Background
As early life interventions for congenital heart disease improve, more patients are living to adulthood and are considering pregnancy. Scoring and classification systems predict the maternal cardiovascular risk of pregnancy in the context of congenital heart disease, but these scoring systems do not assess the potential subsequent risks following pregnancy. Data on the long-term cardiac outcomes after pregnancy are unknown for most lesion types. This limits the ability of healthcare practitioners to thoroughly counsel patients who are considering pregnancy in the setting of congenital heart disease.
Objective
We aimed to evaluate the association between pregnancy and the subsequent long-term cardiovascular health of individuals with congenital heart disease.
Study Design
This was a retrospective longitudinal cohort study of individuals identifying as female who were receiving care in two adult congenital heart disease centers from 2014 to 2019. Patient data were abstracted longitudinally from a patient age of 15 years (or from the time of entry into the healthcare system) to the conclusion of the study, death, or exit from the healthcare system. The primary endpoint, a composite adverse cardiac outcome (death, stroke, heart failure, unanticipated cardiac surgery, or a requirement for a catheterized procedure), was compared between parous (at least one pregnancy >20 weeks’ gestation) and nulliparous individuals. By accounting for differences in the follow-up, the effect of pregnancy was estimated based on the time to the composite adverse outcome in a proportional hazards regression model adjusted for the World Health Organization class, baseline cardiac medications, and number of previous sternotomies. Participants were also categorized according to their lesion type, including septal defects (ventricular septal defects, atrial septal defects, atrioventricular septal defects, or atrioventricular canal defects), right-sided valvular lesions, left-sided valvular lesions, complex cardiac anomalies, and aortopathies, to evaluate if there is a differential effect of pregnancy on the primary outcome when adjusting for lesion type in a sensitivity analysis.
Results
Overall, 711 individuals were eligible for inclusion; 209 were parous and 502 nulliparous. People were classified according to the World Health Organization classification system with 86 (12.3%) being classified as class I, 76 (10.9%) being classified as class II, 272 (38.9%) being classified as class II to III, 155 (22.1%) being classified as class III, and 26 (3.7%) being classified as class IV. Aortic stenosis, bicuspid aortic valve, dilated ascending aorta or aortic root, aortic regurgitation, and pulmonary insufficiency were more common in parous individuals, whereas dextro-transposition of the great arteries, Turner syndrome, hypoplastic right heart, left superior vena cava, and other cardiac diagnoses were more common in nulliparous individuals. In multivariable modeling, pregnancy was associated with the composite adverse cardiac outcome (36.4%% vs 26.1%%; hazard ratio, 1.83; 95% confidence interval, 1.25–2.66). Parous individuals were more likely to have unanticipated cardiac surgery (28.2% vs 18.1%; P =.003). No other individual components of the primary outcome were statistically different between parous and nulliparous individuals in cross-sectional comparisons. The association between pregnancy and the primary outcome was similar in a sensitivity analysis that adjusted for cardiac lesion type (hazard ratio, 1.61; 95% confidence interval, 1.10–2.36).
Conclusion
Among individuals with congenital heart disease, pregnancy was associated with an increase in subsequent long-term adverse cardiac outcomes. These data may inform counseling of individuals with congenital heart disease who are considering pregnancy.
Introduction
As early life interventions for congenital heart disease (CHD) improve, a higher proportion of patients are living to adulthood and are considering pregnancy. The physiological changes that occur as part of pregnancy place hemodynamic stress on individuals with CHD. Scoring and classification systems have been developed to predict the maternal cardiovascular risk of pregnancy in the context of CHD. The most commonly used systems include the modified World Health Organization (WHO) classification, Zwangerschap bij vrouwen met een Aangeboren HARtAfwijking-II (ZAHARA) risk score, Canadian Cardiac Disease in Pregnancy (CARPREG), and CARPREG II scoring systems. These scoring systems aim to predict the cardiovascular risk associated with a pregnancy but do not assess the potential risks that follow pregnancy.
Why was this study conducted?
An increasing number of individuals with congenital heart disease are surviving into adulthood and many are pursuing pregnancy. Although calculators and algorithms exist to predict the risk for cardiac events during pregnancy, little data related to the effects of pregnancy on long-term cardiac outcomes exist. Given the increased cardiac work during pregnancy, it is plausible that pregnancy could have long-term effects on cardiac function.
Key findings
Pregnancy was associated with a higher risk for a long-term composite adverse cardiac outcome among individuals with congenital heart disease who were followed longitudinally.
What does this add to what is known?
This study provides information that extends beyond predicting cardiac events during pregnancy itself by evaluating adverse cardiac outcomes subsequent to delivery. These data can be used for tailored counseling of individuals with congenital heart disease about the anticipated effects of pregnancy on long-term cardiac outcomes.
A previous retrospective cohort study demonstrated an increased risk for long-term adverse cardiac outcomes among parous individuals when compared with nulliparous individuals in cases with a previous pulmonary valve repair or replacement for a tetralogy of Fallot or a pulmonic stenosis. The long-term cardiac outcomes following pregnancy in individuals with other types of CHD remain unknown. Siu et al compared the long-term adverse cardiac events in parous patients with CHD to parous patients without CHD. Parous patients with CHD had a higher risk for adverse long-term cardiac events than those without CHD. However, the long-term effect of pregnancy on cardiovascular function among patients with CHD remains largely unknown, highlighting the importance of ongoing research in this area to optimize patient counseling on the anticipated cardiac outcomes beyond the immediate peripartum period.
Our objective was to evaluate whether there is an association between pregnancy and long-term cardiac outcomes among individuals with CHD. Importantly, this was performed by simulating a longitudinal cohort with ongoing follow-ups over time and taking into account the timing of the pregnancy on the primary cardiovascular event endpoint.
Materials and Methods
This was a retrospective, longitudinal cohort study of all individuals identifying as female, who had CHD, were ≥15 years, and were followed at the Adult Congenital Heart Disease (ACHD) clinic at the University of Utah Health or Intermountain Healthcare, two tertiary care facilities in Salt Lake City, Utah, which serve the majority of patients with CHD in the state. As part of the standard of practice for quality improvement and patient tracking, any patients with CHD seen in the ACHD clinic at either site are included in a clinical database. Individuals were included in the database if they had a visit to the ACHD clinic from August 2014 (creation of institutional clinical database) through April 2019. All individuals in this database were evaluated for eligibility, which included a cursory electronic medical record review to confirm that the age of the patient was ≥15 years and to confirm the presence of a CHD or aortopathy diagnosis. If these criteria were met, a thorough medical record abstraction was performed by trained healthcare personnel. This project was granted institutional review board exemption by both the University of Utah Health and Intermountain Healthcare. The results are reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for observational studies.
The primary exposure was pregnancy. For this study, pregnancy was defined as a delivery at ≥20 weeks’ gestation. Individuals with pregnancies who delivered at ≥20 weeks’ gestation were considered parous. Otherwise, individuals were classified as nulliparous. Adoptions were not included. Data regarding the pregnancy history and specific dates of deliveries were abstracted from the electronic medical records. When antepartum records were available, risk classification scores (WHO, CARPREG, CARPREG II, and ZAHARA) for the pregnancy were also abstracted. The WHO classification was preferentially utilized in the analysis for the following two reasons: (1) some patients did not have CARPREG or ZAHARA scores available because they were not under the care of a physician at our ACHD or Maternal-Fetal Medicine clinics during those pregnancies and (2) the WHO classification has predictive capability exceeding that of the other risk stratification scores.
Once identified, participants were followed longitudinally from 15 years of age (or the time of entry into the healthcare system) until the conclusion of the study, death of the patient, or exit from the healthcare system. The primary outcome was a composite long-term adverse cardiac outcome, which included death, stroke, heart failure, unanticipated cardiac surgery, and the need for a catheterized procedure. These outcomes were chosen a priori based on what were thought to be clinically meaningful endpoints among the authors who are subspecialists in maternal-fetal medicine and ACHD. An age of 15 years was selected as an inclusion criterion for study entry because this is the point at which adult female cardiovascular physiology is considered to have been achieved. Complications and interventions occurring between birth and 15 years of age were collected but not included as an adverse cardiac outcome (eg, surgery at 12 years of age was documented but not noted as part of the composite outcome).
Death was ascertained from reports in the electronic medical records. Cross-linkage to death certificates in the state of Utah occurs on an ongoing basis to ensure that the outcome of death is noted in the electronic medical record of the decedent. Stroke was defined as either a hemorrhagic or ischemic stroke based on brain imaging or surgical pathology reports. Heart failure was defined as either a clinical diagnosis of heart failure or a diagnosis of pulmonary edema by a cardiologist. Unanticipated cardiac surgery included surgical procedures that were not considered to be part of the original or planned staged repair of the congenital lesion. Catheterized procedures were defined as cardiac catheter procedures during which an intervention was performed (eg, not limited to diagnostic evaluation only). These procedures were further described as valvular (eg, balloon valvuloplasty) or other (eg, ablation, collateral vessels requiring coiling).
Baseline characteristics were also abstracted and included race, ethnicity, education history, employment status, insurance status, body mass index (BMI), noncardiac medical conditions (type 1 diabetes mellitus, type 2 diabetes mellitus, chronic hypertension, hypercholesterolemia, neurologic disease or seizures, obstructive sleep apnea [OSA], obesity, or other), baseline medications (anticoagulants, antiarrhythmic agents, antihypertensive agents, lipid-lowering agents, diuretics, other, or unknown), and substance use (tobacco, alcohol, or illicit drugs). Information about the specific cardiac lesion was collected and the cardiac lesion types were grouped into septal defects (ventricular septal defect, atrioseptal defect, atrioventricular septal defect, or atrioventricular canal defect), right-sided valvular lesions, left-sided valvular lesions, complex cardiac anomalies, and aortopathies for comparison between parous and nulliparous participants.
Study data were collected and managed using the Research Electronic Data Capture system, a secure, web-based software platform. , To ensure validity of the detailed manual chart abstraction, a random sample of 5% of the charts were abstracted by both of the data abstractors (S.L.S. and L.L.H.) for the primary outcome with 97% agreement (kappa= 0.85; 95% confidence interval [CI], 0.66–1.0). One discrepancy was corrected, and the updated data were analyzed.
A power estimate was performed based on the number of individuals available in an existing clinical database covering a narrower window than the planned study and the 25% pregnancy rate among those individuals. Assuming a similar distribution of people in the clinic and pregnancies over the years of planned data extraction, we expected a total of 262 parous and 787 nulliparous individuals for inclusion between August 2014 and April 2019. We anticipated that 90% of the individuals would have complete data available for chart abstraction. Thus, with a group sample sizes of 236 parous individuals and 708 nulliparous individuals we would have 80% power to detect a difference between the group proportions of 11%. The rate of morbidity among nulliparous individuals was assumed to be 50% under the null hypothesis and 39% under the alternative hypothesis. The test statistic that was used is the 2-sided Fisher exact test with an α=.05.
Statistical analysis
Demographic and baseline clinical characteristics were summarized for the nulliparous and parous individuals, with between-group differences tested using a 2-sample t test for continuous variable or an exact chi-square test for categorical characteristics. To test the primary hypothesis that among individuals with CHD, parous individuals will have a higher incidence of subsequent long-term adverse cardiac outcomes than nulliparous individuals, the effect of pregnancy on the time to composite adverse outcome was estimated in a proportional hazards regression model to account for the differential duration of observation. In adjusted modeling, the WHO class, use of cardiac medications, and number of previous sternotomies were included as clinically important covariates. Given the small number of patients categorized as WHO class IV, classes III and IV were combined in the model. Difference in number of years of observation (ie, age) between individuals was accounted for in time-to-event modeling. In a time-varying Cox model, time was modeled as maternal age and pregnancy as a time-varying covariate. Parous subjects were excluded if the dates of their deliveries were unknown given that pregnancy was the primary exposure and survival analysis using a time-to-event methodology necessitates knowledge of the timing of exposure.
Given that there may be differences in the outcomes based on cardiac lesion type, we performed a sensitivity analysis in which the cardiac lesion type was included in the multivariable model instead of the WHO class. A test for interaction was first performed for pregnancy and cardiac lesion type. Because there was no significant interaction between the pregnancy and cardiac lesion type, the reported sensitivity analysis model included parity, cardiac lesion type, use of cardiac medications, and the number of previous sternotomies. In a subgroup analysis, we evaluated whether there was a difference in the effect between pregnancy and our primary morbidity endpoint when individuals with Turner syndrome were excluded. We estimated the adjusted Cox proportional hazards model from which our primary endpoint was reported and individuals with Turner syndrome were excluded.
Statistical difference was defined as a 2-sided P value of <.05. All analyses were performed in SAS (SAS Institute Inc, Cary, NC).
Results
In total, 1064 individuals were identified by the institutional clinical database and screened for inclusion. Individuals were most commonly excluded for being younger than 15 years of age or not having a diagnosis of CHD (eg, acquired heart disease such as rheumatic heart disease) ( Figure 1 ). Among otherwise eligible patients, 105 were excluded because of missing dates associated with deliveries (precluding time-to-event analysis), leaving a final sample size of 711 individuals (209 parous and 502 nulliparous).
Parous individuals were more commonly older, were followed longitudinally for longer periods of time, had a higher BMI, were partnered or married, and employed than nulliparous individuals ( Table 1 ). The distribution of baseline WHO class did not differ between parous and nulliparous individuals ( Table 2 ). Parous individuals less commonly had a sternotomy before 15 years of age ( Table 2 ). The frequency of cardiac lesion types for the parous and nulliparous individuals is shown in Figure 2 . Aortic stenosis, bicuspid aortic valve, dilated ascending aorta or aortic root, aortic regurgitation, and pulmonary insufficiency were more common in parous individuals, whereas dextro-transposition of the great arteries, Turner syndrome, hypoplastic right heart, left superior vena cava, and other cardiac diagnoses were more common among nulliparous individuals ( Figure 2 ).
| Characteristic | All, N=711 | Parous, n=209 | Nulliparous, n=502 | P value |
|---|---|---|---|---|
| Age at first visit (y), mean (SE) | 23.9 (10.1) | 26.3 (10.1) | 22.9 (9.9) | <.001 |
| Age at last visit (y), mean (SE) | 28.9 (10.4) | 32.4 (9.0) | 27.4 (10.6) | <.001 |
| Years followed, mean (SE) | 4.7 (5.3) | 6.1 (5.5) | 4.1 (5.2) | <.001 |
| Current or recent overweight or obese BMI | 330 (48.0) | 111 (54.7) | 219 (45.2) | .024 |
| Hispanic ethnicity | 37 (5.2) | 13 (6.2) | 24 (4.8) | .431 |
| Marital status: partnered or married | 270 (39.1) | 164 (80.0) | 106 (21.9) | <.001 |
| Currently employed | 291 (52.2) | 104 (63.0) | 187 (47.7) | <.001 |
| Noncardiac chronic conditions | ||||
| Type 1 diabetes | 4 (0.6) | 2 (1.0) | 2 (0.4) | .364 |
| Type 2 diabetes | 12 (1.7) | 2 (1.0) | 10 (2.0) | .329 |
| Chronic HTN | 49 (6.9) | 17 (8.1) | 32 (6.4) | .399 |
| Hypercholesterolemia | 25 (3.5) | 9 (4.3) | 16 (3.2) | .461 |
| Neurologic disease or seizures | 26 (3.7) | 4 (1.9) | 22 (4.4) | .11 |
| Obstructive sleep apnea | 34 (4.8) | 4 (1.9) | 30 (6.0) | .021 |
| Obesity | 8 (1.1) | 4 (1.9) | 4 (0.8) | .198 |
| Other chronic conditions | 313 (44.0) | 75 (35.9) | 238 (47.4) | .005 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
