The impact of measurement errors on the diagnostic value of bone mass measurements: theoretical considerations

It has become clear over the last decade that correlations between measurements of forearm bone mineral content (BMCarm) by single-photon absorptiometry (SPA) and measurements of spinal bone mineral density (BMDspine) by dual-photon absorptiometry (DPA) in healthy subjects and patients with spinal fractures are invariably significant, but not very powerful (i.e. r = 0.5-0.7). Nonetheless, several recent studies have shown that appendicular bone mass measurements discriminate between spinal fracture and non-fracture at least as well as do spinal DPA measurements. Correlations of a given parameter with measured BMDspine are less important than those with true BMDspine. To establish the latter we made the following assumptions: (1) accuracy errors or SPA BMCarm and DPA BMDspine measurements of 2%-4% and 8%-10%, respectively; and (2) a measured biological variation of SD = 14% for both BMCarm and BMDspine, corresponding to that of healthy women at the menopause. On these assumptions, we found that a correlation between true BMCarm and true BMDspine at about r = 0.8-0.9 yields a correlation between measured BMCarm and measured BMDspine at about r = 0.6--corresponding to experimental data in healthy women at the menopause. Furthermore, we found that the correlation between DPA measured BMDspine and true BMDspine is about the same as that between the SPA measured BMCarm and the true BMDspine. Thus, with the assumptions given above, spinal (DPA) and fore-arm (SPA) measurements appears to predict equally the true BMDspine in healthy perimenopausal women.