We developed a sensitive and specific two-site radioimmunoassay (IRMA) for human osteocalcin using human osteocalcin as a standard and two monoclonal antibodies raised against human osteocalcin purified from human cortical bone, a solid-phase anti-25-37 region and a tracer anti-5-13 sequence of the molecule. A wide range of osteocalcin levels (up to 300 ng/ml) can be measured with a sensitivity of 0.4 ng/ml. The intra- and interassay coefficients of variation are less than 4 and 6%, respectively. The recovery of human osteocalcin from serum samples ranges from 96 to 103%. IRMA was linear for serial sample dilutions in a wide range of serum osteocalcin levels, even in patients with chronic renal failure on hemodialysis. Depletion of serum in intact osteocalcin demonstrated that IRMA detects, in addition to the intact peptide, a large N-terminal midregion fragment that represents about 50% of total osteocalcin levels in normals and patients with Paget's disease and up to 75% in patients with chronic renal failure. This large fragment, previously unrecognized because it cannot be distinguished from intact osteocalcin with gel filtration chromatography, is not generated in vitro by incubation of the serum up to 26 h. We measured osteocalcin in the serum of 309 healthy adults (180 men and 129 women, age range 20-95 years), 36 patients with Paget's disease, 12 patients with primary hyperparathyroidism, 70 patients with chronic renal failure on hemodialysis, and 10 patients on corticosteroid therapy, simultaneously with human IRMA and with a conventional radioimmunoassay (RIA) based on bovine reagents. A tight correlation (r = 0.889) was observed between the two assays in the normal population, but the values obtained with IRMA were about threefold higher (mean 23.3 +/- 10.5 versus 7.5 +/- 3.4 ng/ml) than those obtained with RIA. Reported as Z scores, that is, number of standard deviations from the predicted normal mean adjusted for sex and age, these two assays (IRMA and RIA) gave concordant results in patients with Paget's disease (4.05 +/- 6.21 versus 2.41 +/- 2.53), primary hyperparathyroidism (4.14 +/- 7.17 versus 2.13 +/- 2.28), chronic renal failure (25.32 +/- 24.49 versus 6.93 +/- 5.48), and glucocorticoid treatment (-1.48 +/- 0.78 versus -1.11 +/- 0.57). However, IRMA was more discriminant from controls for all these metabolic bone diseases because the absolute values of mean Z scores with IRMA were significantly higher than those obtained with the RIA (p < 0.05-0.0001).(ABSTRACT TRUNCATED AT 400 WORDS)