CPEP 1997.

Methods	Numbered treatment packs in computer‐generated simple randomisation sequence. Loss to follow‐up: calcium 132/2295 vs placebo 121/2294.
Participants	Setting: medical centres in the USA. Dates of recruitment not clear. Pregnant nulliparas (45% black, 35% non‐Hispanic white, 17% Hispanic white). Passed compliance test (took 75% of placebo over 6‐14 days); BP 134/84 mmHg or less; urine protein dipstick negative or trace; 13‐21 weeks' pregnant. Exclusion criteria: taking medications; obstetric or pre‐existing diseases or personal characteristics which could influence study end points, absorption or metabolism of calcium; any risk associated with calcium supplementation, or compliance; elevated serum creatinine (1.0 mg per dL or more) or calcium (10.6 mg per dL or more); renal disease; haematuria; history or family history of urolithiasis; frequent use of calcium supplements or antacids. Of 11,959 women screened, 5703 were excluded initially and a further 1667 after the compliance test. The remaining 4589 women were enrolled.
Interventions	2 g/day elemental calcium as calcium carbonate, or placebo. Taken until delivery, development of pre‐eclampsia or suspicion of urolithiasis. All women took 50 mg calcium per day as normal supplementation and were asked to drink 6 glasses of water per day. Compliance was 64% in the calcium group and 67% in the placebo group. 20% of women took > 90% of the allocated treatment.
Outcomes	Gestational hypertension (DBP sitting, fifth Korotkoff sound unless zero, 90 mmHg or more on 2 occasions, 4 hours to 1 week apart); severe gestational hypertension (DBP 110 mmHg twice or treated, or complications); proteinuria (300 mg/24 hours or more, 1+ on 2 occasions 4 hours to 1 week apart, 2+ or more, or protein/creatinine ratio 0.35 or more); pre‐eclampsia (gestational hypertension + proteinuria within 7 days of each other); severe pre‐eclampsia (50/2163 vs 59/2173); renal insufficiency (21/2163 vs 23/2173); urolithiasis (1/2163 vs 3/2173); prematurity (< 37 weeks); baby small‐for‐gestational age (124/2163 vs 105/2173); perinatal death. A limited follow‐up of mothers and infants found to have reduced SBP at 2 years of age in the calcium supplementation group (mean 95.4 mmHg, SD 7.6, n = 35 vs 100.2, 7.9, n = 18). The data have not been included in this review because of the low and unequal follow‐up rate (35 and 18 from 497 invited to participate) limit the reliability of the results.
Notes	Multicentre trial, 5 US university centres. Maternal outcomes reported as percentages of the whole number enrolled. In this review, denominators of 2163 (calcium) and 2173 (placebo) have been used. Neonatal outcomes in the report are based on live births (2134 and 2139). Addition of abortions and fetal deaths brings these numbers to 2156 and 2166. It is not clear why a discrepancy in numbers remains. Sources of funding: supported by The National Institute of Child Health and Human Development and the National Heart, Lung and Blood Institute. Study medications were industry provided. CoI: not stated.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Packages of study tablets were prepared and numbered by pharmaceutical manufacturer according to a computer‐generated simple randomisation sequence.
Allocation concealment (selection bias)	Low risk	On enrolment, each woman was assigned the next numbered package of medication at 1 of 5 centres. The blister‐packed tablets were identical in appearance.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Double‐blind. The code was held centrally.
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Double‐blind. The code was held centrally.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Attrition < 10%
Selective reporting (reporting bias)	Unclear risk	Authors used total number of women enrolled to each group as denominator instead of total number minus those lost to follow‐up. Also small discrepancy in overall numbers but unlikely to affect results substantially.
Other bias	Low risk	Baseline characteristics similar