Original Article

Osteoporosis International

, Volume 5, Issue 3, pp 167-173

Better discrimination of hip fracture using bone density, geometry and architecture

  • M. PeacockAffiliated withIndiana University Department of Medicine
  • , C. H. TurnerAffiliated withDepartment of Orthopedics and Mechanical Engineering, Indiana University-Purdue University
  • , G. LiuAffiliated withIndiana University Department of Medicine
  • , A. K. ManatungaAffiliated withIndiana University Department of Medicine
  • , L. TimmermanAffiliated withIndiana University Department of Medicine
  • , C. C. JohnstonJrAffiliated withIndiana University Department of Medicine

Rent the article at a discount

Rent now

* Final gross prices may vary according to local VAT.

Get Access


Bone density predicts the risk of hip fracture. Because hip strength is determined by bone geometry and architecture as well as density, we tested which variables in geometry and architecture were independent discriminators of hip fracture and, if combined with density, improved the discrimination of fracture from non-fracture over bone density alone. The design was a case-control study. The subjects were Caucasian women over the age of 60 years who had sustained a hip fracture after the age of 58 years (n=22), and controls matched for age and weight (n=43) and unmatched controls (n=317) with no history of hip fracture. Variables in density, geometry and architecture were obtained from dual-energy X-ray absorptiometry images and from radiographs of the upper end of the femur. In a univariate model, of the measures of bone mass, the best discriminator of hip fracture was bone mineral density of the neck of femur; of the geometric measurements, it was hip axis length; and of the measurements of bone architecture, it was Singh grade. In a multivariate model, these three variables were shown to be independent discriminators of hip facture. When hip axis length was combined with bone mineral density, there was significant improvement in discrimination of hip fracture (p=0.014), and when Singh grade was combined with hip axis length and bone mineral density there was a further significant improvement (p=0.002). In logistic regression models using hip axis length and Singh grade adjusted for femoral neck bone mineral density, age and weight, the area under the receiver-operating characteristics (ROC) curve for femoral neck density, hip axis length and Singh grade together was significantly greater than for femoral neck density alone (p=0.006). Models that combine bone mass (density), geometry (hip axis length) and architecture (Singh grade) significantly improve the discrimination of hip fracture over bone density by itself. If these models can be shown to be equally useful in predicting hip fracture prospectively and can be obtained from dual-energy X-ray absorptiometry, their use will increase the ability to identify subjects at most risk of hip fracture.


Bone mineral density Hip axis length Hip fracture Osteoporosis Singh grade