Different methods for the estimation of whole body protein synthesis (PS) in healthy adult humans were simultaneously compared in seven young adult subjects (6 males, 1 female) fed for 6 days a diet providing 1 g protein.kg-1.day-1 and approximately 188 kJ.kg-1.day-1. A 24-h intravenous tracer study with L-[1-13C]leucine was performed starting at 6 P.M. on day 6. During fasted (6 h) and fed (6 h) steady states, PS was estimated using an approach based on 13CO2 excretion (PSexcr) and on urinary nitrogen excretion data (corrected for changes in body urea pool). Simultaneously, we used the conventional two-pool model and plasma [13C]ketoisocaproate enrichment for estimating PS. The latter mean estimates of PS were significantly higher than PSexcr during fasting [861 +/- 58 (SD) vs. 663 +/- 160 mg protein.kg-1.6 h-1; P < 0.01] and feeding (985 +/- 63 vs. 779 +/- 127 mg protein.kg-1.6 h-1; P < 0.01) and were much less variable. In hourly small-meal feeding, urinary nitrogen excretion was not a reliable index of body protein oxidation when measured over short periods of 6 h, thereby introducing a lack of precision in PSexcr. We suggest that application of the 13CO2 technique to measure PS in humans is limited by the need for relatively prolonged experimental periods of urine collection and tracer infusion within a given physiological state.