The utility of biomarkers in the prediction of preterm birth: A nested case control study of the ASPIRIN trial

Introduction and Aim: Adverse pregnancy outcomes (APO’s-preterm birth, foetal growth restriction and hypertensive disorders of pregnancy (HDP)) are the primary drivers of perinatal mortality. C-Reactive Protein, Anti-Mullerian Hormone (AMH) and Alpha-FetoProtein (AFP) have shown promise in predicting these APO’s. Aim of this present work is to validate the predictive ability of these biomarkers to identify women at risk for preterm birth. Materials and Methods: A nested observational case-control study was performed drawing nulliparous pregnant women between 10 to 13 weeks and 17 to 21 weeks of gestation in nulliparous women randomized to aspirin or placebo as part of the ASPIRIN (Aspirin Supplementation for Pregnancy Indicated Risk Reduction in Nulliparas) trial. Results: A total of 126 women with these APO’s and 143 controls were selected for analyte analysis. None of the chosen analytes were found to predict preterm birth before 37 weeks or HDP. AFP obtained between 10 to 13 weeks, was able to moderately discriminate against women who had a preterm birth prior to 34 weeks (C-statistic 0.65-95% CI 0.55 to 0.78). No other analytes were found to be predictive of preterm birth prior to 34 weeks. Conclusion: Elevated Alpha-FetoProtein early in pregnancy is associated with early preterm birth (PTB) and may be a marker of Aspirin efficacy.


INTRODUCTION
reterm birth currently affects 15 million children a year; resulting in 1 million neonatal deaths which disproportionately occur in low resource settings (1,2).Beyond mortality, the impact of preterm birth frequently continues throughout the life of offspring resulting in higher rates of chronic medical conditions and developmental delays (3)(4)(5).
Prophylactic strategies such as progesterone (6,7), cervical surveillance (8), care management (9,10) and more recently low dose aspirin (11,12) therapy have been suggested approaches to prevent preterm birth (PTB); however, the ability to accurately identify at risk women to meaningfully direct resources to has been elusive.Several candidate biomarkers for PTB have been suggested including anti-mullerian hormone (AMH) (13)(14)(15), alpha fetoprotein (16) and c-reactive protein (CRP) (17).Nonetheless these studies are noted to be conflicting in their ability to predict preterm birth.We have earlier reported the impact of low-dose aspirin on markers of inflammation and placental function (18).This work was sought to validate the predictive ability of these biomarkers to identify women at risk for preterm birth in a single centre ancillary study of women participating in the global network study testing the effectiveness of low dose aspirin in reducing preterm birth (11).

MATERIALS AND METHODS
We performed a case control study of women participating in the ASPIRIN (Aspirin Supplementation for Pregnancy Indicated Risk Reduction in Nulliparas) Trial conducted at the Global Network site, J N Medical College, Belagavi Karnataka India.Main trial randomized nulliparous women with less than three previous first trimester P pregnancy losses and treated with Low Dose Aspirin (LDA) 81mg daily started between 6 0/7 weeks and 13 6/7 weeks continued through 36 0/7 weeks gestational age (GA) for the prevention of preterm birth (11).This ancillary study was approved by JNMC Institutional ethics committee (IRB00001499) and pregnant women were individually consented for participation.(Clinical Trial Registration: ClinicalTrials.gov,NCT02409680, and Clinical Trial Registry of India, CTRI/2016/05/006970).
As part of the study, serum samples were collected and stored in ultralow temperature freezers (-80 0 C degree) until analysis between 10-13 weeks of gestation.Further, a second sample was collected from the same women during the second trimester of pregnancy (at 17-21 weeks' gestation).Women who had an abortion before 16 weeks of pregnancy were excluded from the analysis.Women delivering before 37 weeks of pregnancy were identified as cases and those women delivering at or after 37 weeks' gestation were considered as controls.
Power analysis was not performed in this analysis due to the exploratory nature of the study.The goal of the analysis was to include about 50 cases (preterm birth) and about 150 controls (Term deliveries).We choose a separate case or control because of limitations in sample, rather than considering a fixed cohort of participants used for all specimens.
The quantitative estimation of AFP (Catalog Number: R&D Systems, kit# DAFP00) and AMH (Catalog Number: DY1737 R&D Systems) were conducted using standardized immunoassays adopted from the protocols of the manufacturer.CRP was estimated by Immunoturbidimetric assay ((Roche Diagnostics) as per the kit manufacturer's protocol.The primary outcome of the study was to determine the predictive value of AMH, CRP or AFP in predicting (C-statistic) either alone or in combination for the prediction of PTB defined as delivery before 37 weeks of gestation.Those variables that demonstrated trend (e.g., had a p-value < 0.1) were in stepwise logistic regression to create a multivariable logistic model.Those that retained a p-value of <0.1 in multivariable modelling were retained.Models were then compared using the ROCCOMP command in Stata 15.0 (College Station Tx).Secondary outcomes of this analysis were PTB defined as birth less than 34 weeks of pregnancy and hypertensive disorders of pregnancy (HDP).Further, we also compared the impact of LDA on the predictive value of these biomarkers as it may have impacted the outcome of PTB, PTB before 34 weeks and HDP.An odd's ratio of each individual analyte was performed as well as subgroups of women who took aspirin or placebo as the use of aspirin may have altered these values.All of the analyses were performed using STATA v15.1 (Colleges Station, TX) (19).

RESULTS
A total of 666 women participated in this exploratory analysis of which 269 had results of the analyses and a total of seven participants experienced an abortion.Finally, 262 women were included (45 preterm cases and 217 term controls) in the analysis.Maternal demographics were similar between the women belonging to aspirin and placebo groups as well as the overall cohort of 269 women (table 1).There was no difference in the analytes relating to either PTB or HDP; excepting that AFP was found to be lower (P=0.02)amongst mothers who delivered before 34 weeks of pregnancy at the first time point of blood collection.Though not statistically significant, CRP trended (P=0.09)lower amongst women who delivered prior to 34 weeks at the first blood draw.Recognizing that LDA therapy may have affected the outcome, the performance of each analyte overall and by their treatment (Aspirin vs. Placebo) using logistic regression is presented in table 3.As anticipated the first draw of AFP in women who received placebo remained statistically significant but not for women who received aspirin.This positive association suggested that higher CRP levels are associated with longer GA.The AUCs of 3 different models of prediction of PTB<34 weeks (AFP1, CRP1, AFP1 and CRP1) are displayed in figure 1. AFP at the first blood draw was noted to be mildly predictive of preterm birth before 34 weeks (C-statistic 0.65-95% CI 0.55 to 0.78).Recognizing CRP at the first blood draw tended to discriminate against women delivering preterm before 34 weeks, three models of prediction using AFP, CRP and AFP and CRP were created to predict preterm birth before 34 weeks.When compared statistically (ROCCOMP function in Stata), none of the models were superior to each other (Pvalues 0.11, 0.77 and 0.09; Fig. 1).Reactive Protein levels at first trimester time point

DISCUSSION
Our study demonstrates that low AFP was associated with early PTB before 34 weeks.Similarly, we saw a trend with CRP but this was not statistically significant and directionally opposite of what we would have clinically anticipated (higher CRP being associated with lower rates of PTB before 34 weeks).AFP has been shown to be associated with bad obstetrical outcomes including both PTB and placental malperfusion disorders (16,(20)(21)(22)(23)(24).However, limited data in this study did not yield a correlation of first trimester AFP with other obstetric outcome parameters, this may be a reflection of smaller sample size.Not surprisingly this difference did not persist when the group of women randomized to ASPIRIN were uniquely examined.These results suggest measurement of AFP during the first trimester may be useful as a biomarker of LDA efficacy in pregnancy; nevertheless, larger studies for validation are required.
The AMH or CRP levels were not different at both the gestational age time points.AMH was associated with PTB in previous studies conducted mostly with infertility issues which may overlap with nonobstetrical conditions (13,15).Finally, CRP, being an acute marker of inflammation, we found a trend with preterm birth before 34 weeks but no other obstetrical outcomes.
The present study has several strengths and weaknesses as an exploratory design.Firstly, this is a nested case control study embedded with ASPIRIN study with a large cohort of women and maintaining the blinding of treatment allocation throughout the study.Secondly, GA was corroborated using early dating ultrasound.Major limitation of the study is that overall numbers of specimens was variable due to participant refusal for blood draw at second time point, technical challenges with the assays (haemolysis/quantity insufficient) and the lack of available controls.Further, due to limitation of the overall sample size adequate power may be lacking in this study to detect meaningful differences.

CONCLUSION
Results of study suggest that AFP during the first trimester may predict preterm birth before 34 weeks' gestation; however, is not predictive of other obstetrical outcomes.Unfortunately, most probably because of the smaller sample size, we were not able to demonstrate the predictive abilities of AFP, CRP and AMH on other important obstetrical outcomes.

CONFLICT OF INTEREST
Authors report no conflicts of interest.

Fig. 1 :
Fig. 1: Comparison of AFP1, CRP1 and Combined.AFP1: Alpha-fetoprotein at first trimester time point, CRP1: C Reactive Protein levels at first trimester time point

Table 1 :
Maternal Demographic parameters & Delivery Outcomes Interquartile range; Cm-Centimeter; Kg-Kilogram; BMI-Body Mass Index; PTB-Preterm Birth; gm-gram.a Wilcoxon Rank-Sum b Fisher's Exact Table 2 depicts the results of the analytes.Please note that the numbers are not equally distributed by analyte due to availability of specimens in each group or technical issues related to individual assays.The N of each specimen are included in the table.Wilcoxon Rank-Sum tests are reported as values for analytes were not normally distributed by the Shapiro-Wilk test.

Table 2 :
Serum analytes and delivery outcomes

Table 3 :
Logistic regression of analytes and outcomes Alpha-fetoprotein at first and second trimester time points, CRP1 and 2: C Reactive Protein levels at first and second trimester time points: AMH1 and 2: Anti Mullerian Hormone levels at first and second trimester time points