Abstract 16405: Prediction of Mortality in Congenital Heart Disease using Cardiopulmonary Exercise Testing: Development and External Validation of a Prognostic Score in 6,071 Patients
Background: Parameters of cardiopulmonary exercise testing (CPET) were identified as strong predictors of mortality in congenital heart disease (CHD). However, such prognostic CPET criteria have not been externally validated.
Methods and Results: 6,071 consecutive CHD patients (28±14 yrs) underwent CPET at one of two tertiary centers between 1999 and 2013. Peak oxygen consumption (peak VO2), anaerobic threshold, VE/VCO2 slope, and heart rate response were measured. During a median follow-up of 3.6 years, 206 patients died. All CPET parameters were strongly related to survival (p<0.0001) and showed good discriminative ability on ROC analysis (AUC 0.76 - 0.84, p<0.0001 for all). A logistic regression model predicting absolute probability of death within 3 years post CPET was constructed based on the 2,860 patients of one center serving as derivation sample. The model was started with age and sex, and additional variables were incrementally added until no further improvement in discriminative and reclassification ability was seen (judged by Akaike’s information criterion and Integrated Discrimination Improvement). The final model included peak VO2 and disease complexity in addition to demographics (c-index 0.85, p<0.0001). The model was well calibrated in the derivation sample, but substantially overestimated mortality in the external validation cohort (n=3,211, Figure - red line). Prediction was substantially improved by adjusting for annual mortality in the target population sample (green line).
Conclusions: Parameters of CPET are strong predictors of survival in CHD pts. The validity of CPET criteria developed at one CHD center may, however, be low when applied to external cohorts. This problem can be rectified by correcting for the annual mortality rate in the target population to adjust for different case mix and to avoid systematically misestimating mortality risks.
- © 2013 by American Heart Association, Inc.