The potential purpose of near-infrared spectroscopy (NIRS) as a clinical application in patients with chronic heart failure (CHF) is the identification of limitations in O delivery or utilization during exercise. The objective of this study was to evaluate absolute and relative test-retest reliability of skeletal muscle oxygenation measurements in patients with CHF. Thirty patients with systolic heart failure (left ventricular ejection fraction 31 ± 8%) performed 6-min constant-load cycling tests at 80% of the anaerobic threshold (AT) with tissue saturation index (TSI) measurement at the vastus lateralis. Tests were repeated after 10 ± 5 days to evaluate reliability. Absolute reliability was assessed with limits of agreement (LoA, expressed as bias ± random error) and coefficients of variation (CV) for absolute values (LoA range: 0·4 ± 6·2% to 0·6 ± 7·9%; CV range: 4·7-7·1%), amplitudes (LoA range -0·5 ± 5·8% to -0·7 ± 6·8%; CV range: 26·2-42·1%), onset and recovery kinetics (mean response times; LoA 0·4 ± 9·5 s, CV 23·5% and LoA -5·8 ± 50·8 s, CV 67·4% respectively) and overshoot characteristics (CV range 45·7-208·6%). Relative reliability was assessed with intraclass correlation coefficients for absolute values (range 0·74-0·90), amplitudes (range 0·85-0·92), onset and recovery kinetics (0·53 and 0·51, respectively) and overshoot characteristics (range 0·17-0·74). In conclusion, absolute reliability of absolute values and onset kinetics seems acceptable for serial within-subject comparison, and as such, for evaluation of treatment effects. Absolute reliability of amplitudes and recovery kinetics is considered unsatisfactory. Relative reliability of absolute values and amplitudes is sufficient for purposes of physiological distinction between patients with CHF. Despite lower relative reliability, kinetics may still be useful for clinical application.