Large variations occur in bone density measurements of children when using different software

Background Paediatric dual X-ray absorptiometry (DXA) studies present a number of technical problems. One of these is that the edge detection algorithms designed for the adult skeleton may fail for paediatric studies. Hologic provide alternative algorithms for low bone density studies. Aim To assess low-density software for the analysis of paediatric DXA studies and to compare with the adult protocol. Methods Our centre has scanned 450 normal children as part of a normal range study. A subgroup of 103 children was selected using a random number generator. The group was distributed evenly between males and females and across the age range 5-17 years. Each individual underwent both a lumbar spine and a whole-body scan on a Hologic QDR-4500W DXA scanner. Both scans were analysed using the standard adult protocol and then re-analysed using the Hologic experimental paediatric protocol for whole body and the Hologic low-density protocol for lumbar spine. Results Both lumbar spine protocols showed an increase in bone mineral density with age; however, the low-density protocol always produced a lower bone mineral density result than the adult protocol. Bland-Altman analysis showed limits of agreement of 0.031 -0.093 g cm - 2 (male, 0.032-0.089 g. cm - 2 ; female, 0.031-0.096 g cm - 2 ). This represents a mean difference of 9%. Five results showed differences greater than the upper limit of agreement. All these cases were children under 11 years of age who had large areas of spine not identified as bone by the adult protocol. These children were all below the 30th percentile for the body mass index. The whole-body protocols showed similar increases in bone mineral density with age; however, the experimental paediatric protocol always produced a lower bone mineral density result than the adult protocol. Paired results showed limits of agreement of 0.0668-0.130 g cm - 2 (male, 0.063-0.124 g cm - 2 ; female, 0.073-0.134 g cm - 2 ). This represents a mean difference of 11%. Five results showed differences greater than the upper limit of agreement. Conclusions For anteroposterior (AP) lumbar spine scans, the use of the paediatric algorithm in children under 11 years of age would prevent the largest failures in analysis. For whole-body scanning, the adult algorithm showed no major failures in children of 11 years or older. It is hoped that forthcoming improvements in whole-body density analysis will improve the results for those under 11 years of age. Normal range data should be generated for any new algorithm to allow proper interpretation of clinical studies.