Effect of Total Hip Bone Area on Osteoporosis Diagnosis and Fractures

Abstract

DXA is affected by skeletal size, with smaller bones giving lower areal BMD despite equal material density. Whether this size effect confounds the use of BMD as a diagnostic and fracture risk assessment tool is unclear. We identified 16,205 women of white ethnicity >or=50 yr of age undergoing baseline hip assessment with DXA (1998-2002) from a population-based database that contains all clinical DXA test results for the Province of Manitoba, Canada. Total hip measurements were categorized according to quartile in total hip bone area (Q1 = smallest, Q4 = largest). Longitudinal health service records were assessed for the presence of nontraumatic osteoporotic fracture codes during a mean of 3.2 yr of follow-up after BMD testing (757 osteoporotic fractures, 186 hip fractures). Total hip bone area strongly affected osteoporosis diagnosis with much higher rates in Q1 (14.4%) than Q4 (8.9%). However, incident fracture rates were constant across all area quartiles, and prevalent fractures were paradoxically fewer in smaller area quartiles (p < 0.001 for trend). Age was a potential confounder that correlated positively with area (r = 0.12, p < 0.0001). When age was not included in a Cox regression model, Q1 seemed to have a lower rate of incident osteoporotic fractures (HR = 0.80, 95% CI = 0.66-0.98, reference Q4) and hip fractures (HR = 0.63, 95% CI = 0.43-0.94) for a given level of BMD. In age-adjusted regression models, total hip BMD was strongly predictive of incident osteoporotic fractures (HR per SD = 1.83, 95% CI = 1.68-1.99) and hip fractures (HR per SD = 2.80, 95% CI = 2.33-3.35), but there was no independent effect of bone area (categorical or continuous). Nested matched subgroup analysis and ROC analysis confirmed that bone area had no appreciable effect on incident fractures. We conclude that total hip areal BMD categorizes a substantially higher fraction of women with smaller bone area as being osteoporotic despite younger age. Incident fracture rates correlate equally well with BMD across all bone area quartiles when adjusted for age.