In this study we assessed the ability of proximal femur and spine DER measurements to discriminate between control (C) and osteoporotic (OP) subjects as well as the reproducibility of the femoral DER technique. DER measurements of the proximal femur and spine (L2-L4) were obtained in 58 control (unfractured) and 54 osteoporotic (vertebral fractured) women. Linear regression analyses for each site with age suggested that density values from the Ward's triangle region explained the majority of variance due to age in the decline of bone density in both control (r = -0.45; P less than 0.0005) and osteoporotic subjects (r = 0.34; P less than 0.05), with similar findings for the femoral neck in OP subjects (r = -0.30; P less than 0.05). No other femoral site produced a significant correlation with age, and vertebral measurements were the lowest in OP subjects (r = -0.11; P = NS). Cross-sectional rates of bone loss in each region implied that for OP subjects the Ward's triangle and femoral neck regions detected the largest amount of bone loss (0.64% and 0.38%/yr, respectively), while vertebral measurements were again the smallest (0.19%/yr). Results indicated that proximal femur measurements can be obtained with relatively good precision (coefficients of variation ranged from 1.9% for femoral neck to 3.0% for Ward's triangle). Finally, receiver operating characteristic curves revealed that both Ward's triangle and femoral neck densities were able to more successfully discriminate control from osteoporotic subjects than vertebral measurements. These data support the contention that measurement of skeletal sites rich in trabecular bone, such as the femur, are more useful for diagnosing osteoporosis than DER measurements of the lumbar spine. The data also suggest that the femur may be a better site than the lumbar spine for the clinical evaluation of osteoporosis with the DER technique.