Integrated Proteomics Pipeline Yields Novel Biomarkers for Predicting PreeclampsiaNovelty and Significance
Preeclampsia, a hypertensive pregnancy complication, is largely unpredictable in healthy nulliparous pregnant women. Accurate preeclampsia prediction in this population would transform antenatal care. To identify novel protein markers relevant to the prediction of preeclampsia, a 3-step mass spectrometric work flow was applied. On selection of candidate biomarkers, mostly from an unbiased discovery experiment (19 women), targeted quantitation was used to verify and validate candidate biomarkers in 2 independent cohorts from the SCOPE (SCreening fOr Pregnancy Endpoints) study. Candidate proteins were measured in plasma specimens collected at 19 to 21 weeks’ gestation from 100 women who later developed preeclampsia and 200 women without preeclampsia recruited from Australia and New Zealand. Protein levels (n=25), age, and blood pressure were then analyzed using logistic regression to identify multimarker models (maximum 6 markers) that met predefined criteria: sensitivity ≥50% at 20% positive predictive value. These 44 algorithms were then tested in an independent European cohort (n=300) yielding 8 validated models. These 8 models detected 50% to 56% of preeclampsia cases in the training and validation sets; the detection rate for preterm preeclampsia cases was 80%. Validated models combine insulin-like growth factor acid labile subunit and soluble endoglin, supplemented with maximally 4 markers of placental growth factor, serine peptidase inhibitor Kunitz type 1, melanoma cell adhesion molecule, selenoprotein P, and blood pressure. Predictive performances were maintained when exchanging mass spectrometry measurements with ELISA measurements for insulin-like growth factor acid labile subunit. In conclusion, we demonstrated that biomarker combinations centered on insulin-like growth factor acid labile subunit have the potential to predict preeclampsia in healthy nulliparous women.
Preeclampsia continues to be a major cause of maternal mortality, resulting in >50 000 maternal deaths worldwide each year, and is the leading cause of iatrogenic preterm birth.1 To prevent preeclampsia, women at high risk of the condition need to be identified early in pregnancy. Although there is significant interest in the prediction of preeclampsia using combinations of clinical risk factors, biophysical measurements, and biochemical tests, to date no screening test has achieved the requisite sensitivity and specificity to be useful and cost-effective in a clinical setting.2–5
Prediction of preeclampsia in healthy nulliparous women is particularly challenging, despite the greatest proportion of cases occurring in this population. The best known combination of markers tested in a low-risk nulliparous population had a sensitivity of 46% for a specificity of 80%, equating to a PPV of around 15.5%.4 Other reports of better prediction have studied general obstetric populations that include high-risk women3 or have used a nested case–control design with controls comprising uncomplicated pregnancies with the consequent overestimation of predictive performance.6
A screening test is likely to require multiple biomarkers that reflect different aspects of the complex pathological processes that culminate in preeclampsia.7 Several proteins indicative of abnormal placentation, such as placental growth factor (PlGF) and pregnancy-associated plasma protein A, have been demonstrated to be predictive of preeclampsia, especially preterm disease.8 Novel plasma biomarkers, representative of placentation or the maternal vascular and inflammatory response in preeclampsia, may be discovered using an unbiased proteomic approach. Unfortunately, to date, most proteomic research, which has aimed to discover biomarkers, has failed to incorporate adequate biomarker validation studies in independent sample sets. These are necessary steps in the translation of potential biomarkers into clinical practice. Recent development of sophisticated mass spectrometry–based quantitation of multiple proteins has enabled the validation of large sets of candidate biomarkers in plasma.9
The objective of this study was to identify, verify, and validate panels of biomarkers, which are predictive of preeclampsia. Given that in current practice women with an estimated ≥20% risk of developing preeclampsia are referred for specialist prenatal care,10 we aimed to develop a test with ≥50% sensitivity for a positive predictive value (PPV) of 20%. Selective reaction monitoring (SRM) was used to verify and validate panels of biomarkers in 2 independent sample sets from a prospective, international cohort of nulliparous women (www.scopestudy.net). In the preeclampsia prediction panels developed, insulin-like growth factor acid labile subunit (IGFALS), a novel preeclampsia biomarker, carries the most predictive weight.
Participants and Specimens
Local ethical committee approval was granted, and written informed consent was obtained from all participants.
Healthy, normotensive, nulliparous, and multiparous women (n=222) were recruited at Ninewells Hospital, Dundee, UK, after assessment of uterine artery Doppler waveform at a routine clinical visit, and an EDTA plasma was obtained at 22 and 26 weeks of gestation.11 Pregnancy outcome data were available in all women, of whom 26 women (12%) developed preeclampsia defined using standard criteria.12 Ten women with preeclampsia were matched for parity, ethnicity, and gestation at sampling to women with uncomplicated pregnancies (n=9).
Biomarker Verification and Validation
Women who were recruited into the SCOPE (SCreening fOr Pregnancy Endpoints) study, a prospective screening study of low-risk nulliparous women recruited in Australia, New Zealand, United Kingdom, and Ireland between November 2004 and July 2011 (ACTRN12607000551493), participated in this study.2 A research midwife interviewed participants at 14 to 16 weeks’ and 19 to 21 weeks’ gestation, and pregnancy outcomes were prospectively tracked. At the time of interview, data were entered on the Internet-accessed central database (MedSciNet). Two consecutive manual blood pressure measurements were recorded. Blood samples were collected on EDTA at 14 to 16 and 19 to 21 weeks, and plasma was stored at −80°C within 4 hours of collection.
One hundred women who developed preeclampsia and 200 controls were randomly selected from the 3182 women recruited in Australia and New Zealand. Controls were selected 2:1 from those who did not have preeclampsia at the same center and included women with uncomplicated pregnancies and those with complications, such as small for gestational age, preterm birth, gestational hypertension, and gestational diabetes mellitus. Preeclampsia was defined as systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg, or both, on ≥2 occasions 4 hours apart after 20 weeks’ gestation but before the onset of labor, or postpartum, with either proteinuria (24-hour urinary protein ≥300 mg or spot urine protein:creatinine ratio ≥30 mg/mmol creatinine or urine dipstick protein ≥++) or any multisystem complication of preeclampsia.2
Fifty cases of preeclampsia and 5:1 controls (no preeclampsia), stratified by center, were randomly selected from women recruited to the European centers (London, Manchester, Leeds, UK, and Cork, Ireland; n=2423).
Mass Spectrometry Methods
N-Terminomics Discovery Platform
An N-terminomics platform, described in Mebazaa et al,9 was used to identify candidate biomarkers in the 22- and 26-week discovery samples. In brief, N-terminomics COFRADIC (COmbined FRActional DIagonal Chromatography)13,14 for complexity reduction and spotting on MALDI (matrix-assisted laser desorption/ionization) targets was used. Experimental details are provided in the online-only Data Supplement.
Quantification of Candidate Biomarkers
The candidate proteins were quantified in the training sample set with targeted mass spectrometry assays based on an SRM peptide quantification method,15 using custom-built assays. In brief, plasma samples were depleted of albumin and IgG, denatured and spiked with a mixture of isotopically labeled peptides serving as internal reference. After tryptic digestion, and peptide separations, quantitative data were obtained with a triple quadrupole mass spectrometry (MS) instrument. The readout of an assay in each sample was the ratio of the analyte signal area (endogenous peptide) over the common internal standard signal area. Comparison of ratios between different samples represents the relative quantification of the protein. Detailed MS methods are described in the online-only Data Supplement.
The sample order was randomized before every analytic step, and laboratory personnel were blinded to the pregnancy outcome related to each sample. Technical variation was estimated by preparing and measuring in duplicate 10% of the samples in a randomized order.
PlGF was measured in all samples using DELFIA time resolve fluorescence technology (PerkinElmer, Turku, Finland). Interassay coefficients of variation were 3% at 16.8 pg/mL and 8% at 852 pg/mL. In the validation samples, only IGFALS was also measured, in duplicate, by ELISA (Mediagnost, Reutlingen, Germany); samples were randomized and blinded to the Mediagnost laboratory. Coefficients of variation (CVs) were <8% for all 5 reference samples.
Biomarker Panel Development
The size of the training set was chosen to achieve an accuracy for the sensitivity of ±10% with a confidence level of 95% for the minimum accepted performance (50% detection rate [sensitivity] at a PPV of 20%). The minimum total number of samples required to achieve this performance was 288.
R and bioconductor were used to perform all statistical analyses.16 The characteristics of the preeclampsia group and controls were compared using Student t test, Wilcoxon rank sum test, and χ2 test. Logistic regression was used to develop multivariable models. The clinical parameters (maternal age and mean arterial pressure [MAP]; no missing values) obtained at 20 weeks’, protein assays (log transformed) with <20% missing values, and a ≤25% CV were used for the multivariable analysis.
The modeling aimed to discover all marker combinations predictive of preeclampsia using a maximum of 6 covariates to limit the risk of overfitting the data. For each combination, a logistic regression model was fitted on the participants with complete data; observations with outlying values were discarded. A conservative stepwise approach was used to select the models. First, the statistical significance of all coefficients was estimated using the Wald test. A model was ignored when the Wald test for one of the coefficients associated with a covariate was P>0.05. For the retained models, the discriminatory power was then estimated using the area under receiver-operator curve (AUC). Models with an AUC below 0.70 were ignored (this AUC corresponds to the AUC of the best univariate predictor; IGFALS). Finally, the sensitivity at 20% PPV was computed for the remaining models, and those with a sensitivity of ≥50%, the preset threshold, were retained for external validation.
Biomarker Panel Validation
The selected models were evaluated in the European samples (validation set). The performance was computed in the validation set using the models developed in the training set without any refitting. Models were considered externally validated if the sensitivity was ≥50% at 20% PPV (Figure S3B in the online-only Data Supplement). Given the validation set comprised 50 cases and 250 controls, the accuracy of the sensitivity observed is expected to be ±14% for the target performance of 50% sensitivity at 20% PPV. The possibility of a validated model occurring by chance was also assessed (see the online-only Data Supplement).
An overview of the steps taken to develop, verify, and validate the prediction models is outlined in Figure 1. The participants from the SCOPE study included in the training and validation data sets for biomarker verification and validation, respectively, are shown in Figure S1.
An N-terminomics platform was used to compare the plasma proteomes of women destined to develop preeclampsia (n=10) with women who had uncomplicated pregnancies (n=9) at 22- and 26-week gestation (Table S1). From this discovery experiment, 64 proteins were selected for verification (Table S2). Previously reported markers for preeclampsia, such as soluble endoglin (sEng), disintegrin, and metalloproteinase domain–containing protein 12 (ADAM12) were identified. In addition to these 64 proteins, 9 proteins previously identified in a cardiovascular biomarker study9 with biology relevant to preeclampsia (Table S2) and 3 proteins (PlGF, soluble fms-like tyrosine kinase-1, and placental protein 13) with a recognized association with preeclampsia were also taken forward to the verification experiments.
Biomarker Verification and Model Development
The characteristics of participants in the training and validation sets are shown in Table 1. SRM assays were successfully developed for 51 proteins from the list of candidate biomarkers. SRM data with a CV ≤25% and ≤20% missing values were obtained for 24 different proteins (Table S2).
Univariate analysis revealed that IGFALS was significantly elevated in 19 to 21-week plasma from women who later developed preeclampsia compared with controls (Table S3; Figure S2). Furthermore, IGFALS was increased before both preterm (<37 weeks; n=30) and term preeclampsia (n=70). IGFALS had the highest performance as a single marker with 48% (95% confidence interval [CI], 37% to 59%) sensitivity at 80% specificity (Table S4). PlGF, sEng, ADAM12, and 20-week MAP also significantly discriminated women destined to develop preterm preeclampsia from control pregnancies (P<0.001; Figure S2).
Development of Models in Training Set
Forty-four models had a prediction performance higher than the predefined cutoff (sensitivity ≥50% at 20% PPV; Figure S3A). There was significant overlap of protein biomarkers in these prediction models, with a small number of biomarkers (PlGF, IGFALS, melanoma cell adhesion molecule [MCAM], sEng, ADAM12, serine peptidase inhibitor Kunitz type 1 [SPINT1]) appearing in the majority of algorithms.
Validation of Prediction Models
Of the 44 models, 8 reached the target performance of 50% sensitivity at 20% PPV for a 5% prevalence in the validation set (Figure S3B). These validated models included combinations of the proteins IGFALS, sEng, ADAM12, SPINT1, MCAM, selenoprotein P, multimerin-2, extracellular matrix protein 1, microtubule-associated protein RP/EB family member 1 or 3, fructose-bisphosphate aldolase A, PlGF, and blood pressure (MAP), Table 2. The likelihood of validating 1 model by chance was computed to be <1%.
The 8 validated models all showed very similar performance for overall preeclampsia prediction (Tables S5 and S6). With the exception of 1 model, these models combine IGFALS and sEng and a selection of 3 or 4 markers out of SPINT1, PlGF, MCAM, selenoprotein P, and MAP. The model that combines the 6 most frequently occurring covariates was selected as an example (Figure 2). Using the model, a risk index (relative risk to develop preeclampsia) was computed for each patient. A risk index cutoff corresponding to 20% PPV was computed on the training set. The cutoff corresponds to a detection rate (sensitivity) of 54% (95% CI, 37% to 66%) in the training set and 50% (95% CI, 36% to 68%) in the validation set. Preterm preeclampsia occurred in 30 women in the training and 12 women in the validation sets. Using the model for all preeclampsia and the same risk index cutoff, the detection of preterm preeclampsia was 72% (95% CI, 48% to 88%) in the training set and 80% (95% CI, 50% to 100%) in the validation set (Figure 2; Tables S5 and S6). Application of this model to a theoretical population of 1000 women would classify 125 women as being high risk of developing preeclampsia. Twenty percent of this high-risk group would later develop preeclampsia, and 10 of the 13 women who would develop preterm preeclampsia would be detected. In the test negative group, 2.9% would develop preeclampsia with 0.29% having preterm disease compared with an unstratified nulliparous population where 1.2% of pregnancies would develop preterm preeclampsia.
The incremental value of the novel biomarkers over the known markers was investigated by calculating the performances of any combination of sENG, PlGF, ADAM12, and MAP: within the training data set. The best model (combination of PLGF, ADAM12, and MAP) had a sensitivity of 30% at 20% PPV; data not shown). A comparison with the performance of the National Institute for Health and Clinical Excellence (NICE) risk factor model17 and the best combination of markers in a comparable population4 are presented in Table 3.
Substitution of SRM Data With ELISA Data for IGFALS
There was good correlation between the SRM and ELISA measurements of IGFALS (r=0.63; P<0.001; Spearman rank correlation; Figure S4). Substitution of IGFALS SRM data with ELISA measurements in the example model did not change its performance. The risk index of the model using the SRM readouts also correlated well with the risk index using the ELISA measurements (r=0.89; P<0.001; Spearman rank correlation; Figure S4), resulting in a detection rate of 59% (95% CI, 41% to 73%) at 20% PPV.
In this study, we identified a number of novel biomarkers associated with the later development of preeclampsia in low-risk nulliparous women. These biomarkers, together with known biomarkers, were then used to develop predictive models that met à priori criteria (detection of ≥50% of preeclampsia cases with a PPV of 20%, given a disease prevalence of 5%). During the development of predictive models, 4 of these novel biomarkers, IGFALS, MCAM, selenoprotein P, and SPINT1, were highly recurrent. In combination with known biomarkers (PlGF and sEng) and MAP, these markers achieved predictive performances with the potential to identify a subgroup of healthy nulliparous women who could receive specialist prenatal care. Overall preeclampsia detection rates ranged from 50% to 56% in the training set and 50% to 54% on external validation, with ≈3 quarters of preterm preeclampsia cases detected.
Fundamental to the identification of novel algorithms to predict preeclampsia was the application of the discovery–verification-validation proteomics pipeline. Our discovery proteomic approach capitalized on sensitivity gains achievable by using N-terminomics13 to identify novel biomarkers associated with later preeclampsia. The samples used for the biomarker discovery experiments were taken from a cohort of samples completely independent to the SCOPE cohort,11 at different gestational ages and with different risk factors for the development of preeclampsia. Although it is probable that a modified list of proteins would have been identified had the discovery experiment been performed in a subset of the SCOPE cohort, independent verification of several biomarkers across these 2 independent populations adds further credibility to the findings. Our quantitative MS assays enabled simultaneous determination of the concentration of 20+ lower abundant plasma proteins (many without established immunoassays) in 25 µL of plasma. Our study highlights the capability of LC-SRM assays to bridge the gap between discovery experiments (many candidates, large number of false positives) and clinical validation studies where fewer markers are studied in 100s of women. The importance of verification and validation of biomarkers in clinical proteomic studies cannot be overstated. Even with the use of highly sensitive unbiased MS techniques, there is a high attrition of biomarkers when measured in a larger independent sample set. Furthermore, in a heterogeneous clinical condition, the univariate performance of individual proteins is superseded by the performance of a combination of proteins within predictive models. Predictive models will always have the highest accuracy in the population in which they were created, even where steps have been taken to minimize overfitting. External validation in an independent sample set provides a much more robust estimate of the predictive performance if translated into clinical settings.
The algorithms devised in this study were selected to enrich, within a population of low-risk nulliparous women where risk factor screening is not adequate,2 a subgroup with a disease prevalence equivalent to current high-risk obstetric clinics,10 that is, a PPV of 20%. Such a screening test would allow stratification of prenatal care, with nulliparous women who screen positive receiving increased antenatal surveillance along with intervention to prevent preeclampsia.18 In comparison with reported multimarker combinations,4 and the NICE risk factor assessment tool, the models devised in this study perform favorably (Table 3) with better detection rates and lower numbers of false positives in healthy nulliparous women. Comparison with other algorithms is problematic as general populations, comprising high-risk and low-risk women, have been studied.3 There are several additional clinical variables, including body mass index,2 which could improve the performance of the biomarkers measured in this study. In future validation studies, which would require the measurement of far fewer biomarkers, it may be appropriate to include body mass index and other clinical variables within the model.
IGFALS, which was increased in the plasma of women who developed preeclampsia, is part of the ternary insulin-like growth factor complex. It is known to complex with IGFBP3 and IGFBP5 (insulin-like growth factor binding protein), proteins that control the bioavailability of IGFs, which are crucial for placental development and growth. The acid labile subunit prevents the transport of the insulin-like growth factor complex across the endothelium into tissues confining them to the circulation. High levels of placental IGFALS mRNA have previously been reported in small for gestational age babies19 and increased serum levels in women with established severe preeclampsia (n=8).20 The other novel biomarkers, MCAM and SPINT1, decreased in women destined to develop preeclampsia consistently improved performance within the multivariable models. MCAM is an endothelial adhesion molecule, important in maintenance of the endothelial monolayer. It is highly expressed by placental trophoblasts with decreased expression in placentas from women with preeclampsia.21 SPINT1 is a cell surface–binding protein of hepatocyte growth factor activator, which regulates hepatocyte growth factor activity. Hepatocyte growth factor contributes to the repair of injured tissues, is abundantly expressed by villous cytotrophoblasts,22 and thought to be essential for placental development.23
Although this study benefits from rigorous protocols related to sample and data collection across 6 centers in 4 countries, it has some limitations. The study size is modest and certain ethnic groups are under-represented. Samples taken at 20-week gestation were used because of their temporal proximity to samples used in the discovery experiment. Although this time point has the advantage of coinciding with a standard antenatal care milestone (fetal anomaly scan), this must be balanced against the greater potential benefit of prophylactic aspirin, if commenced before 20 weeks.18 The benefit of aspirin may well be retained in high-risk nulliparous women if commenced at 20 weeks, but this would need to be assessed in a prospective trial.
Multivariable panels will always be difficult to establish in the context of a low-prevalence disease; the study design used here attempts to limit the chance of over fitting and selection of false positives variables by limiting the number of variables available to the models and using 2 independent sample sets. Validation of only 8 of the 44 models in the test set indicates that over fitting is likely in the training data. The validated models are expected to be generalizable. Given the consistency of the proteins retained in all validated models, it is likely that the combination of proteins identified in this study is robust.
The proteins in this study have been quantified using mass spectrometric methods, which are not currently used in the clinical setting. Commercial ELISA kits are available for several of the markers (PlGF, ADAM12, sEng), and importantly, measurement of IGFALS using ELISA measurements produced equivalent predictive performance to the SRM quantification. Before any clinical application of the biomarker combinations identified in this study, further prospective studies will need to be undertaken with the biomarkers measured on a platform used in clinical laboratories (eg, ELISA) and the predictive algorithms evaluated in an adequately sized cohort of nulliparous women.
By definition, low-risk nulliparous women do not have a history of significant medical disease or previous hypertensive disease in pregnancy, and therefore conventional clinical risk factor models do not perform well in this group. In current high-risk clinic settings (eg, previous preeclampsia, chronic medical disease), the rate of preeclampsia is ≈20%, and therefore, we set out to develop a model that had a PPV of ≥20%. This level of risk would justify referral of women with a positive test for specialist care with a manageable number of false positives. Novel biomarkers relevant to the prediction of preeclampsia were confirmed in 2 independent sample sets in this study, and IGFALS has emerged as a novel marker, predictive of term and preterm preeclampsia. In the future, it is likely that biochemical markers will be combined with a modest number of easily recordable clinical risk markers to improve the prediction of preeclampsia in this low-risk population.
Ethical approval for the longitudinal cohort recruited in Dundee was given by Tayside Medical Research Ethics committee. Ethical approval for the SCOPE study was obtained from local ethics committees [New Zealand AKX/02/00/364; Australia REC 1712/5/2008; London, Leeds, and Manchester 06/MRE01/98; and Cork ECM5 (10) 05/02/08].
We thank Dr Mires, M. Macleod, and all the women who participated in the study at Ninewells Hospital, Dundee. We also thank all the women who participated in the SCOPE Study, the SCOPE PIs, Professor L. Poston (King’s College London), Professor L.M.E. McCowan (University of Auckland), Professor J.J. Walker (University of Leeds), and Professor G.A. Dekker (University of Adelaide); the SCOPE international coordinator, R. Taylor; the SCOPE Country Coordinators, D. Healy (University of Adelaide), A. Briley (Kings College London), and N. Murphy and E. Snapes (University College Cork); and the database and statistical support provided by E.H.Y. Chan, University of Auckland.
Sources of Funding
The SCOPE study was sponsored by the New Enterprise Research Fund, Foundation for Research Science and Technology, New Zealand; Health Research Council 04/198; Evelyn Bond Fund, Auckland District Health Board Charitable Trust; Australia: Premier’s Science and Research Fund, South Australian Government; London: Guy’s and St Thomas’ Charity, United Kingdom, Tommys the Baby Charity; Manchester: UK Biotechnology and Biological Sciences Research Council GT084, UK National Health Services NEAT Grant FSD025, University of Manchester Proof of Concept Funding, Tommy’s the Baby Charity, NIHR; Leeds: Cerebra, UK; and Cork, Ireland: Health Research Board, Ireland CSA/2007/2. The study at Ninewells hospital Dundee was funded by the Chief Scientist Office. J. Myers is also supported by Action Medical Research Endowment Fund and NIHR Manchester Biomedical Research Center. These study sponsors had no role in study design, data analysis, or writing this report. All mass spectrometry data acquisition was funded by Pronota Zwijnaarde, Belgium.
R. Tuytten, G. Thomas, W. Laroy, K. Kas, and G. Vanpoucke are employed by Pronota, which has a commercial interest in the development of predictive tests for preeclampsia. J. Myers, P.N. Baker, and R.A. North have received consultancy fees (paid to their institution) from Pronota. The other authors have no conflicts to report.
The online-only Data Supplement is available with this article at http://hyper.ahajournals.org/lookup/suppl/doi:10.1161/HYPERTENSIONAHA.113.01168/-/DC1.
- Received February 6, 2013.
- Revision received March 1, 2013.
- Accepted April 1, 2013.
- © 2013 American Heart Association, Inc.
- 1.↵World Health Organisation. The world health report 2005—make every mother and child count. 2005. http://www.who.int/whr/2005/whr2005_en.pdf. Accessed March 20, 2013.
- North RA,
- McCowan LM,
- Dekker GA,
- Poston L,
- Chan EH,
- Stewart AW,
- Black MA,
- Taylor RS,
- Walker JJ,
- Baker PN,
- Kenny LC
- Myatt L,
- Clifton RG,
- Roberts JM,
- et al
- Kenny LC,
- Broadhurst DI,
- Dunn W,
- Brown M,
- North RA,
- McCowan L,
- Roberts C,
- Cooper GJ,
- Kell DB,
- Baker PN
- Mebazaa A,
- Vanpoucke G,
- Thomas G,
- et al
- Caritis S,
- Sibai B,
- Hauth J,
- Lindheimer MD,
- Klebanoff M,
- Thom E,
- VanDorsten P,
- Landon M,
- Paul R,
- Miodovnik M,
- Meis P,
- Thurnau G
- Myers J,
- Mires G,
- Macleod M,
- Baker P
- Lange V,
- Picotti P,
- Domon B,
- Aebersold R
- Gentleman R,
- Hornik K,
- Leisch F
- 17.↵NICE. Hypertension in Pregnancy. 2010. http://guidance.nice.org.uk/CG107.
- Duley L,
- Henderson-Smart DJ,
- Meher S,
- King JF
- Tanaka H,
- Nagaike K,
- Takeda N,
- Itoh H,
- Kohama K,
- Fukushima T,
- Miyata S,
- Uchiyama S,
- Uchinokura S,
- Shimomura T,
- Miyazawa K,
- Kitamura N,
- Yamada G,
- Kataoka H
Novelty and Significance
What Is New?
This study identified insulin-like growth factor acid labile subunit, selenoprotein, serine peptidase inhibitor Kunitz type 1, and melanoma cell adhesion molecule as novel biomarkers for preeclampsia.
The combination of insulin-like growth factor acid labile subunit with blood pressure, along with other biomarkers, has the potential to be part of a clinically relevant predictive test for preeclampsia.
What Is Relevant?
Application of this test could improve our ability to identify a subgroup of women at significant risk of preeclampsia among nulliparous women. Women with a positive test would have a 1 in 5 chance of developing preeclampsia, which equates to the risk in current high-risk obstetric clinics.
The algorithm developed in this study could detect up to 50% of all preeclampsia and 80% of preterm preeclampsia cases arising in a low-risk nulliparous population. Detection of the majority preterm preeclampsia cases would allow intervention strategies, such as low-dose aspirin.
A negative test does not adequately risk stratify women at very low risk of preeclampsia to modify management.
This study has identified insulin-like growth factor acid labile subunit as a novel candidate biomarker for preeclampsia; predictive models containing this marker have been validated in 2 independent sample sets.