Comment on Tothill and Hannan: precision and accuracy of measuring changes in bone mineral density by dual-energy X-ray absorptiometry

Abstract

Dear Editors, We read the article by Tothill and Hannan [1] regarding the precision and accuracy of DXA with interest, and acknowledge that Dr. Tothill is one of few researchers who have actually questioned the inerrancy of DXA BMD measurements. However once again we note that the fundamental shortcomings of DXA as an analytical technique are not openly addressed. This latter criticism is quite separate from the ongoing argument of whether bone mass is an appropriate marker of bone quality [2–4]. It is misleading to refer to DXA BMD accuracy, purportedly the difference between the ‘true’ and measured BMD value, when the BMD entity/quantity is both undefined and unknown. Accuracy is conceptually meaningless if no estimate of the ‘true’ value exists. Since bone comprises both organic (mainly type I collagen) and mineral (hydroxyapatite) components, where the latter is a heterogeneous compound, comprising structural analogues of hydroxyapatite (Ca10(PO4)6(OH)2) [5], a ‘bone’ standard for DXA is unlikely. Unfortunately, therefore, no respective surrogate reference material has ever been proposed to allow interinstrument standardisation, nor has default to the most basic logical option, nomination of a ‘consensus method’, been advocated. Although this patent lack of standardisation for DXA-based BMD measurements was known and unequivocally stated a decade ago [6], little if any change has occurred. The plain facts are that DXA is a radiological method lacking in selectivity [7–11], without any agreed reference standard, which attempts to measure an undefined material (‘bone’). Hardly attributes normally associated with reliable analytical methods. As the reliability of any analytical method is determined by both its accuracy and precision, the precision of DXA is intricately affected by its lack of accuracy. Previously we, like Tothill and Hannan [1], conducted an investigation examining long-term BMD changes in women (n=80) who had serial DXA spine and hip scans (n=6–8) during a 4to 8-year period. Subsequent estimates of the measurement error (Se) were more than twofold greater than in our 12-month study [12]. After reviewing individual patient BMD profiles and deleting the value that resulted in minimisation of variance, we found that Se decreased by about 40%, but was still significantly greater than before. This finding re-emphasises our original contention [12] that DXA reliability is limited by ubiquitous ‘outlier’ values directly related to a lack of selectivity. Under these circumstances the least significant change model is inappropriate, since such models require a specific analytical method. We are also wary of precision studies, like that of Tothill and Hannan [1], where spurious data, previously reported, are now retrospectively excluded. Such practices must undoubtedly bias Se. We are dismayed by the persistent claims that DXA is a precise method for BMD estimation, when clearly a myriad of both uncontrollable and undetectable factors may significantly affect any actual in vivo result. The latter must be apparent to all who routinely observe the omnipresent “zig-zag” BMD patterns intrinsic to so many serial patient scan profiles. Yet the litany of all these known deficiencies remains essentially unaddressed, presumably Osteoporos Int (2008) 19:1099–1100 DOI 10.1007/s00198-007-0551-5