Comparison of six calcaneal quantitative ultrasound devices: Precision andhip fracture discrimination

Quantitative ultrasound (QUS) is now accepted as a useful tool in the manag ement of osteoporosis. There are a variety of QUS devices clinically availa ble with a number of differences among them, including their coupling metho ds, parameter calculation algorithms and sites of measurement. This study e valuated the abilities of six calcaneal QUS devices to discriminate between normal and hip-fractured subjects compared with the established method of dual-energy X-ray absorptiometry (DXA). The short-term and mid-term precisi ons of these devices were also determined. Thirty-five women (mean age 74.5 +/- 7.9 years) who had sustained a hip fracture within the past 3 years, a nd 35 age-matched controls (75.8 +/- 5.6 years) were recruited. Ultrasound measurements were acquired using six ultrasound devices: three gel-coupled and three water-coupled devices. Bone mineral density was measured at the h ip using DXA. Discrimination of fracture patients versus controls was asses sed using logistic regression analysis (expressed as age- and BMI-adjusted odds ratios per standard deviation decrease with 95% confidence interval) a nd receiver operating characteristics (ROC) curve analysis. Measurement pre cision was standardized to the biological range (sCV). The sCV ranged from 3.14% to 5.5% for speed of sound (SOS) and from 2.45% to 6.01% for broadban d ultrasound attenuation (BUA). The standardized medium-term precision rang ed from 4.33% to 8.43% for SOS and from 2.77% to 6.91% for BUA. The pairwis e Pearson correlation coefficients between different devices was highly sig nificant (SOS, r = 0.79-0.93; BUA, r = 0.71-0.92). QUS variables correlated weakly, though significantly, with femoral BMD (SOS, r = 0.30-0.55; BUA, r = 0.35-0.61). The absolute BUA and SOS values varied among devices. The ge l-coupled devices generally had a higher SOS than water-coupled devices. Bo ne mineral density (BMD) and BUA were weakly correlated with weight (r = 0. 48-0.57 for BMD and r = 0.18-0.54 for BUA), whereas SOS was independent of weight. All the QUS devices gave similar, statistically significant hip fra cture discrimination for both SOS and BUA measures. The odds ratios for SOS (2.1-2.8) and BUA (2.4-3.4) were comparable to those for femoral BMD (2.6- 3.5), as were the area under the curve (SOS, 0.65-0.71; BUA, 0.62-0.71; BMD , 0.65-0.74) from ROC analysis. Within the limitation of the sample size ai l devices show similar diagnostic sensitivity.