Osteoporosis is frequently undiagnosed before fracture because of the lack of availability of instruments to quantitate bone mass. To evaluate the utility of quantitative ultrasound (QUS) of the calcaneus to diagnose osteoporosis, we determined bone mineral density (BMD) of the posterior-arterior spine, total hip, and femoral neck by dual-energy X-ray absorptiometry (DXA) and QUS in 312 women aged 50 years and older. A risk factor assessment (simple calculated osteoporosis risk estimation [SCORE]) also was quantitated in all women. Ninety-four of the 312 women were diagnosed as osteoporotic based on T scores ≤ −2.5 at the spine, total hip, and/or femoral neck. The sensitivity of the individual central sites for the diagnosis of osteoporosis was 49% at the spine (46 of 94 women), 32% at the total hip (30 of 94 women), and 81% at the femoral neck (76 of 94 women). At a QUS T score ≤ −1, the peripheral technique had a sensitivity of 62% and a specificity of 72%. Combining a QUS T score of ≤ −1 followed by a risk factor assessment of women with a QUS T score ≥ −0.99 using a cut point of 11 increased sensitivity to 81% (comparable with femoral neck DXA) but decreased specificity to 58%. If peripheral QUS measurements and risk factor assessment are the only tools employed before initiation of therapy, the benefits of increased ease of diagnosis will need to be balanced against potentially unnecessary treatment in some normal patients and lack of treatment in some osteoporotic patients.
POSTMENOPAUSAL OSTEOPOROSIS is common. The Third National Health and Nutrition Exam Survey (NHANES III) estimated that 8 million women in the United States have osteoporosis and 15 million have low bone mass.(1) Osteoporotic fractures, especially hip fractures, cause significant morbidity, mortality, and health care expense. There are an estimated 1.3-1.5 million osteoporosis-related fractures per year in the United States, including 700,000 vertebral, 300,000 hip, and 250,000 wrist. Health care expenditures for osteoporosis-related fractures in the United States were estimated at $13.8 billion for 1995; 70% of this cost can be attributed to hip fracture.(2) The 1-year mortality of women suffering hip fracture ranges from 5% to 20%,(3,4) and 50% of those who survive either spend some time in a nursing home(5) or require some assistance with activities of daily living.(6)
Therapies are currently available to slow and/or prevent the bone loss that leads to osteoporosis. The challenge is to identify those women with osteoporosis before fracture. This will allow targeting therapy, decreasing the incidence of osteoporotic fracture, and improving quality of life.
Dual-energy X-ray absorptiometry (DXA) is recognized as the preferred technique to assess bone mass. However, the equipment necessary to perform these tests is expensive, requires either a trained technologist or physician to operate, and is not portable. Consequently, patient access to diagnostic equipment is often limited. The use of lower cost, portable peripheral technologies has been suggested as a means to increase osteoporosis diagnosis before fracture.(7)
One such technique is quantitative ultrasound (QUS) of the calcaneus, which has been reported to predict fracture risk.(8–11) The purpose of this study was to assess the sensitivity and specificity of QUS of the calcaneus for the diagnosis of osteoporosis in women aged 50 years and older in order to determine the utility of this peripheral technology in women at risk of osteoporosis. At a T score of −1, the sensitivity of estimated bone mineral density (BMD) at the calcaneus for the diagnosis of osteoporosis (62%) is better than that of DXA of the posterior-arterior (PA) spine (49%) or total hip (32%) but not as good as DXA of the femoral neck (81%). At a T score of −1, the specificity of QUS for the diagnosis of osteoporosis is 72%. Performing risk factor assessment on individuals with QUS T scores > −0.99 can further increase the sensitivity for osteoporosis diagnosis to 81%, but the specificity decreases to 58%. Osteoporosis therapy often is based on bone mass at the spine and/or hip. A significant limitation of QUS of the calcaneus, either alone or in combination with risk factor assessment, is the relatively low sensitivity reflecting the failure to diagnose some women with osteoporosis and the low specificity for osteoporosis diagnosis reflecting the inclusion of women with low or normal axial bone mass.
MATERIALS AND METHODS
This study was approved by the Human Subjects Committee at the University of Massachusetts Medical School, and informed consent was obtained from all participants. Three hundred and twelve volunteers were recruited by a newspaper advertisement offering free bone density testing for postmenopausal women aged 50 years and older. All participants provided information regarding age; weight; height; age of menopause; diagnosis of rheumatoid arthritis; previous hip, rib, or wrist fracture; and use of medications that affect bone metabolism, from which we calculated a simple calculated osteoporosis risk estimation (SCORE).(12) This instrument was developed by statistical modeling of over 100 patient-reported variables. The result was a series of six questions (age, weight, height, history of fracture, history of rheumatoid arthritis, and estrogen use), which were converted into a numerical score in which values of ≥6.0 predicted low bone mass (T score of ≤ −2.0 at the femoral neck).(12)
Bone turnover was determined by urine cross-linked N-telopeptide of type I collagen (NTx; Osteomark; Ostec, Seattle, WA, U.S.A.). A second-void, morning urine sample was collected; all specimens were frozen, stored at −80°C, and assayed at the end of the study. Raw values were corrected for urinary dilution with urinary creatinine levels, and results were expressed in millimoles bone collagen equivalent per millimoles creatinine.
BMD was measured at the lumbar spine (L1-L4) and the hip (total hip and femoral neck) using DXA (QDR 4500; Hologic, Waltham, MA, U.S.A.).
BMD was estimated at the calcaneus using the Sahara Clinical Bone Sonometer (Hologic). This new portable ultrasound instrument measures transmission of sound through the soft tissue and bone of the calcaneus. The instrument measures both speed of sound (SOS) and broadband ultrasound attenuation (BUA). These are combined into a joint measurement called the quantitative ultrasound index (QUI), which is then converted into an estimated BMD based on the correlation between QUI and calcaneus DXA. The estimated BMD is plotted against an age-dependent reference population, and both the estimated BMD and the T score are displayed.
Quality assurance phantom scans were performed daily on both the QDR 4500 and the Sahara instruments. BMD measurements were reported as T scores based on the mean and SD of the manufacturer's reference population of young white U.S. females using the NHANES database for femoral neck density.
Statistical analysis was done using the SAS statistical software. Correlations are Pearson and Spearman rank correlations as appropriate. All significances are set at 0.05 level. Values are expressed as the mean ± SD.
Patient characteristics are shown in Table 1. Subjects' ages ranged from 50 to 85 years (mean 62 ± 9 years) and age at menopause from 26 to 58 years (mean, 48.5 ± 5.5 years). Height ranged from 142.5 to 177.5 cm (mean, 159 ± 6.5 cm) and weight from 39.5 to 119 kg (mean, 68.4 ± 13.3 kg). SCORE values ranged from −6.0 to 25 (mean, 9 ± 5), and urine NTx ranged from 9 to 244 (mean, 47.8 ± 30.6).
Table Table 1.. Subject Characteristics
Correlation coefficients between T scores at various sites and with SCORE and NTx levels are shown in Table 2. The correlations of BMD at various sites were high as expected, ranging from 0.61 to 0.88. SCORE correlations with axial BMD ranged from −0.33 to −0.52. Urine NTx correlations with spine and hip BMD ranged from −0.28 to −0.31. The correlation between urine NTx and SCORE was a modest 0.13. Estimated BMD at the calcaneus determined by QUS had correlations of 0.44-0.50 with axial BMD. The correlation coefficient of QUS and SCORE was −0.21 and with urine NTx was −0.28.
Table Table 2.. Correlation Coefficients
Osteoporosis was defined as a T score ≤ −2.5 at the spine, femoral neck, or total hip. Normal was defined as T scores ≥ −0.99 at the spine, femoral neck, and total hip. All other volunteers were defined as osteopenic, that is, at least one T score between −1.0 and −2.49 and no T scores ≤ −2.5. Ninety-four of our volunteers (30%) were found to have osteoporosis. One hundred fifty-eight women (51%) had low bone mass, and 60 women (19%) had normal BMD at all axial sites.
The sensitivity of the various central measurements (spine, femoral neck, and total hip) for the prediction of skeletal status is shown in Table 3. The femoral neck had the greatest sensitivity for the assessment of skeletal status detecting 81% of women with osteoporosis and 84% of women with osteopenia. Total hip BMD detected only 32% of the women with osteoporosis.
Table Table 3.. Sensitivity of the Central Sites for Determination of Skeletal Status
The usefulness of QUS to diagnose the skeletal status of our volunteers was evaluated by determining the sensitivity and specificity of the peripheral measurement for the diagnosis of osteoporosis at the axial skeleton (spine, femoral neck, and/or total hip T score ≤ −2.5). As shown in Table 4 at a T score of −2.0, QUS had a sensitivity of 24% and a specificity of 93% for the diagnosis. Only 23 of the 94 women with osteoporosis diagnosed by central DXA had T scores ≤ −2.0 by QUS. Fifteen women who did not have osteoporosis by central DXA had QUS T scores ≤ −2.0. These 15 women all had osteopenia at the spine, femoral neck, and/or total hip. At a T score ≤ −1.0, QUS had a sensitivity of 62% for the diagnosis of osteoporosis (58 of the 94 women with osteoporosis were detected) with a specificity of 72%. Fifty-nine women who did not have osteoporosis had QUS T scores ≤ −1.0. Fifty-four of these women had osteopenia, and 5 women had normal BMD at the three central sites. At a QUS T score ≤ 0, the sensitivity was 89% (84 of 94 women detected), but the specificity was only 32%.
Table Table 4.. Sensitivity and Specificity of QUS for the Diagnosis of Osteoporosis of the Spine and Hip
A SCORE ≥ 6.0 had a sensitivity and specificity of 89% and 29%, respectively, for the diagnosis of osteoporosis, essentially identical to a QUS T score ≤ 0. Neither the simultaneous combination of urine NTx and QUS nor the SCORE value plus QUS improved the sensitivity and specificity for the diagnosis of osteoporosis at the central skeleton.
One hundred and ninety-five women had QUS T scores > −0.99. This included 36 women with osteoporosis at the spine and/or hip. The SCORE values for these 36 women and the remaining 159 women with QUS T scores > −0.99 are shown in Fig. 1. The average SCORE value for the women with osteoporosis was 11.4 ± 4.2, whereas that of the nonosteoporotic women was 7.7 ± 4.1. Eighteen women with osteoporosis had SCORE values of >11, while 31 nonosteoporotic women has SCORE values of >11. Of the latter 31 women, 25 women had low bone mass, and 6 women had normal BMD at the three central sites. QUS of the calcaneus followed by risk factor assessment (SCORE > 11) in women with QUS T scores ≥ −0.99 gave a sensitivity for the diagnosis of osteoporosis of 81% (76 of 94 women) with a specificity of 58% (128 of 218 women).
The World Health Organization has defined osteoporosis as a T score ≤ −2.5. Using a T score ≤ −2.5 at either the PA spine, femoral neck, or total hip, we diagnosed 94 of our 312 women as having osteoporosis. This represents a prevalence of 30%, comparable with the 30% prevalence reported by Melton(13) but less than the 53% prevalence observed by Nelson.(14) Both Melton(13) and Nelson(14) also used a T score of ≤ −2.5 to diagnose osteoporosis. One hundred fifty-eight (51%) of our volunteers were diagnosed as having low bone mass. Therefore, 81% of our study population had either osteoporosis or low bone mass, comparable with the 84% reported by Melton(13) and the 91% reported by Nelson.(14)
Ninety-four of our volunteers were diagnosed as having osteoporosis based on measurements of three central sites. Only 30 women (32%) had osteoporosis at the total hip, and only 46 women (49%) had osteoporosis at the spine. The femoral neck was the one most sensitive site for the diagnosis of osteoporosis in our population with 76 women (81%) having T scores ≤ −2.5 at the femoral neck.
The low sensitivity of the spine measurement may reflect artifactual increases in BMD in this older population. Nevertheless, none of the central measurements alone diagnosed all the women with osteoporosis. However, because osteoporosis was defined by the central measurement, the false-positive rate was 0 (100% specificity).
The SCORE questionnaire was developed from linear regression models and used to identify women with low bone mass at the femoral neck.(12) At a cut point of 6, the screening tool was shown to have a sensitivity of 94% and a specificity of 43% for osteoporosis (T score ≤ −2.0) at the femoral neck. In our population, the SCORE questionnaire had a sensitivity of 89% and a specificity of 29% for the diagnosis of osteoporosis at the spine and/or hip based on a T score of ≤ −2.5. Thus, 71% of those patients with normal BMD would be identified for further testing. At the suggested cut point of 6, a recent study has reported that SCORE had a sensitivity of 90% and a specificity of 32% for the identification of women with T scores at the spine and/or femoral neck ≤ −2.(15) In a population of older white women, mean age of 72.5 years, the sensitivity of SCORE was 98%, but the specificity was only 12.5%.(16) In these older women, increasing the SCORE cut point to 11 decreased the sensitivity to 80% but increased specificity to 45%. At present, SCORE alone has limited value for identifying women with low BMD.(15,16)
QUS is becoming an increasingly popular method to assess bone mass based on its relatively low cost, lack of ionizing radiation, and the ease with which it may be moved from one office site to another. In contrast to DXA of the axial skeleton in which both the spine and the hip are assessed, QUS usually quantitates bone mass at only one skeletal site. This leads to potential errors in diagnosis because of the discordance in bone mass measurements at various sites.(17) This discordance may reflect biological variability, variations in precision of the technologies utilized, or differences in the normative databases for the various instruments. A recent study has suggested that ultrasound measurements at multiple skeletal sites improve fracture discrimination.(18)
Fifty-eight of the 94 women with osteoporosis had a QUS T score ≤ −1.0. This represents a sensitivity of 62%, better than that of lumbar spine or total hip DXA. However, the specificity was only 72%. Of the 117 women with T scores ≤ −1.0, 58 women (49%) had osteoporosis, 54 women (47%) had low bone mass, and 5 women (4%) had normal BMD at the central sites. Thirty-six women with osteoporosis were not detected. It has been suggested that risk factors might be used to further evaluate women determined to have normal bone mass by peripheral technologies.(19) The patients with risk factors but normal peripheral BMD might need central DXA testing.(19) One hundred ninety-five women had QUS T scores ≥ −0.99, including the 36 women with osteoporosis. Eighteen of these osteoporotic women had SCORE values greater than 11. Therefore, combining a QUS SCORE ≤ −1.0 followed by a SCORE evaluation with a cut point of 11 in those women with T scores ≥ −0.99 detected 76 of the 94 women with osteoporosis, comparable with the femoral neck DXA sensitivity of 81%. This combination of QUS followed by SCORE gave a specificity of 58%.
However, of the 166 women detected by QUS and subsequent risk factor evaluation, 76 women (46%) had osteoporosis, 80 women (48%) had low bone mass, and 10 women (6%) had normal BMD at the central sites. Conversely, 18 women with osteoporosis were not detected either by the ultrasound measurement or the risk factor assessment. The combination of QUS followed by risk factor evaluation provides a sensitivity for osteoporosis diagnosis comparable with or better than any one central skeletal measurement. This will facilitate the diagnosis of osteoporosis in women at risk. If QUS and risk factor assessment are the only means of evaluation before therapy, a small percentage of women will be treated despite normal BMD at the central skeleton, and nearly 20% of the women with osteoporosis will remain undiagnosed.
Previous studies have not found QUS to be a good initial osteoporosis diagnostic technique.(20,21) Massie et al.(20) noted that in their study of 1000 women between the age of 45 and 49 years, only 44% of the women in the lowest quartile by DXA were also in the lowest quartile by BUA. Rosenthal et al.(21) reached the same conclusion based on the poor correlation between QUS measurements and DXA at the spine (r = 0.60) and femoral neck (r = 0.59). In the present study, the observed correlations between QUS and DXA T scores ranged between an r value of 0.44 and 0.50. The correlations are weaker than those reported by Greenspan et al. who observed correlations between calcaneal ultrasound and DXA of the spine and hip that ranged between an r value of 0.58 and 0.82.(22) In contrast, Grampp et al.(23) observed correlations of r = 0.31-0.67 between QUS measurements and DXA at the spine and femoral neck. These variations may reflect the use of different QUS instruments and/or different populations studied.
QUS is a significant predictor of fracture.(8–11,24,25) Therefore, a low QUS value is of diagnostic importance regardless of BMD at the central skeleton. Nevertheless, the benefits of increased diagnosis of osteoporosis with single site peripheral skeletal measurements will need to be evaluated considering potentially unnecessary treatment in some women with normal BMD at the central skeleton and lack of treatment in some women with osteoporosis at the central skeleton. The benefits of increased diagnosis for the majority will need to be assessed considering potentially unnecessary treatment in a minority if treatment decisions are based on assessment of bone mass at one peripheral skeletal site.