We conducted a large randomized study (the PPROMEXIL trial) to compare IoL and EM in women with PPROM between 34 and 37 wk of gestational age. Because of the conservative treatment policy and conservative preferences amongst patients in The Netherlands, we had an ideal population in which to perform this trial, with extensive data on all eligible patients including non-participants (31%), the vast majority of whom declined participation because they preferred EM.
We found that in pregnancies complicated by PPROM between 34 and 37 wk of gestation, IoL does not substantially reduce the incidence of neonatal sepsis compared to EM. The number of neonates with respiratory distress was comparable in both arms, and IoL did not increase the risk of a cesarean section, findings that were confirmed in meta-analysis. However, in our study IoL increased the risk of hypoglycemia and hyperbilirubinemia, as well as the use of epidural or spinal analgesia during labor.
Our findings are in line with the results of the TERMPROM trial, which compared IoL with EM in 5,041 women with prelabor rupture of membranes at term 
. The TERMPROM trial showed that IoL did not reduce the risk of neonatal sepsis as compared to EM (2.5% versus 2.8%) 
In contrast to earlier studies 
, our pragmatic protocol did not include routine cultures from all neonates to diagnose sepsis. Because of the lack of consensus amongst Dutch neonatologists on whether to take blood samples routinely after prolonged premature rupture of membranes, neonatal blood samples and liquor cultures were taken only on clinical indication. All cases with any possible sign of neonatal sepsis were adjudicated by a panel of neonatologists. In consensus they decided whether or not a newborn had suffered neonatal sepsis (suspected or proven). Despite of the lack of blood culture from all neonates in the trial, we believe that no cases of neonatal sepsis were missed and that the incidence of neonatal sepsis was not overestimated.
IoL reduced the risk of chorioamnionitis. Several studies have demonstrated a relationship between chorioamnionitis and adverse neonatal outcome 
. In a large study of very premature neonates (<28 wk; the ELGAN study) 
, a relationship between cerebral palsy and/or white matter damage and positive bacteriological cultures from the placenta was demonstrated. Other studies have also described a relationship between chorioamnionitis and increased risk for sepsis, respiratory distress, pneumonia, and even neonatal death 
. We doubt, however, that these findings can be extrapolated to our population. The incidence of cerebral palsy is significantly lower in the near-term-birth population, and reported incidences of adverse neonatal outcome in a near-term and term newborns are low (maximum reported incidences of 1.9% in the chorioamnionitis group) 
In line with the TERMPROM trial, the number of cesarean sections in our study was comparable in the IoL and EM groups 
. We could not confirm the trend for increased risk on cesarean section in the EM group that was reported in the previous systematic review 
The risks of hypoglycemia and hyperbilirubinemia were decreased in children of women treated expectantly. These findings have, to our knowledge, not previously been reported for a prospective study. In a recent retrospective study in three tertiary hospitals in France, a similar incidence of hyperbilirubinemia (37% for IoL versus 33% for EM) and a slightly lower incidence of hypoglycemia (12% for IoL versus 6% for EM) was found, but due to a lack of power, differences were not statistically significant in the French study 
We doubt whether prolongation of pregnancy, which was on average 3.3 d longer in the EM group in our trial, will have solely contributed to the decreased risk of hypoglycemia and hyperbilirubinemia. Maybe spontaneous onset of labor enhances the speed of physiological maturing by means of a still unknown adaptational process, as is known to happen in the lungs, reducing the incidence of RDS in spontaneous delivery compared to elective cesarean section 
The clinical importance of these findings for later (cognitive and motor) development in children is not clear at present for our study group, although it is known that following symptomatic neonatal hypoglycemia, more than 50% of infants demonstrate cognitive and motor impairments at the age of 18 mo 
. In low-birth-weight infants, even an asymptomatic moderate hypoglycemia may lead to cognitive and motor impairments at the age of 18 mo 
. Hyperbilirubinemia is potentially neurotoxic, especially in infants born preterm 
. When treated appropriately, however, bilirubin levels under 30 mg/dl are not associated with adverse neurodevelopmental outcome 
MacKay et al. 
reported on the increased need for special education in preterm-delivered infants. In a retrospective cohort study of 407,503 schoolchildren, they showed that gestational age at delivery had a strong, dose-dependent relationship with special educational need. Until further evidence becomes available, the decreased risk of special educational need with advancing gestational age should be taken into account when considering how best to manage PPROM.
Within the Dutch Consortium for Women's Health and Reproductivity Studies (http://www.studies-obsgyn.nl
), the PPROMEXIL trial is the largest trial so far with regard to the number of participating hospitals (60 out of 98 eligible hospitals, 61%). The 207 non-randomized women in our study who allowed data collection differed from the randomized women. Similar to two other Dutch consortium trials (HYPITAT and DIGITAT) 
, the women who agreed to be randomized differed in level of education, smoking habits, maternal age, and preferred management from those who did not agree to be randomized. The non-randomized subgroup of women who preferred IoL was too small to draw any conclusions from. In the EM subgroup, no differences were seen in the primary outcome or the secondary neonatal and maternal outcomes. Even though some women eligible to participate in the trial did not, we believe that we did not miss a significant group at a higher (or lower) risk for neonatal sepsis who were treated expectantly. Despite some differences in baseline characteristics, we assume that the results of our study can be generalized to at least the Dutch/Western European population. Because of wide differences in general health care and availability of antibiotics, it is likely that these results cannot be generalized to low-income countries.
The main limitation of our study is that it proved to be underpowered. We hypothesized a decrease in neonatal sepsis rate of 7.5% to 2.5%, but found a difference of only 1.5% (2.6% in the IoL group versus 4.1% in the EM group). The liberal use of antibiotic therapy before or during labor (41% of all participating women received antibiotics) might have contributed to a lower incidence compared to the other trials in which antibiotics were not administered prophylactically 
. These previous studies showed high rates of neonatal sepsis with EM. Similarly, improved health care may have contributed to a reduction of the incidence of neonatal sepsis in women with PPROM over the last decades.
In our study the observed difference in sepsis rates between the EM and IoL groups did not reach statistical significance. Although this study is one of the largest published so far, our sample size was insufficient to demonstrate small differences. In retrospect, anticipating a risk reduction of 66% (a difference in neonatal sepsis rate of 7.5% versus 2.5%) might have been too optimistic. However, several previous studies 
showed neonatal sepsis incidences up to 9.5% with EM, and we did observe an incidence near 2.5% in the IoL group. Although optimistic, we feel that our hypothesized risk reduction was not unrealistic. Furthermore, we used a one-sided test for the power calculation. We considered it not plausible that IoL in women with PPROM near term would increase the proportion of cases of neonatal sepsis. In retrospect, considering the results of the meta-analysis, one might question this choice of a one-sided test, as several studies in the meta-analysis show an increased risk for sepsis in the IoL group. However, the analysis was executed exactly as planned in advance. Two-sided testing would have required a sample size that would not have been feasible in our setting, in view of the limitations set by our funding body.
A second potential limitation is that EM prolonged gestation by just 4 d. This rather small difference might be partly due to the fact that the median gestational age at rupture of membranes was 35+4 wk and median gestational age at randomization was 35+6 wk. The overrepresentation of women beyond the 35th completed week of gestation was caused by the fact that women in their 35th week of gestation more often refused to participate (mean gestational age at PPROM in the non-randomized group was 34+6 wk), and before our study an expectant policy was the standard. Furthermore, hesitation of clinicians to induce labor before 35 completed weeks of gestation, which prior to the start of the PPROMEXIL trial was not recommended in the Dutch guidelines, might also have influenced this outcome.
A third limitation of this study is that we reported many secondary, mostly neonatal, outcomes. Although this is not uncommon in studies in maternal–fetal medicine, it is possible that a significant difference can be found by chance. If one applies a Bonferroni correction to the p
-value, the adjusted threshold is p
<0.001. By applying this threshold, the incidence of hypoglycemia (p
0.0008) remains the only statistically significant difference between the groups.
We are aware of the ongoing multicenter PPROMT trial (ISRCTN44485060), which has a design similar to that of our study. That trial may possibly answer the question whether IoL in women with near-term PPROM reduces the risk of neonatal sepsis 
. However, the updated meta-analysis clearly demonstrates that the incidence of neonatal sepsis is comparable in both treatment strategies 
. We therefore plan to perform an individual patient data meta-analysis on the management of PPROM. Combining large trials in an individual patient data meta-analysis would, in our opinion, produce the best currently available evidence for the management of PPROM. We have already planned such an analysis with the PPROMT study group, and will contact researchers from other published trials to collaborate in an individual patient data meta-analysis.
We conclude that in pregnancies complicated by PPROM between 34 and 37 wk of gestation the incidence of neonatal sepsis is low. Neither our trial nor the updated meta-analysis shows that IoL substantially improves pregnancy outcomes compared with EM.