A pilot study of 107 women aged 60-69 years recently suggested that the measurement of broadband ultrasound attenuation (BUA) provides a superior cost effective pre-screen referral method for bone mineral density (BMD) measurement by DXA (dual-energy X-ray absorptiometry) than can be achieved by clinical criteria (CC). The aim of this study was to compare the accuracy and cost effectiveness of BUA and clinical criteria in a younger cohort. 599 women aged 50-54 years (52.18 +/- 1.35) had previously been measured by DXA at lumbar spine and right femoral neck, along with BUA measurement of the right calcaneus. Each subject had also completed an extensive clinical and social questionnaire to ascertain those who would have met one or more of the six general clinical criteria adopted by our Centre. Each subject was classified by DXA using the WHO criteria as normal, osteopenic or osteoporotic, defined at lumbar spine or femoral neck. Sensitivity, specificity and accuracy were calculated for BUA and the clinical criteria, noting that analysis was undertaken with and without the oestrogen deficiency clinical criterion (CC1): "Any oestrogen deficient woman who would want to be treated or would want to continue treatment if found to be osteopenic or osteoporotic". The accuracy for identifying osteoporotic subjects was 72.8% for BUA (at the point of matched sensitivity and specificity, 75 dB MHz(-1)), 30.7% for CC(1-6) and 64.3% for CC(2-6). When osteopenic subjects were incorporated, the accuracies were 63.8% for BUA (at the point of matched sensitivity and specificity, 82 dB MHz(-1)), 60.3% for CC(1-6) and 55.7% for CC(2-6). The minimum cost per osteoporotic subject correctly identified was pound sterling 573.50 by DXA alone, pound sterling 325 by BUA, pound sterling 458 by CC(1-6) and pound sterling 416 by CC(2-6). When osteopenic subjects were incorporated, the costs were pound sterling 87, pound sterling 83.50, pound sterling 78 and pound sterling 74, respectively. The overall cost, dependent upon the prevalence of osteoporosis (or osteopenia) within the population, more accurately indicates the feasibility of a population-based screening programme. For the identification of either osteoporotic or osteopenic subjects from the general population by DXA, the prevalence-compensated cost (cost per subject correctly identified multiplied by prevalence) is pound sterling 45, irrespective of age cohort. If CC(2-6) were adopted for the identification of osteoporotic subjects alone, the prevalence-compensated cost would be pound sterling 32 and pound sterling 42 for the 50-54 and 60-69 aged cohorts, respectively. For BUA, the prevalence-compensated cost falls to pound sterling 25 and pound sterling 43 for the 50-54 and 60-69 aged cohorts, respectively. If osteoporotic or osteopenic subjects were to be identified in the 50-54 aged cohort, both CC(2-6) (pound sterling 38) and BUA (pound sterling 43) perform similarly to DXA alone. BUA appears to provide a valuable population pre-screen for the identification of osteoporotic subjects, less so for osteopenic. It is suggested that if both osteopenic and osteoporotic women are to be identified for clinical management incorporating DXA, then neither BUA nor clinical criteria are satisfactory referral methods. An unanswered question from this study, however, is whether ultrasound has an independent role in the assessment of fracture risk for perimenopausal women who do not have the benefit of referral for DXA.