Comparison of risk stratification models for pregnancy in congenital heart disease

Comparison of risk stratification models for pregnancy in congenital heart disease

View Article

Denayer N, Troost E, Santens B, De Meester P, Roggen L, Moons P, Van Calsteren K, Budts W, Van De Bruaene A.
Int J Cardiol. 2020 Sep 12:S0167-5273(20)33814-6. doi: 10.1016/j.ijcard.2020.09.033. Online ahead of print.
PMID: 32931856

Take Home Points:

  • There are several risk stratification tools which are utilised to ‘predict’ adverse outcomes in women with congenital heart disease who become pregnant.
  • There are 4 commonly employed maternal risk predictive tools – CARPREG, CARPREG II, ZAHARA and the modified WHO score.
  • This study randomly selected 100 women and for each woman, the 4 risk prediction tools were calculated and summarised in a weighted average risk for each stratification model.
  • To evaluate accuracy of each model, the weighted average risk was plotted against the actual observed number of “cardiac events” as defined in the respective risk models.
  • Maternal adverse events occurred in 8% of the study population – all in patients with at least moderately complex CHD.
  • All the risk models over-estimated maternal cardiac risk but in this selected cohort, the ZAHARA risk model appeared to be a closer reflection of maternal risk.

Dr Damien Cullington

Commentary from Dr. Damien Cullington (Liverpool, UK), section editor of ACHD Journal Watch:  Neil’s Bohr rightly pointed out: “Prediction is very difficult, especially if it’s about the future” and predicting risks to women with CHD during pregnancy is no different. The CARPREG, CARPREG II, ZAHARA and modified WHO (mWHO) score are the main risk predictive scores employed to estimate risk in pregnant women with CHD – the most recently updated ESC guidelines for the management of cardiovascular disorders during pregnancy advocate the use of the mWHO score. Head-to-head comparison is challenging “since the different models have been constructed in different patient populations with different outcome measures”. This study sought to compare the 4 scores against each other to assess which predicts outcome best.

The characteristics of the patient cohort (n=100) are shown in Table 1. Over half of patients were primigravida (n=52) and the mean maternal age was 30.4 ± 3.7 yrs. 98% had a biventricular circulation. The four commonest CHD conditions were VSD (n=34), coarctation (n=16), TOF (n=15) and ASD (n=14). Pre-pregnancy, 97% had no symptoms of heart failure. Only 15% of the cohort were taking medication pre-pregnancy.


The actual rate of maternal cardiac complications was 8% – (Tables 2a and 2b). For each of the 4 models, a composite risk was calculated. The predictive accuracy for each risk model is shown in Table 3. There was one sudden cardiac death in a patient with Marfan syndrome 4 days post-partum but aside from this case, all CV complications occurred in patients with at least moderate complexity CHD – (4 TOF, 2 atrial switch and 1 Fontan). No complications were seen in patients with simple CHD lesions.

table 1 baselinetable 1 continued

table 2atable 2b


All models tended to over-estimate risk but ZAHARA was the closest in making an accurate prediction of CV event for this selected dataset. Work continues to refine how we predict risk. The authors suggest that work should also be focused on developing a risk score for patients referred and managed in cardio-obstetric centres.