Villar 2001 (WHO 2001).

Methods	Parallel arm cluster‐RCT conducted in 53 clinics in Argentina, Cuba, Saudi Arabia and Thailand between May 1996 and April 1998.
Participants	Sample size: 53 clusters (24526 individuals). Clusters: clinics serving 300 new patients within 24 months. The clinics had to be part of a public or semi‐public health system and not require direct fee‐for‐services payment. Clinics had to have an ANC system in place with adequate staffing and be able to implement and fund tests or activities required by the protocol. Individuals: all women attending prenatal care for the first time at any participating clinic were eligible. Women later found not to be pregnant were excluded. Multiple births were excluded from some outcomes (specifically low birthweight outcomes). Women had to be traceable at delivery, including women transferred to hospitals as high‐risk.
Interventions	Target: health system (reorganisation of services). Arm 1 (27 clusters, 12,568 women): a reduced visits regime of ANC. Women classified as higher risk received standard ANC but were analysed according to ITT. The new model of care included 4 antenatal visits for low‐risk women. The visits were goal‐oriented and focused on scientifically evaluated components of ANC. Arm 2 (26 clusters, 11,958 women): standard ANC.
Outcomes	Trial primary outcomes: low birthweight (< 2500 g), pre‐eclampsia/eclampsia, severe postpartum anaemia (< 90 g/L Hb), treated urinary tract infection. Review outcomes reported: Primary: maternal mortality. Secondary: ANC coverage (< 5 visits), tetanus protection, syphilis treatment, preterm birth, low birthweight infants, perinatal mortality, neonatal mortality, stillbirth. Other: median no. of ANC visits, pre‐eclampsia, antepartum haemorrhage, mode of delivery, and others. Follow‐up: schedule of antenatal visits and data collection at delivery.
Notes	Funders: UNDP/UNFPA/WHO/World Bank Special Programme of Research, Development and Research Training in Human Reproduction of WHO. Additional support from City of Rosario, Argentina, Ministry of Health, Cuba, National Institute of Public Health, Mexico, The Population Council ‐ Regional Office for Latin America and the Caribbean, Ministry of Health, Saudia Arabia, Swedish Agency of Research Cooperation with Developing Countries, Ministry of Public Health and Faculty of Medicine, Khon Kaen University, Thailand, Department for International Development, UK; Mother Care ‐ John Snow, Inc; National Institute for Child Health and Human Development, National Institutes of Health, USA, and The World Bank.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Sequence was computer‐generated. Randomisation was stratified according to study site and clinic characteristics.
Allocation concealment (selection bias)	Low risk	Allocation kept centrally until each site had completed the basic introductory training of study personnel, which took place in both intervention and control clinics.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	Not blinded.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Blinding not possible. Staff recording outcome data after the birth were not aware of group allocation but outcomes were recorded by staff providing ANC (not blinded).
Recruitment bias (for cluster RCTs)	Low risk	None noted.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Loss of follow‐up varied for delivery outcomes but was reasonably low and balanced across groups (i.e. loss to follow‐up in reduced visits group was 253/12,568 (2.0%) and standard ANC group was 290/11958 (2.4%). For the low birthweight outcome 138/11672 single births were missing for the new model clinics and 81/11121 in the standard care clinics.
Selective reporting (reporting bias)	Low risk	Relevant outcomes were reported.
Analysis bias	Low risk	Analysis appropriate for clusters; ICC reported; ITT analysis was performed.
Other bias	Low risk	There was some evidence of imbalance at baseline. Women in the new model were less likely to smoke during pregnancy (10.4% versus 12.5%) but it was more likely that women in the new model clinics to have lower levels of education (17.5% education less than primary versus 15.7%). The impact of these differences at baseline are not clear and the differences are taken into account in the adjusted analyses.
Overall risk assessment	Low risk	No serious risk of bias concerns.