Correspondence: Dr Y. Ville, Senior Lecturer, Fetal Medicine Unit, St George's Hospital Medical School, Cranmer Terrace, London SW 17 ORE, UK.
Objectives To test the association between cytokine levels in the amniotic fluid and (i) the vascular invasion phase of intrauterine infection, (ii) the occurrence of periventricular leukomalacia; to assess the correlation between C-reactive protein levels, a recognised biological marker of inflammation in maternal serum and cytokine levels in the amniotic fluid.
Design Prospective clinical study.
Setting Fetal medicine unit and neonatal intensive care unit, Antoine Beclere Hospital, Clamart, France.
Sample Thirty-one pregnancies complicated by chorioamnionitis leading to birth before 32 weeks of gestation.
Methods Interleukin 1-β, Interleukin 6 and TNF-α prospectively measured in the amniotic fluid. Histological examination of the placenta. Ultrasound examination and magnetic resonance imaging of the brains of the newborn infants performed within the first week of life.
Main outcome measures The occurrence of periventricular leukomalacia was assessed by trans fontanellar ultrasound and magnetic resonance imaging.
Results There was a significant positive correlation between the occurrence of histological chorioamnionitis, vascular extension of infection of the membranes, maternal inflammatory syndrome and neonatal sepsis. A strong association was found between maternal serum C-reactive protein concentrations and cytokine levels in the amniotic fluid. Interleukin-1β was the best predictor of vascular extension of chorioamnionitis, and TNF-α was the best predictor of the development of severe early neonatal infection. There was no association betwecn the amniotic fluid levels of cytokines and the development of periventricular leukomalacia.
Conclusions These data suggest that IL-1β, IL-6 and TNF-α are produced in relation to intrauterine inflammation and infection, but cannot be directly implicated in the development of fetal cerebral white matter lesions.
Over the past two decades, improvements in the neonatal intensive care of very preterm infants has resulted in a marked reduction in both perinatal morbidity and mortality, but has not reduced the prevalence of cerebral palsy1. The incidence of severe intraventricular haemorrhage has decreased with the widespread use of antenatal steroids, and periventricular leukomalacia has become the main cause of cerebral palsy, accounting for 38% to 93% of cases of cerebral palsy after a diagnosis of cystic periventricular leukomalacia has been made2–4. Periventricular leukomalacia is defined as a progressive lesion of the white matter adjacent to the external angle of the lateral ventricles occumng in the perinatal period'. Its prevalence ranges from 5% to 15% among infants born before 32 weeks of gestation5. Cystic periventricular leukomalacia is characterised by the appearance of cystic lesions on cranial ultrasounds performed during the first weeks of life. Its occurrence is almost always associated with serious subsequent neuromotor impairment (diplegia or tetraplegia)2,3. Non cystic periventricular leukomalacia is suspected when hyperechoic ultrasonographic densities persist for more than three weeks3. The incidence of non cystic periventricular leukomalacia is not well established and the neurodevelopmental outcome of affected infants is less predictable6.
The pathophysiology of periventricular leukomalacia remains poorly understood. However in recent studies, chorioamnionitis and prolonged preterm prelabour amniorrhexis were associated with its development7 and chorioamnionitis has recently been shown to be the main independent risk factor for developing cerebral palsy following prematurity9–11. Endotoxins have been suggested to be the mediators of white matter damage12, and increased levels of cytokines have been found in the amniotic fluid in preterm labour with intrauterine infection13–15. Leviton16 and, recently, Yoon et al.17 found a strong correlation between the high plasma levels of tumour necrosis factor (TNF-α) and Interleukin-6 (IL-6) in the umbilical cord at birth and impaired neurological development. In order to overcome the potential bias of retrospective studies, we measured prospectively the amniotic fluid levels of three cytokines and assessed their value both as markers of intrauterine infection and as predictors of the occurrence of periventricular leukomalacia in very preterm infants born after pregnancies complicated by severe chorioamnionitis.
The aims of the study were, therefore, firstly to test the association between cytokine levels in the amniotic fluid and (i) the vascular invasion phase of intrauterine infection and (ii) the occurrence of periventricular leukomalacia in the neonates; and second to assess the correlation between C-reactive protein levels, a commonly used marker of inflammation in maternal serum, and cytokine levels in the amniotic fluid.
SUBJECTS AND METHODS
Thirty-one women were included in this study which took place between February 1994 and February 1997. Inclusion criteria were pregnancy complicated by chorioamnionitis and preterm delivery (before 32 weeks of gestation) of infants who survived for at least seven days. The upper limit of inclusion was set at 32 weeks of gestation because periventricular leukomalacia is rarely encountered at a later gestation.
Chorioamnionitis was defined by the association of preterm labour and at least two of the following criteria at presentation: maternal temperature > 38°C; duration of rupture of the membranes >24 h; maternal serum C-reactive protein level >20 mg/L; maternal leukocytosis >15,000/mm3; fetal tachycardia >160 bpm without maternal tachycardia.
Gestational age was defined by the date of the last menstrual period and by ultrasound examination performed before 14 weeks. Neonatal outcome was analysed by clinical, biological and ultrasound assessment. On admission, all infants underwent routine investigations for neonatal sepsis detection. Biological tests (white blood cell count and C-reactive protein) were performed on admission and repeated after 24 hours. Cultures from neonatal blood, gastric aspirate, anus and skin swab cultures were also performed on admission. Bacterial cultures were considered positive if a single organism recognised as a pathogen in congenital sepsis was isolated from blood culture or from three other sites. Congenital neonatal sepsis was defined by the association of positive bacterial cultures together with an elevated neonatal C-reactive protein level, a total neutrophil count outside the reference range for preterm infants or an immature neutrophil rate >7%.
Amniotic fluid was obtained by transabdominal amniocentesis as part of the routine management of suspected chorioamnionitis on admission. This was sent for gram stain and culture and 5mL of fluid were stored in polypropylene tubes at-20°C. Interleukin-1 β (IL1 -β) (Immunotech, Marseille, France), Interleukind (IL-6) (Diclanone, Besangon, France) and TNF-α (Immunotech, Marseille, France) were measured on each sample by ELISA. The sensitivities of the kits were 2 pg, 5 pg and 5 pg. Intra and interassay variability was (0.52%, 2.01%), (1.6%, 5.4%) and (4%, 6.5%) for IL-6, TNF-α and IL-1, respectively. Cut-off levels of amniotic fluid cytokine levels were derived from previous studies to define normal and high concentrations13–15,17. The population was accordingly divided into four groups. Group 1 had a TNF- α > 300 pg/mL, group 2 an IL-6 > 5000 pg/mL, group 3 an IL-1β > 1000 pg/mL and group 4 normal cytokine levels, below these cut-off levels.
Delivered placentae were placed in clean plastic bags and taken to the laboratory within 12 hours. Histological examination of the placentae was performed blind by a single pathologist.
Histological chorioamnionitis was defined by the presence of acute inflammatory change in any tissue sample on the basis of criteria previously described14. Histological lesions were classified accordingly to the presence or absence of extension of the inflammatory process through the vascular wall in the umbilical cord section.
All infants were routinely examined by cerebral ultrasound scanning in the coronal and sagital planes (Acuson XP128,7 MHz transducer, Mountain View, USA) at least twice during the first week of life and then at least weekly until 34 weeks of post-conceptional age. Cystic and noncystic periventricular leukomalacia were diagnosed before infant discharge or subsequent neonatal death. Infants with microcystic periventricular leukoma- lacia or persisting hyperechoic periventricular densities after three weeks underwent magnetic resonance imaging at a postconceptional age of 32 to 36 weeks. Trans- fontanellar ultrasonographic scans and magnetic resonance images were reviewed by radiologists unaware of the obstetrical or biological findings. A diagnosis of non- cystic periventricular leukomalacia was made in the presence of linear periventricular zones of increased intensity corresponding to glial scars or ventriculomegaly with irregular contour of the lateral ventricles in the absence of intraventricular haemorrhage18,19.
Comparison of continuous data was made using the Mann-Whitney U test. The correlation coefficient between continuous variables which do not follow a normal distribution was estimated using the Spear mann's rank correlation coefficient. Proportions in different groups were compared using the χ2 test, with Yates' correction, or Fisher's exact test when the groups were small, as appropriate; P values were considered significant below 0.05.
Thirty-one mothers (24 singletons and 7 twin pregnancies) with a diagnosis of chorioamnionitis and preterm labour were delivered of 34 infants before 32 weeks of gestation. One infant died from pulmonary haemorrhage after 48 hours and 33 survived the first week and were kept for analysis. Thirty-four samples of amniotic fluid were obtained from these 31 cases of suspected intrauterine infection (24 singletons, 6 dichorionic diamniotic twins and 4 monochorionic diamniotic twins). Tocolysis was stopped as soon as the woman met the criteria for chorioamnionitis; positive gram stain and amniotic fluid culture were found in seven cases. Chorioamnionitis indicated delivery within 24 hours after diagnosis. All infants were routinely examined by ultrasound scanning and by magnetic resonance imaging. None of the infants included died during the first ten days of life. Periven- tricular leukomalacia associated lesions were diagnosed in six (18%) cases including 5 cystic and one non-cystic periventricular leukomalacia. Among the five with cystic pervientricular leukomalacia, one late onset form (diagnosed after six weeks of life) was excluded from the correlations with antenatal events. The demographics and perinatal outcome measures are presented in Table 1. The women were allocated to four groups according to the amniotic fluid concentrations of the three cytokines (Table 1). There was a strong correlation between amnotic concentrations of the different cytokines with coefficient correlation of 0.004,0.0074 and <0.0001 between IL-1 and IL-6, IL-6 and TNF-α, and IL-1 and TNF-α, respectively.
Table 1. Perinatal data on infants according to amniotic fluid cytokine concentrations (pg/mL). Values are given as median [range], determined by Fisher's exact test or the Mann-Whitney U test. CRP = C-reactive protein; RDS = respiratory distress syndrome; PVL = penventricular leukomalacia; IVH = inhaventricular haemonhage.
TNF-α >300 IL-6 >5000
*Comparison between any elevated and normal levels of amniotic fluid cytokines.
**plus one case of late onset cystic PVL. excluded from correlation studies.
Gestational age (weeks)
No endothelium invasion
Time between amniocentesis and delivery (days)
Maternal CRP (mg/L)
Maternal antibiotics > 24 h belore delivery
Prolonged rupture of membranes > 24 h
Total PVL (cystic PVL)
IVH (Papile grade 3 and 4)
All 31 placentae were examined. All but 5 showed lesions due to chorioamnionitis. Among the 26 affected placentae, vascular and parietal dissemination of the inflammatory process were found in 17 (65%) and 5 (21 %) respectively. No correlation was found between the severity of histological chorioamnionitis and the development of periventricular leukomalacia (Table 1).
There was no significant difference in antenatal treatment, interval between diagnosis and delivery, gestational age at delivery or birthweight between women with normal or elevated cytokine levels in the amniotic fluid. The association between relevant perinatal events, amniotic fluid levels of TNF-α, IL-6 and IL-1β, or with any of these three cytokine and the development of periventricular leukomalacia is shown in Table 1. There was a positive correlation between the amniotic fluid cytokine levels and both maternal serum C-reactive protein concentration and histological extension of intrauterine infection, but not with the occurrence of white matter lesions in the neonatal brain (P < 0.01, P < 0.005 and P= 0.36, respectively). All three cytokines were detected with higher concentrations in amniotic fluid when there was a proven neonatal sepsis during the first 48 hours of life (P= 0.02).
Table 2 shows the close association between amniotic fluid levels of cytokmes and maternal serum C- reactive protein concentration produced in response to intrauterine infection, especially during the vascular invasion of the process.
Table 2. Association of cytokine levels in amniotic fluid and maternal C-reactive protein levels in plasma with the vascular invasion phase of infection and the occurrence of white matter damage. Values are given as median [range]. determined by the Mann-Whitney U test. PVL = periventricular leukomalacia; CRP = C-reactive protein.
(n= 27) (n= 6)
(n= 6) (n= 6)
Figure 1 illustrates the correlation between amniotic fluid cytokine levels (IL-6, IL-1β and TNF-α) and the severity of histological chorioamnionitis.
There was a high correlation coefficient between maternal serum C-reactive protein concentration and the amniotic fluid concentration of each cytokine (P= 0.0021, P = 0.0006 and P= 0.0004 for Interleukin-6, Interleukin-1β and TNF-α, respectively).
There is growing evidence that periventricular leukomalacia and cerebral palsy are associated with preterm rupture of the membranes and chorioamnionitis8–11,20. Two retrospective studies among preterm infants born with periven tricular leukomalacia before 36 weeks of gestation have reported that high levels of IL-6, in either umbilical cord plasma or amniotic fluid, were associated both with severe histological chorioamnionitis and subsequent development of white matter lesions in the brains of the infants. This suggests a role for IL-6 as a mediator in both placenta and white matter response to inflammation14,17. We chose to study IL-1β and TNF-α because they may be involved in the pathogenesis of periventricular leukomalacia16. IL-6 was studied because it has been identified in the amniotic fluid as a suitable marker for microbial invasion of the amniotic cavity14,21,22. Our results were well above what is considered to be normal values, even in preterm labour14,21,22. Our data have refined the association between increased cytokine levels and the severity of placental histological lesions of chorioamnionitis14,21. We have also identified IL-1β as being a more sensitive marker of chorioamnionitis and a better predictor of vascular invasion in the placenta than IL-6. Amniotic fluid levels of TNF-α appear to be the best predictor of early neonatal sepsis. This does not concur with previous studies, which considered IL-6 levels in the amniotic fluid or umbilical cord plasma of newborn infants to be the best markers of intrauterine infection14,16,21 However, since the increase in both TNF-α and IL-1β plasma concentrations occur before that of IL-6 in response to infection21, this discrepancy could be explained by the fact that our samples were taken early in the course of clinical and or biological intrauterine infection as part of our standardised protocol to diagnose chorioamnionitis in pretenn labour. Moreover, IL-1β has also been identified as a valuable marker of preterm parturition even without evidence of intrauterine infection24–27.
Although the incidence of periventricuar leukomalacia in our population was similar to that reported by Yoon et al.,17 stringent criteria for the diagnosis of chorioamnionitis and that of periventricular leukomala- cia, as well as stratification of the severity of the intrauterine inflammatory process failed to identify any correlation between the amniotic fluid levels of cytolunes and the development of periventricular leuko- malacia in the newborns. Normal levels of TNF-α, IL-6 or IL-1β were found in the amniotic fluid of all but one infants (n= 5) who subsequently developed cystic periventricular leukomalacia. The excluded infant who developed late onset periventricular leukomalacia was delivered from a mother with normal amniotic fluid cytokine levels. Conversely, all infants born from mothers with very high levels of the three selected cytokines had normal serial brain examination. The timing in cytokine production in response to intrauterine infection cannot explain the favourable neurological outcome following high levels of amniotic fluid. A prospective study is more likely to control for confounding variables between intrauterine infection and other risk factors for white matter lesions in very preterm infants such as vaginal bleeding, prolonged premature rupture of the membranes or labour8,28. Our diagnostic criteria of chorioamnionitis might have been more stringent and stratified, and the studied population is therefore likely to be more homogeneous than those described in previous series. Tocolysis was used more often in patients with chorioamnionitis, and this is likely to reflect the severity of preterm labour with intrauterine infection. However tocolysis was stopped as soon as the result of the gram stain examination of the amniotic fluid sampled by amniocentesis on admission was known to be positive.
In conclusion, this study has refined the value of amniotic fluid cytokine levels, particularly IL-1β, in the prenatal diagnosis of chorioamnionitis and the prediction of severe neonatal sepsis in a high risk population. However, elevated amniotic fluid cytokine levels were not associated with the occurrence of white matter damage in preterm infants as previously suggested. This should stimulate the study of other markers to predict the development of white matter lesions involving apoptosis and/or excitotoxic process. Cytokine levels in the amniotic fluid are unlikely to have a role in clinical obstetrics until cost and technical issues allowing for quick results can be overcome. However the close correlation between cytokine levels in the amniotic fluid and maternal serum C-reactive protein concentrations should strengthen the role of this marker in the diagnosis and management of chorioamniotis.
This study was funded by La Délégation la Recherche Clinique de l'Assistance Publique-Hôpitaux de Paris. We are indebted to Alain Portier for his technical assistance.