Prediction of microbial invasion of the amniotic cavity in women with preterm labour: analysis of multiple proteins in amniotic and cervical fluids

Authors

  • R-M Holst,

    1. Perinatal Centre, Department of Obstetrics and Gynaecology, Institute for Clinical Sciences, Sahlgrenska Academy, Sahlgrenska University Hospital/East, Göteborg, Sweden
    Search for more papers by this author
  • H Hagberg,

    1. Perinatal Centre, Department of Obstetrics and Gynaecology, Institute for Clinical Sciences, Sahlgrenska Academy, Sahlgrenska University Hospital/East, Göteborg, Sweden
    2. Institute of Reproductive and Developmental Biology, Hammersmith Campus, Imperial College, London, UK
    Search for more papers by this author
  • U-B Wennerholm,

    1. Perinatal Centre, Department of Obstetrics and Gynaecology, Institute for Clinical Sciences, Sahlgrenska Academy, Sahlgrenska University Hospital/East, Göteborg, Sweden
    Search for more papers by this author
  • K Skogstrand,

    1. Department of Clinical Biochemistry, Statens Serum Institut, Copenhagen, Denmark
    Search for more papers by this author
  • P Thorsen,

    1. Perinatal Centre, Department of Obstetrics and Gynaecology, Institute for Clinical Sciences, Sahlgrenska Academy, Sahlgrenska University Hospital/East, Göteborg, Sweden
    Search for more papers by this author
  • B Jacobsson

    1. Perinatal Centre, Department of Obstetrics and Gynaecology, Institute for Clinical Sciences, Sahlgrenska Academy, Sahlgrenska University Hospital/East, Göteborg, Sweden
    2. Department of Obstetrics and Gynaecology, Rikshospitalet, Oslo, Norway
    Search for more papers by this author

Dr B Jacobsson, Department of Obstetrics and Gynaecology, Sahlgrenska Academy, Sahlgrenska University Hospital/Östra, SE-41685 Göteborg, Sweden. Email bo.jacobsson@obgyn.gu.se

Abstract

Please cite this paper as: Holst R-M, Hagberg H, Wennerholm U-B, Skogstrand K, Thorsen P, Jacobsson B. Prediction of microbial invasion of the amniotic cavity in women with preterm labour: analysis of multiple proteins in amniotic and cervical fluids. BJOG 2011;118:240–249.

Objective  Microbial invasion of the amniotic cavity is a major cause of preterm delivery and the diagnosis is dependent on invasive amniocentesis. The objective was to determine whether specific proteins in amniotic and cervical fluids alone, or in combination, could identify bacterial invasion.

Design  A prospective follow-up study.

Population  Women with singleton pregnancies presenting with preterm labour between 22 and 33 weeks of gestation (n = 89).

Setting  Sahlgrenska University Hospital, Gothenburg, Sweden.

Methods  Amniotic and cervical fluid was analysed with polymerase chain reaction for Mycoplasmas, and was cultured for aerobic and anaerobic bacteria. Twenty-seven proteins were analysed using multiplex technology. Individual levels of each protein were compared in order to find associations between different proteins and microbial invasion of the amniotic cavity. Predictive models based on multiple proteins were created using stepwise binary logistic regression.

Main outcome measure  The main outcome measure was microbial invasion of the amniotic cavity.

Results  Microbial invasion of the amniotic cavity was present in 17% (15/89) of the women. Concentration levels of several amniotic and cervical proteins were significantly higher in women with microbial invasion of the amniotic cavity. Three multivariate predictive models were found. The predictive power of the non-invasive model (73% sensitivity, 88% specificity, 55% positive predictive value, 94% negative predictive value) was as good as the invasive models. Area under the receiver operating characteristic (ROC) curve and likelihood ratio were 0.87 and 6.0, respectively.

Conclusions  Prediction of intra-amniotic infection using selected cervical proteins was equally good as prediction using the same proteins collected from amniotic fluid, or a combination of cervical and amniotic proteins.

Introduction

Spontaneous preterm delivery is considered to be a syndrome caused by multiple pathological processes that activate the terminal pathway of parturition.1 Evidence suggests that infection and inflammation are implicated in the mechanisms of spontaneous preterm delivery and preterm labor (PTL), as well as in fetal injury.2,3 Histologic chorioamnionitis is inversely related to gestational age and is highest among pregnancies ending between 20 and 24 weeks of gestation.4 In the majority of cases this condition is of subclinical nature and cannot be diagnosed until after delivery, or by using invasive methods such as amniocentesis or cordocentesis, which can be hazardous to the pregnancy. The gold standard for identification of intrauterine infection to date has been the isolation of microbes in amniotic fluid sampled by amniocentesis. Less than 1% of the amniotic fluid of women in labour at term contains bacteria.3 Therefore, the isolation of any microbes in amniotic fluid is considered a pathological finding, known as microbial invasion of the amniotic cavity (MIAC). Microbiological studies suggest that intrauterine infection accounts for as much as 25–45% of spontaneous preterm deliveries.5 Using molecular microbiological techniques additional bacterial footprints have been detected in as many as 60% of women delivering preterm.6 Normal functions of the female reproductive tract such as parturition and cervical ripening are dependant on inflammatory processes that are regulated by pro- and anti-inflammatory proteins.7 Data indicate firmly that the colonisation of microbes and/or inflammation of the chorio–decidual interface induce the production of a cascade of cytokines and other proteins, leading to the activation of neutrophils and the release of uterotonins such as prostaglandins, resulting in spontaneous preterm delivery.8 It is therefore of great importance to study such proteins and their relationship with MIAC. It is of special interest to find markers in the cervical fluid that can predict intra-amniotic infection/inflammation in order to avoid invasive sampling, as few biomarkers are available today, and have been limited in their predictive capacity.9,10

A new, multiplexed sandwich immunoassay has been developed based on flowmetric xMAP technology. This technology makes it possible to analyse an array of proteins simultaneously using only small sample volumes. The xMAP technology has been used to analyse multiple inflammatory markers and neurotrophins in neonatal dried blood spots, and from maternal plasma and cervical–vaginal mucous in women with PTL.11,12 In the current study we used this technology to analyse a panel of 27 candidate proteins in amniotic and cervical fluid.

The aim of the study was to determine whether these selected proteins (cytokines, chemokines and neurotrophins) in amniotic and cervical fluid are related to MIAC, and whether alone or in combination with each other, and/or with risk factors like cervical length, cigarette smoking and previous spontaneous preterm delivery, and/or late abortion, they could predict MIAC.

Methods

In this prospective cohort study, 89 women with singleton pregnancies presenting in PTL at Sahlgrenska University Hospital/East, Gothenburg, Sweden, between 1996 and 2005, and between 22 weeks and 0 days and 33 weeks and 6 days of gestation, were enrolled. PTL was defined as regular uterine contractions (at least two uterine contractions every 10 minutes for 30 minutes or longer, confirmed by external tocometry), in combination with one of three cervical changes documented by digital examination (1, ≤2 cm in length and ≥1 cm of dilatation; 2, ≤2 cm in length and cervical softening; 3, ≥1 cm dilatation and cervical softening) and/or a cervical length of <30 mm measured by transvaginal sonography (TVS). All women had intact membranes at enrollment.

Cervical ripening was assessed by digital examination. Cervical length was measured by TVS in a standardised way, as described in our previous work.13 Women with preterm prelabour rupture of membranes (PPROM), known uterine malformations, fetal malformations, significant vaginal bleeding, imminent delivery, cervical cerclage or fetal distress were excluded.

Gestational age was determined by routine ultrasound in the second trimester (at 16–19 weeks of gestation) in all women, with the exception of three who had gestational age determined by the date of their last menstrual period. Tocolytic therapy (intravenous terbutaline and/or indomethacin, the latter if the gestational age was <28 weeks) was administered according to local protocol. Corticosteroids were administered (betamethasone 12.5 mg × 2) to encourage fetal lung maturity in 75 of the 89 women. Twenty-four patients were administered with corticosteroids before sampling the amniotic and cervical fluid. Cervical and amniotic samples were collected at the same time in all patients. Cervical mucus was obtained from the women with a Cytobrush (Cytobrush Plus GT; Medscan Medical AB, Malmö, Sweden) from the external cervical os. The cervical mucus was weighed and kept in a refrigerator (+4°C) until processesing, which occurred within 5 hour. The Cytobrush with the cervical mucus was submerged in 1.0 ml of NaCl (9 mg/ml), shaken for 30 minutes at +4°C, followed by centrifugation at 855 × g at +4°C for 10 minutes and stored at −80°C until analysis.

Ultrasound-guided transabdominal amniocentesis, aspirating 30–50 ml of amniotic fluid, was performed under antiseptic conditions within 12 hours of admission. After retrieval, the AF was immediately placed in a refrigerator (+4°C) and centrifuged, within 5 hours of sampling, at 855 × g in +4°C for 10 minutes. The supernatant was stored at −80°C until analysis.

A sample of uncentrifuged amniotic fluid was immediately transported to the microbiological laboratory for polymerase chain reaction (PCR) analysis of Ureaplasma urealyticum and Mycoplasma hominis, and for aerobic and anaerobic culture. MIAC was defined as a positive PCR and/or growth of any bacteria in the amniotic fluid, except for coagulase-negative Staphylococcus, which was considered to be a contamination from skin. However, coagulase-negative Staphylococcus in amniotic fluid from patients with high intra-amniotic levels of interleukin-6 (IL-6) was considered to indicate MIAC.9,13–17 No women were induced or subjected to caesarean section purely for MIAC.

Concentrations of inflammatory markers IL-1β, IL-2, IL-4, IL-5, IL-6, IL-8, IL-10, IL-12, IL-17, IL-18, soluble IL-6 receptor α, interferon-γ (IFN-γ), tumour necrosis factor-α (TNF-α), TNF-β, monocyte chemotactic protein-1 (MCP-1), transforming growth factor-β (TGF-β), macrophage inflammatory protein-1α (MIP-1α), MIP-1β, matrix metalloproteinasis-9 (MMP—9), triggering receptor expressed on myeloid cells-1 (TREM-1), brain-derived neurotrophic factor, granulocyte macrophage colony stimulating factor (CSF), neurotrophin-3, neurotrophin-4, soluble TNF receptor I, migration inhibitory factor (MIF), regulated on activation, normal T-expressed and secreted (RANTES) in amniotic and cervical fluid were analysed at Statens Serum Institute (Department of Clinical Biochemistry, Copenhagen, Denmark) using a multiplex sandwich immunoassay based on flowmetric xMAP technology, as previously described.11 The xMAP technology is based on flowmetric analysis of microbeads that act as solid support for individual assay reactions incorporating a common fluorophore reporter. Assays for several analytes can be made simultaneously on different sets of beads with unique fluorescence characteristics. This technique makes it possible to measure a broad panel of proteins with small sample volumes.

The amniotic and cervical samples were measured undiluted in duplicate. A 50-μl portion of the sample were added to each filter plate well with 50 μl of a suspension of capture-antibody-conjugated beads, with 1500 beads per analyte. After 1.5 hours of incubation, the beads were washed twice and subsequently reacted for 1.5 hours with a mixture (50 μl) of relevant biotinylated detection antibodies, each diluted 1:1000; next, 50 μl of streptavidin-phycoerythrin, 20 μg/ml, was added to the wells. Incubation was then continued for an additional 30 minutes. Finally, the beads were washed twice and re-suspended in 125 μl of buffer and analysed.

In both amniotic and cervical fluid the mean intra-assay coefficient of variation (CV) was 6%, and the mean inter-assay CV was 12%. We used a defined working range as described by Skogstrand et al.11 instead of the more commonly used signal-to-noise ratio (limit of detection), as this was considered a more precise way of defining the sensitivity as it was not possible to obtain amniotic and cervical fluid depleted of cytokines. Thus the detection level was set as half the lowest concentrations in the working range (in both amniotic and cervical fluid): IL-β (40 pg/ml), IL-2 (4 pg/ml), IL-4 (4 pg/ml), IL-5 (4 pg/ml), IL-6 (40 pg/ml), IL-8 (40 pg/ml), IL-10 (10 pg/ml), IL-12 (4 pg/ml), IL-17 (4 pg/ml), IL-18 (40 pg/ml), soluble IL-6 receptor α (2500 pg/ml), INF-γ (4 pg/ml), TNF-α (4 pg/ml), TNF-β (4 pg/ml), MCP-1 (156 pg/ml), TGF-β (4 pg/ml), MIP-1α (40 pg/ml), MIP-1β (40 pg/ml), MMP-9 (5000 pg/ml), TREM-1 (100 pg/ml), brain-derived neurotrophic factor (BDNF) (10 mg/ml), granulocyte-macrophage-CSF (4 pg/ml), neurotrophin-3 (40 pg/ml), neurotrophin-4 (4 pg/ml), soluble TNF receptor-I (156 pg/ml), MIF (100 pg/ml) and RANTES (40 pg/ml).12,14

The study was approved by the local ethics committee at the University of Gothenburg (nos 349-95, 476-05). All patients gave their informed consent before enrollment.

For comparisons between women with and without MIAC, the Mann–Whitney U-test was used for continuous variables, and Fisher’s exact test was used for dichotomous variables. Receiver operating characteristic (ROC) curves, and the area under the curve for prediction of MIAC, were computed for each marker and for demographic variables. Each of these variables was dichotomised from the ROC curve to get the optimal prediction of MIAC. Logistic regressions with these dichotomised independent variables were used to obtain odds ratios (ORs). The significant univariate variables (< 0.01) were entered into a stepwise logistic regression with a maximum of two variables in the final model. A Spearman’s rank correlation test was used to measure associations between continuous variables. P < 0.05 or a 95% confidence interval (95% CI) not including 1 was considered to be statistically significant. All analyses were computed using sas v9.1 or v9.2 (SAS Institute Inc., Cary, NC, USA) or StatView v5.01 (SAS Institute Inc.).

Results

The study population consisted of 89 women with singleton pregnancies. The majority of the women were socially and financially strong, and included white (84%), Asian (10%), Hispanic (4%) and African (2%) women. Patients were included in the study at a median of 30 weeks plus 5 days of gestation, and the median gestational age at delivery was 34 weeks plus 3 days of gestation. Delivery within 7 days occurred in 38% (34/89) of women, and 47% (42/89) of the women delivered before 34 weeks of gestation.

Demographic characteristics are presented in Table 1. Seventeen percent (15/89) of the women had MIAC and two of the 15 patients were positive for two different microbes. The microbes isolated in the 15 cases are presented in Table 2. Women who had MIAC did not differ from those who did not have MIAC regarding maternal age, parity, number of previous gestations, number of previous spontaneous preterm deliveries, smoking habits, cervical length and gestational age at study inclusion. The median gestational age at delivery was significantly lower in the women with MIAC (31 weeks plus 2 days) than in those without (34 weeks plus 6 days) (P < 0.01), and the number of women delivering within 7 days of sampling was significantly higher among patients with MIAC (P = 0.02). The interval from sampling to spontaneous preterm delivery was strongly related to MIAC: the median latency in days from sampling to delivery in the group of women with MIAC was 2 days, versus 24.5 days in the group without MIAC (P < 0.01).

Table 1.   Clinical background variables in women with MIAC compared with women with no MIAC
VariableMIAC n = 15No MIAC n = 74P
  1. Statistical evaluations were performed with a Mann–Whitney U-test or Fisher’s exact test.

Maternal age (years, median, range)28 (21–35)29 (17–44)0.55
Nulliparous13/1541/740.09
Number of previous gestations (median, range)1 (0–7)1 (0–8)0.95
Gestational age at sampling (weeks + days, median, range)30 + 6 (23 + 2 to 33 + 2)30 + 4 (22 + 2 to 33 + 5)0.81
Gestational age at delivery (weeks + days, median, range)31 + 2 (23 + 3 to 39 + 5)34 + 6 (23 + 2 to 43 + 0)<0.01
Latency (days, median, range)2 (0–70)24.5 (1–125)<0.01
Numbers delivering within 7 days from sampling10240.02
Numbers delivering before 32 weeks of gestation8210.07
Numbers delivering before 34 weeks of gestation9330.40
Cervical length (mm, median, range)11.5 (0–28)17.0 (0–43)0.13
Previous spontaneous preterm delivery or late abortion1/1520/740.11
Smoking at the beginning of pregnancy5/1510/740.12
Corticosteroids given before sampling3210.75
Table 2.   Microbes isolated from the amniotic fluid in 15 patients: two patients had two microbes isolated in the amniotic fluid
Microbe
  1. *All Coagulase-negative Staphylococcus had intra-amniotic inflammation, defined as high levels of IL-6.

2 Ureaplasma urealyticum
4 Anaerobic gram-negative rods
1 Corynebacterium
1 Peptostreptococcus
1 Listeria monocytogenes
1 Alpha streptococcus
1 Actinomyces odontolyticus
1 Snethia sanguinegens
1 Streptococcus mitis
1 Difteriodic rods
3 Coagulase-negative Staphylococcus*

The median concentrations of all proteins in amniotic and cervical fluid in relation to MIAC, and no MIAC, and area under the curve together with OR, 95% CI and P values are presented in Tables 3 and 4. The diagnostic indices for a single protein in the prediction of MIAC in this study (data not shown) were not as good as for the combination of various markers in the multivariate models.

Table 3.   Comparison of levels of cytokines and neuropeptides (median, range, ng/ml) found in amniotic fluid in the group with MIAC and in the group without MIAC using the Mann–Whitney U-test
VariableMIAC (n = 15)No MIAC (n = 74)PAUCCut-off value ng/ml% above cut-offMIAC OR (95% CI)P
  1. ROC curve analysis calculated the best cut-off value for every single protein in the prediction of MIAC with the area under the curve (AUC) and the odds ratio (95% CI). Significant results are marked in bold.

IL-1β0.479 (0.04–14.921)0.040 (0.040–7.431)<0.010.711.7561215.3 (3.7–64.0)<0.01
IL-20.164 (0.022–3.343)0.190 (0.011–3.796)0.960.500.107661.5 (0.4–5.1)0.53
IL-40.015 (0.004–0.06)0.011 (0.004–0.493)0.320.420.06110.5 (0.1–4.4)0.55
IL-50.052 (0.004–0.687)0.033 (0.004–2.454)0.140.620.17193.8 (1.1–12.9)0.03
IL-626.161 (0.173–40.0)10.096 (0.123–59.393)0.010.7124.673010.0 (2.8–35.5)<0.01
IL-103.264 (0–24.795)0.472 (0.01–25.324)0.060.663.2641613.0 (3.5–48.2)<0.01
IL-120.049 (0.004–0.416)0.024 (0.004–0.601)0.560.550.049313.1 (1.0–9.6)0.05
IL-170.004 (0.004–2.308)0.004 (0.004–1.818)0.360.570.209213.1 (0.9–10.3)0.06
IL-180.438 (0.04–1.309)0.107 (0–2.156)<0.010.760.298219.6 (2.8–32.8)<0.01
sIL-6Rα6.269 (0–32.947)7.150 (0–564.897)0.690.537.836460.7 (0.2–2.3)0.61
IFN-γ0.317 (0.004–1.901)0.082 (0.004–5.029)0.260.590.302344.1 (1.3–12.8)0.02
TNF-α0.236 (0.004–2.515)0.078 (0.004–1.609)<0.010.750.626120.7 (4.5–96.3)<0.01
TNF-β0.004 (0.004–0.232)0.004 (0.004–1.203)0.790.480.23271.0 (0.1–9.1)0.99
MCP-16.732 (1.213–57.855)4.721 (0.313–160.0)0.330.4214.67220.8 (0.2–3.3)0.80
TGF-β0.004 (0.004–1.165)0.004 (0.004–1.337)0.440.550.018241.8 (0.5–6.1)0.33
MIP-1α4.256 (0.040–40.0)0.538 (0–38.688)0.040.674.256225.9 (1.8–19.4)<0.01
MIP-1β4.301 (0.193–14.542)0.467 (0.04–44.354)0.020.693.537247.7 (2.3–25.8)<0.01
TREM-15.919 (0.098–146.882)3.551 (0–96.506)0.360.585.919332.9 (0.9–9.0)0.07
BDNF2.568 (0.13–14.133)2.420 (0.010–76.159)0.450.4414.13380.8 (0.1–7.3)0.85
GM-CSF0.39 (0.03–3.178)0.281 (0.004–18.858)0.180.610.33483.6 (1.1–12.4)0.04
NT-40.064.0 (0.005–0.244)0.053 (0.004–0.926)0.720.470.006980.2 (0.0–3.3)0.25
NT-36.651 (0.095–97.169)6.875 (0.04–279.861)0.990.500.142980.2 (0.0–3.3)0.25
sTNF RI22.668 (2.64–71.34)28.152 (1.793–115.146)0.600.5416.715690.6 (0.2–2.0)0.44
MIF54.639 (7.035–128.802)62.919 (5.324–118.033)0.880.51114.705418.2 (1.8–190.2)0.02
RANTES4.250 (0.382–39.465)1.058 (0.04–57.278)<0.010.712.729345.4 (1.6–17.7)0.01
Table 4.   The levels of cytokines and neuropeptides (median, range, ng/ml) in cervical fluid were compared in the MIAC group and the no-MIAC group using the Mann–Whitney U-test
VariableMIAC (n = 15)No MIAC (n = 74)PAUCCut-off value ng/ml% above cut-offMIAC OR (95% CI)P
  1. ROC curve analysis calculated the best cut-off value for every single protein in the prediction of MIAC with the AUC and the odds ratio (95% CI). Significant results are marked in bold.

IL-1β12.635 (0.56–108.51)6.423 (0–40.0)0.040.6714.431177.2 (2.1–25.2)<0.01
IL-20.087 (0.004–0.794)0.038 (0.004–0.626)0.030.680.064384.3 (1.3–14.2)0.01
IL-40.008 (0.004–0.10)0.004 (0.004–0.201)<0.010.710.008275.9 (1.8–19.2)<0.01
IL-50.004 (0.004–0.074)0.004 (0–0.281)0.240.580.017192.6 (0.7–8.9)0.13
IL-640.0 (1.403–40.0)5.096 (0.04–40.0)<0.010.789.6074412.0 (2.5–57.3)<0.01
IL-100.143 (10.0–1.442)0.080 (0.01–2.626)0.150.620.115493.4 (1.0–11.7)0.05
IL-120.017 (0.004–0.228)0.004 (0.004–0.335)0.010.680.015304.7 (1.5–14.9)<0.01
IL-170.049 (0.004–0.649)0.013 (0.004–0.768)<0.010.750.0253412.4 (3.2–49.1)<0.01
IL-186.165 (0.735–40.0)4.594 (0.04–40.0)0.580.4512.817310.5 (0.1–1.9)0.30
sIL-6Rα16.558 (2.500–96.402)2.5 (0–47.333)<0.010.7610.498376.5 (1.9–22.6)<0.01
IFN-γ0.048 (0.004–1.026)0.031 (0.004–2.002)0.050.660.005698.0 (1.0–64.6)0.05
TNF-α0.149 (0.004–1.150)0.052 (0.004–2.794)0.060.660.081473.8 (1.1–13.1)0.03
TNF-β0.045 (0.004–3.594)0.016 (0.004–0.641)0.040.670.028537.6 (1.6–36.3)0.01
MCP-15.170 (0.548–160.0)2.110 (0.313–160.0)<0.010.712.6904920.5 (2.6–164.5)<0.01
TGF-β0.048.0 (0.004–0.303)0.046 (0.004–1.081)0.950.510.303160.3 (0.0–2.8)0.31
MIP-1α7.087 (0.526–24.841)2.464 (0.04–56.394)0.080.654.674464.0 (1.2–13.9)0.03
MIP-1β8.250 (0.360–15.037)2.609 (0.04–40.0)0.070.653.079534.5 (1.2–17.1)0.03
TREM-12.026 (0–18.333)0.624 (0–14.223)0.030.681.380376.5 (1.9–22.6)<0.01
BDNF2.487 (0.172–23.895)1.492 (0.227–24.506)0.270.591.673513.2 (0.9–11.1)0.06
GM-CSF0.318 (063–1.191)0.139 (0.004–1.7)0.030.680.291295.0 (1.6–16.1)<0.01
NT-40.026 (0.004–0.241)0.014 (0.004–0.403)0.110.630.025304.7 (1.5–14.9)<0.01
NT-33.876 (0.146–28.167)1.093 (0.040–37.545)0.020.690.907576.2 (1.3–29.2)0.02
sTNF RI17.547 (0.493–43.699)9.616 (0.313–39.843)<0.010.739.897547.2 (1.5–34.4)0.01
MIF21.106 (2.4–100.0)18.012 (0–100.0)0.460.5611.597642.6 (0.7–9.9)0.17
RANTES3.162 (0.634–44.332)2.667 (0.04–53.027)0.170.612.347573.6 (0.9–13.8)0.06

The IL-8 and MMP-9 results from both amniotic and cervical fluid were excluded from statistical calculations because of methodological problems. Their levels were in the upper range of the standard curve, but the limited amount of residual sample volume did not permit further testing.

Amniotic fluid

The women who had MIAC had significantly higher median levels of IL-1β, IL-6, IL-18, TNF-α, MIP-1α, MIP-1β and RANTES (Table 3) than did women without MIAC. The highest values of the area under the ROC curve were detected for IL-18 (area under the curve = 0.76), TNF-α (area under the curve = 0.75), IL-1β (area under the curve = 0.71), IL-6 (area under the curve = 0.71) and RANTES (area under the curve = 0.71).

Cervical fluid

In cervical fluid IL-1β, IL-2, IL-4, IL-6, IL-12, IL-17, soluble IL-6 receptor α, IFN-γ, TNF-β, MCP-1, TREM-1, granulocyte-macrophage-CSF, neurotrphin-3 and soluble TNF receptor I (Table 4) were significantly higher in the group of women with MIAC than in the women without MIAC. The highest values of the area under the curve were detected for IL-6 (area under the curve = 0.78), IL-17 (area under the curve = 0.75), soluble IL-6 receptor α (area under the curve = 0.76), sTNF RI (area under the curve = 0.73) and MCP-1 (area under the curve = 0.71).

Comparison between amniotic and cervical fluid

The correlations between individual proteins in the two different compartments were generally low. However, there were significant correlations between the levels in amniotic and cervical fluid of IL-1β (ρ = 0.42, P < 0.01), IL-5 (ρ = 0.21, P = 0.05), IL-6 (ρ = 0.38, < 0.01), IL-10 (ρ = 0.23, = 0.03), IL-17 (ρ = 0.25, = 0.02), soluble IL-6 receptor α (ρ = 0.22, = 0.05), TNF-β (ρ = 0.24, = 0.03), TGF-β (ρ = 0.21, = 0.05), MIP-1α (ρ = 0.45, < 0.01) and MIP-1β (ρ = 0.414, < 0.01).

Models for predicting MIAC

To predict which women with symptoms of PTL also had MIAC, protein levels from both amniotic and cervical fluid along with background variables (cervical length, smoking, previous spontaneous preterm delivery) were included in a multivariable analysis. Variables significant at < 0.01 in the univariable analyses were entered into a forward selection logistic regression, and a maximum of two variables in the final model were used to construct predictive models using variables in amniotic (Table 5) and cervical fluid (Table 6) separately, and in a model combining proteins from the two different compartments (Table 7). None of the demographic variables, cervical length, previous preterm delivery, and/or late abortion or maternal smoking, were significant. In the multivariate model analysing proteins in amniotic fluid (Table 5), TNF-α (0.62 ng/ml; OR 15.9, 95% CI 2.1–122.2), and IL-6 (24.7 ng/ml; OR 8.3, 95% CI 1.7–40.6) contributed significantly to the prediction of MIAC. The area under the curve was 0.86 (Figure 1). In the non-invasive prediction model using cervical fluid proteins (Table 6), IL-17 (0.025 ng/ml; OR 6.1, 95% CI 1.2–31.8) and MCP-1 (2.7 ng/ml; OR 32.3, 95% CI 2.4–433.9) contributed significantly to the prediction of MIAC. The area under the curve was 0.87 (Figure 2) and the likelihood ratio was 6.0. In the combined prediction model using proteins from both amniotic and cervical fluid (Table 7), amniotic TNF-α (0.62 ng/ml; OR 29.4, 95% CI 1.5–551.2) and cervical MCP-1 (2.7 ng/ml; OR 44.4, 95% CI 3.0–648.4) contributed to the prediction of MIAC (area under the curve = 0.85, Figure 3). Thus we found that the results of the non-invasive model using markers from the cervical fluid were comparable with the amniotic fluid or the mixed amniotic and cervical models.

Table 5.   Odds ratio (95% CI) for prediction of MIAC using a multivariate model with high levels (ng/ml) of TNF-α and IL-6 in amniotic fluid. Predictive values including the likelihood ratio for a prediction model combining these two markers are also shown
VariableMIAC OR (95% CI)
High TNF-α amniotic fluid (0.62 ng/ml)15.9 (2.1–122.2)
High IL-6 amniotic fluid (24.7 ng/ml)8.3 (1.7–40.6)
 Predictive values
Correctly predicted/total number70/89 (79%)
Sensitivity13/15 (87%)
Specificity57/74 (77%)
Positive predictive value13/30 (43%)
Negative predictive value57/59 (97%)
Likelihood ratio (sensitivity/1-specificity)3.8
Table 6.   Odds ratio (95% CI) for prediction of MIAC using a multivariate model combining high levels (ng/ml) of IL-17 and MCP-1 in cervical fluid. Predictive values including the likelihood ratio for a prediction model combining these two markers are also shown
VariableMIAC OR (95% CI)
High IL-17 cervical fluid (0.025 ng/ml)6.1 (1.2–31.8)
High MCP-1 cervical fluid (2.7 ng/ml)32.3 (2.4–433.9)
 Predictive values
Correctly predicted/total number76/89 (85%)
Sensitivity11/15 (73%)
Specificity65/74 (88%)
Positive predictive value11/20 (55%)
Negative predictive value65/69 (94%)
Likelihood ratio (sensitivity/1-specificity)6.0
Table 7.   Odds ratio (95% CI) for prediction of MIAC using a multivariate model combining high levels (ng/ml) of TNF-α in amniotic fluid and MCP-1 in cervical fluid. Predictive values including the likelihood ratio for a prediction model combining these two markers from the two compartments are also shown
VariableMIAC OR (95% CI)
High TNF-α amniotic fluid (0.62 ng/ml)29.4 (1.5–551.2)
High MCP-1 cervical fluid (2.69 ng/ml)44.4 (3.0–648.4)
 Predictive values
Correctly predicted/total number58/89 (65%)
Sensitivity14/15 (93%)
Specificity44/74 (60%)
Positive predictive value14/44 (32%)
Negative predictive value44/45 (98%)
Likelihood ratio (sensitivity/1-specificity)2.3
Figure 1.

 Receiver operating characteristic curve analysis for the combination of amniotic fluid levels of TNF-α (cut-off level 0.062 ng/ml) and IL-6 (cut-off level 24.67 ng/ml). The area under the curve was 0.86.

Figure 2.

 Receiver operating characteristic curve analysis for the combination of cervical fluid concentrations of IL-17 (cut-off level 0.025 ng/ml) and MCP-1 (cut-off level 2.7 ng/ml). The area under the curve was 0.87.

Figure 3.

 Receiver operating characteristic curve analysis for the combination of concentrations of TNF-α (cut-off level 0.62 ng/ml) in amniotic fluid and of MCP-1 (cut-off level 2.69 ng/ml) in cervical fluid. The area under the curve was 0.85.

Discussion

In the current study, we examined 27 specific inflammatory proteins in two compartments (amniotic and cervical fluid), together with demographic variables (cervical length, maternal smoking, and previous spontaneous preterm delivery and/or late abortion) and analysed their ability to predict MIAC in women with singleton pregnancies with intact membranes and symptoms of PTL. We found three multivariate predictive models of MIAC that could be validated for clinical use. Our main finding was that the non-invasive prediction model using only IL-17 and MCP-1 in cervical fluid was equally good for predicting MIAC as the same proteins analysed from amniotic fluid. This finding needs to be replicated in a second larger cohort. The detection of bacteria in the amniotic fluid still remains the gold standard definition of MIAC.

The importance of detecting MIAC during pregnancy is supported by evidence that MIAC and chorioamnionitis give rise to preterm delivery, a fetal inflammatory response syndrome,18–20 and that inflammation contributes to neonatal brain injury and subsequent cystic periventricular leukomalacia and cerebral palsy.21 The gold standard for detecting MIAC in an ongoing pregnancy has been an invasive amniocentesis performed in order to obtain amniotic fluid for culture and/or PCR analyses. Histological chorioamnionitis and culture of the membranes are also a measure of MIAC, but cannot be performed in an ongoing pregnancy. There are very few studies available on the safety of third-trimester amniocentesis. A recent study of 111 women with a median gestational age of 36 weeks who underwent amniocentesis for assessment of lung maturity or MIAC were followed-up within 24 hours of amniocentesis, and it was found that that third trimester amniocentesis carried a complication rate of 3.6%.22 Although all complications were self-limited and no perinatal or maternal morbidity was directly linked to the procedure, and no patients with complications required urgent delivery, it is reasonable to assume that the patients suffered some anxiety and discomfort. It is therefore of special interest to find markers that can be collected non-invasively and can predict intra-amniotic infection/inflammation with at least the same precision as markers collected from amniotic fluid. A speculum examination and collection of analytes from the cervical os is a method that offers a simple way of assessing the cervix, and is associated with few if any risks, and has a high acceptance rate with the women involved.23

Several of the proteins (TGF-1β, IL-17, soluble IL-6 receptor α, neurotrophins, IL-4, IL-5, TREM-1, IL-12, IFN-γ) that we analysed from amniotic and cervical fluid have never, to our knowledge, been studied in relation to MIAC, whereas others (e.g. Il-6, TNF-α, IL-1β) have been extensively studied.

The volumes of both amniotic and cervical fluid (especially in the case of PPROM) are often limited. The xMAP technology has opened new possibilities to study an array of inflammatory proteins using only small quantities of analytes. The same technique has been used by others to find proteins that could predict preterm delivery and infection.12,24,25 In these studies maternal serum or plasma were analysed, and the samples were taken in the second trimester. Gargano et al.24 analysed the levels of IL-1β, IL-2, IL-12, IL-18, INF-γ, TNF-α, TNF-β, IL-4, IL-6, IL-10, IL-17, TGF-β and granulocyte-macrophage-CSF in maternal plasma in mid-pregnancy (15–22 weeks of gestation) in 926 women, and related the levels to preterm delivery and histologic chorioamnionitis. They found that IL-1β, IL-2, IL-12, INF-γ, IL-4, IL-6 and TGF-β were significantly associated with preterm delivery before 35 weeks of gestation in women with histologic chorioamnionitis.

We found (using the non-invasive model) that high cervical fluid concentrations of the proteins IL-17 and MCP-1 predicted MIAC with a reasonably high correctly predicted proportion of cases (85%; likelihood ratio, 6.0; negative predictive value, 94%). In a clinical setting where a positive test result will result in delivery (via the induction of labour or caesarean section), the use of the model is questionable as the positive predictive value was still limited (55%). But in a different clinical setting a 55% risk of MIAC may be sufficiently high to justify the administration of antibiotics.

We found previously that IL-17 in both amniotic and cervical fluid was significantly associated with delivery in <7 days, and in these studies, we demonstrate that IL-17 in cervical fluid, but not in amniotic fluid, is strongly associated with MIAC, and that the levels also contribute significantly to a prediction of MIAC using the non-invasive model.14 This is a novel finding that may be important, as IL-17 could play a critical role in the pathophysiology of PTL. It is produced by a novel type of CD4-positive T cell, TH17 cells, under the influence of IL-23 production by macrophages. IL-17 promotes the expansion and recruitment of innate immune cells such as neutrophils,26 and also cooperates with the Toll-like receptor ligands, IL-1β and TNF-α, to enhance inflammatory reactions.27 Its receptor, IL-17RA, is ubiquitously expressed and shares many features with classical innate immune receptors. IL-17 seems to be of particular importance not only in relation to autoimmunity but also to chronic inflammation. This suggests that this cytokine could contribute to prolonging and enhancing the inflammatory response in connection with preterm birth.26,28,29 Secondly, IL-17 expression indicates that both innate and adaptive immunity is activated in preterm labour, as TH17 cells do not respond to IL-23 unless they have been activated by antigen-presenting cells in secondary lymphoid organs.27,30 Thirdly, the data imply that not only the Th1/Th2 balance but also Th17 is of importance in preterm labor.27,31

Monocyte chemotactic protein-1 (MCP-1 or CCL2) in amniotic fluid has been associated with preterm labour, histologic chorioamnionitis, intra-amniotic inflammation and MIAC.9,16,32 Cervical MCP-1 has also been shown to be elevated in preterm parturition,14,16,33 intra-amniotic inflammation and MIAC.16 In the current study we confirm these previous findings and demonstrate that cervical fluid MCP-1 contributed significantly to the prediction of MIAC in the non-invasive logistic regression model. MCP-1 is a pro-inflammatory chemokine produced by monocytes/macrophages, fibroblasts, B-cells, endothelial cells and smooth muscle cells. It increases in response to infection/inflammation, and regulates cytokines and adhesion molecules preferentially in macrophages. Experimental studies suggest that MCP-1 plays an important role in preterm labour.34

The invasive model (including amniotic MIF, TNF-α and IL-6) predicted MIAC with a higher sensitivity (93%) and negative predictive value (98%), but inferior specificity (76%) and overall predictive ability (79%), as compared with predictions based on cervical measurements. The fact that MIF contributes to MIAC prediction was a surprising finding because the median levels of this cytokine did not differ between the MIAC and non-MIAC groups (Table 3). However, when the cases were dichotomised there was a significantly higher proportion of cases with very high MIF in the MIAC group (OR 18.2, 95% CI 1.8–190.2), and this difference in distribution apparently also turned out to contribute substantially to the amniotic fluid multiple regression model (Table 5). Cervical MIF was also found to be significantly associated with delivery within 7 days,14 and amniotic fluid MIF was previously associated with intra-amniotic inflammation, histologic chorioamnionitis and shorter amniocentesis-to-delivery interval in preterm labour with intact membranes. Furthermore, an increased expression of MIF protein and mRNA in chorioamniotic membranes was observed in patients with histologic chorioamnionitis.35

MIF is involved in regulating macrophage trafficking at the fetal–maternal interface,36 but the precise role MIF has in preterm delivery is not yet known. MIF is a critical upstream regulator of innate and acquired immune responses of major importance in relation to sepsis. The administration of MIF increases lethality during endotoxaemia, whereas neutralisation of this cytokine prevents endotoxic shock and death associated with bacterial infection. It is not surprising that IL-6 and TNF-α in amniotic fluid qualified as predictors of MIAC. Both these cytokines have been extensively studied and found to be associated with intra-amniotic infections in patients with preterm labour, as well as with subsequent neonatal morbidity.19,37,38

Conclusions

Using xMAP technology we examined an array of inflammatory proteins in small volumes of amniotic and cervical fluid. Many previously known and unknown proteins were significantly associated with MIAC. Based on this information three different prediction models were calculated using multiple regression with information from amniotic fluid, amniotic and cervical fluid combined (invasive models) or cervical fluid only (non-invasive model). We believe that the non-invasive model based on cervical analysis of IL-17 and MCP-1 is promising, and can be used, after validation by a new independent study, to guide the clinician in distinguishing women at high risk from women at low risk of MIAC, without requiring amniocentesis. This makes it possible to target high-risk women for treatment and care, and to avoid unnecessary potentially harmful treatments.

Disclosure of interests

There are no conflicts of interests among the authors.

Contribution to authorship

All included authors have fulfilled both of the requirements for authorship outlined in the instructions for authors. R-MH and BJ both contributed to the conception and design of the study, to the interpretation of the results and to the writing of the article. R-MH, HH, UBW and BJ recruited and consented the subjects. PT and KS contributed to the statistical analysis, the writing of the article and the interpretation of the results. All authors took part in the evaluation of the data and approved the final article.

Details of ethics approval

The regional ethical committee approved the study (349-95, Ö506-99, EPN Gbg dnr 476/05).

Funding

This work was supported by grants from the Swedish Medical Research Council (VR 2006-3396), Swedish Medical Society (SLS 2008-21198), Swedish government grants to researchers in the public health service (ALFGBG-2863, ALFGBG-11522) and The Göteborg Medical Society.

Acknowledgement

We thank Mattias Molin for his assistance with the statistical analyses.

Ancillary