Incremental and independent value of cardiopulmonary exercise test measures and the Seattle Heart Failure Model for prediction of risk in patients with heart failure

Multivariable risk scores and exercise measures are well-validated risk prediction methods. Combining information from a functional evaluation and a risk model may improve accuracy of risk predictions. We analyzed whether adding exercise measures to the Seattle Heart Failure Model (SHFM) improves risk prediction accuracy in systolic heart failure.We used a sample of patients from the Heart Failure and A Controlled Trial Investigating Outcomes of Exercise TraiNing (HF-ACTION) study (http://www.clinicaltrials.gov; unique identifier: NCT00047437) to examine the addition of peak oxygen consumption, expired volume per unit time/volume of carbon dioxide slope, 6-minute walk distance, or cardiopulmonary exercise duration to the SHFM. Multivariable Cox proportional hazards models were used to test the association between the combined end point (death, left ventricular assist device, or cardiac transplantation) and the addition of exercise variables to the SHFM.The sample included 2,152 patients. The SHFM and all exercise measures were associated with events (all p0.0001) in proportional hazards models. There was statistically significant improvement in risk estimation when exercise measures were added to the SHFM. However, the improvement in the C index for the addition of peak volume of oxygen consumption (+0.01), expired volume per unit time/volume of carbon dioxide slope (+0.02), 6-minute walk distance (-0.001), and cardiopulmonary exercise duration (+0.001) to the SHFM was small or slightly worse than the SHFM alone. Changes in risk assignment with the addition of exercise variables were minimal for patients above or below a 15% 1-year mortality.Exercise performance measures and the SHFM are independently useful for predicting risk in systolic heart failure. Adding cardiopulmonary exercise testing measures and 6MWD to the SHFM offers only minimal improvement in risk reassignment at clinically meaningful cut points.