DOI: 10.1111/1471-0528.12993 www.bjog.org
New biomarkers for the prediction of spontaneous preterm labour in symptomatic pregnant women: a comparison with fetal fibronectin S Liong,a,b MKW Di Quinzio,a,b G Fleming,b M Permezel,a,b GE Rice,c HM Georgioua,b a Department of Obstetrics and Gynaecology, University of Melbourne, Parkville, Vic., Australia b Mercy Perinatal Research Centre, Mercy Hospital for Women, Heidelberg, Vic., Australia c Centre for Clinical Research, Royal Brisbane and Women’s Hospital, University of Queensland, Herston, Qld, Australia Correspondence: Dr HM Georgiou, Department of Obstetrics and Gynaecology, Mercy Hospital for Women, University of Melbourne, 163 Studley Road, Heidelberg, Vic. 3084, Australia. Email [email protected]
Accepted 3 June 2014. Published Online 24 July 2014.
Objective To identify cervicovaginal fluid (CVF) biomarkers
predictive of spontaneous preterm birth in women with symptoms of preterm labour. Design Retrospective cohort study. Setting Melbourne, Australia. Population Women with a singleton pregnancy admitted to the
Emergency Department between 22 and 36 weeks of gestation presenting with symptoms of preterm labour. Methods Two-dimensional electrophoresis was used to analyse the CVF proteome. Validation of putative biomarkers was performed using enzyme-linked immunosorbent assay (ELISA) in an independent cohort. Optimal concentration thresholds of putative biomarkers were determined and the predictive efficacy for preterm birth was compared with that of fetal fibronectin. Main outcome measures Prediction of spontaneous preterm
labour within 7 days. Results Differentially expressed proteins were identified by proteomic analysis in women presenting with ‘threatened’ preterm labour without cervical change who subsequently delivered
preterm (n = 12 women). ELISA validation using an independent cohort (n = 129 women) found albumin and vitamin D-binding protein (VDBP) to be significantly altered between women who subsequently experienced preterm birth and those who delivered at term. Prediction of preterm delivery within 7 days using a dual biomarker model (albumin/VDBP) provided 66.7% sensitivity, 100% specificity, 100% positive predictive value (PPV) and 96.7% negative predictive value (NPV), compared with fetal fibronectin yielding 66.7, 87.9, 36.4 and 96.2%, respectively (n = 64). Using the maximum number of screened samples, the predictive utility of albumin/VDBP yielded a sensitivity of 77.8%, specificity and PPV of 100% and NPV of 98.0% (n = 109). Conclusions The dual biomarker model of albumin/VDBP is more
efficacious than fetal fibronectin in predicting spontaneous preterm delivery in symptomatic women within 7 days. A clinical diagnostic trial is required to test this model on a larger population to confirm these findings and to further refine the predictive values. Keywords Biomarkers, cervicovaginal fluid, pregnancy, preterm labour, proteomics.
Please cite this paper as: Liong S, Di Quinzio MKW, Fleming G, Permezel M, Rice GE, Georgiou HM. New biomarkers for the prediction of spontaneous preterm labour in symptomatic pregnant women: a comparison with fetal fibronectin. BJOG 2014; DOI: 10.1111/1471-0528.12993.
Introduction Preterm birth (7 days, n = 105
IL-1ra (n = 117)
0.652 (0.504–0.800)
0.819 (0.711–0.926) ≤7 days, n = 12
Albumin (n = 117)
20.8 (7.2–42.2)
98.2 (90.4–99.7) 13.1 (5.9–24.2)
26.3 (13.4–43.1) 73.3 (63.8–81.5) 83.3 (51.6–97.4) 0.001
≥19 000
(ng/ml) (95% CI)
94.7 (87.1–98.5)
0.22 (0.03–1.40) 1.81 (1.34–2.45)
8.3 (1.0–68.7)
13.8 (2.8–66.7) 0.23 (0.06–0.81) 97.5 (91.1–99.6)
3.12 (2.08–4.69)
OR
(95% CI)
LR
(95% CI)
LR+
(95% CI)
NPV (%)
(95% CI) (95% CI)
PPV (%) Specificity (%)
(95% CI) (95% CI)
Sensitivity (%) Threshold
concentration
P-value the curve
Area under
Table 2. Optimal concentration threshold to determine the efficacy of individual putative cervicovaginal fluid (CVF) biomarkers for the prediction of spontaneous preterm birth within 7 days
Liong et al.
gestation) was determined by binary logistic regression. ROC curves were constructed and AUC values and optimal concentration thresholds for predictive utility were determined. Based on these thresholds, the sensitivities, specificities, PPVs and NPVs for the prediction of preterm birth within 7 days from sampling were determined (Supporting Information Figure S2 and Table 2). Albumin and VDBP displayed the highest AUC, with albumin yielding the higher sensitivity (83.3%; optimal concentration threshold ≥19 000 ng/ml) and VDBP yielding the higher specificity (98.1%; optimal concentration threshold ≥90 ng/ml).
Comparison of candidate biomarkers with fFN for the prediction of spontaneous preterm birth Where it was deemed clinically appropriate, a QuikCheckTM fFN test was performed on 64 women between 22 and 3 cm dilation, Group 3) were not tested with the QuikCheckTM fFN kit. Utilising the same 64 women in whom albumin and VDBP were both assayed, a comparison of the efficacy of fFN with a combined biomarker model (albumin and VDBP) for the prediction of labour within 7 days was made. Based on the previously determined optimal concentration thresholds for each biomarker, the data were transformed into a binary format (positive or negative outcome) and ROC curves were constructed (Figure 2). The results indicate that the combined model yields a higher AUC than the fFN test, with a superior specificity and PPV of 100% (Table 3). This compares with an fFN PPV of 36.4% in the same cohort of women. As an additional analysis, the utility of different multimarker combinations (fFN/albumin; fFN/VDBP and fFN/ albumin/VDBP) for the prediction of preterm birth within 7 days was assessed using a contingency table based on the optimal threshold for each biomarker. In combinations that included fFN, the sensitivity was not improved relative to the combined albumin/VDBP model. There was also a concurrent reduction in the specificity and PPV (data not shown).
ª 2014 Royal College of Obstetricians and Gynaecologists
CI, confidence interval; LR+, positive likelihood ratio; LR , negative likelihood ratio; NPV, negative predictive value; OR, odds ratio; PPV, positive predictive value. Concentration threshold values for albumin (≥19 000 ng/ml) and VDBP (≥90 ng/ml) are based on the results presented in Table 2. *The QuikCheckTM fFN is a qualitative indicator test with a positive colour change threshold of ≥50 ng/ml. **LR+ and OR values were determined using a specificity of >98%, as specificities of 100% result in undefined LR+ and OR values.
>114** (8.4–1544) 0.33 (0.11–1.01) 96.7 (88.5–99.5) 100 (40.2–100) 100 (93.8–100) 66.7 (22.7–94.7) 0.008
>38.7* (5.11–293)
14.6 (2.2–94.8) 0.38 (0.12–1.18) 96.2 (87.0–99.4) 36.4 (11.2–69.1) 87.9 (76.7–95.0) 66.7 (22.7–94.7)
Within 7 days (n = 6) QuikCheckTM 0.773 (0.543–1.000) fFN* Albumin 0.833 (0.600–1.000) and VDBP
0.029
5.52 (2.25–13.53)
OR (95% CI) LR (95% CI) LR+ (95% CI) NPV (%) (95% CI) PPV (%) (95% CI) Specificity (%) (95% CI) Sensitivity (%) (95% CI) Pvalue Area under the curve (95% CI) Delivery
Table 3. Utility of a combined biomarker model of albumin and vitamin D-binding protein (VDBP) for the prediction of spontaneous preterm birth in symptomatic women compared with the QuikCheckTM fetal fibronectin (fFN) test. All three biomarkers were measured in the same women (n = 64)
Predictive biomarkers of threatened preterm labour
ª 2014 Royal College of Obstetricians and Gynaecologists
Figure 2. Receiver operating characteristic (ROC) curves of cervicovaginal fluid (CVF) biomarkers for the prediction of preterm birth within 7 days in symptomatic women with no cervical change (n = 64); QuikCheckTM fetal fibronectin (fFN) test (bold black line) compared with the albumin/vitamin D-binding protein (VDBP) multimarker model (green line).
Utility of the combined biomarker model of albumin/VDBP for the prediction of spontaneous preterm birth Using the maximum number of samples available in which both albumin and VDBP were screened (n = 109), the utility of the combined albumin/VDBP model for the prediction of spontaneous preterm birth of 347* (27.9–4306)
CI, confidence interval; LR+, positive likelihood ratio; LR , negative likelihood ratio; NPV, negative predictive value; OR, odds ratio; PPV, positive predictive value. Concentration threshold values for albumin (≥19 000 ng/ml) and VDBP (≥90 ng/ml) are based on the results presented in Table 2. *LR+ and OR values were determined using a specificity of >98% as specificities of 100% result in undefined LR+ and OR values.
Focusing on women presenting with symptoms of PTL in whom no observable cervical change was evident, a panel of four biomarkers was selected, optimal concentration thresholds were determined and the utility for the prediction of spontaneous PTL within 7 days was assessed. Albumin and VDBP proved to be superior to IL-1ra and thioredoxin. Subsequently, a direct comparison of the performance of a combined albumin/VDBP biomarker ‘test’ was made with that of the QuikCheckTM test in 64 women who were deemed to be eligible for fFN testing on presentation to the Emergency Department. The combined biomarker predictive model of albumin/VDBP yielded a superior specificity, PPV and NPV compared with fFN. Using the maximum number of CVF samples available (n = 109), the combined albumin/VDBP predictive model yielded a sensitivity of 77.8%, specificity of 100%, PPV of 100% and NPV of 98.0%.
Limitations and strengths Although overall sample numbers were small, the strength of this study was the direct comparison between our dual biomarker ‘test’ (using raw concentration data) and the performance of fFN in the same cohort of women. Uncorrected raw data were used to simulate a ‘bedside test’ designed to provide a definitive positive or negative result. Our test outperformed the fFN test for the prediction of PTL within 7 days. A predictive bedside test needs to be robust to confounders such as multifetal gestation, bacterial vaginosis, recent sexual intercourse and recent bleeding. Women presenting with these potential confounders were excluded from the study. Future studies will need to determine the impact of these confounders on the CVF expression of albumin and VDBP. No correlation was made between the clinical diagnosis of ‘no observable cervical change’ and transvaginal ultrasonographic measurement of cervical length. A shortened cervical length may reflect some degree of cervical ripening in
8
preparation for labour. Although the cervix may have appeared ‘normal’ in these women, a shortened cervical length may have been present. Although ultrasonographic findings may improve the predictive utility of a test for PTL, access to this resource may not always be feasible. An additional consideration was the use of tocolytic therapy. All women who tested positive for fFN in the study (n = 11) received tocolytic therapy with oral nifedipine. Of these women, seven delivered at term and four delivered preterm. The possible impact of tocolytic therapy on the study findings remains to be established.
Interpretation In symptomatic women who subsequently delivered preterm, it is hypothesised that the reduction of thioredoxin and IL-1ra in the CVF may indirectly reflect the upregulated pro-inflammatory processes that are essential to parturition. Thioredoxin is a potent antioxidant and has been implicated in the process of cervical remodelling.15 IL-1ra has anti-inflammatory properties by acting as an antagonist against the pro-inflammatory IL-1a and IL-1b cytokines by competitive binding of the IL-1 receptor on the cell surface of leucocytes.16 Albumin is an abundant plasma protein involved in the regulation of capillary osmotic pressure, but also functions as a transporter protein and possesses antioxidant properties.17–19 VDBP, although acting as a carrier protein for vitamin D metabolites, also plays an important role in the clearance of actin released during tissue injury and the augmentation of the pro-inflammatory response.20–23 As both albumin and VDBP are abundant serum proteins, an increase in these proteins in the CVF may reflect increased endothelial permeability which may occur in preparation for labour. These biomarkers have been implicated previously in the CVF in association with spontaneous term labour,7 spontaneous preterm birth9,10,24 and preterm prelabour rupture of the fetal membranes.25 For women who spontaneously labour at a preterm gestation, it is speculated that the
ª 2014 Royal College of Obstetricians and Gynaecologists
Predictive biomarkers of threatened preterm labour
similarities between the expression profiles of the CVF proteins in symptomatic and asymptomatic women indicate that both groups share similar increased inflammation, oxidative stress and matrix remodelling associated with preterm birth. The development of a predictive test for spontaneous preterm birth with high accuracy continues to remain a challenge, most probably because of the multifactorial and complex aetiology of labour. It is proposed that a multimarker model may offer the best chance for more accurate prediction. The current gold standard used to predict preterm birth is cervicovaginal fFN. fFN present in the CVF at preterm gestations is thought to be a result of the disruption or separation of the choriodecidual interface, and is associated with an increased risk of preterm birth.26 However, a positive fFN result does not necessarily indicate that a woman will deliver preterm owing to its relatively poor PPV. A positive fFN result may be confounded by other factors, including recent sexual intercourse, bleeding, digital vaginal examination and amniotic fluid contamination. Our previous studies have shown that recent unprotected sexual intercourse does not affect the expression of albumin (Liong S, Di Quinzio MKW, Fleming G, Permezel M, Georgiou HM, unpublished data), VDBP,24 IL-1ra12 or thioredoxin11 in the CVF. A comprehensive systematic review of 27 published studies recently assessed the accuracy of fFN for the prediction of spontaneous preterm birth in symptomatic women.27 The pooled estimates of sensitivity and specificity for fFN for the prediction of preterm birth within 7–10 days after sampling were 76.7 and 82.7%, respectively. Our dual biomarker model, assessing 109 symptomatic women without cervical change, provided a sensitivity of 77.8% and specificity of 100% for the prediction of spontaneous PTL within 7 days.
Contribution to authorship
Conclusion and research recommendations
Additional Supporting Information may be found in the online version of this article: Figure S1. Two-dimensional difference in-gel electrophoresis (2D-DIGE) analysis of the cervicovaginal fluid (CVF) proteome from women symptomatic of preterm labour (PTL) with no observable cervical change who spontaneously laboured and delivered preterm. Figure S2. Receiver operating characteristic (ROC) curves of putative cervicovaginal fluid (CVF) biomarkers for the prediction of preterm birth within 7 days in symptomatic women with no cervical change. Figure S3. Receiver operating characteristic (ROC) curves of cervicovaginal fluid (CVF) biomarkers for the prediction of preterm birth within 7 days in symptomatic women with no cervical change. Table S1. Obstetric data of participants with symptoms of preterm labour used for two-dimensional difference in-gel electrophoresis (2D-DIGE).
The present study has demonstrated distinctive changes that occur within the CVF proteome between gestation-matched women with symptoms of PTL who either deliver at term or preterm. Our validation study using a larger sample size of symptomatic women has shown that a dual biomarker model consisting of albumin and VDBP is more accurate than the QuikCheckTM fFN test. A clinical diagnostic trial is required to test this model on a larger population to confirm these findings and to further refine the predictive values. The identification of women with a highly reliable test for preterm birth could provide the opportunity for early therapeutic and interventional care for both mothers and babies.
Disclosure of interests
Conception and design of the experiments: SL, HMG, MKWD, MP. Performance of the experiments: SL. Analyses of the data: SL, HMG, MKWD, MP. Contribution of reagents/materials/analysis tools: HMG, MKWD, GER. Drafting and editing of the manuscript: SL, HMG, MKWD. Recruitment and collection of CVF samples: GF. Clinical consultation: MKWD, MP.
Details of ethics approval This study was approved by the Mercy Health Research Ethics Committee (R56/06).
Funding Funding was provided by the Medical Research Foundation for Women and Babies. SL is a recipient of an Australian Postgraduate Award (Australian Government Department of Industry).
Acknowledgements The authors thank all participants involved in this study and the Mercy Hospital for Women medical and midwifery staff for obtaining the cervicovaginal fluid (CVF) samples. Dr Yujing Heng is acknowledged for processing the CVF samples, and mass spectrometry assistance was provided by Karen Oliva, Dr Matthew Knight (Department of Primary Industries, Bundoora, Vic., Australia) and Drs Paul O’Donnell and Nick Williamson (Bio21 Institute, Parkville, Vic., Australia). Microbiological assessment was performed by the Microbiology Department, Austin Pathology, Austin Health, Heidelberg, Vic., Australia.
Supporting Information
None of the authors have any conflict of interest to report.
ª 2014 Royal College of Obstetricians and Gynaecologists
9
Liong et al.
Table S2. Peptide sequences of characterised proteins. Table S3. Obstetric data of symptomatic participants in the validation cohort. Table S4. QuikCheckTM fetal fibronectin (fFN) test results of symptomatic women (n = 64) with no observable cervical change. &
References 1 Beck S, Wojdyla D, Say L, Betran A, Merialdi M, Requejo J, et al. The worldwide incidence of preterm birth: a systematic review of maternal mortality and morbidity. Bull World Health Org 2010;88:31–8. 2 Bastek JA, Sammel MD, Pare E, Srinivas SK, Posencheg MA, Elovitz MA. Adverse neonatal outcomes: examining the risks between preterm, late preterm, and term infants. Am J Obstet Gynecol 2008;199:e1–8. 3 Dekker G, Lee S, North R, McCowan L, Simpson N, Roberts C. Rick factors for preterm birth in an international prospective cohort of nulliparous women. PLoS ONE 2012;7:e39154. 4 Mercer B, Goldenberg R, Das A, Moawad A, Iams J, Meis P, et al. The preterm prediction study: a clinical risk assessment system. Am J Obstet Gynecol 1996;174:1885–93. 5 Honest H, Bachmann L, Gupta J, Kleijnen J, Khan K. Accuracy of cervicovaginal fetal fibronectin test in predicting risk of spontaneous preterm birth: systematic review. BMJ 2002;325:301. 6 Cooper S, Lange I, Wood S, Tang S, Miller L, Ross S. Diagnostic accuracy of rapid phIGFBP-1 assay for predicting preterm labor in symptomatic patients. J Perinatol 2012;32:460–5. 7 Di Quinzio MKW, Georgiou HM, Holdsworth-Carson SJ, Ayhan M, Heng YJ, Walker SP, et al. Proteomic analysis of human cervicovaginal fluid displays differential protein expression in association with labor onset at term. J Proteome Res 2008;7:1916–21. 8 Heng YJ, Di Quinzio MKW, Permezel M, Ayhan M, Rice GE, Georgiou HM. Temporal proteomic analysis of human cervicovaginal fluid with impending term labor. J Proteome Res 2010;9: 1344–50. 9 Pereira L, Reddy AP, Jacob T, Thomas A, Schneider KA, Dasari S, et al. Identification of novel protein biomarkers of preterm birth in human cervical-vaginal fluid. J Proteome Res 2007;6:1269–76. 10 Liong S, Di Quinzio MK, Fleming G, Permezel M, Rice GE, Georgiou HM. Prediction of spontaneous preterm labour in at-risk pregnant women. Reproduction 2013;146:335–45. 11 Heng YJ, Di Quinzio MKW, Permezel M, Rice G, Georgiou H. Temporal expression of antioxidants in human cervicovaginal fluid associated with spontaneous labor. Antioxid Redox Signal 2010;13:951–7.
10
12 Heng YJ, Liong S, Permezel M, Rice GE, Di Quinzio MK, Georgiou HM. The interplay of the interleukin 1 system in pregnancy and labor. Reprod Sci 2014;21:122–30. 13 Sanchez-Ramos L, Delke I, Zamora J, Kaunitz A. Fetal fibronectin as a short-term predictor of preterm birth in symptomatic patients: a meta-analysis. Obstet Gynecol 2009;114:631–40. 14 Daskalakis G, Papantoniou N, Koutsodimas N, Papapanagiotou A, Antsaklis A. Fetal fibronectin as a predictor of preterm birth. Obstet Gynecol 2000;20:347–53. 15 Lysell J, Stjernholm Vladic Y, Ciarlo N, Holmgren A, Sahlin L. Immunohistochemical determination of thioredoxin and glutaredoxin distribution in the human cervix, and possible relation to cervical ripening. Gynecol Endocrinol 2003;17:303–10. 16 Arend WP, Malyak M, Guthridge C, Gabay C. Interleukin-1 receptor antagonist: role in biology. Annu Rev Immunol 1998;16:27–55. 17 Quinlan G, Mumby S, Martin G, Bernard G, Gutteridge J, Evans T. Albumin influences total plasma antioxidant capacity favorably in patients with acute lung injury. Crit Care Med 2004;32:755–9. 18 Margarson M, Soni N. Serum albumin: touchstone or totem? Anaesthesia 1998;53:789–803. 19 Fanalia G, di Masib A, Trezzab V, Marinob M, Fasanoa M, Ascenzib P. Human serum albumin: from bench to bedside. Mol Aspects Med 2012;33:209–90. 20 McLeod JF, Kowalski MA, Haddad JG. Interactions among serum vitamin D binding protein, monomeric actin, proflin, and profilactin. J Biol Chem 1989;264:1260–7. 21 Haddad JG. Human serum binding protein for vitamin D and its metabolites (DBP): evidence that actin is the DBP binding component in human skeletal muscle. Arch Biochem Biophys 1982;213:538–44. 22 Haddad JG. Plasma vitamin D-binding protein (Gc-globulin): multiple tasks. J Steroid Biochem Mol Biol 1995;53:579–82. 23 Perez HD. Gc-globulin (vitamin D-binding protein) increases binding of low concentrations of C5a des Arg to human polymorphonuclear leukocytes: an explanation for its co chemotaxin activity. Inflammation 1994;18:215–20. 24 Liong S, Di Quinzio MK, Fleming G, Permezel M, Georgiou HM. Is vitamin D binding protein a novel predictor of labour? PLoS ONE 2013;8:e76490. 25 Liong S, Di Quinzio MK, Heng YJ, Fleming G, Permezel M, Rice GE, et al. Proteomic analysis of human cervicovaginal fluid collected prior to preterm premature rupture of the fetal membranes. Reproduction 2013;145:137–47. 26 Lockwood CJ, Senyei AE, Dische MR, Casal D, Shah KD, Thung SN, et al. Fetal fibronectin in cervical and vaginal secretions as a predictor of preterm delivery. N Engl J Med 1991;325:669–74. 27 Deshpande SN, van Asselt AD, Tomini F, Armstrong N, Allen A, Noake C, et al. Rapid fetal fibronectin testing to predict preterm birth in women with symptoms of premature labour: a systematic review and cost analysis. Health Technol Assess 2013;17:1–138.
ª 2014 Royal College of Obstetricians and Gynaecologists
New biomarkers for the prediction of spontaneous preterm labour in symptomatic pregnant women: a comparison with fetal fibronectin S Liong,a,b MKW Di Quinzio,a,b G Fleming,b M Permezel,a,b GE Rice,c HM Georgioua,b a Department of Obstetrics and Gynaecology, University of Melbourne, Parkville, Vic., Australia b Mercy Perinatal Research Centre, Mercy Hospital for Women, Heidelberg, Vic., Australia c Centre for Clinical Research, Royal Brisbane and Women’s Hospital, University of Queensland, Herston, Qld, Australia Correspondence: Dr HM Georgiou, Department of Obstetrics and Gynaecology, Mercy Hospital for Women, University of Melbourne, 163 Studley Road, Heidelberg, Vic. 3084, Australia. Email [email protected]
Accepted 3 June 2014. Published Online 24 July 2014.
Objective To identify cervicovaginal fluid (CVF) biomarkers
predictive of spontaneous preterm birth in women with symptoms of preterm labour. Design Retrospective cohort study. Setting Melbourne, Australia. Population Women with a singleton pregnancy admitted to the
Emergency Department between 22 and 36 weeks of gestation presenting with symptoms of preterm labour. Methods Two-dimensional electrophoresis was used to analyse the CVF proteome. Validation of putative biomarkers was performed using enzyme-linked immunosorbent assay (ELISA) in an independent cohort. Optimal concentration thresholds of putative biomarkers were determined and the predictive efficacy for preterm birth was compared with that of fetal fibronectin. Main outcome measures Prediction of spontaneous preterm
labour within 7 days. Results Differentially expressed proteins were identified by proteomic analysis in women presenting with ‘threatened’ preterm labour without cervical change who subsequently delivered
preterm (n = 12 women). ELISA validation using an independent cohort (n = 129 women) found albumin and vitamin D-binding protein (VDBP) to be significantly altered between women who subsequently experienced preterm birth and those who delivered at term. Prediction of preterm delivery within 7 days using a dual biomarker model (albumin/VDBP) provided 66.7% sensitivity, 100% specificity, 100% positive predictive value (PPV) and 96.7% negative predictive value (NPV), compared with fetal fibronectin yielding 66.7, 87.9, 36.4 and 96.2%, respectively (n = 64). Using the maximum number of screened samples, the predictive utility of albumin/VDBP yielded a sensitivity of 77.8%, specificity and PPV of 100% and NPV of 98.0% (n = 109). Conclusions The dual biomarker model of albumin/VDBP is more
efficacious than fetal fibronectin in predicting spontaneous preterm delivery in symptomatic women within 7 days. A clinical diagnostic trial is required to test this model on a larger population to confirm these findings and to further refine the predictive values. Keywords Biomarkers, cervicovaginal fluid, pregnancy, preterm labour, proteomics.
Please cite this paper as: Liong S, Di Quinzio MKW, Fleming G, Permezel M, Rice GE, Georgiou HM. New biomarkers for the prediction of spontaneous preterm labour in symptomatic pregnant women: a comparison with fetal fibronectin. BJOG 2014; DOI: 10.1111/1471-0528.12993.
Introduction Preterm birth (7 days, n = 105
IL-1ra (n = 117)
0.652 (0.504–0.800)
0.819 (0.711–0.926) ≤7 days, n = 12
Albumin (n = 117)
20.8 (7.2–42.2)
98.2 (90.4–99.7) 13.1 (5.9–24.2)
26.3 (13.4–43.1) 73.3 (63.8–81.5) 83.3 (51.6–97.4) 0.001
≥19 000
(ng/ml) (95% CI)
94.7 (87.1–98.5)
0.22 (0.03–1.40) 1.81 (1.34–2.45)
8.3 (1.0–68.7)
13.8 (2.8–66.7) 0.23 (0.06–0.81) 97.5 (91.1–99.6)
3.12 (2.08–4.69)
OR
(95% CI)
LR
(95% CI)
LR+
(95% CI)
NPV (%)
(95% CI) (95% CI)
PPV (%) Specificity (%)
(95% CI) (95% CI)
Sensitivity (%) Threshold
concentration
P-value the curve
Area under
Table 2. Optimal concentration threshold to determine the efficacy of individual putative cervicovaginal fluid (CVF) biomarkers for the prediction of spontaneous preterm birth within 7 days
Liong et al.
gestation) was determined by binary logistic regression. ROC curves were constructed and AUC values and optimal concentration thresholds for predictive utility were determined. Based on these thresholds, the sensitivities, specificities, PPVs and NPVs for the prediction of preterm birth within 7 days from sampling were determined (Supporting Information Figure S2 and Table 2). Albumin and VDBP displayed the highest AUC, with albumin yielding the higher sensitivity (83.3%; optimal concentration threshold ≥19 000 ng/ml) and VDBP yielding the higher specificity (98.1%; optimal concentration threshold ≥90 ng/ml).
Comparison of candidate biomarkers with fFN for the prediction of spontaneous preterm birth Where it was deemed clinically appropriate, a QuikCheckTM fFN test was performed on 64 women between 22 and 3 cm dilation, Group 3) were not tested with the QuikCheckTM fFN kit. Utilising the same 64 women in whom albumin and VDBP were both assayed, a comparison of the efficacy of fFN with a combined biomarker model (albumin and VDBP) for the prediction of labour within 7 days was made. Based on the previously determined optimal concentration thresholds for each biomarker, the data were transformed into a binary format (positive or negative outcome) and ROC curves were constructed (Figure 2). The results indicate that the combined model yields a higher AUC than the fFN test, with a superior specificity and PPV of 100% (Table 3). This compares with an fFN PPV of 36.4% in the same cohort of women. As an additional analysis, the utility of different multimarker combinations (fFN/albumin; fFN/VDBP and fFN/ albumin/VDBP) for the prediction of preterm birth within 7 days was assessed using a contingency table based on the optimal threshold for each biomarker. In combinations that included fFN, the sensitivity was not improved relative to the combined albumin/VDBP model. There was also a concurrent reduction in the specificity and PPV (data not shown).
ª 2014 Royal College of Obstetricians and Gynaecologists
CI, confidence interval; LR+, positive likelihood ratio; LR , negative likelihood ratio; NPV, negative predictive value; OR, odds ratio; PPV, positive predictive value. Concentration threshold values for albumin (≥19 000 ng/ml) and VDBP (≥90 ng/ml) are based on the results presented in Table 2. *The QuikCheckTM fFN is a qualitative indicator test with a positive colour change threshold of ≥50 ng/ml. **LR+ and OR values were determined using a specificity of >98%, as specificities of 100% result in undefined LR+ and OR values.
>114** (8.4–1544) 0.33 (0.11–1.01) 96.7 (88.5–99.5) 100 (40.2–100) 100 (93.8–100) 66.7 (22.7–94.7) 0.008
>38.7* (5.11–293)
14.6 (2.2–94.8) 0.38 (0.12–1.18) 96.2 (87.0–99.4) 36.4 (11.2–69.1) 87.9 (76.7–95.0) 66.7 (22.7–94.7)
Within 7 days (n = 6) QuikCheckTM 0.773 (0.543–1.000) fFN* Albumin 0.833 (0.600–1.000) and VDBP
0.029
5.52 (2.25–13.53)
OR (95% CI) LR (95% CI) LR+ (95% CI) NPV (%) (95% CI) PPV (%) (95% CI) Specificity (%) (95% CI) Sensitivity (%) (95% CI) Pvalue Area under the curve (95% CI) Delivery
Table 3. Utility of a combined biomarker model of albumin and vitamin D-binding protein (VDBP) for the prediction of spontaneous preterm birth in symptomatic women compared with the QuikCheckTM fetal fibronectin (fFN) test. All three biomarkers were measured in the same women (n = 64)
Predictive biomarkers of threatened preterm labour
ª 2014 Royal College of Obstetricians and Gynaecologists
Figure 2. Receiver operating characteristic (ROC) curves of cervicovaginal fluid (CVF) biomarkers for the prediction of preterm birth within 7 days in symptomatic women with no cervical change (n = 64); QuikCheckTM fetal fibronectin (fFN) test (bold black line) compared with the albumin/vitamin D-binding protein (VDBP) multimarker model (green line).
Utility of the combined biomarker model of albumin/VDBP for the prediction of spontaneous preterm birth Using the maximum number of samples available in which both albumin and VDBP were screened (n = 109), the utility of the combined albumin/VDBP model for the prediction of spontaneous preterm birth of 347* (27.9–4306)
CI, confidence interval; LR+, positive likelihood ratio; LR , negative likelihood ratio; NPV, negative predictive value; OR, odds ratio; PPV, positive predictive value. Concentration threshold values for albumin (≥19 000 ng/ml) and VDBP (≥90 ng/ml) are based on the results presented in Table 2. *LR+ and OR values were determined using a specificity of >98% as specificities of 100% result in undefined LR+ and OR values.
Focusing on women presenting with symptoms of PTL in whom no observable cervical change was evident, a panel of four biomarkers was selected, optimal concentration thresholds were determined and the utility for the prediction of spontaneous PTL within 7 days was assessed. Albumin and VDBP proved to be superior to IL-1ra and thioredoxin. Subsequently, a direct comparison of the performance of a combined albumin/VDBP biomarker ‘test’ was made with that of the QuikCheckTM test in 64 women who were deemed to be eligible for fFN testing on presentation to the Emergency Department. The combined biomarker predictive model of albumin/VDBP yielded a superior specificity, PPV and NPV compared with fFN. Using the maximum number of CVF samples available (n = 109), the combined albumin/VDBP predictive model yielded a sensitivity of 77.8%, specificity of 100%, PPV of 100% and NPV of 98.0%.
Limitations and strengths Although overall sample numbers were small, the strength of this study was the direct comparison between our dual biomarker ‘test’ (using raw concentration data) and the performance of fFN in the same cohort of women. Uncorrected raw data were used to simulate a ‘bedside test’ designed to provide a definitive positive or negative result. Our test outperformed the fFN test for the prediction of PTL within 7 days. A predictive bedside test needs to be robust to confounders such as multifetal gestation, bacterial vaginosis, recent sexual intercourse and recent bleeding. Women presenting with these potential confounders were excluded from the study. Future studies will need to determine the impact of these confounders on the CVF expression of albumin and VDBP. No correlation was made between the clinical diagnosis of ‘no observable cervical change’ and transvaginal ultrasonographic measurement of cervical length. A shortened cervical length may reflect some degree of cervical ripening in
8
preparation for labour. Although the cervix may have appeared ‘normal’ in these women, a shortened cervical length may have been present. Although ultrasonographic findings may improve the predictive utility of a test for PTL, access to this resource may not always be feasible. An additional consideration was the use of tocolytic therapy. All women who tested positive for fFN in the study (n = 11) received tocolytic therapy with oral nifedipine. Of these women, seven delivered at term and four delivered preterm. The possible impact of tocolytic therapy on the study findings remains to be established.
Interpretation In symptomatic women who subsequently delivered preterm, it is hypothesised that the reduction of thioredoxin and IL-1ra in the CVF may indirectly reflect the upregulated pro-inflammatory processes that are essential to parturition. Thioredoxin is a potent antioxidant and has been implicated in the process of cervical remodelling.15 IL-1ra has anti-inflammatory properties by acting as an antagonist against the pro-inflammatory IL-1a and IL-1b cytokines by competitive binding of the IL-1 receptor on the cell surface of leucocytes.16 Albumin is an abundant plasma protein involved in the regulation of capillary osmotic pressure, but also functions as a transporter protein and possesses antioxidant properties.17–19 VDBP, although acting as a carrier protein for vitamin D metabolites, also plays an important role in the clearance of actin released during tissue injury and the augmentation of the pro-inflammatory response.20–23 As both albumin and VDBP are abundant serum proteins, an increase in these proteins in the CVF may reflect increased endothelial permeability which may occur in preparation for labour. These biomarkers have been implicated previously in the CVF in association with spontaneous term labour,7 spontaneous preterm birth9,10,24 and preterm prelabour rupture of the fetal membranes.25 For women who spontaneously labour at a preterm gestation, it is speculated that the
ª 2014 Royal College of Obstetricians and Gynaecologists
Predictive biomarkers of threatened preterm labour
similarities between the expression profiles of the CVF proteins in symptomatic and asymptomatic women indicate that both groups share similar increased inflammation, oxidative stress and matrix remodelling associated with preterm birth. The development of a predictive test for spontaneous preterm birth with high accuracy continues to remain a challenge, most probably because of the multifactorial and complex aetiology of labour. It is proposed that a multimarker model may offer the best chance for more accurate prediction. The current gold standard used to predict preterm birth is cervicovaginal fFN. fFN present in the CVF at preterm gestations is thought to be a result of the disruption or separation of the choriodecidual interface, and is associated with an increased risk of preterm birth.26 However, a positive fFN result does not necessarily indicate that a woman will deliver preterm owing to its relatively poor PPV. A positive fFN result may be confounded by other factors, including recent sexual intercourse, bleeding, digital vaginal examination and amniotic fluid contamination. Our previous studies have shown that recent unprotected sexual intercourse does not affect the expression of albumin (Liong S, Di Quinzio MKW, Fleming G, Permezel M, Georgiou HM, unpublished data), VDBP,24 IL-1ra12 or thioredoxin11 in the CVF. A comprehensive systematic review of 27 published studies recently assessed the accuracy of fFN for the prediction of spontaneous preterm birth in symptomatic women.27 The pooled estimates of sensitivity and specificity for fFN for the prediction of preterm birth within 7–10 days after sampling were 76.7 and 82.7%, respectively. Our dual biomarker model, assessing 109 symptomatic women without cervical change, provided a sensitivity of 77.8% and specificity of 100% for the prediction of spontaneous PTL within 7 days.
Contribution to authorship
Conclusion and research recommendations
Additional Supporting Information may be found in the online version of this article: Figure S1. Two-dimensional difference in-gel electrophoresis (2D-DIGE) analysis of the cervicovaginal fluid (CVF) proteome from women symptomatic of preterm labour (PTL) with no observable cervical change who spontaneously laboured and delivered preterm. Figure S2. Receiver operating characteristic (ROC) curves of putative cervicovaginal fluid (CVF) biomarkers for the prediction of preterm birth within 7 days in symptomatic women with no cervical change. Figure S3. Receiver operating characteristic (ROC) curves of cervicovaginal fluid (CVF) biomarkers for the prediction of preterm birth within 7 days in symptomatic women with no cervical change. Table S1. Obstetric data of participants with symptoms of preterm labour used for two-dimensional difference in-gel electrophoresis (2D-DIGE).
The present study has demonstrated distinctive changes that occur within the CVF proteome between gestation-matched women with symptoms of PTL who either deliver at term or preterm. Our validation study using a larger sample size of symptomatic women has shown that a dual biomarker model consisting of albumin and VDBP is more accurate than the QuikCheckTM fFN test. A clinical diagnostic trial is required to test this model on a larger population to confirm these findings and to further refine the predictive values. The identification of women with a highly reliable test for preterm birth could provide the opportunity for early therapeutic and interventional care for both mothers and babies.
Disclosure of interests
Conception and design of the experiments: SL, HMG, MKWD, MP. Performance of the experiments: SL. Analyses of the data: SL, HMG, MKWD, MP. Contribution of reagents/materials/analysis tools: HMG, MKWD, GER. Drafting and editing of the manuscript: SL, HMG, MKWD. Recruitment and collection of CVF samples: GF. Clinical consultation: MKWD, MP.
Details of ethics approval This study was approved by the Mercy Health Research Ethics Committee (R56/06).
Funding Funding was provided by the Medical Research Foundation for Women and Babies. SL is a recipient of an Australian Postgraduate Award (Australian Government Department of Industry).
Acknowledgements The authors thank all participants involved in this study and the Mercy Hospital for Women medical and midwifery staff for obtaining the cervicovaginal fluid (CVF) samples. Dr Yujing Heng is acknowledged for processing the CVF samples, and mass spectrometry assistance was provided by Karen Oliva, Dr Matthew Knight (Department of Primary Industries, Bundoora, Vic., Australia) and Drs Paul O’Donnell and Nick Williamson (Bio21 Institute, Parkville, Vic., Australia). Microbiological assessment was performed by the Microbiology Department, Austin Pathology, Austin Health, Heidelberg, Vic., Australia.
Supporting Information
None of the authors have any conflict of interest to report.
ª 2014 Royal College of Obstetricians and Gynaecologists
9
Liong et al.
Table S2. Peptide sequences of characterised proteins. Table S3. Obstetric data of symptomatic participants in the validation cohort. Table S4. QuikCheckTM fetal fibronectin (fFN) test results of symptomatic women (n = 64) with no observable cervical change. &
References 1 Beck S, Wojdyla D, Say L, Betran A, Merialdi M, Requejo J, et al. The worldwide incidence of preterm birth: a systematic review of maternal mortality and morbidity. Bull World Health Org 2010;88:31–8. 2 Bastek JA, Sammel MD, Pare E, Srinivas SK, Posencheg MA, Elovitz MA. Adverse neonatal outcomes: examining the risks between preterm, late preterm, and term infants. Am J Obstet Gynecol 2008;199:e1–8. 3 Dekker G, Lee S, North R, McCowan L, Simpson N, Roberts C. Rick factors for preterm birth in an international prospective cohort of nulliparous women. PLoS ONE 2012;7:e39154. 4 Mercer B, Goldenberg R, Das A, Moawad A, Iams J, Meis P, et al. The preterm prediction study: a clinical risk assessment system. Am J Obstet Gynecol 1996;174:1885–93. 5 Honest H, Bachmann L, Gupta J, Kleijnen J, Khan K. Accuracy of cervicovaginal fetal fibronectin test in predicting risk of spontaneous preterm birth: systematic review. BMJ 2002;325:301. 6 Cooper S, Lange I, Wood S, Tang S, Miller L, Ross S. Diagnostic accuracy of rapid phIGFBP-1 assay for predicting preterm labor in symptomatic patients. J Perinatol 2012;32:460–5. 7 Di Quinzio MKW, Georgiou HM, Holdsworth-Carson SJ, Ayhan M, Heng YJ, Walker SP, et al. Proteomic analysis of human cervicovaginal fluid displays differential protein expression in association with labor onset at term. J Proteome Res 2008;7:1916–21. 8 Heng YJ, Di Quinzio MKW, Permezel M, Ayhan M, Rice GE, Georgiou HM. Temporal proteomic analysis of human cervicovaginal fluid with impending term labor. J Proteome Res 2010;9: 1344–50. 9 Pereira L, Reddy AP, Jacob T, Thomas A, Schneider KA, Dasari S, et al. Identification of novel protein biomarkers of preterm birth in human cervical-vaginal fluid. J Proteome Res 2007;6:1269–76. 10 Liong S, Di Quinzio MK, Fleming G, Permezel M, Rice GE, Georgiou HM. Prediction of spontaneous preterm labour in at-risk pregnant women. Reproduction 2013;146:335–45. 11 Heng YJ, Di Quinzio MKW, Permezel M, Rice G, Georgiou H. Temporal expression of antioxidants in human cervicovaginal fluid associated with spontaneous labor. Antioxid Redox Signal 2010;13:951–7.
10
12 Heng YJ, Liong S, Permezel M, Rice GE, Di Quinzio MK, Georgiou HM. The interplay of the interleukin 1 system in pregnancy and labor. Reprod Sci 2014;21:122–30. 13 Sanchez-Ramos L, Delke I, Zamora J, Kaunitz A. Fetal fibronectin as a short-term predictor of preterm birth in symptomatic patients: a meta-analysis. Obstet Gynecol 2009;114:631–40. 14 Daskalakis G, Papantoniou N, Koutsodimas N, Papapanagiotou A, Antsaklis A. Fetal fibronectin as a predictor of preterm birth. Obstet Gynecol 2000;20:347–53. 15 Lysell J, Stjernholm Vladic Y, Ciarlo N, Holmgren A, Sahlin L. Immunohistochemical determination of thioredoxin and glutaredoxin distribution in the human cervix, and possible relation to cervical ripening. Gynecol Endocrinol 2003;17:303–10. 16 Arend WP, Malyak M, Guthridge C, Gabay C. Interleukin-1 receptor antagonist: role in biology. Annu Rev Immunol 1998;16:27–55. 17 Quinlan G, Mumby S, Martin G, Bernard G, Gutteridge J, Evans T. Albumin influences total plasma antioxidant capacity favorably in patients with acute lung injury. Crit Care Med 2004;32:755–9. 18 Margarson M, Soni N. Serum albumin: touchstone or totem? Anaesthesia 1998;53:789–803. 19 Fanalia G, di Masib A, Trezzab V, Marinob M, Fasanoa M, Ascenzib P. Human serum albumin: from bench to bedside. Mol Aspects Med 2012;33:209–90. 20 McLeod JF, Kowalski MA, Haddad JG. Interactions among serum vitamin D binding protein, monomeric actin, proflin, and profilactin. J Biol Chem 1989;264:1260–7. 21 Haddad JG. Human serum binding protein for vitamin D and its metabolites (DBP): evidence that actin is the DBP binding component in human skeletal muscle. Arch Biochem Biophys 1982;213:538–44. 22 Haddad JG. Plasma vitamin D-binding protein (Gc-globulin): multiple tasks. J Steroid Biochem Mol Biol 1995;53:579–82. 23 Perez HD. Gc-globulin (vitamin D-binding protein) increases binding of low concentrations of C5a des Arg to human polymorphonuclear leukocytes: an explanation for its co chemotaxin activity. Inflammation 1994;18:215–20. 24 Liong S, Di Quinzio MK, Fleming G, Permezel M, Georgiou HM. Is vitamin D binding protein a novel predictor of labour? PLoS ONE 2013;8:e76490. 25 Liong S, Di Quinzio MK, Heng YJ, Fleming G, Permezel M, Rice GE, et al. Proteomic analysis of human cervicovaginal fluid collected prior to preterm premature rupture of the fetal membranes. Reproduction 2013;145:137–47. 26 Lockwood CJ, Senyei AE, Dische MR, Casal D, Shah KD, Thung SN, et al. Fetal fibronectin in cervical and vaginal secretions as a predictor of preterm delivery. N Engl J Med 1991;325:669–74. 27 Deshpande SN, van Asselt AD, Tomini F, Armstrong N, Allen A, Noake C, et al. Rapid fetal fibronectin testing to predict preterm birth in women with symptoms of premature labour: a systematic review and cost analysis. Health Technol Assess 2013;17:1–138.
ª 2014 Royal College of Obstetricians and Gynaecologists
