Engstrom 2008 (C).

Methods	Cluster‐randomised trial. 3 arm design in which health facilities were the unit of randomisation.
Participants	391 children, both sexes (184 females (47%)), 6 months old. Study carried out through primary healthcare units in Rio de Janeiro, Brazil. 15 health care centres (6 intervention, 9 control). Inclusion criteria: absence of iron supplementation in the month preceding recruitment and negative for sickle cell anaemia. Baseline prevalence of anaemia (taken from the control group): 60.4%. Socioeconomic status:approximately 30% of the mothers worked outside the home; most families (> 90%) had access to radio and television, but < 20% had access to a car.
Interventions	Health facilities were allocated to one of the following groups: Group 1 (n = 188): children received weekly supplementation with 25 mg of elemental iron (as oral ferrous sulphate) per week in syrup and education on anaemia and diet; Group 2 (n = 188): children received daily supplements containing 12.5 mg elemental iron dailyand education on anaemia and diet; Group 3 (n = 94): children received received no intervention and was recruited retrospectively. Length of the intervention: 24 weeks. For the purposes of this review we only compared groups 1 and 2.
Outcomes	Haemoglobin, anaemia (Hb <110 g/L) and adherence.
Notes	Analyses were performed taking into account cluster sampling. Non‐malaria area.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Healthcare units were randomly selected. Method of sequence generation not described.
Allocation concealment (selection bias)	Low risk	Not reported. Since the intervention was allocated at health care unit level, it is unlikely there was a selection bias at the individual level.
Blinding (performance bias and detection bias) All outcomes	High risk	Participants: Mothers were aware of supplements Personnel: Clinic staff were aware of supplements Outcome assessors: Unlikely Control group identified retrospectively so they were not aware of trial during treatment phase.
Incomplete outcome data (attrition bias) All outcomes	High risk	38/188 (20.2%) were lost to follow up the daily group and 41/188 (21.8%) in the weekly group.
Selective reporting (reporting bias)	Unclear risk	There is insufficient information to permit judgement.
Other bias	Unclear risk	Baseline characteristics were similar for most variables. Regression analysis was carried out to identify possible confounders and where possible confounders accounted for at least 10% of variation they were entered into the final model. However for anaemia no confounders were maintained in the final regression analysis.