Meta-Analysis: Accuracy of Quantitative Ultrasound for Identifying Patients with Osteoporosis

Context Can calcaneal quantitative ultrasound accurately identify adults with osteoporosis? Contribution This meta-analysis of 25 studies summarizes current knowledge about the accuracy of calcaneal quantitative ultrasound for identifying adults with a dual-energy x-ray absorptiometry (DXA) T-score of 2.5 or less at the hip or spine. The authors found no quantitative ultrasound thresholds at which sensitivity or specificity was sufficiently high to rule out or rule in DXA-determined osteoporosis. Cautions These studies did not evaluate benefits or harms of including quantitative ultrasound in screening programs. Implications Calcaneal quantitative ultrasound results at commonly used thresholds do not definitively exclude or confirm DXA-determined osteoporosis. The Editors Osteoporosis, a systemic skeletal disease characterized by low bone mass and microarchitectural deterioration of bone tissue, with a consequent increase in bone fragility and susceptibility to fracture (1), affects approximately 200 million people worldwide (2). In the United States, osteoporosis affects approximately 10 million persons, contributes to 1.5 million fractures annually, and accounted for direct costs of $18 billion in 2002 (3). Although medical therapies for patients with osteoporosis are available and reduce fracture risk (4-10), most affected individuals are asymptomatic, undiagnosed, and untreated (11). Several organizations, including the U.S. Preventive Services Task Force (12), recommend screening; however, there is no consensus on how to screen patients for osteoporosis. Dual-energy x-ray absorptiometry (DXA) is the most widely used method for diagnosing osteoporosis in most countries (13). This test involves positioning the body site of interest in the path of an x-ray beam and measuring beam attenuation, which is related to bone mineral content. Bone mineral density (BMD) is calculated as the ratio of bone content to the scanned area (14). The World Health Organization's (WHO) operational definition for osteoporosis is a BMD that is 2.5 SDs (T-scores) or more below the mean for young healthy adult women; the WHO's operational definition of osteopenia is a T-score between 1 and 2.5 (15). Numerous DXA devices are currently in use. Correlation between DXA BMD measurements obtained at the same central site (lumbar spine or femoral neck) with different devices has been reported to be 0.92 to 0.99 in several studies (16-19). Recently, there has been increased interest in the use of quantitative ultrasound for osteoporosis screening. Calcaneal quantitative ultrasound for bone assessment typically involves placing ultrasound transducers on either side of the calcaneus; one acts as a wave transmitter, and the other acts as the receiver (20). These devices assess 3 main types of parameters: broadband ultrasound attenuation, speed of sound or velocity of sound, and quantitative ultrasound index stiffness. Broadband ultrasound attenuation measures the frequency dependence of attenuation of the ultrasound signal that occurs as energy is removed from the wave, primarily by absorption and scattering in the bone and soft tissue (21). Speed of sound and velocity of sound measure the distance the ultrasound signal travels per unit of time (22). Quantitative ultrasound index and stiffness are composite parameters derived from broadband ultrasound attenuation and speed of sound or velocity of sound (21, 22). Ultrasound parameter values are typically lower in osteoporotic bone than in healthy bone (22). There are numerous calcaneal quantitative ultrasound devices in use, but there are no universal guidelines establishing normal versus abnormal measurement values. In addition, studies have reported correlation coefficient values between 0.44 and 0.93 for measurements of the same parameters by different quantitative ultrasound devices (23, 24). Several large prospective studies have shown that calcaneal quantitative ultrasound can predict future fracture risk nearly as well as DXA (25-28). Quantitative ultrasound also has several potential advantages over DXA: It is less expensive, is portable, does not involve ionizing radiation, and does not require specially trained personnel (29-32). Also, unlike DXA, quantitative ultrasound may be able to assess bone quality in addition to BMD (33-35). However, 2 key gaps in the evidence limit the use of quantitative ultrasound as a first-line diagnostic tool in clinical practice. First, there are no consensus diagnostic criteria for osteoporosis using this technique. The WHO's operational definition for osteoporosis was derived in the context of DXA and has typically been applied to DXA (36). Direct application of this definition to quantitative ultrasound is not advisable (37, 38). Second, clinical trials of the efficacy of medical therapies for reducing fracture risk in persons without a history of osteoporotic fracture have used DXA rather than quantitative ultrasound to select patients (39). It is not known whether the results of these trials can be generalized to patients identified by quantitative ultrasound as having high risk for fracture (39). Some evidence suggests that women selected for osteoporosis therapy on the basis of fracture risk factors rather than low DXA BMD may not benefit similarly from treatment (7). In the absence of direct evidence of treatment efficacy for patients identified by quantitative ultrasound as having high risk for fracture, the clinical utility of this test for improving osteoporosis outcomes lies with its degree of correlation with DXA results (40). Correlation coefficients between calcaneal quantitative ultrasound measurements and DXA BMD at the spine or the hip have ranged between 0.27 and 0.7 in several larger studies (41-51). Thus, several researchers have suggested that quantitative ultrasound could be used as a prescreening test to reduce the number of patients who require additional DXA testing (52-61). We performed a systematic review to address 3 questions relevant to such a strategy. First, what are the sensitivity and specificity of calcaneal quantitative ultrasound for identifying patients who meet WHO DXA osteoporosis criteria at the hip or the spine? Second, given a pretest probability of osteoporosis (for example, on the basis of risk factors, such as age and sex) and quantitative ultrasound results, what is the post-test probability of DXA-determined osteoporosis? Third, what do these findings tell us about the strength of the evidence supporting the use of calcaneal quantitative ultrasound to screen for osteoporosis? Methods Data Sources We searched MEDLINE (1966 to October 2005), EMBASE (1993 to May 2004), Cochrane Central Register of Controlled Trials and Cochrane Database of Systematic Reviews (1952 to March 2004), and the Science Citation Index (1945 to April 2004) with assistance from a professional research librarian (Figure 1). We supplemented our searches by manually reviewing bibliographies of eligible studies and relevant review articles. Figure 1. Literature search strategies and reasons for study exclusion. Study Selection We included English-language studies that evaluated the sensitivity and specificity of calcaneal quantitative ultrasound for identifying adults with DXA T-scores of 2.5 or less at the hip or spine. We required that both sites be tested, with a T-score of 2.5 or less at either site indicative of osteoporosis (62, 63), because T-scores can differ at the lumbar spine and the hip (62-64) and the spine is often affected by bone loss earlier than the femoral neck. Thus, if only hip BMD was tested, some individuals at risk for vertebral fracture may have been missed. We chose to focus on studies with a DXA T-score of 2.5 or less as the reference standard because we felt this was the clinical population of most interest. Most of the randomized, controlled trials that have demonstrated efficacy of pharmacologic therapies for reducing fracture risk in persons without a history of fracture have done so in this population. In addition, several guidelines agree that persons with T-scores of 2.5 or less should be treated (65-67), although there is more controversy surrounding treatment of those without a history of fracture and T-scores greater than 2.5. We excluded studies that did not use DXA as the reference standard because the bulk of the evidence showing that medical therapy reduces fracture risk in persons without a history of osteoporotic fracture has been based on patient selection by DXA criteria. We limited inclusion to studies that performed quantitative ultrasound and DXA testing in all participants, had at least 10 participants with and 10 participants without DXA-determined osteoporosis, and reported at least 1 pair of sensitivity and specificity values (Figure 1). Data Extraction Two authors independently abstracted study design information, study participant information, results, and information about potential sources of bias from included studies (Appendix Table 1). We resolved abstraction discrepancies by repeated review and discussion. Appendix Table 1. Study Characteristics Data Synthesis All of the studies that met our inclusion criteria reported quantitative ultrasound thresholds (cutoff values used to separate positive results from negative results) corresponding to each pair of sensitivity and specificity values. We used this information to determine the relationship between threshold and sensitivity and specificity. We computed random-effects regression models with sensitivity or specificity as the dependent variable and threshold as the independent variable, as will be explained. We also calculated summary receiver-operating characteristic (ROC) curves using the sensitivity and specificity estimates reported by the included studies. We used MATLAB, version 7.0, release 14 (The MathWorks, Inc., Natick, Massachusetts), and STATA, version 8.1 (StataCorp LP, College Station, Texas), to perform our data an