Recent studies suggest that fracture risk is associated with increased undercarboxylated osteocalcin. Methods use differences in binding of undercarboxylated and fully carboxylated osteocalcin to hydroxyapatite or barium sulfate. We evaluated these methods and found that results varied with the amount and preparation of the salts. Furthermore, patient samples with differing amounts of total osteocalcin could not be directly compared. Errors in the determination of undercarboxylated osteocalcin were minimized by expressing data as the percent of the total osteocalcin in the sample, and correcting for the basal level of osteocalcin using a polynomial equation derived from multiple binding curves. Errors from 5-15% in estimation of undercarboxylated osteocalcin were observed without both of these corrections. When differing types of assays were employed (RIA, intact, N-terminal), results also were affected. In normal adults and children and in patients on long-term warfarin therapy, the percent osteocalcin not bound to hydroxyapatite was lower when measured with an intact assay than by a polyclonal RIA. Differences were related to the amount of N-terminal osteocalcin fragments, which had low affinity for hydroxyapatite and resulted in variable overestimation of undercarboxylated osteocalcin. In a kit specific for uncarboxylated osteocalcin, we found good discrimination between carboxylated and uncarboxylated intact osteocalcin. However, the assay detected large osteocalcin fragments and overestimated their concentration by up to 350%. Values for uncarboxylated osteocalcin were not different in patients on coumadin compared with normal adults with this kit, but when normalized to the total intact osteocalcin, percent uncarboxylated osteocalcin was greater in patients on coumadin than in controls, as would be expected. Kit values for uncarboxylated osteocalcin in normal children were higher than intact values in the same subject, because of the increased reactivity of the kit toward circulating fragments that were elevated in children. Thus, for estimation of undercarboxylated osteocalcin, care must be taken to standardize the hydroxyapatite or barium sulfite used for binding, to correct for the basal level of osteocalcin in the sample, to use immunoassays that do not detect small fragments, and to express the results as the percent of the total osteocalcin in the sample. Without these precautions, the assessment of undercarboxylated osteocalcin is not reliable.