Advanced Analysis Protocol Improves Quality of Pediatric Hip Structural Analysis

To determine the precision, accuracy, and unique analysis challenges of HSA in children. Hip structural analysis (HSA) variables, a collection of 10 measures including cross-sectional area (CSA), cross-sectional inertia (CSI), and buckling ratio (BR), have been shown to be independent risk factors in determining fracture risk in adults, but there have been few studies reporting the utility and accuracy of HSA in children. Previous work has described the precision of HSA in adults, but the precision and unique challenges of the HSA protocol in children is unexplored. Here we describe the unique challenges, precision, and quality assurance protocol of pediatric HSA measures in a large cohort of over 2,500 children. This is a retrospective analysis of prospectively collected DXA scans acquired as part of two studies, the Bone Mineral Density in Childhood Study (BMDCS) and the Genome-wide Analysis Study (GWAS). The combined sample consisted of 2,514 children (10,787 scans, 1,271 girls) aged from 5 to 21 years. The proximal femur DXA scans were acquired on five Hologic systems (Hologic, Inc., Marlborough, MA) of similar models (A and W) with up to eight years of annual follow-up between 2002 and 2009. All scans were analyzed centrally by the authors using one technologist using APEX 3.4 software. A unique and comprehensive quality assurance check was completed for all scans including a review of the acquisition criteria set by ISCD and a review of the automatically placed HSA region's narrow neck (NN), intertrochanteric (IT), and femoral shaft (FS) region of interests. During processing, regions were either repositioned or eliminated on DXA imaging. Duplicate scans were performed on 150 children (71 girls) for precision assessment. Specific HSA quality control (QC) codes were generated for this particular analysis in accordance with the author's criteria. Short-term precision estimates were calculated as the RMSE and %CV. QA codes were assigned to the NN, IT, and FS boxes that were either incorrectly positioned or invalidated. Of the entire dataset under 10% of NN and FS boxes needed to be repositioned and none were invalidated. Figure 1 provides an example of proper placement of the IT box (at a 45-degree angle) in between the greater and lesser trochanter. If the angle of the IT box is either < 10 or >25 degrees, the IT box was invalidated. In this study, 100% of the IT boxes needed to be repositioned and 54% remained invalid. Multiple reasons were identified for an invalid scan region including the unavoidable presence of a growth plate in the hip scans for participants less than 15 years old, as shown in Figure 1. All HSA precision over all age groups ranged was less than 6% CV except for the NN Buckling ratio and Cross-sectional Inertia. In general, the precision error was lower in the older ages versus the younger participants. See Table 1. We conclude that HSA creates precise estimates in children that are comparable to that in adults for the femur neck and shaft but not the intertrochanteric region. Thorough quality assurance procedures must be in place to safeguard against poor region placement due to the size of the bone.