Subjects

We enrolled women seeking prenatal care who were 12 ^0/7 –19 ^6/7 weeks pregnant and diagnosed with nonproteinuric chronic hypertension or a history of preeclampsia in their most recent pregnancy that progressed >20 weeks’ gestation. Chronic hypertension was defined as the presence of hypertension prior to pregnancy or <20 weeks’ gestation. For women receiving antihypertensive medications at time of enrollment, the presence of sustained hypertension <20 weeks’ gestation was not mandated. A diagnosis of prior preeclampsia was based on review of the medical record for the previous pregnancy by the research team. In the absence of medical records, the diagnosis was according to the best judgment of the research team. This required that the team confirm that the patients’ reported events of their pregnancies are highly consistent with a diagnosis of prior preeclampsia. Exclusion criteria were planned delivery elsewhere, multifetal gestation, allergy to vitamin C or vitamin E, requirement for aspirin or anticoagulant medication, 24-hour urinary protein ≥300 mg, prepregnancy diabetes mellitus, known fetal anomaly incompatible with life, or prior participation in the study.

Women were assigned randomly to receive daily vitamin C 1000 mg and vitamin E 400 IU or placebo. Each active treatment gel cap contained 500 mg of ascorbic acid, 100 IU of d-alpha tocopherol, 100 IU of d-alpha tocopherol acetate, and excipients (gelatin, soybean oil, glycerin, water lecithin, and caramel color). The placebo gel caps contained excipients only and were externally identical to the active drug. Participants were instructed to ingest 2 gel caps daily from enrollment until delivery or until the diagnosis of preeclampsia. Study participants were discouraged from the use of antioxidant vitamins, calcium supplements, and chronic use of aspirin. The women were followed up at routine prenatal visits, typically every 4 weeks until 26-28 weeks of gestation, every 2-3 weeks until 36 weeks of gestation, and then weekly until delivery or the onset of preeclampsia.

The primary outcome of the trial was the development of preeclampsia. For women with chronic hypertension, (superimposed) preeclampsia was defined by onset of proteinuria (either 300 mg/24 hour or ≥2+ by dipstick) or thrombocytopenia (platelet count <100 × 10 ³ /mm ³ ), elevated liver enzymes (aspartate aminotransferase or alanine aminotransferase >70 IU/L), or pulmonary edema. Women were followed up through the 14th day postpartum for the occurrence of preeclampsia.

In this analysis we compared the rates of superimposed preeclampsia and other adverse outcomes such as perinatal deaths, abruptio placentae, preterm delivery at <37 weeks and <34 weeks, small for gestational age (SGA), and neonatal respiratory distress syndrome, between women with and without a history of preeclampsia. We also analyzed other known risk factors that may influence the rate of superimposed preeclampsia such as maternal age, duration of hypertension, use of antihypertensive medications, BMI, systolic and diastolic blood pressures at enrollment, and smoking during pregnancy. SGA was defined as a birth weight >10th percentile according to the growth tables of Alexander et al. Abruptio placentae was diagnosed according to clinical findings and/or placental examination.

Because no differences were found in the rates of superimposed preeclampsia or other adverse outcomes between the treatment groups in the trial, the data were pooled across treatment groups for this analysis. Parallel analyses adjusting for treatment group yielded virtually identical results and are not reported. Assuming a rate of superimposed preeclampsia of 15% in those without prior preeclampsia, and 25% in those with prior preeclampsia with a 2-sided α = 0.05, our sample size had a power of 61% to detect a difference in the rate of superimposed preeclampsia. Statistical comparisons were stratified by site using Cochran-Mantel-Haenszel general association χ ² statistics for categorical variables and analysis of variance for continuous variables. Analyses that adjusted for additional covariates used logistic regression models. Data were analyzed using software (SAS/STAT, version 9.1.3; SAS Institute Inc, Cary, NC). Exact Cochran-Mantel-Haenszel procedures in software (StatXact, version 8; Cytel Inc, Cambridge, MA) were used when binary outcomes were sparse, yielding odds ratios instead of risk ratios. Analyses were performed using 2-sided α = 0.05 without adjustment for multiple comparisons.