Serial measurement of bone density is important in the clinical management of osteoporosis because it helps to confirm therapeutic response and identify individuals who continue to lose bone density. The lumbar spine is often stated to be the preferred site for monitoring purposes, but much of the supporting information derives from the relatively poor precision of the femoral neck, which is known to be less reproducible than the total hip. The Manitoba Bone Density Program delivers clinical services to the Province of Manitoba, Canada and maintains a detailed quality assurance program. We compared lumbar spine and total hip reproducibility using a pencil-beam scanner (n = 82), a fan-beam scanner (n = 92), and a cross-calibration between the two scanners (n = 61). Each comparison showed significantly lower measurement error at the total hip than at the lumbar spine (p < 0.001). We assessed the likelihood that follow-up measurements would show an absolute difference exceeding the least significant change in routine clinical practice (n = 2412 paired examinations). The total hip measurement detected change more frequently than the lumbar spine in each of these comparisons. In summary, we believe that the total hip site has been undervalued for monitoring purposes and might actually be the preferred site in older individuals with a high prevalence of lumbar spine artifact.