• An investigation of the test-retest reliability of an ultrasound densitometer

      Lamb, Kevin L.; Owen, David G. (University of Liverpool (Chester College of Higher Education), 18/10/1998)
      Army recruits undertake a rapidly increasing amount of exercise in their initial basic training period. Injuries due physical training forces many recruits out of the Army and costs the Ministry of Defence millions of pounds. Stress fractures are one of the most commonly diagnosed injuries amongst Army recruits. Low bone mineral density has been identified as a risk factor for stress fractures. A technique which can measure bone mineral density is Ultrasound Densitometry (US). This study will address a gap in the research by assessing the inter-observer and intra-observer reliability of the two US measurements, broadband ultrasound attenuation (BUA) and the velocity of sound (VOS). Ninety eight white male recruits, median aged 18 (I.Q. range 1yr) were measured at the calcanea of the non-dominant foot. A repeated measures design was used, BUA and VOS were measured in 55 subjects by both researcher 'A1 and 'B' for inter-observer (inter-BUA & inter-VOS), and 43 subjects were measured for BUA and VOS twice by researcher 'A' for the intra-observer analysis (intra-BUA & intra-VOS). The results from this study found that a coefficient of variation (CV) analysis was not appropriate for assessing measurement error, this was due to the homoscedasity of the data. An alternative method the '95% limits of agreement' found that only VOS was reliable. The '95% limits of agreement1 results (bias ±1.96 x s) were 0.74 ±22.77 m/s for intra-VOS and 4.85 ±23.44 m/s for inter-VOS, the variance in scores were judged to be acceptable, f-test confirmed this with a non-significant difference between measurements (t=0.83, p=0.477; t=0.42, p=0.677, respectively). The '95% limits of agreement1 results for BUA were -0.22±11.56 dB/MHz (inter-BUA) and -1.39 ±11.11 dB/MHz (intra-BUA). These results represent an unacceptable variability in the range of scores obtained. This is highlighted when expressed as a proportion of the mean measurement: inter-BUA ±11.41% and intra-BUA ± 11.91%. However, the West's for inter- and intra-BUA indicate no significant difference (t = -0.07, p = 0.091; t = 1.60, p = 0.116). This insignificance may be the result of the inappropriateness of a statistical method that reliance on a comparison of means. The CV results for BUA indicate that both inter- and intra-BUA are reliable (4.08% & 4.38%, respectively), even though as already stated that the BUA measurements are not deemed reliable when analysed by the '95% limits of agreement'. The results of this study suggest that VOS measurements are reliable and that BUA measurements are non-reliable. As both BUA and VOS would have been used to assess those at risk of suffering stress fractures it was essential that both were found to be reliable. Thus US's appropriateness in individual diagnosis is questioned. This study has also highlighted how the use of an inappropriate statistical method, in this case the CV, can effect the interpretation of data and cause false claims over e.g. reliability.