Search

Abstract: The role of quantitative ultrasound (QUS) in clinical practice is debatable. An unanswered question is that whether combining QUS and BMD measurements could improve the prediction of fracture risk. We examined this in a sample of men and women in the European Prospective Investigation into Cancer (EPIC)-Norfolk who had both heel QUS and hip DXA between 1995 and 1997 and were followed for any incident fracture up to 2007. From 1455 participants (703 men) aged 65–76 years at baseline, 79 developed a fracture over 10.3±1.4 years of follow-up. Two separate sex-stratified Cox proportional-hazard models were used including clinical risk factors and total hip BMD. Heel broadband ultrasound attenuation (BUA) was also included in the second model. Global measures of model fit, area under ROC curve, and the Hosmer–Lemeshow statistic showed relative superiority of the model including BUA. Using each model, we calculated 10-year absolute risk of fracture for all participants and categorized them in groups of <5%, 5% to <15%, and ≥15%. Comparison of groupings showed a total re-classification of 16.6% of participants after inclusion of BUA with the greatest re-classification (30.7%) among the group with intermediate risk. Adding a QUS measurement to models based on clinical risk factors and BMD improves the predictive power of models and suggests that further attention should be paid to QUS as a clinical tool for fracture risk assessment. [Copyright &y& Elsevier]

Abstract: A sample size of 30 degrees of freedom (df) for bone mineral density (BMD) precision studies may be insufficient for reliably categorizing change. Monte Carlo simulation was used to evaluate the effect of precision study sample size on identifying change in clinical patients. Least significant change (LSC) from 198 spine and 193 total hip scan-pairs was used to categorize change for 1420 patients undergoing BMD monitoring. Relative to this reference change fraction (RCF), LSC limits were identified that gave specified deviations from the RCF (−25% to +25%). Confidence limits (95% and 80%) for these LSC values (5 to 500 df) were estimated using ‘bootstrap’ samplings. A sample size providing 140 df is needed to avoid overdetecting spine change by 5% and 150 df to avoid underdetecting spine change by 5% with 95% confidence limits. A sample size of 30 df resulted in up to a 12.5% overdetection and 10.0% underdetection of spine or hip change based upon 95% confidence limits. In conclusion, assessing the effect of precision study sample size on classifying change in monitored patients is an important element of the precision assessment that is neglected in current recommendations. Sample sizes larger than 30 df are required if low levels of categorization error are to be achieved. [Copyright &y& Elsevier]