|Methods||Cluster RCT conducted in 8 villages in an area of rural Malawi with a high incidence of kwashiorkor. Unit of randomisation was the household. Trial was conducted for 5 months between November 2003 and March 2004, and follow up was for 20 weeks.|
Computer generated randomisation.
Investigators and caregivers were blinded to allocation.
20 week follow up.
|Participants||2372 healthy children aged between 12 and 48 months from 2156 Households. Mean ages were 28.2 and 28.5 months in intervention and control groups respectively. Indices of baseline nutrition status were mean and standard deviation of the weight for age in z-scores of the participants (-1.4 and 1.0 respectively in both groups); and the mean and standard deviation of the circumference of middle upper arm in centimetres (14.8 and 1.3 respectively in both groups).|
Exclusion criteria were the presence of severe chronic illness and oedema at the time of enrolment.
|Interventions||INTERVENTION GROUP: The intervention consisted of daily administration of a citrus flavoured powder containing 1.8 mg of riboflavin, 23 mg of vitamin E (alpha tocopherol), 55 mcg of selenium and 300 mg of N acetyl cysteine. Intervention was administered at home by the caregiver.|
CONTROL GROUP: The placebo consisted of a similar tasting powder with no active ingredient. The children in both groups were to drink the powder daily for 20 weeks which was the entire trial duration.
Follow up was conducted at two-weekly intervals in the community by a study team comprising two investigators and two study nurses.
|Outcomes||Participants were assessed for the development of oedema by compressing the skin over both of the child’s fourth metatarsal bones with a firm finger for 10 seconds; and looking for evidence of pitting. Development of oedema was the primary outcome measure of the trial. Secondary outcomes were:|
number of deaths due to any cause;
rate of change in weight;
gain in circumference of the mid upper arm and length of the arm;
number of days of fever, cough and diarrhoea.
|Risk of bias|
|Adequate sequence generation?||Yes||Allocation sequence for the included trial was generated by a computerised random number generator blocked in groups of 200. These numbers were noted on the cards with which children were randomised. Siblings received data cards with the same letter of allocation as the unit was the household and not the individual child.|
|Allocation concealment?||Unclear||It was clearly stated that all investigators and the caretakers of the participants were blind as to which letters (of allocation) were placebo and which contained antioxidants. The authors do not explicitly state the method used to conceal the allocation.|
|Yes||All the investigators (who assessed the outcomes) and child caretakers were blinded as to the content of the allocation (either placebo or active powder) each child had received.|
|Incomplete outcome data addressed? |
|Yes||40 children were reported to have been lost to follow up. The main reasons for this, where it was known, were family moving from trial location, child not liking the taste of the intervention and carer unable to bring the child for follow up. The reasons were similar in both groups and the number of losses were similar in both groups, both in terms of the clusters and individual children. The authors reported performing an intention-to-treat analysis of the data. They reported outcomes for all the randomised participants in the groups to which they were randomised and measured outcome data in all participants up to the end of the study for most of the participants and up to the last follow up for those who died or were lost to follow up.|
|Free of selective reporting?||Unclear||We were not able to compare the study report with the protocol but have no reason to believe the trial report is biased by selective outcome reporting after an appraisal of the study report. However, because we are not able to exhaustively investigate this, we would rate this as an unclear risk of bias.|
|Free of other bias?||Yes||We identified no particular issues that could bias a cluster RCT namely recruitment bias, baseline imbalance, loss of a whole cluster and statistical analysis allowing for the cluster design.|