Characterisation of cervicovaginal fluid (CVF) biomarkers across gestation for the identification of spontaneous preterm birth (sPTB)

Ph.D.BackgroundSpontaneous preterm birth (sPTB) is defined as birth occurring < 37 weeks' gestation; this obstetric problem accounts for 11% of deliveries worldwide. Though efforts are being made to decrease the occurrence of sPTB, it is still a major cause perinatal morbidity and mortality. A myriad of obstetrics strategies to screen for sPTB, such as the measurement of cervical length, fetal fibronectin and phosphorylated insulin-like growth factor binding protein-1 have emerged. However, due to the heterogeneity of sPTB, there is unlikely to be a single test that will be able to predict this pregnancy complication, but a combination of tests may improve the clinical predictive performance. Therefore, our objectives in this study were to 1) evaluate the screening performance of the CVF biomarkers panel developed by Carmentix for the prediction of sPTB at 16-24 weeks’ gestation in an unselected Hong Kong Chinese population. 2) establish multi-marker prediction models (maternal characteristics, CVF biomarkers, cervical length) to increase predictive performance for spontaneous PTB and 3) characterise the temporal changes of the CVF biomarkers and cervical length across gestation in women complicated by sPTB and to compare the differences between those with sPTB and term births.MethodsThis prospective study included a cross-sectional validation study (N=890) and a longitudinal study (N=464) in asymptomatic singleton pregnancies that attended routine Down’s syndrome screening clinic between 11-13 weeks of gestation at Prince of Wales Hospital, Hong Kong, SAR. After recruitment, pregnant women were asked if they could come back for follow up visits at 16-20, 21-24- and 28-32-weeks’ gestation. We obtained maternal characteristics, measured cervical length by transvaginal ultrasonography and collected CVF samples. Through enzyme-linked immunosorbent assay, CVF samples were analysed for the levels of seven previously reported candidate biomarkers that included Interleukin 1 receptor antagonist (IL-1Ra), Vitamin D Binding Protein (VDBP), Tissue inhibitor of matrix metalloproteinases-1 (TIMP-1), Pigment epithelial-derived factor (PEDF), Extra Cellular Matrix protein1 (ECM1), Laminin Subunit Gamma 2 (LAMC2) and Gamma-Glutamyl Hydrolase (GGH).In the cross-sectional study, receiver operating characteristic (ROC) analysis was used to determine the predictive performance of the Carmentix algorithms B and C. Logistic regression analysis was then used to develop four predictive models based on various combinations of maternal characteristics, obstetric history, cervical length and CVF biomarkers for the prediction of sPTB at <37 weeks’ gestation. ROC curves were then generated to determine the performance of each model. In the longitudinal study, the mixed-effects modelling was used to characterise the biomarkers across gestation between preterm and term births. The outcome was sPTB <37 weeks’ gestation.ResultsIn the cross-sectional validation study, the Carmentix biomarker algorithm B achieved an area under ROC (AUC) of 0.505 (95% CI 0.472-0.538) (p=0.924), with detection rates (DRs) of 17.1%, 24.4% and 34.4% at a fixed false positive rates (FPRs) 10%, 20% and 30% respectively for prediction of sPTB at <37 weeks’ gestation. The algorithm C achieved an area under ROC (AUC) of 0.532 (95% CI 0.436-0.630 (p=0.522), with DRs of 19.5%, 24.4% and 29.3% at a fixed FPRs of 10%, 20% and 30% respectively for prediction of sPTB at <37 weeks’ gestation.The combination of maternal characteristics and past obstetrics history alone achieved an area under ROC (AUC) of 0.558 (95% CI 0.524-0.591) (p=0.033), with DRs of 21.0%, 29.0% and 37.1%, at fixed FPRs of 10%, 20% and 30%, respectively for prediction of sPTB at <37 weeks’ gestation. The addition of CVF biomarkers at 16-20 weeks’ gestation to maternal characteristics and past obstetrics history improved the predictive performance, achieving an area under ROC (AUC) of 0.628 (95% CI 0.594-0.660) (p=0.0039), with DRs of 26.1%, 37.0% and 47.8% at fixed FPRs of 10%, 20% and 30%, respectively, for the prediction of sPTB at <37 weeks’ gestation. The addition of cervical length at 21-24 weeks’ gestation to maternal characteristics and past obstetrics history further improved the predictive performance, achieving an AUC of 0.666 (95% CI 0.633-0.698) (p<0.001), with DRs of 28.3%, 43.5% and 56.5% at fixed FPR of 10%, 20% and 30%, respectively, for the prediction of sPTB at <37 weeks’ gestation. The combination of maternal characteristics, past obstetrics history and Carmentix biomarker algorithm B at 16-20 weeks’ gestation achieved an AUC 0.589 (95% CI 0.555-0.623) (p= 0.069), with DR of 29.3%, 34.1% and 43.9%, at fixed FPR of 10%, 20% and 30%, respectively, for the prediction of sPTB at <37 weeks’ gestation.In the longitudinal study, cervical length, IL-1Ra, VDBP, ECM1 and LAMC2 decreases with increasing gestation in pregnant women with sPTB compared to pregnant women with term birth but did not achieve statistical significance, with p-values of0.77, 0.916, 0.639, 0.143 and 0.807 respectively. There were no apparent differences between pregnant women with sPTB and pregnant women with term birth for TIMP1, PEDF and GGH with p-values of 0.311, 0.574 and 0.759 respectively.ConclusionsWe have demonstrated that the Carmentix novel CVF biomarkers, which were initially evaluated in the Melbourne Australian population, have been successfully measured in the Hong Kong Chinese population but failed to achieve the expected performance. We have also demonstrated that the temporal changes of these biomarkers, IL-1Ra, VBBP, ECM1, GGH and LAMC2 across three trimesters have the potential to differentiate between women who deliver preterm spontaneously from those who deliver term.背景自發性早產（sPTB）是指妊娠期 <37周的分娩；該產科問題占世界範圍內分娩人數的 11%。儘管在降低 sPTB發病率方面做出了相當大的努力，它仍然是圍產期發病率和死亡率的主要原因。為了篩查 sPTB，已經出現了多種產科策略，如宮頸長度、胎兒纖維連接蛋白和磷酸化胰島素樣因子結合蛋白 -1。然而，由於 sPTB病因的多樣性，單一測試即可預測 sPTB不太可能實現，而組合測試可能會提高 sPTB的臨床預測性能。因此，我們在本研究中的目標有， 1. 評估由 Carmentix創建的 CVF生物標誌物組在妊娠 16-24周的非選擇性的香港中國人群中對 於 sPTB的預測篩查性能； 2. 建立多指標預測模型（母體特徵、 CVF生物標誌物、宮頸長度）以提高對 sPTB的預測能力； 3. 探討發生 sPTB的孕婦的 CVF生物標誌物和宮頸長度的時間變化特徵，並比較其在自發性早產與足月妊娠間的差異。方法這項前瞻性研究針對孕11-13周在香港威爾斯親王醫院進行常規唐氏綜合征篩查的無症狀單胎孕婦，包括一項橫斷面驗證研究（ N 890）和一項縱向研究 N 464）。招募後，孕婦被詢問是否可以在妊娠 16-20周、 21-24周和 28-32周回來進行隨訪。我們記錄了產婦特徵及產科史等 相關資料，通過陰道超聲測量宮頸長度，並收集CVF樣本。通過酶聯免疫吸附法（ ELISA）對 CVF樣品進行了 7種先前報導的候選生物標誌物的分析，包括白細胞介素 1受體拮抗劑（ IL-1Ra）、維生素 D結合蛋白VDBP）、基質金屬蛋白酶組織抑制劑 1 TIMP-1）、色素上皮源性因 子 PEDF）、細胞外基質蛋白1 ECM1）等，層粘連蛋白亞單位 γ2 LAMC2 γ-穀氨醯水解酶GGH）。在橫斷面研究中 通過 受試者操作特性曲線（ ROC 來 確定 Carmentix算法 B和 C的 預測 性能 。 採用 Logistic回歸分析建立了 4個預測模型，根據母體特徵、產科史、宮頸長度和 CVF生物標誌物的不同組合， 來 預測妊娠 <37周的 sPTB。 並進一步 生成 ROC曲線，以確定各個模型的性能。在縱向研究中，採用混合效應模型來描述早產組和足月組之間的生物標誌物和宮頸長度在妊娠期間的變化。結局變量為妊娠 37周前的 sPTB。結果在橫斷面驗證研究中，Carmentix生物標誌物演算法 B的 AUC為 0.505 95%CI 0.472-0.538; p=0.924），在 FPR為 10%、 20%和 30%時， sPTB<37周的 DR分別為17.1%、 24.4% 和 34.4%。 Carmentix生物標誌物演算法 C的 AUC為 0.532 95%CI 0.0.436-0.630; p=0.522），在 FPR為 10%、 20%和 30%時， sPTB<37周的 DR分別為19.5%、 24.4% 和 29.3%。僅結合母體特徵和既往產科病史，ROC曲線下面積（ AUC）為 0.558 95%CI 0.524-0.591 p=0.033），在 10%、 20%和 30%的固定假陽性率（ FPR）下，對 sPTB<37周的檢出率（ DR）分別為 21.0%、 29.0%和 37.1%。在母體特徵和既往產科史的基礎上，加入妊娠 16-20周的 CVF生物標誌物可提高預測性能， AUC為 0.628 95%CI 0.594-0.660 p=0.0039 FPR為 10%、 20%和 30%時， sPTB<37周的 DR分別為26.1%、 37.0%和 47.8%。將妊娠 21-24周的宮頸長度與母體特徵和既往產科病史相結合，可進一步提高預測性能，AUC為 0.666 95%CI 0.633-0.698 p<0.001 FPR為 10%、 20%和 30%時， sPTB<37周的 DR分別 為 28.3%、 43.5%和 56.5%。結合母體特徵、既往產科史和妊娠 16-20周的 Carmentix生物標誌物演算法 B AUC為0.589 95%CI 0.555-0.623; p=0.069），在 FPR為 10%、 20%和 30%時， sPTB