Osteoporosis, an increasingly common health problem, is the result of the continuous loss of bone during adult life, which is accelerated in women after the menopause. This additional bone loss is of variable length and intensity, resulting in 'fast-losers' and 'normal-losers'. It is important to detect those fast-losers with a normal to low bone mass in order to prevent the development of osteoporosis related fractures. . Biochemical markers of bone resorption are potential candidates for predicting future bone loss. I These markers, such as urinary hydroxyproline and the pyridinium cross-links, are higher in postmenopausal women, and decrease after treatment.' However, the predictive power of a single measurement of bone resorption markers in individuals is poor. A combination of two bone resorption markers with a bone formation marker, or combination of three values of the resorption marker deoxypyridinoline (DPD) obtained in samples taken at 6 months intervals, gave a potentially workable predictive modeJ.2 The low predictive power might be due, at least partly, to intra-individual variation. If the excretion of markers fluctuates from day to day, the measurement of one overnight value is a poor guideline in classifying subjects or monitoring therapy. Apart from the day-to-day variation, the excretion of markers show a diurnal rhythm,' requiring uniformly timed or 24 h urine collections. Frequently used samples are the 2 h fasting second-void and the morning first-void (or overnight) urine collections. The results are expressed per unit creatinine. We investigated the intra-individual variation for two recently launched immunoassays deoxypyridinoline (DPD, Pyrilinks-D, Metra Biosystems, Palo Alto, CA, USA) and N-terminal