A new cardiopulmonary exercise testing score for predicting heart failure events

Abstract

Abstract Background Cardiopulmonary exercise test (CPET) provides several variables (V) that are strong predictors of events in patients (pts) with heart failure (HF). Purpose Our aim was to develop and evaluate the predictive power (PP) of a CPET-based score, comparing to that of the Heart Failure Survival Score – HFSS. Methods Retrospective evaluation of adult pts with HF submitted to CPET in a tertiary center. Pts were followed up for at least 1 year for the primary endpoint of cardiac death and urgent heart transplantation/ ventricular assist device implantation (MTV). The PP of several CPET V's was assessed using ROC curve analysis, which was used to define optimal threshold values for each V. A multiple regression analysis was performed to identify independent prognostic predictors and to determine the regression coefficient (β) for the Vs included in the model, each expressed dichotomously using the threshold value. According to β, a weight was assigned to each V and summed to calculate the composite score. ROC curves were compared using the Hanley and McNeil test. Results CPET was performed in 487 HF pts, with a mean age of 56±13 years, 79% were male. 46% of pts were of ischemic etiology, with a mean LVEF of 30±8%, a mean HFSS of 8.6±1.1 and a mean BNP value of 509±668pg/ml. The mortality rate during a mean follow-up of 21 months, was 19% (93pts) with 23 pts (5%) undergoing HT. The primary endpoint was reached by 55 (11%) pts. The variables with higher predictive power were OUES (AUC 0.796), ventilatory power (AUC 0.790), the partial pressure of end-tidal CO2 at the anaerobic threshold (PETCO2L – AUC of 0.787), the pVO2 (AUC 0.767) and heart rhythm (HR) during the test (AUC – 0.640). LVEF also presented a high predictive power with AUC of 0.755. The multivariate analysis revealed that pVO2, PETCO2L, AF, and LVEF were independent prognostic predictors. According to the β of these Vs, the equation was calculated as follows: (pVO2 × 2.194) + (PETCO2L × 1.545) + (LVEF × 1.134) + (HR × 1.055; 0 if AF, 1 if sinus rhythm). The score presented a high PP with an AUC of 0.866. A cut-off of 120 had an 83.6% sensitivity and a 75% specificity for MTV, and pts with a score value of <120 had a markedly lower rate of MTV (log-rank p<0.001). When compared to HFSS, our score presented a higher PP (0.866 vs 0.774, p=0.011). Conclusion A multivariable score based on readily available CPET Vs provides a simple, integrated and powerful method to predict HF events. Funding Acknowledgement Type of funding sources: None.