Indonesia 1999.

Methods	A double‐blinded, randomised placebo‐controlled trial. A factorial design 4‐arm (Zibuvita study) trial and then included a follow‐up study of the infant (the Zinak and Pronak study).
Participants	5736 women who live in Purworejo district of central Java were identified as pregnant from the Indonesian Ministry of Health surveillance. Of these pregnant women, 2173 women at a gestational age of less than (120 days) 17 weeks were eligible for the trial study. After losses to follow‐up, only 2098 delivery and 1956 neonates were analysed in the Zinak and Pronak study. The follow‐up of the children were from birth up to 2 years of age.
Interventions	Women were randomly allocated to 3 treatment groups and placebo group. The treatment groups were given micronutrient capsules from the date of inclusion in the study until delivery. The capsule contained either 2400 RE of vitamin A (as retinyl palmitate) or 20 mg of ZnSO4, or the same dose of both vitamin A and ZnSO4, or placebo. All capsules also contained 2 mg of DL‐a‐tocopherol as an antioxidant and 350 mg of soya bean oil, 20 mg of beeswax and 8 mg of lecithin as capsule filler. The supplements were manufactured by Tishcon Corp. (Westbury, NY, USA) and they were packaged in plastic strips in identical, opaque pink capsules that was sufficient supplements for 2 weeks or 1 month. Fieldworkers distributed capsules and monitored compliance at the home of the women by counting the unused capsule. Vitamin A (2400RE): n = 484/527 (91.8%). Zinc (20 mg): n = 477/531 (89.8%). Vitamin A (2400RE) + zinc (20 mg): n = 495/543 (91.2%). Placebo: n = 500/523 (95.6%).
Outcomes	Maternal outcome Pregnancy weight. Neonatal outcomes Birthweight; low birthweight; stillbirth/neonatal death; blue/floppy (neonatal hypoxia); fever/not drinking; umbilical infection; 6‐month Z‐scores; 6‐month haemoglobin, plasma retinol, plasma zinc; birth size (weight and length); small‐for‐gestational age.
Notes	Adherence: mean adherence ranged from 71% to 73% across the 4 arms of the study.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Pseudo‐random number generator in blocks of 12.
Allocation concealment (selection bias)	Low risk	Treatment allocation sequence was prepared and held at a remote site.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	All investigators, field and laboratory staff and participants were blinded to the treatment code.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Not reported but likely to have been done.
Incomplete outcome data (attrition bias) All outcomes	Low risk	519 of the 1008 women had pregnancies ending between 1 April and 31 October 1997; data available for 503/519 (97%) of these women.
Selective reporting (reporting bias)	Unclear risk	The protocol was not available and there was not enough information to make this judgement.
Other bias	Low risk	No apparent risk of other bias.