Validation of the CARPREG II risk stratification model and the WHOm scale in pregnant women with heart disease

Abstract The physiological changes of pregnancy imply a state of haemodynamic stress, which increases the risk of maternal-fetal complications in women with cardiac pathology. Risk stratification models allow clinical decisions to be guided and optimal diagnostic, follow-up and management strategies to be established according to each estimated category in order to reduce adverse outcomes. Objective To validate the CARPREG II and WHOm risk prediction models in a population of pregnant women with heart disease. Materials and methods Validation and comparison study of a risk prediction model in a prospective cohort of pregnant women with heart disease assessed between 2016 and 2019 by a cardiobstetric team in an average income country. The CARPREG II score was established and patients were classified according to the WHOm risk scale. We assessed cardiovascular and perinatal outcomes and determined the calibration and level of discrimination of these tools. Results In a cohort of 328 pregnant women (27±7 years), 33% (n=110) had congenital heart disease, followed by arrhythmias in 30% (n=98), valvular pathologies in 14% (n=46) and cardiomyopathies in 9% (n=29). In 56% of the pregnancies, the route of delivery was caesarean section, 84% of these by obstetric indication. A cardiac event occurred in 15%, with left heart failure (5.3%) and arrhythmias (2.3%) being the most frequent. The frequency of maternal death of cardiac origin during the study was 1.6%. Neonatal outcomes occurred in 37% of gestations (preterm delivery (16%) and low weight for gestational age (8.4%)) and obstetric events in 12.5%: pregnancy-induced hypertension (9%) and postpartum haemorrhage (2.3%). NYHA functional class III-IV or cyanosis (OR 12 95% CI 3.1 - 46.4) and left ventricular dysfunction (LVEF <55%) (OR 3 95% CI 1 - 10.9) were the most statistically significant risk predictors. Discrimination of both models was adequate (AUC-ROC of 0.74 95% CI 0.64 - 0.84) for the CARPREG II risk index and 0.77 for the WHOm scale (95% CI 0.69 - 0.86) (Figure 2). Calibration is also good in the study population (Hosmer- Lemeshow goodness-of-fit 0.6 and 0.1, respectively). By including in the CARPREG II model the variables ejection fraction and pulmonary artery systolic pressure in their numerical and not dichotomised form, a discrete improvement in the predictive ability of the scale is evident (AUC-ROC 0.81 95% CI 0.71–0.91). Conclusions The CARPREG II and WHOm risk stratification models have good ability to discriminate the risk of adverse cardiac outcomes in pregnant women with heart disease and fit our population. To improve the predictive power of CARPREG II, the variables pulmonary hypertension and left ventricular dysfunction could be used numerically and not dichotomised as in the original model. Funding Acknowledgement Type of funding sources: Private hospital(s). Main funding source(s): San Vicente Foundation University Hospital, Cardio-obstetric Center