BACKGROUND: The first (VT1) and second (VT2) ventilatory thresholds are landmarks to tailor exercise rehabilitation in coronary artery disease (CAD) or chronic heart failure (CHF). Methods allowing VT1 and VT2 determination based on heart rate variability (HRV) have been proposed but not yet evaluated in these patients. DESIGN: To determine the heart rate (HR) associated with VT1 and VT2 by three methods of HRV analyses. METHODS: Fourteen CHF and 24 CAD patients performed an exercise test on a cycle ergometer (10 -W every minute until exhaustion). VT1 and VT2 were determined with the \'respiratory equivalent\' method. HR at VT1 was determined with the standard deviation (SD) of R minus sign in boxR intervals (VT(SD)) and of the instantaneous beat-to-beat variability of the Poincare plot method (VT(Poincare)). HR at VT1 and VT2 was determined through a time-varying HRV analysis method (VT(TV1) and VT(TV2), respectively). RESULTS: HR at VT(SD) was significantly higher than HR at VT1. No significant differences were observed between HR at VT(Poincare), VT(TV1), and at VT1, nor between HR at VT(TV2) and VT2. HR at VT(SD), VT(Poincare), and VT1 were significantly correlated, but with a low r (2) value and a large mean HR difference. With the time-varying method, the mean HR difference was lower than 5% and the correlation coefficients were higher (especially for VT(TV2)). CONCLUSIONS: SD and Poincare plot methods lead to substantial inaccuracy in HR estimates. The time-varying HRV analysis led to strong correlation coefficients and low limits of agreement. Therefore, this method may be a promising, low-cost tool for non-invasive assessment of the ventilatory thresholds in cardiac disease.