You are here

Can neonatal sepsis be predicted in late preterm premature rupture of membranes? Development of a prediction model

European Journal of Obstetrics & Gynecology and Reproductive Biology, pages 90 - 95

Abstract

Objective

Women with late preterm premature rupture of membranes (PROM) have an increased risk that their child will develop neonatal sepsis. We evaluated whether neonatal sepsis can be predicted from antepartum parameters in these women.

Study design

We used multivariable logistic regression to develop a prediction model. Data were obtained from two recent randomized controlled trials on induction of labor versus expectant management in late preterm PROM (PPROMEXIL trials, (ISRCTN29313500 and ISRCTN05689407). Data from randomized as well as non-randomized women, who consented to the use of their medical data, were used. We evaluated 13 potential antepartum predictors for neonatal sepsis. Missing data were imputed. Discriminative ability of the model was expressed as the area under the receiver operating characteristic (ROC) curve and a calibration with both a calibration plot and the Hosmer and Lemeshow goodness-of-fit test. Overall performance of the prediction model was quantified as the scaled Brier score.

Results

We studied 970 women. Thirty-three (3.4%) neonates suffered neonatal sepsis. Maternal age (OR 1.09 per year), maternal CRP level (OR 1.01 per mmol/l), maternal temperature (OR 1.80 per °C) and positive GBS culture (OR 2.20) were associated with an increased risk of neonatal sepsis. The model had an area under the ROC-curve of 0.71. The model had both a good calibration and accuracy.

Conclusions

Antepartum parameters aid in the more precise prediction of the risk of neonatal sepsis in women with late preterm PPROM.

Keywords: Late preterm premature rupture of membranes, PPROM, Predicting, Prediction model, Neonatal sepsis.

1. Introduction

Preterm premature rupture of the membranes (PROM) is an important clinical problem. It complicates 1% to 5% of all pregnancies, and precedes 30–40% of all preterm deliveries [1], [2], and [3]. Preterm PROM is associated with increased perinatal and maternal morbidity and mortality [4], [5], and [6]. The risk of infection, resulting in chorioamnionitis and neonatal sepsis, is considered to be the largest threat in patients with late preterm PROM.

While there is no longer a need to administer antenatal corticosteroids for lung maturity, induction of labor is recommended from 34+0 weeks onwards [7] and [8]. However, two randomized controlled trials in The Netherlands showed that induction of labor (IoL) in asymptomatic women did not reduce the risk of neonatal sepsis compared to expectant management (EM) (combined Relative Risk (RR) 0.66, 95% Confidence Interval (CI) 0.30 to 1.5; neonatal sepsis rate of 2.7% (IoL) and 3.8% (EM) [9] and [10].

In the light of this, expectant management could be advocated in late preterm PROM, while identification of a high-risk subgroup could be helpful in the management of these women. Inducing labor and close monitoring of identified neonates at risk may reduce the incidence of neonatal sepsis even further, whereas EM and less frequent monitoring in low risk women will potentially lead to a reduction of neonatal complications due to prematurity, and of costs.

In this secondary analysis, we assessed the ability of several antepartum clinical parameters to predict neonatal sepsis. The aim was to develop a prediction model based on parameters easily available in routine practice.

2. Materials and methods

For this study we used the data from the PPROMEXIL trials; PPROMEXIL (ISRCTN29313500) [10] and PPROMEXIL-2-trial (ISRCTN05689407) [9] . The PPROMEXIL trials were multicenter, open-label, randomized controlled trials in The Netherlands in which all eight academic and 52 non-academic hospitals participated. Briefly, women with a singleton or twin pregnancy were eligible for the PPROMEXIL trial when they were not in labor 24 h after PPROM between 34+0 and 37+0 weeks of gestational age. Fig. 1 outlines the study design and selection of patients.

gr1

Fig. 1 Trial profile.

PPROM had to occur after 26+0 weeks. Women with a monochorionic multiparty, an abnormal (non-reassuring) cardiotocogram, meconium stained amniotic fluid, major fetal anomalies, signs of intrauterine infections, HELLP syndrome or severe preeclampsia were not included.

Women who consented to participation in the trial were randomly allocated to either induction of labor (IoL-group) or expectant management (EM-group). Patients who did not consent to randomization, but who provided authorization for the use of their medical data, were treated according to their preference (non-randomization group), and had either immediate delivery or expectant monitoring. The management of these women was similar to that of randomized patients. In the present study we combined data from randomized women and women who refused randomization in the same cohort.

On admission a vaginal swab was taken to culture vaginal microorganisms. Immediate delivery was mostly due to induction of labor, but in some cases, for example for breech presentation, planned Cesarean delivery was performed. Expectant management consisted of at least daily maternal temperature monitoring and twice weekly blood sampling for signs of infection. If signs or symptoms of an (intra-uterine) infection developed labor was induced before 37 + 0 weeks of gestational age. Otherwise, if a woman reached 37+0 weeks of gestational age, labor was induced or a planned Cesarean section was done. In non-randomized women preferring expectant management labor was induced at 37+0 weeks of gestational age, according the national guideline [11] . The details of the trials have been published [9], [10], and [12].

The outcome of interest was early onset neonatal sepsis. This was defined as: (1) positive blood culture taken at birth or (2) two or more symptoms of infection within 72 h plus one of three items: (a) positive blood culture; (b) CRP > 20; (c) positive surface culture of a known virulent pathogen [9] and [10]. Each case of suspected sepsis was judged by an independent panel of pediatricians (AM and RM). They were unaware of the allocation of randomization or preferred management in non-randomization, and individually adjudicated either neonatal sepsis or no sepsis. In case of disagreement this was resolved by a panel discussion.

Guidelines about the number of potential predictors that can be included in a prediction model have been proposed, including a requirement of 10 events per predictor variable [13] . However, since this is the first study to assess the predictive value of a multivariable model on the probability of neonatal sepsis, with many more potential predictors, we allowed five events per predictor variable to be included in the final model, dictating a maximum of six predictors (Ncases = 33). We assumed the following variables as candidate predictors: maternal age, parity (nulliparity versus multiparity), ethnicity, maternal smoking, maternal body mass index (BMI) at start of pregnancy, gestational age (GA) at PROM, antenatal administration of steroids, antepartum administration of antibiotics, and the following variables ascertained on admission: maternal C-reactive protein (CRP) level, maternal white blood cell count, maternal temperature, positive vaginal culture (any pathogenic specimen), positive group B streptococcus (GBS) culture. Antibiotics were given according to local protocol on admission, before signs or symptoms of chorioamnionitis.

Data were incomplete for some variables. The omission of patients from the analysis who have one or more predictor values missing could lead to a loss of precision and subsequently a reduction in statistical power [14], [15], and [16]. More seriously, using only complete cases could potentially bias results. We assumed these data were missing at random (MAR) [17] . Therefore, we imputed the dataset using regression imputation.

All potential predictors were introduced in a multivariable logistic regression model, and were step-by-step excluded using the Wald test. We used a liberal p-value of 0.20 to keep a predictor in the model, according to current guidelines for developing prediction models [13] .

Ultimately, only a maximum of 6 predictors could be included in the model, so if there were more than six significant predictors, the six predictors with the lowest p-value would be retained in the model.

There were 14 twin pregnancies in the entire study group, of which three of the 28 neonates developed sepsis, in one case a first-born child and in one case both children. Adding both twins to the analysis would introduce high correlation between these pairs, introducing much more complexity to the prediction model. Therefore, in case of a twin pregnancy, we included only the firstborn in the analysis.

In deriving our prediction model, we have chosen to include more potential predictors than advised, increasing the risk of severe over-fitting of our model. To account for this, we internally validated the model using bootstrapping techniques, and applied shrinkage to the original prediction model, a method which has extensively been described by Steyerberg [16] .

Discriminative ability of the model (after shrinkage) was determined by calculating the area under the receiver operating characteristic (ROC) curve, which is a measure of the models’ ability to distinguish between cases with and without neonatal sepsis [18] . We assessed the calibration of the model with a calibration plot. To quantify the calibration, we performed a Hosmer and Lemeshow test for goodness-of-fit, where a low p-value (p < 0.05) indicates evidence of poor calibration and lack of fit. Furthermore, we visually inspected the calibration plot. For overall performance, we computed the Brier score, which is a measure of the accuracy of the predictions. Because the maximum Brier score for a given set of patients depends on the incidence of the outcome, we scaled the Brier score by its maximum possible score for the incidence of neonatal sepsis in our population, so that it ranges from 0 to 100%, and can be interpreted as Pearson's R2 statistic.

We used SPSS for Windows (version 19.0; SPSS Inc., Chicago, IL) for the imputation of missing values. All statistical modeling steps, including internal validation were performed using R, a language and environment for statistical computing (version 2.12.2; http://www.rproject.org/ ).

3. Results

A total of 970 women were identified as eligible for the present study, of whom 727 had been randomized and 243 had not been randomized but had been managed according to the policy of their preference. Non-randomized women differed significantly from randomized women in maternal age (older), smoking habit (fewer smoked), educational level (more highly educated), gestational age at PROM (earlier gestational age), positive vaginal culture (any specimen or group B streptococcus; higher incidence), antenatal administration corticosteroids (higher incidence), antepartum antibiotic treatment (more frequently).

From these 970 women, 590 were initially managed expectantly while in 380 the aim had been immediate delivery. Baseline characteristics and distribution of potential predictors between women with and women without a neonate suffering from neonatal sepsis before imputation are given in Table 1 . Of all (n = 970) neonates 33 (3.4%) suffered neonatal sepsis.

Table 1 Baseline characteristics and potential predictors.

Characteristics a Women with a newborn with neonatal sepsis (n = 33) Women with a newborn without neonatal sepsis (n = 937) p-Value Missing values n (%)
Maternal age (range) [±SD], y 33.2 (22.9–46.7) [±6.0] 30.2 (18.1–45.5) [±5.2] 0.001 0 (0%)
Number of nulliparous, n (%) 17 (52%) 415 (44%) 0.415 0 (0%)
Twin pregnancy, n (%) 2 (6.1%) 12 (1.3%) 0.024 0 (0%)
Intended management at PPROM        
Immediate delivery 11 (33%) 369 (39%) 0.484 0 (0%)
Expectant management 22 (67%) 568 (61%) 0.484  
Ethnic origin, white, n (%) 25 (83%) 715 (83%) 0.913 74 (7.6%)
Maternal smoking, n (%) 4 (12%) 187 (20%) 0.275 49 (5.1%)
Higher educational level (higher professional school or university), n (%) 8 (38%) 195 (37%) 0.883 422 (43.5%)
Body mass index at booking (range) [±SD], kg/m 2 25.1 (17.9–43.4) [±6.0] 24.8 (15.8–53.3) [±5.4] 0.799 157 (16.2%)
Gestational age at PPROM        
<34 weeks, n (%) 7 (21%) 168 (18%) 0.895 0 (0%)
34 + 0 to 34 + 6 weeks, n (%) 3 (9.1%) 141 (15%)
35 + 0 to 35 + 6 weeks, n (%) 10 (30%) 286 (31%)
36 + 0 to 36 + 6 weeks, n (%) 13 (39%) 338 (36%)
Gestational age at PPROM, mean [±SD] (median) [IQR], d 244.0 [±14]

(247) [241–253]
244.8 [±11]

(248) [241–253]
0.693 0 (0%)
Gestational age at delivery, mean [±SD] (median) [IQR], d 251.1 [±7.1]

(254) [246–256]
252.5 [±7.9]

(254) [249–258]
0.317 0 (0%)
Maternal temperature (°C), [±SD] 37.5 [±1.0] 37.0 [±0.6] <0.0001 280 (28.9%)
Maternal CRP (mmol/l) [±SD] 36.2 [±62] 15.7 [±28] 0.001 251 (25.9%)
Maternal WBC (×10 9 /l) [±SD] 16.0 [±6.6] 13.2 [±5.2] 0.010 192 (8.0%)
Positive vaginal culture (any specimen), n (%) 13 (40%) 205 (22%) 0.018 2 (0.2%)
Positive vaginal culture (GBS), n (%) 11 (33%) 143 (15%) 0.005 2 (0.2%)
Antenatal corticosteroids administration, n (%) 8 (28%) 165 (18%) 0.229 58 (6.0%)
Antepartum antibiotic treatment, n (%) 16 (48%) 364 (39%) 0.265 3 (0.3%)

a Percents given are related to available data per characteristic before imputation and may differ from total number of patients.

Characteristics indicated in bold are potential predictor characteristics.

CRP – C-reactive protein; GBS – group B streptococcus; IQR – inter quartile range; SD – standard deviation; WBC – white blood count.

Table 2 shows both the original prediction model for neonatal sepsis and the validated model after the shrinkage. The bootstrap validation yielded a shrinkage factor of 0.89, and this was uniformly applied to all regression coefficients. Maternal age, CRP level and maternal temperature at admission and positive vaginal culture for Group B streptococcus were included in the final model. After internal validation the odds ratios were 1.09 per year for maternal age, 1.01 per mmol/l for maternal CRP level, 1.80 per °C for maternal temperature and 2.20 for a positive Group B streptococcus culture.

Table 2 Prediction model for the estimation of the individual risk for neonatal sepsis.

  Original model Model after internal validation
Variable Regression coefficient Odds ratio (95% CI) p-Value Regression coefficient a Odds ratio
Intercept −31.17 −28.06
CRP (mmol/l) 0.007 1.01 (1.00–1.01) 0.07 0.006 1.01
Age (years) 0.094 1.10 (1.03–1.17) 0.01 0.083 1.09
GBS culture (pos/neg) 0.886 2.42 (1.12–5.25) 0.02 0.788 2.20
Temperature (°C) 0.658 1.93 (1.17–3.19) 0.01 0.586 1.80

a Regression coefficients after adjustment for overfitting by shrinkage (shrinkage factor = 0.83), the intercept was re-estimated.

To calculate the absolute risk for neonatal sepsis

FORMULA:

stripin: si1.gif

CI – confidence interval; CRP – C-reactive protein; GBS – group B Streptococcus.

The individual risk of neonatal sepsis can be estimated with the formula provided in Table 2 .

Prediction models for randomized women only and non-imputed women resulted in similar outcome (data not shown). Fig. 2 shows the ROC curve of the prediction model for neonatal sepsis in women with PPROM. The area under the curve was 0.71 (95% CI = 0.61–0.82), indicating good discriminative ability.

gr2

Fig. 2 ROC curve of the prediction model after the internal validation step.

Fig. 3 shows the calibration plot of the prediction model. It shows that for all risk deciles the average predicted risk is approximately the same as the observed frequency, indicating very good calibration. It does show, however, that most predicted risks are relatively low, because of the low prevalence of sepsis in our sample. Predicted risks of neonatal sepsis of over 20% are rare. Furthermore, the Hosmer and Lemeshow goodness-of-fit test yielded a p-value of 0.32, verifying that the model is well calibrated. Overall accuracy was high, quantified by a scaled Brier score of 29.5%.

gr3

Fig. 3 Calibration plot of the model.

4. Comments

In the present study we evaluated potential antenatal predictors of neonatal sepsis in women with late preterm PROM. Despite the fact that the overall incidence of neonatal sepsis was low in our study population [9] and [10], we were able to develop a prediction model based on four predictors, i.e. maternal age, maternal CRP level, maternal temperature and a positive GBS culture. The model had a good discriminative ability as well as a good calibration and overall fit of the model. Still the most predicted risks of neonatal sepsis were low and a predicted risk of more than 20% was rare.

Based on this model, one can individualize a neonate's risk, which can be helpful to guide the obstetrician whether or not to monitor the woman more closely and induce labor. The PPROMEXIL trials showed induction of labor will lead to an overall risk of neonatal sepsis of about 2.5%. So with a predicted risk of 2.5% or less, induction of labor will not contribute to better neonatal outcome. Whereas with an increasing predicted risk of neonatal sepsis there will be a point at which the risk of sepsis outweighs the risk of complications of prematurity itself, such as hypoglycemia (16% for IoL vs 9% in EM), hyperbilirubinemia (33% for IoL vs 25% in EM). In our opinion a predicted risk of neonatal sepsis between 10 and 15% outweighed the risk of complications due to prematurity. So even though the current model is incapable of predicting a risk above 20% this is sufficient to guide clinical decision-making.

We decided to include non-randomized women although they differ from randomized patient in several aspects. The most important differences are related to the earlier age at which PPROM occurred in the non-randomized women. Because of the low rate of women declining any form of participation (3.7%, Table 1 ), we felt that by including non-randomized women this model is even more generalizable than a model based on randomized patients only. A model based on randomized women only gave similar outcome with less predictive power.

Our study has several limitations. First, we imputed missing variables because omission of patients with missing variables could lead to a loss in statistical power [14], [15], and [16]. Although there are no guidelines about the maximum percentage of missing data that can be imputed, we decided to leave out erythrocyte sedimentation rate (ESR) and maternal education level because we considered the proportion of missing data of these variables as too high (33 and 44%, respectively). Since there is always an uncertainty introduced by imputing missing values, regression estimates for these variables could be unstable, and the probability of biased estimates is higher than would be the case if (many) more values would be observed. By doing this we missed out the possible predictive ability of these variables.

Conversely, maternal CRP level and maternal temperature, which were both included in the model, were amongst the 13 potential predictors the two with the largest proportion of missing data (26 and 29%, respectively). To estimate the effect of our imputation we performed a sensitivity analysis for the complete non-imputed sample (n = 564). This analysis identified the same predictors with a similar effect size.

Secondly, we had only five events per predictor while 10 events per predictor variable is recommended [13] . This might have influenced the validity of the model. We considered 13 variables as potential predictors, and by including only three of them in the prediction model we would, in our opinion, have missed the predictive ability of the other variables.

Thirdly, developing a prediction model as a secondary analysis from data of a RCT might to some extent limit the generalizability of the results to the whole population. Therefore external validation is needed to confirm or refute our findings.

From the four identified factors, maternal CRP level, maternal temperature and positive Group B streptococcus culture seem to have a pathophysiological explanation, whereas the increased risk of increasing maternal age is more difficult to explain. As there is no plausible biological mechanism explaining why age is predictive of neonatal infection, the associations may be caused by other factors related to maternal age, which we did or did not document in this study. Further research is needed to explore the role of maternal age in the causal pathway of neonatal sepsis.

The predictive capacity of CRP to identify neonatal sepsis antepartum is variable. Van de Laar et al. could not identify any paper of sufficient quality that had studied the accuracy of CRP in predicting neonatal sepsis in women with PROM prior to 36 weeks gestational age [19] . A few recent studies showed different predictive abilities of CRP, which ranged between an AUC of 0.59 in preterm women (GA 24–36 weeks) and 0.80 in late preterm and term women (mean GA 38.5 weeks) [20], [21], and [22]. Our model shows a moderate predictive ability for CRP (AUC 0.67 [95%CI 0.57 to 0.77]), but in the preterm population one can in our opinion not rely only on antepartum CRP levels.

A positive vaginal culture for either Group-B streptococcus (GBS) or other microorganisms was in both cases a predictor of neonatal sepsis. Because the predictive value of both types of positive vaginal culture were strongly correlated, we decided to include GBS positive culture in the model, because GBS is one of the most frequently identified microorganisms causing neonatal sepsis [23] . A recent Canadian study, however, showed a significant reduction in GBS as a cause for early onset sepsis over time and a trend toward an increase in other microorganisms [24] . Unfortunately, we were unable to extract data for specific microorganisms other than GBS from the database. We therefore could not use for example coagulase-negative Staphylococcus or Escherichia coli as individual bacteria for the prediction of neonatal sepsis. On admission GBS status was unknown in a large proportion of our asymptomatic study population. Because we did not use a real-time PCR method on GBS status and antibiotics were given according to local protocol, 20% of GBS positive women did not receive antibiotics during admission or labor. In these cases GBS status was only known after delivery.

In both the PPROMEXIL trials, and in the meta-analysis performed after those studies, we observed no benefit from immediate delivery in the prevention of neonatal sepsis. In view of the prognosticators that we identified, one can hypothesize that immediate delivery, maybe after antibiotic prophylaxis, might be useful in women with a high-risk profile. As our study is hypothesis-generating, this should be assessed in future studies. As we are in close contact with the investigators of the on-going Australian led PROMT study, this study can be used as such [25] .

Indeed, external validation is needed to confirm the model's ability in other populations or settings to identify women with an increased risk in whom induction of labor is advised, as opposed to women with a low risk in whom the risks of complications due to prematurity outweighs the risk of neonatal sepsis and where expectant management is the best treatment strategy. Furthermore we have planned to investigate interaction between biomarkers and the treatment allocation in the PPROMEXIL trials [9] and [10] in the near future.

We developed a prediction model of the risk of neonatal sepsis in women with late preterm PROM with a good discriminative ability as well as a good calibration and overall validity. If externally validated, this prediction model might be used to decide in which women with late preterm PROM labor should be induced.

Acknowledgments

We thank the research staff of the Dutch Consortium for Women's health and reproductivity studies ( www.studies-obsgyn.nl ), and the residents, midwives, nurses and gynecologists of the participating centers for their help with recruitment and data collection. And finally, we would thank all women who participated in the PPROMEXIL trial.

References

  • [1] S.M. Cox, M.L. Williams, K.J. Leveno. The natural history of preterm ruptured membranes: what to expect of expectant management. Obstet Gynecol. 1988;71:558-562
  • [2] B.M. Mercer, R.L. Goldenberg, P.J. Meis, et al. The Preterm Prediction Study: prediction of preterm premature rupture of membranes through clinical findings and ancillary testing. The National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network. Am J Obstet Gynecol. 2000;183:738-745 Crossref.
  • [3] R.S. Gibbs, J.D. Blanco. Premature rupture of the membranes. Obstet Gynecol. 1982;60:671-679
  • [4] B. Furman, I. Shoham-Vardi, A. Bashiri, O. Erez, M. Mazor. Clinical significance and outcome of preterm prelabor rupture of membranes: population-based study. Eur J Obstet Gynecol Reprod Biol. 2000;92:209-216 Crossref.
  • [5] R.L. Goldenberg, K.G. Nelson, R.O. Davis, J. Koski. Delay in delivery: influence of gestational age and the duration of delay on perinatal outcome. Obstet Gynecol. 1984;64:480-484
  • [6] B.M. Mercer. Preterm premature rupture of the membranes. Obstet Gynecol. 2003;101:178-193 Crossref.
  • [7] ACOG. ACOG Practice Bulletin No. 80: premature rupture of membranes. Clinical management guidelines for obstetrician-gynecologists. Obstet Gynecol. 2007;109:1007-1019
  • [8] Royal College of Obstetricians and Gynaecologists. Preterm prelabour rupture of membranes (Green Top Guideline No. 44). (RCOG, London, 2006)
  • [9] D.P. van der Ham, J.L. van der Heyden, B.C. Opmeer, et al. Management of late-preterm premature rupture of membranes: the PPROMEXIL-2 trial. Am J Obstet Gynecol. 2012;207:276.e1-10
  • [10] D.P. van der Ham, S.M. Vijgen, J.G. Nijhuis, et al. Induction of labor versus expectant management in women with preterm prelabor rupture of membranes between 34 and 37 weeks: a randomized controlled trial. PLoS Med. 2012;9:e1001208 Crossref.
  • [11] NVOGNet. Breken van de vliezen voor het begin van de baring. (NVOG, 2002)
  • [12] D.P. van der Ham, J.G. Nijhuis, B.W. Mol, et al. Induction of labour versus expectant management in women with preterm prelabour rupture of membranes between 34 and 37 weeks (the PPROMEXIL-trial). BMC Pregnancy Childbirth. 2007;7:11 Crossref.
  • [13] F.J. Harrell. Regression modeling strategies. (Springer, New York, 2001)
  • [14] P. Allison. Missing data. (Sage, Iowa, 2001)
  • [15] A.R. Donders, G.J. van der Heijden, T. Stijnen, K.G. Moons. Review: a gentle introduction to imputation of missing values. J Clin Epidemiol. 2006;59:1087-1091 Crossref.
  • [16] E.W. Steyerberg. Clinical prediction models. (Springer, New York, 2009)
  • [17] D. Rubin. Inference and missing data. BIometrika. 1976;63:581-592 Crossref.
  • [18] J.A. Hanley, B.J. McNeil. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology. 1983;148:839-843
  • [19] R. van de Laar, D.P. van der Ham, S.G. Oei, C. Willekes, C.P. Weiner, B.W. Mol. Accuracy of C-reactive protein determination in predicting chorioamnionitis and neonatal infection in pregnant women with premature rupture of membranes: a systematic review. Eur J Obstet Gynecol Reprod Biol. 2009;147:124-129 Crossref.
  • [20] T. Popowski, F. Goffinet, F. Maillard, T. Schmitz, S. Leroy, G. Kayem. Maternal markers for detecting early-onset neonatal infection and chorioamnionitis in cases of premature rupture of membranes at or after 34 weeks of gestation: a two-center prospective study. BMC Pregnancy Childbirth. 2011;11:26 Crossref.
  • [21] J.L. van der Heyden, S.S. van Teeffelen, A.C. Coolen, et al. Is it useful to measure C-reactive protein and leukocytes in patients with prelabor rupture of membranes?. Am J Perinatol. 2010;27:543-547 Crossref.
  • [22] A. Torbe, M. Sokolowska, S. Kwiatkowski, R. Rzepka, B. Torbe, R. Czajka. Maternal plasma lipopolysaccharide binding protein (LBP) concentrations in pregnancy complicated by preterm premature rupture of membranes. Eur J Obstet Gynecol Reprod Biol. 2011;156:153-157 Crossref.
  • [23] E.J. Weston, T. Pondo, M.M. Lewis, et al. The burden of invasive early-onset neonatal sepsis in the United States, 2005–2008. Pediatr Infect Dis J. 2011;30:937-941 Crossref.
  • [24] M. Sgro, P.S. Shah, D. Campbell, A. Tenuta, S. Shivananda, S.K. Lee. Early-onset neonatal sepsis: rate and organism pattern between 2003 and 2008. J Perinatol. 2011;31:794-798
  • [25] J.M. Morris, C.L. Roberts, C.A. Crowther, S.L. Buchanan, D.J. Henderson-Smart, G. Salkeld. Protocol for the immediate delivery versus expectant care of women with preterm prelabour rupture of the membranes close to term (PPROMT) Trial [ISRCTN44485060]. BMC Pregnancy Childbirth. 2006;6:9

Footnotes

a Department of Obstetrics and Gynecology, Maastricht University Medical Centre, GROW – School for Oncology and Developmental Biology, PO Box 5800, 6202 AZ Maastricht, The Netherlands

b Department of Obstetrics and Gynecology, Martini Hospital Groningen, PO Box 30033, 9700 RB Groningen, The Netherlands

c Clinical Research Unit, Academic Medical Centre Amsterdam, PO Box 22700, 1105 DE Amsterdam, The Netherlands

d Department of Obstetrics and Gynecology, VieCuri Medical Centre, Tegelseweg 210, 5912 BL Venlo, The Netherlands

e Department of Pediatrics, Maastricht University Medical Centre, GROW – School for Oncology and Developmental Biology, PO Box 5800, 6202 AZ Maastricht, The Netherlands

f Department of Obstetrics and Gynecology, University Medical Centre Groningen, Hanzeplein 1, 9713 GZ Groningen, The Netherlands

g Department of Obstetrics and Gynecology, Isala Klinieken Zwolle, Dokter van Heesweg 2, 8025 AB Zwolle, The Netherlands

h Department of Obstetrics and Gynecology, Leiden University Medical Centre, Albiniusdreef 2, 2333 ZA Leiden, The Netherlands

i Department of Obstetrics and Gynecology, Maxima Medical Centre Veldhoven, De Run 4600, 5504 DB Veldhoven, The Netherlands

j Department of Obstetrics and Gynecology, University Medical Centre Utrecht, Heidelberglaan 100, 3584 CX Utrecht, The Netherlands

k Department of Obstetrics and Gynecology, Albert Schweitzer Hospital Dordrecht, Albert Schweitzerplaats 25, 3318 AT Dordrecht, The Netherlands

l Department of Obstetrics and Gynecology, Amphia Hospital Breda, Langendijk 75, 4819 EV Breda, The Netherlands

m Department of Obstetrics and Gynecology, Ikazia Hospital Rotterdam, Montessoriweg 1, 3083 AN Rotterdam, The Netherlands

n Department of Obstetrics and Gynecology, School of Paediatrics and Reproductive Health, University of Adelaide, 5000 SA, Australia

lowast Corresponding author at: Department of Obstetrics and Gynecology, Martini Hospital Groningen, PO Box 30033, 9700 RB Groningen, The Netherlands. Tel.: +31 648 342140.