Results
1. Demographic characteristics of the participants subgroups
A total of 565 singleton pregnancies were included in the study. Of the participants with newly-onset symptom of elevated BP, 262 cases remained hypertensive without PE (GH group) while 201 cases progressed to PE eventually (PE group). Among the 102 participants with chronic hypertension, superimposed PE complicated 48 mothers during pregnancy (chrHBP with PE). Baseline characteristics were summarized inTable 1 . The maternal age of chrHBP with PE and chrHBP was significantly higher than GH and PE group (P<0.05), while there were no significant differences in maternal BMI and pregnancy complications (P>0.05).
2. sFlt-1/PLGF ratios in the participants subgroups
The sFlt-1/PlGF ratio in the chrHBP patients were significantly lower than the other three groups; Patients with PE had higher sFlt-1/PlGF ratios with or without the presence of chrHBP [i.e., PE (92.6) versus GH (19.2), P<0.001; chrHBP with superimposed PE (92.3) versus chrHBP (8.6), P=0.038] (Figure 1a, Table 2 ). The ratio in early-onset PE women were substantially higher than the late-onset subsets. (Figure 1b ) Significant differences were also observed in the ratios between the early- and late-onset participants with chrHBP with PE (P<0.05).
3. Performance of sFlt-1/PLGF ratio on the early prediction of progression from GH to PE
GH patients with progression to PE within 1 week had higher sFlt/PLGF ratio compared to corresponding GH patients without progression (184.5 vs 36.2 ). Patients who progressed within 2, 3 or 4 weeks also showed higher sFlt/PIGF ratios compared to corresponding GH group (81.5 vs 26.6, 180.7 vs 18.9, 180.7 vs 18.7, respectively). The predictive properties of the sFlt-1/PlGF ratio were presented via the Receiver Operating Characteristics (ROC) curves for different clinical subsets of PE. (Figure 2a ) The areas under the curve (AUCs) of the ROC were 77.98%, 76.90%, 77.20 and 78.39% for the ratio on predicting the progression to PE within 1, 2, 3 and 4 weeks, respectively(Figure 2b ) . For the selected single-cutoff model, the cutoff point 36.44 for 1 to 4 weeks showed the best sensitivity (0.85-0.94) and specificity (0.61-0.71), with negative predictive values (NPV) of 0.55-0.67 and positive predictive values (PPV) of 0.91-0.94(Table 3 ).
Comparing to the corresponding GH group, the median sFlt-1/PlGF ratios were 62.5 vs 25.9, 62.5 vs 17.2, 61.5 vs 16.8 and 61.5 vs 16.8 for the participants developing late-onset PE within 1 to 4 weeks, respectively (Figure 2a ) . The AUCs of the ROC were 89.88%, 88.58%, 86.25% and 86.54% for the ratio on predicting the occurrence of PE within 1, 2, 3 and 4 weeks, respectively. The cutoff point 51.52 for 1 to 4 weeks had NPVs of 0.55-0.67 and PPVs of 0.91-0.94(Figure 2b,Table 3 ).
4. Performance of sFlt-1/PLGF ratio on the early prediction of progression from chrHBP to PE
For the development of superimposed PE <34 weeks, the median sFlt-1/PlGF ratio for progression within 1 week was 270.9 vs 47.5 comparing to the corresponding chrHBP group, and 255.9 vs 20.1 both within 2 and 3 weeks and 213.1 vs 20.0 within 4 weeks (Figure 3a ) . The AUCs of ROC for the ratios predicting superimposed PE were 83.89%, 82.50%, 82.50% and 86.36% for progression within 1, 2, 3 and 4 weeks, respectively. For the selected single-cutoff model, the NPVs for cutoff at 47.78 for 2-4 weeks were 1.0 and PPVs were 0.90-0.96(Figure 3b,Table 4 ).
With regards to the chrHBP with superimposed PE after 34 weeks, the median sFlt-1/PlGF ratios were 32.9 vs 13.8, 32.9 vs 11.8, 32.1 vs 11.4 and 32.1 vs 11.4 for the progression within 1 to 4 weeks, respectively compared to the corresponding chrHBP group(Figure 3a ). The AUCs of the ROC were 72.41%, 72.41%, 73.03% and 73.03% on predicting the occurrence of PE within 1, 2, 3 and 4 weeks, respectively. The cutoff point 28.61 for 1 to 4 weeks had NPVs of 0.88 and PPVs of 0.43-0.79(Figure 3b,Table 4 ).