AbstractBackgroundClinical risk factors for preeclampsia (PE), including previous PE, chronic hypertension, and pregestational diabetes, are poorly predictive of PE. Preterm PE, defined as diagnosis of PE with delivery prior to 37 weeks’ gestational age (GA), is more likely to be associated with serious morbidities and difficult clinical decision making. Therefore, there remains an urgent clinical need to develop a safe, feasible, and accurate predictor of preterm PE that integrates molecular biomarkers and relevant clinical factors into a single risk assessment score that can be used to guide clinical management.Objective(s)To discover, verify, and validate a mid-trimester proteomic biomarker risk predictor for preterm PE, comprised of a composite clinical variable and a small number of maternal serum analytes.Study DesignThis was a secondary analysis of data from two large clinical trials (PAPR,NCT02787213; TREETOP,NCT01371019). PAPR subjects’ eligibility was limited to those who had consented to research into preterm birth and pregnancy complications and who had blood drawn between 180/7– 226/7weeks’ gestation. TREETOP subjects were limited to those who had blood drawn between 180/7– 206/7weeks’ gestation. PAPR subjects were assigned to a discovery cohort, and TREETOP subjects were randomly assigned to a first-phase cohort for verification (comprised of one-third of eligible subjects) and to a separate second-phase cohort for validation (comprised of the remaining two-thirds of eligible subjects). Peptides were analyzed by liquid chromatography-multiple reaction monitoring mass spectrometry measuring 77 pregnancy-related proteins and quality control proteins. Models were limited to a maximum of one additional protein ratio and a composite clinical variable, referred to as Clin3, which was deemed positive if any of three factors was true for the subject: prior PE; pre-existing hypertension; and/or pregestational diabetes. Overall classifier performance was assessed via area under the receiver operating characteristic curve (AUC).ResultsVerification yielded nine multi-component classifier models for prediction of preterm PE, all of which were subsequently validated. Classifiers exhibited greater predictive performance than clinical factors alone. Example performance metrics across a range of classifier score thresholds and GA at birth cutoffs of 37, 34 and 32 weeks for the Clin3 + inhibin subunit beta c (INHBC)/SHBG classifier, which showed the highest AUC, demonstrating a sensitivity of 89% at a specificity of 75% for prediction of early-onset preeclampsia (<34 weeks’ GA).Conclusion(s)Here, we report on discovery, verification, and validation of models for prediction of preterm PE. The log ratio of INHBC/SHBG along with any one of three clinical risk factors demonstrated high sensitivity and specificity. This combination of protein biomarkers and clinical factors has the potential to be used in the mid-trimester of pregnancy to guide clinical management to avoid both unnecessary medical procedures and the most serious complications of early-onset PE.

Preterm births are the leading cause of neonatal death in the United States. Previously, a spontaneous preterm birth (sPTB) predictor based on the ratio of two proteins, IBP4/SHBG, was validated as a predictor of sPTB in the Proteomic Assessment of Preterm Risk (PAPR) study. In particular, a proteomic biomarker threshold of -1.37, corresponding to a ~two-fold increase or ~15% risk of sPTB, significantly stratified earlier deliveries. Guidelines for molecular tests advise replication in a second independent study. Here we tested whether the significant association between proteomic biomarker scores above the threshold and sPTB, and associated adverse outcomes, was replicated in a second independent study, the Multicenter Assessment of a Spontaneous Preterm Birth Risk Predictor (TREETOP). The threshold significantly stratified subjects in PAPR and TREETOP for sPTB (p = 0.041, p = 0.041, respectively). Application of the threshold in a Kaplan-Meier analysis demonstrated significant stratification in each study, respectively, for gestational age at birth (p < 001, p = 0.0016) and rate of hospital discharge for both neonate (p < 0.001, p = 0.005) and mother (p < 0.001, p < 0.001). Above the threshold, severe neonatal morbidity/mortality and mortality alone were 2.2 (p = 0.0083,) and 7.4-fold higher (p = 0.018), respectively, in both studies combined. Thus, higher predictor scores were associated with multiple adverse pregnancy outcomes.

The clinical management of pregnancy and spontaneous preterm birth (sPTB) relies on estimates of gestational age (GA). Our objective was to evaluate the effect of GA dating uncertainty on the observed performance of a validated proteomic biomarker risk predictor, and then to test the generalizability of that effect in a broader range of GA at blood draw. In a secondary analysis of a prospective clinical trial (PAPR; NCT01371019), we compared two GA dating categories: both ultrasound and dating by last menstrual period (LMP) (all subjects) and excluding dating by LMP (excluding LMP). The risk predictor's performance was observed at the validated risk predictor threshold both in weeks 19^1/7-20^6/7 and extended to weeks 18^0/7-20^6/7. Strict blinding and independent statistical analyses were employed. The validated biomarker risk predictor showed greater observed sensitivity of 88% at 75% specificity (increases of 17% and 1%) in more reliably dated (excluding-LMP) subjects, relative to all subjects. Excluding dating by LMP significantly improved the sensitivity in weeks 19^1/7-20^6/7. In the broader blood draw window, the previously validated risk predictor threshold significantly stratified higher and lower risk of sPTB, and the risk predictor again showed significantly greater observed sensitivity in excluding-LMP subjects. These findings have implications for testing the performance of models aimed at predicting PTB.

Background

Preterm birth remains a common and devastating complication of pregnancy. There remains a need for effective and accurate screening methods for preterm birth. Using a proteomic approach, we previously discovered and validated (Proteomic Assessment of Preterm Risk study, NCT01371019) a preterm birth predictor comprising a ratio of insulin-like growth factor-binding protein 4 to sex hormone-binding globulin.

Objective

To determine the performance of the ratio of insulin-like growth factor-binding protein 4 to sex hormone-binding globulin to predict both spontaneous and medically indicated very preterm births, in an independent cohort distinct from the one in which it was developed.

Study design

This was a prospective observational study (Multicenter Assessment of a Spontaneous Preterm Birth Risk Predictor, NCT02787213) at 18 sites in the United States. Women had blood drawn at 17^0/7 to 21^6/7 weeks' gestation. For confirmation, we planned to analyze a randomly selected subgroup of women having blood drawn between 19^1/7 and 20^6/7 weeks' gestation, with the results of the remaining study participants blinded for future validation studies. Serum from participants was analyzed by mass spectrometry. Neonatal morbidity and mortality were analyzed using a composite score by a method from the PREGNANT trial (NCT00615550, Hassan et al). Scores of 0-3 reflect increasing numbers of morbidities or length of neonatal intensive care unit stay, and 4 represents perinatal mortality.

Results

A total of 5011 women were enrolled, with 847 included in this planned substudy analysis. There were 9 preterm birth cases at <32^0/7 weeks' gestation and 838 noncases at ≥32^0/7 weeks' gestation; 21 of 847 infants had neonatal composite morbidity and mortality index scores of ≥3, and 4 of 21 had a score of 4. The ratio of insulin-like growth factor-binding protein 4 to sex hormone-binding globulin ratio was substantially higher in both preterm births at <32^0/7 weeks' gestation and there were more severe neonatal outcomes. The ratio of insulin-like growth factor-binding protein 4 to sex hormone-binding globulin ratio was significantly predictive of birth at <32^0/7 weeks' gestation (area under the receiver operating characteristic curve, 0.71; 95% confidence interval, 0.55-0.87; P=.016). Stratification by body mass index, optimized in the previous validation study (222), resulted in an area under the receiver operating characteristic curve of 0.76 (95% confidence interval, 0.59-0.93; P=.023). The ratio of insulin-like growth factor-binding protein 4 to sex hormone-binding globulin ratio predicted neonatal outcomes with respective area under the receiver operating characteristic curve of 0.67 (95% confidence interval, 0.57-0.77; P=.005) and 0.78 (95% confidence interval, 0.63-0.93; P=.026) for neonatal composite morbidity and mortality scores of ≥3 or 4. In addition, the ratio of insulin-like growth factor-binding protein 4 to sex hormone binding globulin significantly stratified neonates with increased length of hospital stay (log rank P=.023).

Conclusion

We confirmed in an independent cohort the ratio of insulin-like growth factor-binding protein 4 to sex hormone-binding globulin ratio as a predictor of very preterm birth, with additional prediction of increased length of neonatal hospital stay and increased severity of adverse neonatal outcomes. Potential uses of the ratio of insulin-like growth factor-binding protein 4 to sex hormone-binding globulin predictor may be to risk stratify patients for implementation of preterm birth preventive strategies and direct patients to appropriate levels of care.

Objectives

Preterm birth occurs in more than 10% of U.S. births and is the leading cause of U.S. neonatal deaths, with estimated annual costs exceeding $25 billion USD. Using real-world data, we modeled the potential clinical and economic utility of a prematurity-reduction program comprising screening in a racially and ethnically diverse population with a validated proteomic biomarker risk predictor, followed by case management with or without pharmacological treatment.

Methods

The ACCORDANT microsimulation model used individual patient data from a prespecified, randomly selected sub-cohort (N = 847) of a multicenter, observational study of U.S. subjects receiving standard obstetric care with masked risk predictor assessment (TREETOP; NCT02787213). All subjects were included in three arms across 500 simulated trials: standard of care (SoC, control); risk predictor/case management comprising increased outreach, education and specialist care (RP-CM, active); and multimodal management (risk predictor/case management with pharmacological treatment) (RP-MM, active). In the active arms, only subjects stratified as higher risk by the predictor were modeled as receiving the intervention, whereas lower-risk subjects received standard care. Higher-risk subjects' gestational ages at birth were shifted based on published efficacies, and dependent outcomes, calibrated using national datasets, were changed accordingly. Subjects otherwise retained their original TREETOP outcomes. Arms were compared using survival analysis for neonatal and maternal hospital length of stay, bootstrap intervals for neonatal cost, and Fisher's exact test for neonatal morbidity/mortality (significance, p < .05).

Results

The model predicted improvements for all outcomes. RP-CM decreased neonatal and maternal hospital stay by 19% (p = .029) and 8.5% (p = .001), respectively; neonatal costs' point estimate by 16% (p = .098); and moderate-to-severe neonatal morbidity/mortality by 29% (p = .025). RP-MM strengthened observed reductions and significance. Point estimates of benefit did not differ by race/ethnicity.

Conclusions

Modeled evaluation of a biomarker-based test-and-treat strategy in a diverse population predicts clinically and economically meaningful improvements in neonatal and maternal outcomes.

AbstractObjectivesPreterm birth occurs in more than 10% of U.S. births and is the leading cause of U.S. neonatal deaths, with estimated annual costs exceeding $25 billion USD. Using real-world data, we modeled the potential clinical and economic utility of a prematurity-reduction program comprising screening in a racially and ethnically diverse population with a validated proteomic biomarker risk predictor, followed by case management with or without pharmacological treatment.MethodsThe ACCORDANT microsimulation model used individual patient data from a prespecified, randomly selected sub-cohort (N=847) of a multicenter, observational study of U.S. subjects receiving standard obstetric care with masked risk predictor assessment (TREETOP; NCT02787213). All subjects were included in three arms across 500 simulated trials: standard of care (SoC, control); risk predictor/case management comprising increased outreach, education and specialist care (RP-CM, active); and risk predictor/case management with pharmacological treatment (RP-MM, active). In the active arms, only subjects stratified as higher-risk by the predictor were modeled as receiving the intervention, whereas lower-risk subjects received standard care. Higher-risk subjects’ gestational ages at birth were shifted based on published efficacies, and dependent outcomes, calibrated using national datasets, were changed accordingly. Subjects otherwise retained their original TREETOP outcomes. Arms were compared using survival analysis for neonatal and maternal hospital length of stay, bootstrap intervals for neonatal cost, and Fisher’s exact test for neonatal morbidity/mortality (significance, p<0.05).ResultsThe model predicted improvements for all outcomes. RP-CM decreased neonatal and maternal hospital stay by 19% (p=0.029) and 8.5% (p=0.001), respectively; neonatal costs’ point estimate by 16% (p=0.098); and moderate-to-severe neonatal morbidity/mortality by 29% (p=0.025). RP-MM strengthened observed reductions and significance. Point estimates of benefit did not differ by race/ethnicity.ConclusionsModeled evaluation of a biomarker-based test-and-treat strategy in a diverse population predicts clinically and economically meaningful improvements in neonatal and maternal outcomes.Plain language summaryPreterm birth, defined as delivery before 37 weeks’ gestation, is the leading cause of illness and death in newborns. In the United States, more than 10% of infants is born prematurely, and this rate is substantially higher in lower-income, inner-city and Black populations. Prematurity associates with substantially increased risk of short- and long-term medical complications and can generate significant costs throughout the lives of affected children. Annual U.S. health care costs to manage short- and long-term prematurity complications are estimated to exceed $25 billion.Clinical interventions, including case management (increased patient outreach, education and specialist care), pharmacological treatment and their combination, can provide benefit to pregnancies at higher risk for preterm birth. Early and sensitive risk detection, however, remains a challenge.We have developed and validated a proteomic biomarker risk predictor for early identification of pregnancies at increased risk of preterm birth. The ACCORDANT study modeled treatments with real-world patient data from a racially and ethnically diverse U.S. population to compare the benefits of risk predictor testing plus clinical intervention for higher-risk pregnancies versus no testing and standard care. Measured outcomes included neonatal and maternal length of hospital stay, associated costs and neonatal morbidity and mortality. The model projected improved outcomes and reduced costs across all subjects, including ethnic and racial populations, when predicted higher-risk pregnancies were treated using case management with or without pharmacological treatment. The biomarker risk predictor shows high potential to be a clinically important component of risk stratification for pregnant women, leading to tangible gains in reducing the impact of preterm birth.

Development of effective prevention and treatment strategies for pre-eclampsia is limited by the lack of accurate methods for identification of at-risk pregnancies. We performed small RNA sequencing (RNA-seq) of maternal serum extracellular RNAs (exRNAs) to discover and verify microRNAs (miRNAs) differentially expressed in patients who later developed pre-eclampsia. Sera collected from 73 pre-eclampsia cases and 139 controls between 17 and 28 weeks gestational age (GA), divided into separate discovery and verification cohorts, are analyzed by small RNA-seq. Discovery and verification of univariate and bivariate miRNA biomarkers reveal that bivariate biomarkers verify at a markedly higher rate than univariate biomarkers. The majority of verified biomarkers contain miR-155-5p, which has been reported to mediate the pre-eclampsia-associated repression of endothelial nitric oxide synthase (eNOS) by tumor necrosis factor alpha (TNF-α). Deconvolution analysis reveals that several verified miRNA biomarkers come from the placenta and are likely carried by placenta-specific extracellular vesicles.