Bone turnover markers are subject to day-to-day and within-day variability, which may influence clinical interpretation. We examined the effect of fasting vs. feeding on the concentration and between-day variability of several markers. Twenty healthy premenopausal women were studied on 10 consecutive weekdays. Subjects were studied either in the fasting (no breakfast) or fed (breakfast at 08:00 h) state on alternate days, and were randomized to begin either fasting or fed. Two hour urine collections were obtained each day between 08:00 h and 10:00 h, and blood samples were collected daily at 09:00 h. The N-telopeptide cross-link of type I collagen in urine (uNTX) and serum (sNTX), the C-telopeptide in urine (uCTX) and serum (sbetaCTX), and immunoreactive free deoxypyridinoline (uifDPD) in urine were measured as resorption markers. Procollagen type I N-terminal propeptide (PINP), osteocalcin (OC), and bone alkaline phosphatase (bone ALP) were measured as formation markers. All bone formation and resorption markers were significantly lower in the fed state with the exception of bone ALP. The magnitude of the decrease ranged from 3.8 +/- 0.9% for PINP (p < 0.0001) to 17.8 +/- 2.6% (p < 0.0001) for sbetaCTX. Measurement variability was partitioned into analytical variability based on replicate assays (CV(a)) and within-subject variability (CV(i)). The CV(i) was greater (p < 0.05) for some markers in the fasting state (uifDPD, uNTX, and sNTX) but greater in the fed state for other markers (OC and sbetaCTX). In conclusion, the clinical impact of feeding vs. fasting is small with the exception of sbetaCTX; however, in clinical practice, collection of samples in the fasting state may be necessary to minimize the unpredictable effects of feeding. The mechanism of the acute effect of feeding on bone turnover remains uncertain.