reported outcomes after 105 days of supplementation, but this was the only study in which the participants were all < 1 year of age and so received liquid formula only. No significant difference between the groups was apparent when z-scores for H/A, W/A and W/H were analysed (see Table 56 and Appendix 12). However, there was a statistically significant difference in the proportion of infants in the zinc-supplemented group whose percentile H/A score increased after 30 and 45 days of nutritional rehabilitation, indicating that the zinc group began to grow earlier than the control group [58% (at 30 days) and 79% (at 45 days) of the 15 mg/l zinc group had an increase in their H/A percentile score in relation to their admission score in comparison with only 20% (at 30 days) and 45% (at 45 days) in the 3.2 mg/l zinc group; p < 0.002 and p < 0.03, respectively]. When these data were analysed by sex, the authors found that the effect held for males, but not for females (see Appendix 12). This difference was no longer significant after 60 days of nutritional rehabilitation (see Table 56).

Doherty and colleagues68,69 reported on the changes in W/A, W/H and H/A z-scores and MUAC (see Table 56), as well as knemometry and skinfold thickness (see Appendix 12). There were three groups in this study and two comparisons were made, the first between the 1.5 mg/kg and the 6 mg/kg zinc dose for 15 days and the second between the 6 mg/kg zinc dose for 15 days or for 30 days. The results reported after 90 days of follow-up indicated that no significant differences were reported for either comparison for any of the anthropometric measures reported. Doherty and colleagues68,69 did find that, overall, good catch-up growth was achieved over 90 days, with the average intragroup W/H z-score improved from 1.54 to 1.67 units, and the H/A z-score improved from 0.44 to 0.49 units.

Finally, Bhutta and colleagues79 reported on just the anthropometric measure of MUAC, finding no significant differences in this measure after 14 days of inpatient therapy (mean MUAC: zinc group 12.0 ± 1.4 cm vs 12.4 ± 1.8 cm in the placebo group; p = 0.66), and no significant differences emerged after a further 14 days of home-based supplementation (see Table 56).

Comorbidities

The effect of zinc supplementation on comorbidities was an outcome reported by five studies.70,72,7476,79 In three studies,70,72,79 at least some of these comorbidities had been present at baseline and the study authors report on the effect of the intervention in resolving these. In the study76 in which comorbidities were recorded daily for 105 days and the study74,75 reporting outcomes at 90-day follow-up, it is likely that new incidences of comorbidities are captured.

Three studies70,72,74,75 reported oedema as an outcome (Table 57). Gatheru and colleagues70 reported that the duration of oedema ranged between 2 and 18 days for both groups. However, a greater proportion of participants in the zinc group had lost their oedema by day 7 (77% vs 55%), and the mean number of days taken to lose oedema was statistically significantly lower in the zinc group (mean ± SD: zinc group 6.3 ± 4.6 days vs control group 8.1 ± 4.4 days; p < 0.05).In contrast, Hemalatha and colleagues72 found no statistically significant difference in the mean number of days taken to lose oedema (mean ± SE: zinc group 9.0 ± 2.0 days vs control group 15.7 ± 2.7 days). In the final study to report oedema, Makonnen and colleagues74,75 found a trend for the zinc group to recover more quickly during the first 3 weeks of hospitalisation, but the difference between the groups was not statistically significant (see Appendix 12) and by the 90-day follow-up no patient in either group had any oedema.

TABLE 57. Comorbidities in children receiving zinc supplements.

TABLE 57

Comorbidities in children receiving zinc supplements.

Diarrhoea was reported by four studies70,7476,79 and the results were mixed, with two studies70,74,75 finding statistically significant differences in favour of zinc supplementation, one study76 finding a statistically significant difference in favour of the control group, and one79 finding no statistically significant differences between the groups. Gatheru and colleagues70 reported that the duration of diarrhoea was statistically significantly lower in the zinc group than in the control group (zinc group 3.62 ± 2.78 days vs control group 10.8 ± 3.4 days; p < 0.001). It is not clear whether the participants contributing data to this outcome all had diarrhoea at baseline or whether new episodes of diarrhoea arising during treatment are included. Makonnen and colleagues74,75 also found in favour of zinc supplementation when reporting on the proportion of participants with diarrhoea at the 90-day follow-up. A statistically significant difference was observed, with fewer participants having diarrhoea in the zinc-supplemented group than in the placebo group (zinc group 2.9% vs placebo group 36.7%, 95% CI for difference −32% to −15%). In contrast, Schlesinger and colleagues,76 who analysed infectious episodes using three indices (mean episodes/infant, mean duration of each episode and mean percentage days infected during the 105 days of rehabilitation), found a statistically significant difference in the average number of acute diarrhoeal episodes that favoured the formula with low zinc content (zinc 15 mg/l formula two episodes vs zinc 3.2 mg/l zero episodes; p-value not reported). Schlesinger and colleagues76 stated that each episode lasted 1–2 days and had no impact on nutritional rehabilitation. Finally, Bhutta and colleagues,79 who focused their study on the treatment of children with diarrhoea at baseline, recorded two measures for the 14-day inpatient phase of their study: stool frequency and stool volume. No statistically significant differences were found between the groups for either measure (see Table 57).

The remaining morbidity outcomes reported by the studies (see Table 57) provide mixed results. Two studies70,74,75 reported statistically significant differences in favour of the zinc-supplemented group: Gatheru and colleagues70 for duration of anorexia and time taken for skin lesions to heal and Makonnen and colleagues74,75 for the proportion of participants with vomiting, fever, acute respiratory infections, skin infections or pallor at the 90-day follow-up. One study76 reported that the mean number of otitis media episodes came close to reaching a statistically significant difference between the groups in favour of the group receiving a higher concentration of zinc (zinc 15 mg/l formula 0.73 ± 0.9 vs zinc 3.2 formula 1.85 ± 2.3; p-value between 0.05 and 0.1), but the same authors stated that for number or duration of upper and lower respiratory infection, purulent conjunctivitis, and skin and mucous candidiasis, no differences were observed between groups (see Appendix 12). Finally, for the outcome duration of morbidity because of infections, Hemalatha and colleagues72 state that the groups were comparable, but no p-value is reported.

Adverse effects

Oral exposure to large doses of supplemental zinc is associated with some known adverse events, such as gastrointestinal effects (e.g. nausea) and copper deficiency.83 Two studies specifically checked for the adverse effects of zinc on plasma copper levels.72,79 Hemalatha and colleagues72 reported that, although plasma copper levels rose significantly in both groups, the zinc supplementation did not adversely affect plasma copper levels. In contrast, Bhutta and colleagues79 found that serum copper levels fell during zinc supplementation, whereas in the placebo group serum copper levels significantly increased. Both Simmer and colleagues77 and Vasudevan and colleagues78 reported that no adverse effects were noted in the zinc-supplemented groups. Simmer and colleagues77 reported that tube feeding was required for one patient in each group and two patients in each group received a blood transfusion. This study additionally reported on anorexia, which can develop because of zinc deficiency and might have been expected to affect the control group, but there was no difference in calorie or protein intakes between the groups. Two trials68,69,76 reported on comorbidity events which have already been noted in the sections above. Doherty and colleagues68,69 had to suspend enrolment to their trial when the groups receiving 6 mg/kg zinc supplements were found to be at a significantly greater risk of death than the group receiving 1.5 mg/kg zinc supplements. Schlesinger and colleagues76 found that infants receiving formula with a zinc content of 15 mg/l had on average two acute diarrhoeal episodes, whereas infants receiving 3.2 mg/l zinc formula had no such episodes. However, this statistically significant difference in diarrhoeal episodes had no impact on nutritional rehabilitation. The remaining studies did not report adverse events.70,71,7375

Zinc status

Eight70,7279 of the 10 studies reported on zinc status, and the results are briefly summarised here, with details provided in Appendix 12. It should be noted, however, that although serum or plasma zinc may be the best available biomarker at a population level to reflect dietary zinc intake, and although it changes in response to zinc supplementation, it does not necessarily reflect individual zinc status.84 Five studies70,7375,78,79 found a statistically significant difference in serum zinc concentrations between the groups, with serum zinc being higher following supplementation than it was in the non-supplemented group. One study found a statistically significant difference in the proportion of participants defined as having a low plasma zinc, which was in favour of the zinc-supplemented group.76 Two studies72,77 did not report differences between groups, but did report statistically significant increases in plasma zinc in comparison with baseline values in either the zinc group or both study groups. Three studies72,76,77 also reported the difference from baseline in leucocyte zinc. One study72 found that leucocyte zinc was statistically significantly increased from baseline in both groups, one77 reported a statistically significant difference for the zinc-supplemented group only, and one76 found no significant differences in either group.

Other outcomes

Additional outcomes, such as duration of hospital stay, calorie intake, nitrogen intake and results of biochemical assays, were also reported by some studies, but have not been presented here. Further details are available in the data extraction forms in Appendix 12.

Summary

  • Ten trials investigated zinc supplements as part of a treatment regimen for SAM.6879 The trials took place in seven countries, and employed different criteria for identifying SAM; in addition, the age range of enrolled children differed and most reported on fewer than 100 participants. One study focused on participants who also had persistent diarrhoea.79 The interventions also varied in many aspects. More than half of the studies were of weak methodological quality,7073,77,78 two were of moderate quality7476 and only two were judged to have strong methodological quality.68,69,79 Only three trials conducted an ITT analysis of the data.71,76,79
  • Only two studies68,69,74,75 reported mortality as an outcome. One study68,69 of children aged from 6 to 36 months suspended enrolment to the trial after an interim analysis found a significant risk of death for participants receiving 6 mg/kg/day zinc in comparison with those receiving 1.5 mg/kg/day zinc. In contrast, in the other study74,75 that enrolled children aged from 6 to 60 months, significantly more deaths were reported in the group receiving placebo than in the group receiving a zinc dose of 10 mg/day. One study provided zinc according to body weight68,69 and the other as a fixed dose,74,75 and although neither study reports on the weight of participants, the 10 mg/day dose is likely to be one of the lowest provided whereas the 6 mg/kg/day dose is likely to be one of the highest. It is difficult to know how similar the participants were because the studies do not report on the same anthropometric characteristics at baseline, although the criteria for defining SAM were comparable.
  • Intervention effects on weight were reported directly by seven studies7073,7779 (e.g. as absolute gains in weight or rate of weight gain) or as anthropometric measures by five studies68,69,7376,79 (e.g. W/H, W/A), with two of these studies reporting weight both directly and by anthropometry. The results for weight and anthropometry outcomes are conflicting, with three studies70,73,77 reporting statistically significant effects in favour of zinc supplementation, two studies reporting mixed results7476 (in favour of zinc or no significant differences between the groups) and five studies68,69,71,72,78,79 finding no significant differences between the groups. None of the studies reported a statistically significant effect in favour of the comparator/control group for a weight-related outcome.
  • Five studies70,72,7476,79 reported the effect of zinc supplementation on comorbidities, either those present at baseline or also including new incidences of comorbidities. Again, the results present a mixed picture. One study70 found that zinc significantly reduced the duration of four comorbidities present at baseline, whereas another study74,75 reported a statistically significant beneficial effect of zinc at the 90-day follow-up for six of the seven comorbidities assessed. In contrast, two other studies found no statistically significant effects on diarrhoea (one study79) or oedema (one study72), and a third study76 stated that there was a statistically significant difference in the average number of acute episodes of diarrhoea that favoured the control group (i.e. lower dose zinc).
  • Four studies70,71,7375 did not report on adverse events and two77,78 reported that no adverse events occurred because of zinc supplementation. Two studies specifically reported the impact of zinc supplementation on plasma copper because exposure to large doses of supplemental zinc is known to cause copper deficiency. In one study72 plasma copper levels rose in both groups, whereas in the other study79 copper levels significantly increased in the placebo group but fell in the group receiving zinc, and the study authors suggest that it may be more appropriate to provide a mix of micronutrients as a supplement rather than zinc alone. As already noted above, statistically significant differences in mortality were reported by one study67,68 in which more deaths occurred in the 6 mg/kg/day zinc-supplemented group, and in another study more episodes of acute diarrhoea75 occurred in the zinc group, although in this latter case this was stated not to have affected nutritional rehabilitation.

Quantity and quality of research available: other supplements

There were three trials examining supplements other than zinc: one evaluating potassium,80 one evaluating nicotinic acid81 and one evaluating NTs82 (Table 58). All three trials were conducted in single centres. One trial was a double-blind RCT,82 whereas the other two trials were judged to be CCTs during quality assessment (see Table 59).80,81 The trials were set in India,81 Malawi80 and Mexico,82 with none reporting any external funding.

TABLE 58. Characteristics of the included studies providing other supplements.

TABLE 58

Characteristics of the included studies providing other supplements.

TABLE 59. Summary of methodological quality: studies providing other supplements.

TABLE 59

Summary of methodological quality: studies providing other supplements.

There was no consistency in the type of supplement compared by the three trials. Manary and Brewster80 evaluated a high dose of potassium (total dose 7.7 mmol/kg/day) versus a standard dose (total dose 4.7 mmol/kg/day) for the first 7 days of therapy. Philip and colleagues81 compared the addition of three doses of nicotinic acid per day (25 mg/kg /day) versus none for 1 month. Vásquez-Garibay and colleagues82 compared a milk-based formula with added NTs versus a formula of the same energy density, but without the addition of NTs. All trials took place in the hospital inpatient setting.

The sample sizes were small, varying from 2582 to 116 children.80 Mean age ranged from 7.6 to 8.1 months in one trial82 and from 27.9 to 29.3 months in another.80 However, both these trials reported baseline characteristics for children completing the trial only. Philip and colleagues81 did not provide any baseline characteristics, but included children aged from 0 to 4 years. Sex was reported by only one trial,82 with almost double the number of boys as girls.

Only one trial, by Vásquez-Garibay and colleagues,82 provided a definition for severe malnutrition, this being W/A or W/H of < −3 SD from the median NCHS reference. Manary and Brewster80 included children with kwashiorkor, all of whom had oedema, whereas Philip and colleagues81 included children fulfilling what they called ‘the standard criteria’ for marasmus.

Baseline mean W/H z-scores suggested a greater severity of malnutrition in the children in the trial by Vásquez-Garibay and colleagues82 than in those in the trial by Manary and Brewster.80 However, it appears that the W/H z-scores in the latter study were calculated before loss of oedema (z-scores would be expected to be lower, indicating more severe malnutrition after loss of oedema). Also, in the trial by Vásquez-Garibay and colleagues,82 mean baseline arm circumference (assumed to be MUAC) appears to be very low, falling well below the threshold for SAM. Without baseline characteristics, it is unclear if the severity of malnutrition was comparable with the population in the trial by Philip and colleagues with that in the other two trials.81

Summary of quality assessment

Two of the included trials were rated overall as weak for their methodological quality and quality of reporting (Table 59),81,82 with the third being rated overall as moderate.80

For selection bias, two trials were rated as strong,80,82 whereas the study of Philip and colleagues81 was rated as weak and therefore at potential risk of selection bias. A strong rating indicates that the selected individuals are likely to be representative of the target population and ≥ 80% of selected individuals participated in the trial. A weak rating indicates that participants may not be representative of the target population, or that the selection method and/or levels of participation were unclear/not described. All three trials were rated as strong for their study design. Two of the trials were described as RCTs80,82 and one as a CCT;81 however, the trial by Manary and Brewster80 was judged to be a CCT, as it provided no details of the randomisation method/procedure.

Only one trial had no important differences in baseline characteristics between the trial arms, and without potentially confounding variables was rated as strong.80 The remaining two trials81,82 were rated as weak. Philip and colleagues81 only reported age for baseline characteristics and it is, therefore, unclear if there were any confounding variables, whereas Vásquez-Garibay and colleagues82 acknowledged some differences between the treatment arms. It is unclear if these differences were between or within treatment arms. For blinding, only one trial employed a double-blind method and was therefore rated as strong.80 The two remaining trials provided no details and were therefore rated as weak.81,82 Both trials could therefore be at risk of bias in either the care provided (performance bias) or how the outcomes were assessed (measurement or detection bias) or both. Not blinding children/parents to the research question could lead to reporting bias. It may not always be possible to blind children/parents to the intervention, but the potential bias needs to be kept in mind when interpreting the results.

Two trials were rated as weak80,81 and one as moderate82 for their data collection methods. Only the trial by Vásquez-Garibay and colleagues82 used valid data collection tools, but it was not possible to judge if these tools were reliably employed. For the trials rated as weak, it was not possible to assess if either the data collection tools were valid or reliable.80,81 All three trials were rated as strong for withdrawals and dropouts, providing both numbers and reasons, as well as having ≥ 80% of participants completing the study. For intervention integrity, there was no information in one trial on either the percentage of participants who had received the intervention or if the consistency of the intervention had been measured.81 In the other two trials,80,82 ≥ 80% of participants received the allocated intervention, but only one of these measured the consistency of the intervention.82 There appeared to be no contamination of the interventions (i.e. all children received the allocated intervention only) in any of the three trials. All trials used the infant/child as the unit of allocation and for statistical analysis of the results, and were judged to use appropriate methods of statistical analysis for the research question. However, only one of the trials reported on how missing data were dealt with in the analysis (i.e. ITT).81

Assessment of effectiveness: other supplements

Mortality

Only Manary and Brewster80 reported mortality as an outcome (Table 60). There were a total of 34 (34%) deaths in hospital during the trial, 14 out of 48 deaths in the high-dose potassium intervention group compared with 20 out of 51 in the standard potassium control group. However, although the case fatality rate reduced by 33% in the intervention group, the difference between the groups was not statistically significant (p = 0.40). Twenty-one of the 34 deaths were early deaths (within 5 days) and 13 were late deaths (after 5 days). However, 11 children (intervention n = 3, control n = 8) were taken from hospital after completing the 7-day trial before discharge, resolution of oedema and clinical improvement. Only the percentage of late deaths was statistically significant after adjusting it to include three children (all from the control group) who had left hospital and who were not expected to have survived at home, with a lower number of deaths in the high-dose potassium intervention group (8%) than in the standard potassium control group (32%; p = 0.02) [OR 5.3 (95% CI 1.2 to 31.0)]. The five children whose deaths between days 9 and 13 were classified as unexpected all had persisting diarrhoea.

TABLE 60. Mortality in children receiving other supplements.

TABLE 60

Mortality in children receiving other supplements.

Weight gain

Two out of the three trials reported weight gain, but only one trial81 reported weight gain relative to initial weight (thus removing effects owing to the starting differences in body weight).81,82 The trial by Manary and Brewster80 only reported weight loss, although it is not particularly clear if this was because of resolution of oedema.

Philip and colleagues81 calculated separately for each week, with both groups showing maximum gain during week 2, followed by week 3, with the lowest gain in weeks 1 and 4 (no data reported). For both groups, the rate of weight gain was slightly higher in those children with a greater initial weight deficit. Mean weight gain in 1 month was statistically significantly higher in the intervention group with added nicotinic acid (p = 0.001; Table 61). In the trial by Vásquez-Garibay and colleagues,82 mean weight gain per day was similar between groups regardless of the addition of NT (no p-value reported). Typical weight gain was said to be five times higher than that of normal infants aged around 8 months.

TABLE 61. Weight gain in children receiving other supplements.

TABLE 61

Weight gain in children receiving other supplements.

Anthropometric measures

Only Vásquez-Garibay and colleagues82 reported on this outcome. The authors state that there was a significant improvement in W/A in both groups from the first week regardless of the addition of NTs, but presented no data. The same trend was reported for W/L (Table 62), but only p-values for within-group differences were reported. The pace of linear growth was said to be double that of normal infants aged around 8 months.

TABLE 62. Anthropometric outcomes in children receiving other supplements.

TABLE 62

Anthropometric outcomes in children receiving other supplements.

Vásquez-Garibay and colleagues82 also included total upper-arm area, upper-arm muscle area, upper-arm fat area and the arm fat index as outcomes, but differences were not significant between groups at week 4 regardless of the addition of NTs, although there were statistically significant within-group differences (see Appendix 12).

Additional outcomes

Manary and Brewster80 reported weight loss as one of their clinical outcomes (although not explicitly stated, this is presumed to be reported as an indication of the resolution of oedema following the start of treatment). There were no statistically significant differences between the treatment arms by day 7 (p = 0.36) or after discharge (p = 0.61) regardless of the added potassium (Table 63), nor were there any significant differences between treatment arms in the number of days children stayed in hospital (p = 0.21). However, the high-potassium intervention group suffered significantly fewer presumed septic episodes (3 vs 18) [OR 8.9 (95% CI 2.2 to 50.9)], respiratory symptoms and new skin ulcerations than the standard-potassium control group, as illustrated in Table 64.

TABLE 63. Additional outcomes in children receiving other supplements.

TABLE 63

Additional outcomes in children receiving other supplements.

TABLE 64. Comorbidities in children receiving other supplements.

TABLE 64

Comorbidities in children receiving other supplements.

Mean urea concentration and mean alkaline phosphatase were statistically significantly lower for the intervention with added NTs (p = 0.009 and p = 0.041, respectively).82 Treatment arms were combined for initial versus final outcome comparisons of white blood cell count, creatinine, glucose, calcium and phosphorus levels, with all showing significant improvements for the whole group apart from changes in white blood cell count.85 Philip and colleagues81 included an outcome for calories consumed for 1 g gain in weight. Although this was lower in the intervention group (14.2 vs 19.3), which had the added nicotinic acid; no p-value was reported.

Summary

  • Three trials investigated supplements other than zinc as part of a treatment regimen for SAM. Each trial examined a different supplement either potassium,80 nicotinic acid81 or NTs.82 The trials took place in different countries and enrolled participants meeting different criteria. Two of the studies81,82 were of weak methodological quality; the third80 was of moderate quality. Only one trial81 conducted an ITT analysis of the data.
  • Only the study80 providing high-dose potassium to the intervention group reported mortality as an outcome. This trial found no difference in early deaths, but found a statistically significant benefit for the intervention group receiving high-dose potassium for late deaths, which meant that there was a 33% reduction in the case fatality rate overall in this group.
  • Intervention effects on weight were reported directly by two studies as mean weight gained per kg of body weight81 or mean weight gained per day.82 One of these studies82 also reported weight gain in terms of body mass index (BMI), alongside other anthropometric measures. Participants receiving nicotinic acid supplementation81 gained statistically significantly more weight per kg in 1 month than participants on the standard diet. In the other study,82 there was no difference in daily weight gain between groups receiving milk-based formula either with or without added NTs, and reporting of anthropometry outcomes was unclear so it was not possible to determine which (if any) p-values indicated a statistically significant difference between the groups.
  • One study80 reported on comorbidities, finding that the high-potassium intervention group suffered significantly fewer presumed septic episodes, respiratory symptoms and new skin ulcerations, whereas there was no statistical difference between the groups for irritability, diarrhoea and oedema.

Ongoing studies

The search for ongoing studies identified 41 records. Of these, 10 appear (from the limited details available) to meet the inclusion criteria for the review. These 10 studies seem to map to questions 7 (two studies), 8 (one study), 10 (four studies), 14 (two studies) and 22 (one study, which may also map to Q8). Summary details of these ongoing studies are presented in Appendix 13.

Image ch2f1
© 2012, Crown Copyright.

Included under terms of UK Non-commercial Government License.

Bookshelf ID: NBK98565

Views

  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (1.6M)

Other titles in this collection

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...