Abstract
Background
Pregnancy in women with cardiac disease carries substantial maternal risk. The modified WHO classification, Cardiac Disease in Pregnancy Study II (CARPREG II) and Zwangerschap bij Aangeboren HARtAfwijkingen (ZAHARA; Pregnancy in Congenital Heart Disease) are the three principal risk stratification tools in this setting. External validation in contemporary practice is essential to assess its performance in specialist programmes.
Methods
Single-centre retrospective cohort study of all pregnancies in women with cardiac disease delivering ≥24+0 gestational weeks at a level IV maternal centre between 2009 and 2023. The primary outcome was maternal cardiac complication within 12 months postpartum; the secondary outcome was obstetric complication during pregnancy or the postpartum period. Model accuracy was assessed by comparing the weighted average predicted risk with the observed number of cardiac events, using each scoring system’s original published definitions. Additional performance measures included observed-to-expected (O:E) ratios and logistic-regression area under the curve (AUC).
Results
Among 85 473 deliveries, 407 pregnancies in 282 women were included. The observed maternal cardiac complication rate was 4.18% (17 events); no maternal deaths occurred. All three models overestimate risk: weighted average predicted rates were 12.78% (CARPREG II), 8.3% (mWHO) and 6.7% (ZAHARA), yielding O:E ratios of 0.33, 0.50 and 0.62, respectively. Discrimination was moderate across all scores (AUC 0.69 (95% CI 0.55 to 0.80), 0.69 (95% CI 0.53 to 0.83) and 0.74 (95% CI 0.61 to 0.85) for mWHO, CARPREG II and ZAHARA, respectively. Obstetric complication rates were similar across mWHO classes (8.0%, 7.0% and 7.4% for classes I–III). No score predicted obstetric complications (AUC ≈0.5 for all models).
Conclusions
Cardiac risk stratification scores might overestimate maternal cardiac risk in this contemporary specialist cohort, consistent with calibration drift, while preserving a clinically meaningful cardiac risk gradient. None captured obstetric risk. These findings highlight the role of individualised assessment. While our findings provide useful preliminary insight, further evaluation in larger and more diverse cohorts is needed to refine risk prediction.
Keywords: CONGENITAL HEART DISEASE, Pregnancy, Congenital Abnormalities, Women
WHAT IS ALREADY KNOWN ON THIS TOPIC
The mWHO classification, CARPREG II and ZAHARA are the three principal tools for maternal cardiac risk stratification in pregnancy, yet all were derived more than a decade ago in populations with substantially higher cardiac event rates than those seen in contemporary specialist programmes. Systematic overestimation of risk has been reported across multiple external validation studies, though no single tool has demonstrated consistent superiority.
WHAT THIS STUDY ADDS
In a 14-year tertiary cohort of 407 pregnancies managed under a standardised multidisciplinary protocol, all three scores substantially overestimate the observed cardiac complication rate of 4.18%. A cardiac risk gradient was preserved across all models, but none predicted obstetric complications, with area under the curve values near 0.5 for all scores.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
These findings suggest that absolute risk estimates from existing scores should be interpreted cautiously in contemporary specialist settings, while their relative risk rankings retain clinical utility for triage and surveillance planning.
Introduction
Pregnancy imposes a progressive haemodynamic load on the cardiovascular system, changes that begin early in the first trimester and reach their peak at the third trimester, delivery and postpartum.1 2 Women with structural or arrhythmic cardiac disease who cannot adequately adapt to these demands are at risk of heart failure, symptomatic arrhythmia, thromboembolic events and, in the most severe cases, death.3 4 Adverse fetal and neonatal outcomes, including prematurity, growth restriction and neonatal mortality, are also elevated compared with the general obstetric population.5
Over time, pregnancies complicated by pre-existing cardiac disease have become more prevalent, making cardio-obstetric risk stratification an increasingly important clinical skill.6 Preconception risk stratification is central to the care of these women.7 Accurate risk estimation supports evidence-based counselling about the advisability of pregnancy, guides decisions about prepregnancy cardiac optimisation, determines the appropriate level and setting of antenatal surveillance and underpins delivery planning within a multidisciplinary pregnancy heart team.8 9
Three risk scores currently form the foundation of clinical risk stratification in this setting: mWHO classification, Cardiac Disease in Pregnancy Study II (CARPREG II) and Zwangerschap bij Aangeboren HARtAfwijkingen (ZAHARA; Pregnancy in Congenital Heart Disease) score.5 8 10 Variable findings across settings have shown that no single risk score is superior, and that population-specific validation is necessary.11–13
Several studies have reported overestimation of cardiac risk by these tools in contemporary cohorts, though findings have varied, with some cohorts reporting underestimation in lower-risk groups.14–18 The ability to rank patients by relative risk (discrimination) has been more consistently preserved, though no single tool has demonstrated clear superiority across settings.
The present study was, therefore, designed to: (1) externally validate the mWHO classification, CARPREG II score and ZAHARA score against observed maternal cardiac and obstetric complications in our contemporary cohort; (2) compare calibration and discrimination across both endpoint types for all three scores simultaneously.
Methods
Study design and setting
This retrospective cohort study was conducted at Rabin Medical Center (Petach Tikva, Israel), a tertiary referral centre and level IV maternal care centre according to the American College of Obstetricians and Gynecologists (ACOG).19 The pregnancy heart team operates weekly joint antenatal clinics and multidisciplinary delivery planning meetings for all women with known cardiac disease. The present study uses the same institutional cohort as previously described of peripartum anaesthetic management and postpartum haemorrhage complications in women with heart diseases.20
Study population
We identified all pregnancies in women with a documented diagnosis of cardiac disease delivering at our centre between January 2009 and December 2023; 2009 was chosen as the year the electronic medical record system was introduced.
Patients were identified by searching electronic medical records for relevant International Classification of Diseases codes. Two physicians independently reviewed each patient’s cardiac diagnoses and assigned them to the appropriate mWHO class based on the severity of their lesion, in accordance with standard definitions.8 A senior cardiologist then reviewed the diagnosis and confirmed the classifications for each patient.
Inclusion criteria were: (1) maternal age ≥18 years at delivery; (2) confirmed diagnosis of maternal cardiac disease, including congenital heart disease (CHD), valvular heart disease, cardiomyopathy, aortopathy or significant arrhythmic heart disease; (3) delivery ≥24+0 gestational weeks and (4) follow-up data available through at least 12 months postpartum. Exclusion criteria included isolated mitral valve prolapse without haemodynamic significance; patent foramen ovale as the sole cardiac diagnosis; gestational hypertension or preeclampsia without a pre-existing cardiac diagnosis; pregnancy termination or fetal loss before 24 weeks and incomplete medical records precluding risk score calculation or outcome ascertainment. Where a woman had more than one eligible pregnancy during the study period, all pregnancies were included as independent observations. Given the very small number of patients with mWHO class IV disease (n=2) and the clinical heterogeneity and extreme-risk profile of this subgroup, these patients were not included in the mWHO-stratified statistical analyses. Both patients with mWHO class IV had severe aortic stenosis with left ventricular dysfunction. The first patient had complex CHD and a history of prior cerebrovascular accident. She delivered by caesarean section and required aortic valve replacement several months postpartum. The second patient had an uneventful peripartum course.
Baseline data extraction
Extracted variables included maternal age, body mass index (kg/m²), gravidity, parity, primary cardiac diagnosis (free-text name and diagnosis code), New York Heart Association (NYHA) functional class before pregnancy and before delivery, history of prior cardiac events (arrhythmia, atrial fibrillation, heart failure, thromboembolic event, pulmonary embolism, cerebrovascular accident, endocarditis, cardiac arrest/sudden death reanimation), presence of implanted device (pacemaker, defibrillator), use of cardiac medications before pregnancy (digoxin, beta-blockers, calcium channel blockers, angiotensin-converting-enzyme (ACE) inhibitors, antidiabetic agents, furosemide and hydralazine) and anticoagulation during pregnancy. Obstetric history included prior caesarean deliveries, gestational diabetes mellitus, fetal growth restriction and mode of delivery (vaginal, caesarean).
Risk score calculation
The three scores applied were: CARPREG II, ZAHARA, and the mWHO classification as specified in the 2018 ESC pregnancy guidelines.5 8 10 For women classified as mWHO II-III, assignment to class II or III was made by a senior cardiologist based on additional clinical features. mWHO classification was assigned based on cardiac diagnosis according to the 2018 ESC Guidelines.8 CARPREG II assigns points based on 10 weighted predictors: prior cardiac events or arrhythmia (3 points), poor functional class or cyanosis (3 points), mechanical valve (3 points), ventricular dysfunction (2 points), high-risk valve lesion or LVOTO (2 points), pulmonary hypertension (2 points), coronary artery disease (2 points), high-risk aortopathy (2 points), no prior cardiac intervention (1 point) and late pregnancy assessment (1 point). The predicted cardiac complication rate increases with total score: 5% for 0-1 points, 10% for 2 point, 15% for 3 points 22% for 4 points and 41% for >4 points, as per the original published thresholds.10 Given the small number of events in our cohort, score categories 2 and ≥3 were combined into a single ≥2 group for tabular presentation and group comparisons. ZAHARA assigns weighted points across eight predictors: history of arrhythmia (1.5), poor functional class (0.75), left heart obstruction (2.5), mechanical valve prostconsistentlyhesis (4.25), systemic atrioventricular valve regurgitation (0.75), pulmonary atrioventricular valve regurgitation (0.75), cardiac medication use before pregnancy (1.5) and cyanotic heart disease (1.0). Risk groups are defined as low (<2 points), intermediate (2–3.9 points) and high (≥4 points), with published event rates of 2.9%, 7.5% and 17.2%, respectively.5
Expected event rates for the mWHO classification were taken from the 2018 ESC guidelines: 2.5–5% for class I, 5.7–10.5% for class II and 19–27% for class III.8 Women initially classified as mWHO II–III were assigned to class II or III by a senior cardiologist, so the II–III band was not used.4 21
Outcome definitions
The primary outcome was maternal cardiac complication during pregnancy or within 12 months postpartum, defined as any of the following: new or worsening heart failure requiring medical therapy or hospitalisation; sustained symptomatic arrhythmia requiring treatment; thromboembolic event (pulmonary embolism, deep vein thrombosis, or valve thrombosis); cerebrovascular accident or transient ischaemic attack; acute coronary syndrome; pacemaker implantation; cardiac catheter-based or surgical intervention for a cardiac indication; cardiac arrest or cardiac-related death.
Secondary complications included: (1) obstetric complication during pregnancy or postpartum period, defined as hypertensive disorder of pregnancy (gestational hypertension, pre-eclampsia or eclampsia), postpartum haemorrhage (blood loss >500 mL for vaginal delivery or >1000 mL for caesarean delivery), need for blood transfusion, haemorrhagic shock, disseminated intravascular coagulopathy, pulmonary embolism, sepsis, unplanned hysterectomy or relaparotomy and (2) a composite endpoint defined as cardiac or obstetric complication.
Statistical analysis
All analyses were performed using Python (V.3.12.3; pandas V.2.2.2, numpy V.1.26.4, scipy V.1.13.1, matplotlib V.3.8.4). Continuous variables are presented as mean±SD or median (IQR); categorical variables as frequency and percentage for each risk score, observed complication rates were tabulated per score category and compared with published predicted rates from each score’s original derivation cohort: CARPREG II, ZAHARA and mWHO from the 2018 ESC guidelines and the ROPAC registry.4 5 8 21 Calibration was assessed by computing a weighted average predicted risk (sum of the proportion of patients in each category multiplied by the published midpoint event rate for that category) and comparing it to the observed event rate. The O:E ratio was defined as the observed event rate divided by the weighted average predicted (expected) event rate, where 1.0 indicates perfect calibration, values below 1.0 indicate overestimation of risk by the score, and values above 1.0 indicate underestimation.
Discrimination was assessed using logistic regression (scikit-learn Logistic Regression) with each score entered as a continuous predictor of the primary cardiac outcome. Bootstrap CIs for AUC estimates were wide and overlapping across all three scores, precluding firm conclusions about superiority between tools. Clinically prespecified cut-offs: mWHO≥III, CARPREG II≥1 and ZAHARA≥3. A two-sided p value <0.05 was considered statistically significant.
Patient and public involvement
Patients and the public were not involved in the design, conduct, reporting or dissemination plans of this research. This study was a retrospective analysis of routinely collected clinical data; prospective patient or public involvement was, therefore, neither feasible nor appropriate given the study design.
Results
Study population
Between 2009 and 2023, 85 473 deliveries were recorded, of which 409 occurred in 284 women with cardiac disease. After exclusion of two mWHO class IV pregnancies, the analysis cohort comprised 407 pregnancies in 282 women: 125 (30.7%) in mWHO class I, 228 (56.0%) in class II and 54 (13.3%) in class III. The underlying cardiac diagnoses are summarised in online supplemental table S1. Demographic and baseline cardiac data are listed in table 1.
Table 1. Baseline characteristics, peripartum management, postpartum haemorrhage and complications stratified by mWHO risk class (N=407 pregnancies in 282 women).
| Variable | mWHO class I (n=125 deliveries 96 patients) |
mWHO class II (n=228 deliveries 151 patients) |
mWHO class III (n=54 deliveries 35 patients) |
P value |
|---|---|---|---|---|
| Section 1. Maternal demographic and baseline cardiac characteristics | ||||
| Age, years | 32.4±4.9 | 31.4±5.0 | 29.7±4.4 | 0.004b |
| Body mass index, kg/m² | 24.5±4.7 | 23.7±4.4 | 24.4±3.4 | 0.178 |
| Gravidity | 3.0((2.0–4.0) | 2.0((1.0–4.0) | 2.0((1.0–3.0) | 0.001c |
| Parity | 2.0((0.0–3.0) | 1.0((0.0–2.0) | 0.5((0.0–1.0) | <0.001bc |
| Cardiac medication before pregnancy*, n (%) | 3 (2.4) | 19 (8.3) | 18 (33.3) | <0.001bc |
| Anticoagulation during pregnancy, n (%) | 22 (17.6) | 6 (2.6) | 14 (25.9) | <0.001ac |
| NYHA functional class before pregnancy | — | — | — | <0.001abc |
| Class I, n (%) | 124 (99.2) | 205 (89.9) | 21 (38.9) | — |
| Class II, n (%) | 1 (0.8) | 23 (10.1) | 31 (57.4) | — |
| Class III, n (%) | 0 (0.0) | 0 (0.0) | 2 (3.7) | — |
| NYHA functional class before delivery | — | — | — | <0.001abc |
| Class I, n (%) | 123 (98.4) | 197 (86.4) | 16 (29.6) | — |
| Class II, n (%) | 2 (1.6) | 31 (13.6) | 28 (51.9) | — |
| Class III, n (%) | 0 (0.0) | 0 (0.0) | 10 (18.5) | — |
| Cardiac complications before pregnancy†, n (%) | 25 (20.0) | 21 (9.2) | 23 (42.6) | <0.001abc |
| Heart failure | 0 (0.0) | 4 (1.8) | 9 (16.7) | — |
| Cerebrovascular accident | 13 (10.4) | 8 (3.5) | 1 (1.9) | — |
| Atrial fibrillation | 0 (0.0) | 2 (0.9) | 4 (7.4) | — |
| Pulmonary embolism | 4 (3.2) | 0 (0.0) | 3 (5.6) | — |
| Thromboembolic event | 5 (4.0) | 0 (0.0) | 1 (1.9) | — |
| Endocarditis | 2 (1.6) | 2 (0.9) | 3 (5.6) | — |
| Pacemaker | 0 (0.0) | 4 (1.8) | 1 (1.9) | — |
| ARDS | 0 (0.0) | 1 (0.4) | 0 (0.0) | — |
| Section 2. Peripartum and obstetric | ||||
| Gestational age at delivery, weeks | 38.3±2.1 | 37.2±2.7 | 35.8±3.2 | <0.001bc |
| Prior caesarean delivery, n (%) | 18 (14.4) | 39 (17.1) | 10 (18.5) | 0.733 |
| Gestational diabetes mellitus, n (%) | 12 (9.6) | 30 (13.2) | 4 (7.4) | 0.375 |
| Fetal growth restriction, n (%) | 2 (1.6) | 11 (4.8) | 12 (22.2) | <0.001bc |
| Vaginal delivery, n (%) | 102 (81.6) | 174 (76.3) | 32 (59.3) | 0.005bc |
| Epidural analgesia (vaginal deliveries), n (%) | 67 (65.7) | 128 (73.6) | 27 (84.4) | 0.097 |
| Caesarean delivery, n (%) | 23 (18.4) | 54 (23.7) | 22 (40.7) | 0.005bc |
| Planned caesarean delivery, n (%) | 16 (12.8) | 40 (17.5) | 16 (29.6) | 0.025b |
| Non-elective caesarean delivery, n (%) | 7 (30.4) | 14 (25.9) | 6 (27.2) | — |
Data are presented as mean±SD, median (IQR), or n (%) as appropriate. mWHO class IV (n=2 deliveries) excluded from statistical comparisons and described separately in text.
Superscripts denote post-hoc pairwise comparisons: a=mWHO class I vs II; b=mWHO class I vs III; c=mWHO class II vs III.
Fetal growth restriction defined as estimated fetal weight or birth weight below the 10th percentile according to local institutional growth charts.
Cardiac medication before pregnancy includes digoxin, beta-blockers, calcium channel blockers, ACE inhibitors, anti-diabetics, furosemide and hydralazine.
Individual subcategory counts may exceed the group total due to one patient with two co-existing diagnoses.
ACE, angiotensin-converting-enzyme inhibitors; ARDS, acute respiratory distress syndrome; BMI, body mass index; CVA, cerebrovascular accident; GDM, gestational diabetes mellitus; mWHO, modified WHO; NYHA, New York Heart Association.
Maternal complications
The overall observed rate of maternal cardiac complications within 1-year postpartum was 4.18% (17 events out of 407 pregnancies). No maternal deaths were recorded within the first postpartum year. The most common cardiac event was heart failure. Obstetric complications occurred in 30 pregnancies (7.4%), and 46 pregnancies (11.3%) experienced at least one cardiac or obstetric complication (table 2).
Table 2. Maternal cardiac and obstetric complications (N=407 pregnancies).
| Complications | N |
|---|---|
| A. Maternal cardiac complications within 12 months postpartum | |
| Cardiac complication (primary composite endpoint)* | 17 |
| Heart failure requiring treatment or hospitalisation | 7 |
| Symptomatic arrhythmia requiring treatment | 5 |
| Non-sustained ventricular tachycardia | 2 |
| Sustained ventricular tachycardia | 1 |
| Atrial fibrillation/flutter | 2 |
| Thromboembolic event (DVT, PE or valve thrombosis) | 2 |
| Stroke or transient ischaemic attack | 1 |
| New pacemaker implantation | 1 |
| Cardiac surgical or catheter-based intervention | 1 |
| Cardiac arrest | 0 |
| Cardiac-related death within 12 months postpartum | 0 |
| B. Obstetric complications (during pregnancy and postpartum) | |
| Obstetric complication (secondary endpoint)† | 30 |
| Postpartum haemorrhage | 26 (6.4%) |
| Blood transfusion | 22 (5.4%) |
| Hypertensive disorder of pregnancy (any) | 14 (3.4%) |
| Gestational hypertension | 8 (1.7%) |
| Preeclampsia | 4 (1.0%) |
| Haemorrhagic shock | 2 (0.5%) |
| Disseminated intravascular coagulopathy | 3 (0.7%) |
| Pulmonary embolism (postpartum)‡ | 1 (0.2%) |
| Sepsis | 1 (0.2%) |
| Unplanned hysterectomy or re-laparotomy | 3 (0.7%) |
| C. Composite and delivery complications | |
| Composite endpoint (cardiac or obstetric complication) | 46 (11.3%) |
| mWHO class I | 12 (9.6%) |
| mWHO class II | 23 (10.1%) |
| mWHO class III | 11 (20.4%) |
Data are presented as n or n (%) unless stated otherwise.
Primary cardiac endpoint: composite of new or worsening heart failure requiring treatment, sustained symptomatic arrhythmia requiring treatment, thromboembolic event (DVT, pulmonary embolism or valve thrombosis), stroke or TIA, pacemaker implantation, cardiac catheter-based or surgical intervention, cardiac arrest, or death occurring during pregnancy or within 12 months postpartum. Event counts by individual risk model definition use the predictor and outcome definitions of each respective original derivation study.
Secondary obstetric endpoint: composite of postpartum haemorrhage (>500 mL for vaginal delivery or >1000 mL for caesarean delivery), haemorrhagic shock, hypertensive disorder of pregnancy (gestational hypertension, pre-eclampsia, HELLP syndrome), disseminated intravascular coagulopathy, postpartum pulmonary embolism, sepsis or unplanned hysterectomy/re-laparotomy.
No pulmonary embolism event contributed to both composite endpoints.
DVT, deep vein thrombosis; EBL, estimated blood loss; HELLP, hemolysis, elevated liver enzymes and low platelets; mWHO, modified WHO; PE, pulmonary embolism; TIA, transient ischemic attack.
Risk distribution and complications by mWHO classification
The distribution of pregnancies and maternal complications stratified by mWHO risk class is summarised in table 3A. A clear gradient of cardiac risk was observed across mWHO classes: the cardiac complication rate was 1.6% in class I, 3.5% in class II and 13.0% in class III (p=0.002). In contrast, obstetric complication rates did not increase significantly across classes (8.0%, 7.0% and 7.4% for classes I, II and III, respectively), indicating that the mWHO classification captures cardiac but not obstetric risk.
Table 3. Distribution of women and complications according to (A) mWHO maternal cardiac risk class, (B) ZAHARA risk group and (C) CARPREG II score.
| mWHO class | N | % of cohort | Cardiac complications, n (%) | Obstetric complications, n (%) | Any complication*, n (%) |
|---|---|---|---|---|---|
| A | |||||
| I | 125 | 30.7 | 2 (1.6%) | 10 (8.0%) | 12 (9.6%) |
| II | 228 | 56.0 | 8 (3.5%) | 16 (7.0%) | 23 (10.1%) |
| III | 54 | 13.3 | 7 (13.0%) | 4 (7.4%) | 11 (20.4%) |
| Total | 407 | 100.0 | 17 (4.2%) | 30 (7.4%) | 46 (11.3%) |
| ZAHARA risk group | N | % of cohort | Cardiac complications, n (%) | Obstetric complications, n (%) | Any complication, n (%) |
|---|---|---|---|---|---|
| B | |||||
| Low (<2) | 342 | 84.0 | 10 (2.9%) | 26 (7.6%) | 35 (10.2%) |
| Intermediate (2–3.9) | 49 | 12.0 | 5 (10.2%) | 3 (6.1%) | 8 (16.3%) |
| High (≥4) | 16 | 3.9 | 2 (12.5%) | 1 (6.2%) | 3 (18.8%) |
| Total | 407 | 100.0 | 17 (4.2%) | 30 (7.4%) | 46 (11.3%) |
| CARPREG category | N | % of cohort | Cardiac complications, n (%) | Obstetric complications, n (%) | Any complication, n (%) |
|---|---|---|---|---|---|
| C | |||||
| 0 points | 204 | 50.1 | 5 (2.5%) | 14 (6.9%) | 19 (9.3%) |
| 1 point | 84 | 20.6 | 1 (1.2%) | 9 (10.7%) | 10 (11.9%) |
| ≥2 points | 119 | 29.2 | 11 (9.2%) | 7 (5.9%) | 17 (14.3%) |
| Total | 407 | 100.0 | 17 (4.2%) | 30 (7.4%) | 46 (11.3%) |
ZAHARA score groups: low<2; intermediate 2–3.9; high≥4.
CARPREG II categories: 0 points; 1 point; ≥2 points.
Any maternal complication=cardiac or obstetric complication. Cardiac complications defined as any cardiac event within 1 year postpartum. Obstetric complications defined as postpartum complications per model definitions.
CARPREG II, Cardiac Disease in Pregnancy Study II; ZAHARA, Zwangerschap bij Aangeboren HARtAfwijkingen (Pregnancy in Women With Congenital Heart Disease).
Risk distribution and complications by ZAHARA score
Table 3B presents complications by ZAHARA risk group. Most patients (84.0%) were classified in the low-risk group (score<2), with an observed cardiac complication rate of 2.9%. Rates increased substantially in the intermediate (2–3.9) and high (≥4) groups, at 10.2% and 12.5%, respectively, consistent with an escalating risk gradient. Composite complication rates followed a similar pattern: 10.2%, 16.3% and 18.8% for the low, intermediate and high groups, respectively (p=0.002).
Risk distribution and complications by CARPREG II score
Among the three score groups, patients with a CARPREG II score of ≥2 had the highest cardiac complication rate at 9.2% (11/119), compared with 2.5% (5/204) for those with a score of 0 and 1.2% (1/84) for those with a score of 1 (p trend=0.007). Notably, the score-1 group had a slightly lower cardiac event rate than the score-0 group, although absolute numbers were small. Obstetric complication rates did not follow a consistent pattern across CARPREG II groups (6.9%, 10.7% and 5.9% for scores 0, 1 and ≥2, respectively), suggesting limited discriminatory value for non-cardiac endpoints.
Comparison of predicted versus observed risk and model discrimination
All three risk models overestimated the observed maternal cardiac complication rate of 4.18%. Predicted rates were 12.78% for CARPREG II, 8.3% for mWHO and 6.7% for ZAHARA, corresponding to O:E ratios of 0.33, 0.50 and 0.62 (table 4, figure 1). Among the three models, ZAHARA demonstrated the closest calibration and numerically highest discrimination (AUC 0.74, 95% CI 0.61 to 0.85), compared with CARPREG II (AUC 0.69, 95% CI 0.53 to 0.83) and mWHO (AUC 0.69, 95% CI 0.55 to 0.80); CIs overlapped substantially across all three scores.
Table 4. Predicted versus observed cardiac complication rates and model discrimination.
| Model | Predicted risk (%) | Observed risk (%) | O:E ratio | AUC (95% CI) |
|---|---|---|---|---|
| CARPREG II | 12.78% | 4.18% | 0.33 | 0.69 (0.53 to 0.83) |
| mWHO | 8.30% | 4.18% | 0.50 | 0.69 (0.55 to 0.80) |
| ZAHARA | 6.70% | 4.18% | 0.62 | 0.74 (0.61 to 0.85) |
O:E ratio = observed events/expected events (ratio closer to 1.0 indicates better calibration). AUC = area under the receiver operating characteristic curve for cardiac complications within 1 year postpartum.
AUC, area under the receiver operating characteristic curve; CARPREG II, Cardiac Disease in Pregnancy Study II; mWHO, modified WHO; mWHO, modified WHO; O:E, observed-to-expected; ZAHARA, Zwangerschap bij Aangeboren HARtAfwijkingen (Pregnancy in Women With Congenital Heart Disease).
Figure 1. Distribution of patients and predicted versus observed cardiac complications across CARPREG II, ZAHARA and modified WHO (mWHO) risk categories. Grey bars represent the number of patients within each risk category for the CARPREG II, ZAHARA and mWHO models. Green diamonds indicate the observed cardiac complication rate (% within each class), while red squares indicate the predicted complication risk according to each model. The red line connects predicted values to illustrate the trend across score categories. AUC, area under the curve; CARPREG II, Cardiac Disease in Pregnancy Study II; ROC, receiver operating characteristic; ZAHARA, Zwangerschap bij Aangeboren HARtAfwijkingen — Pregnancy in Women With Congenital Heart Disease.
Discussion
In this retrospective cohort of 407 pregnancies managed over 14 years under a standardised multidisciplinary protocol, several principal findings emerged. First, all three risk scores substantially overestimated the observed cardiac complication rate of 4.18%, with predicted rates of 12.78% (CARPREG II), 8.3% (mWHO) and 6.7% (ZAHARA), yielding O:E ratios of 0.33, 0.50 and 0.62, respectively. No maternal deaths occurred within 12 months of delivery. Second, despite this calibration drift, all three scores preserved a clinically meaningful cardiac risk gradient across risk categories, confirming their continued value for relative risk stratification. Third, ZAHARA provided the closest calibration and highest discrimination for cardiac complications (AUC 0.74), followed by CARPREG II (AUC 0.69) and mWHO (AUC 0.69). Fourth, none of the three scores predicted obstetric complications, with AUC values near 0.5 across all models and flat obstetric event rates.
The cardiac complication rate of 4.18% among our cohort is among the lowest reported in the validation literature and is consistent with complications expected from a contemporary specialist programme. Balci et al reported 10.3% in a Dutch multicentre cohort; Kim et al observed 19.7% in a North American tertiary centre; and Denayer et al reported 8% in a Belgian single-centre study.14 15 18 Several factors likely explain our low event rate. Our centre operates a formal multidisciplinary pregnancy heart team with standardised preconception counselling, an approach consistently associated with improved complications in large registry studies including ROPAC.21 The derivation cohorts of these scores span eras different from contemporary practice: ZAHARA drew from patients enrolled between 1966 and 2005, and CARPREG II from 1994 to 2011, periods predating routine fetal echocardiography, catheter-based valve interventions and the widespread adoption of formal multidisciplinary pregnancy heart teams.5 Both obstetric and cardiac care have advanced substantially over these intervals, reducing event rates across the board.
All three risk stratification tools overestimated the observed complication rate in our cohort. This is consistent with the concept of calibration drift, which is the tendency of prediction models to overestimate risk when applied in settings where background event rates are lower than those of the derivation population. Similar findings have been reported across the external validation literature, suggesting this is a recognised limitation of these tools rather than a finding specific to our study and centre.15–18 However, given the modest size of our cohort and the small number of events, this observation should be interpreted cautiously. These findings suggest that the absolute risk estimates provided by these tools may be less reliable in contemporary specialist settings than their relative risk rankings. Clinicians may, therefore, find these tools most useful for identifying which patients warrant closer surveillance, rather than as a source of precise individual probability estimates.
The near equivalence of AUC values across all three scores (0.69, 0.69 and 0.74 for CARPREG II, mWHO and ZAHARA respectively) is consistent with published external validation literature. This convergence suggests that discrimination, the ability to rank patients by relative risk, is similarly limited across all three tools in contemporary populations, and that the meaningful differentiator between them is calibration, not discrimination. ZAHARA’s closer calibration in our cohort may partly reflect population alignment: ZAHARA was derived exclusively in a CHD population, and our cohort is CHD-predominant. CARPREG II, by contrast, included approximately 36% non-CHD diagnoses in its derivation cohort, and mWHO encompasses all forms of cardiovascular disease including acquired and arrhythmic conditions. The better fit of ZAHARA may, therefore, reflect diagnostic case-mix rather than inherent methodological superiority, and this interpretation should temper strong conclusions about ZAHARA’s generalisability to populations with a higher proportion of acquired cardiac disease.
These findings also highlight a broader limitation of score-based risk stratification in this population. Each woman with cardiac disease presents a unique combination of lesion type and severity, functional reserve, obstetric history and individual circumstances that no single score fully captures. The consistent inability of all three models to predict obstetric risks, alongside their tendency to overestimate cardiac risk in contemporary specialist settings, suggests that numerical scores are best understood as one input into a broader clinical assessment rather than a standalone decision-making tool. An experienced multidisciplinary pregnancy heart team, integrating cardiology, maternal-fetal medicine and anaesthesiology, might be better positioned to perform the individualised holistic assessment that population-derived scores cannot replicate.
None of the three scores incorporates obstetric risk predictors since these tools were derived and validated exclusively against cardiac endpoints. It is, therefore, unsurprising that none predicted obstetric complications in our cohort, with AUC values near 0.5 across all models and flat obstetric event rates across mWHO classes. Our data make a compelling cumulative case: cardiac and obstetric risks are independent dimensions that cannot be captured by refinement of existing cardiac-only scores. Taken together, these observations suggest that cardiac and obstetric risks may represent distinct clinical dimensions that existing cardiac-focused tools are not designed to capture.
Several limitations merit acknowledgement. The retrospective single-centre design introduces selection bias and mWHO class IV patients were not included in this cohort. The absolute number of primary cardiac events limits statistical power for subgroup analyses and temporal heterogeneity across 14 years may have introduced unmeasured confounding in management practices. Single-centre findings may not generalise to settings with different disease spectra or cardiac event rates. In addition, the modified WHO classification used in this study reflects the version in use during the study period; the recently published mWHO 2.0 update (2025) was not evaluated, as it postdates data collection. Notably, several variables incorporated in mWHO 2.0 overlap with those included in CARPREG II, which was validated in this cohort, providing partial indirect coverage of the updated risk framework
Conclusion
In this contemporary single-centre cohort, all three cardiac risk stratification scores overestimated observed maternal cardiac complication rates, suggesting calibration drift in a specialist setting with low background event rates. A cardiac risk gradient was nonetheless preserved across all models, consistent with their continued utility for relative risk stratification, though absolute risk estimates should be interpreted with caution. This finding requires confirmation in larger and more diverse populations.
Supplementary material
Acknowledgements
The authors thank the staff of the Helen Schneider Hospital for Women and the Cardio-Obstetrics Clinic at Rabin Medical Center for their support in data collection.
Footnotes
Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Provenance and peer review: Not commissioned; externally peer-reviewed.
Patient consent for publication: Not applicable.
Ethics approval: The study was approved by the Institutional Review Board of Rabin Medical Center (RMC-0115-19). The requirement for individual patient consent was waived given the retrospective, observational and non-interventional design. All procedures were conducted in accordance with the principles of the Declaration of Helsinki.
Data availability free text: The data that support the findings of this study are available from the corresponding author upon reasonable request. The data are not publicly available due to privacy and ethical restrictions.
Data availability statement
Data sharing not applicable as no datasets generated and/or analysed for this study.
References
- 1.Hunter S, Robson SC. Adaptation of the maternal heart in pregnancy. Heart. 1992;68:540–3. doi: 10.1136/hrt.68.12.540. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Cornette J, Ruys TPE, Rossi A, et al. Hemodynamic adaptation to pregnancy in women with structural heart disease. Int J Cardiol. 2013;168:825–31. doi: 10.1016/j.ijcard.2012.10.005. [DOI] [PubMed] [Google Scholar]
- 3.Siu SC, Sermer M, Colman JM, et al. Prospective multicenter study of pregnancy outcomes in women with heart disease. Circulation. 2001;104:515–21. doi: 10.1161/hc3001.093437. [DOI] [PubMed] [Google Scholar]
- 4.Roos-Hesselink JW, Ruys TPE, Stein JI, et al. Outcome of pregnancy in patients with structural or ischaemic heart disease: results of a registry of the European Society of Cardiology. Eur Heart J. 2013;34:657–65. doi: 10.1093/eurheartj/ehs270. [DOI] [PubMed] [Google Scholar]
- 5.Drenthen W, Boersma E, Balci A, et al. Predictors of pregnancy complications in women with congenital heart disease. Eur Heart J. 2010;31:2124–32. doi: 10.1093/eurheartj/ehq200. [DOI] [PubMed] [Google Scholar]
- 6.Bottega N, Malhamé I, Guo L, et al. Secular trends in pregnancy rates, delivery outcomes, and related health care utilization among women with congenital heart disease. Congenit Heart Dis. 2019;14:735–44. doi: 10.1111/chd.12811. [DOI] [PubMed] [Google Scholar]
- 7.De Backer J, Haugaa KH, Hasselberg NE, et al. 2025 ESC Guidelines for the management of cardiovascular disease and pregnancy. Eur Heart J. 2025;46:4462–568. doi: 10.1093/eurheartj/ehaf193. [DOI] [PubMed] [Google Scholar]
- 8.Regitz-Zagrosek V, Roos-Hesselink JW, Bauersachs J, et al. 2018 ESC Guidelines for the management of cardiovascular diseases during pregnancy. Eur Heart J. 2018;39:3165–241. doi: 10.1093/eurheartj/ehy340. [DOI] [PubMed] [Google Scholar]
- 9.Canobbio MM, Warnes CA, Aboulhosn J, et al. Management of Pregnancy in Patients With Complex Congenital Heart Disease: A Scientific Statement for Healthcare Professionals From the American Heart Association. Circulation. 2017;135:e50–87. doi: 10.1161/CIR.0000000000000458. [DOI] [PubMed] [Google Scholar]
- 10.Silversides CK, Grewal J, Mason J, et al. Pregnancy Outcomes in Women With Heart Disease: The CARPREG II Study. J Am Coll Cardiol. 2018;71:2419–30. doi: 10.1016/j.jacc.2018.02.076. [DOI] [PubMed] [Google Scholar]
- 11.Lu C-W, Shih J-C, Chen S-Y, et al. Comparison of 3 Risk Estimation Methods for Predicting Cardiac Outcomes in Pregnant Women With Congenital Heart Disease. Circ J. 2015;79:1609–17. doi: 10.1253/circj.CJ-14-1368. [DOI] [PubMed] [Google Scholar]
- 12.Fu Q, Lin J. Predictive accuracy of three clinical risk assessment systems for cardiac complications among Chinese pregnant women with congenital heart disease. Intl J Gynecology & Obste. 2016;134:140–4. doi: 10.1016/j.ijgo.2016.02.010. [DOI] [PubMed] [Google Scholar]
- 13.Pijuan-Domènech A, Galian L, Goya M, et al. Cardiac complications during pregnancy are better predicted with the modified WHO risk score. Int J Cardiol. 2015;195:149–54. doi: 10.1016/j.ijcard.2015.05.076. [DOI] [PubMed] [Google Scholar]
- 14.Balci A, Sollie-Szarynska KM, van der Bijl AGL, et al. Prospective validation and assessment of cardiovascular and offspring risk models for pregnant women with congenital heart disease. Heart. 2014;100:1373–81. doi: 10.1136/heartjnl-2014-305597. [DOI] [PubMed] [Google Scholar]
- 15.Kim YY, Goldberg LA, Awh K, et al. Accuracy of risk prediction scores in pregnant women with congenital heart disease. Congenit Heart Dis. 2019;14:470–8. doi: 10.1111/chd.12750. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kha R, Melov SJ, Alahakoon TI, et al. Predicting cardiac and pregnancy outcomes in women with adult congenital heart disease using the Anatomic and Physiological (AP) Classification System: How much does physiology matter? Int J Cardiol Congenit Heart Dis. 2024;15:100486. doi: 10.1016/j.ijcchd.2023.100486. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.van Hagen IM, Roos-Hesselink JW. Pregnancy in congenital heart disease: risk prediction and counselling. Heart. 2020;106:1853–61. doi: 10.1136/heartjnl-2019-314702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Denayer N, Troost E, Santens B, et al. Comparison of risk stratification models for pregnancy in congenital heart disease. Int J Cardiol. 2021;323:54–60. doi: 10.1016/j.ijcard.2020.09.033. [DOI] [PubMed] [Google Scholar]
- 19.Levels of Maternal Care: Obstetric Care Consensus No, 9. Obstetrics & Gynecology. 2019;134:e41–55. doi: 10.1097/AOG.0000000000003383. [DOI] [PubMed] [Google Scholar]
- 20.Iluz-Freundlich D, Vikhorova Y, Azem K, et al. Peripartum anesthesia management and outcomes of patients with congenital heart disease: a single-center retrospective analysis (2009-2023) Int J Obstet Anesth. 2024;60:104241. doi: 10.1016/j.ijoa.2024.104241. [DOI] [PubMed] [Google Scholar]
- 21.Roos-Hesselink J, Baris L, Johnson M, et al. Pregnancy outcomes in women with cardiovascular disease: evolving trends over 10 years in the ESC Registry Of Pregnancy And Cardiac disease (ROPAC) Eur Heart J. 2019;40:3848–55. doi: 10.1093/eurheartj/ehz136. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data sharing not applicable as no datasets generated and/or analysed for this study.