The performance of the Hologic QDR-2000 DXA osteodensitometer was critically evaluated at four centers, using at all four centers one bone equivalent humanoid spine phantom supplied by the manufacturer. Results were compared with results from Hologic QDR-1000/W using that phantom tested at the same centers. It appears that the concept of fan-beam scanning--as used in the QDR-2000: a fan-beam, a linear array detector above the phantom, and an x-ray tube located rather close to the spine below the phantom--creates problems due to the magnification effect of the fan beam. The effect of decreasing the distance between the "vertebrae" of the phantom and the couch are: bone mineral content (BMC) increases by 2.8% per cm, projected area (Area) by 2.8% per cm, and bone mineral density (BMD) is unchanged. When QDR-1000/W is upgraded to QDR-2000, BMD is relatively constant, but there are shifts of BMC and Area which are partly due to the magnification effect of the fan-beam. Replacement of a QDR-1000/W with a QDR-2000 can invalidate longitudinal measurements, even for BMD, unless the proportionality factors of the QDR-2000 are checked and, if necessary, changed. This is true for switching from QDR-1000/W to pencil-beam mode of QDR-2000 or to fan-beam mode of QDR-2000. Even with pencil-beam mode, the long-term precision error with phantoms is higher for QDR-2000 than for QDR-1000/W (for BMD, 0.47% versus 0.35%).