Mixed venous oxy-hemoglobin saturation (M(v)O2) is a physiological variable with several features that might be desirable as a control parameter for rate adaptive pacing. Despite these desirable characteristics, the long- term reliability of the M(v)O2 sensor in viva is uncertain. We, therefore, designed a study to prospectively evaluate the long-term performance of a permanently implanted M(v)O2 saturation sensor in patients requiring VVIR pacing. Under an FDA approved feasibility Study, eight patients were implanted with a VVIR pulse generator and a right ventricular pacing lead incorporating an M(v)O2 sensor. In order to accurately assess long-term stability of the sensor, patients underwent submaximal treadmill exercise using the Chronotropic Assessment Exercise Protocol (CAEP) at 2 weeks, 6 weeks, and 3, 6, 9, 12, 18, and 24 months following pacemaker implantation. Paired maximal exercise testing using the CAEP was also performed with the pacing system programmed to the VVI and VVIR modes in randomized sequence with measurement of expired gas exchange after 6 weeks and 12 months of follow-up. During maximal treadmill exercise the peak exercise heart rate (132 ± 9 vs 71.5 ± 5 beats/min, P < 0.00001) and maximal rate of oxygen consumption (1,704 ± 633 vs 1382 ± 407 mL/min, P = 0.01) were significantly greater in the VVIR than in the VVI pacing mode. Similarly, the duration of exercise was greater in the VVIR than the VVI pacing mode (8.9 ± 3.6 min vs 7.6 ± 3.7 min, P = 0.04). The resting M(v)O2 and the M(v)O2 at peak exercise were similar in the VVI and VVIR pacing modes (P = NS). However, the M(v)O2 at each comparable treadmill exercise stage was significantly higher in the VVIR mode than in the VVI mode (CAEP stage 1 (P = 0.005), stage 2 (P = 0.04), stage 3 (P = 0.008), and stage 4 (P = 0.04). The correlation between M(v)O2 and oxygen consumption (VO2) was excellent (r = -0.93). Telemetry of the re, flectance of red and infrared light and M(v)O2 in the right ventricle during identical exercise workloads revealed no significant change over the first 12 months of follow-up (ANOVA, P = NS). The chronotropic response to exercise remained proportional to VO2 in all patients over the first 12 months of follow-up. The time course of change in M(v)O2 during maximal exercise was significantly faster than for VO2. At the 18- and 24- month follow-up exercise tests, a significant deterioration of the sensor signal with attenuation of chronotropic response was noted for 4 of the 8 subjects with replacement of the pacing system required in one patient because of lack of appropriate rate modulation. Rate modulated VVIR pacing controlled by right ventricular M(v)O2 provides a chronotropic response that is highly correlated with VO2. This parameter responds rapidly to changes in workload with kinetics that are more rapid than those of VO2. Appropriate rate modulation provides a higher M(v)O2 at identical workloads than does VVI pacing. Although the M(v)O2 sensor remains stable and accurate over the first year following implantation, significant deterioration of the signal occurs by 18-24 months in many patients.