Background: The ventilatory threshold (VT) is usually determined by visual assessment of the point where the rate of elimination of carbon dioxide (VCO(2)) increases nonlinearly with respect to oxygen uptake (VO(2)) (the V-Slope method). We quantified the reliability of VT determination using data from a multicenter study in patients with heart failure.
Methods and results: The Fix-Heart Failure-5 study of cardiac contractility modulation enrolled 428 patients from 50 centers in the United States. Cardiopulmonary exercise tests were performed at baseline and 12, 24, and 50 weeks after randomization, which provided 1679 tests. The VT was determined from each test in a core laboratory by 2 independent readers. VT could not be determined for 276 tests (16.4% indeterminate). Inter-observer variability (quantified by the 95% limits of agreement, LoA, expressed as a percent of the mean value) was 20.2% between the 2 readers, with a coefficient of variation (CV) of 7.3%. Intra-observer variability was assessed by resubmitting (blinded) 179 tests to the same readers; the LoA was 24.7% for reader 1 and 16.9% for reader 2, with CVs of 6.1 and 8.9%, respectively. Ninety-one tests were submitted to 2 additional readers at a second core lab. Inter-observer variability in the second lab was 26.7% with a CV of 9.6%. Inter-laboratory variability was 21.4%, with a CV of 7.7%.
Conclusions: Inter-observer, intra-observer, and inter-site variation in determining the VT should be considered when using the VT as an end point in clinical trials of heart failure.