2015 Volume 79 Issue 7 Pages 1609-1617
Background: Three risk estimation methods for predicting the cardiac outcomes of pregnancy in women with heart disease have been proposed. This study was designed to compare their prediction performance in an Asian cohort with congenital heart disease (CHD).
Methods and Results: This study enrolled pregnant women with CHD who delivered their babies after the 20th gestational week between 1985 and 2011. Of 268 pregnancies in 190 women with CHD, 18 (6.7%) had cardiac complications. The incidence of maternal cardiac events among women with a CARPREG index of 0, 1 or 2 was 3.4%, 27.3% and 100%. The incidence was 2.7%, 8.6%, 11.1%, 40% and 17.6% for those with a ZAHARA score 0–0.5, 0.51–1.5, 1.51–2.5, 2.51–3.5 and >3.5. Among patients with a modified World Health Organization (WHO) classification I, II, III and IV, the incidence of maternal cardiac events was 0%, 4.0%, 12.2% and 25.7%. The c-statistic was 0.732 (95% confidence interval (CI): 0.589, 0.876; P<0.001) for the CARPREG score, 0.737 (95% CI: 0.611, 0.864; P=0.001) for the ZAHARA score and 0.827 (95% CI: 0.745, 0.909; P<0.001) for the WHO classification.
Conclusions: All 3 risk estimation methods had good performance in predicting maternal cardiac outcomes; however, the modified WHO classification demonstrated superior discrimination and calibration. (Circ J 2015; 79: 1609–1617)
With the advances in cardiac interventions, more patients with congenital heart disease (CHD) can survive into adulthood,1 and the number of women with CHD who reach childbearing age is increasing.2 The presence of CHD may increase the risk of maternal cardiovascular complications during pregnancy, as well as the risk of fetal and neonatal complications related to the altered hemodynamic status in pregnancy.3–7 Risk identification and providing counseling before planning pregnancy are therefore essential for women with CHD. Given that the risk of pregnancy may be affected by the general cardiac status and specific heart disease, several approaches have been proposed for estimating the risk. Guidelines from the American College of Cardiology/American Heart Association, the Canadian Cardiovascular Society and the European Society of Cardiology (ESC) have summarized the pregnancy suggestions according to specific congenital cardiac lesions.8–10 Because considerable variation in the clinical status of patients with the same specific heart lesion occurs frequently and suggestions of these guidelines are usually based on small retrospective study series, 2 risk scores that considered the maternal clinical cardiac status, not the specific disease, were proposed.5,11 First, the CARPREG investigators in Canada have provided a risk index for predicting maternal pregnancy outcome in women with heart disease (74% of their study cohort were women with CHD).11 They defined 4 risk predictors: (1) prior cardiac event (heart failure, transient ischemic attack, stroke or arrhythmia); (2) New York Heart Association (NYHA) functional classification >II or cyanosis at baseline; (3) impaired systemic ventricular function; and (4) severe left-sided heart obstruction. The estimated risks of cardiac events during pregnancy were 5%, 27% and 75% in those women with numbers of risk factors of 0, 1 and >1, respectively. The CARPREG risk index was further validated by several studies.3,12 The risk factors of severe pulmonary regurgitation and/or reduced subpulmonary ventricular systolic function were added to improve the overall risk assessment.3 The investigators of the ZAHARA study in Europe proposed another risk scoring system for predicting the pregnancy complications in women with CHD.5 Their scoring system consists of 8 risk factors with a variable weighted coefficient, including a history of arrhythmia, cardiac medications before pregnancy, NYHA functional class ≥II, severe left heart obstruction, severe systemic atrioventricular (AV) valve regurgitation, severe pulmonary AV valve regurgitation, mechanical valve prosthesis and cyanotic heart disease. They advocated that the modification of the risk index might enhance discrimination and calibration. The third method, named the modified World Health Organization (WHO) risk classification, was recommended by the ESC to estimate the risk of pregnancy in women with heart disease.13,14 Four risk classes (WHO I, II, III and IV) were defined according to the conditions of specific heart lesions and the clinical cardiac risk status such as mechanical valve, systemic right ventricle, Fontan circulation or pulmonary arterial hypertension. Women classified as WHO class IV were advised against pregnancy because of the extremely high maternal cardiovascular risks. The CARPREG index and modified WHO classification can be applied to both congenital and acquired heart conditions; however, the ZAHARA score is exclusively for women with CHD.
Relevant data regarding the outcome and risk analyses of pregnancy in women with CHD in the Asian population are still scarce.15–17 In addition, the prediction performance of the ZAHARA score for women with CHD has not yet been examined. Therefore, based on a longitudinal database from a tertiary care center in Taiwan, we investigated the outcomes of pregnancy in women with CHD. The predictive performances of the 3 risk prediction methods were compared in this Asian patient cohort.
The study was approved by the Institutional Review Board of the National Taiwan University Hospital. Pregnant women who had CHD and delivered after at least 20 weeks of gestation from 1985 to 2011 at the hospital were enrolled in this retrospective cohort study. Data collection included age, height, weight, educational and marital status, smoking, alcohol consumption, medications, obstetric history, systemic medical history, prior cardiac event history (arrhythmias, congestive heart failure, stroke or transient ischemic attack), NYHA functional class before pregnancy, blood pressure, heart rate, presence of cyanosis at rest before pregnancy (oxygen saturation measured by pulse oximetry <90%), electrocardiography results, echocardiographic results, and other cardiac imaging findings. Systemic ventricular dysfunction was defined as an ejection fraction <40% measured by echocardiography and left heart obstruction was defined by a mitral valve area <2 cm2, an aortic valve area <1.5 cm2, or a peak left ventricular outflow gradient >30 mmHg detected by Doppler echocardiography. In addition, severe left ventricular outflow tract (LVOT) obstruction was defined as a peak LVOT gradient >40 mmHg on Doppler echocardiography. An estimated or measured systolic pressure of the main pulmonary artery between 30 and 40 mmHg denoted mild pulmonary hypertension, between 40 and 70 mmHg was defined as moderate pulmonary hypertension and pressures above 70 mmHg indicated severe pulmonary hypertension. The underlying CHD in these patients were classified into 3 groups based on the complexity (simple, moderate complexity and great complexity), according to the definition by the Bethesda Conference classification.1 In addition, cardiac, obstetric, and neonatal events were recorded as described by Siu et al.11
1. Cardiac events consisted of cardiac death, cardiac arrest, stroke, symptomatic sustained arrhythmias requiring therapy, pulmonary edema, a decline in at least 2 NYHA functional classes, and the need for urgent invasive cardiac interventions during pregnancy or within 6 weeks postpartum.
2. Obstetric events consisted of pre-eclampsia (defined as a sustained systolic or diastolic blood pressure ≥140 or ≥90 mmHg, respectively, with concurrent proteinuria), profound postpartum hemorrhage (defined as an estimated blood loss ≥500 ml for vaginal delivery or ≥1 L for cesarean delivery accompanied by ≥20 g/L drop in hemoglobin or requiring transfusion), and noncardiac death.
3. Fetal and neonatal events were defined as preterm delivery (before 37 weeks of gestation), birth weight small for gestational age (<10th percentile), respiratory distress syndrome, cerebral intraventricular hemorrhage, fetal death after 20 weeks of gestation, neonatal death (within the first month after birth) or the presence of CHD.
Application of the Risk Estimation MethodsThe risk estimation methods were applied according to the clinical status immediately before the pregnancy by reviewing the medical records of each patient.
CARPREG Index The score was calculated by assigning 1 point to each of the following variables: cardiac events prior to pregnancy (arrhythmia, congestive heart failure or stroke/transient ischemic attack), a baseline NYHA functional class >II or cyanosis, systemic ventricular dysfunction, left heart obstruction as defined by Siu et al11 and reduced subpulmonary ventricular function and/or severe pulmonary regurgitation as defined by Khairy et al.3
ZAHARA Score The score was calculated by summing the 8 risk factors with different weighted coefficients.5
Modified WHO Risk Classification The estimated maternal cardiac risk associated with pregnancy was classified into 4 categories according to the recommendation of the Task Force of the ESC. The categories ranged from the very low risk of WHO class I to the highest risk of WHO class IV.13,14
Statistical MethodsThe statistical software SAS version 9.2 (SAS Institute, Cary, NC, USA) was used for all analyses. Continuous variables are expressed as the mean±standard deviation. Because some patients experienced more than 1 pregnancy, we conducted a general estimating equation (GEE) analysis to identify the potential significant maternal risk factors for maternal cardiac, fetal/neonatal, and obstetric events separately. Variables with a P value <0.1 in the univariable analyses were included in the multivariable GEE models. Variables with 2-tailed values of P<0.05 in the multivariable GEE model were considered statistically significant. The correlations between the categorical variables were analyzed by Kendall’s tau-b method, and the results expressed as r2. To access the discrimination of the 3 risk estimation methods, we computed the c-statistic, which is equivalent to the area under the receiver-operating characteristic (ROC) curve.18 A model with a c-statistic >0.75 is generally considered to have a meaningful discriminatory ability and a c-statistic <0.5 denotes a poor discriminative capacity. We also examined the calibration of the score systems using the Hosmer-Lemeshow goodness-of-fit test and a significant P value indicates a lack of fit.19
There were 201 pregnant women with CHD identified for this study. After excluding 7 patients who underwent therapeutic abortion and 4 patients who had spontaneous abortions before the gestational age of 20 weeks, we enrolled 190 mothers who had delivered 271 babies after at least 20 weeks of gestation (including 3 deliveries of twins). The maternal age at delivery ranged from 16 to 42 years (30.4±4.7 years). The distribution of the patients’ CHD is summarized in Table 1. Among them, 91 (47.9%) patients had simple CHD, 73 (38.4%) had CHD of moderate complexity, 17 (8.9%) had CHD of great complexity, and 9 (4.7%) had Marfan syndrome (Table 2). Cesarean sections were performed in 138 deliveries (51.5%) (Table 2). Because our study was retrospective, we could only obtain the record of NYHA functional status before pregnancy in 233 of 268 (87%) deliveries (Table 2).
| Parity | |||
|---|---|---|---|
| 1 | 125 (65.8%) | ||
| 2 | 53 (27.9%) | ||
| 3 | 11 (5.8%) | ||
| 4 | 1 (0.5%) | ||
| CHD | Native | Repaired | Total |
| Ventricular septal defect | 41 | 25 | 66 |
| Atrial septal defect | 19 | 26 | 45 |
| Tetralogy of Fallot | 0 | 22* | 22 |
| Patent ductus arteriosus | 8 | 13 | 21 |
| Marfan syndrome | 9 | 3† | 12 |
| Pulmonary valve stenosis | 4 | 7 | 11 |
| Aortic stenosis | 2 | 5 | 7 |
| Atrioventricular septal defect | 0 | 5 | 5 |
| Ebstein’s anomaly | 3 | 0 | 3 |
| Aortic coarctation | 1 | 1 | 2 |
| Complete TGA | 0 | 2‡ | 2 |
| Congenitally corrected TGA | 1 | 1§ | 2 |
| Other | 2¶ | 5# | 7 |
*The 21 patients with total repair, 1 patient with palliative central shunt. †Three patients with Bentall’s operation plus aortic valve replacement. ‡One patient with Mustard operation, 1 patient with Senning operation. §Repair of ventricular septal defect plus Rastelli operation. ¶One with coronary artery fistula, 1 with interruption of aortic arch. #One with double chambered right ventricle, 1 with cor triatriatum, 1 with hemitruncus, 1 with double outlet of right ventricle, 1 with pulmonary atresia with intact ventricular septum after bidirectional Glenn shunt. CHD, congenital heart disease; TGA, transposition of the great arteries.
| Patients, n | Deliveries, n | Cardiac events, n (%) |
Obstetric events, n (%) |
Fetal/neonatal events, n (%) |
|
|---|---|---|---|---|---|
| Total | 190 | 268 | 18 (6.7) | 10 (3.7) | 53 (19.8) |
| C-section | 138 | 14 (10.1) | 8 (5.8) | 32 (23.2) | |
| Twin pregnancy | 3 | 3 | 1 (33.3) | 0 | 3 (100) |
| CHD complexity | |||||
| Simple | 91 | 120 | 3 (2.5) | 5 (4.2) | 18 (15.0) |
| Moderate | 73 | 114 | 5 (4.4) | 3 (2.6) | 19 (16.7) |
| Great | 17 | 22 | 9 (40.9) | 1 (4.6) | 13 (59.1) |
| Marfan syndrome | 9 | 12 | 1 (8.3) | 1 (8.3) | 3 (25.0) |
| Pulmonary hypertension | 22 | 32 | 9 (28) | 2 (6.3) | 10 (31.0) |
| Mild | 10 | 16 | 3 (18.8) | 1 (6.3) | 3 (18.8) |
| Moderate | 4 | 4 | 0 | 0 | 0 |
| Severe | 8 | 12 | 6 (50.0) | 1 (8.3) | 7 (58.3) |
| SaO2 <90% | 5 | 6 | 4 (66.7) | 1 (16.7) | 5 (83.3) |
| NYHA functional class | |||||
| I–II | 159 | 227 | 12 (5.3) | 8 (3.5) | 42 (18.5) |
| III–IV | 5 | 6 | 6 (100) | 1 (16.7) | 6 (100) |
| Prior cardiac events | 12 | 16 | 4 (25.0) | 0 | 4 (25.0) |
| Previous arrhythmias | 11 | 14 | 4 (28.6) | 0 | 4 (28.6) |
| Mechanical valve | 9 | 11 | 2 (18.2) | 2 (18.2) | 4 (36.4) |
| Severe AR | 9 | 14 | 2 (14.3) | 0 | 4 (28.6) |
| Severe PR | 17 | 22 | 1 (4.6) | 1 (4.6) | 3 (13.6) |
| Moderate/severe SAVR | 7 | 9 | 2 (22.2) | 0 | 1 (11.1) |
| Moderate/severe PAVR | 23 | 34 | 6 (17.7) | 2 (5.9) | 7 (20.6) |
| Severe LVOT obstruction | 3 | 6 | 0 | 0 | 3 (50.0) |
| Severe systemic ventricular dysfunction | 1 | 2 | 2 (100) | 0 | 1 (50.0) |
| Cardiac medication | 14 | 22 | 3 (13.6) | 2 (9.1) | 7 (31.8) |
| Maternal smoking | 3 | 6 | 1 (16.7) | 0 | 3 (50.0) |
| Marfan syndrome, Ao >40 mm | 5 | 7 | 1 (14.3) | 1 (14.3) | 3 (42.9) |
| CARPREG index | |||||
| 0 | 165 | 234 | 8 (3.4) | 9 (3.9) | 39 (16.7) |
| 1 | 24 | 33 | 9 (27.3) | 1 (3.0) | 13 (39.4) |
| 2 | 1 | 1 | 1 (100) | 0 | 1 (100) |
| ZAHARA score | |||||
| 0–0.5 | 106 | 149 | 4 (2.7) | 6 (4.0) | 22 (14.8) |
| 0.51–1.5 | 51 | 70 | 6 (8.6) | 2 (2.9) | 16 (22.9) |
| 1.51–2.5 | 18 | 27 | 3 (11.1) | 0 | 8 (29.6) |
| 2.51–3.5 | 3 | 5 | 2 (40) | 0 | 2 (40.0) |
| ≥3.51 | 12 | 17 | 3 (17.6) | 2 (11.8) | 5 (29.4) |
| WHO classification | |||||
| I | 62 | 91 | 0 | 3 (3.3) | 12 (13.2) |
| II | 75 | 101 | 4 (4.0) | 3 (3.0) | 15 (14.9) |
| III | 29 | 41 | 5 (12.2) | 2 (4.9) | 15 (36.6) |
| IV | 24 | 35 | 9 (25.7) | 2 (5.7) | 11 (31.4) |
Ao, aorta; AR, aortic regurgitation; C-section, caesarian section; CHD, congenital heart disease; LVOT, left ventricular outflow tract; NYHA, New York Heart Association; PAVR, pulmonary atrioventricular valve regurgitation; PR, pulmonary regurgitation; SaO2, systemic oxygen saturation; SAVR, systemic atrioventricular valve regurgitation; WHO, World Health Organization.
The pregnancy outcomes were assessed for the occurrence of 3 types of complications (maternal cardiac, obstetric and fetal/neonatal events). Among the 268 pregnancies, no complication occurred in 196 (73.1%). Only 2 pregnancies (1 patient with Eisenmenger’s syndrome and 1 patient with Marfan syndrome) were complicated with all 3 types of events, and 19 (7.1%) pregnancies were accompanied by 2 types of complication. The occurrence of maternal cardiac events was associated with the occurrence of fetal/neonatal events (r2=0.391, P<0.001).
Maternal Cardiac Events Cardiac complications occurred in 18 (6.7%) of 268 deliveries, including 2 (0.7%) cases of cardiac death. Both deaths occurred immediately after delivery in patients with unrepaired CHD and severe pulmonary hypertension. The most common cardiac event was aggravated congestive heart failure, which was detected in 13 deliveries (4.9%), followed by symptomatic arrhythmias in 4 deliveries (1.5%). Infective endocarditis, pulmonary embolism and pulmonary edema each occurred once in different patients (Table 3).
| No. of pregnancies (%) | |
|---|---|
| Cardiac events | 18 (6.7%) |
| Aggravated congestive heart failure | 13 |
| Symptomatic arrhythmia | 4 |
| Death | 2 |
| Infective endocarditis | 1 |
| Pulmonary embolism | 1 |
| Pulmonary edema | 1 |
| Obstetric events | 10 (3.7%) |
| Postpartum hemorrhage | 8 |
| Pre-eclampsia | 2 |
| Fetal/neonatal events | 53 (19.8%) |
| Prematurity | 27 |
| Small for gestation age | 15 |
| Respiratory distress syndrome | 6 |
| Neonatal death | 3 |
| Neonatal CHD | 5 |
| Fetal death | 2 |
CHD, congenital heart disease.
Because we found that NYHA functional class >II highly correlated with cyanosis (r2: 0.766, P<0.001), we combined these 2 variables for the GEE analysis. Additionally, the variable of moderate to severe pulmonary AV valve regurgitation was removed because of its high correlation with great CHD complexity, previous arrhythmia, NYHA class >II or cyanosis and pulmonary hypertension. The risk factors associated with a maternal cardiac event, according to the multivariable GEE analysis, were NYHA functional class >II or cyanosis (odds ratio (OR): 17.88; 95% confidence interval (CI): 1.06–301.2; P=0.045), moderate or severe systemic AV valve regurgitation (OR: 6.04; 95% CI: 1.08–33.89; P=0.041) and pulmonary hypertension (OR: 4.47; 95% CI: 1.12–17.86; P=0.034) (Table 4).
| Univariable analysis | Multivariable analysis | |||||
|---|---|---|---|---|---|---|
| OR | 95% CI | P value | OR | 95% CI | P value | |
| Maternal cardiac events | ||||||
| Great CHD complexity | 15.34 | (4.7, 50.04) | <0.0001 | 2.28 | (0.45, 11.6) | 0.319 |
| Age (per 1 year of increase) | 1.08 | (0.95, 1.23) | 0.23 | |||
| C-section | 4.28 | (1.17, 15.68) | 0.028 | 2.76 | (0.78, 9.81) | 0.116 |
| Cardiac medication | 2.79 | (0.68, 11.40) | 0.152 | |||
| Mechanical valve | 3.39 | (0.64, 18.05) | 0.152 | |||
| Previous arrhythmia | 6.85 | (1.87, 25.01) | 0.004 | 4.11 | (0.99, 17.07) | 0.052 |
| NYHA class >II or cyanosis | 47.1 | (7.65, 290.1) | <0.0001 | 17.88 | (1.06, 301.2) | 0.045 |
| Moderate/severe SVVR | 4.51 | (0.79, 25.67) | 0.09 | 6.04 | (1.08, 33.89) | 0.041 |
| Moderate/severe PVVR | 3.91 | (1.22, 12.57) | 0.022 | |||
| Pulmonary hypertension | 11.01 | (3.6, 33.68) | <0.0001 | 4.47 | (1.12, 17.86) | 0.034 |
| Maternal obstetric events | ||||||
| C-section | 3.94 | (0.82, 18.87) | 0.086 | 3.4 | (0.7, 16.4) | 0.128 |
| Cardiac medication | 2.98 | (0.59, 14.98) | 0.186 | |||
| Mechanical valve | 6.91 | (1.42, 33.76) | 0.017 | 5.21 | (1.08, 25.13) | 0.040 |
| NYHA class >II or cyanosis | 4.64 | (0.69, 31.33) | 0.116 | |||
| Fetal/neonatal events | ||||||
| Great CHD complexity | 7.03 | (2.32, 21.3) | 0.0006 | 8.96 | (1.89, 42.51) | 0.0057 |
| Marfan syndrome with aortic dilatation | 2.97 | (0.98, 9.01) | 0.055 | |||
| Maternal smoking | 4.68 | (1.56, 14.07) | 0.006 | 6.94 | (2.24, 21.53) | 0.0008 |
| C-section | 1.6 | (0.82, 3.14) | 0.17 | |||
| Cardiac medication | 2.25 | (0.83, 6.07) | 0.111 | |||
| Severe LVOT obstruction | 4.07 | (0.75, 22.12) | 0.105 | |||
| Mechanical valve | 2.41 | (0.7, 8.32) | 0.163 | |||
| NYHA class >II or cyanosis | 25.75 | (2.89, 229) | 0.004 | |||
| Pulmonary hypertension | 2.49 | (0.95, 6.49) | 0.062 | 1.96 | (0.56, 6.88) | 0.29 |
CI, confidence interval; OR, odds ratio. Other abbreviations as in Table 2.
Obstetric Events There were 10 (3.7%) obstetric events in the study cohort, including postpartum hemorrhage in 8 patients and pre-eclampsia in 2 patients (Table 3). The multivariable GEE analyses identified having a mechanical valve as the only independent risk factor related to maternal obstetric events (OR: 5.21; 95% CI: 1.08–25.13; P=0.04) (Table 4).
Fetal and Neonatal Events There were 53 (19.8%) documented fetal or neonatal events, including 2 intrauterine fetal deaths and 3 neonatal deaths (Table 3). The most common complication was premature delivery (10.1%), followed by being small for gestational age (5.6%) and respiratory distress syndrome (2.2%). Five newborns presented with CHD (1.9%). The risk factors for fetal and neonatal events, according to multivariable GEE analysis, were great CHD complexity (OR: 8.96; 95% CI: 1.89–42.51; P=0.0057) and maternal smoking (OR: 6.94; 95% CI: 2.24–21.53; P=0.0008) (Table 4).
Maternal Cardiac Outcome Prediction by the 3 Different Risk Estimation MethodsThe maternal cardiac risk in women with a CARPREG index of 0 was 3.4%. The risk increased to 27.3% in women with a score of 1 and to 100% in women with a score of 2 (Figure 1, Table 2). The CARPREG index was a significant predictor of cardiac events. A CARPREG index increasing from 0 to 1 and 2 predicted an 11.94-fold increase in the number of maternal cardiac events (95% CI: 3.94–36.15; P<0.0001) (Table 5).
Occurrence of maternal cardiac events with respect to 3 risk estimation methods. WHO, World Health Organization.
| Risk estimation method | OR | 95% CI | P value |
|---|---|---|---|
| CARPREG index | |||
| 0 | 1 | – | – |
| 1 and 2 | 11.94 | (3.94, 36.15) | <0.0001 |
| ZAHARA score (per 1 point of increase) | 1.47 | (1.19, 1.82) | 0.0004 |
| WHO classification | |||
| I and II | 1 | – | – |
| III | 6.86 | (1.71, 27.51) | 0.007 |
| IV | 16.85 | (4.54, 62.59) | <0.0001 |
Abbreviations as in Tables 2,4.
When we applied the risk score proposed by the ZAHARA study to our study cohort, the incidence of a cardiac event was 2.7% for patients with a score of 0–0.5, 8.6% for a score of 0.51–1.5, 11.1% for a score of 1.51–2.5 and 40% for a score of 2.51–3.5 (Figure 1, Table 2). However, in patients with a ZAHARA score >3.51, the incidence of maternal events decreased to 17.6%. The ZAHARA score was also a significant predictor of maternal cardiac events, and an increase of 1 point in the ZAHARA score predicted a 1.47-fold increase in the number of maternal cardiac events (95% CI: 1.19–1.82, P=0.0004) (Table 5).
No maternal cardiac event occurred in patients in WHO class I. The cardiac risk was 4.0% for those women inWHO class II, 12.2% in WHO class III and 25.7% in WHO class IV (Figure 1, Table 2). A WHO class increase from I and II to class III predicted a 6.86-fold increase in maternal cardiac risk (95% CI: 1.71–27.51; P=0.007) (Table 5). A WHO class increase from I and II to class IV predicted a 16.85-fold increase in maternal cardiac risk (95% CI: 4.54–62.59; P<0.0001) (Table 5). These results indicated that the WHO classification was also a significant predictor of maternal cardiac events.
The ROC curves for predicting maternal cardiac events using these 3 risk estimation methods in our cohort are depicted in Figure 2. The CARPREG index and ZAHARA score both demonstrated good discrimination, with a c-statistic of 0.732 (95% CI: 0.589–0.876; P<0.001) and 0.737 (95% CI: 0.611–0.864; P<0.001), respectively (Figure 2, Table 6). However, the Hosmer-Lemeshow P value of the ZAHARA score was 0.024, indicating lack-of-fit for the ZAHARA score in predicting the maternal cardiac events in our cohort. The ZAHARA score had a tendency to overestimate the risk in patients with a score >3.51 (Figure 1).
Receiver-operating characteristic curves of 3 risk estimation methods of predicting maternal cardiac events. WHO, World Health Organization.
| Risk estimation method | ROC curve | Hosmer-Lemeshow test | ||
|---|---|---|---|---|
| c-statistic | 95% CI | P value | P value | |
| CARPREG index | 0.732 | (0.589, 0.876) | <0.001 | – |
| ZAHARA score | 0.737 | (0.611, 0.864) | 0.001 | 0.024 |
| WHO classification | 0.827 | (0.745, 0.909) | 0.001 | 1.0 |
ROC, receiver-operating characteristic. Other abbreviations as in Tables 2,4.
Comparing the other 2 risk scores, the WHO classification demonstrated the best discrimination, with a c-statistic (area under the ROC curve) of 0.827 (95% CI: 0.745–0.909; P=0.001) (Figure 2, Table 6). The Hosmer-Lemeshow goodness-of-fit P value of the WHO classification was 1.0, suggesting that the WHO classification had good calibration.
This is the first study to compare the performance of 3 risk estimation methods for predicting maternal cardiac risk of pregnancy in patients with CHD. Our results indicated that all 3 methods were significant predictors of maternal cardiac risk, but the modified WHO classification had the best discrimination and calibration.
The modified WHO classification is the only method that takes into account both specific heart lesions (eg, Marfan syndrome, bicuspid aortic valve, tetralogy of Fallot, aortic coarctation, Fontan circulation or systemic right ventricle) and the clinical cardiac status (eg, systemic ventricular dysfunction).13,14 This strategy may be a better approach to predict the cardiac risks in pregnant women with CHD. No cardiac event occurred in patients in WHO class I, but maternal cardiac events happened in 3.4% of patients with CARPREG score of 0. This implies that WHO class I is more reassuring in a pregnant patient with CHD than a CARPREG score of 0. The WHO classification is the only method with a c-statistic >0.75 (0.827), slightly higher than the c-statistic of 0.732 for the CARPREG index and 0.737 for the ZAHARA score.
The rate of cardiac events increased to 25.4% for patients in WHO class IV, including 2 patients with pulmonary hypertension and Eisenmenger’s syndrome who died of cardiac cause. In addition, the WHO classification is the only risk score clearly suggesting the contraindication of pregnancy in women with heart disease (WHO class IV), which is very helpful in standard clinical practice. In our study, pulmonary hypertension was a significant independent predictor of maternal cardiac events (Table 4). Among 32 deliveries in patients with maternal pulmonary hypertension, the occurrence of maternal cardiac events was 28%. The modified WHO classification is the only risk score to take into account pulmonary hypertension as a risk factor. The recently reported incidence of pulmonary hypertension in adult patients with CHD is approximately 3.2%.20 Patients with pulmonary hypertension might be advised against pregnancy, so they were underrepresented in both the CARPREG study (0.50%, 3 in 599 deliveries)11 and the ZAHARA study (0.31%, 4 in 1,302 deliveries).5 The incidence of pulmonary hypertension in our study was much higher (11.9%, 32 in 268 deliveries), possibly because most of these pregnancies were in the early years of our study and pre-pregnancy counseling was not provided at that time. The mortality rate of the present patients with severe pulmonary hypertension was 16.7% (2/12). Although this rate is lower than that in the recent report (28%), mortality is still a major issue.21 Pregnancy should be avoided in women with CHD and severe pulmonary hypertension.13,14,21,22
We noticed that the rate of cardiac events was only 17.6% among patients with ZAHARA score >3.5. Of the 12 pregnancies in patients with ZAHARA score >3.5, 11 of the women had mechanical valves, which would result in a ZAHARA score of at least 4.25. Among the 11 pregnancies of 9 patients with mechanical valves, only 1 patient experienced a ventricular arrhythmia and 1 patient with aggravated heart failure required medical treatment. No events, such as thromboembolism or valve dysfunction, occurred in these 9 patients during pregnancy. The rate of cardiac events in the present patients with mechanical valves was much lower than that reported by Drenthen et al (18% vs. 100%).5 Their study included only 4 pregnancies in women with maternal mechanical prosthesis, and all experienced cardiac events. The small sample size in their study may have resulted in an overestimation of the risk. In our previous study of 47 pregnant patients with mechanical valves, the rates of valvular dysfunction and thromboembolism were both 6.4% (3/47).23 These data are similar to those reported by Sbarouni et al.24 The rates of valvular dysfunction and thromboembolism were 8.6% (13/151) and 5.2% (8/151), respectively. The weighted coefficient for a mechanical valve in the ZAHARA score system may lead to an overestimation of the risk. Although a state-of-the-art regimen of anticoagulation therapy during the pregnancy may account for the decreased number of cardiac events, the risk of having a mechanical valve in pregnancy is still significant.14,21,25–27
The present study investigated the pregnancy outcomes in women with CHD in Taiwan. In this Asian cohort, cardiac events (6.7%), obstetric complications (3.7%) and neonatal complications (19.8%) were frequently encountered. Nevertheless, three-quarters of the 268 pregnancies were free of any events during the pregnancy and the delivery. We identified the association of NYHA functional class >II or cyanosis, pulmonary hypertension and moderate/severe systemic AV valve regurgitation with adverse maternal cardiac events. Recently reported incidences of maternal cardiac complications range from 4.5% to 25.0%, and the variation is mostly explained by the different distributions of disease severity in the reported series.3–6,11,15,16,28 Compared with the patients in the study by Siu et al,11 we had a smaller proportion of patients with CARPREG scores of 1 and 2 (Figure 1), and the occurrence of maternal cardiac events in our patients was lower (6.7% vs. 13%). In contrast, the distributions of the ZAHARA scores were quite similar between our patients and the patient group in the ZAHARA study (Figure 1),5 and the maternal cardiac complication rates were similar (6.7% in our series vs. 7.1% in the ZAHARA group). The CARPREG index5 and the ZAHARA score11 were derived from cohorts with heart disease or CHD, respectively. The performance of risk prediction may be affected by the disease distribution of the study cohort. For example, patients with pulmonary hypertension were underrepresented in both the CARPREG and ZAHARA studies. Therefore, the risk of pulmonary hypertension in pregnancy may be underestimated by these 2 scores. The better prediction performance of the WHO classification may be attributed to the inclusion of some important cardiac lesions or clinical conditions such as systemic right ventricle or pulmonary hypertension, which are rather uncommon. The WHO classification has better integrality of known maternal cardiac risks in predicting the cardiac outcomes of pregnancy in women with CHD.
Study LimitationsOur study was retrospective and some missing values were inevitable in our cohort. For example, the record of NYHA functional status before pregnancy was obtained in 233 of 268 (87%) deliveries in our series (Table 2). As mentioned earlier, heterogeneity in the distribution of disease severity could affect the overall outcome or the prediction performance of the risk scores in different cohorts of pregnant women with CHD. In addition, a few patients may have undergone therapeutic termination or avoided pregnancy because of the presence of known risk factors related to pregnancy. The development of new treatment strategies for CHD, such as the evolution of Fontan procedures or new device therapies, may further alter the disease patterns of pregnant patients with CHD in the future.
A cardiopulmonary exercise test with measurement of chronotropic response prior to pregnancy can also provide an objective method of predicting pregnancy outcomes for women with CHD.17,29 A larger scale, prospective study in the future incorporating exercise testing with known maternal risk factors during pre-pregnancy counseling may be helpful in further refining the prognosis of pregnancy in patients with CHD.
The hemodynamic burden during pregnancy may increase the cardiovascular risk in women with CHD. One-quarter of the pregnancies in this study had complications, including maternal cardiac, obstetric and fetal/neonatal events. We discovered that all 3 risk estimation methods were effective, but the modified WHO classification may be better in predicting maternal cardiac outcomes. Women with CHD who are considering pregnancy should receive careful evaluation and risk stratification.
This work was supported by a grant from the National Taiwan University Hospital (NTUH098-N1245). We thank Chin-Hao Chang, PhD and Hui-Wen Chang, MS. from the National Translational Medicine and Clinical Trial Resource Center (which is funded by National Research Program for Biopharmaceuticals at the National Science Council of Taiwan; NSC101-2325-B-002-078) and the Department of Medical Research at the National Taiwan University Hospital for their expert statistical assistance.
None of the authors has a conflict of interest.
