To examine the effect of caffeine ingestion on muscle glycogen utilization and the neuroendocrine axis during exercise, we studied 20 muscle glycogen-loaded subjects who were given placebo or caffeine (6 mg/kg) in a double blinded fashion 90 min before cycling for 2 h at 65% of their maximal oxygen consumption. Exercise-induced glycogen depletion in the thigh muscle was noninvasively measured by means of 13C nuclear magnetic resonance spectroscopy (NMR) spectroscopy, and plasma concentrations of substrates and neuroendocrine hormones, including beta-endorphins, were also assessed. Muscle glycogen content was increased 140% above normal values on the caffeine trial day (P < 0.001). After cycling for 2 h, caffeine ingestion was associated with a greater increase in plasma lactate (caffeine: +1.0 +/- 0.2 mmol/L; placebo, +0.1 +/- 0.2 mmol/L; P < 0.005), epinephrine (caffeine, +223 +/- 82 pg/mL; placebo, +56 +/- 26 pg/mL; P < 0.05), and cortisol (caffeine, +12 +/- 3 mg/mL; placebo, +2 +/- 2 mg/mL; P < 0.001) levels. However, plasma free fatty acid concentrations increased (caffeine, +814 +/- 133 mmol/L; placebo, +785 +/- 85 mmol/L; P = NS), and muscle glycogen content decreased (caffeine, -57 +/- 6 mmol/L muscle; placebo, -53 +/- 5 mmol/L muscle; P = NS) to the same extent in both groups. At the same time, plasma beta-endorphin levels almost doubled (from 30 +/- 5 to 53 +/- 13 pg/mL; P < 0.05) in the caffeine-treated group, whereas no change occurred in the placebo group. We conclude that caffeine ingestion 90 min before prolonged exercise does not exert a muscle glycogen-sparing effect in athletes with high muscle glycogen content. However, these data suggest that caffeine lowers the threshold for exercise-induced beta-endorphin and cortisol release, which may contribute to the reported benefits of caffeine on exercise endurance.