Gestational systolic blood pressure trajectories and risk of adverse maternal and perinatal outcomes in Chinese women

The study was based on a registered-based cohort study conducted in Taicang, a small but developed city in Jiangsu Province. Details of the Taicang register-based cohort was described previously [31, 32]. In brief, a total of 24,458 pregnant women who delivered live births between Jan. 1, 2014 and Nov. 30, 2019 in any hospital or community healthcare center in Taicang were enrolled. Subsequently, exclusion was successively made for 3635 individuals with fewer than three BP measurements between the 10th week and 40th gestational week, 367 subjects with polyembryony, and 103 pregnant women with chronic hypertension (≥140/90 mmHg) [31, 32]. Finally, a total of 20,353 pregnant women were included in the present study.

The flowchart of the exclusion and inclusion process of our study population is presented in Fig. 1. All participants provided written informed consent. This study was approved by the ethics committee of Soochow University and Maternal and Child Healthcare Center of Taicang.

Antenatal examination for general pregnant women in the current study was based on the policy of antenatal care in Taicang city: initiated at around 12 weeks of gestation, thereafter once every 4 weeks at < 28 weeks of gestation, once every 2 weeks at 28–37 weeks of gestation, and once per week at ≥37 weeks of gestation. The measurements of BP were taken as part of routine prenatal care by physicians, and BPs values were retrieved from the computerized tracking system maintained by the study clinical institutions.

BP was measured using a calibrated mercury sphygmomanometer following a standardized protocol. In brief, all participants were seated in an upright position with back support and instructed to relax for 5 min. A cuff was placed around the nondominant upper arm, which was supported at the level of the heart, with the bladder midline over the brachial artery pulsation. We assigned an average of two sequential BPs to each record, with a minimum 2-min rest period between measurements.

The main outcome variables in our study included maternal (GH, PE/eclampsia) and fetal outcomes [LBW, SGA, PTD and early-term delivery (ETD)]. Based on the JNC7 Guideline, the 2013 American College of Obstetricians and Gynecologists (ACOG) statement defines GH as hypertension (≥140/90 mmHg) that manifested after 20 weeks’ gestation without proteinuria, PE as GH accompanied by proteinuria or other symptoms [29]. Eclampsia was defined by a patient experiencing convulsions, who had PE or severe PE, with or without albuminuria, where another cause had been ruled out [33]. PTD was defined as a live-singleton birth that occurred no later than 36 weeks of gestation, whereas ETD was defined as a live birth with a gestational age between 37 and 38 weeks [34]. SGA was defined as a gestational-age-adjusted birth weight below the tenth percentile [35]. LBW was described as an infant weight < 2500 g at delivery [36].

When pregnant women started their first antenatal examination, information including maternal age, early pregnancy body mass index (BMI), BP, and obstetrical history (e.g., gestation, parity, and abortion in previous pregnancy) was collected. During the following antenatal visits, status of gestational diabetes mellitus (GDM) and thyroid disease were also examined by obstetricians as a clinical routine.

GDM was considered at approximately 25 weeks of gestation when any of the following criteria were met on the 75-g oral glucose tolerance test: fasting plasma glucose level ≥ 5.1 mmol/L, ≥10 mmol/L at 1 h, and ≥ 8.5 mmol/L at 2 h [37]. Thyroid disease during pregnancy was diagnosed according to the Guidelines of the American Thyroid Association [38].

The number of SBP measurements achieved at < 10 weeks of gestation or at ≥40 weeks of gestation was too small to be analyzed in the current study. Therefore, we used SBP values measured between the period of 10 weeks 0 days and 40 weeks 0 days.

In our study, the change patterns of SBP were fitted by LCGM model by using Proc Traj in statistical analysis system (SAS) software 9.4. A censored normal model (CNORM) was considered appropriate due to the continuity of SBP. We mainly considered Bayesian Information Criterion (BIC) and posterior class-membership probabilities to determine the optimal trajectory model. First, the closer the BIC value is to zero, the better the model fits the data. Second, for each model involving latent trajectories, posterior class-membership probabilities were used to obtain a posterior classification of the participants in each latent trajectories to evaluate goodness-of-fit and to characterise the discrimination of latent trajectories. The higher the mean posterior class-membership probabilities within each latent trajectories, the better the model is. Third, we also retrieved the proportion of subjects classified in each latent trajectories with a posterior probability above a threshold of 0.7, indicating the proportion of subjects unambiguously classified in each latent trajectories. Proportion of subjects with high posterior probabilities(i.e. > 0.7) reaches 65%, illustrating a good classification [28]. According to the above model selection criteria, we compared 2 to 7 trajectory models, and selected six trajectories as the optimal model. The parameter estimates of each trajectory model of model with 2 to 7 trajectories are shown in Supplementary Table 1. Finally, cubic, quadratic, and linear terms were evaluated based on their statistical significance after starting with the highest polynomial. In our final model, all of the six trajectories had cubic order terms.

Continuous and categorical variables were presented as mean ± standard deviation (SD) and frequency (percentage), respectively. Maternal and neonatal characteristics across SBP trajectories were compared by using analysis of variance (ANOVA) for normal distributed variables and Kruskal Wallis test for skewed data, respectively. We calculated odds ratio (OR) [95% Confidence level (95% Cl)] in four logistic models to evaluate the associations between SBP trajectories and adverse maternal and births outcomes. Model 1 was unadjusted. Model 2 controlled for possible influence of maternal age at delivery (in years, continuous), early pregnancy BMI (Kg/m², continuous), gestation, parity, and presence of GDM. Based on Model 2, Model 3 additionally adjusted for SBP (mmHg, continuous) at the first visit for antenatal care, and SBP measurement times (continuous) during pregnancy. Model 4 further included infant sex (boys, girls) and presence of HDP (including GH, PE and eclampsia), on the basis of Model 3. We also performed sensitivity analyses to assess the robustness of our findings. What’s more, we expanded the recruited population to women with more than two times of SBP records during the antenatal examination period to avoid selection bias of the population. Eventually, the improvement in risk identification [represent as c-statistics, continuous net reclassification index (NRI), and integrated discrimination improvement (IDI)] of adding SBP trajectories over established risk model (composed by variables in Models 3 or 4) was evaluated [39, 40]. All statistical tests were performed using SAS software (version 9.4, SAS Institute, Cary, NC, USA), and differences were considered statistically significant when two-sided P ≤ 0.05.

As shown in Figs. 2, 20,353 participants were assigned into six different subgroups: the numbers of subjects were 1521 (7.47%), 4454 (21.88%), 3893 (19.13%), 3352 (16.47%), 4404 (21.64%) and 2729 (13.41%) for the low delayed-increasing, low reverse-increasing, low-stable, medium reverse-increasing, medium-stable and high-stable patterns, respectively. Low, moderate, and high refers to SBP < 110 mmHg, 110–120 mmHg and > 120 mmHg in the 10th week of gestation, respectively. The high-stable trajectory had SBP around 125 mmHg in the 10th gestational week, and increased slightly onwards.

Mean maternal and gestational ages were 27.10 ± 4.38 years and 38.66 ± 1.30 weeks, respectively. Maternal and fetal demographic and baseline characteristics according to SBP trajectories are illustrated in Tables 1 and 2. In this study, the SBP measurement times during pregnancy were up to 15 (Median = 9, interquartile range = 7–10). Pregnant women belonged to the high-stable pattern presented the most adverse risk factor profile: they were older, and had the highest BMI and proportions of gestational diseases (GDM and thyroid disease). Fetuses whose mothers grouped into the high-stable were more likely to suffer ETD, PTD, LBW and SGA, and tended to have lower Apgar score at 1 min and 5 min.

As shown in Tables 2, 0.53, 1.28, 0.39, 1.14, 6.12 and 13.08% of women in the low delayed-increasing, low reverse-increasing, low-stable, medium-stable, medium reverse-increasing and high-stable patterns were defined as GH, respectively. The corresponding proportion was 0.66, 0.63, 0.64, 1.04, 0.89 and 1.58% for PE/eclampsia. Compared to women with the low-stable pattern, the medium reverse-increasing (OR = 3.16, 95% CI = 1.66–6.02) and the high-stable patterns (OR = 5.28, 95% CI = 2.76–10.10) were more likely to experience GH, after adjusting for variables in logistic regression Model 3. However, the associations between PE/eclampsia and SBP patterns were not statistically significant in any multivariate logistic regression models.

The incidences of ETD, PTD, SGA and LBW across different SBP trajectories are present in Table 3. Women demonstrated the high-stable trajectory had the highest risk of poor fetal outcomes among all the trajectories. Adjusted for variables in Model 4 and compared to women belonged to the low-stable pattern, mothers displayed the high-stable pattern had increased risk of averse birth outcomes, with OR (95% CI) of 1.30 (1.13–1.50) for ETD, 1.53 (1.12–2.08) for PTD, 1.32 (1.06–1.65) for SGA and 1.64 (1.08–2.48) for LBW, respectively. Mothers with the medium-stable trajectory also showed increased risk of ETD (OR = 1.17, 95% CI = 1.05–1.30) and PTD (OR = 1.31, 95% CI = 1.03–1.67). Meanwhile, fetuses whose mothers displayed the low reverse-increasing (OR = 1.25, 95% CI = 1.12–1.39) and the medium reverse-increasing patterns (OR = 1.27, 95% CI = 1.12–1.44) were more likely to have ETD newborns. Nevertheless, mothers with the low delayed-increasing pattern (OR = 0.57, 95% CI = 0.34–0.94) had reduced risk of LBW.

We performed sensitivity analyses by expanding SBP records into more than twice between 10 and 40 gestational weeks. The results were similar with the analyses with at least three times of SBP records. (Supplementary Fig. 1 and Supplementary Table 2–4).

Table 4 illustrates whether adding SBP trajectories to a logistic regression model consisting of other confounding factors could improve discriminative ability of individuals at poor pregnant outcomes. SBP trajectories slightly but significantly improved risk discrimination of GH, ETD and LBW, over traditional risk factors (all P values < 005). Specially, the incorporation of SBP trajectories to Model 3, resulted in significantly improved predictive value for GH (c-statistics increased from 0.835 to 0.859, P < 0.0001; NRI = 14.25%, P < 0.0001; IDI = 2.98%, P < 0.0001).