s12884 018 1908 9 PDF

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Murtoniemi et al.

BMC Pregnancy and Childbirth (2018) 18:279


https://doi.org/10.1186/s12884-018-1908-9

RESEARCH ARTICLE Open Access

Prediction of pre-eclampsia and its


subtypes in high-risk cohort:
hyperglycosylated human chorionic
gonadotropin in multivariate models
Katja Murtoniemi1* , Pia M. Villa2, Jaakko Matomäki3, Elina Keikkala4, Piia Vuorela5,2, Esa Hämäläinen6,
Eero Kajantie7, Anu-Katriina Pesonen8, Katri Räikkönen8, Pekka Taipale9, Ulf-Håkan Stenman6
and Hannele Laivuori10,11

Abstract
Background: The proportion of hyperglycosylated human chorionic gonadotropin (hCG-h) to total human chorionic
gonadotropin (%hCG-h) during the first trimester is a promising biomarker for prediction of early-onset pre-eclampsia.
We wanted to evaluate the performance of clinical risk factors, mean arterial pressure (MAP), %hCG-h, hCGβ,
pregnancy-associated plasma protein A (PAPP-A), placental growth factor (PlGF) and mean pulsatility index of
the uterine artery (Uta-PI) in the first trimester in predicting pre-eclampsia (PE) and its subtypes early-onset,
late-onset, severe and non-severe PE in a high-risk cohort.
Methods: We studied a subcohort of 257 high-risk women in the prospectively collected Prediction and Prevention of
Pre-eclampsia and Intrauterine Growth Restriction (PREDO) cohort. Multivariate logistic regression was used to
construct the prediction models. The first model included background variables and MAP. Additionally, biomarkers
were included in the second model and mean Uta-PI was included in the third model. All variables that improved the
model fit were included at each step. The area under the curve (AUC) was determined for all models.
Results: We found that lower levels of serum PlGF concentration were associated with early-onset PE, whereas lower
%hCG-h was associated with the late-onset PE. Serum PlGF was lower and hCGβ higher in severe PE, while %hCG-h
and serum PAPP-A were lower in non-severe PE. By using multivariate regression analyses the best prediction for all PE
was achieved with the third model: AUC was 0.66, and sensitivity 36% at 90% specificity. Third model also
gave the highest prediction accuracy for late-onset, severe and non-severe PE: AUC 0.66 with 32% sensitivity,
AUC 0.65, 24% sensitivity and AUC 0.60, 22% sensitivity at 90% specificity, respectively. The best prediction for
early-onset PE was achieved using the second model: AUC 0.68 and 20% sensitivity at 90% specificity.
Conclusions: Although the multivariate models did not meet the requirements to be clinically useful screening tools,
our results indicate that the biomarker profile in women with risk factors for PE is different according to the subtype of
PE. The heterogeneous nature of PE results in difficulty to find new, clinically useful biomarkers for prediction of PE in
early pregnancy in high-risk cohorts.
(Continued on next page)

* Correspondence: [email protected]
1
University of Helsinki and Turunmaa District Hospital, Gynaecological
Outpatient Clinic, Hospital District of Southwest Finland, Kaskenkatu 13,
20700 Turku, Finland
Full list of author information is available at the end of the article

© The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver
(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Murtoniemi et al. BMC Pregnancy and Childbirth (2018) 18:279 Page 2 of 10

(Continued from previous page)


Trial registration: International Standard Randomised Controlled Trial number ISRCTN14030412, Date of registration
6/09/2007, retrospectively registered.
Keywords: Pre-eclampsia, Screening, Biomarkers, Early-onset pre-eclampsia, Late-onset pre-eclampsia, Severe pre-
eclampsia, Placental growth factor, hCGβ, Hyperglycosylated human chorionic gonadotropin, Pregnancy associated
plasma protein a

Background pre-eclampsia, 66 were excluded from the analyses be-


Pre-eclampsia is a pregnancy-specific multisystem dis- cause they took prophylactic aspirin during pregnancy as
order, the pathogenesis of which is incompletely under- a part of a randomised trial [9]. Originally the subcohort
stood. It occurs in 2–8% of pregnancies and is one of of the present study comprised of 267 women. Ten
the leading causes of maternal and fetal morbidity and women were not included to the analyses for various rea-
mortality globally [1]. The early identification of women sons: one woman had miscarriage at 19 weeks of gesta-
at high risk for pre-eclampsia would guide in planning tion, one woman discontinued the study due to a
of their follow-up during pregnancy and in the applica- non-medical reason and eight women were excluded due
tion of preventive measures. There is evidence that to missing data.
low-dose aspirin, started at 12–16 weeks of gestation, re-
duces the risk of pre-eclampsia [2, 3].
Clinical risk factors, e.g. antiphospholipid antibody syn- Biophysical measurements
drome, prior pre-eclampsia, chronic hypertension and Gestational age was confirmed by first trimester ultra-
pregestational diabetes have been used in early pregnancy sound measurement of the fetal crown-rump length.
in identifying those women at high risk of developing Doppler ultrasound measurements were performed
pre-eclampsia [4, 5]. A predictive test with high sensitivity transvaginally at 11 0/7–13 6/7 weeks of gestation. The
and positive predictive value that incorporates maternal uterine artery flow was measured from the level of the
risk factors, biomarkers and biophysical measurements is inner os of the cervix, as it approaches the uterus
needed to implement prophylaxis strategies [6]. Recently, laterally. The mean PI was calculated and then the mul-
we have shown, that serum hyperglycosylated human tiple of the median (MoM) mean PI was calculated with
chorionic gonadotropin (hCG-h) is a promising marker of the equation: Loge mean Uta-PI = 1.39–0.012 × GA +
early-onset pre-eclampsia [7], but this was a case-control GA2 × 0.0000198 MoM [10], where GA = gestational age
study. Thus, our aim in the present study was to test in days. The MAP was calculated from the first trimester
hCG-h or proportion of hCG-h to hCG (%hCG-h) in a co- visit blood pressure measurement with the equation:
hort of high-risk women. We wanted to investigate if MAP = diastolic blood pressure + (systolic blood pres-
hCG-h could predict pre-eclampsia when combined with sure – diastolic blood pressure)/3. The bilateral mea-
other biomarkers including serum free hCG beta (hCGβ), surements of the Uta-PI were available for 83.7% (215/
pregnancy-associated plasma protein-A (PAPP-A) and 257) of high-risk women.
placental growth factor (PlGF), as well as biophysical mea-
surements, mean arterial pressure (MAP) and Doppler
ultrasound measurement of the mean pulsatility index of Biomarkers
the uterine artery (Uta-PI). For this purpose, we constructed Fasting blood samples were drawn from antecubital
multivariate regression models and tested their predictive veins at 11 0/7–13 6/7 (mean 13 0/7) weeks of gestation.
value to detect the different subtypes of pre-eclampsia. Serum was separated within an hour by centrifugation
and stored in − 80 °C until analysis. Serum hCG-h con-
centrations were determined using a time-resolved
Methods immunofluorometric assay. To eliminate the interaction
Study cohort between complement and B152 antibody in the hCG-h
The present study is a part of the multidisciplinary ‘Predic- assay, serum samples were diluted 100-fold prior to ana-
tion and Prevention of Pre-eclampsia and Intrauterine lysis with EDTA-containing buffer (5 mmol/L) [11].
Growth Restriction’ (PREDO) project [8]. A subcohort with Serum hCG, hCGβ, PAPP-A and PlGF levels were
known risk factor status for pre-eclampsia was prospect- measured using time-resolved immunofluorometric
ively collected between September 2005 and June 2009 in assay according to the manufacturer’s instructions
ten participating maternity clinics in Finland. The inclusion (AutoDelfia, Perkin Elmer, Wallac, Turku, Finland). The
and exclusion criteria are described in Additional file 1: intra- and inter-assay coefficients of variation were 4,9%
Table S1. Of the women with clinical risk factors for (mean) and 6,7% (mean) for PlGF, < 2.3 and < 3.7% for
Murtoniemi et al. BMC Pregnancy and Childbirth (2018) 18:279 Page 3 of 10

PAPP-A, < 3.3 and < 6.2% for hCGβ, 1.8 and < 8.8% for were included in model 1, since these variables are ob-
hCG and 2.2 and < 10.5% for hCG-h, respectively. tainable and economical. Biomarkers, which were the
main interest of the study, were added in model 2 and
Outcome measures MoM of the mean Uta-PI was added in model 3 to clar-
Primary outcome was pre-eclampsia, defined as a systolic ify if it still improves prediction rates. All variables that
blood pressure ≥ 140 mmHg and/or a diastolic blood pres- improved the model fit in the multivariate logistic re-
sure ≥ 90 mmHg occurring after 20 weeks of gestation gression analyses were included at each step, because we
combined with a urinary 24-h protein excretion of ≥ 0.3 g wanted to investigate the maximum prediction potential
or the dipstick equivalent in two consecutive measure- of each model. We used 10-time cross validation to
ments [12]. Secondary outcomes were: early-onset compensate the lack of replication cohort.
pre-eclampsia (diagnosed before 34 0/7 weeks of gesta- The background variables used for model fitting were
tion), late-onset pre-eclampsia (diagnosed at or after 34 0/ maternal age, primiparity, pre-pregnancy body mass
7 weeks of gestation), severe pre-eclampsia (systolic blood index (BMI), infertility treatment before the present
pressure ≥ 160 mmHg and/or diastolic blood pressure ≥ pregnancy, pre-eclampsia in a previous pregnancy, a
110 mmHg and/or proteinuria ≥5 g/24 h), non-severe SGA infant or gestational hypertension, type 1 diabetes
pre-eclampsia (pre-eclampsia not fulfilling the criteria of mellitus and fetus mortus in a previous pregnancy. The
severe pre-eclampsia) and small for gestational age (SGA) area under the receiver operating characteristic (ROC)
(birthweight < − 2 standard deviations (SD) i.e. approxi- curve (AUC) value was determined for each model and
mately < 2,5% percentile) [13]. All diagnoses were inde- models were compared using the R package pROC [15].
pendently confirmed by a jury of two physicians and one Results of multivariate logistic regression analyses are
research nurse, as described previously [9]. presented without confidence intervals, as confidence
intervals obtained from regularised methods are prob-
Statistical analyses lematic [16]. In addition to AUROC comparison, models
Binary logistic regression was used to compare the char- were compared by calculating the p-values with DeLongs
acteristics of the study groups. The mean of continuous method. [17]. As our aim was to assess the screening
variables was calculated to compare the differences be- performance of the models in clinical practice, the refer-
tween the groups, if the variable was normally distrib- ence group was all women who did not develop the out-
uted. If the number of subjects in a subgroup was low or come of interest, as in the study by Kenny et al. [18]. For
the continuous variable was not normally distributed, example, for severe pre-eclampsia the reference group
the median and interquartile range was reported. consisted of women who did not develop severe
The concentrations of hCG-h, hCG, hCGβ, PAPP-A, pre-eclampsia: women without pre-eclampsia and
and PlGF, as well as %hCG-h were normally distributed women who developed non-severe pre-eclampsia.
after log-transformation and were adjusted for gesta- Calculating study power is problematic regarding regu-
tional age using linear regression analysis. The median larized logistic regression analyses used in present study,
regression equation was calculated from the respective because power calculations are based on statistical signifi-
concentrations measured from 107 women without clin- cance and it is not possible to calculate significance with
ical risk factors in the PREDO cohort and from a screen- the statistical method used in this study. To give an indi-
ing cohort from the Kuopio University Hospital [7]. cation of the power, power analysis was done for univari-
Then, concentrations measured from women with ate logistic regression model. The power decreases when
clinical risk factors in the PREDO cohort were com- OR approaches 1, e.g. for pre-eclampsia (prevalence
pared to the gestational age-adjusted median and 13.2%, N 257) power is 96% with OR 0.5, 79% with OR
expressed as MoM. 0.6, 49% with OR 0.7 and 22% with OR 0.8 [19].
The binary logistic regression was used to evaluate
univariate associations between measured variables and Results
outcomes. The results were presented with odds ratios Baseline and pregnancy characteristics
(OR). Since the number of investigated predictor vari- The total cohort comprised of 257 women with risk fac-
ables was large compared to the number of women who tors for pre-eclampsia. Pre-eclampsia occurred in 34
developed pre-eclampsia, prediction models were built (13.2%) of the women in the study cohort. Of those who
using regularised logistic regression with the L1/ developed pre-eclampsia, 9 (26.5%) had early-onset
L2-norm using the R package glmnet [14]. Cross valid- pre-eclampsia and 17 (50%) had a severe form of the
ation was used to select regularisation variables and sep- disease. Clinical characteristics of the high-risk women
arately to assess model fit. Three separate models were who did or did not develop pre-eclampsia are presented
fitted for all outcomes on clinical bases. Maternal clinical in Table 1 and characteristics by pre-eclampsia subtype
background variables and MAP in the first trimester in Table 2.
Murtoniemi et al. BMC Pregnancy and Childbirth (2018) 18:279 Page 4 of 10

Table 1 Characteristics of women with clinical risk factors for pre-eclampsia (PE), according to their PE status
Women affected by PE (N = 34) Women not affected by PE (N = 223) p-value ORa 95% CI Lower Upper
b
Age, years (SD) 31.5 (5.8) 32.2 (5.7) 0.58 0.98 0.92 1.05
BMI, pre-pregnancy (kg/m2) (SD)b 29.7 (7.3) 28.1 (6.7) 0.19 1.04 0.98 1.09
Primiparous, n (%) 9 (26.5) 52 (23.2) 0.55 0.76 0.31 1.85
Infertility treatment, n (%) 4 (13.3) 24 (11.5) 0.88 0.91 0.28 2.99
Chronic disease, n (%) 18 (52.9) 93 (42.3) 0.67 0.84 0.38 1.88
Education, n (%)
Elementary or less 3 (9.7) 4 (2.1) 0.04 4.98 1.06 23.44
High school or vocational school 8 (25.8) 57 (30.0) 0.64 0.81 0.34 1.92
Intermediate 14 (45.2) 81 (42.6) 0.79 1.12 0.52 2.38
University 6 (19.4) 48 (25.3) 0.49 0.71 0.28 1.83
a
Binary logistic regression
b
Mean
OR odds ratio, BMI body mass index, CI confidence interval, SD standard deviation,

Twelve (4.7%) women gave birth to a SGA newborn. Of pre-eclampsia compared to other women. In women with
these women, eight (67%) had developed pre-eclampsia. non-severe pre-eclampsia, both the median PAPP-A and
The prevalence of each inclusion criterion in women who %hCG-h levels were lower than in other women.
developed pre-eclampsia and in women who did not
develop pre-eclampsia is presented in Additional file 2:
Table S2. There was one stillbirth at the 27th week of Multivariate logistic regression models
gestation. Pregnancy characteristics are presented in The predictive models were constructed for pre-eclampsia
Table 3 and the median or mean values of measured and its subtypes as outcomes. All variables that improved
variables in Additional file 3: Table S3. the model fit in the multivariate logistic regression analyses
were included, therefore each regression model is individ-
ual: for example model 1 for early-onset pre-eclampsia
Univariate analyses consists of different variables than model 1 for late-onset
All results of univariate analyses are summarised in pre-eclampsia. The multivariate models, AUC values and
Table 4. None of the biomarkers were different between diagnostic characteristics of pre-eclampsia and its subtypes
pre-eclamptic and non-pre-eclamptic women. The me- are presented in Table 5.
dian Uta-PI and MAP were higher in women who devel- The effect of each risk factor on the risk of developing
oped pre-eclampsia than in women who did not develop pre-eclampsia was estimated from the model 1. The
pre-eclampsia. Women who developed early-onset most significant factor that increased the risk of develop-
pre-eclampsia had lower median PlGF and higher MAP ing all pre-eclampsia (OR 1.69) and the late-onset (OR
compared to women who did not develop early-onset 2.40) or the non-severe (OR 2.32) subtypes was a history
pre-eclampsia. Women who developed late-onset of pre-eclampsia. Primiparity (OR 3.34) was the most
pre-eclampsia had lower %hCG-h and higher MAPs than significant factor for the early-onset subtype.
other women. Median hCGβ, Uta-PI and MAP were For all pre-eclampsia, the best validated AUC value of
higher, and median PlGF lower, in women with severe 0.66 at sensitivities of 36 and 16% were achieved with 90

Table 2 Characteristics of women with clinical risk factors for pre-eclampsia (PE) who developed PE, by subtype
Maternal Characteristics Early-onset PE n = 9 Late-onset PE n = 25 p-value* Severe PE n = 17 Non-severe PE n = 17 p-value*
Age, mean years (SD) 32.4 (4.9) 31.2 (6.1) 0.22 32.6 (5.5) 30.4 (6.0) 0.05
BMI, pre-pregnancy (kg/m2) (SD) 29.5 (8.1) 29.8 (7.1) 0.96 28.0 (7.2) 31.5 (7.1) 0.27
Primiparous, n (%) 2 (22.2) 2 (9.5) 1.00 7 (41.2) 2 (11.8) 0.02
Infertility treatment, n (%) 3 (33.3) 15 (60.0) 1.00 3 (17.6) 1 (7.7) 0.43
Chronic disease, n (%) 5 (56.6) 4 (16.0) 1.00 8 (47.1) 10 (58.8) 0.28
*Binary logistic regression
BMI body mass index, SD standard deviation
Early-onset PE = diagnosed before 34 0/7 weeks of gestation, late-onset PE = diagnosed at or after 34 0/7 weeks of gestation, severe PE = systolic blood pressure ≥
160 mmHg and/or diastolic blood pressure ≥ 110 mmHg and/or proteinuria ≥5 g/24 h, Non-severe PE = PE not fulfilling the criteria of severe pre-eclampsia
Murtoniemi et al. BMC Pregnancy and Childbirth (2018) 18:279 Page 5 of 10

Table 3 Pregnancy characteristics


Pregnancy characteristics of high-risk women Women not affected PE Late-onset PE Early-onset PE Non-Severe Severe PE
by PE (N = 223) (N = 34) (n= 25) (n=9) PE (n=17) (n=17)
Antihypertensive medication n (%) 17 (7.6) 22 (64.7) 16 (64.0) 6 (66.7) 8 (47.1) 14 (82.4)
Before 20 weeks of gestation 5 (2.2) 2 (5.9) 2 (8.0) 0 1 (5.9) 1 (5.9)
After 20 weeks of gestation 12 (5.5) 20 (58.8) 14 (56.0) 6 (66.7) 7 (41.2) 13 (76.5)
Weight gain during pregnancy, kg 12.8 (6.7) 12.4 (6.6) 11.0 (4.0) 11.1 (10.0) 12.8 (6.6) 12.0 (8.6)
Gestational diabetes, n (%) 56 (25.2) 9 (26.5) 7 (28.0) 2 (22.2) 4 (23.5) 5 (29.4)
Oral glucose tolerance test not performed 17 (7.6) 2 (5.9) 1 (4.0) 1 (11.1) 1 (5.9) 1 (5.9)
Highest systolic blood pressure, mmHg 136 (26) 168 (28) 165 (24) 183 (27) 161 (20) 182 (27)
Highest diastolic blood pressure, mmHg 90 (17) 110 (8) 110 (15) 105 (14) 10 (13) 111 (10)
Highest proteinuria, g/day 0.2 (0.0) 1.2 (1.8) 0.8 (1.3) 2.4 (5.2) 0.7 (0.9) 1.9 (3.1)
Gestational weeks at birth 39.9 (6.8) 38.4 (6.6) 38.7 (1.6) 31.9 (6.2) 38.6 (2.0) 36.6 (7.2)
Mode of delivery, n (%)
Vaginal 151 (68.6) 18 (52.9) 16 (64.0) 2 (22.2) 11 (64.7) 8 (41.2)
Vacuum extraction 20 (9.1) 1 (2.9) 1 (4.0) 0 0 0
Elective caesarean section 17 (7.7) 1 (2.9) 1 (4.0) 0 1 (8.3) 0
Caesarean section during labour 31 (14.2) 14 (41.2) 7 (28.0) 7 (77.8) 5 (29.4) 9 (52.9)
Apgar score at 5 min 9 (0) 9 (2) 9 (1) 8 (1) 9 (1) 8.5 (2)
Umbilical artery pH 7.26 (0.10) 7.25 (0.10) 7.25 (0.14) 7.27 (0.20) 7.24 (0.13) 7.25 (0.16)
Newborn birthweight, g 3615 (685) 3109 (1260) 3370 (663) 1350 (1240) 3570 (629) 2340 (1890)
Placental weight, g 610 (180) 545 (185) 580 (335) 350 (40) 570 (105) 440 (215)
With continuous variables median and interquartile range in parenthesis is presented
n = number of cases, PE = pre-eclampsia
Early-onset PE = pre-eclampsia diagnosis < 34 weeks of gestation; Late-onset PE = pre-eclampsia diagnosis ≥ 34 weeks of gestation; Severe pre-eclampsia = systolic
blood pressure ≥ 160 mmHg and/or diastolic blood pressure ≥ 110 mmHg and/or proteinuria ≥5 g/24 h, Non-severe pre-eclampsia = PE not fulfilling the criteria of
severe pre-eclampsia

and 95% specificity, respectively, when combining mater- Discussion


nal characteristics, MAP, biomarkers and Uta-PI MoM. To our knowledge, this is a first study investigating first tri-
With this model, positive predictive value (PPV) was mester hCG-h as a potential pre-eclampsia predictor in a
33% and negative predictive value was 88%. prospectively recruited high-risk cohort. In univariate ana-
The best multivariate model for prediction of lyses, %hCG-h levels were lower in women with late-onset
early-onset pre-eclampsia was achieved by combining and non-severe pre-eclampsia. Lower levels of serum PlGF
maternal characteristics, MAP and biomarkers. The vali- were associated with early-onset and severe pre-eclampsia,
dated AUC value was 0.68 with 20% sensitivity at both and lower serum PAPP-A levels with non-severe
90 and 95% specificity. The PPV for early-onset pre-eclampsia. Higher serum hCGβ levels were associated
pre-eclampsia was 25% and NPV was 88%. with severe pre-eclampsia. The first trimester MAP was
For prediction of late-onset pre-eclampsia, model 3 gave higher in pre-eclampsia and its subtypes, except in the
the highest prediction rates with an AUC value of 0.66, non-severe subtype, when compared to all other partici-
with 32 and 16% sensitivity at 90 and 95% specificity, re- pants. Uta-PI MoM was higher in women who developed
spectively. For prediction of severe pre-eclampsia, the best pre-eclampsia compared to women who did not, and in
validated AUC value was 0.65 with 24% sensitivity at 90% women who developed severe pre-eclampsia compared to
specificity. It was achieved by combining MAP, the a priori all other participants. Despite the abovementioned differ-
risk factor of having a previous fetus mortus, biomarkers ences in serum %hCG-h, PlGF, PAPP-A, hCGβ, MAP and
and Uta-PI MoM (model 3). A sensitivity of 23% at 95% Uta-PI between the groups, multivariate models gave only
sensitivity was achieved by combining MAP and bio- a modest prediction of the disease and did not meet the
markers (model 2). The best multivariate model for pre- requirement of a clinically useful screening test.
dicting non-severe pre-eclampsia, with a validated AUC Pre-eclampsia in a previous pregnancy was the most
value of 0.60, 22 and 15% sensitivity at 90 and 95% specifi- important risk factor associated with pre-eclampsia in a
city, respectively, was attained with model 3. subsequent pregnancy. This is in line with a recent
Murtoniemi et al. BMC Pregnancy and Childbirth (2018) 18:279 Page 6 of 10

Table 4 Results of the univariate analyses


Variable Participants Median (IQR) / Mean (SD) OR (95% CI) p-value
PE, n No PE, n PE No PE
hCG-h MoM 34 223 0.27 (0.11–0.61) 0.30 (0.15–055) 0.98 (0.49–1.97) 0.958
%hCG-h MoM 34 223 0.77 (0.64–1.24) 0.99 (0.71–1.45) 0.55 (0.26–1.14) 0.107
hCGβ MoM 34 223 0.60 (0.35–0.89) 0.56 (0.35–0.91) 1.21 (0.70–2.10) 0.503
PAPP-A MoM 34 222 0.69 (0.45–0.92) 0.84 (0.52–1.26) 0.50 (0.24–1.05) 0.069
PlGF MoM 34 222 0.82 (0.65–1.06) 0.99 (0.81–1.22) 0.32 (0.10–1.00) 0.051
Uta-PI MoM 29 186 1.15 (0.83–1.25) 0.92 (0.71–1.12) 3.81 (1.12–12.95) 0.032
MAPa 34 211 102.1 (10.9) 94.7 (14.4) 1.05 (1.02–1.09) 0.001
Variable EOPE, n No EOPE, n EOPE No EOPE OR (95% CI) p-value
hCG-h MoM 9 248 0.61 (0.23–0.75) 0.29 (0.14–0.53) 0.65 (1.61–3.96) 0.304
%hCG-h MoM 9 248 1.32 (1.15–1.50) 0.92 (0.68–1.39) 1.58 (0.57–4.39) 0.381
hCGβ MoM 9 248 0.84 (0.61–1.04) 0.56 (0.34–0.91) 0.61 (1.41–3.26) 0.417
PAPP-A MoM 9 247 0.56 (0.50–0.85) 0.83 (0.51–1.21) 0.11 (0.46–2.01) 0.305
PlGF MoM 9 247 0.63 (0.47–0.73) 0.98 (0.80–1.22) < 0.001 (< 0.001–0.04) < 0.001
Uta-PI MoM 6 209 1.17 (0.97–1.25) 0.93 (0.73–1.14) 0.59 (5.56–52.2) 0.134
MAPa 9 236 104.3 (15.0) 95.4 (11.5) 1.06 (1.01–1.12) 0.028
Variable LOPE, n No LOPE, n LOPE No LOPE OR (95% CI) p-value
hCG-h MoM 25 232 0.20 (0.10–0.33) 0.30 (0.15–0.57) 0.26 (0.70–1.87) 0.478
%hCG-h MoM 25 232 0.76 (0.64–0.89) 1.02 (0.71–1.46) 0.30 (0.11–0.81) 0.018
hCGβ MoM 25 232 0.54 (0.31–0.79) 0.57 (0.35–0.91) 0.56 (1.09–2.13) 0.796
PAPP-A MoM 25 231 0.72 (0.43–0.95) 0.84 (0.51–1.22) 0.23 (0.54–1.24) 0.144
PlGF MoM 25 231 0.86 (0.80–1.18) 0.98 (0.80–1.21) 0.91 (0.34–2.41) 0.846
Uta-PI MoM 23 192 1.06 (0.78–1.33) 0.94 (0.72–1.13) 2.96 (0.78–11.23) 0.110
MAPa 25 220 101.3 (14.5) 95.0 (11.2) 1.04 (1.01–1.08) 0.013
Variable Severe PE, n No Severe PE, n Severe PE No Severe PE OR (95% CI) p-value
hCG-h MoM 17 240 0.47 (0.12–0.75) 0.30 (0.15–0.53) 1.63 (0.82–3.27) 0.165
%hCG-h MoM 17 240 0.90 (0.64–1.50) 0.95 (0.69–1.39) 0.94 (0.39–2.25) 0.885
hCGβ MoM 17 240 0.85 (0.35–1.13) 0.56 (0.35–0.90) 1.78 (1.01–3.16) 0.048
PAPP-A MoM 17 239 0.77 (0.45–1.08) 0.83 (0.51–1.21) 0.81 (0.37–1.80) 0.612
PlGF MoM 17 239 0.80 (0.61–0.83) 0.98 (0.80–1.22) 0.11 (0.02–0.69) 0.018
Uta-PI MoM 17 239 1.21 (0.96–1.43) 0.93 (0.73–1.13) 6.97 (1.4–34.2) 0.017
MAPa 17 228 105.8 (16.0) 94.9 (11.0) 1.08 (1.03–1.12) 0.001
Variable Non-severe PE, n No Non-severe PE, n Non-severe PE No Non-severe PE OR (95% CI) p-value
hCG-h MoM 17 240 0.15 (0.30–0.57) 0.22 (0.10–0.33) 0.22 (0.03–1.55) 0.127
%hCG-h MoM 17 240 0.78 (0.68–0.89) 0.98 (0.69–1.46) 0.28 (0.09–0.95) 0.041
hCGβ MoM 17 240 0.56 (0.37–0.75) 0.57 (0.35–0.92) 0.41 (0.10–1.63) 0.207
PAPP-A MoM 17 239 0.55 (0.49–0.90) 0.84 (0.51–1.26) 0.25 (0.07–0.94) 0.041
PlGF MoM 17 239 0.86 (0.73–1.08) 0.98 (0.80–1.21) 0.76 (0.22–2.65) 0.663
Uta-PI MoM 15 200 1.02 (0.82–1.21) 0.94 (0.73–1.14) 1.57 (0.30–8.31) 0.596
MAPa 17 228 98.3 (12.0) 95.5 (11.7) 1.02 (0.98–1.06) 0.336
a
mean and SD
Binary logistic regression was used to evaluate the association of measured variables to pre-eclampsia and its subtypes
IQR interquartile range; SD standard deviation; CI confidence interval; PE pre-eclampsia; EOPE early-onset pre-eclampsia (diagnosed before 34 0/7 weeks of
gestation); LOPE late-onset pre-eclampsia (diagnosed at or after 34 0/7 weeks of gestation); BMI body mass index; SGA small for gestational age; DM diabetes
mellitus; MAP mean arterial pressure; Uta-PI pulsatility index of the uterine artery; hCG human chorionic gonadotropin; hCG-h hyperglycosylated hCG; %hCG-h the
ratio of hCG-h to hCG; PAPP-A pregnancy-associated plasma protein a; PlGF placental growth factor; MoM multiple of the median
Early-onset PE = pre-eclampsia diagnosis < 34 weeks of gestation; Late-onset PE = pre-eclampsia diagnosis ≥ 34 weeks of gestation; Severe pre-eclampsia = systolic
blood pressure ≥ 160 mmHg and/or diastolic blood pressure ≥ 110 mmHg and/or proteinuria ≥5 g/24 h, Non-severe pre-eclampsia = PE not fulfilling the criteria of
severe pre-eclampsia
Murtoniemi et al. BMC Pregnancy and Childbirth (2018) 18:279 Page 7 of 10

Table 5 Multivariate regression models and screening test characteristics for pre-eclampsia and its subtypes
Outcome Variables in model Prevalence, AUC Valid.AUC At 90% Specificity for At 95% Specificity for
and % Validated AUC Validated AUC
model
PE 13.2 Sensitivity, PPV, NPV, PLR Sensitivity, PPV, NPV, PLR
% % % % % %
Model 1 Age, prior PE, prior SGA, DM type-I, MAP 0.70 0.55 23 27 88 2.4 13 29 88 2.6
Model 2 Model 1 variables + hCG MoM, %hCG-h 0.79 0.60 20 24 88 2.0 17 33 88 3.2
MoM, free beta hCG MoM, PAPP-A MoM,
PlGF MoM
Model 3 Model 1 variables + CH, hCG MoM, 0.85 0.66 36 36 90 3.7 16 33 88 3.3
%hCG-h MoM, free beta hCG MoM, PlGF
MoM, Uta-PI MoM
EOPE 3.5
Model 1 Primiparity, CH, prior SGA, DM type-I, MAP 0.84 0.52 7 10 86 0.7 6.7 17 87 1.3
Model 2 Primiparity, CH, DM type-I, MAP, hCG 0.96 0.68 20 24 88 2.0 20 38 88 3.9
MoM, %hCG-h MoM, free, PlGF MoM
Model 3 Primiparity, prior PE, prior SGA, CH, MAP, 0.95 0.62 11 14 87 1.1 11 25 88 2.2
hCG MoM, PlGF MoM
LOPE 9.7
Model 1 Age, prior PE, prior SGA, CH, DM type-I, 0.75 0.54 10 14 87 1.0 6.7 17 87 1.3
MAP
Model 2 Age, prior PE, prior SGA, CH, MAP, %hCG- 0.84 0.62 27 30 89 2.7 6.7 17 87 1.3
h MoM, free beta hCG MoM
Model 3 Model 1 variables, prior FM, hCG MoM, 0.89 0.66 32 33 90 3.3 16 33 88 3.3
%hCG-h MoM, free beta hCG MoM, PlGF
MoM, Uta-PI MoM
Severe PE 6.6
Model 1 MAP 0.68 0.58 20 24 88 2.0 17 33 88 3.2
Model 2 MAP, hCG MoM, free beta hCG MoM, PlGF 0.78 0.62 23 27 88 2.4 23 41 89 4.5
MoM
Model 3 MAP, prior FM, hCG MoM, %hCG-h MoM, 0.86 0.65 24 27 89 2.5 20 38 89 4.1
PlGF MoM, Uta-PI MoM
Non- 6.6
Severe PE
Model 1 Prior PE 0.71 0.54 3.3 5.6 86 0.4 3.3 7.1 86 0.5
Model 2 Age, BMI, prior PE, prior SGA, CH, %hCG-h 0.86 0.58 17 21 88 1.7 10 23 87 1.9
MoM, PAPP-A MoM
Model 3 Age, BMI, prior PE, prior SGA, CH, DM 0.89 0.60 22 25 88 2.2 15 31 88 2.9
type-I, prior FM, %hCG-h MoM, free beta
hCG MoM, PlGF MoM, Uta-PI MoM
Valid., validated; PPV, positive predictive value; NPV, negative predictive value; PLR, positive likelihood ratio; PE, pre-eclampsia; EOPE, early-onset PE; LOPE,
late-onset PE;
BMI, body mass index; SGA, small for gestational age; FM, fetus mortus; DM, diabetes mellitus; CH, chronic hypertension; MAP, mean arterial pressure; Uta-PI,
pulsatility index of the uterine artery; hCG, human chorionic gonadotropin; hCG-h, hyperglycosylated hCG; %hCG-h, the ratio of hCG-h to hCG; PAPP-A, pregnancy-
associated plasma protein A;
PlGF, placental growth factor; MoM, multiple of the median.
Early-onset PE = pre-eclampsia diagnosis < 34 weeks of gestation; Late-onset PE = pre-eclampsia diagnosis ≥ 34 weeks of gestation; Severe pre-eclampsia = systolic
blood pressure ≥ 160 mmHg and/or diastolic blood pressure ≥ 110 mmHg and/or proteinuria ≥5 g/24 h, Non-severe pre-eclampsia = PE not fulfilling the criteria of
severe pre-eclampsia.
Prediction models were built using regularised logistic regression. Cross validation was used to select regularisation variables and separately to assess model fit.
The AUC values are expressed before and after 10-fold cross validation. Three separate models were fitted for all outcomes. First, background variables and MAP
in the first trimester were included in the model (model 1). Next, biomarkers were added (model 2). Finally, MoM of the mean Uta-PI was added (model 3). All
variables that improved the model fit in the multivariate logistic regression analyses were included.

meta-analysis, wherein women with prior pre-eclampsia and non-severe pre-eclampsia, whereas primiparity had the
had the greatest pooled relative risk for developing highest OR for early-onset pre-eclampsia in multivariate
pre-eclampsia [5]. Interestingly, in our study, prior analyses. There are a few studies that distinguish between
pre-eclampsia had the strongest association with late-onset pre-eclampsia subtypes [20, 21]. In contrast to our study,
Murtoniemi et al. BMC Pregnancy and Childbirth (2018) 18:279 Page 8 of 10

other studies have been conducted in unselected popula- the definition in most of other studies is pre-eclampsia
tions. In accordance with our results, in the study of Ode- requiring delivery before 34 weeks of gestation.
gård et al., prior pre-eclampsia and primiparity had the Strength of our study is a carefully characterised, pro-
highest ORs for predicting pre-eclampsia, but primiparity spectively collected cohort of women with clinical risk fac-
did not appear to be specifically associated with either of tors for pre-eclampsia. A jury of two physicians and one
the clinical subtypes [21]. research nurse independently confirmed all pre-eclampsia
The impaired invasion of cytotrophoblasts into the spiral diagnoses. Furthermore, our prospective study reflects the
arteries is thought to be one pathophysiological mechanism true incidence of early-onset and late-onset pre-eclampsia
of pre-eclampsia [22]. The exact role of hCG-h in the in high-risk women.
pathophysiology of pre-eclampsia is not known. However, A limitation of our study is the relatively small sample
there is strong evidence that extravillous cytotrophoblasts size, but the high incidence of PE cases (13.2%) allowed
initiate the production of hCG-h during their differentiation us to obtain some interesting results. Another limitation
from proliferative cytotrophoblasts to invasive cytotropho- is that only six Doppler measurements were available in
blasts in normal pregnancies and that hCG-h circulating in the early-onset group.
maternal serum reflects the invasion process of tropho- The lack of a replication cohort is a limitation. Therefore
blasts during the first trimester [23]. Thus, hCG-h or its ra- we used 10-fold cross validation to obtain more realistic
tio to total hCG might represent a biomarker of early performance estimates for the models. Initially our models
placentation [24]. Lower %hCG-h in late-onset and reached quite promising AUC values for predicting
non-severe forms of pre-eclampsia may indicate a mild de- pre-eclampsia but after validation the values decreased. A
velopmental insufficiency of the placenta [25]. The results recent study using a combination of maternal risk factors,
of the present study stand in contrast to a previous report, PAPP-A, PlGF, MAP and Uta-PI for predicting
in which %hCG-h was lower in early-onset pre-eclampsia pre-eclampsia in the first trimester in a screening popula-
[7]. One possible explanation is that the median gestational tion also reported lower detection rates after five-fold cross
age at the time of sampling was higher in this study (13.0 validation [30] than in some earlier studies where there
vs. 10.3 gestational weeks). In very early pregnancy (at 4 to were not any kind of validation of the prediction rates [31,
5 weeks of pregnancy) virtually all (90–100%) of the hCG in 32]. The reason for the modest prediction rates of our
serum consists of hCG-h. The concentration decreases to study compared to the studies conducted in low-risk or
5–10% at 10 weeks and to 3% after 20 weeks [11, 24]. In screening populations may be, that the role of impaired
the present study, the regression line reflecting median placentation could be less obvious in high-risk women,
%hCG-h against gestational weeks in the scatterplot showed while maternal predisposing factors for vascular injury may
lower MoM values in early-onset pre-eclampsia before become more important. It should be noted that high-risk
12 weeks gestation than in the other groups. This is in conditions per se multiply the risk for pre-eclampsia.
agreement with previous publications [7, 26]. However, only The negative result of our attempt to increase the pre-
nine women had early-onset pre-eclampsia. Thus, it is not dictive power of a multivariate model with a new bio-
possible to draw any definitive conclusions. marker as well as the similar results from studies of others
Adding the Doppler measurement of Uta-PI to the raises the question: Why are we not finding a good predict-
multivariate regression models increased the AUCs of ive model? Heterogeneous nature of pre-eclampsia poses a
the models (from model 2 to model 3) for all challenge. Myatt and co-workers [33] have suggested strat-
pre-eclampsia and its subtypes except for early-onset egy to pre-eclampsia research including standardised data
pre-eclampsia. This stands in contrast to earlier studies collection to hasten our understanding of the cause of the
conducted in high-risk cohorts, where Uta-PI predicted disease and to improve the early recognition.
early-onset pre-eclampsia and severe pre-eclampsia bet-
ter than all pre-eclampsia or late-onset pre-eclampsia. Conclusions
However, our results are in accordance to the results This study was a preliminary study investigating the po-
from the same studies showing that mean Uta-PI is tential of hCG-h or %hCG-h to improve the prediction
much less useful for prediction of pre-eclampsia and its of pre-eclampsia in a high-risk cohort in the first trimes-
subtypes in the first trimester in a high-risk population ter using a multivariate regression model. There was a
than in a low-risk or screening population [27–29]. significant reduction of %hCG-h levels concentrations in
women with late-onset and non-severe pre-eclampsia,
Strength and limitations but in combination with other biomarkers, maternal
It should be noted that the definition of early-onset characteristics, MAP and Uta-PI, the sensitivity and the
pre-eclampsia was different in the present study than in positive predictive values of the multivariate regression
most studies. We defined early-onset pre-eclampsia as models did not meet the requirements for a clinically
cases diagnosed before 34 weeks of gestation, whereas useful screening test among high-risk women.
Murtoniemi et al. BMC Pregnancy and Childbirth (2018) 18:279 Page 9 of 10

Additional files Author details


1
University of Helsinki and Turunmaa District Hospital, Gynaecological
Outpatient Clinic, Hospital District of Southwest Finland, Kaskenkatu 13,
Additional file 1: Table S1. Inclusion and exclusion criteria of the risk
20700 Turku, Finland. 2Obstetrics and Gynecology, University of Helsinki and
group. The inclusion and exlusion criteria for the risk group in PREDO
Helsinki University Hospital, P.O. Box 140, FI-00029 Helsinki, Haartmaninkatu 2,
project. (DOCX 67 kb)
Finland. 3Department of Biostatistics, University of Turku, FI-20014 TURUN
Additional file 2: Table S2. The number of each inclusion criterion in YLIOPISTO, Turku, Finland. 4Department of Obstetrics and Gynaecology, Oulu
women with and without pre-eclampsia. (DOCX 77 kb) University Hospital and University of Oulu, Kajaanintie 50, 90220 Oulu,
Additional file 3: Table S3. Concentrations of biomarkers, mean arterial Finland. 5Finnish Medical Society Duodecim / Current Care, PL 713,
pressure and uterine artery pulsatility index. The median concentrations Kalevankatu 11 A, 00101 HELSINKI, Finland. 6HUSLAB and Department of
of each biomarker, mean arterial pressure and uterine artery pulsatility Clinical Chemistry, University of Helsinki and Helsinki University Hospital, PO
index in women with and without pre-eclampsia and by pre-eclampsia BOX 720, 00029 Helsinki, Finland. 7Hospital for Children and Adolescents,
subtype. (DOCX 78 kb) University of Helsinki and Helsinki University Hospital, Stenbäckinkatu 11, P.O.
Box 281, FI-00029 Helsinki, Finland. 8Department of Psychology and
Logopedics, University of Helsinki, Siltavuorenpenger 1-5, P.O. Box 9, FI-00014
Abbreviations Helsinki, Finland. 9Suomen Terveystalo Oy, Asemakatu 22-24, 70100 Kuopio,
%hCG-h: serum proportion of hCG-h to total hCG; AUC: area under the Finland. 10Institute for Molecular Medicine and Medical and Clinical Genetics,
curve; BMI: body mass index; GA: gestational age in days; hCG: human University of Helsinki, P.O. Box 63, FI-00014 Helsinki, Finland. 11University of
chorionic gonadotropin; hCG-h: hyperglycosylated human chorionic Helsinki and Helsinki University Hospital, P.O. Box 63, FI-00014 Helsinki,
gonadotropin; MAP: mean arterial pressure; MoM: multiple of the median; Finland.
OR: odds ratio; PAPP-A: pregnancy-associated plasma protein A; PlGF: placental
growth factor; PREDO: Prediction and Prevention of Pre-eclampsia and Intrauterine Received: 19 April 2017 Accepted: 19 June 2018
Growth Restriction cohort; ROC: receiver operating characteristic; SGA: small for
gestational age; Uta-PI: mean pulsatility index of the uterine artery

Acknowledgements References
We thank docent Eeva Ekholm, University of Turku and Turku University Hospital, 1. Duley L. The global impact of pre-eclampsia and eclampsia. Semin Perinatol.
for the valuable comments on the article. This work was supported by 2009;33(3):130–7.
PerkinElmer Finland Oy (PlGF assays). The PREDO project was supported by 2. Roberge S, Villa P, Nicolaides K, Giguère Y, Vainio M, Bakthi A, Ebrashy A,
Finska Läkaresällskapet, Foundation of Turku University Hospital, Päivikki and Bujold E. Early administration of low-dose aspirin for the prevention of
Sakari Sohlberg Foundation, University of Helsinki Research Funds, Finnish preterm and term preeclampsia: a systematic review and meta-analysis.
Medical Foundation, Academy of Finland, Signe and Ane Gyllenberg Foundation, Fetal Diagn Ther. 2012;31(3):141–6.
Sigrid Juselius Foundation, Foundation for Paediatric Research, Novo Nordisk 3. Roberge S, Nicolaides KH, Demers S, Villa P, Bujold E. Prevention of perinatal
Foundation and Jane and Aatos Erkko Foundation. We thank the doctors and death and adverse perinatal outcome using low-dose aspirin: a meta-analysis.
study nurses for their work on the PREDO Project. We are grateful to the women Ultrasound Obstet Gynecol. 2013;41(5):491–9.
who participated in the PREDO Project. 4. AGOC Committee Opinion No. 638. First-trimester risk assessment for early-
onset preeclampsia. Obstet Gynecol. 2015;126:e25–7.
Availability of data and material 5. Bartsch E, Medcalf KE, Park AL, Ray JG. Clinical risk factors for pre-eclampsia
The datasets used to analyse the current study are available from the determined in early pregnancy: systematic review and meta-analysis of
corresponding author on reasonable request. large cohort studies. BMJ. 2016;353:i1753.
6. Nicolaides KH. Turning the pyramid of prenatal care. Fetal Diagn Ther. 2011;
Funding 29(3):183–96.
Finska Läkaresällskapet, Foundation of EVO research funding (A special 7. Keikkala E, Vuorela P, Laivuori H, Romppanen J, Heinonen S, Stenman UH.
Finnish state subsidy for health science research), Academy of Finland, Signe First trimester hyperglycosylated human chorionic gonadotrophin in serum
and Ane Gyllenberg Foundation, Sigrid Juselius Foundation, University of - a marker of early-onset preeclampsia. Placenta. 2013;34(11):1059–65.
Helsinki Research Funds, Finnish Medical Foundation, Jane and Aatos Erkko 8. Girchenko P, Hämäläinen E, Kajantie E, Pesonen AK, Villa P, Laivuori H,
Foundation, and Päivikki and Sakari Sohlberg Foundation. Räikkönen K. Prediction and prevention of preeclampsia and intrauterine
growth restriction (PREDO) study. Int J Epidemiol. 2016; https://doi.org/10.
1093/ije/dyw154.
Authors’ contributions
9. Villa PM, Kajantie E, Räikkönen K, Pesonen AK, Hämäläinen E, Vainio M,
The authors made the following substantial contributions to this work:
Taipale P, Laivuori H. Aspirin in the prevention of pre-eclampsia in high-risk
concept and design: PV, EH, KR, AKP, PT, EK, HL; analysis and interpretation of
women: a randomised placebo-controlled PREDO trial and a meta-analysis
data: KM, JM; writing and drafting the article: KM, JM, PV, HL; and revising the
of randomised trials. BJOG. 2013;120(1):64–74.
manuscript for intellectual content: all authors. Final approval of the version
10. Gómez O, Figueras F, Fernández S, Bennasar M, Martínez JM, Puerto B,
to be published was given by all authors.
Gratacós E. Reference ranges for uterine artery mean pulsatility index at 11-41
weeks of gestation. Ultrasound Obstet Gynecol. 2008;32(2):128–32.
Ethics approval and consent to participate 11. Stenman UH, Birken S, Lempiäinen A, Hotakainen K, Alfthan H. Elimination
An Ethics Committee of the Helsinki and Uusimaa Hospital District approved of complement interference in immunoassay of hyperglycosylated
the study and written informed consent was obtained from all participants. human chorionic gonadotropin. Clin Chem. 2011;57(7):1075–7.
12. Practice Bulletin No ACOG. 33: diagnosis and Management of Preeclampsia
Consent for publication and Eclampsia. Obstet Gynecol. 2002;99(1):159–67.
Not applicable. 13. Pihkala J, Hakala T, Voutilainen P, Raivio K. Characteristic of recent fetal
growth curves in Finland. Duodecim. 1989;105(18):1540–6.
Competing interests 14. Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear
Consulting fees from Roche Diagnostics (Villa). Finox Biotech Nordics AB, models via coordinate descent. J Stat Softw. 2010;33(1):1–22.
unconditional support for the Meeting of the Nordic Expert Group (Laivuori). 15. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, Müller M.
pROC: an open-source package for R and S+ to analyze and compare ROC
curves. BMC Bioinformatics. 2011;12:77.
Publisher’s Note 16. Pavlou M, Ambler G, Seaman S, De Iorio M, Omar RZ. Review and evaluation
Springer Nature remains neutral with regard to jurisdictional claims in of penalised regression methods for risk prediction in low-dimensional data
published maps and institutional affiliations. with few events. Stat Med. 2016;35(7):1159–77.
Murtoniemi et al. BMC Pregnancy and Childbirth (2018) 18:279 Page 10 of 10

17. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under
two or more correlated receiver operating characteristic curves: a
nonparametric approach. Biometrics. 1988;44(3):837–45.
18. Kenny LC, Black MA, Poston L, Taylor R, Myers JE, Baker PN, McCowan LM,
Simpson NA, Dekker GA, Roberts CT, et al. Early pregnancy prediction of
preeclampsia in nulliparous women, combining clinical risk and biomarkers:
the screening for pregnancy endpoints (SCOPE) international cohort study.
Hypertension. 2014;64(3):644–52.
19. Hsieh FY, Bloch DA, Larsen MD. A simple method of sample size calculation
for linear and logistic regression. Stat Med. 1998;17(14):1623–34.
20. Poon LC, Kametas NA, Chelemen T, Leal A, Nicolaides KH. Maternal risk
factors for hypertensive disorders in pregnancy: a multivariate approach.
J Hum Hypertens. 2010;24(2):104–10.
21. Odegård RA, Vatten LJ, Nilsen ST, Salvesen KA, Austgulen R. Risk factors and
clinical manifestations of pre-eclampsia. BJOG. 2000;107(11):1410–6.
22. Meekins JW, Pijnenborg R, Hanssens M, McFadyen IR, van Asshe A. A study
of placental bed spiral arteries and trophoblast invasion in normal and
severe pre-eclamptic pregnancies. Br J Obstet Gynaecol. 1994;101(8):669–74.
23. Guibourdenche J, Handschuh K, Tsatsaris V, Gerbaud P, Leguy MC, Muller F,
Brion DE, Fournier T. Hyperglycosylated hCG is a marker of early human
trophoblast invasion. J Clin Endocrinol Metab. 2010;95(10):E240–4.
24. Cole LA. Hyperglycosylated hCG. Placenta. 2007;28(10):977–86.
25. Soto E, Romero R, Kusanovic JP, Ogge G, Hussein Y, Yeo L, Hassan SS, Kim
CJ, Chaiworapongsa T. Late-onset preeclampsia is associated with an
imbalance of angiogenic and anti-angiogenic factors in patients with and
without placental lesions consistent with maternal underperfusion. J Matern
Fetal Neonatal Med. 2012;25(5):498–507.
26. Keikkala E, Ranta JK, Vuorela P, Leinonen R, Laivuori H, Väisänen S, Marttala J,
Romppanen J, Pulkki K, Stenman UH, et al. Serum hyperglycosylated human
chorionic gonadotrophin at 14-17 weeks of gestation does not predict
preeclampsia. Prenat Diagn. 2014;34(7):699–705.
27. Cnossen JS, Morris RK, ter Riet G, Mol BW, van der Post JA, Coomarasamy A,
Zwinderman AH, Robson SC, Bindels PJ, Kleijnen J, et al. Use of uterine
artery Doppler ultrasonography to predict pre-eclampsia and intrauterine
growth restriction: a systematic review and bivariable meta-analysis. CMAJ.
2008;178(6):701–11.
28. Herraiz I, Arbues J, Camano I, Gomez-Montes E, Graneras A, Galindo A.
Application of a first-trimester prediction model for pre-eclampsia based on
uterine arteries and maternal history in high-risk pregnancies. Prenat Diagn.
2009;29(12):1123–9.
29. Herraiz I, Escribano D, Gomez-Arriaga PI, Hernindez-Garcia JM, Herraiz MA,
Galindo A. Predictive value of sequential models of uterine artery Doppler
in pregnancies at high risk for pre-eclampsia. Ultrasound Obstet Gynecol.
2012;40(1):68–74.
30. O'Gorman N, Wright D, Syngelaki A, Akolekar R, Wright A, Poon LC, et al.
Competing risks model in screening for preeclampsia by maternal factors
and biomarkers at 11-13 weeks gestation. Am J Obstet Gynecol. 2016;214(1):
103.e1–e12.
31. Crovetto F, Figueras F, Triunfo S, Crispi F, Rodriguez-Sureda V, Dominguez C,
et al. First trimester screening for early and late preeclampsia based on
maternal characteristics, biophysical parameters, and angiogenic factors.
Prenat Diagn. 2015;35(2):183–91.
32. Akolekar R, Syngelaki A, Poon L, Wright D, Nicolaides KH. Competing risks
model in early screening for preeclampsia by biophysical and biochemical
markers. Fetal Diagn Ther. 2013;33(1):8–15.
33. Myatt L, Redman CW, Staff AC, Hansson S, Wilson ML, Laivuori H, Poston L,
Roberts JM, CoLaboratory GP. Strategy for standardization of preeclampsia
research study design. Hypertension. 2014;63(6):1293–301.

You might also like