Prediction of spontaneous preterm delivery in women with threatened preterm labour: a prospective cohort study of multiple proteins in maternal serum

Authors

  • P Tsiartas,

    1. Department of Obstetrics and Gynaecology, Papageorgiou University Hospital, Thessaloniki, Greece
    2. Department of Obstetrics and Gynaecology, Institute for Clinical Sciences, Sahlgrenska Academy, Sahlgrenska University Hospital/Östra, Göteborg, Sweden
    Search for more papers by this author
  • RM Holst,

    1. Department of Obstetrics and Gynaecology, Institute for Clinical Sciences, Sahlgrenska Academy, Sahlgrenska University Hospital/Östra, Göteborg, Sweden
    Search for more papers by this author
  • UB Wennerholm,

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

    1. Department of Obstetrics and Gynaecology, Institute for Clinical Sciences, Sahlgrenska Academy, Sahlgrenska University Hospital/Östra, Göteborg, Sweden
    2. Institute of Reproductive and Developmental Biology, Hammersmith Campus, Imperial College, London, UK
    Search for more papers by this author
  • DM Hougaard,

    1. Department of Clinical Biochemistry and Immunology, Statens Serum Institute, Copenhagen, Denmark
    Search for more papers by this author
  • K Skogstrand,

    1. Department of Clinical Biochemistry and Immunology, Statens Serum Institute, Copenhagen, Denmark
    Search for more papers by this author
  • BD Pearce,

    1. Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, GA, USA
    Search for more papers by this author
  • P Thorsen,

    1. Department of Obstetrics and Gynaecology, Lillebaelt Hospital, Kolding, Denmark
    Search for more papers by this author
  • M Kacerovsky,

    1. Department of Obstetrics and Gynaecology, Institute for Clinical Sciences, Sahlgrenska Academy, Sahlgrenska University Hospital/Östra, Göteborg, Sweden
    2. Department of Obstetrics and Gynaecology, Charles University in Prague, Faculty of Medicine Hradec Kralove, University Hospital Hradec Kralove, Kralove, Czech Republic
    Search for more papers by this author
  • B Jacobsson

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

Dr P Tsiartas, Department of Obstetrics and Gynaecology, Papageorgiou University Hospital, Ring Road, Nea Efkarpia, Thessaloniki 56429, Greece. Email tsiartaspanos@gmail.com

Abstract

Please cite this paper as: Tsiartas P, Holst R, Wennerholm U, Hagberg H, Hougaard D, Skogstrand K, Pearce B, Thorsen P, Kacerovsky M, Jacobsson B. Prediction of spontaneous preterm delivery in women with threatened preterm labour: a prospective cohort study of multiple proteins in maternal serum. BJOG 2012;119:866–873.

Objective  To analyse whether specific proteins in maternal serum and cervical length, alone or in combination, can predict the likelihood that women with intact membranes with threatened preterm labour will deliver spontaneously within 7 days of sampling.

Design  Cohort study.

Setting  Sahlgrenska University Hospital, Gothenburg, Sweden.

Population  Women at between 22 and 33 weeks of gestation with threatened preterm labour (n = 142) admitted to the Sahlgrenska University Hospital, Gothenburg, Sweden, in 1995–2005.

Methods  Maternal serum was tested for 27 proteins using multiplex xMAP technology. Individual levels of each protein were compared, and calculations were performed to investigate potential associations between different proteins, cervical length and spontaneous preterm delivery. Receiver operating characteristic curves were used to find the best cut-off values for continuous variables in relation to spontaneous preterm delivery within 7 days of sampling. Prediction models were created based on a stepwise logistic regression using binary variables.

Main outcome measure  Spontaneous preterm delivery within 7 days.

Results  In order to determine the best prediction model, we analysed models of serum proteins alone, cervical length alone, and the combination of serum proteins and cervical length. We found one multivariable combined model through the data analysis that more accurately predicted spontaneous preterm delivery within 7 days. This model was based on serum interleukin-10 (IL-10) levels, serum RANTES levels and cervical length (sensitivity 74%, specificity 87%, positive predictive value 76%, negative predictive value 86%, likelihood ratio 5.8 and area under the curve 0.88).

Conclusions  A combination of maternal serum proteins and cervical length constituted the best prediction model, and would help determine whether women with threatened preterm labour are likely to deliver within 7 days of measurement.

Introduction

Preterm delivery (PTD) (birth before 37 weeks of gestation) remains a cause of substantial perinatal mortality and long-term morbidity.1 Fetal and neonatal morbidity and mortality rates are strongly associated with gestational age at birth. Specifically, infants born before 32 weeks of gestation are at risk of sequelae.2 In many countries, like the USA, the rates of PTD (12–13%) continue to rise.3 However, in Sweden, the rate has been stable at around 5.5% for the past 30 years.4,5 PTD aetiological factors operate through multiple pathophysiological pathways that overlap and include molecular factors, creating pathophysiological heterogeneity.6

More than 70% of women presenting with symptoms of preterm labour (PTL) do not progress to active labour and delivery.7,8 It is essential to identify pregnant women with threatened PTL who will deliver preterm, and differentiate them from the women who will continue their pregnancies to full term. Unnecessary hospital stays and potential harmful treatments can thus be avoided, and attention can instead be focused on women who are truly at high risk of spontaneous PTD. Attempts to predict PTD based on maternal and biochemical data, and interventions to reduce PTD rates, have been largely unsuccessful.9–11 The necessity of finding reliable prediction models is urgent, and a multiple-markers test indicative of the multifactorial aetiology of PTD is likely to be more successful.12,13

Studies indicate that an increased production of cytokines in maternal and fetal serum, and amniotic and cervical fluid, is involved in both labour at term and PTL.14,15 The study of such proteins is of great importance because the pathophysiology of PTD most likely involves the disruption of molecular networks, as opposed to an isolated abnormality in an individual gene or protein. Thus, it is important to measure multiple biomarkers simultaneously and place these into a biological context. A multiplexed sandwich immunoassay has been developed based on flowmetric xMAP technology. This technology makes it possible to analyse an array of proteins simultaneously using small sample volumes. The xMAP technology has been used to analyse multiple inflammatory markers and neurotrophins in maternal serum in women with threatened PTL.11,15,16

In the current study, we aimed to test the predictive ability of 27 maternal serum proteins alone and in combination with the cervical length for delivery within 7 days of sampling in women with threatened PTL at various gestational ages, using xMAP technology.

Methods

In this prospective cohort study, we enrolled 142 healthy women without major medical problems (i.e. diabetes mellitus, hypertension or pre-eclampsia) and with singleton pregnancies presenting with threatened PTL between 22 weeks plus 0–7 days and 33 weeks plus 6–7 days of gestation at Sahlgrenska University Hospital, Gothenburg, Sweden, from 1996 to 2005. Threatened PTL was defined as regular uterine contractions (at least two uterine contractions every 10 minutes for ≥30 minutes, as confirmed by external tocometry), in combination with one of three cervical changes, documented by digital examination (1, length ≤2 cm and dilatation ≥1 cm; 2, length ≤2 cm and cervical softening; 3, dilatation ≥1 cm and cervical softening), and/or cervical length <30 mm measured by transvaginal sonography. All women had intact membranes at enrolment.

Cervical ripening was assessed by digital examination and cervical length was measured by transvaginal sonography using a standardised method. The cervical length was measured with the woman adopting the dorsal lithotomy position with an empty bladder. When the cervical canal was visualised, the probe was withdrawn to avoid pressure, distortion and elongation of the cervix. A sagittal view showing the entire cervix (endocervix and vaginal cervix), with the echogenic endocervical mucosa along the length of the cervical canal, was obtained. Callipers were used to measure the distance between the notches made by the junction of the anterior and posterior cervical walls at the internal and external os. Three measurements were performed (in millimetres), and the shortest distance was noted.

Women with preterm prelabour rupture of membranes, known uterine malformations, fetal malformations, significant vaginal bleeding, imminent delivery, cervical cerclage or fetal distress were excluded from the study. Because of limited resources, we did not recruit women during the night, at weekends, at Christmas or in the summer months, as well as during busy hours on the delivery ward, which limited the number of women enrolled in the study.

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 in which the gestational ages were determined by the date of their last menstrual periods. Tocolytic therapy (intravenous terbutaline and/or indomethacin, the latter used at <28 weeks of gestation) was administered according to local protocol. A corticosteroid was administered (betamethasone 12 mg × 1 dose × 2 days) to stimulate fetal lung maturity.

Serum samples were drawn from the mother at the time of hospital admission, after informed consent to participate in the study was obtained. The serum samples were placed in a refrigerator (+4°C) within 10 minutes, and were processed within 6 hours before freezing (−80°C) for later analyses.

Concentrations of the following markers were measured (symbols assigned by the HUGO Gene Nomenclature Committee are in parenthesis): interleukin-1β (IL1β), IL-2 (IL2), IL-4 (IL4), IL-5 (IL5), IL-6 (IL6/IFNβ2), IL-8 (IL8), IL-10 (IL10), IL-12 (interleukin 12A – natural killer cell stimulatory factor 1), IL-17 (IL17A), IL-18 (IL18 – interferon-gamma-inducing factor), soluble IL-6 receptor α (sIL6R), interferon-γ (IFNG), tumour necrosis factor-α (TNF, TNF superfamily, member 2), tumour necrosis factor-β (LTA – lymphotoxin alpha – TNF superfamily, member 1), monocyte chemotactic protein-1 [CCL2 – chemokine (C–C motif) ligand 2], transforming growth factor-β (TGFB1), macrophage inflammatory protein-1α [CCL3 – chemokine (C–C motif) ligand 3], macrophage inflammatory protein-1β [CCL4 – chemokine (C–C motif) ligand 4], matrix metalloproteinasis-9 (MMP9), triggering receptor expressed on myeloid cells-1 (TREM1), brain-derived neurotrophic factor (BDNF), granulocyte-macrophage-colony-stimulating factor (CSF2 – colony stimulating factor 2), neurotrophin-3 (NTF3), neurotrophin-4 (NTF4), soluble TNF receptor I (sTNFR1A), migration inhibitory factor (MIF), regulated on activation normal T-expressed and secreted RANTES [CCL5 – chemokine (C–C motif) ligand 5]. Assays of maternal serum were analysed at Statens Serum Institute, Department of Clinical Biochemistry, Denmark, using a multiplex sandwich immunoassay based on flowmetric xMAP technology, as previously described.17,18

The maternal serum was measured in duplicates in a diluted assay buffer. A sample was added to each filter plate well with a suspension of capture-antibody-conjugated beads. After an incubation of 1.5 hours the beads were washed twice and subsequently reacted for 1.5 hours with a mixture of relevant detection antibodies. Next, a quantity of streptavidin-phycoerythrin was added to the wells. Incubation then was continued for an additional 30 minutes. Finally, the beads were washed twice, re-suspended in buffer and analysed.

The detection level in the maternal serum was set as half the lowest concentrations in the working range:17 IL-1β (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), sIL-6Rα (2500 pg/ml), IFN-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), BDNF (10 pg/ml), GM-CSF (4 pg/ml), NT-3 (40 pg/ml), NT-4 (4 pg/ml), soluble TNF RI (156 pg/ml), MIF (100 pg/ml) and RANTES (40 pg/ml). The staff member who performed the analyses (K.S.) did not have any clinical information on the measure of outcome.

The study was approved by the local ethics committee at the University of Gothenburg (nos 349-95, 476-05).

Statistical analysis

Women who did and did not deliver within 7 days were compared using a Mann–Whitney U-test for continuous variables and Fisher’s exact test for dichotomous variables. Receiver operating characteristic (ROC) curves and the area under the curve (AUC) were computed for each protein, cervical length and the background variables for the prediction of delivery within 7 days. The best cut-off values for every protein and cervical length were determined with spss 17.0 and medcalc 11.0.1.0 for ROC curve analysis, based mainly on the calculation of the highest summary of sensitivity plus specificity for each coordinate point in the ROC curve, but at the same time a higher weight was attributed to the sensitivity for the determination of the best cut-off point for every variable. Continuous variables were dichotomised from the ROC curve to obtain the optimal prediction of delivery within 7 days. Crosstabs were used to obtain odds ratios (ORs). The significant variables in univariate analyses (P < 0.05) were entered into a backward stepwise logistic regression. We used 0.05 as our limit in the univariate analyses in order to avoid the problem of mass significance and the use of a large number of predictors exceeding 10. Choosing a different limit than 0.05, for example 0.25, would result in the inclusion of 20 variables as possible predictors in the stepwise selection, which would not be appropriate to common statistical standards. All analyses were performed for each serum marker, the cervical length and the combinations. We also used a Hosmer and Lemeshow test for the goodness of fit of the models. A P value of less than 0.05 or a 95% confidence interval (95% CI) that did not include 1.0 was considered to be statistically significant. All analyses were computed using spss 17.0 (SPSS Inc., Chicago, IL, USA) and medcalc 11.0.1.0 (MedCalc Software, Mariakerke, Belgium).

Results

In our cohort of 142 women with singleton pregnancies, 40% (57/142) delivered within 7 days. Demographic characteristics are presented in Table 1. Women who delivered within 7 days of sampling did not differ from those who delivered later with regard to the number of previous PTDs, body mass index (BMI), smoking status or gestational age at study inclusion. The women who delivered within 7 days were more often nulliparous (70%) than women who delivered later (56%).

Table 1.   Clinical background variables in women delivering within or after 7 days of assessment
VariableWomen delivering ≤7 days n = 57Women delivering >7 days n = 85 P
  1. P value calculated by a Mann–Whitney U-test or a Fisher’s exact test. Q1–Q3 represents the values of the 25th and 75th percentiles. Significant results are marked in bold.

Maternal age (years, median, Q1–Q3)30.5 (28–33.5)29 (24–33)0.190
Nulliparous (n = 88) (61.9%)40 (70%)48 (56%)0.079
Gestational age at sampling (weeks + days, median, range)30 + 0 (22 + 3 to 33 + 5)30 + 5 (22 + 2 to 33 + 5)0.077
Gestational age at delivery (weeks + days, median, range)30 + 5 (23 + 2 to 34 + 0)36 + 6 (24 + 2 to 43 + 0) 0.000
Latency in days (median, range, Q1–Q3)2 (0–7, 1–3.5)42 (8–125, 23–61.5) 0.000
Previous spontaneous preterm delivery (n = 27) (19%)8 (14%)19 (22%)0.278
Smoking at the beginning of pregnancy (n = 22) (15.4%)8 (14%)14 (16.4%)0.815
BMI at the beginning of pregnancy (median, Q1–Q3)22 (20–24)21 (20–26)0.787

The median concentrations of all proteins in maternal serum and the cervical length in relation to delivery within 7 and after 7 days, together with the P values, are presented in Table 2. Only proteins with detectable maternal serum levels in more than 50% of the samples were included in the analyses. Thus, IL-1β, IL-2, IL-5, IL-6 and IL-8 were excluded from further analyses.

Table 2.   Levels of cytokines and neuropeptides (median, Q1–Q3, 25th and 75th percentiles, pg/ml) in maternal serum and the cervical length (median, Q1–Q3, 25th and 75th percentiles, mm) measured at sampling, in relation to delivery within 7 days. P1 refers to the P value from the comparison of the two groups using a Mann–Whitney U-test. Receiver operator characteristic (ROC) curve analysis calculated the best cut-off value and the area under the curve (AUC) for every single protein and cervical length. Crude odds ratios (Ors, 95% CI) for the delivery within 7 days and P2 were calculated using cross tabs. Significant results are marked in bold
VariableDelivery within 7 days median, (Q1–Q3) (n = 57)Delivery after 7 days median, (Q1–Q3) (n = 85)% above detection limit P 1 AUCCut-off value% above cut-offDelivery within 7 days crude OR (95% CI) P 2
IL-45 (4–7)5 (4–7)590.8270.481 pg/ml470.9 (0.4–1.7)0.864
IL-1093 (38.5–176.5)30 (4–71.5)74 <0.001 0.6948 pg/ml50 4.6 (2.2–9.7) <0.001
IL-1211 (4–25)13 (4.5–32.5)740.3570.4519 pg/ml350.6 (0.3–1.3)0.288
IL-1718 (4–51)14 (4–39)610.3310.5434 pg/ml321.8 (0.8–3.7)0.104
IL-18389 (242–528)353 (264–484.5)1000.9500.50268 pg/ml690.5 (0.2–1.1)0.139
Soluble IL-6Rα26 545 (23 000.5–38912.5)24 941 (18 879.5–31 172.5)100 0.050 0.5922 576 pg/ml64 2.7 (1.2–5.8) 0.008
IFN-γ12 (4–34.5)10 (4–36.5)630.3360.5410 pg/ml501.6 (0.8–3.1)0.175
TNF-α18 (4–27)14 (4–33.5)720.9200.5017 pg/ml431.6 (0.8–3.2)0.171
TNF-β64 (19–99)45 (19.5–68)100 0.043 0.6068 pg/ml33 2.9 (1.4–6) 0.004
MCP-1176 (10–414.5)178 (58–433)570.6990.4825 pg/ml760.5 (0.2–1)0.108
TGF-β106 (43–188)89 (40–161.5)1000.2170.56111 pg/ml391.7 (0.8–3.4)0.120
MIP-1α157 (98–274)141 (92.5–229.5)950.3790.54197 pg/ml331.7 (0.8–3.4)0.149
MIP-1β279 (191–381)223 (166–306)100 0.024 0.61271 pg/ml40 2.5 (1.2–5.1) 0.009
MMP-9542 061 (372 503–750 501)301 680 (167 851.5–482 200.0)100 <0.001 0.74469 645 pg/ml41 6.0 (2.9–12.7) <0.001
TREM-11525 (517.0–2563.5)1204 (610–2894)1000.7730.481472 pg/ml461.5 (0.7–2.9)0.236
BDNF5483 (3969.5–8195.0)4581 (3317.0–6156.5)100 0.008 0.636950 pg/ml25 3.7 (1.6–8.2) 0.001
GM-CSF13 (10–36)24 (10–53)100 0.042 0.4021 pg/ml55 0.4 (0.2–0.8) 0.016
NT-410 (4–19)12 (4–25)650.3820.4519 pg/ml290.4 (0.2–1)0.091
NT-382 (37.5–178.5)52 (10–144)660.0760.5860 pg/ml54 2.3 (1.1–4.7) 0.017
Soluble TNF RI1129 (790.5–1129.0)902 (622.5–1253.5)99 0.004 0.64749 pg/ml69 3.7 (1.6–8.5) 0.002
MIF56 320 (19 715.5–208 809.5)21934 (4806–43 160)95 <0.001 0.7145 695 pg/ml35 4.7 (2.2–9.9) <0.001
RANTES42 013 (34 434–68 594)38 504 (28 556.0–46 428.5)100 0.018 0.6149 293 pg/ml27 3.9 (1.8–8.6) <0.001
Cervical length5.5 (0–15)21 (15–26) <0.001 0.7718 mm57 21.2 (6.0–74.7) <0.001

Women who delivered within 7 days had significantly higher levels of IL-10, soluble IL-6Rα, TNF-β, MIP-1β, MMP-9, BDNF, soluble TNF receptor I, MIF and RANTES than did the women who delivered later (Table 2). The highest values of AUC were detected for IL-10 (AUC = 0.69, < 0.001), MMP-9 (AUC = 0.74, < 0.001), MIF (AUC = 0.71, < 0.001) and cervical length (AUC = 0.77, < 0.001).

Based on a stepwise multivariable logistic regression of dichotomous variables, we first analysed the most significant serum proteins from the univariate analyses alone, then cervical length alone and, lastly, the combination of serum proteins along with cervical length to construct the best prediction model. We found that the combined model with proteins and cervical length had the best prediction capacity (Tables 3 and 4). In the combined multivariable model of maternal serum proteins and cervical length, high levels of IL-10 (≥48 pg/ml; OR 5.0, 95% CI 1.7–14.3), RANTES (≥49293 pg/ml; OR 9.9, 95% CI 2.5–39.2) and short cervical length (≤18 mm; OR 43.5, 95% CI 8.9–211.3) were statistically significant contributions to the prediction of spontaneous PTD within 7 days of sampling (Table 3). Using this combined model, the ROC curve (Figure 1) had an AUC of 0.88, sensitivity 74%, specificity 87%, positive predictive value 76%, negative predictive value 86%, positive likelihood ratio 5.83 and negative likelihood ratio 0.30 (Table 3).

Table 3.   Odds ratios (ORs and 95% CIs) for the prediction of delivery within 7 days of sampling, using binary multivariate models with IL-10 and RANTES in maternal plasma alone, cervical length alone and a combined model with maternal plasma proteins and cervical length. Predictive values including the likelihood ratio for the prediction models combining these three markers are also shown. P values were calculated using the Hosmer and Lemeshow test for the goodness of fit of the models
VariableSPTD ≤ 7 days of sampling OR (95% CI)
Serum proteins onlyCervical length onlyCombined model
  1. The regression equation for the combined model is logit P = −4.71 + 1.62·IL-10 + 2.29·RANTES + 3.77·CL, where CL is cervical length. In the regression equation all the variables are binary values.

Cervical Length (≤18 mm) 21.2 (6–74.7)43.5 (8.9–211.3)
High RANTES plasma (≥49 293 pg/ml)3.9 (1.7–9) 9.9 (2.5–39.2)
High IL-10 plasma (≥48 pg/ml)4.6 (2.1–10) 5.0 (1.7–14.3)
   Predictive values  
Correctly predicted69.0%72.7%82.6%
Sensitivity84.2%92.8%73.8%
Specificity54.1%62.0%87.3%
Positive predictive value55.2%56.5%75.6%
Negative predictive value83.6%94.2%86.2%
Positive likelihood ratio1.842.455.83
Negative likelihood ratio0.290.120.30
AUC (CI)0.74 (0.66–0.82)0.77 (0.71–0.84)0.88 (0.84–0.95)
P 0.990.91
Table 4.   Integrated discrimination improvement (IDI) and net reclassification improvement (NRI) indexes for the comparison of the discrimination ability between the three prediction models
 IDINRI
Serum proteins only vs cervical length only0.150.04
Serum proteins only vs combined model<0.001<0.001
Cervical length only vs combined model<0.001<0.001
Figure 1.

 Receiver operating characteristic curve analysis for the prediction model combining maternal plasma concentrations of interleukin-10 (cut-off level ≥ 48 pg/ml), RANTES (cut-off level ≥ 49 293 pg/ml) and cervical length (cut-off level ≤ 18 mm). The area under the curve for this model was 0.88.

Discussion

The necessity of finding good prediction models for imminent PTD is urgent, and a multiple-markers test indicative of the multifactorial aetiology of PTD is likely to be the most successful. Our best multivariable prediction model was based on maternal serum IL-10, RANTES and cervical length, and demonstrated a good capacity for the prediction of delivery within 7 days from the beginning of labour.

Few studies have investigated the relationship between inflammatory cytokines in the maternal serum and spontaneous PTD in women before and during PTL. Vogel et al.15 developed a combined prediction model using both serum and vaginal inflammatory markers in the early second trimester (at 12–25 weeks of gestation) in women with a prior PTD and without symptoms at enrolment: this model predicted 69% of all recurrent spontaneous PTD at a 5% false-positive rate. In a previous study, we analysed the same 27 proteins from amniotic and cervical fluid in 89 women, and found one combined multivariable prediction model based on amniotic macrophage inflammatory protein-1beta, cervical interferon-gamma and monocyte chemotactic protein-1.18 We showed that a combination of proteins from amniotic fluid and cervical fluid could help determine which women will deliver within 7 days. In the present study we used maternal serum, which is easier to collect and less invasive for the patient compared with amniotic fluid and cervical fluid.

Transvaginal cervical length measurement is commonly used to predict PTD in women with threatened PTL. A cervical length measurement of 25 mm or less is generally considered to be an excellent indicator of an increased risk of PTD in symptomatic women.19–21 The fetal fibronectin test is also used in predicting spontaneous PTD within 7–10 days of testing among women with symptoms of threatened PTD.22 Several studies have reported that the fetal fibronectin test and cervical length measurement provided similar results in predicting the risk of PTD.23,24 However, it remains unclear whether the combination of the fetal fibronectin test and cervical length measurement improves the prediction of PTD, and how these tests should be combined.25

In our multivariable prediction model, levels of maternal serum IL-10 were significantly higher in women who delivered preterm (within 7 days of sampling) compared with women who delivered later. Using explant cultures of gestational membranes, Mitchell et al.26 found that IL-10 exerts anti-inflammatory effects on choriodecidua, but that in the adjacent amnion it has a remarkable pro-inflammatory action. These findings suggest that the fetal membranes can exhibit opposing responses to IL-10, depending on whether the inflammation occurs at the maternal or fetal surface. Aaltonen et al.27 also found that in vitro stimulation of choriodecidual tissue with high levels of Ureaplasma urealyticum induced an inflammatory response, followed by the secretion of IL-10.

Another cytokine in our prediction model was RANTES. We found that levels of maternal serum RANTES were statistically higher in women who delivered preterm (within 7 days of sampling), compared with women who delivered later. However, there are few studies describing the association between RANTES in maternal serum and the PTD. Chow et al.28 showed that RANTES levels in maternal serum were elevated in the second trimester in asymptomatic women. Using a murine model, Yang et al.29 also showed that the production of several pro-inflammatory cytokines, including RANTES, occurred very early in the cascade of events that lead to PTL.

The interval of 7 days was chosen to represent an average number of days of hospital stay and monitoring of such women with symptoms of preterm labour. During this period of time the clinician has the possibility to admit the woman, provide corticosteroids and tocolytics, and let her deliver.

Conclusion

In our study population of women with threatened PTL, we defined a combined prediction model that uses small serum volumes and cervical length measurements. This model has the potential to be useful to clinicians to determine which symptomatic women will deliver preterm. This model would, however, need to be tested in a new cohort of high-risk women in order to confirm its predictive ability. If such studies confirm the moderately high positive predictive value and negative predictive value, this could help to identify women with a considerable risk of delivering preterm, who can be targeted for treatment, and likewise women who are at low risk that could be sent home without treatment. A non-invasive, low-cost test with a high positive predictive value could assist in the timing of corticosteroid administration, for example. On the market today, there are tests with high negative predictive values that make it possible to refrain from hospitalisation, bed rest and repeated monitoring; however, none are available with a high positive predictive value. Furthermore, improved tests with high positive predictive value can help to target novel treatment for high-risk groups in clinical trials, as we have so far not been able to improve the overall neonatal outcome in this group of women with symptoms of threatened PTL.

Disclosure of interests

All authors declare that there are no conflicts of interest to disclose.

Contribution to authorship

The contributions of individual authors to this paper are as follows: planning research (PT and BJ); performing research (PT, BJ, R-MH, HH, UBW); analysing data (PT, BJ, KS, DH, PTH, MK); interpreting data (PT, BJ, BP, KS, DH, PTH, MK); and writing the first draft of the article (PT and BJ).

Details of ethics approval

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

Funding

This work was supported by grants from the Swedish Medical Society (SLS 2008-21198), Swedish government grants to researchers in the public health service (ALFGBG-2863 and ALFGBG-11522) and Swedish Medical Research Council (VR 2006-3396), NIMH grant 5R21-MH68513-3 and by the Medical Research Council in the UK, grant P19381.

Acknowledgement

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

Ancillary