In cardiac rehabilitation, different endurance exercises such as walking and cycling are often performed. The training intensity for these modes is determined from a single treadmill or bicycle test by ventilatory threshold (VT). In this study, differences of VT between walking and cycling and agreement of VT between training modes were assessed in cardiac patients. A total of 46 cardiac rehabilitation patients (mean age 59.5+/-8.4 years, 45 males) (31 untrained and 15 trained) completed a maximal exercise test on bicycle and treadmill, with breath-by-breath analysis of oxygen uptake (VO2), carbon dioxide production and expiratory volume. VT was determined by V-slope method. Correlations of VT and VO2peak were calculated between exercise modes. Bland-Altman plots were made for determining VT agreement between modes. VT was significantly different between walking and cycling in trained patients (P<0.05), but not in untrained patients (P>0.05). When untrained and trained patients were compared, VT correlation was lower (r=0.50) in the former group, as compared to the latter group (r=0.78). Also, Bland-Altman plots showed smaller limits of agreement for VT in trained (2 SD -1.6 to 7.8 ml/min/kg), as compared to untrained patients (2 SD -7.0 to 9.6 ml/min/kg). In trained patients, VT correlates well between training methods, but is highly exercise mode specific. In untrained patients, VT is not exercise mode specific, but the VT has a low correlation between training modes. This study shows that VT should be assessed by the appropriate exercise model for determining exercise intensity in cardiac rehabilitation.