This is a secondary analysis using data from the World Health Organization Calcium Supplementation for the Prevention of Preeclampsia Trial. The trial was approved by the Institutional Review Boards of all participating institutions. Detailed description of the trial is provided elsewhere.4
Briefly, a total of 8,325 normotensive nulliparous women were enrolled before 20 weeks of gestation from clinics serving women with low dietary calcium intake (<600 mg/day) in 6 countries (Argentina, Egypt, India, Peru, South Africa and Vietnam) between 2001 and 2003. Women were randomized to receive either 1.5 g calcium/day supplement or placebo tablets throughout pregnancy. Demographic and clinical information was recorded at enrollment.
Blood pressure was measured at each antenatal care visit with a mercury sphygmomanometer twice (the average was used) at 3-minute intervals, with the subject seated for at least 5 minutes before measurement with the cuff positioned at the level of the heart on the right arm. Great efforts were made on training, certification and retraining of clinic staff every 3 months to obtain the blood pressure measures as accurate as possible.5
There was no difference between the treatment and the control groups at baseline on maternal demographic and clinical characteristics, gestational age at enrollment, mean systolic and diastolic blood pressure at randomization.4
The trial found that calcium supplementation did not significantly reduce the risk of overall preeclampsia. The mean systolic and diastolic blood pressure levels at 30 – 34 weeks were identical at 108 and 64 mmHg, respectively, for the treatment and the control groups (Student t-tests on systolic and diastolic pressure, both p > 0.05). Thus, we assumed that calcium supplementation did not influence mean blood pressure and combined the two groups for the current analysis. However, calcium supplementation was found to mitigate the severity of the disease, including preterm delivery.4
Therefore, we controlled for calcium supplementation in the multivariable analysis as a potential confounder.
Baseline blood pressure was designated as the average pressure recorded between 12 and 19 weeks of gestation and mid 3rd trimester blood pressure as the average of any measurement between 30 and 34 weeks of gestation. Rise (or fall) in blood pressure was calculated as the difference between the mid 3rd trimester and baseline pressures, and pulse pressure, difference between systolic and diastolic measurements.
Gestational age at entry was determined with the “best obstetric estimate”, which included the last menstrual period, uterine size and ultrasound examination if required by the attending physician.4
At each prenatal visit and admission to delivery, date and estimated gestational age were recorded. To ensure accuracy of gestational age, we calculated a gestational age at delivery as follows: Gestational age at enrollment + (delivery date – enrollment date)/7, then took the integer and compared the calculated gestational age at delivery with the recorded gestational age at delivery as completed weeks. Any discrepancy greater than one week resulted in exclusion of the subject due to the possibility that the gestational age was incorrect. Type of labor onset was categorized as spontaneous or medically indicated (induced or prelabor cesarean section).
There were a total of 8,002 nulliparous women with a singleton pregnancy who were normotensive at enrollment and had complete delivery information. 6,547 (82%) women had an accurate gestational age. 9 had postterm deliveries (>42 weeks) and were excluded, as were 1,323 women with any of the following conditions: complaint of any health problems at the first visit, stillbirth in present pregnancy, or delivered a newborn with any congenital malformations. These exclusions left 5,167 subjects for analysis.
Because blood pressure measurements were all normally distributed, descriptive statistics were expressed as mean and standard deviations. Analysis of variance and chi-square test were used for continuous and categorical variables, respectively. We used multiple linear regression and Cox proportional hazard model to control for potential confounders for continuous and time-to-event outcomes, respectively. The potential confounders included maternal age, education, maternal body mass index, gestational age at admission, smoking status, race, gravidity and calcium treatment. Data analysis was done in SAS (Statistical Analysis System version 9.0. Cary, NC).