Stata version 11.2 for windows

To harmonise measurements between different anatomical sites and devices, we computed sex-specific BMC and BMD Z-scores and used them as continuous variables for agreement analysis with the Bland-Altman method. Mean differences and limits of agreement were estimated, because assumptions were met (10) . Also, Z-scores were categorised using cut-offs at each unit of standard deviation (SD) to estimate concordance between the categories. Observed agreement (%) was calculated as a measure of overall concordance, and Cohen's linear weighted kappa (j) with 95% confidence intervals (95% CIs) was estimated to account for the agreement expected by chance.
Receiver operator characteristic (ROC) curves were plotted, and the areas under the curves (AUC) with their 95% CIs were calculated to assess the accuracy of measures to predict fracture. Likelihood ratios with 95% CIs were computed for Z-score categories. For this particular analysis, two children with missing information for lifetime fracture history were excluded. We also excluded 16 fractures that occurred up to two years of age, which are usually thought to result from severe trauma, regardless of bone quality, such as delivery injuries, child abuse or accidental falls due to the caregiver's negligence (11, 12) .
The statistical analysis was performed using Stata version 11.2 for Windows (Stata Corp LP, College Station, TX, USA).