Expectant Management of Hypertensive Disorders of Pregnancy and Future Cardiovascular Morbidity

Abstract

Objective:

To test the hypothesis that a longer length of time between diagnosis of hypertensive disorders of pregnancy and delivery is associated with increased risk of cardiovascular morbidity in the years after delivery.

Methods:

This is a retrospective cohort study based in the New York State Inpatient Database. The first delivery for all patients from 2005 to 2014 who delivered preterm with an ICD-9 CM code for hypertensive disorders of pregnancy (excluding isolated chronic hypertension) was included. The duration between diagnosis and delivery was divided into ≤7 days or >7 days. The primary outcome was admission for a composite of cardiovascular disease, stroke, or death after the index delivery through December 31, 2014.

Results:

There were 22,594 patients with a median follow-up period of 5.2 years: 19,750 (87.4%) were delivered within 7 days of diagnosis and 2,844 (12.6%) were delivered >7 days from diagnosis. The primary outcome occurred in 216 (1.1%) patients in the 0–7 days group (21 events per 10,000 person-years) and 67 (2.4%) patients in the >7 days group (46 events per 10,000 person-years), adjusted hazard ratio 1.45 (95% confidence interval 1.09 to 1.93). The findings were robust in a number of sensitivity analyses.

Conclusions:

Prolonged expectant management of preterm hypertensive disorders of pregnancy is associated with an increased risk of maternal cardiac disease in the ensuing years.

Précis:

Prolonged expectant management of preterm hypertensive disorders of pregnancy is associated with an increased risk of maternal cardiac disease in the ensuing years.

Introduction

Recently, attention has been focused on the link between hypertensive disorders in pregnancy and long-term cardiovascular outcomes in women.^1–4 A history of preeclampsia or gestational hypertension approximately doubles a woman’s lifetime risk of ischemic cardiac disease or stroke, and clinicians are advised to inform patients of these risks and ensure appropriate cardiovascular follow-up after pregnancy.^{2, 3, 5, 6} However, little is known about how management of these disorders during pregnancy itself may affect subsequent cardiovascular risk. In particular, the relationship between the length of time from diagnosis to delivery in patients with preeclampsia or gestational hypertension and long-term cardiovascular outcomes is not known. If these disorders cause or modify the risk for subsequent cardiovascular dysfunction, as opposed to only serving as markers of poor cardiovascular health at baseline, then it is plausible that a longer duration from diagnosis to delivery could increase the future risk of cardiovascular disease.

Currently, clinically stable patients with preeclampsia or gestational hypertension are typically expectantly managed from diagnosis until 34 weeks in severe cases, or 37 weeks in cases without severe features, at which point they are delivered.⁶ Although generally considered safe for maternal and neonatal health in the short term, the long-term maternal implications of this management strategy are unknown.^7–9

Therefore, we used a large state database to test the hypothesis that an increasing duration from diagnosis to delivery in patients with hypertensive disorders of pregnancy is associated with increased risk of cardiovascular disease or death after pregnancy.

Methods

We constructed a retrospective cohort from the New York State Inpatient Database (SID), Healthcare Cost and Utilization Project (HCUP), from the Agency for Healthcare Research and Quality including all in-hospital deliveries from the third quarter of 2005 through December 31, 2014. The New York SID captures approximately 97% of all acute care hospital discharges in the state of New York from 1990–2016.¹⁰ Information in the database includes basic demographic data such as age, race, and payer, as well as International Classification of Diseases, 9^th Revision, Clinical modification (ICD-9-CM) billing codes from the hospital discharge. Furthermore, ICD-9-CM procedure codes as well as the day of the hospitalization on which the procedure occurred are included. Data are collected at the state level and are sent to the Agency for Healthcare Research and Quality which cleans and restructures the data to a common format before providing it for analysis. The years 2005–2014 were chosen to include a period with adequate time for follow-up. We did not use data from 2015 and beyond due to the adoption of the ICD-10 coding system in that year. Deliveries were identified using an algorithm that utilizes ICD-9 diagnosis and procedure codes.^{11, 12} Women residing in New York State with hypertensive disorders of pregnancy including eclampsia, severe preeclampsia, mild or unspecified preeclampsia, gestational hypertension, or superimposed preeclampsia or eclampsia on chronic hypertension were identified using diagnosis codes from the ICD-9-CM (Appendix 1). Patients with codes for only chronic hypertension were excluded, as were patients coded for both chronic hypertension and a pregnancy-related hypertensive disorder other than superimposed preeclampsia or eclampsia, due to the uncertainty of the underlying diagnosis.¹³ The index pregnancy was defined as the first delivery for that patient within the study timeframe, regardless of parity or of subsequent deliveries. Because hypertensive disorders of pregnancy are not normally managed expectantly at term, only deliveries at <37 weeks gestation (identified with ICD-9-CM diagnosis code 644.21) were included.

Patient comorbidities were extracted using ICD-9-CM diagnosis codes in the 6 months prior to delivery or during the delivery hospitalization. The Elixhauser comorbidity index was used ^{12, 14}, and additional comorbidities were coded including gestational diabetes, peripartum cardiomyopathy, and cardiovascular disease in pregnancy (such as valvular disease, coronary artery disease, or pre-pregnancy cardiomyopathy). Sociodemographic data included age, race–ethnicity (non-Hispanic black, non-Hispanic white, Hispanic, and other), and payer (private, public, or other).

For each patient, the interval between the diagnosis of a hypertensive disorder of pregnancy and delivery was calculated. The date of diagnosis was taken as the date of the index delivery admission. However, if patients had an antecedent hospital admission with a diagnosis of a hypertensive disorder of pregnancy within 17 weeks of their delivery, then the date of this preceding admission was used as the date of diagnosis. Seventeen weeks was chosen because preeclampsia and gestational hypertension are normally diagnosed no earlier than 20 weeks, and typically patients with a preexisting diagnosis of pregnancy-related hypertensive disease are not expectantly managed beyond 37 weeks were excluded, as described above.

The database includes the number of days between the admission date and each procedure occurring in the hospital. Thus, unique procedure codes associated with delivery (e.g., cesarean, operative vaginal delivery) can be identified and the duration between admission and delivery procedures was defined as the primary exposure (duration of expectant management of the hypertensive disorder). For non-operative spontaneous vaginal deliveries, the delivery date and duration were defined in a hierarchical fashion. First, procedures commonly associated with the exact day of delivery (episiotomy, manual rotation, perineal repair) were used. If those codes were not available, procedure codes related to induction of labor were used (artificial rupture of membranes, surgical or medical induction of labor), with the latest date used as this was assumed to be the most proximate to delivery. If no procedure codes were available, then the date of delivery and primary exposure could not be calculated and the patient was not included. The duration of expectant management was considered as both a continuous variable (number of weeks) and empirically categorized as 0–7 days or >7 days.

Patients who died during the delivery hospitalization, who were only diagnosed postpartum, who had no delivery date that could be calculated, who had a lag from diagnosis to delivery >17 weeks, who delivered prior to admission, who did not reside in New York State, or who were listed as “male” were excluded.

The primary outcome was a composite of death or hospital readmission for acute cardiovascular or cerebrovascular disease (hypertensive heart disease, acute myocardial infarction, cardiomyopathy, cardiac arrest, heart failure, stroke, or intracerebral hemorrhage) after the index hospitalization (through the end of 2014). The ICD-9-CM diagnosis codes and complete list of outcomes are shown in Appendix 1. Only the first readmission meeting criteria for the primary outcome was counted. Patients with occurrence of the primary outcome before or during the delivery admission were excluded, as the purpose of this study was to examine longer-term cardiovascular risks, as opposed to events within the pregnancy itself. Thus, we excluded primary outcomes that occurred during the delivery hospital admission but a sensitivity analysis was performed with these patients included and censored on the day of discharge, as has been done previously.¹⁵

Demographic and baseline clinical data were compared across the two exposure groups (0–7 days and >7 days) using the Mann-Whitney-U test for continuous variables and the chi-square test for categorical variables. Observations were censored at the end of data availability (December 31, 2014). The median follow-up time was estimated as the Kaplan-Meier estimate of potential follow-up.¹⁶ Time-to-event analysis with Kaplan-Meier curves and Cox proportional hazards was used to examine the association between increasing duration from diagnosis to delivery and development of cardio- and cerebro-vascular composite morbidity. The logrank test was used to compare the survival functions. The proportional hazards assumption was assessed using the Kolmogorov-type supremum test based on 1,000 simulations. Hazard ratios (HR) and 95% confidence intervals (CI) were calculated. Cox proportional hazards models were adjusted for age, race/ethnicity, payer, cesarean delivery, multiple pregnancy, gestational diabetes, pre-gestational diabetes, tobacco use, alcohol and drug use, obesity, anemia, chronic hypertension, chronic kidney disease, chronic lung disease, and depression. Cox models were used treating weeks from diagnosis to delivery as both a categorical variable (0–7 days vs. >7 days) and a continuous variable.

A number of sensitivity analyses were performed to evaluate the robustness of the findings. First, the analysis was restricted to only patients with severe preeclampsia, since expectant management in this population may have different effects than in a population with more mild disease. Second, since patients with an extremely long duration between diagnosis and delivery may have atypical presentations compared to patients delivered more expediently, we repeated the analysis only in patients delivered within 2 weeks (treating days from diagnosis to delivery in a continuous fashion) as well as in all patients delivered within 6 weeks (treating weeks from diagnosis to delivery as a continuous variable). Third, in order to account for subsequent pregnancies, we repeated the primary analysis including time-dependent variables for subsequent delivery and subsequent hypertensive disorders of pregnancy during the study period. Fourth, to see if there were any differences in outcomes in patients with longer periods of expectant management compared to somewhat shorter ones, we re-categorized the duration from diagnosis to delivery as ≤7 days, 8–14 days, >14 days. Fifth, we repeated the primary analysis including patients with occurrence of the primary outcome during the hospitalization, censoring on the day of discharge. Sixth, because patients with superimposed preeclampsia have cardiovascular disease at baseline by definition, we repeated the primary analysis excluding all such patients. Finally, as there were patients with no delivery date calculable who were otherwise eligible for the analyses, we repeated the primary analysis assigning all of these patients in each of the two groups (≤7 days and >7 days). Similarly, we repeated the primary analysis including patients with a diagnosis of chronic hypertension as well as a hypertensive disorder of pregnancy other than superimposed preeclampsia.

Due to the many differences in demographic and clinical characteristics between the two groups and the large number of covariates, we repeated the primary analysis using a propensity score model. Briefly, we used logistic regression to create the propensity score for the primary exposure (≤7 days and >7 days from diagnosis to delivery). Variables in the propensity score model included age, race, payer, multiple pregnancy, cesarean delivery, gestational diabetes, pre-gestational diabetes, chronic kidney disease, chronic hypertension, anemia, obesity, depression, chronic lung disease, year of delivery, and tobacco use. We then performed a 1:3 match using a greedy matching algorithm with the matching caliper set to 0.2 x the logit of the standard deviation of the propensity scores. We assessed for balance in the matched cohort by examining the standardized differences of the variables between the two groups, with a standardized difference <0.1 considered evidence for balance. After the matching process was complete, we repeated the primary Cox analysis in the matched sample stratified by the matching variable.

SAS version 9.3 (SAS Institute, Cary NC) was used for analysis. A two-sided p-value <0.05 was considered statistically significant. The study was deemed exempt from review by the Washington University in St. Louis Human Research Protection Office because the data lack any personally identifying information. Due to restrictions on presentation of results from HCUP data, any counts <11 are reported as “<11” to preserve patient privacy.

Results

A total of 23,199 patients who delivered preterm with a hypertensive disorder of pregnancy were identified. Of these, 19 died during the delivery hospitalization, <11 were diagnosed only postpartum, 183 had a mixed diagnosis of chronic hypertension and a diagnosis other than superimposed preeclampsia, 147 had no delivery date calculable, <11 had a lag from diagnosis to delivery of >17 weeks, 11 delivered prior to admission, 38 had the primary outcome prior to delivery admission, and 194 had the primary outcome during the delivery admission. This left 22,594 patients for the primary analysis (Figure 1).

Figure 1:

Study flow diagram. *Due to restrictions on presentation of results from the Healthcare Cost and Utilization Project data, any counts <11 are reported as “<11” to preserve patient privacy.

Demographic and clinical information on the final delivery population are displayed in Table 1. As expected, prolonged duration from diagnosis to delivery was uncommon: there were 19,750 (87.4%) patients with time from diagnosis to delivery of 0–7 days and 2,844 (12.6%) with duration >7 days. Patients with prolonged expectant management were older, more likely to be black, less likely to have severe preeclampsia but more likely to have superimposed preeclampsia, and had more comorbidities compared to patients with shorter duration of expectant management.

Table 1.

Characteristics of the women with a diagnosis of a pregnancy-related hypertensive disorder who delivered preterm in New York from 2005–2014

Variable	≤7 days (n=19,750)	>7 days (n=2,844)	P*
Median age (interquartile range)- yr	30 (24, 35)	32 (26, 36)	<0.001
Race or ethnicity —no.(%)			0.052
White	6,843 (34.7)	956 (33.6)	Referent category
Black	5,711 (28.9)	887 (31.2)	0.013
Hispanic	3,610 (18.3)	481 (16.9)	0.076
Other †	3,586 (18.2)	520 (18.3)	0.73
Payer—no. (%)			0.45
Private	10,266 (52.0)	1,495 (52.6)	Referent category
Public	9,092 (46.0)	1,302 (45.8)	0.34
Other	392 (2.0)	47 (1.7)	0.23
Hypertensive diagnosis at delivery—no.(%) ‡			<0.001
Gestational   Hypertension	2,545 (13.0)	316 (11.1)	Referent category
Mild preeclampsia	5,363 (27.2)	859 (30.2)	<0.001
Severe preeclampsia	9,017 (45.7)	1,046 (36.8)	0.79
Eclampsia	263 (1.3)	11 (0.4)	<0.001
Superimposed   Preeclampsia	2,562 (13.0)	612 (21.5)	<0.001
Cesarean Delivery—no. (%)	13,555 (68.6)	2,133 (75.0)	<0.001
Multiple pregnancy—no. (%)	3,156 (16.0)	606 (21.3)	<0.001
Comorbidities —no. (%)
Chronic hypertension	1,979 (10.0)	533 (18.7)	<0.001
Chronic kidney disease	98 (0.5)	35 (1.2)	<0.001
Chronic lung disease	1,479 (7.5)	318 (11.2)	<0.001
Gestational diabetes	2,213 (11.2)	508 (17.9)	<0.001
Pre-gestational   Diabetes mellitus	825 (4.2)	266 (9.4)	<0.001
Obesity	1,687 (8.5)	413 (14.5)	<0.001
Anemia	3,439 (17.4)	627 (22.1)	<0.001
Depression	506 (2.6)	129 (4.5)	<0.001
Tobacco use	937 (4.7)	157 (5.5)	0.071
Alcohol use	41 (0.2)	<11 a	0.68
Drug use	371 (1.9)	47 (1.7)	0.40
Year of delivery no. (%)			0.88
2014	1,549 (7.8)	222 (7.8)	0.92
2013	1,746 (8.8)	243 (8.5)	0.58
2012	1,925 (9.8)	290 (10.2)	0.47
2011	2,088 (10.6)	305 (10.7)	0.83
2010	2,233 (11.3)	319 (11.2)	0.87
2009	2,258 (11.4)	305 (10.7)	0.26
2008	2,158 (10.9)	313 (11.0)	0.93
2007	2,293 (11.6)	312 (11.0)	0.31
2006	2,399 (12.2)	367 (12.9)	0.78
2005	1,101 (5.6)	168 (5.9)	Referent category

^*from Chi-square test, Fisher Exact Test, Mann-Whitney-U, or, for categorical variables with >2 categories, logistic regression. Percentages may not total 100 because of rounding.

^†including 398 patients with missing race

^‡each patient only counted once, most severe category used

^adue to privacy restrictions counts <11 are not displayed

The median follow-up period for the 0–7 days group was 5.2 years (interquartile range 3.0, 7.8), and the median follow-up period for the >7 days group was similar at 5.2 years (interquartile range 3.0, 7.5) (p=0.34). The median time from diagnosis to delivery was 1 day (interquartile range 0–3 days, range 0–117 days). The primary outcome, readmission for acute cardiovascular disease or death, occurred in 216 (1.1%) patients in the 0–7 days group (21 events per 10,000 person-years) and 67 (2.4%) patients in the >7 days group (46 events per 10,000 person-years). There were 32 deaths during the follow-up period (3 deaths per 10,000 person-years. Cardiomyopathy and heart failure were the most common outcomes (n=336), other acute cardiovascular diseases were rare (n=14), and cerebrovascular disease or stroke accounted for 88 cases (patients may have had more than one diagnosis but were only counted once in the composite outcome).

We found a statistically significant association between duration from diagnosis to delivery and the development of the primary outcome with a HR of 2.16 (95% CI 1.64, 2.84) for duration >7 days compared to ≤7 days. The association remained significant after adjusting for clinical covariates, with an adjusted HR of 1.45 (95%CI 1.09, 1.93) (Figure 2 and Table 2). The logrank test for homogeneity of the survival curves was significant (p<0.001). When duration from diagnosis to delivery was considered as a continuous variable, the crude HR for each increase in week between diagnosis and delivery was 1.19 (95%CI 1.13, 1.25), with an adjusted HR of 1.07 (95% CI 1.01, 1.13).

Figure 2:

Time-to-event curves for cardiovascular readmission or death. Patients with a lag >7 days from diagnosis of a hypertensive disorder of pregnancy to delivery (red) had an increased risk of death or cardiovascular admission in the years after delivery compared to patients who were delivered within 7 days (blue).

Table 2.

Hazard Ratio and 95% Confidence Interval for Readmission for Cardiovascular disease, Cerebrovascular disease, or Death for Covariates

Variable	Category	Crude Hazard Ratio (95% confidence interval)	Adjusted Hazard Ratio (95% Confidence Interval)	Hazard ratio (95% confidence interval) from propensity score matched analysis (n=11,185)
Time from diagnosis to delivery	>7 days	2.16 (1.64, 2.84)	1.45 (1.09, 1.93)	1.55 (1.11, 2.17)
	≤7 days	Referent	Referent	Referent
Age (per year)		1.03 (1.01, 1.04)	1.03 (1.01, 1.04)	0.99 (0.95, 1.01)
Race/ethnicity	Black	2.31 (1.72, 3.09)	1.77 (1.29, 2.44)	1.57 (0.97, 2.52)
	Hispanic	1.40 (0.97, 2.02)	1.19 (0.81, 1.77)	1.21(0.65, 2.24)
	Other	1.08 (0.72, 1.61)	1.03 (0.69, 1.56)	0.80 (0.40, 1.59)
	White	Referent	Referent	Referent
Payer	Public	1.73 (1.36, 2.19)	1.42 (1.09, 1.86)	1.47 (1.00, 2.15)
	Other	0.74 (0.24, 2.33)	0.77 (0.24, 2.43)	5.4 (0.46, 62.07)
	Private	Referent	Referent	Referent
Multiple pregnancy		0.81 (0.58, 1.14)	1.05 (0.74, 1.50)	0.58 (0.35, 0.96)
Cesarean delivery		1.35 (1.03, 1.76)	1.06 (0.80, 1.40)	1.12 (0.65, 1.95)
Gestational diabetes		1.28 (0.92, 1.78)	1.02 (0.72, 1.43)	0.64 (0.40, 1.02)
Pre-gestational diabetes		5.57 (4.19, 7.41)	3.15 (2.29, 4.34)	1.80 (1.04, 3.12)
Chronic hypertension		3.49 (2.69, 5.43)	1.86 (1.14, 2.50)	1.10 (0.67, 1.81)
Chronic kidney disease		10.21 (6.07, 17.18)	3.12 (1.76, 5.52)	2.15 (0.73, 6.37)
Obesity		2.50 (1.83, 3.41)	1.49 (1.06, 2.09)	0.97 (0.59, 1.60)
Chronic lung disease		2.57 (1.89, 3.50)	1.82 (1.32, 2.51)	0.95 (0.56, 1.63)
Anemia		2.38 (1.85, 3.05)	1.81 (1.39, 2.35)	1.36 (0.90, 2.05)
Depression		2.24 (1.35, 3.70)	1.41 (0.83, 2.38)	1.66 (0.73, 3.77)
Tobacco use		1.60 (1.02, 2.53)	1.16 (0.71, 1.89)	1.75 (0.82, 3.74)
Alcohol use		3.49 (0.87, 14.03)	0.99 (0.22, 4.43)	2.56 (0.23, 29.12)
Illicit drug use		2.00 (1.06, 3.76)	1.12 (0.56, 2.25)	2.32 (0.74, 7.29)
Year of Delivery	2014	0.47 (0.14, 1.62)	0.31 (0.09, 1.08)	1.30 (0.29, 5.82)
	2013	0.89 (0.40, 2.00)	0.63 (0.28, 1.42)	1.81 (0.49, 6.69)
	2012	1.33 (0.79, 2.54)	0.95 (0.49, 1.83)	2.32 (0.80, 6.73)
	2011	1.31 (0.71, 2,42)	0.96 (0.52, 1.78)	3.75 (1.37, 10.26)
	2010	0.87 (0.46, 1.64)	0.68 (0.36, 1.28)	1.65 (0.59, 4.56)
	2009	1.25 (0.71, 2.20)	1.02 (0.57, 1.80)	2.60 (0.98, 6.85)
	2008	1.67 (0.98, 2.84)	1.45 (0.85, 2.48)	3.53 (1.40, 8.86)
	2007	1.37 (0.80, 2.32)	1.27 (0.75, 2.17)	1.98 (0.79, 4.94)
	2006	1.39 (0.84, 2.31)	1.29 (0.78, 2.14)	1.68 (0.71, 4.01)
	2005	Referent	Referent	Referent

In the analysis restricted to only patients with severe preeclampsia (n=10,063), the association of expectant management with risk of the composite outcome persisted, with an adjusted HR of 1.82 (95% CI 1.10, 3.01) for a duration of >7 days from diagnosis to delivery compared to ≤7 days. When restricting to only patients delivered within two weeks of diagnosis (n=20,993), each additional day between diagnosis and delivery was associated with an adjusted HR for the composite outcome of 1.04 (95%CI 1.00, 1.08). When excluding latency >6 weeks (n=22,325 in analysis), the adjusted HR for each additional week of expectant management was 1.13 (95% CI 1.02, 1.24). An analysis including time-dependent variables for subsequent delivery and subsequent hypertensive disorder did not alter the results (data not shown).

Next, when considering the time from diagnosis to delivery as ≤7 days, 8–14 days, >14 days, the adjusted HR for 8–14 days (compared to ≤7 days) was 1.43 (95% CI 0.95, 2.16), while the adjusted HR for >14 days (compared to ≤7 days) was 1.46 (95% CI 1.03, 2.06). When including all patients who had the primary outcome during their delivery hospitalization (n=194), the adjusted HR for the primary outcome was 1.28 (95% CI 0.97, 1.69). When excluding all patients with superimposed preeclampsia, the adjusted HR for the primary outcome was 1.65 (95% CI 1.16, 2.35). Finally, including the 147 patients with no calculable delivery date calculable (assigning all such patients to each of the two exposure groups) did not materially alter the effect estimate for the primary outcome (when including all these patients in the >7 days group, the aHR and 95%CI for the primary outcome was 1.44 (1.08, 1.90), when including them all in the ≤7 days group the aHR and 95%CI were 1.44 (1.09, 1.92). Similarly, including the 176 patients with a “mixed” diagnosis of chronic hypertension and a hypertensive disorder of pregnancy other than superimposed preeclampsia did not change the results.

The propensity score matching algorithm matched 2,836 of the 2,844 patients in the >7 days group (143 patients matched to at least one but <3 controls and were included in the analysis). The covariates were balanced between the two groups with all standardized differences <0.1. The demographic data and the standardized differences are presented in the Appendix 2. The HR and 95%CI for the primary outcome in the matched model were 1.55 (1.11, 2.17), similar to the estimates from the adjusted Cox model (Table 2).

Discussion

We found that prolonged exposure to hypertensive disorders of pregnancy was associated with an increased risk of subsequent cardiovascular readmission or death with a median follow-up of over 5 years, with a 7% increased risk for each week of expectant management. These findings suggest that the vascular insult of preeclampsia and related obstetric hypertensive disorders appear to increase maternal cardiovascular morbidity and may exert a dose-response effect related to duration of expectant management from diagnosis to delivery.

There have been only limited studies specifically focusing on interval from diagnosis to delivery and outcomes. In a follow-up of the HYPITAT trial of expectant management versus delivery of patients with mild preeclampsia at or after 36 weeks’ gestation, there were no differences in cardiovascular outcomes at a mean 2.5 years postpartum.¹⁷ These findings are important but have limited applicability to patients with early-onset hypertensive disorders of pregnancy who may be managed expectantly for far longer than in the HYPITAT trial (3.3 days in the immediate delivery group vs. 10.3 days in the expectant group in that trial). For instance, in one retrospective cohort of patients with early-onset preeclampsia without severe features, the mean latency was 7.6 days; however, the subgroup of patients who successfully achieved at least 7 days of latency had a mean latency of 24 days.¹⁸ However, follow-up of preterm patients from the HYPITAT-II trial as well as earlier trials of expectant management versus delivery will be important for understanding the cardiovascular risk conferred by prolonged management of these disorders.^{9, 19}

To date the pathophysiological mechanisms underlying the association between hypertensive disorders of pregnancy and long-term maternal cardiovascular outcomes are not known. On the one hand, hypertensive disorders of pregnancy may reflect underlying maternal risk factors, rather than representing new physiologic insults that modify the future risk.^{20, 21} On the other hand, these disorders may cause endothelial dysfunction or other physiologic alterations that either cause permanent injury during pregnancy or that persist after delivery.^{22, 23} In the case of expectant management of obstetric hypertensive disease, a prolonged exposure to this endothelial dysfunction may cause long-lasting cardiovascular injury that is manifested in cardiovascular disease later in life.

Our study has a number of strengths. First, the data are derived from a large, comprehensive database that includes nearly all in-patient admissions in the state of New York. Second, ICD-9 coding has been tested and validated for critical variables in this analysis including induction of labor, cesarean delivery, and preterm delivery^{11, 13, 24, 25} Although coding accuracy for hypertensive disorders of pregnancy may be more prone to diagnostic error, and the diagnostic criteria for preeclampsia and gestational hypertension changed during the study period, we included all hypertensive disorders of pregnancy. Additionally, it is unknown whether the maternal risks differ based on slight differences in diagnostic criteria; it is for this reason that we included all hypertensive disorders of pregnancy without distinction.^{6, 26, 27} Third, by using time-to-event models we were able to include the maximum amount of potential follow-up available for each patient and had a median of >5 years of follow-up time available.

Limitations of this analysis should be considered. First, the HCUP dataset is based on ICD-9 codes, which limits the amount of available clinical information. Furthermore, because the database only includes inpatient admissions, patients who were diagnosed in the outpatient setting and managed as outpatients may be coded as having less time of exposure than they actually did. However, the findings persisted in patients coded for severe preeclampsia, which is typically managed in the inpatient setting. Perhaps the biggest limitation of the dataset is the limited obstetrical information available, specifically gestational age at delivery and diagnosis, as this information is not available in ICD-9 coding. We limited this analysis to only preterm deliveries to ensure roughly comparable gestational ages, but further investigations that include gestational age will be important for determining if our findings are due primarily to the early-onset of the hypertensive disorders or to the duration of expectant management regardless of gestational age of exposure. Finally, patients who moved out of New York State or who sought care in other states were lost to follow-up.

In conclusion, this study demonstrates that prolonged exposure to hypertensive disorders of pregnancy is associated with an increased risk of maternal cardiovascular disease or death in the ensuing years after delivery. Although expectant management of these conditions is considered safe in the short term, and is often done to optimize neonatal outcomes, the long-term maternal implications of this strategy deserve further attention. Even if the overall balance of neonatal morbidity and maternal morbidity leans in favor of expectant management according to current guidelines, interventions including maternal cardiology follow-up and risk minimization interventions should be studied and prioritized during and after pregnancies complicated by obstetric hypertensive disease.