Pregnant women presenting for vaginal delivery in two tertiary Egyptian hospitals were screened for participation in this open-label, three-arm, parallel, randomized controlled trial. Approval was obtained from the research ethics committees of both hospitals: El Galaa Teaching Hospital in Cairo (the largest maternity hospital in Cairo), and Shatby Maternity Hospital in Alexandria (the university hospital of Alexandria University), where all three routes of oxytocin administration were routinely used.
Women were eligible to participate if they delivered a live birth vaginally, did not receive pre-delivery uterotonics to induce or augment labor, and were able to provide informed consent. Written consent was obtained after admission, upon arrival to the labor ward. Blood pressure and pre-delivery hemoglobin were subsequently measured and recorded, the latter using HemoCue® hb 201+ (HemoCue, Ängelholm, Sweden).
Women were randomized to receive 10 IU of oxytocin by IM injection, IV infusion, or IV bolus immediately after delivery of the baby. IM injection was usually administered in the thigh. For IV infusion, oxytocin was mixed in 500 ml of fluid and administered through gravity-driven infusion with the roller clamp fully open, most often using an 18 gauge needle. IV bolus was pushed directly into the IV port over approximately 1 min.
Information on other prophylactic measures provided in the third stage of labor, including controlled cord traction and uterine massage, was recorded on standardized data collection forms. Postpartum blood pressure and any side effects or adverse events experienced after oxytocin administration were also recorded. Postpartum blood loss was measured at 1 h post-delivery using a plastic blood collection drape funneled into a calibrated container. For women diagnosed with PPH, blood loss was also recorded at the time of PPH diagnosis and at active bleeding cessation. Women diagnosed with PPH received standard of care treatment at each hospital. Interventions, including administration of additional uterotonics or blood transfusion, were documented. Postpartum hemoglobin was measured at least 24 h after delivery and at least 12 h after removal of the IV for women receiving IV fluids, if possible, or just before discharge if women were discharged sooner.
Our primary outcomes were mean blood loss and proportion of women with blood loss ≥500 ml. Secondary outcomes included proportion of women with blood loss ≥350 ml and ≥ 1000 ml, change in pre- to post-delivery hemoglobin, time to placental delivery, administration of additional oxytocin or other uterotonics, and observed side effects within 1 h postpartum.
The sample size calculation was derived from the expected rate of women with blood loss ≥500 ml in the two comparisons of this three-arm study: IM injection vs. IV infusion and IM injection vs. IV bolus. Based on previous studies, we expected a slightly larger difference of blood loss outcomes for the IV bolus vs. IM injection comparison, thus a smaller sample size was required for that comparison [23, 29]. We augmented sample sizes to compensate for conducting two 80% correlated comparisons (equivalent to requiring a significance level of 0.0435 for each test) and to account for a 2% attrition rate. The resulting sample size requirement was 4900 women, at a 3:3:1 ratio (2100 in each of the IM injection and IV infusion groups and 700 in the IV bolus group), with 80% power for the comparison of IM injection to IV infusion and 85% power for the comparison of IM injection to IV bolus administration. The sample size was also sufficient to detect a 50 ml mean difference in blood loss between study groups.
The simple randomization code was computer-generated in blocks of seven at Gynuity Health Projects in New York, and each assignment was contained in a sequentially numbered, sealed, opaque envelope. Each hospital was independently randomized. Hospital study staff had no access to the randomization code and were instructed to open the next envelope prior to the woman’s delivery, during the second stage of labor.
Analysis was done using the intent to treat approach. P values for baseline characteristics were calculated using the chi-square test of association for categorical variables and one-way analysis of variance (ANOVA) for continuous variables. Differences were considered significant at α = 0.0435, to account for the multiple comparisons made in this three-arm study. Log-binomial regression was used to calculate relative risks (RRs) and associated 95% confidence intervals (CIs) for categorical outcomes. Linear regression was used to calculate regression coefficients and associated 95% CIs for continuous outcomes. We first assessed the assumption of normal distribution all continuous secondary outcomes (including postpartum blood loss, time to placental delivery in minutes, total blood loss, and change in pre-to post-delivery hemoglobin). None were normally distributed, thus transformation (using the natural log ln) was done on all continuous outcomes. To facilitate interpretation of estimates obtained from linear regression of these log-transformed outcomes, we used the following formula to produce an estimate of the percent change in the mean outcome (y) associate with treatment group in question (d): y = 100·[exp(βd) − 1], where β is equal to the regression coefficient for the log-transformed outcome. Analyses were conducted using Stata 12 (StataCorp. 2011. Stata Statistical Software: Release 12. College Station, TX: StataCorp LP).