Liu 2012.

Methods	3‐arm double‐blind randomised controlled trial, individually‐randomised. Randomisation was stratified by county, and random block sizes of 3, 6, and 9 were used to ensure geographical balance with an approximately equal distribution of treatments within and across study counties.
Participants	18775 nulliparous pregnant women 20 years of age or older, with mild or no anaemia (Hb level greater than 100 g/L),with no more than 20 weeks of gestation attending prenatal care in 5 rural counties in Hebei Province, northern China, where basic health services were provided through 3‐tier (county, township, and village) healthcare networks of northern China, from May 2006 through April 2009. Additionally, eligible women had not consumed micronutrient supplements other than folic acid in the prior 6 months. Women were followed monthly from early pregnancy through delivery and at 4–8 weeks postpartum. Their infants were followed monthly from birth until 1 year of age.
Interventions	Participants were randomly assigned to 1 of 3 groups: group 1 (n = 6261) received 400 µg (0.4 mg) folic acid daily (control); group 2 (n = 6252) received 30 mg elemental iron (as ferrous fumarate) plus 400 µg (0.4 mg) folic acid daily; group 3 (n = 6262) received 30 mg elemental iron (as ferrous fumarate) with 400 µg (0.4 mg) folic acid daily and 800 μg vitamin A, 10 mg vitamin E, 5 μg vitamin D, 70 mg vitamin C, 1.4 mg thiamine, 1.4 mg riboflavin, 1.9 mg vitamin B6, 2.6 μg vitamin B12, 18 mg niacin, 15 mg zinc, 2 mg copper, 150 μg iodine, and 65 μg selenium. The supplements were provided before 20 weeks of gestation to delivery. Each woman received 2 bottles of supplements at enrolment and 1 at monthly follow‐up visits. Each bottle contained 31 supplements, including the type of the supplements per group according to lot number. Only groups 1 and 2 are considered in this review (folic acid vs iron + folic acid). Health worker cadre: trained county or township physicians completed relevant measurements and collected data based on a perinatal and child healthcare surveillance system. At enrolment, the physician assigned women the next lot number on the randomisation schedule and provdied the supplements.
Outcomes	Perinatal mortality, neonatal deaths, infant deaths, maternal Hb concentration and anaemia at 24 to 28 weeks of gestation, birthweight, birth length, duration of gestation, preterm delivery, compliance.
Notes	By gestational age: early, if supplementation started before 20 weeks' gestation; by anaemic status at start of intervention: non‐anaemic; by dose of iron: low daily dose of iron if 30 mg or less of elemental iron; by type of formulation: normal release iron supplement/not specified; by iron compound bioavailability in comparison to ferrous sulphate: equivalent bioavailability: ferrous fumarate; by malaria risk setting: not reported. It is presumed to be malaria‐free.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"A statistician external to the study randomly assigned ten 4‐digit lot numbers to each of the 3 supplement types (masked to the formulation and allocation) and generated the assignment list for each county proportional to the expected number of participants; within each county and block, lot numbers were randomly assigned using RANUNI in SAS statistics software (SAS Institute Inc)."
Allocation concealment (selection bias)	Low risk	Ten 4‐digit lot numbers to each of the 3 supplement types (masked to the formulation and allocation). At enrolment, the physician assigned women the next lot number on the randomisation schedule.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	"Aside from a pharmaceutical engineer who ensured allocation of lot numbers to the correct supplement formulations, all others (ie, participants, local physicians, study personnel, and investigators) were masked to the identity of the supplements. Treatment codes were broken after completion of the study and main analyses."
Blinding of outcome assessment (detection bias) All outcomes	Low risk	"Aside from a pharmaceutical engineer who ensured allocation of lot numbers to the correct supplement formulations, all others (ie, participants, local physicians, study personnel, and investigators) were masked to the identity of the supplements. Treatment codes were broken after completion of the study and main analyses."
Incomplete outcome data (attrition bias) All outcomes	Low risk	There were 299/6261 (4.77%) losses to follow‐up in the group 1; 280/6252 (4.47%) in group 2: and 299/6262 (4.77%) in the group 3 for various reasons: permanently moved, induced abortions, spontaneous abortions, dropped out or maternal death. The attrition was balanced among groups.
Selective reporting (reporting bias)	Low risk	There appears to be no selective reporting. Trial Registration: clinicaltrials.gov Identifier: NCT00133744.
Other bias	Low risk	No other bias apparent.