significance mated that with a sample size of 10,000 women considered to indicate statistical the study would have 90s power to show a 30% However, since the adjustment is minimal, we re- reduction in the rate of the primary outcome, from port 95% confidence intervals. Data from all women were analyzed according 4% in the placebo group to 2.8% in the vitamin p, with a two-sided type I error rate of 5%. to the group to which they were randomly as- An independent data and safety monitoring signed, regardless of whether they took the study committee monitored the trial and reviewed in capsules. Continuous variables were compared tenm results. A group sequential method was used with the use of the Wilcoxon rank-sum test, and to characterize the rate at which the type I error categorical variables with the use of the chi-square was spent the chosen spending function was the test. For all secondary outcomes, nominal p values generalization of the OBrien-Flem of less than 0.05 were considered to indicate sta Lan-DeMets ing boundary, Three interim analyses were per- tistical significance: p values have not been ad. formed. In the final analysis of the primary out justed for multiple comparisons. Unless indicated come, two-tailed P values of less than 0.045 were all analyses that are presented were prespecified.