Improved prediction of gestational hypertension by inclusion of placental growth factor and pregnancy associated plasma protein-a in a sample of Ghanaian women

This study was nested in a prospective cohort of 1010 adult pregnant women with a singleton pregnancy and without known pre-existent hypertension recruited between July 2012 and March 2014 at Ridge Regional Hospital and Maamobi General Hospital in Accra. Accra, the capital city of Ghana, is cosmopolitan with high, middle and low-income persons from different ethnic backgrounds living and working in the city [18]. Persons from all the social strata access health services, including antenatal and delivery care in these public hospitals. These hospitals were also chosen because they have a high attendance so the recruitment of pregnant women into the study could be completed in a shorter time. Eligibility criteria for this study were gestational age at enrollment of between between 8 and 13 weeks, based on ultrasound scan. This specific subset of women was selected based on evidence that prediction with these biomarkers is most appropriate at this gestational age [7‐10, 19‐21]. Women with gestational age at enrollment of less than 8 weeks or more than 13 weeks (n = 411), without PlGF MoM values (n = 95) or women without outcome data (n = 131) were excluded. We used the principle of 10 outcome events per variable for logistic and Cox regression analysis [22‐25] to obtain a sample size adequate for our analysis. With an incidence of gestational hypertension of 10% in the Ghanaian population [26], and eight variables in the prediction model, a sample size of 393 women was considered adequate for the analysis.

The women were included in the study after they had given written informed consent and were interviewed by trained research assistants using a structured questionnaire for socio-demographic characteristics and obstetric history. They were followed up at each antenatal clinic visit till they delivered. None of the women who developed gestational hypertension progressed to preeclampsia. Pregnancy outcomes were obtained at delivery and from the hospital maternity register.

Maternal weight (measured in kilogrammes with a bathroom scale), height (measured in centimeters with a stadiometer), blood pressure (measured in millimeters of mercury) and urine protein (defined as 2+ or more on urine dipstick) were obtained at the initial and subsequent antenatal clinic visits from the maternal health record books.

Blood pressure measurements were performed by trained midwives using a mercury sphygmomanometer. The appropriate adult sized cuff was placed on the bare left upper arm with the woman comfortably seated and her back supported and legs uncrossed. The arm was at the level of the heart and neither the patient nor the observer talked during the measurement. Korotkoff phase V sounds were used [27]. Two readings were taken at interval of five minutes and the average used as the woman’s blood pressure.

Blood specimen was obtained from women on the day of their enrollment into the study by a phlebotomist. After the blood had coagulated, it was centrifuged to obtain the serum which was stored at a temperature of -20 °C in a freezer at the Maamobi General Hospital. Serum samples from the Ridge Hospital were stored temporarily in a fridge at 4 °C and transported daily in a cold box with ice packs to the laboratory at Maamobi General Hospital for storage. The frozen serum samples were air-freighted on dried ice to the Dutch Institute for Public Health and Environment (RIVM) in Bilthoven, the Netherlands, where they were stored at a temperature of − 80 °C until they were analyzed for PIGF and PAPP-A. PAPP-A and PlGF concentrations were determined using commercially available immunoassays and the AutoDelfia automated analyzer (PerkinElmer, Turku, Finland). Details of the assay method are described elsewhere by Browne et al. [28]. PAPP-A concentrations were corrected for gestational age and maternal weight and expressed as multiple of the median (MoM) using the reference equations from the Dutch national prenatal screening programme for Down syndrome based on PAPP-A measurements between 8 to 13 weeks gestation of more than 10,000 pregnancies [29]:

PAPP-A MoM gestational age correction

where x = gestational age at blood sampling in days.

PAPP-A MoM weight correction; Exp (1.23234075–0.0181912*x), where x = weight in kilograms.

PlGF concentrations were also corrected for gestational age and expressed as MoM [28] by using the manufacturer’s (Perkin Elmer) reference equation for gestational age in days (between 9 to 13 weeks gestation) as follows:

where x = gestational age at blood sampling in days.

PlGF was not corrected for maternal weight because serum PlGF concentration is not correlated with maternal weight [29].

The outcome, gestational hypertension, was defined as a systolic BP of 140 mmHg or more and or a diastolic BP of 90 mmHg or more on at least two separate occasions, and present for the first time after 20 weeks of pregnancy [30].

Ethical approval for the study was granted by the Ethical Review Committee of the Ghana Health Service (GHS-ERC 07/09/11). All participating women gave written informed consent before they were enrolled in the study.

SPSS software (version 20.0, IBM SPSS Statistics Inc., Chicago, Illinois, USA) and R statistical software (R version 3.1.0 (2014–04-10). The R Foundation for Statistical Computing Platform: x86_64-w64-mingw32/× 64 (64-bit)) were used for statistical analysis. The mean and standard deviation of continuous predictors were calculated for women who developed gestational hypertension and those who did not. Means were compared using the Student’s t-test; percentages for categorical data were assessed by Chi-square test. The median with interquartile range was reported for non-normally distributed variables.

Logistic regression was used to derive the original prediction model using gestational hypertension as the outcome and the following maternal clinical characteristics as the predictors: maternal height, weight, parity, previous history of gestational hypertension, family history of hypertension and diastolic blood pressure. The maternal weight, height, diastolic blood pressure, parity, PAPP-A MoM and PlGF MoM were included in the logistic regression model as continuous variables. The principle of 10 events per variable for logistic and Cox regression analysis [31] was applied in model building. A history of hypertension in parents and history of gestational hypertension in a previous pregnancy were included in the logistic regression as dichotomous variables. As the variable ‘previous history of gestational hypertension’ was not applicable to primigravid women, a separate model was fitted for them.

PAPP-A MoM and PlGF MoM were included in the model as continuous variables so as not to lose power through categorization, and also because the appropriate cut-off value of these biomarkers for the Ghanaian population is not known [28]. The PAPP-A and PlGF as MoM values were included in turns and then together to the logistic regression. The predictive ability of each model (PAPP-A only, PlGF only, combined) was assessed. The models were internally validated using the bootstrapping technique. The resulting shrinkage factor after bootstrapping was used to adjust the regression coefficients, thus correcting for model overfitting.

The performance of the models was assessed by the area under the receiver operating characteristic curve (AUC) or c-statistic. The AUC of the original model with only maternal clinical characteristics was compared to that of the models with PAPPA and maternal clinical characteristics, PlGF with maternal characteristics and both PAPP-A, PlGF and maternal characteristics.

Hypertensive disorders of pregnancy, including gestational hypertension and preeclampsia, are among the leading causes of maternal morbidity in LMICs. In Ghana they rank as the third leading cause of mortality, having overtaken hemorrhage [26]. The ability to predict this in women at increased risk (of the disorder) and thereby institute preventive measures to minimize their 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 larger 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 settings. 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 immunodeficiency virus screening in newborns and for other disorders [57‐66]. It is cheaper than conventional serum assay and logistically simpler to implement in screening programmes because samples can be obtained and transported from remote locations where the laboratory infrastructure is limited. The technique for sample taking is also simpler and requires less training compared to venepuncture. In using DBS however, an issue to be considered is how well the concentration of the biomarkers in whole blood correlates with that of DBS. Pennings et al. [67] and Browne et al. [68] have shown that the correlation coefficient between serum and DBS concentrations for PAPP-Aand ß-hCG were both greater than 0.94. Cowans et al also reported that ß-hCG stability is improved in DBS as compared to serum storage. This makes the collection, storage, transport and assay of biomarkers using DBS feasible in low resource settings.

It is recommended that this study should be replicated locally and externally in similar settings using larger sample sizes to validate the findings of this study before possible translation to clinical practice.

The feasibility and sustainability of any planned introduction and eventual scale-up in the use of biomarkers to improve prediction of hypertensive disorders has to be assessed using a cost-benefit analysis.