Second-generation high-resolution peripheral quantitative computed tomography (HR-pQCT) provides the highest resolution in vivo to assess bone density and microarchitecture in 3D. Although strong agreement of most outcomes measured with first- (XCTI) and second- (XCTII) generation HR-pQCT has been demonstrated, the ability to use the two systems interchangeably is unknown. From in vivo measurements, we determined the limits of estimating XCTII data from XCTI scans conducted in vivo and whether that estimation can be improved by linear cross-calibration equations. These data are crucial as the research field transitions to the new technology. Our study design established cross-calibration equations by scanning 62 individuals on both systems on the same day and then tested those cross-calibrations on the same cohort 6 months later so that estimated (denoted as XCTII*) and "true" XCTII parameters could be compared. We calculated the generalized least-significant change (GLSC) for those predictions. There was strong agreement between both systems for density (R2  > 0.94), macroarchitecture (R2  > 0.95), and most microarchitecture outcomes with the exception of trabecular thickness (Tb.Th, R2  = 0.51 to 0.67). Linear regression equations largely eliminated the systematic error between XCTII and XCTII* and produced a good estimation of most outcomes, with individual error estimates between 0.2% and 3.4%, with the exception of Tt.BMD. Between-system GLSC was similar to within-XCTI LSC (eg, 8.3 to 41.9 mg HA/cm3 for density outcomes). We found that differences between outcomes assessed with XCTI and XCTII can be largely eliminated by cross-calibration. Tb.Th is poorly estimated because it is measured more accurately by XCTII than XCTI. It may be possible to use cross-calibration for most outcomes when both scanner generations are used for multicenter and longitudinal studies. © 2017 American Society for Bone and Mineral Research.