Decreased bone mineral density (BMD) in the elderly increases the risk of hip fracture. Measurement of proximal femoral BMD can help us predict the risk of hip fracture, especially in the elderly. Since the BMD of proximal femur is usually measured on the unilateral side, we studied the risk of underestimation with measurement of unilateral proximal femur BMD in 266 normal Chinese women. In order to evaluate the effect of age, these subjects were divided into group A (18-59 years, n = 189) and group B (60-88 years, n = 77). BMDs of both proximal femurs were assessed with Norland 2600 dual photon absorptiometry. Using a cutoff T score of -1, the negative predictive value (NPV) in the entire group was 86.9% for femoral neck BMD and 85.7% for trochanter, and 82.2% for Ward's triangle: in group A, the NPV was 88.9% for femoral neck, 88.8% for trochanter, and 97.2% for Ward's triangle, but in group B, the NPV was 60.0% for femoral neck, 71.0% for trochanter, and 24.1% for Ward's triangle. The accuracy in the entire group was 86.1% for femoral neck, 84.2% for trochanter, and 86.3% for Ward's triangle: in group A the accuracy was 84.6% for femoral neck, 84.9% for trochanter, and 92.8% for Ward's triangle, but in group B, the accuracy was 89.6% for femoral neck, 81.7% for trochanter, and 90.0% for Ward's triangle. In general, NPV and accuracy increased at the expense of positive predictive value when the cutoff T score was changed to -2.5. This study suggested that measurement of unilateral proximal femur BMD was sufficient for screening the contralateral hip BMD in group A at a cutoff T score of -1. However, a T score of -2.5 was recommended for group B, and one should be careful in its application to Ward's triangle.