Abstract

Osteoporosis is a common disease with a strong genetic component characterized by reduced bone mass and an increased risk of fragility fractures. Bone mineral density (BMD) is the most important determinant of osteoporotic fracture risk, but the genes responsible for BMD regulation and fracture are incompletely defined. To enable multi-center studies to examine the genetic influences on BMD there is a requirement to standardize measurements across different manufacturers of bone densitometers, different versions of machines and different normative ranges. This paper describes a method developed to allow near-identical subjects with low age-adjusted BMD (based on Z-scores) to be recruited in 17 centers using 27 different densitometers. Cross-calibration was based on measurements using a European spine phantom circulated to all centers and measured ten times on each individual machine. From theses values an individual exponential curve, based on nominal versus observed BMD, was derived for each machine. As expected, there were large and significant variations in nominal BMD values, not only between scanners from different manufacturers but also between different versions of scanners from the same manufacturer. Hologic scanners tended to underestimate the nominal BMD, while Lunar scanners overestimated the value. Norland scanners gave mixed values over estimating BMD at the lower nominal value (0.5 g/cm2) while underestimating the value at the higher value (1.5 g/cm2). The validity of the exponential equations was tested using hip and spine measurements on 991 non-proband women from a familial osteoporosis study (FAMOS). After cross-calibration there was a considerable reduction in variation between machines. This observation, coupled with the absence of a similar reduction in variation attributable to a linear regression on age, demonstrated the validity of the cross-calibration approach. Use of the cross-calibration curves along with a standard normative range (in the case of this study, the Hologic normative range) allowed age-specific Z-scores to be used as an inclusion criterion in this genetic study, a method that will be useful for other trials where age-specific BMD inclusion criteria are required.