Bone mineral density (BMD) measurement with dual-energy X-ray absorptiometry (DXA) is widely used to assess osteoporosis and monitor BMD in untreated and treated individuals. Systematic sources of error can occur with DXA scanners, including calibration drift. Publications suggest a drift tolerance in the range 0.5 % to 1.5 %, but are not evidence- based. The current study was performed to directly determine how varying degrees of simulated DXA calibration drift would affect misclassification of BMD change in routine clinical practice. Using data from the Manitoba Bone Density Program, we accessed results for 14,942 individuals age 40 years and older undergoing baseline and repeat fan-beam DXA measurements of the total hip within an interval of 1-5 years. A small amount of simulated drift (absolute 0.003 g/cm(2) or relative 0.25 %) had little effect, and resulted in less than 5 % BMD change misclassification. Misclassification exceeded 10 % with absolute BMD drift greater than 0.006 g/cm(2) or relative drift over 0.75 %, and was greater than 35 % for absolute BMD drift of 0.024 g/cm(2). Similar trends were seen when results were stratified according to use of anti-osteoporosis medication, when varying the least significant change (LSC), and for evaluating lumbar spine BMD change. In summary, relatively small degrees of DXA calibration drift can have large effects on misclassifying BMD change. Our results support a calibration drift tolerance of 0.006 g/cm(2) or 0.5 %. These findings may help to guide timing for DXA scanner servicing and repair.