Inclusion/exclusion criteria

Inclusion criteria were singleton pregnancy at 23 ^0/7 –30 ^6/7 weeks of gestation, mother at least 18 years old, and spontaneous PROM. PROM was defined by (1) a positive placental alpha-1 microglobulin test (AmniSure; QIAGEN, Germantown, MD) from a vaginal swab, (2) vaginal leakage of indigo carmine dye that had been instilled by amniocentesis, or (3) ≥2 of the following findings: a positive nitrazine test of a vaginal swab; ferning observed on a microscope slide of vaginal fluid; gross pooling of clear fluid in posterior vaginal fornix on speculum examination; and/or ultrasound examination that showed oligohydramnios.

Exclusion criteria included any of the following conditions: contraindications to expectant management (such as suspected intraamniotic infection or inflammation, active preterm labor, nonreassuring fetal heart rate tracing, intrauterine fetal death, preeclampsia, active uterine bleeding, documented fetal lung maturity (FLM), or other condition that required immediate delivery); fetal conditions likely to cause serious neonatal morbidity (such as malformations that involved vital organs or likely to require surgical repair, fetal viral infection, hydrops); cervical cerclage present at the time of PROM; medical conditions that had been treated with systemic steroids; and contraindications to 17OHP-C (such as allergy to drug or vehicle, current or past thromboembolic disorder, current or past hormone-sensitive cancer, undiagnosed vaginal bleeding, cholestatic jaundice of pregnancy, or other active liver disease, uncontrolled hypertension).

Patients with PROM who were excluded initially because they were having contractions or bleeding could become eligible later if their contractions or bleeding subsided. We did not discriminate whether PROM had occurred before the onset of contractions (prelabor PROM, premature PROM) or after. This was based partly on the belief that the distinction between spontaneous PROM after preterm labor vs prelabor PROM is somewhat arbitrary, because the 2 may merely be different manifestations of a common “spontaneous parturition syndrome,” sharing common risk factors, antecedents, and covariates. It also was based partly on the belief that most of the adverse obstetrics events that occur in a quiescent patient after PROM are attributable directly to the consequences of PROM, such as oligohydramnios (which leads to cord compression or abruption) or loss of barrier function (which leads to ascending infection or cord prolapse), or attributable to the underlying factors that cause PROM in the first place (such as infection or inflammation that leads to labor or chorioamnionitis or cervical insufficiency that leads to silent dilation and early preterm birth), regardless of whether PROM occurred before or after the onset of contractions. Finally, it was based partly on the observation that the clinical treatment of patients with threatened preterm parturition typically is guided more by whether the membranes are intact than by which came first, contractions or PROM.

Patients who had been on treatment for the prevention of preterm birth with the use of 17OHP-C or vaginal progesterone at the time of PROM were not excluded. We recommended that vaginal progesterone be discontinued after PROM because of the risk of ascending infection. A patient on 17OHP-C treatment was eligible if she was willing to stop the previous treatment and be assigned randomly for the trial. If she was not willing, she was continued on 17OHP-C and not enrolled in the trial.

Procedures

Eligible patients who gave written, informed consent were allocated randomly to receive a weekly 1-mL intramuscular injection of either 17OHP-C (Makena, 250 mg; Lumara Health, Chesterfield, MO) or an identical-appearing placebo (castor oil vehicle). Randomization followed a computer-generated block scheme, with 1:1 group allocation, stratified by site and by gestational age (23 ^0/7 –25 ^6/7 , 26 ^0/7 –28 ^6/7 , or 29 ^0/7 –30 ^6/7 weeks). Study medications were dispensed by the hospital pharmacy in syringes that appeared identical for 17OHP-C and placebo. Subjects, research personnel, and caregivers had no knowledge of the group assignment, either before or after randomization. The randomization code was not broken until the database had been finalized for analysis. Study medication was continued weekly until 34 ^0/7 weeks of gestation, unless delivery occurred earlier.

Expectant management of PROM was attempted according to the principles that the investigators had agreed on in advance. All patients were admitted to the hospital until delivery. Prophylactic antibiotics were given. Each site had a “usual” antibiotic regimen similar to that of Mercer et al ; however, the investigators could not agree on a single regimen for all sites or regimens to cover all contingencies, such as allergies. A single course of antenatal corticosteroids was given, usually betamethasone 12 mg intramuscularly (2 doses, 24 hours apart). A “rescue course” of corticosteroids was used at some sites for patients undelivered after 2 weeks, although the protocol discouraged this because of a lack of data to support efficacy of repeat steroid dosing in women with PROM and because of a concern that repeat dosing might increase the rate of chorioamnionitis. A third course of corticosteroids was not permitted. Tocolytic therapy was permitted at the caregiver’s discretion and according to local custom during the first 48 hours after corticosteroid dosing, but not thereafter. Magnesium sulfate for prevention of cerebral palsy, when used, followed each hospital’s protocol.

Amniocentesis to rule out intraamniotic infection was encouraged. If the amniocentesis was performed before randomization, then randomization was postponed until preliminary results indicated low probability of infection. Fetal heart rate monitoring was performed at least daily. Caregivers were advised to avoid digital examination of the cervix before the onset of labor. Amniotic fluid for FLM testing was collected from vaginal secretions or during amniocentesis, if possible, at 32 ^0/7 –33 ^6/7 weeks of gestation. The type of FLM test and the criteria for diagnosing maturity were based on local custom. Peripartum and perioperative antibiotic prophylaxis were encouraged and followed local custom. If there was no fluid leakage for >72 hours, diagnosis of membrane resealing could be entertained after a trial of ambulation if ultrasound examination showed normal amniotic fluid volume and if at least 1 confirmatory test showed no evidence of amniotic fluid leakage (either placental alpha-1 microglobulin test, fetal fibronectin, indigo carmine leakage, or absence of pooling on examination with negative fern and nitrazine tests). On diagnosis of resealing, study medication was discontinued, and the patient was discharged from the hospital.

Data and statistics

Case report forms were completed based on the maternal medical record and the neonatal record up to the first 60 days of life. Infant outcomes after the initial hospitalization or after 60 days of life were not captured. Data were entered into a centralized computer database via an online data entry system. Data entries were audited to compare their accuracy against source documents in the medical record. Data queries that were generated from the audits and from quality checks were resolved before the randomization code was broken.

The primary outcome was the continuation of pregnancy either until 34 ^0/7 or until 32 ^0/7 –33 ^6/7 weeks of gestation with documentation of FLM testing, which are endpoints that are widely accepted as indications to proceed with delivery rather than continue expectant management. The 2 prespecified secondary outcomes were (1) the interval from randomization to delivery and (2) composite adverse perinatal outcome , which was defined as ≥1 of the following diagnoses: stillbirth, neonatal death, infant death, respiratory distress syndrome, intracranial hemorrhage (grade 3 or 4), necrotizing enterocolitis (stage 2 or 3), and/or culture-proven neonatal sepsis within 72 hours of birth. Each component of the composite was defined as per Mercer et al. Periventricular leukomalacia was not included in the definitions of Mercer et al, but we included it in the composite outcome if the diagnosis was made on ultrasound scanning or magnetic resonance imaging within 96 hours of birth; the case report form did not distinguish cystic vs noncystic subtypes.

For statistical analysis, comparisons between the 2 treatment groups on categoric variables were made using Fisher exact test or Cochran-Mantel-Haenszel χ ² test stratified by gestational age at randomization. Wilcoxon rank sum test was used for continuous variables. The randomization-to-delivery interval was tested using Cox regression with adjustment for gestational age at randomization. Analyses were based on the intent-to-treat principle, that is, subjects were analyzed with their allocated group regardless of whether they withdrew or discontinued treatment, with the exception of 2 subjects (1 from each treatment group) who were lost to follow-up evaluation and had no available maternal or neonatal outcome data. Two-sided probability values were reported; a probability value of < .049 was considered significant. All analyses were performed with SAS software (version 9.2; SAS Institute, Cary, NC).

The study sample size was planned to detect an increase in the primary outcome from 30% in the placebo group to 50% in the 17OHP-C group. The anticipated rate in the placebo group was based on a query of the national neonatal database of Pediatrix division of Mednax, Inc, our parent company. We calculated that 105 subjects per group (210 total) would yield 80% power to detect this difference with the use of a 2-tailed alpha at .049, adjusted for 1 planned interim analysis based on the O’Brien-Fleming alpha spending function. The sample size was adjusted upward to 111 per group (222 total) to allow for up to 5% loss-to-follow-up. We noted a priori that this sample size would also yield 80% power to detect a 4-day difference in randomization-to-delivery interval between the groups (assuming a standard deviation of 10 days) and 80% power to detect a reduction in composite adverse perinatal outcome from 60% in the placebo group to 40% in the 17OHP-C group.

Interim analysis

A planned interim analysis was conducted on completion of 50% of the case reports. The plan was to stop the trial if there was a significant difference between groups in the primary outcome (with a nominal alpha of .0031 based on the Lan and DeMets group sequential method and O’Brien-Fleming alpha spending function), if there was a safety signal that suggested an increased risk of adverse events in the 17OHP-C group or if analysis of the observed trends indicated a low conditional power to show a significant difference between groups in the primary outcome, unless mitigated by favorable trends in the 2 prespecified secondary outcomes.

The results of the analysis suggested that it was futile to continue the trial. Based on the observed trend, the conditional power to show a between-groups difference in the primary outcome was <10%, and no favorable trend was noted in the secondary outcomes. With these findings, the Data Safety and Monitoring Board recommended discontinuation of the trial. All further enrollment was stopped, but patients who were already enrolled continued to receive weekly study medication until delivery. This report includes the results from all the enrolled subjects.