A group of 116 postmenopausal women, 58 femoral neck fracture cases and 58 agematched nonfracture controls, was studied. In each subject, the second metacarpal of the right hand and the femur of the left or nonfracture side were measured by radiographic morphometry, and the trabecular patterns in the femoral neck graded by Singh’s method. Results were examined by discrimination function analysis to identify the most useful combinations of variables for the classification of cases. The most useful single measurement was found to be the Singh grade (misclassification, 2 1 %) and the least useful measurement site, the second metacarpal, for which the best variable (the cortical area/total area ratio) produced a misclassification of 40%. Better discrimination was obtained by using two-variable combinations, that involving Singh grade and width of the calcar femorale being the best (misclassification, 17%). There was only one three-variable combination in which all variables contributed significantly to the discrimination;​ this combination involved Singh grade, calcar width and femoral midshaft cortical width and resulted in a misclassification of less than 16% of all subjects.

Because the Singh grade provides the best discrimination between fracture and nonfracture groups, fracture risk must be predominantly determined by the trabecular structure and integrity in the femoral neck. It may be that the apparent relationship between the age-related changes in the metacarpal and the femur fracture incidence is actually owing to the correspondence (on a population basis) between the loss of cortical bone at that site and the loss of trabecular bone from the femur, and does not in itself necessarily imply that cortical bone loss is of major importance in the femoral neck. However, the results indicate that cortical bone loss in the femur does indeed significantly increase fracture risk, albeit less than the destruction of the trabecular architecture.