Variations in diagnostic performances of dual-energy X-ray absorptiometry in the northwest of The Netherlands.

Between-center variation in bone densitometry may influence the frequency of the diagnosis of osteoporosis. To evaluate this problem, dual-energy X-ray absorptiometry (DXA) machines of the medical centers in the northwest of The Netherlands were evaluated. Four phantoms were used to test the 17 DXA machines of 16 participating centers. Each phantom was measured 10 times and the data were analyzed on the corresponding DXA machine using the software delivered by the manufacturer. The analyses were done with the reference population as used in daily practice. There were DXA devices of seven different brands and types, using four different reference populations for the lumbar spine and seven for the hip. The observed differences in bone mineral densities (BMD) were up to 0.20 g/cm(2) for the lumbar spine, 0.15 g/cm(2) for the femoral neck, and 0.12 g/cm(2) for the total hip. The coefficients of variation (CV) of the repeated phantom measurements ranged between 0.3% and 1.3% for the lumbar spine, 1.6% and 4.6% for the femoral neck, and 0.3% and 0.9% for the total hip. The mean female T-scores of 10 phantom measurements differed up to 0.6 SD between the DXA machines for the lumbar spine and up to 0.8 SD for the total hip. Mathematically, replacing a Hologic 2000 DXA machine with a newer type of the same brand (a Hologic 4500) caused a shift in diagnosis from osteoporosis to osteopenia of +1.1% for the lumbar spine and -4.5% for the total hip. When the Hologic 2000 was replaced by a Hologic 4500 with NHANES reference values, the shift from osteoporosis to osteopenia was also +1.1% for the lumbar spine, and -13.4% for the total hip. The clinical impact of the observed differences is difficult to estimate. One may conclude that the differences of the tested DXA devices are partly based on differences in DXA machines, but for the most part on the use of different reference populations. It is recommended to standardize the reference population, although the consequent shift in diagnosis will be confusing for physicians and patients, and adaptation of the reference values on the DXA devices of different brands with different technical qualities and measurement specifications will be difficult.