Bhandari 2004.

Methods	Design: cluster‐RCT Unit of randomisation: communities Intention to treat: yes Adjustment for clustering: yes. Quote: "All results reported are adjusted for cluster randomisation (using the “cluster” option of the “regress” command)" (p 2344)
Participants	Number: 8 communities with 1025 newborn infants (intervention: 552; control: 473) Inclusion criteria: newborns enrolled if they were local residents and informed written consent was obtained Exclusion criteria: not stated Age: newborns enrolled and followed up every 3 months up to the age of 18 months Gender: intervention: 52.2% male, 47.8% female; control: 53.5% male, 46.5% female Ethnicity: not reported Settings: State of Haryana Country: India Attrition: intervention: 117/552 (21.2%); control: 79/473 (16.7%)
Interventions	Intervention (see Table 4 for detailed description): large group education feeding demonstrations Control: treatment as usual (routine services) Duration of each intervention session: not reported
Outcomes	Effect on physical growth (weights and lengths) Complementary feeding practices (effects of the types of food fed to children, responsive feeding, hygiene practices) Prevalence of diarrhoea Not used in this review: prevalence of cough prevalence of fever Time points reported: weights and lengths at 6, 12 and 18 months, and complementary feeding practices at 9 and 18 months
Notes	Study start and end dates: not reported Study duration: 18 months Conflict of interest: not stated Source of funding: "supported by the Department of Child and Adolescent Health and Development, World Health Organization, Geneva, Switzerland" (quote, p 2342)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Comment: sequence was generated using random numbers table (see p 2344)
Allocation concealment (selection bias)	Low risk	Quote: "a statistician, not involved with the study, generated 4 single‐digit random numbers using a random numbers table" (p 2344)
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Comment: not described, probably not done
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Comment: not described, probably not done
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: more than 10% loss (196/1025) but "all analyses were by intention to treat" (quote, p 2344)
Selective reporting (reporting bias)	Unclear risk	Comment: none observed. No protocol available
Other bias	Low risk	Comment: none observed