See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/324061003
Improved prediction of gestational hypertension by inclusion of placental
growth factor and pregnancy associated plasma protein-a in a sample of
Ghanaian women
Article  in  Reproductive Health · March 2018
DOI: 10.1186/s12978-018-0492-9
CITATIONS READS
0 73
7 authors, including:
Edward Antwi Kerstin Klipstein-Grobusch
Utrecht University University Medical Center Utrecht
17 PUBLICATIONS   102 CITATIONS    260 PUBLICATIONS   4,777 CITATIONS   
SEE PROFILE SEE PROFILE
Joyce L. Browne Peter C J I Schielen
Utrecht University National Institute for Public Health and the Environment (RIVM)
78 PUBLICATIONS   375 CITATIONS    118 PUBLICATIONS   1,591 CITATIONS   
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Autoimmune effects of hexachlorobenzene View project
CF screening View project
All content following this page was uploaded by Peter C J I Schielen on 28 March 2018.
The user has requested enhancement of the downloaded file.
Antwi et al. Reproductive Health  (2018) 15:56 
https://doi.org/10.1186/s12978-018-0492-9
RESEARCH Open Access
Improved prediction of gestational
hypertension by inclusion of placental
growth factor and pregnancy associated
plasma protein-a in a sample of Ghanaian
women
Edward Antwi1,2* , Kerstin Klipstein-Grobusch1,4, Joyce L. Browne1, Peter C. Schielen5, Kwadwo A. Koram3,
Irene A. Agyepong2 and Diederick E. Grobbee1
Abstract
Background: We assessed whether adding the biomarkers Pregnancy Associated Plasma Protein-A (PAPP-A) and
Placental Growth Factor (PlGF) to maternal clinical characteristics improved the prediction of a previously developed
model for gestational hypertension in a cohort of Ghanaian pregnant women.
Methods: This study was nested in a prospective cohort of 1010 pregnant women attending antenatal clinics in two
public hospitals in Accra, Ghana. Pregnant women who were normotensive, at a gestational age at recruitment of
between 8 and 13 weeks and provided a blood sample for biomarker analysis were eligible for inclusion. From serum,
biomarkers PAPP-A and PlGF concentrations were measured by the AutoDELFIA immunoassay method and multiple of
the median (MoM) values corrected for gestational age (PAPP-A and PlGF) and maternal weight (PAPP-A) were
calculated. To obtain prediction models, these biomarkers were included with clinical predictors maternal weight,
height, diastolic blood pressure, a previous history of gestational hypertension, history of hypertension in parents and
parity in a logistic regression to obtain prediction models. The Area Under the Receiver Operating Characteristic Curve
(AUC) was used to assess the predictive ability of the models.
Results: Three hundred and seventy three women participated in this study. The area under the curve (AUC) of the
model with only maternal clinical characteristics was 0.75 (0.64–0.86) and 0.89(0.73–1.00) for multiparous and
primigravid women respectively. The AUCs after inclusion of both PAPP-A and PlGF were 0.82 (0.74–0.89) and
0.95 (0.87–1.00) for multiparous and primigravid women respectively.
Conclusion: Adding the biomarkers PAPP-A and PlGF to maternal characteristics to a prediction model for
gestational hypertension in a cohort of Ghanaian pregnant women improved predictive ability. Further
research using larger sample sizes in similar settings to validate these findings is recommended.
Keywords: Prediction model, Gestational hypertension, Biomarkers, Hypertensive disorders of pregnancy
* Correspondence: ed_antwi@yahoo.com
1Julius Global Health, Julius Center for Health Sciences and Primary Care,
University Medical Center Utrecht, Utrecht University, Utrecht, the
Netherlands
2Ghana Health Service, P.M.B, Ministries, Accra, Greater Accra, Ghana
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.
Antwi et al. Reproductive Health  (2018) 15:56 Page 2 of 10
Plain English summary maternal clinical characteristics (diastolic blood pres-
Gestational hypertension and preeclampsia affect between sure, family history of hypertension in parents, history of
5 to 10% of all pregnancies and can result in complications gestational hypertension (GH) in a previous pregnancy,
in the mother and the fetus. Early prediction of pregnant parity, height and weight) improved prediction of gesta-
women at risk of these conditions will lead to better moni- tional hypertension.
toring and appropriate management. This study was con-
ducted in antenatal clinic settings in Ghana to investigate Methods
whether adding two biomarkers, placental growth factor Study design and study population
and pregnancy associated plasma protein A, to a previ- This study was nested in a prospective cohort of 1010
ously developed prediction model based on maternal clin- adult pregnant women with a singleton pregnancy and
ical characteristics improved the performance of the without known pre-existent hypertension recruited
model. between July 2012 and March 2014 at Ridge Regional
Logistic regression was used to derive a prediction Hospital and Maamobi General Hospital in Accra.
model. Adding biomarkers to a previously validated pre- Accra, the capital city of Ghana, is cosmopolitan with
diction model improved the performance of the model high, middle and low-income persons from different eth-
for gestational hypertension. nic backgrounds living and working in the city [18].
We recommend further research using larger sample Persons from all the social strata access health services,
sizes in similar settings to validate our findings. including antenatal and delivery care in these public
hospitals. These hospitals were also chosen because they
Background have a high attendance so the recruitment of pregnant
Hypertensive disorders of pregnancy (HDP) are leading women into the study could be completed in a shorter
causes of maternal morbidity and mortality globally and time. Eligibility criteria for this study were gestational
affect about 5 to 10% of all pregnancies [1, 2]. The burden age at enrollment of between between 8 and 13 weeks,
of these conditions is greatest in low and middle income based on ultrasound scan. This specific subset of women
countries (LMICs) [3, 4]. Early identification of pregnant was selected based on evidence that prediction with
women at risk of developing these conditions result in these biomarkers is most appropriate at this gestational
better monitoring and management to minimize compli- age [7–10, 19–21]. Women with gestational age at
cations to the mother and the fetus. Prediction models enrollment of less than 8 weeks or more than 13 weeks
have been used to identify women at high risk of HDPs, (n = 411), without PlGF MoM values (n = 95) or women
particularly preeclampsia [3–6]. In addition, prevention without outcome data (n = 131) were excluded. We used
interventions could be started such as calcium and aspirin the principle of 10 outcome events per variable for logis-
supplementation that have been shown to reduce the risk tic and Cox regression analysis [22–25] to obtain a
of HDPs, particularly preeclampsia [7–12]. For example, sample size adequate for our analysis. With an incidence
in the ASPRE (Combined Multimarker Screening and of gestational hypertension of 10% in the Ghanaian
Randomized Patient Treatment with Aspirin for population [26], and eight variables in the prediction
Evidence-Based Preeclampsia Prevention) trial with risk model, a sample size of 393 women was considered
selection based on screening, a reduction in the incidence adequate for the analysis.
of preterm preeclampsia in the aspirin arm by 62% was The women were included in the study after they had
observed [12]. given written informed consent and were interviewed by
PAPP-A is a protease that is involved in the local trained research assistants using a structured question-
release of insulin-like growth factors. Low first trimester naire for socio-demographic characteristics and obstetric
levels of PAPP-A is associated with hypertensive disor- history. They were followed up at each antenatal clinic
ders of pregnancy [13–15]. Placental growth factor visit till they delivered. None of the women who devel-
(PIGF) is an angiogenic factor and low concentrations oped gestational hypertension progressed to preeclamp-
have been observed in pregnant women who develop sia. Pregnancy outcomes were obtained at delivery and
preeclampsia. Suboptimal secretion of PlGF between 8 from the hospital maternity register.
to 14 weeks gestation as a result of placental dysfunction
has been associated with disorders such as preeclampsia, Variables
intrauterine growth restriction, small-for-gestational age Independent variables
and still births [16]. Maternal weight (measured in kilogrammes with a bath-
The aim of this study was to assess whether the room scale), height (measured in centimeters with a
addition of the biomarkers, placental growth factor stadiometer), blood pressure (measured in millimeters of
(PIGF) and pregnancy-associated protein A (PAPP-A) to mercury) and urine protein (defined as 2+ or more on
a previously developed prediction model [17] based on urine dipstick) were obtained at the initial and
Antwi et al. Reproductive Health  (2018) 15:56 Page 3 of 10
subsequent antenatal clinic visits from the maternal PlGF was not corrected for maternal weight because
health record books. serum PlGF concentration is not correlated with mater-
Blood pressure measurements were performed by nal weight [29].
trained midwives using a mercury sphygmomanometer.
The appropriate adult sized cuff was placed on the bare Outcome
left upper arm with the woman comfortably seated and The outcome, gestational hypertension, was defined as a
her back supported and legs uncrossed. The arm was at systolic BP of 140 mmHg or more and or a diastolic BP
the level of the heart and neither the patient nor the ob- of 90 mmHg or more on at least two separate occasions,
server talked during the measurement. Korotkoff phase and present for the first time after 20 weeks of preg-
V sounds were used [27]. Two readings were taken at nancy [30].
interval of five minutes and the average used as the
woman’s blood pressure. Ethical considerations
Ethical approval for the study was granted by the Ethical
PAPP-A and PlGF assay Review Committee of the Ghana Health Service (GHS-
Blood specimen was obtained from women on the day ERC 07/09/11). All participating women gave written
of their enrollment into the study by a phlebotomist. informed consent before they were enrolled in the study.
After the blood had coagulated, it was centrifuged to
obtain the serum which was stored at a temperature of Statistical analysis
-20 °C in a freezer at the Maamobi General Hospital. SPSS software (version 20.0, IBM SPSS Statistics Inc.,
Serum samples from the Ridge Hospital were stored Chicago, Illinois, USA) and R statistical software (R ver-
temporarily in a fridge at 4 °C and transported daily in a sion 3.1.0 (2014–04-10). The R Foundation for Statistical
cold box with ice packs to the laboratory at Maamobi Computing Platform: x86_64-w64-mingw32/× 64 (64-
General Hospital for storage. The frozen serum samples bit)) were used for statistical analysis. The mean and
were air-freighted on dried ice to the Dutch Institute for standard deviation of continuous predictors were calcu-
Public Health and Environment (RIVM) in Bilthoven, lated for women who developed gestational hypertension
the Netherlands, where they were stored at a and those who did not. Means were compared using the
temperature of − 80 °C until they were analyzed for Student’s t-test; percentages for categorical data were
PIGF and PAPP-A. PAPP-A and PlGF concentrations assessed by Chi-square test. The median with interquar-
were determined using commercially available immuno- tile range was reported for non-normally distributed
assays and the AutoDelfia automated analyzer (PerkinEl- variables.
mer, Turku, Finland). Details of the assay method are Logistic regression was used to derive the original pre-
described elsewhere by Browne et al. [28]. PAPP-A con- diction model using gestational hypertension as the out-
centrations were corrected for gestational age and ma- come and the following maternal clinical characteristics
ternal weight and expressed as multiple of the median as the predictors: maternal height, weight, parity, previ-
(MoM) using the reference equations from the Dutch ous history of gestational hypertension, family history of
national prenatal screening programme for Down syn- hypertension and diastolic blood pressure. The maternal
drome based on PAPP-A measurements between 8 to weight, height, diastolic blood pressure, parity, PAPP-A
13 weeks gestation of more than 10,000 pregnancies MoM and PlGF MoM were included in the logistic
[29]: regression model as continuous variables. The principle
PAPP-A MoM gestational age correction of 10 events per variable for logistic and Cox regression
 analysis [31] was applied in model building. A history ofy ¼ 12; 605:9606–552:53697
  hypertension in parents and history of gestational hyper-þ7:42649 2–0:0278 3; tension in a previous pregnancy were included in the lo-
gistic regression as dichotomous variables. As the
where x = gestational age at blood sampling in days. variable ‘previous history of gestational hypertension’
PAPP-A MoM weight correction; Exp (1.23234075– was not applicable to primigravid women, a separate
0.0181912*x), where x = weight in kilograms. model was fitted for them.
PlGF concentrations were also corrected for gestational PAPP-A MoM and PlGF MoM were included in the
age and expressed as MoM [28] by using the manufac- model as continuous variables so as not to lose power
turer’s (Perkin Elmer) reference equation for gestational through categorization, and also because the appropriate
age in days (between 9 to 13 weeks gestation) as follows: cut-off value of these biomarkers for the Ghanaian
y ¼ 75:08–1:7769  þ0:015892 population is not known [28]. The PAPP-A and PlGF as
MoM values were included in turns and then together
where x = gestational age at blood sampling in days. to the logistic regression. The predictive ability of each
Antwi et al. Reproductive Health  (2018) 15:56 Page 4 of 10
model (PAPP-A only, PlGF only, combined) was between women with and without gestational hyperten-
assessed. The models were internally validated using the sion (159.1 cm (SD 7.1) vs. 161.4 cm (SD 6.3), p = 0.08).
bootstrapping technique. The resulting shrinkage factor However, there was a difference in the mean weight of
after bootstrapping was used to adjust the regression co- women with and without gestational hypertension
efficients, thus correcting for model overfitting. (72.9 kg (SD 16.3) vs. 66.0 kg (SD 12.9), p = 0.013). The
The performance of the models was assessed by the mean diastolic blood pressure differed between women
area under the receiver operating characteristic curve who developed gestational hypertension and those who
(AUC) or c-statistic. The AUC of the original model did not (74.3 mmHg (SD 13.6) vs. 68.5 mmHg (SD 9.9),
with only maternal clinical characteristics was compared p = 0.006).
to that of the models with PAPPA and maternal clinical Table 3 presents the median and interquartile range of
characteristics, PlGF with maternal characteristics and MoM of PAPP-A and PlGF by gestational week. The
both PAPP-A, PlGF and maternal characteristics. median MoM PAPP-A (adjusted for gestational age and
maternal weight) ranged between 1.68 and 4.36. The
Results median MoM PlGF ranged between 0.90 and 1.68.
Characteristics of the 373 study participants are pre- Table 4 shows the regression coefficients and the AUC
sented in Table 1. Most of the women (81%) were mul- of the various models for multiparous women. The AUC
tiparous. The mean age was 28.3 (SD 4.9) years and the of the model with only maternal characteristics was 0.75
mean gestational age at booking was 11.6 weeks (SD (0.64–0.86). The AUC of the model with maternal char-
1.4). acteristics and PAPP-A was 0.78 (0.70–0.87), with ma-
The flow chart for selection of study participants is ternal characteristics, and PlGF was 0.76 (0.64–0.87),
shown in Fig. 1. Of 1010 women in the original cohort, and maternal characteristics with both biomarkers 0.82
373 women met the inclusion criteria. (0.74–0.89). Figure 2 shows the Receiver Operating
Table 2 compares characteristics of women who devel- Characteristic curves for the prediction models for mul-
oped gestational hypertension to those who did not. tiparous women. Table 5 shows the regression coeffi-
Twenty-five women (6.7%) developed gestational hyper- cients and the AUC of the models for primigravid
tension. There was a difference in mean age between women. The AUC of the model with only maternal char-
women who developed gestational hypertension and acteristics was 0.89 (0.73- 1.00). The AUC of the model
those who did not (30.3 (SD 5.3) years vs. 28.2 (SD 4.9) with maternal characteristics and both biomarkers was
years, p = 0.04). There was no difference in mean height 0.95 (0.87-1.00).
Table 1 Baseline characteristics of the study population Discussion
(n = 373) The addition of PlGF and PAPP-A together to the model
Variable Mean (SD) or N (%) markedly improved its predictive ability, with an
Age (years) 28.3 (4.9) increase in AUC from 0.75 to 0.82 for multiparous
women and 0.89 to 0.95 for primigravid women, whereas
Height (cm) 161.2 (6.3)
adding either one of the two had only marginal effect.
Weight (kg) 66.5 (13.3) These findings are in line with other studies that
Systolic blood 110.5 (12.9) reported improved prediction by the addition of bio-
pressure (mmHg) markers to maternal characteristics [5, 19, 32–34].
Diastolic blood 68.9 (10.3) Several issues arise in comparing this study to other
pressure (mmHg)
prediction studies. The first is that most prediction
Gestational age at 11.6 (1.4) models predict preeclampsia rather than gestational
booking (weeks)
hypertension [35]. Hence there were fewer prediction
PlGF MoM corrected Median 1.28, IQR
for gestational age (0.96–1.88) models for gestational hypertension to which we could
directly compare our models. Therefore we included
PAPP-A MoM corrected Median 2.29, IQR
for gestational age (1.15–3.86) models for preeclampsia as well in the comparison of
the model performance.
PAPP-A MoM corrected Median 2.34, IQR
for gestational age and (1.19–3.82) The second issue is that we derived separate models
maternal weight for multiparous and primigravid women. This was
Parity: because the primigravid women could not respond to
Primigravid women 71 (19.0%) the question of ”a previous history of gestational hyper-
tension or preeclampsia”. Being an important predictor,
2–3 pregnancies 116 (31.1%)
we maintained that variable in the model and in a sub
> 4 pregnancies 186 (49.9%) analysis fitted a different model for primigravid women
Antwi et al. Reproductive Health  (2018) 15:56 Page 5 of 10
Fig. 1 Flow chart illustrating participant selection
(n = 71). However because of the relatively small number maternal factors, biophysical and biomarkers compared
of primigravid women and outcome events on which with using only maternal factors [19].
these estimates are based, they should be interpreted Poon et al also reported that PAPP-A and PlGF in
with caution The third issue is that the same types of combination with maternal characteristics and uterine
biomarkers are not used across prediction studies. artery pulsatility index improved detection rates of pre-
Hence finding studies with the same predictors as in this eclampsia [21]. We did not include uterine artery pulsa-
study was a challenge. A number of prediction studies tility index in our study because uterine artery Doppler
also added uterine artery pulsatility index to biomarkers is not readily available in low resource settings.
and maternal characteristics [19, 21, 32] because it im- Another issue is that most of the prediction studies
proves prediction. For instance, Kuc et al. reported that have been conducted in Europe and North America.
the best detection rates for preeclampsia were obtained There are few studies in Sub Saharan African popula-
when maternal characteristics, biomarkers and uterine tions to which we could directly compare our results.
artery pulsatility index were combined [32]. Akolekar et Ukah et al in a prospective cohort study of pregnant
al. also reported a three-fold increase in detection rates women attending antenatal care in Maputo,
in screening for preeclampsia by the combination of Mozambique, measured the serum PlGF concentration
Table 2 Baseline characteristics of the study population by the outcome, gestational hypertension
Variable (Mean (SD)) Gestational hypertension (No) Gestational hypertension (Yes) p-value
N = 348 N = 25
Age (years) 28.2 (4.9) 30.3 (5.3) 0.04
Height(cm) 161.4 (6.3) 159.1 (7.1) 0.08
Weight (kg) 66.0 (12.9) 72.9 (16.3) 0.013
Systolic blood pressure 110.1 (12.7) 116.4 (14.2) 0.018
(mmHg)
Diastolic blood pressure 68.5 (9.9) 74.3 (13.6) 0.006
(mmHg)
Gestational age at 11.6 (1.4) 11.3 (1.5) 0.38
booking (weeks)
Antwi et al. Reproductive Health  (2018) 15:56 Page 6 of 10
Table 3 Median and interquartile range of MoM of PAPP-A and PlGF by gestational week (n = 373)
Gestational week Number of MoM PAPP-A, median (IQR), MoM PAPP-A, median MoM PlGF, median (IQR),
women (%) adjusted for gestational age (IQR), adjusted for adjusted for gestational
and maternal weight maternal weight age
8 17 (4.5) 4.36 (1.06–8.47) 4.46 (1.19–6.42) 1.17 (0.85–1.51)
9 40 (10.7) 1.68 (1.04–4.64) 2.04 (0.86–4.25) 0.90 (0.73–1.36)
10 86 (23.1) 2.39 (1.45–3.83) 2.33 (1.44–4.12) 1.15 (0.97–1.66)
11 71 (19.3) 1.76 (0.85–3.05) 1.96 (0.88–3.01) 1.21 (0.95–1.49)
12 66 (17.6) 2.21 (1.06–3.65) 2.26 (1.20–3.34) 1.29 (1.03–1.91)
13 93 (24.8) 2.63 (1.49–4.51) 2.55 (1.57–4.05) 1.68 (1.34–2.94)
Total 373 2.29 (1.15–3.86) 2.34 (1.19–3.82) 1.28 (0.96–1.88)
IQR Interquartile range, MoM multiple of the median
The median MoM value of the reference population by default is 1. The gestational age and weight adjusted PAPP-A median MoM was 2.29. The weight adjusted
PAPP-A median MoM was 2.34 and the median PlGF MoM was 1.28
in women suspected of having preeclampsia after study, for the multiparous women, the AUC of the pre-
20 weeks of gestation. This study had as its primary out- diction model with only maternal clinical characteristics
come, the time-to-delivery after confirmation of pre- was 0.75 and this increased to 0.82 upon the addition of
eclampsia [36]. This study differed from ours in terms of both PlGF and PAPP-A to the prediction model. For the
being a diagnostic study rather than a prediction study. primigravid women, the AUC of the prediction model
The AUC is used to quantify the overall ability of a with only maternal clinical characteristics was 0.89 and
test or a logistic regression model to discriminate this increased to 0.95 upon the addition of both PlGF
between two outcomes such as disease or non-disease and PAPP-A to the prediction model This is an indica-
[37–40]. It generally ranges from 0.5 to 1 and represents tion that the addition of both biomarkers simultaneously
the prediction model’s ability to correctly classify a ran- to the models improved the prediction performance.
domly selected individual as being from one of two The higher median MoM values of PlGF (1.28) and
hypothetical populations [40–43]. An AUC value of 1.0 PAPP-A (2.29) in our study compared to the reference
is considered perfect, 0.9–0.99 excellent, 0.8–0.89 good, population of Dutch women (median MoM of 1 by
0.7–0.79 fair and 0.51–0.69 poor. An AUC of 0.5 is con- default) is consistent with other studies that have shown
sidered non-informative. Hence the AUC of 0.82 racial and ethnic differences in the levels of these bio-
obtained in our study shows that the model with mater- markers, particularly in women of African and Asian
nal characteristics and both PAPP-A and PlGF has good decent [45–54]. The median MoM of PAPP-A between
predictive ability. 8 weeks gestation to 13 weeks gestation ranged between
Pencina et al. [44] and Peters et al. [33] have also indi- 1.68 and 4.36. That of PlGF MoM ranged from 0.90 at
cated that increase in the AUC upon the addition of a gestational week 9 to 1.68 at gestational week 13. Differ-
predictor to a model shows that the predictor has ences in the median MoM PlGF and PAPP-A levels
improved the predictive ability of the model. In our between some ethnic groups in Ghana have also been
Table 4 Regression coefficients and AUC of prediction models for multiparous women (n = 302)
Variable Model with only maternal Model with addition Model with addition Model with addition of PlGF
characteristics of PlGF MoM of PAPP-A MoM MoM and PAPP-A MoM
Intercept 9.68 10.0 10.46 12.18
History of hypertension in −1.52 −1.50 − 1.60 −1.65
parents
Previous history of 0.47 0.55 0.42 0.72
hypertension in pregnancy
Weight 0.026 0.025 0.024 0.023
Height −0.097 −0.099 −0.102 − 0.112
Parity 0.29 0.29 0.33 0.34
Diastolic BP 0.036 0.036 0.037 0.042
PlGF MoM – −0.15 – −0.713
PAPP-A MoM – 0.033 0.098
AUC 0.75 (0.64–0.86) 0.76 (0.64–0.87) 0.78 (0.70–0.87) 0.82 (0.74–0.89)
Antwi et al. Reproductive Health  (2018) 15:56 Page 7 of 10
impact is a useful strategy to improving maternal and
perinatal outcomes.
Biomarkers have shown some promise in improving
the prediction of gestational hypertension and other
hypertensive disorders in pregnancy, although a lot
more research is still needed. Future studies using lar-
ger sample sizes should be conducted to confirm the
findings of this study. When confirmed, one factor to
be considered in the use of biomarkers in prediction
models in the clinical setting would be the cost of
carrying out biomarker tests, especially in LMIC set-
tings. A feasible approach in this regard would be the
use of dried blood spot samples (DBS) instead of
serum which requires refrigeration during storage and
transport. DBS have been widely used in newborn
screening for sickle cell disease [55, 56], human im-
Fig. 2 Receiver operating characteristic (ROC) curves of prediction munodeficiency virus screening in newborns and for
models for multiparous women. Model 1 (black line): Maternal
characteristics only, Model 2 (red line): Maternal characteristics and other disorders [57–66]. It is cheaper than conven-
PlGF MoM, Model 2 (red line): Maternal characteristics and PlGF tional serum assay and logistically simpler to imple-
MoM, Model 2 (red line): Maternal characteristics and PlGF MoM ment in screening programmes because samples can
be obtained and transported from remote locations
where the laboratory infrastructure is limited. The
reported in this population [28]. As a result of the higher technique for sample taking is also simpler and requires less
MoM values, there is the need for a correction factor for training compared to venepuncture. In using DBS however,
the Ghanaian population and sub populations to prevent an issue to be considered is how well the concentration of
the under estimation of risk calculations for placental the biomarkers in whole blood correlates with that of DBS.
disorders and aneuploidies. Pennings et al. [67] and Browne et al. [68] have shown that
the correlation coefficient between serum and DBS concen-
trations for PAPP-Aand ß-hCG were both greater than
Clinical and research implications 0.94. Cowans et al also reported that ß-hCG stability is
Hypertensive disorders of pregnancy, including gesta- improved in DBS as compared to serum storage. This
tional hypertension and preeclampsia, are among the makes the collection, storage, transport and assay of
leading causes of maternal morbidity in LMICs. In biomarkers using DBS feasible in low resource settings.
Ghana they rank as the third leading cause of mortality, It is recommended that this study should be replicated
having overtaken hemorrhage [26]. The ability to predict locally and externally in similar settings using larger
this in women at increased risk (of the disorder) and sample sizes to validate the findings of this study before
thereby institute preventive measures to minimize their possible translation to clinical practice.
Table 5 Regression coefficients and AUC of prediction models for primigravid women (n = 71)
Variable Model with only maternal Model with addition of Model with addition of Model with addition of
characteristics PlGF MoM PAPP-A MoM PlGF MoM and PAPP-A MoM
Intercept 17.64 21.96 19.41 14.92
History of hypertension in parents −1.47 −1.63 −1.49 −1.92
Previous history of hypertension in – – – –
pregnancy
Weight 0.123 0.154 0.134 0.148
Height −0.216 −0.264 −0.237 −0.214
Parity – – – –
Diastolic BP 0.110 0.118 0.116 0.118
PlGF MoM – 0.323 – 0.834
PAPP-A MoM – – 0.098 −0.373
AUC 0.899 (0.732–1.000) 0.925 (0.808–1.000) 0.903 (0.749–1.000) 0.951 (0.870–1.000)
Antwi et al. Reproductive Health  (2018) 15:56 Page 8 of 10
The feasibility and sustainability of any planned intro- Sciences, University of the Witwatersrand, Johannesburg, South Africa.
5
duction and eventual scale-up in the use of biomarkers Center for Infectious Diseases Research, Diagnostics and Screening (IDS),
National Institute for Public Health and the Environment (RIVM), Bilthoven,
to improve prediction of hypertensive disorders has to the Netherlands.
be assessed using a cost-benefit analysis.
Received: 21 December 2017 Accepted: 9 March 2018
Conclusion
The addition of PAPP-A and PlGF to prediction models
based on maternal clinical characteristics (diastolic blood References
pressure, family history of hypertension in parents, his- 1. Hutcheon JA, Lisonkova S, Joseph KS. Epidemiology of pre-eclampsia and
the other hypertensive disorders of pregnancy. Best Pract Res Clin Obstetr
tory of gestational hypertension in a previous pregnancy, Gynaecol. 2011;25(4):391–403.
parity, height and weight) markedly improved prediction 2. Peters RM, Flack JM. Hypertensive disorders of pregnancy. J Obstet Gynecol
of gestational hypertension. This study should be repli- Neonatal Nurs. 2004;33(2):209–20.
3. North RA, McCowan LME, Dekker GA, Poston L, Chan EHY, Stewart AW,
cated using a larger sample size. et al. Clinical risk prediction for pre-eclampsia in nulliparous women:
development of model in international prospective cohort. BMJ.
Abbreviations 2011;342:d1875.
AUC: Area under the receiver operating characteristic curve; BP: Blood 4. Park H, Kim S, Jung Y, Shim S, Kim J, Cho Y, et al. Screening models
Pressure; DBS: Dried Blood Spot Sample; GH: Gestational Hypertension; using multiple markers for early detection of late-onset preeclampsia in
HDP: Hypertensive Disorder of Pregnancy; IQR: Inter Quartile Range; low-risk pregnancy. BMC Pregnancy Childbirth. 2014;14(1):35.
LMIC: Low and Middle Income Country; MoM: Multiple of the Median; PAPP- 5. Nijdam ME, Janssen KJ, Moons KG, Grobbee DE, van der Post JA, Bots ML,
A: Pregnancy Associated Plasma Protein-A; PE: Preeclampsia; PlGF: Placental Franx A. Prediction model for hypertension in pregnancy in nulliparous
Growth Factor; RIVM: Dutch Institute for Public Health and Environment; women using information obtained at the first antenatal visit. J Hypertens.
ROC: Receiver Operating Characteristic Curve; SD: Standard Deviation 2010;28(1):119–26.
6. Payne B, Hodgson S, Hutcheon JA, Joseph KS, Li J, Lee T, et al.
Acknowledgements Performance of the fullPIERS model in predicting adverse maternal
We acknowledge the midwives and the laboratory staff who played a role in outcomes in pre-eclampsia using patient data from the PIERS (pre-
the study. We also thank Dr. Justice Ahetor for assisting with aspects of the eclampsia integrated estimate of RiSk) cohort, collected on admission
statistical analysis. 3560. BJOG Int J Obstet Gynaecol. 2013;120(1):113–8.
7. Bujold E, Roberge SP, Lacasse Y, Bureau M, Audibert F, Marcoux S, et al.
Funding Prevention of preeclampsia and intrauterine growth restriction with
This research received funding from the UMC Utrecht Global Health Support aspirin started in early pregnancy: a meta-analysis. Obstet Gynecol.
program. The funders played no role in the study design, data collection, 2010;116(2, Part 1):402–14.
data analysis and interpretation as well as writing of the manuscript. 8. Duley L, Henderson‐Smart DJ, Meher S, King JF. Antiplatelet agents for
preventing pre‐eclampsia and its complications. The Cochrane Library. 2007;
Availability of data and materials (2):CD004659.
The datasets used and/or analysed in this study are available from the 9. Hofmeyr GJ, Lawrie TA, Atallah AN, Duley L. Calcium supplementation
corresponding author upon reasonable request. during pregnancy for preventing hypertensive disorders and related
problems. Cochrane Database Syst Rev. 2010;8(8):CD001059.
Authors’ contributions
10. Nicolaides KH. Turning the pyramid of prenatal care. Fetal Diagn Ther. 2011;
EA designed the study, collected data, carried out statistical analysis and
29(3):183–96.
wrote the initial draft of the manuscript. KK-G assisted with data analysis.
11. Roberge Sp VP, Nicolaides K, Giguire Y, Vainio M, Bakthi A, et al. Early
KK-G, JLB, PCS, KAK, IAA and DEG provided scientific guidance and were also
administration of low-dose aspirin for the prevention of preterm and term
actively involved in the preparation and review of the manuscript. All the
preeclampsia: a systematic review and meta-analysis. Fetal Diagn Ther. 2012;
authors read and approved the final manuscript.
31(3):141–6.
Ethics approval and consent to participate 12. Rolnik DL, Wright D, Poon LCY, Syngelaki A, O'Gorman N, de Paco Matallana
Ethics approval for this study was granted by the Ghana Health Service C, Akolekar R, Cicero S, Janga D, Singh M, Molina FS, Persico N, Jani JC,
Plasencia W, Papaioannou G, Tenenbaum-Gavish K, Nicolaides KH. ASPRE
Ethics Review Committee (Committee Reference Number: GHS-ERC 07/09/
trial: performance of screening for preterm pre-eclampsia. Ultrasound
11). All participating women gave written informed consent before they
Obstet Gynecol. 2017;50:492–95. https://doi.org/10.1002/uog.18816.
were enrolled in the study.
13. Bonno M, Oxvig C, Kephart GM, Wagner JM, Kristensen T, Sottrup-Jensen L,
Consent for publication et al. Localization of pregnancy-associated plasma protein-a and
Not applicable. colocalization of pregnancy-associated plasma protein-a messenger
ribonucleic acid and eosinophil granule major basic protein messenger
Competing interests ribonucleic acid in placenta. Lab Investig. 1994;71(4):560–6.
The authors declare that they have no competing interests. 14. Lawrence JB, Oxvig C, Overgaard MT, Sottrup-Jensen L, Gleich GJ, Hays LG,
et al. The insulin-like growth factor (IGF)-dependent IGF binding protein-4
protease secreted by human fibroblasts is pregnancy-associated plasma
Publisher’s Note protein-a. Proc Natl Acad Sci. 1999;96(6):3149–53.
Springer Nature remains neutral with regard to jurisdictional claims in 15. Bersinger NA, Smárason AK, Muttukrishna S, Groome NP, Redman CW.
published maps and institutional affiliations. Women with preeclampsia have increased serum levels of pregnancy-
associated plasma protein a (PAPP-A), inhibin a, activin a and soluble E-
Author details selectin. Hypeter Pregnan. 2003;22(1):45–55.
1Julius Global Health, Julius Center for Health Sciences and Primary Care, 16. Odibo AO, Patel KR, Spitalnik A, Odibo L, Huettner P. Placental pathology,
University Medical Center Utrecht, Utrecht University, Utrecht, the first-trimester biomarkers and adverse pregnancy outcomes. J Perinatol.
Netherlands. 2Ghana Health Service, P.M.B, Ministries, Accra, Greater Accra, 2014;34(3):186–91.
Ghana. 3Noguchi Memorial Institute for Medical Research, College of Health 17. Antwi E, Groenwold RH, Browne JL, Franx A, Agyepong IA, Koram KA, et al.
Sciences, University of Ghana, Legon, Accra, Ghana. 4Division of Development and validation of a prediction model for gestational
Epidemiology & Biostatistics, School of Public Health, Faculty of Health hypertension in a Ghanaian cohort. BMJ Open. 2017;7(1):e012670.
Antwi et al. Reproductive Health  (2018) 15:56 Page 9 of 10
18. Agyei-Mensah S, Owusu G. Segregated by neighbourhoods? A portrait of 40. Hosmer DW Jr, Lemeshow S, Sturdivant RX. Applied logistic regression. 3rd
ethnic diversity in the neighbourhoods of the Accra metropolitan area, Edition. Wiley; 2013. p. 528. ISBN: 978-0-470-58247-3.
Ghana. Popul Space Place. 2010;16(6):499–516. 41. Simpson AJ, Fitter MJ. What is the best index of detectability? Psychol Bull.
19. Akolekar R, Syngelaki A, Sarquis R, Zvanca M, Nicolaides KH. Prediction 1973;80(6):481.
of early, intermediate and late pre-eclampsia from maternal factors, 42. Bamber D. The area above the ordinal dominance graph and the area below
biophysical and biochemical markers at 11 to13 weeks. Prenat Diagn. the receiver operating characteristic graph. J Math Psychol. 1975;12(4):387–415.
2011;31(1):66–74. 43. Zhang DD, Zhou X, Freeman DH, Freeman JL. A Nonparametric method for
20. Akolekar R, Syngelaki A, Poon L, Wright D, Nicolaides KH. Competing the comparison of partial areas under ROC curves and its application to
risks model in early screening for preeclampsia by biophysical and large health care data sets. Stat Med. 2002;21(5):701–15.
biochemical markers. Fetal Diagn Ther. 2013;33(1):8–15. 44. Pencina MJ, D'agostino RB, Vasan RS. Statistical methods for assessment of
21. Poon LC, Nicolaides KH. First-trimester maternal factors and biomarker added usefulness of new biomarkers. Clin Chem Lab Med. 2010;48(12):1703–11.
screening for preeclampsia. Prenat Diagn. 2014;34(7):618–27. 45. Ball S, Wright D, Sodre D, Lachmann R, Nicolaides KH. Temporal effect of afro-
22. Concato J, Peduzzi P, Holford TR, Feinstein AR. Importance of events per Caribbean race on serum pregnancy-associated plasma protein-a at 9 to 13
independent variable in proportional hazards analysis I. Background, goals, weeks gestation in screening for aneuploidies. Fetal Diagn Ther. 2012;31(3):162–9.
and general strategy. J Clin Epidemiol. 1995;48(12):1495–501. 46. Cowans NJ, Spencer K. Effect of gestational age on first trimester maternal
23. Harrell FE Jr, Lee KLFAU, Mark DB. Multivariable prognostic models: issues in serum prenatal screening correction factors for ethnicity and IVF
developing models, evaluating assumptions and adequacy, and measuring conception. Prenat Diagn. 2013;33(1):56–60.
and reducing errors. Stat Med. 1996;15(0277–6715 (Print)):361–87. 47. Kagan KO, Wright D, Spencer K, Molina FS, Nicolaides KH. FirstGÇÉtrimester
24. Peduzzi P, Concato J, Feinstein AR, Holford TR. Importance of events per screening for trisomy 21 by free beta human chorionic gonadotropin and
independent variable in proportional hazards regression analysis II. Accuracy pregnancyGÇÉassociated plasma protein-A: impact of maternal and
and precision of regression estimates. J Clin Epidemiol. 1995;48(12):1503–10. pregnancy characteristics. Ultrasound Obstet Gynecol 2008;31(5):493–502.
25. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. A simulation study 48. Leung TY, Spencer K, Leung TN, Fung TY, Lau TK. Higher median levels of
of the number of events per variable in logistic regression analysis. J Clin free [beta]-hCG and PAPP-A in the first trimester of pregnancy in a Chinese
Epidemiol. 1996;49(12):1373–9. ethnic group. Fetal Diagn Ther. 2006;21(1):140.
26. Ghana Statistical Service (GSS), Ghana Health Service (GHS), and Macro 49. Manotaya S, Zitzler J, Li X, Wibowo N, Pham TM, Kang MS, et al. Effect of ethnicity
International. Ghana Maternal Health Survey 2007. Calverton: GSS, GHS, and on first trimester biomarkers for combined trisomy 21 screening: results from a
Macro International; 2009. multicenter study in six Asian countries. Prenat Diagn. 2015;35(8):735–40.
27. Pickering TG, Hall JE, Appel LJ, Falkner BE, Graves J, Hill MN, et al. 50. Nørgaard P, Wright D, Ball S, Newell P, Kirkegaard I, OBr P, et al. Autocorrelation
Recommendations for blood pressure measurement in humans and and cross-correlation between hCGߦ and PAPP-A in repeated sampling during
experimental animals part 1: blood pressure measurement in humans: a first trimester of pregnancy. Clin Chem Lab Med. 2013;51(9):1781–8.
statement for professionals from the Subcommittee of Professional and 51. Pandya P, Wright D, Syngelaki A, Akolekar R, Nicolaides KH. Maternal serum
Public Education of the American Heart Association Council on high placental growth factor in prospective screening for aneuploidies at 8 to13
blood pressure research. Hypertension. 2005;45(1):142–61. weeks gestation. Fetal Diagn Ther. 2012;31(2):87–93.
28. Browne JL, Klipstein-Grobusch K, Koster MP, Ramamoorthy D, Antwi E, 52. Spencer K, Heath V, El Sheikhah A, Ong CYT, Nicolaides KH. Ethnicity and
Belmouden I, et al. Pregnancy associated plasma protein-a and The need for correction of biochemical and ultrasound markers of
placental growth factor in a sub-Saharan African population: a nested chromosomal anomalies in the first trimester: a study of oriental, Asian and
cross-sectional study. PLoS One. 2016;11(8):e0159592. afro-Caribbean populations. Prenat Diagn. 2005;25(5):365–9.
29. Kuc S, Koster MP, Franx A, Schielen PC, Visser GH. Maternal 53. Tsiakkas A, Duvdevani N, Wright A, Wright D, Nicolaides KH. Serum placental
characteristics, mean arterial pressure and serum markers in early growth factor in the three trimesters of pregnancy: effects of maternal
prediction of preeclampsia. PLoS One. 2013;8(5):e63546. characteristics and medical history. Ultrasound Obstet Gynecol. 2015;45(5):591–8.
30. Report of the National High Blood Pressure Education Program Working Group 54. Wright D, Silva M, Papadopoulos S, Wright A, Nicolaides KH. Serum
on High Blood Pressure in Pregnancy. Am J Obstet Gynecol 183(1):s1-s22. pregnancyassociated plasma protein-A in the three trimesters of pregnancy:
31. Harrell FE, Lee KL, Mark DB. Multivariable prognostic models: issues in effects of maternal characteristics and medical history. Ultrasound Obstet
developing models, evaluating assumptions and adequacy, and measuring Gynecol. 2015;46(1):42–50.
and reducing errors. Stat Med. 1996;15(4):361–87. 55. Jinks DC, Minter M, Tarver DA, Vanderford M, Hejtmancik JF, McCabe ER.
32. Kuc S, Wortelboer EJ, van Rijn B, van Rijn BB, Franx A, Franx AF, Visser GH, Molecular genetic diagnosis of sickle cell disease using dried blood specimens
Schielen P, Schielen PC. Evaluation of 7 serum biomarkers and uterine artery on blotters used for newborn screening. Hum Genet. 1989;81(4):363–6.
Doppler ultrasound for first-trimester prediction of preeclampsia: a 56. Streetly A, Latinovic R, Hall K, Henthorn J. Implementation of universal
systematic review. Obstet Gynecol Surv. 2011;66(4):225-39. newborn bloodspot screening for sickle cell disease and other clinically
33. Peters SA, den Ruijter HM, Bots ML, Moons KG. Improvements in risk significant haemoglobinopathies in England: screening results for 2005-07.
stratification for the occurrence of cardiovascular disease by imaging J Clin Pathol. 2009;62(1):26–30.
subclinical atherosclerosis: a systematic review. Heart. 2011. https://doi.org/ 57. Therrell BL Jr, Hannon WH, Bailey DB Jr, Goldman EB, Monaco J, Norgaard-
10.1136/heartjnl-2011-300747. Pedersen B, et al. Committee report: considerations and recommendations for
34. Poon LCY, Kametas NA, Chelemen T, Leal A, Nicolaides KH. Maternal risk national guidance regarding the retention and use of residual dried blood spot
factors for hypertensive disorders in pregnancy: a multivariate approach. specimens after newborn screening. Obstet Gynecol Surv. 2011;66(11):687–9.
J Hum Hypertens. 2010;24(2):104–10. 58. Mei JV, Alexander JR, Adam BW, Hannon WH. Use of filter paper for the
35. Kleinrouweler CE, Cheong-See FM, Collins GS, Kwee A, Thangaratinam S, collection and analysis of human whole blood specimens. J Nutr. 2001;
Khan KS, Mol BW, Pajkrt E, Moons KG, Schuit E. Prognostic models in 131(5):1631S–6S.
obstetrics: available, but far from applicable. American Journal of Obstetrics 59. Hallack R, Doherty LE, Wethers JA, Parker MM. Evaluation of dried blood
& Gynecology. 2016;214(1):79-90. spot specimens for HIV-1 drug-resistance testing using the Trugene- HIV-1
36. Ukah UV, Mbofana F, Rocha BM, Loquiha O, Mudenyanga C, Usta M, Urso M, genotyping assay. J Clin Virol. 2008;41(4):283–7.
Drebit S, Magee LA, von Dadelszen P. Diagnostic Performance of Placental 60. Cassol SA, Lapointe N, Salas T, Hankins C, Arella M, Fauvel M, et al. Diagnosis
Growth Factor in Women With Suspected Preeclampsia Attending of vertical HIV-1 transmission using the polymerase chain reaction and dried
Antenatal Facilities in Maputo, MozambiqueNovelty and Significance. blood spot specimens. JAIDS. 1992;5(2):113–9.
Hypertension. 2017;69(3):469-74. 61. Bellisario R, Colinas RJ, Pass KA. Simultaneous measurement of thyroxine and
37. Brown CD, Davis HT. Receiver operating characteristics curves and related thyrotropin from newborn dried blood-spot specimens using a multiplexed
decision measures: a tutorial. Chemom Intell Lab Syst. 2006;80(1):24–38. fluorescent microsphere immunoassay. Clin Chem. 2000;46(9):1422–4.
38. Cook NR. Use and Misuse of the receiver operating characteristic curve in 62. Brambilla D, Jennings C, Aldrovandi G, Bremer J, Comeau AM, Cassol SA, et al.
risk prediction. Circulation. 2007;115(7):928. Multicenter evaluation of use of dried blood and plasma spot specimens in
39. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver quantitative assays for human immunodeficiency virus RNA: measurement,
operating characteristic (ROC) curve. Radiology. 1982;143(1):29–36. precision, and RNA stability. J Clin Microbiol. 2003;41(5):1888–93.
Antwi et al. Reproductive Health  (2018) 15:56 Page 10 of 10
63. Edelbroek PM, van der Heijden J, Stolk LM. Dried blood spot methods in
therapeutic drug monitoring: methods, assays, and pitfalls. Ther Drug Monit.
2009;31(3):327–36.
64. Bellisario R, Colinas RJ, Pass KA. Simultaneous measurement of antibodies to
three HIV-1 antigens in newborn dried blood-spot specimens using a
multiplexed microsphere-based immunoassay. Early Hum Dev. 2001;64(1):21–5.
65. McCabe ER, Huang SZ, Seltzer WK, Law ML. DNA Microextraction from dried
blood spots on filter paper blotters: potential applications to newborn
screening. Hum Genet. 1987;75(3):213–6.
66. Parker SP, Cubitt WD. The use of the dried blood spot sample in
epidemiological studies. J Clin Pathol. 1999;52(9):633.
67. Pennings JL, Siljee JE, Imholz S, Kuc S, de Vries A, Schielen PC, et al.
Comparison of different blood collection, sample matrix, and immunoassay
methods in a prenatal screening setting. Dis Markers. 2014;2014.509821.
68. Browne JL, Schielen PCJI, Belmouden I, Pennings JLA, Klipstein-Grobusch K.
Dried blood spot measurement of pregnancyassociated plasma protein a
(PAPP-A) and free ßsubunit of human chorionic gonadotropin (ßhCG) from
a low resource setting. Prenat Diagn. 2015;35(6):592–7.
Submit your next manuscript to BioMed Central 
and we will help you at every step:
•  We accept pre-submission inquiries 
•  Our selector tool helps you to find the most relevant journal
•  We provide round the clock customer support 
•  Convenient online submission
•  Thorough peer review
•  Inclusion in PubMed and all major indexing services 
•  Maximum visibility for your research
Submit your manuscript at
www.biomedcentral.com/submit
View publication stats