In research settings, longitudinal measurements of bone mineral density have become an integral part of the assessment of patients with metabolic skeletal disorders. To adequately utilize longitudinal measures, confidence in the long-term precision of the measurement technique must be very high. Dual-energy x-ray absorptiometry (DXA) has become commonly utilized in this context, and to better understand its long-term precision and to develop quality assurance protocols for its use, we examined the performance of eight DXA machines over a 3 year period. Anthropomorphic spine phantoms were measured frequently on each machine during the period of observation, and precision was estimated from the consistency of these determinations. Overall precision was excellent (mean longitudinal coefficient of variation, 0.4%). Nevertheless, by using a series of objective quality control criteria, small alterations in the performance of each machine were identified (mean number of changes, 4.6 in 3 years; mean magnitude, 0.0039 g/cm2, or 0.4%). The cumulative effects of those changes were sufficient to cause a significant (albeit minor) change in the regression slopes (phantom mineral density versus time) of most machines. The same quality control rules were also used to quantitate the magnitude of change and to adjust retrospectively machine performance during the period of observation, such that alterations were minimal and regression slopes were not significantly different from zero. Although the precision of DXA is excellent, alterations in machine function must be anticipated during longitudinal use. The development of quality control protocols provides the means to detect change objectively and to adjust for alterations in performance during the course of longitudinal evaluations.