NTN-PW: normotensive pregnant women; BOS: biomarkers of oxidative stress; AGMs: angiogenic growth mediators; O&G: obstetrics and gynaecology; W1, wave 1 or visit 1, W2, wave 2 or visit 2; W3, wave 3 or visit 3
A qualified consultant obstetrician/gynaecologist physically examined all participants. PE was defined as systolic blood pressure (SBP) /diastolic blood pressure (DBP) greater than or equal to 140/90mmHg with visible proteinuria (≥1+ dipstick) or 24-hour proteinuria of ≥300mg/day on two (2) occasions at least four (4) hours apart detected after 20 weeks gestation in previously normotensive pregnant women. Early-onset PE (EO-PE) and late-onset PE (LO-PE) were defined as PE that occurred before and at or after 34 weeks gestation, respectively (American College of Obstetricians and Gynecologists & Task Force on Hypertension in Pregnancy, 2013; Raymond & Peterson, 2011)

Laboratory assays

Serum, plasma and urine samples were obtained from all participants up to a total of 3 visits or waves, at 10-week intervals throughout gestation (median 17, 27, and 37 weeks). Samples were stored at -80oC (Thermo Scientific Ultra-Low Freezer) until the biomarkers of OS and AGMs were analysed.
Following the manufacturer’s instructions, urinary and serum 8-OHdG were analysed in duplicate using highly sensitive and competitive ELISA kits (ab201734, Abcam, China). Serum concentrations of 8-OHdG were measured immediately after sample collection to avoid autoxidation during long storage. The inter-and intra- assay coefficients of variation (CV) were 3.5% and 4.5%, respectively. Urinary 8-OHdG concentrations were normalised to creatinine (Cr) concentrations and recorded as ng/mg Cr. Serum 8-epi-PGF2α was analysed in duplicate using competitive ELISA kits from ELabscience, China (cat. log E-EL-0041). The intra-and-inter assay coefficients of variation (CV) were 5.6% and 6.4%, respectively.
TAC reagents were obtained from Sigma-Aldrich (Hong Kong, China). Plasma samples were thawed to measure TAC spectrophotometrically at 593 nm using Mindray BA-88A, China. The estimation of TAC was based on the Ferric Reducing Ability of Plasma (FRAP) and the protocol as described by Benzie and Strain (1996). The absorbance was used to obtain the concentrations after comparison to standard curves and recorded in µmol/l.
AGMs including serum concentrations of VEGF-A, sFlt-1, PlGF, and sEng were measured in duplicate using competitive Quantikine ELISA kits from R&D System Inc. (Minneapolis, MN USA). Absorbance was measured at 450 nm wavelength using a microplate ELISA reader (Bio-Tek ELx808 microplate reader, Hayward, CA, USA). The inter-and intra- assay coefficient of variation obtained in our laboratory was 1.1 and 1.3 for VEGF-A, 1.5 and 3.8 for sFlt-1, 4.6 and 3.3 for PlGF and 2.8 and 5.2 for sEng, respectively.
All laboratory assays were performed at the Molecular Medicine Laboratory of the Kwame Nkrumah University of Science and Technology and the Biochemistry and Immunology Department of the Komfo Anokye Teaching Hospital, Ghana.

Statistical analyses

The normality of the data was tested using the Kolmogorov-Smirnov test. Data were presented as median (interquartile ranges) for non-parametric continuous variables and frequency (percentages) for categorical variables. A Chi-square test was performed to test associations between the proportions of variables among the study groups. Median comparisons between more than two independent variables were performed using Kruskal-Wallis one-way ANOVA followed by a Bonferroni posthoc multiple comparison test and adjusted p-values were recorded. A receiver operating characteristic (ROC) curve and area under the curve (AUC) were generated to evaluate the diagnostic performance of the model. P< 0.05 was considered statistically significant. Data were analysed using SPSS version 24 (IBM Corp, NY, USA), XLSTAT Premium version 2018.1 and R version 3.4.3 (R core Team 2017).

Results

Unlike OHS groups, there was a statistically significant difference between the median maternal ages between SHS pregnant group who developed PE compared to those who did not (p <0.001). There was a significantly increased SBP, DBP, sEng, sFlt-1, 8-epiPGF2α, serum 8-OHdG, urinary 8-OHdG and combined ratios of sFlt-1/PlGF ratio, 8-epiPGF2alpha/PlGF ratio, 8-OHdG/PlGF ratio and sEng/PlGF ratio, and correspondingly reduced PIGF, VEGF-A and TAC among PE groups compared to NTN-PW group (p <0.001). Unlike the OHS groups, the degree of imbalance in biomarkers of OS and AGMs was higher in SHS who developed EO-PE followed by LO-PE compared to NTN-PW (p <0.001). Although no statistical significance was observed, the clinically significant difference indicated by the high level of imbalances in favour of SHS rather than the OHS group was observed in biomarkers of OS and AGMs. Meanwhile, there was a significant difference in median SBP between SHS-associated NTN-pregnancy and OHS- associated NTN-pregnancy (p =0.038)(Table 1) .
Overall, SBP, DBP, and biomarkers of OS and AGMs increased from baseline to mid-pregnancy among SHS women who later developed PE and NTN-pregnancies rather than OHS pregnant women who later developed PE and NTN-pregnancies. At both early 2nd trimester (10-20 weeks gestation, W1) and mid-pregnancy (21-31 weeks gestation, W2), the median maternal serum levels of PlGF, VEGF-A, and plasma TAC were significantly decreased whereas those of sEng, sFlt-1, 8-epiPGF2α, 8-OHdG, urinary 8-OHdG and the ratios: sFlt-1/PlGF, 8-epiPGF2α/PlGF, 8-OHdG/PlGF and sEng/PlGF were significantly increased among the SHS who later developed EO-PE followed by LO-PE compared to NTN-PW (p <0.001). Similar observations occurred among the OHS group (p <0.001) even though the trend of imbalance was higher among the SHS group. There was a clinically significant difference between the SHS group and the OHS who later developed PE and those who did not (Table 2).
Within each visit, there was no difference in gestational age across the groups. Meanwhile, there was a significant difference in SBP and DBP at both visit 1 or W1 and visit 2 or W2 across the study groups (p<0.05) (Table 2).
Compared to the individual biomarkers at visit 1 or W1 and visit 2 or W2, the combined biomarkers of OS and AGMs, particularly the mid-pregnancy (W2) 8-OHdG/PIGF ratio yielded the highest discriminating power or AUC (0.93, p <0.001) (Figure 2c) with the best sensitivity (85.6%), specificity (92.4%), positive predictive value (PPV) (86.6%), negative predictive value (NPV) (85.2%), positive likelihood ratio (LR+) (9.9) and negative likelihood ratio (LR-) (0.1) at a cut-off value ≥0.80. At the cut-off value for 8-OHdG/PIGF ratio, NTN-PW had 4.8-fold increased odds of developing PE (adjusted odds ratio (aOR) =4.8 95%CI (1.5-11.5), p <0.001).
Except for W1 TAC levels, all the single and combined biomarkers of OS and AGMs yielded a significant (all p <0.05) discriminating power and adjusted odds ratios for predicting PE(Table S1) .