We assessed the analytical and biological variation of pyridinium crosslinks in early morning, 2 h fasting, and 24 h urine specimens from 14 healthy adults over a 1 month period. The results were expressed both in terms of pyridinoline concentration and pyridinoline/creatinine ratio. The data obtained were used to select the optimum specimen for clinical purposes. We found that: (a) early morning specimens are preferred; (b) results should be expressed as pyridinoline/creatinine ratio; (c) reference intervals should be stratified according to gender; (d) the necessary analytical imprecision (CV < or = 9%), derived from biological variation, is not easily achieved by current methods; (e) the difference between serial results from an individual must be > 50% to be statistically significant; and (f) assessment of risk for osteoporotic fracture by means of the pyridinium crosslink assay would, in a significant number of patients, require analysis of multiple urine specimens.