Birungi 2015.

Methods	Cluster‐RCT (randomisation by community unit): NCT00397150 (PROMISE‐EBF trial, Ugandan site), with follow‐up study of infants at 5 years
Participants	765 pregnant women and their fetuses from 24 community clusters were randomised. Inclusion criteria: women residing in a selected cluster, ≥ 6 months pregnant, with no plans to move outside of the cluster within 1 year Exclusion criteria: reduced ability to collaborate for psychological/mental reasons, severely ill, given birth more than 7 days ago, multiple birth, newborn with severe malformation Setting: Mbale district, Eastern Uganda (data collection, including for the follow‐up study, from 2006 to 2011)
Interventions	Group 1 (n = 396 pregnant women from 12 clusters randomised) Women received individual tailored home‐based peer counselling focused on promoting exclusive breastfeeding. The intervention was delivered by workers from the community who were trained in the intervention protocol. The counselling included one prenatal visit followed by four postpartum visits. Group 2 (n = 369 pregnant from 12 clusters randomised) Women received the standard care delivered by public health services. Timing: commenced towards the end of pregnancy, and continued through weeks 1 to 10 after birth (≤ 6 months intervention duration) Theory or model used as a basis for intervention: not reported
Outcomes	Data in meta‐analysis for: primary outcome: caries in primary teeth, dmft index; secondary outcomes: none Narrative text for: none Tabulated data for: none Additional outcomes that had not been prespecified: Child: none reported. Mother: breastfeeding
Notes	Funding: Quote: "This work was supported by European Union Sixth Framework International Cooperation–Developing Countries, Research Council of Norway, Swedish International Development Cooperation Agency, Norwegian Programme for Development, Research and Education, South African National Research Foundation, and Rockefeller Brothers Foundation". Declarations of interest: none declared
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "A total of 24 clusters were stratified into urban and rural and allocated at random (computer generated with an allocation ration 1:1)".
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not reported; unlikely participants and personnel were blinded considering the nature of the intervention assessed
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: “The interviewers and dentists were aware of the children’s involvement in the PROMISE EBF trial but were blinded with respect to their group allocation”.
Incomplete outcome data (attrition bias) All outcomes	High risk	Overall, 417/765 (55%) children of the mothers randomised were available for caries and other outcome assessment at 5 years, of which 215/396 (54.3%) were children of mothers randomised to the intervention group, and 202/369 (54.7%) were children of mothers randomised to the control group. Therefore, very high loss to follow‐up. Though the losses were relatively evenly distributed across the two groups, the authors reported differences in the characteristics of the children in the two groups at the 5‐year assessment (see other bias below).
Selective reporting (reporting bias)	Unclear risk	Prespecified outcomes reported, but caries in infants/children not included as specified outcome in the study protocol.
Other bias	High risk	The authors reported that the intervention and control group child cohorts from the PROMISE‐EBF study, used for the included follow‐up study, "differed substantially with respect to the prevalence of EBF at 24 weeks of infant's age (59% versus 12%). Additionally, they reported significant differences between the groups at the 5‐year follow‐up data collection point (when caries were assessed), in socioeconomic status".