Early Pregnancy Blood Pressure Patterns Identify Risk of Hypertensive Disorders of Pregnancy Among Racial and Ethnic Groups

Abstract

BACKGROUND:

Hypertensive disorders of pregnancy are a leading cause of severe maternal morbidity and mortality and confer 4-fold higher perinatal mortality in Black women. Early pregnancy blood pressure patterns may differentiate risk of hypertensive disorders of pregnancy.

METHODS:

This study identified distinct blood pressure trajectories from 0 to 20 weeks’ gestation to evaluate subsequent pregnancy-related hypertension in a retrospective cohort of 174925 women with no prior hypertension or history of preeclampsia, prenatal care entry ≤14 weeks, and a stillborn or live singleton birth delivered at Kaiser Permanente Northern California hospitals in 2009 to 2019. We used electronic health records to obtain clinical outcomes, covariables, and longitudinal outpatient blood pressure measurements ≤20 weeks’ gestation (mean 4.1 measurements). Latent class trajectory modeling identified 6 blood pressure groups: ultra-low-declining(referent), low-declining, moderate-fast-decline, low-increasing, moderate-stable, and elevated-stable. Multivariable logistic regression evaluated trajectory group-associations with the odds of preeclampsia/eclampsia and gestational hypertension‚ and effect modification by race-ethnicity and prepregnancy body size.

RESULTS:

Compared with ultra-low-declining, adjusted odds ratios (95% confidence intervals [CIs]) for low-increasing, moderate-stable, and elevated-stable groups were 3.25 (2.7–3.9), 5.3 (4.5–6.3), and 9.2 (7.7–11.1) for preeclampsia/eclampsia‚ and 6.4 (4.9–8.3), 13.6 (10.5–17.7), and 30.2 (23.2–39.4) for gestational hypertension. Race/ethnicity, and prepregnancy obesity modified the trajectory-group associations with preeclampsia/eclampsia (interaction P<0.01), with highest risks for Black, then Hispanic and Asian women for all blood pressure trajectories, and with increasing obesity class.

CONCLUSIONS:

Early pregnancy blood pressure patterns revealed racial and ethnic differences in associations with preeclampsia/eclampsia risk within equivalent levels and patterns. These blood pressure patterns may improve individual risk stratification permitting targeted surveillance and early mitigation strategies.

Preeclampsia/eclampsia and gestational hypertension are serious hypertensive disorders of pregnancy (HDP) affecting up to 5% to 6% of all pregnant women and are the leading causes of serious maternalfetal morbidity and mortality worldwide.¹ From 1979 to 2018, the overall hypertension-related maternal mortality rate in the United States was 2.1 per 100000 live births; with Black women having almost 4-fold higher rates than White women (5.4 versus 1.4 per 100000 live births).² Despite decades of research, the pathophysiology of pregnancy-related hypertensive disorders remains poorly understood. Traditional risk factors include maternal prepregnancy overweight or obesity, smoking, Black race, prior chronic hypertension, diabetes, or history of preeclampsia/eclampsia.^3–6 Yet, many pregnant women, particularly nulliparas, do not possess known risk factors and enter pregnancy with normal blood pressure (BP) levels, and therefore, are generally considered low risk. In 2013, the American College of Obstetrics and Gynecology identified the need for early risk stratification of pregnant women and development of new methods for improved prediction of pregnancy-related hypertensive disorders.¹

There has been limited success in previous studies that have examined the risk of HDP associated with BP measurements in early pregnancy. Several studies have assessed the 2017 American College of Cardiology/American Heart Association (ACC/AHA) guidelines, which define BP as normal (systolic <120 mm Hg and diastolic <80 mm Hg), elevated (systolic 120–129 mm Hg and diastolic <80 mm Hg), stage 1 hypertension (systolic 130–139 mm Hg or diastolic 80–89 mm Hg), and stage 2 hypertension (systolic ≥140 mm Hg or diastolic ≥90 mm Hg).⁷ Studies assessing these guidelines found 2.06- to 2.54-fold higher risk of HDP associated with stage 1 hypertension and 1.29- to 1.54-fold higher risk of HDP associated with elevated BP compared with normal BP.^8–11 These modest effect sizes show that using only one or 2 BP measurements in early gestation has a limited ability to predict risk of HDP. Furthermore, most studies using ACC/AHA thresholds have assessed HDP as a combined end point and lack specific risk estimates for preeclampsia/eclampsia and gestational hypertension. Thus, we propose a novel approach to fully characterize longitudinal BP measurements to develop latent class BP trajectory models that capture distinct patterns affecting risk of HDP.

To address these gaps, we selected a diverse sample of 174925 women with a singleton pregnancy, prenatal care entry ≤14 weeks, no prior chronic hypertension, other serious disease, or history of preeclampsia/eclampsia‚ and having outpatient prenatal clinical BP measurements available from the comprehensive electronic health records (EHR) data from 0 to 20 weeks’ gestation within a community-based integrated healthcare system. This study sought to identify distinct early pregnancy longitudinal BP patterns and evaluate their associations with the development of preeclampsia/eclampsia and gestational hypertension after 20 weeks’ gestation, and possible effect modification by race and ethnicity, or prepregnancy body size.

METHODS

AHA Journal TOP Guidelines

Requests to access the dataset from qualified researchers trained in human subjects’ confidentiality protocols may be sent to the Dr Erica P. Gunderson, Principal Investigator, at the Division of Research, erica.gunderson@kp.org. The patient data is owned by the Kaiser Foundation Health Plan, Inc, Kaiser Foundation Hospitals, Inc, and The Permanente Medical Group, Inc. Because of their third-party rights, it is not possible to make the data publicly available without restriction.

Research Setting

The Kaiser Permanente Northern California (KPNC) is an integrated health care delivery system providing care to >4.5 million members through >10000 physicians, >255 medical facilities, and 21 hospitals. The KPNC service area spans 22 counties of the greater Bay Area, as well as the California central valley from Sacramento to Fresno and includes urban and rural areas. The KPNC membership covers 30% of the population in the service area, is socio-demographically diverse, and highly representative of the surrounding and statewide population.¹² There are 16 delivery hospitals and ≈44000 births per year with prenatal care standardized across centers. In the KPNC region, 89.5% of all pregnant women enter prenatal before 14 weeks of gestation.¹³ We used EHR data to obtain clinical outpatient prenatal BP measurements, prepregnancy information, clinical variables, socio-demographic factors, and perinatal outcomes for pregnancies delivered from January 1, 2009, to December 31, 2019. This data-only project was approved by the KPNC Institutional Review Board, which waived the requirement for informed consent from the patients, given the retrospective, data-only, minimal risk study design.

Eligibility Criteria

For each woman, we selected the first index singleton live or stillbirth delivered at a KPNC hospital in 2009 to 2019 that met study eligibility criteria. Among 308775 women with births, we excluded prenatal care entry after 14 weeks’ gestation, KP membership gap ≥4 months or no prenatal care, delivery outside KP hospital, or prior serious medical condition (ie, cancer, kidney disease, cardiovascular disease, etc International Classification of Diseases (ICD)-codes; see Appendix I in the Supplemental Material). This yielded a sample size of 267887 women (Figure S1). Next, we identified and excluded women with chronic hypertension up to 2 years before conception of the index birth with a validated algorithm¹⁴ applied to the EHR system: (1) ICD-9/10 codes for hypertension identified on 2 separate dates (Appendix II in the Supplemental Material), (2) stage 2 BP elevations (systolic ≥140 mm Hg or diastolic ≥90 mm Hg) from outpatient records on 2 separate consecutive occasions at least 3 months apart, or (3) ICD-9/10 codes for hypertension identified one time and filling a prescription for one or more antihypertensive medications (Appendix III in the Supplemental Material). We also excluded women identified with chronic hypertension during pregnancy by using ICD-9/10 codes, BP measurements, and antihypertensive medications (Detailed Text in the Supplemental Material). We excluded all women with diagnosed hypertension or receiving medications for hypertension before conception or during early pregnancy (0–20 weeks). Thus, these criteria identified and excluded 13626 women with prior chronic hypertension or hypertension during the index pregnancy <20 weeks; 5.1% of 267887 women.

We also excluded 47 women missing all BP measurements before 20 weeks’ gestation and 4322 parous women with a history of preeclampsia/eclampsia based on ICD-9/10 codes (see Appendix IV in the Supplemental Material) and our natural language processing algorithm applied to clinician notes for previous KPNC births. The final sample included 249892 pregnant women without prior chronic hypertension, kidney or other serious disease, early pregnancy hypertension, or history of preeclampsia/eclampsia that met all other selection criteria. This study includes N=174925 women randomly selected as 70% of the eligible sample for the development of early BP trajectory groups during the first 20 weeks of gestation. The remaining 30% of the sample will be reserved for future studies assessing internal validation of BP trajectory groups and prediction models.

Primary Outcomes: Classification of preeclampsia/eclampsia and Gestational Hypertension

We classified preeclampsia/eclampsia and gestational hypertension after 20 weeks’ gestation using ICD-9/10 codes in the second and third trimester. For women with both gestational hypertension and preeclampsia/eclampsia ICD-9/10 codes, we chose the diagnosis from an inpatient admission closest to the delivery date. If both gestational hypertension and preeclampsia/eclampsia codes were assigned on the same day, then women were classified with preeclampsia/eclampsia. We conducted a validation study (Table S1) to assess the accuracy of ICD-9/10 codes to classify preeclampsia/eclampsia and gestational hypertension. For preeclampsia/eclampsia‚ both the sensitivity and specificity were 94%. Gestational hypertension had a sensitivity of 85% and specificity of 91%. For any HDP, the sensitivity was 97% and specificity was 85%.

Clinical BP Measurements During 20 Weeks' Gestation

To develop the BP trajectory groups, we obtained BP measurements from outpatient clinical prenatal care visits from 0 to 20 weeks’ gestation from the EHR. We selected one BP measurement per day defined as the last clinical BP measurement from an outpatient prenatal care visit within each 24-hour period after exclusion of invalid systolic BP <70 mm Hg or diastolic BP <40 mm Hg measurements. We used the BP measurements from prenatal outpatient visits obtained by trained medical assistants using automated oscillometers to capture BP under routine conditions rather than variation under other conditions that may involve acute illness. The number of BP measurements for each woman ranged from 1 to 10 measurements on separate days during 0 to 20 weeks’ gestation. These longitudinal BP measurements were used to develop the trajectory groups and to assign each individual to one mutually exclusive group.

Latent Class BP Trajectory Modeling

We first used group-based trajectory modeling to identify distinct early pregnancy BP trajectory groups during the first 20 weeks’ gestation. Group-based trajectory modeling (also called latent class growth modeling) is a statistical approach that combines finite mixture modeling and growth curve modeling into a unified model to identify latent classes of individuals with similar patterns of change over time.¹⁵ Specifically, longitudinal measurements of BP collected at multiple time points during the first 20 weeks’ gestation were assumed to follow a parametric distribution with latent classes where the parameter values that determine the underlying distribution vary by class membership. We fit each trajectory curve with third order polynomial terms to allow for curvature and tested for statistical significance of quadratic and cubic terms to determine the appropriate shape of each trajectory pattern (Enhanced Methods in the Supplemental Material). To determine the optimal number of groups, we fit several models ranging from 3 to 6 latent classes and chose the best model based on the Bayes factor, which compares the Bayesian Information Criterion for each model.¹⁶ Model parameters were estimated by maximum likelihood. Each woman was then assigned to the trajectory group that best reflected her profile of change using a maximum probability assignment rule for the posterior probabilities for group membership obtained from the model.¹⁵ The trajectory group modeling was conducted in SAS using PROC TRAJ.¹⁷

Covariables

We obtained clinical, sociodemographic, and lifestyle covariables from the KPNC outpatient and inpatient EHR, including maternal age, self-reported race and ethnicity, height, parity for the index pregnancy, measured prepregnancy weight within 1 year before conception or weight ≤14 weeks’ gestation, smoking status during pregnancy, mode of delivery, gestational age, last weight measured before delivery, diabetes status (none, pregestational, or gestational diabetes), and neonatal outcomes (live or stillbirth, birth weight, size-for-gestational-age,¹⁸ admission to the neonatal intensive care unit, and neonatal and perinatal morbidity and mortality).

Race and ethnicity were self-identified by patients to their health care providers as follows: African American or Black (hereafter Black), Asian, American Indian or Alaskan Native, Native Hawaiian or Other Pacific Islander, Hispanic or Latino (hereafter Hispanic), or White, based on the US Office of Management and Budget’s Revisions to the Standards for the Classification of Federal Data on Race and Ethnicity (2007). Women who self-identified as belonging to multiple groups, did not identify race or ethnicity (unknown), or groups of small size (American Indian, Alaskan native, native Hawaiian, or other Pacific Islander), were combined as other /unknown. For this analysis, the racial and ethnic groups were categorized as Asian, Black, Hispanic, White, and other/unknown. These specific groups represent the constellation of social and other factors that make up the lived experience of individuals within these groups in the United States. These categories reflect not only geographic ancestry but also social characteristics including shared history, language, beliefs, and customs and have been linked with social determinants of health (education, socio-economic disadvantage, and discrimination).¹⁹ Jointly, these factors may contribute to the disparities in observed racial and ethnic health outcomes.²⁰ These groups have broad importance in health research for not only identifying, but monitoring, understanding and intervening on health inequities.²¹

Maternal mortality was obtained from the US Centers for Disease Control and Prevention data files on mortality with linkage to the social security records. We calculated body mass index (BMI) in kg/m² by dividing prepregnancy or early first trimester weight (kg) by height in meters squared, and classified BMI categories based on standard definitions,²² and total gestational weight gain as the difference between the last measured weight before delivery and prepregnancy or <14 weeks’ gestation weight. The neighborhood deprivation index was derived using zip codes for each Census tract in the United States and included in the Census Bureau’s American Community Survey data (https://www.census.gov/programs-surveys/acs/,10/2021).

Statistical Methods

Maternal and infant characteristics and outcomes were described among the HDP outcome groups (preeclampsia/eclampsia, gestational hypertension, or no HDP), and the 6 BP trajectory groups using χ² statistics for categorical variables, ANOVA methods for continuous variables, and the Kruskal-Wallis test of the equality of medians for variables with skewed distributions. Potential confounders (prepregnancy BMI, parity, age, race and ethnicity, smoking habit, and diabetes status) were evaluated based on a priori hypotheses and statistical significance (P<0.05) for the associations with the BP trajectory groups and HDP outcomes. All P values are for 2-sided tests with statistical significance P<0.05. All analyses used Statistical Analysis Software for Windows 9.4 (SAS Institute Inc, Cary, NC).

We estimated overall percentages of women with preeclampsia/eclampsia and gestational hypertension separately within each BP trajectory group and evaluated percentages among prepregnancy BMI and race and ethnicity groups for each BP trajectory group. We modeled each outcome in separate logistic regression models, with and without adjustment for established risk factors including clinical variables (parity, prepregnancy BMI, diabetes status [pregestational diabetes, gestational diabetes, none], lifestyle behaviors: smoking status [never, former, current, or unknown], and socio-demographic characteristics (age, racial and ethnic groups, and neighborhood deprivation index [quartiles]), as previously described for KPNC.^23,24 Since both preeclampsia/eclampsia and gestational hypertension outcomes were rare (<10% incidence), the odds ratio can be interpreted as an estimate of the relative risk.²⁵

For the main effect of BP trajectory group on the risk of each outcome, we reported the adjusted odds ratio (aOR) and 95% CI of preeclampsia/eclampsia, and gestational hypertension versus no HDP in separate models for the pairwise contrast of each BP trajectory group relative to the ultra-low-declining group. We then examined potential effect modification by prepregnancy BMI categories (after exclusion of underweight category), and by race and ethnicity (after exclusion of other/unknown). Following Knol and VanderWeele’s recommendations for presenting analyses of effect modification and interaction, we reported the primary results as the aORs and 95% CIs for the pairwise contrast of each BP trajectory group relative to a single reference group.²⁶ We used interaction terms to test for statistical significance on the multiplicative scale. We also fit stratified models to estimate the BP trajectory group associations within each BMI strata and within each racial and ethnic group. We also conducted sensitivity analyses in which models included the initial BP as a separate covariate in each model to estimate the aORs for the 6 BP trajectory groups.

RESULTS

Among 174925 eligible women with a singleton live or still birth from 2009 to 2019 and no prior hypertension or history of preeclampsia/eclampsia (Figure S1), 169,528 women (96.9%) had one or more BP measurements through 12 weeks’ gestation, and 99804 (57.1%) had 4 or more BP measurements obtained on different days from 0 to 20 weeks’ gestation. We identified 8342 (4.8%) and 8767 (5.0%) women who were diagnosed with preeclampsia/eclampsia and gestational hypertension after 20 weeks’ gestation, respectively (Table 1). Compared with women who did not develop preeclampsia/eclampsia or gestational hypertension (no HDP), women who developed preeclampsia/eclampsia were more likely to be younger, Black race or Hispanic ethnicity, nulliparous, overweight or obese, and had higher gestational weight gain, perinatal mortality, preterm delivery, small-for-gestational-age infants, admission to the neonatal intensive care unit, and to be undergoing C-section delivery, and had shorter gestations and lower mean birthweight (Table 1). The preeclampsia/eclampsia and no HDP groups had higher stillbirth rates than the gestational hypertension group (P<0.001). Women who developed gestational hypertension were slightly older, White or Black race, current or former smokers, had prepregnancy obesity, C-section delivery, and higher gestational weight gain versus no HDP.

Table 1.

Maternal and Newborn Characteristics and Perinatal Outcomes Associated with Preeclampsia/Eclampsia and Gestational Hypertension; (N=174 925 Women With a Singleton Live or Still Birth)

Characteristics	All, N=174 925	Preeclampsia/eclampsia, N=8342	Gestational hypertension, N=8767	No HDP (referent), N=157 816	P value
Maternal age, y, mean (SD)	30.9 (5.3)	30.6 (5.8)	31.0 (5.4)	30.9 (5.3)	<0.001
Age categories, n (%)					<0.001
18–25y	32 163 (18.4)	1860 (22.3)	1627 (18.6)	28 676 (18.2)
26–30y	54 617 (31.2)	2427 (29.1)	2728 (31.1)	49 462 (31.3)
31–35y	58 060 (33.2)	2526 (30.3)	2823 (32.2)	52 711 (33.4)
36–40y	25 541 (14.6)	1257 (15.1)	1307 (14.9)	22 977 (14.6)
41–45y	4544 (2.6)	272 (3.3)	282 (3.2)	3990 (2.5)
Race and ethnicity, n (%)					<0.001
Asian	44 216 (25.3)	1824 (21.9)	1672 (19.1)	40 720 (25.8)
Black	12 708 (73)	825 (9.9)	740 (8.4)	11 143 (71)
Hispanic	46 289 (26.5)	2413 (28.9)	1927 (22.0)	41 949 (26.6)
Other/unknown	7785 (4.5)	365 (4.4)	388 (4.4)	7032 (4.5)
White	63 927 (36.6)	2915 (34.9)	4040 (46.1)	56 972 (36.1)
Smoking status, n (%)					<0.001
Current	9918 (5.7)	510 (6.1)	585 (6.7)	8823 (5.6)
Former	21 300 (12.2)	1104 (13.2)	1313 (15.0)	18 883 (12.0)
Never	142 836 (81.7)	6680 (80.1)	6821 (77.8)	129 335 (82.0)
Missing	871 (0.5)	48 (0.6)	48 (0.6)	775 (0.5)
Prenatal parity, n (%)					<0.001
Nulliparous	99 972 (572)	6542 (78.4)	6495 (74.1)	86 935 (55.1)
Primiparous	47 215 (270)	1130 (13.5)	1446 (16.5)	44 639 (28.3)
Biparous	18 947 (10.8)	451 (5.4)	566 (6.5)	17 930 (11.4)
Multiparous	8791 (5.0)	219 (2.6)	260 (3.0)	8312 (5.3)
Neighborhood deprivation index, n (%)					<0.001
<-1 (least deprived)	20 135 (11.5)	848 (10.2)	997 (11.4)	18 290 (11.6)
>-1 and <0	87 984 (50.4)	4130 (49.6)	4502 (51.4)	79 352 (50.4)
>0 and <1	45 318 (25.9)	2255 (271)	2257 (25.8)	40 806 (25.9)
>1 (most deprived)	21 214 (12.1)	1094 (13.1)	1000 (11.4)	19 120 (12.1)
Diabetes status, n (%)					<0.001
Gestational diabetes	19 882 (11.4)	1270 (15.2)	1218 (13.9)	17 394 (11.0)
Pregestational diabetes	1078 (0.6)	165 (2.0)	87 (1.0)	826 (0.5)
None	153 965 (88.0)	6907 (82.8)	7462 (85.1)	139 596 (88.5)
Prepregnancy BMI Categories, n (%)					<0.001
Underweight (<18.5)	4666 (2.7)	153 (1.8)	90 (1.0)	4423 (2.8)
Normal weight (18.5–24.9)	86 172 (49.6)	3231 (38.9)	3013 (34.5)	79 928 (51.0)
Overweight (25–29.9)	47 046 (27.1)	2343 (28.2)	2495 (28.6)	42 208 (26.9)
Obesity class I (30–34.9)	21 515 (12.4)	1393 (16.8)	1639 (18.8)	18 483 (11.8)
Obesity class II (35–39.9)	9020 (5.2)	713 (8.6)	829 (9.5)	7478 (4.8)
Obesity class III (≥40)	5248 (3.0)	465 (5.6)	655 (7.5)	4128 (2.6)
Prepregnancy weight, kg, mean (SD)	68.9 (16.6)	73.0 (18.8)	77.1 (20.0)	68.2 (16.1)	<0.001
Height, cm, mean (SD)	162.5 (7.0)	161.9 (7.0)	163.6 (7.2)	162.5 (7.0)	<0.001
Prepregnancy BMI, kg/m2, mean (SD)	26.0 (5.8)	27.8 (6.6)	28.7 (6.9)	25.8 (5.6)	<0.001
Total gestational weight gain, kg, mean (SD)	13.5 (6.5)	15.0 (7.3)	14.7 (7.4)	13.4 (6.3)	<0.001
Gestational age, wk, first prenatal care, mean (SD)	8.2 (2.0)	8.0 (2.0)	8.0 (1.9)	8.2 (2.0)	<0.001
No. of BP measurements up to 20 wk GA, mean (SD)	4.07 (1.73)	4.25 (1.84)	4.14 (1.76)	4.06 (1.73)	<0.001
Delivery gestational age, wk, mean (SD)	39.2 (2.0)	38.3 (2.5)	39.3 (1.7)	39.3 (2.0)	<0.001
Mode of delivery, n (%)*					<0.001
Normal spontaneous	118 064 (68.1)	4418 (53.5)	5591 (64.1)	108 055 (69.0)
Assisted vaginal delivery	10 451 (6.0)	597 (7.2)	575 (6.6)	9279 (5.9)
C-section	44 900 (25.9)	3246 (39.3)	2554 (29.3)	39 100 (25.0)
Missing	92 (0.1)	5 (0.0)	2 (0.02)	85 (0.1)
Maternal mortality, n (%)†
Within 42 days postdelivery	13 (0.01)	2 (0.02)	0 (0.00)	11 (0.01)	0.19
Within 1 y postdelivery, by race or ethnicity	30 (0.02)	3 (0.04)	2 (0.02)	25 (0.02)	0.18
Asian	10	2	0	8
Black	6	1	0	5
Hispanic	7	0	0	7
Other/unknown	0	0	0	0
White	7	0	2	5
All birth outcomes, n (%)					<0.001
Still births	890 (0.5)	44 (0.5)	20 (0.2)	826 (0.5)
Live births	174 035 (99.5)	8298 (99.5)	8747 (99.8)	156 990 (99.5)
Gestational age, n (%)					<0.001
Early preterm <34 wk	2954 (1.7)	479 (5.7)	88 (1.0)	2387 (1.5)
Late preterm 34–36 6/7 wk	8438 (4.8)	1148 (13.8)	392 (4.5)	6898 (4.4)
37 to 40 wk	105 788 (60.5)	4822 (57.8)	5344 (61.0)	95 622 (60.6)
>40 wk	57 745 (33.0)	1893 (22.7)	2943 (33.6)	52 909 (33.5)
Birthweight, g, mean (SD)	3360 (540)	3074 (733)	3347 (550)	3376 (523)	<0.001
Live birth outcomes
Size-for-GA, n (%)					<0.001
Small (SGA)	4594 (2.6)	675 (8.1)	331 (3.8)	3588 (2.3)
Appropriate (AGA)	139 286 (80.1)	6413 (77.3)	6846 (78.3)	126 027 (80.3)
Large (LGA)	30 092 (173)	1210 (14.6)	1570 (18.0)	27 312 (174)
Newborn nursery, n (%)					<0.001
Normal newborn	154 884 (89.7)	6276 (76.5)	7643 (88.2)	140 965 (90.5)
Intermediate level	9899 (5.7)	811 (9.9)	523 (6.0)	8565 (5.5)
NICU	7801 (4.5)	1118 (13.6)	501 (5.8)	6182 (4.0)
Postdelivery death	144 (0.1)	2 (0.0)	0 (0.0)	142 (0.1)
Neonatal mortality (live births age up to 42 days)	296 (0.17)	15 (0.18)	5 (0.06)	276 (0.18)	0.03
Stillbirth by GA, n (%)†					0.02
Early preterm <34 wk	650 (73.0)	30 (68.2)	8 (40.0)	612 (74.1)
Late preterm 34 to 36 6/7 wk	69 (7.8)	3 (6.8)	3 (15.0)	63 (7.6)
37 to 40 wk	114 (12.8)	8 (18.2)	7 (35.0)	99 (12.0)
>40 wk	57 (6.4)	3 (6.8)	2 (10.0)	52 (6.3)
Total	890 (100)	44 (100)	20 (100)	826 (100)
Perinatal mortality, n (%)	1186 (0.7)	59 (0.7)	25 (0.3)	1102 (0.7)	<0.001
By gestational age					<0.001
Early preterm <34 wk	2954 (1.7)	479 (5.7)	88 (1.0)	2387 (1.5)
Late preterm 34 to 36 6/7 wk	8438 (4.8)	1148 (13.8)	392 (4.5)	6898 (4.4)
37 to 40 wk	105 788 (60.5)	4822 (57.8)	5344 (61.0)	95 622 (60.6)
>40 wk	57 745 (33.0)	1893 (22.7)	2943 (33.6)	52 909 (33.5)

Perinatal mortality=fetal+neonatal deaths. AGA indicates average-for-gestational age; BMI, body mass index; C-section, cesarean section delivery; GA, gestational age; HDP, hypertensive disorders of pregnancy; LGA, large-for-gestational age; NICU, neonatal intensive care unit; and SGA, small-for-gestational age.

^*Mode of delivery for live births only

^†Fisher exact test.

Development of BP Trajectory Groups

Trajectory modeling identified 6 distinct BP trajectory groups: ultra-low-declining, low-declining, low-increasing, moderate-fast-decline, moderate-stable, and elevated-stable (Figure 1). The initial mean systolic BP and changes up to 20 weeks’ gestation were as follows: ultralow-declining of 100.0 mm Hg decreased by 6 mm Hg; low-declining of 107.4 mm Hg decreased by 5.2 mm Hg; moderate-fast-declining of 120.9 mm Hg decreased by 11.9 mm Hg; low-increasing of 109.8 mm Hg increased by 2.2 mm Hg, moderate-stable of 120.4 mm Hg decreased by 2.6 mm Hg; and elevated-stable of 128.7 mm Hg decreased by 2.8 mm Hg. The mean (95% CI) for the first and the last systolic BPs are shown for each BP trajectory group in the Materials-Detailed Methods in the Supplemental Material (SPage 3). We selected the 6-group trajectory model as the final model because it showed substantially better model fit than the models with only 3, 4, or 5 groups; specifically, the Bayes factor was >1000 when comparing the 6-group model to each of the other models, whereas a Bayes factor >10 suggests very strong evidence in favor of the 6-group model. The overall mean and interquartile ranges for systolic BP and diastolic BP levels are shown by trajectory groups (Table S2).

Figure 1. Six early pregnancy systolic blood pressure trajectory groups from 0 to 20 weeks’ gestation.

Six early pregnancy blood pressure trajectory groups: ultra-low-declining (–light blue), low-declining (–green), moderate-fast-decline (–red), low-increasing (–yellow), moderate-stable (–black), and elevated-stable (–purple).

Association of BP Trajectory Groups and Covariables With preeclampsia/eclampsia and gestational hypertension

Black, Hispanic, and Asian women had higher prevalence (%) of preeclampsia/eclampsia among 4 BP trajectory groups (elevated-stable, moderate-stable, low-increasing, and moderate-fast-decline) compared with White women (Figure 2). The percentages of women who developed preeclampsia/eclampsia and gestational hypertension increased from the ultra-low-declining (referent) through the highest BP trajectory group, elevated-stable (Table 2 and Table S3). For preeclampsia/eclampsia, 6127 women (73.4% of all preeclampsia/eclampsia outcomes) were found in the 3 highest BP trajectory groups (elevated-stable, moderate-stable, and low-increasing) and 4711 (56.5%) of preeclampsia/eclampsia outcomes were within the 2 next highest BP trajectory groups (moderate-stable and low-increasing) which were found among 50.7% and 43.7% of the total sample, respectively. For gestational hypertension, there were 7200 women (82.1% of all gestational hypertension outcomes) classified within the 3 highest BP trajectory groups, and of these, 5954 women (67.9%) overall were within the 2 highest trajectory groups (elevated-stable and moderate-stable). Higher BP and the stable trajectory groups also had higher percentages of Black women and prepregnancy obesity (Tables S4 and S5).

Figure 2.

Percentage (%) of women who developed preeclampsia/eclampsia by race and ethnicity groups among the 6 early pregnancy systolic blood pressure trajectory groups.

Table 2.

Additive Models for Early Pregnancy (≤20 weeks’ Gestation) Systolic BP Trajectory Groups and Hypertensive Disorders of Pregnancy; aOR (95% CI) of Preeclampsia/Eclampsia, and of Gestational Hypertension vs No HDP (N=174 925 Women With a Singleton Live or Still Birth)

Early BP trajectory groups—main effect	Preeclampsia/eclampsia, N (%)	aOR (95% CI) of preeclampsia/eclampsia	Gestational hypertension, N (%)	aOR (95% CI) of gestational hypertension
Elevated-stable	1416 (11.7)	9.2 (7.7–11.1)	2135 (17.7)	30.2 (23.2–39.4)
Moderate-stable	3275 (7.3)	5.3 (4.5–6.3)	3819 (8.5)	13.6 (10.5–17.7)
Low-increasing	1436 (4.5)	3.3 (2.7–3.9)	1246 (3.9)	6.4 (4.9–8.3)
Moderate-fast-decline	735 (3.8)	2.7 (2.3–3.3)	652 (3.3)	5.4 (4.1–7.1)
Low-declining	1343 (2.4)	1.8 (1.5–2.2)	857 (1.5)	2.6 (2.0–3.5)
Ultra-low-declining	137 (1.3)	1.0 (Referent)	58 (0.53)	1.0 (Referent)
Covariable adjustment
Parity
Nulliparous (0 prior births)		3.16 (2.99–3.34)		2.32 (2.20–2.45)
Parous (1 or more prior births)		1.0 (Referent)		1.0 (Referent)
Race and ethnicity
Asian		1.15 (1.08–1.22)		0.90 (0.84–0.95)
Black		1.47 (1.35–1.60)		0.89 (0.82–0.97)
Hispanic		1.35 (1.27–1.43)		0.78 (0.73–0.83)
Other/unknown		1.14 (1.02–1.28)		0.92 (0.82–1.03)
White		1.0 (Referent)		1.0 (Referent)
Age, y		1.02 (1.01–1.02)		1.02 (1.02–1.03)
Prepregnancy BMI, kg/m2		1.02 (1.02–1.03)		1.04 (1.03–1.04)
Diabetes status
Pregestational		2.95 (2.47–3.52)		1.27 (1.00–1.60)
Gestational diabetes		1.29 (1.20–1.38)		1.01 (0.95–1.08)
None		1.0 (Referent)		1.0 (Referent)
Smoking (cigarette) habit
Current		1.05 (0.95–1.16)		1.11 (1.02–1.22)
Former		1.06 (0.99–1.14)		1.14 (1.07–1.22)
Unknown		1.07 (0.79–1.45)		1.13 (0.83–1.53)
Never		1.0 (Referent)		1.0 (Referent)
Neighborhood deprivation index
≤−1 (least deprived)		1.0 (Referent)		1.0 (Referent)
>−1 and ≤0		1.05 (0.97–1.13)		0.95 (0.88–1.02)
>0 and ≤1		1.06 (0.97–1.15)		0.92 (0.84–0.99)
>1 (most deprived)		1.10 (1.00–1.21)		0.94 (0.85–1.03)

All aORs (95%CIs) from multivariable models adjusted for all covariates. aOR indicates adjusted odds ratio; BMI, body mass index; BP blood pressure; and HDP hypertensive disorders of pregnancy.

In multivariable logistic regression additive models adjusted for key clinical and socio-demographic covariables, the aORs showed a gradient of increasing risk of gestational hypertension and preeclampsia/eclampsia from the low-declining group up through the elevated-stable group compared with the ultra-low-declining group (referent; Table 2). For preeclampsia/eclampsia, compared with the ultra-low-declining group, the aORs (95% CI) ranged from 1.82 (1.53–2.18) for the low-declining group up to 9.24 (7.70–11.09) for the elevated-stable group among the consecutive trajectory groups. Model covariables independently associated with higher odds of preeclampsia/eclampsia (aORs [95% CI]) included nulliparity (3.16 [2.99–3.34]), older maternal age, Black race (1.47 [1.35–1.60]), Hispanic ethnicity (1.35 [1.27–1.43]), and Asian (1.15 [1.08–1.22]) compared with White race, increasing prepregnancy BMI (1.02 [1.02–1.03]), and pregestational diabetes (2.95 [2.47–3.52]) and gestational diabetes (1.29 [1.20–1.38]) compared with none (Table 2). For gestational hypertension, compared with the ultra-low-declining group, aORs (95% CI) ranged from 2.6 (2.0–3.5) for the low-declining group up to 30.2 (23.2–39.4) for the elevated-stable group.

Effect Modification of BP Trajectory Groups by Race and Ethnicity and BMI Groups

Race-ethnicity was an effect modifier of the BP trajectory-association with preeclampsia/eclampsia (P<0.001) but not for gestational hypertension (P=0.94). Black women had the highest relative odds of preeclampsia/eclampsia followed in order by Hispanic and Asian women compared with White women in the ultra-low-declining group, for most BP trajectory groups (Table 3). The 3 trajectory groups (elevated-stable, moderate-stable, and low-increasing) comprised 663 (80.4%) of the preeclampsia/eclampsia outcomes among 6823 Black women. Prepregnancy BMI was an effect modifier of the BP trajectory-associations with preeclampsia/eclampsia (P=0.007), but not for gestational hypertension (P=0.052) (Table 3). Obesity classes had higher prevalence of preeclampsia/eclampsia for the 3 highest trajectory groups (elevated-stable, moderate-stable, and low-increasing) than overweight or normal BMI groups (Figure 3). Overall, race-ethnicity effect modification of the trajectory-association with preeclampsia/eclampsia was stronger than prepregnancy BMI. For gestational hypertension, percentages for White women (19.2%) were highest followed by Asians, Hispanics, and Black women (16.1%) for the elevated-stable group (Figure S2). Overall, the prevalence (%) of women with gestational hypertension within the 6 distinct trajectory groups, and aORs of gestational hypertension did not differ significantly by race-ethnicity and prepregnancy BMI (Table S6 and Figure S3). We also present the results of the BP trajectory group associations within each BMI strata (low-declining trajectory as referent) and within each race-ethnicity group (ultra-low-declining trajectory as referent) for the aORs of preeclampsia/eclampsia versus no HDP (Table S7). Sensitivity analyses in which models included the initial systolic BP measurement as a separate covariate in each model demonstrated that the BP trajectory groups remained strongly associated with development of preeclampsia/eclampsia and gestational hypertension (Table S8).

Table 3.

Interaction Models for Racial and Ethnic Groups and Prepregnancy BMI Categories for Early Pregnancy Systolic BP Trajectory Groups and aORs (95% CIs) of Preeclampsia/Eclampsia vs No HDP; Race and Ethnicity Group-Interaction Model (White and Ultra-Low-Declining Group as the Referent) and Prepregnancy BMI Group-Interaction Model (Normal BMI and Ultra-Low-Declining Group as the Referent)

Early pregnancy systolic BP trajectory groups	Race and ethnicity groups
	White, N=63 927	Asian, N=44 216	Hispanic, N=46 289	Black, N=12 708
aOR (95% CI) of preeclampsia/eclampsia
Elevated-stable	6.0 (4.3–8.4)	7.4 (5.2–10.5)	8.4 (6.0–11.8)	9.1 (6.4–13.1)
Moderate-stable	3.6 (2.6–5.0)	4.5 (3.2–6.2)	4.5 (3.2–6.2)	5.0 (3.5–7.0)
Low-increasing	2.3 (1.6–3.2)	2.5 (1.8–3.5)	2.8 (2.0–4.0)	3.1 (2.1–4.4)
Moderate-fast-decline	1.7 (1.2–2.3)	2.3 (1.6–3.2)	2.6 (1.8–3.7)	2.8 (1.9–4.3)
Low-declining	1.2 (0.9–1.7)	1.2 (0.9–1.7)	1.9 (1.3–2.6)	2.2 (1.5–3.2)
Ultra-low-declining	1.0 (Referent)	0.6 (0.4–0.9)	1.0 (0.7–1.6)	1.2 (0.5–2.7)
Early pregnancy systolic BP trajectory groups	Prepregnancy BMI categories
	Normal, N=86 172	Overweight, N=47 046	Obesity class I, N=21 515	Obesity class II, N=9020	Obesity class III, N=5248
aOR (95% CI) of preeclampsia/eclampsia
Elevated-stable	10.1 (8.0–12.8)	10.3 (8.2–13.0)	13.6 (10.8–17.2)	14.9 (11.5–19.1)	14.0 (10.8–18.2)
Moderate-stable	5.1 (4.1–6.2)	6.5 (5.3–8.0)	7.5 (6.0–9.3)	8.5 (6.7–10.7)	9.2 (7.2–11.8)
Low-increasing	3.6 (2.9–4.4)	3.6 (2.9–4.5)	4.5 (3.5–5.7)	4.5 (3.3–6.1)	4.6 (3.1–6.7)
Moderate-fast-decline	3.1 (2.5–3.9)	2.7 (2.1–3.4)	3.3 (2.5–4.4)	4.9 (3.4–7.0)	5.6 (3.5–8.9)
Low-declining	1.9 (1.6–2.4)	2.2 (1.8–2.8)	2.0 (1.5–2.7)	3.3 (2.2–4.8)	2.3 (1.2–4.5)
Ultra-low-declining	1.0 (Referent)	1.4 (0.8–2.2)

Models are adjusted for parity (nulliparous vs primi- or multiparous), maternal age (continuous), prepregnancy BMI (continuous), diabetes status (3 groups: pregestational, gestational or none), smoking habit, and Neighborhood Deprivation Index (quartiles). Race and ethnicity group-interaction overall P values: preeclampsia/eclampsia <0.001, race and ethnicity group interaction models exclude the other/unknown group (N=7785), Prepregnancy BMI category-interaction overall P values: preeclampsia/eclampsia=0.007, prepregnancy BMI interaction models exclude underweight women (N=4666). Ultra-low-declining blood pressure trajectory group aORs cannot be estimated for obesity class I, II or III because there are 0 cases of preeclampsia/eclampsia for these subgroups. Estimates are not displayed for subgroups with fewer than 5 cases due to small cell sizes. aOR indicates adjusted odds ratio; BP, blood pressure; BMI, body mass index; and HDP, hypertensive disorders of pregnancy.

Figure 3.

Percentage (%) of women who developed preeclampsia/eclampsia by prepregnancy body mass index (BMI) categories among the 6 early pregnancy systolic blood pressure trajectory groups.

DISCUSSION

In our study, we demonstrate that early pregnancy BP trajectories, an inexpensive and readily available measure, may have clinical implications to improve early risk stratification of women for development of preeclampsia/eclampsia and gestational hypertension. The pattern of increasing BP among subgroups with low normal BP levels, or elevated early pregnancy BP levels revealed higher risk of the HDP, while declining BP patterns within both elevated or normal BP ranges revealed much lower risk, reflecting the expected physiological decline by mid-gestation for healthy pregnancies. The prognostic value can be further increased when race and ethnicity and prepregnancy BMI groups are additionally considered in addition to early pregnancy BP trajectories within currently acceptable ranges. The novel findings from this study are the development of 6 distinct longitudinal early pregnancy BP patterns using latent class trajectory modeling. This statistical method uses all routine prenatal outpatient BP measurements during the first 20 weeks’ gestation to differentiate subsequent risk of preeclampsia/eclampsia and gestational hypertension especially for a high proportion of women with normal systolic BP levels (100 to <120 mm Hg) during early gestation. Our findings demonstrate that BP patterns considered normal (<120 mm Hg) that rapidly increase or show minimal or no decline during the first half of pregnancy reflect substantially higher relative risks of developing preeclampsia/eclampsia and gestational hypertension later in pregnancy, compared with associations identified using the initial BP levels. Consistent with our findings, the Avon Longitudinal Study of Parents and Children (n=13016) found that higher initial systolic BPs (>120 mm Hg or ≈120 mm Hg) in early pregnancy and smaller declines up to 18 weeks’ gestation were directly associated with risks of gestational hypertension and preeclampsia/eclampsia.²⁷ Our findings delineate differences in risk of HDP within BP levels below 120 mm Hg and the importance of marked patterns of decline and increase in BP through 20 weeks’ gestation in women without recognized clinical risk factors.

Second, this study is among the first to evaluate these differences among racial and ethnic groups within equivalent BP levels and change patterns. Black women had the highest relative risk of preeclampsia/eclampsia (compared with White women) among racial and ethnic groups within equivalent BP levels and similar patterns, and especially for subgroups with normal BP trajectories, or elevated BP per ACC/AHA criteria. Compared with White women in the ultra-low-declining BP trajectory group, the 5 highest trajectory groups showed highest risk of preeclampsia/eclampsia among Black women, followed by, Hispanic and Asian women. There was a graded increase in the relative risk of preeclampsia/eclampsia within all 6 BP trajectory groups for prepregnancy BMI categories and obesity classes I to III. The BP trajectory-associations with gestational hypertension increased with prepregnancy obesity class, but did not vary among racial and ethnic groups. Three early pregnancy BP trajectory groups (ie, elevated-stable, moderate-stable, and low-increasing) contained 73.4% and 82.1% of preeclampsia/eclampsia and gestational hypertension outcomes, respectively. By contrast, women classified among the ultra-low-declining, low-declining, or moderate-fast-decline trajectory groups comprised 26.6% and 17.9% of overall preeclampsia/eclampsia and gestational hypertension outcomes, respectively. The elevated-stable and moderate-stable BP trajectory groups differed in the risk of preeclampsia/eclampsia by race and ethnicity, with higher absolute prevalence among Black (22%) and Hispanic (21%) versus White women (16.6%), but lower prevalence of gestational hypertension in Black (23%) versus White women (29%).

Pregnancy-related hypertensive disorders complicate up to 10% of all pregnancies and are a leading cause of severe maternal morbidity, and perinatal morbidity and mortality,^1,2,28 that disproportionately affect minority women.²⁹ Preeclampsia/eclampsia is a leading cause of maternal morality worldwide.³⁰ In the United States, Black women have a 4-fold higher risk of mortality from pregnancy-related hypertension compared with White women.² Higher risk of adverse pregnancy outcomes among certain racial and ethnic groups may arise from the complex interactions of social and environmental factors (eg, discrimination, customs, economic disadvantage, barriers to healthcare access), epigenetics, and other influences on health outcomes that are poorly understood. Geronimus’ weathering hypothesis is apropos in theorizing that socially structured, repeated stress process activation can accumulate and increase disease vulnerability across the life course in marginalized groups.³¹ The disparities in adverse perinatal and maternal outcomes for US Black women,^29,32–35 including preterm birth,³⁶ preeclampsia/eclampsia,^37,38 and serious maternal morbidity and mortality have been well-documented.^2,30,39 The Black-White outcome gap has persisted, even after accounting for lifestyle behaviors and socioeconomic position. Some evidence indicates that racial discrimination, as a psychosocial stressor, increases the risk of preterm and low birth weight. Mustillo et al³² found that the Black-White difference in these perinatal outcomes was partially attributed to self-reported experiences of racial discrimination. Thus, the health disparities are hypothesized to be determined in part by social determinants of health rather than necessarily inherent biologic differences. Our findings add to this body of evidence in that early BP patterns are independently associated with higher risk of preeclampsia/eclampsia and gestational hypertension in a graded manner among all racial and ethnic groups. Yet, the direct associations for preeclampsia/eclampsia with the elevated- or moderate- stable BP patterns were strongest for Black women. The implications for health care systems include identification of actionable areas to mitigate the disparities.

HDP has increased by 25% to 50% within the past 2 decades,⁴⁰ possibly related to secular trends for older age at first birth, and increasing obesity, diabetes, and chronic hypertension among US women of childbearing age.^41,42 Hypertensive disorders are the most common reasons for puerperal admissions to the intensive care unit.⁴³ Serious adverse maternal outcomes related to preeclampsia/eclampsia include hemorrhage, placental abruption, pulmonary edema, kidney or liver failure, stroke and in rare cases, death.^44–47 Perinatal morbidity and mortality from preeclampsia/eclampsia is primarily due to still birth, fetal growth restriction, and early preterm birth (frequently iatrogenic as management).^40,48–50

Unfortunately, the mainstay of management remains delivery to begin to reverse the acute pathophysiology of pregnancy-related hypertension. Due to limited strategies for prevention, the American College of Obstetrics and Gynecology Task Force¹ has emphasized the need for clinical research to develop early risk stratification for hypertensive disorders that would afford sufficient time for interventions to improve perinatal outcomes. Currently, low dose aspirin appears to be only minimally effective and is recommended for pregnant women with moderate risk or greater (ie, stage 2 chronic hypertension, history of preeclampsia/eclampsia).^51–53 Universal use is not recommended by most experts because of uncertainty about their effects on long-term infant outcomes.⁵³ A current clinical protocol incorporating biomarkers for risk prediction seems to identify women at increased risk for early onset preeclampsia who may receive subsequent benefit from low dose aspirin.⁵⁴ Enthusiasm for this approach is tempered by its complexity and the fact that it identifies only a small proportion (early onset) of women with subsequent preeclampsia/eclampsia.

To address this challenge, our findings suggest that early pregnancy BP trajectories may be a beneficial and pragmatic approach to identifying women at higher risk for preeclampsia/eclampsia and gestational hypertension. Women with elevated-stable, moderate-stable, and low-increasing longitudinal systolic BP patterns, especially the highest risk subgroups, such as Black and Hispanic women and women with obesity, may benefit from preventative strategies. Our latent class BP trajectory groups may also improve risk stratification for gestational hypertension. However, another important consideration is that a history of preeclampsia/eclampsia is associated with increased risk of later life cardiovascular disease,⁵⁵ although this excess risk associated with gestational hypertension is less clear. Thus, longitudinal studies across the reproductive life-course are needed to evaluate the longer-term impact of BP patterns from before and throughout pregnancy, as well as clinically defined hypertensive disorders (eg, stage 1 and stage 2 hypertension) and preclinical BP defined by ACC/AHA as elevated, and their longer-term relationship to future risk of cardiovascular disease outcomes in women during mid to late life.

Comparison to Previous Literature

One previous study of 300000 live births in Southern California Kaiser Permanente classified 23% of the sample with stage 1 hypertension based on a single BP elevation (systolic BP of 130 mm Hg or higher, diastolic BP of 80 mm Hg or higher, or both) before 20 weeks’ gestation. This single elevation was associated with 2-fold higher odds of any HDP compared with being normotensive. Our risk estimates based on multiple BPs differentiated and identified 6- to 9-fold higher relative risks for women by identifying a very low risk BP group among normotensive women. Further, this previous large study did not characterize BP patterns during early pregnancy, nor risk differences individually for preeclampsia/eclampsia and gestational hypertension outcomes, or effect modification by race and ethnicity. The inclusion of >1 birth per woman potentially biases risk estimates with higher parity. Although maternal obesity is associated strongly with risk of HDP,⁸ maternal BMI has not been evaluated in prior studies as an effect modifier. Finally, longitudinal BP patterns and risk differences within the full range of normotensive BP levels were not previously studied, and very few studies had included sociodemo-graphically diverse populations.

Many other studies have developed prediction models for preeclampsia/eclampsia by selecting a single BP during the first or second trimesters, or configured BP measures extending up through 28 to 32 weeks, in addition to clinical risk factors (ie, BMI, age, parity) and biomarkers.^56,57 The heterogeneity of the study samples limited their clinical utility for prediction of HDP.⁵⁸ Although some models identified up to ≈80% of women who developed preeclampsia/eclampsia‚ they demonstrated low sensitivity with ≈50% of women classified as high risk, which severely limited the predictive ability to discriminate subgroups at highest risk. The addition of biomarkers and second trimester Doppler scans had minimal impact and did not improve prediction, while being cost prohibitive or unavailable in most clinical settings. Overall, these models showed limited applicability and lacked calibration data to support routine implementation.⁵⁹ Other limitations includes sample sizes <20000 women, inability to determine prior chronic hypertension status or history of preeclampsia/eclampsia objectively, low racial and ethnic diversity, and no independent validation cohorts.

Strengths and Limitations

Our study had many strengths including the large, community-based sample within a single integrated health care system, 10 years of comprehensive EHR data with repeated (mean of 4) clinical BPs during early pregnancy (all outpatient BPs ≤20 weeks’ gestation), prepregnancy BMI, medical, selected social determinants of health (neighborhood deprivation index) and lifestyle factors, and broad racial and ethnic diversity. We also identified and excluded women with known clinical risk factors: serious medical conditions (ie, kidney disease, cancer, cardiovascular disease, etc), prior chronic or early pregnancy hypertension, and history of preeclampsia/eclampsia. Models accounted for age, parity, diabetes, smoking, self-reported racial and ethnic groups, and sociodemographic covariables. There were some limitations to our study; assessment of history of preeclampsia/eclampsia based on natural language processing was limited to previous births delivered within the KPNC health care system. This technique improved the identification of most women, but potentially underestimated this particular risk factor. Finally, this study was unable to evaluate individual-level social determinants of health (ie, education, economic resources, health care barriers, discrimination, nativity) that may affect the model estimates of the risk of preeclampsia/eclampsia among racial and ethnic groups. The preeclampsia/eclampsia and gestational hypertension outcomes were identified by ICD-9/10 codes within a single integrated health care system for high validity. The chart validation showed that EHR classification had excellent sensitivity and specificity (94%) for preeclampsia/eclampsia, as well as for gestational hypertension (85% and 91%).

Conclusions and Implications of Findings

Our current data show that early pregnancy longitudinal BP patterns represented by trajectory groups may provide an effective and scalable approach to identifying women at increased risk for preeclampsia/eclampsia and gestational hypertension in early pregnancy above and beyond existing patient sociodemographic, behavioral, and clinical characteristics. For equivalent early pregnancy BP patterns, there were higher risks of HDP among some racial and ethnic groups. Therefore, future studies are needed to elucidate modifiable factors predicting HDP and test potential interventions that would modify BP patterns for early mitigation of HDP risk. These findings hold promise to improve risk prediction beyond current models relying on clinical features, biomarker testing, or a single BP measurement. Next steps are to develop prediction models for preeclampsia/eclampsia and gestational hypertension using the 6 early pregnancy BP trajectory groups, and to conduct a validation study based on the separate internal sample of 74967 (30%) women that were reserved from our overall eligible sample. We will contrast the model with BP trajectory groups with and without clinical risk factors and prenatal screening biomarkers, as well as BP categories based on the AHA/ACC clinical criteria (normal, elevated, and stage 1 hypertension) defined for nonpregnant adults.

Perspectives

Finally, early pregnancy BP trajectory groups for prediction of HDP may form the basis of new clinical tools to provide alerts in real time for clinicians and health care systems, including evaluation of optimal frequency and timing of routine prenatal BP measurements across gestation. Future research will test the performance of these automated tools in comparison to a simple observation of clinical risk factors, and risk stratification to efficiently conduct clinical trials that evaluate new interventions.

NOVELTY AND RELEVANCE.

What Is New?

Clinical blood pressure measurements up to 20 weeks’ gestation revealed distinct longitudinal patterns with subsequent higher prevalence of new onset hypertensive disorders of pregnancy in low risk women.

Early pregnancy pressure patterns within acceptable (ie, normal) ranges were associated with higher risk of preeclampsia/eclampsia and gestational hypertension.

What Is Relevant?

Three systolic blood pressure patterns during the first 20 weeks of pregnancy identified 73.4% of preeclampsia/eclampsia and 82.1% of gestational hypertension outcomes, independent of initial blood pressure and other risk factors.

Early blood pressure patterns identify women who may benefit from management and interventions to ameliorate risk of pregnancy-related hypertension.

Clinical/Pathophysiological Implications?

Early blood pressure levels within the normal range (<120 mm Hg) that showed rapid increases, or minimal or no decline from 0 to 20 weeks of gestation represented several times higher risk of preeclampsia/eclampsia and gestational hypertension later in pregnancy. Blood pressure trajectories may improve risk stratification for development of hypertensive disorders of pregnancy above and beyond initial blood pressure levels.

Nonstandard Abbreviations and Acronyms

ACC/AHA

American College of Cardiology/American Heart Association

aOR

adjusted odds ratio

BMI

body mass index

BP

blood pressure

EHR

electronic health record

HDP

hypertensive disorders of pregnancy

ICD

International Classification of Diseases

KPNC

Kaiser Permanente Northern California