Quantitative heel ultrasound as a predictor for osteoporosis

Vasi Naganathan, Lyn March, David Hunter, Nick A Pocock, Joanna Markovey and Philip N Sambrook
Med J Aust 1999; 171 (6): 297-300.
Published online: 20 September 1999

Quantitative heel ultrasound as a predictor for osteoporosis

Vasi Naganathan, Lyn March, David Hunter, Nick A Pocock,
Joanna Markovey and Philip N Sambrook

MJA 1999; 171: 297-300
For related articles, see Prince, Maguire & Lobb et al

Abstract - Introduction - Methods - Results - Discussion - References - Authors' details
- - More articles on Rheumatology

Abstract Objective: To determine the diagnostic value of quantitative ultrasound (QUS) to predict bone mineral density (BMD) categories as defined by dual-energy x-ray absorptiometry.
Design: Cross-sectional survey.
Setting: Rheumatology department of a tertiary care hospital (Royal North Shore Hospital, Sydney, NSW), 1997-1998.
Subjects: 326 healthy women aged 45-80 years who had volunteered for a twin study. Our study included both members of non-identical twin pairs but only one randomly selected member of identical twin pairs.
Main outcome measures: BMD categories as defined by dual-energy x-ray absorptiometry of lumbar spine and left hip, and QUS of calcaneus; sensitivity, specificity and likelihood ratios (LRs) of QUS parameters to diagnose osteoporosis as defined by BMD.
Results: The sensitivity of QUS to diagnose BMD osteoporosis varied between 9% and 47%, depending on the QUS parameter. The specificity of QUS was high (88%-100%). If all QUS parameters were normal, osteoporosis was unlikely (LR, 0-0.2). One QUS parameter, broadband ultrasound attenuation (BUA), was highly predictive of osteoporosis by BMD when in the osteoporotic range (LR, Infinity), but had low sensitivity (9%). QUS results in the osteoporotic range for other parameters and all QUS results in the osteopenic range were less predictive (LR, 1.0-5.2) of osteoporotic BMD.
Conclusion: These results suggest that, for most of those tested for osteoporosis by QUS in the community, uncertainty remains about expected BMD.

Introduction Quantitative heel ultrasound has recently been introduced in Australian pharmacies as a "screening" tool for osteoporosis. The technology is relatively cheap, radiation-free and portable, but its accuracy in diagnosing osteoporosis is unclear.

Bone mineral density (BMD), measured by dual energy x-ray absorptiometry (DEXA), is the best predictor of fracture risk and is currently considered the "gold standard" for diagnosing osteoporosis. Although prospective studies have shown that quantitative ultrasound (QUS) predicts future fracture risk independently of BMD,1,2 most women with abnormal results will proceed to formal BMD measurement to assess the need for therapeutic intervention; women with normal QUS results may be reassured they do not need BMD measurement. It is unclear how many women have unnecessary further investigations or are falsely reassured. When used in this way, QUS has diagnostic value only if it can accurately predict BMD categories as determined by DEXA.

Our aim was to examine the role of QUS in predicting BMD diagnostic categories. We determined conventional sensitivity and specificity, as well as likelihood ratios (LRs). These have the advantage of allowing test results to be assessed for several diagnostic categories rather than only at a single cut-off between "normal" and "abnormal".3

Subjects and setting
The study was a cross-sectional survey of healthy women aged 45-80 years who had volunteered to take part in a twin study. They were recruited from the Australian Twin Registry and media advertising. Our ultrasound study included both members of each non-identical twin pair and one randomly selected member of each identical twin pair.

The study was conducted in the Rheumatology Department of the Royal North Shore Hospital, Sydney, NSW (a tertiary care hospital), in 1997 and 1998. It was approved by the hospital's Human Research Ethics Committee.  

Assessment Subjects had BMD measurements of their lumbar spine (L1-L4) and left hip (neck of femur and total hip) by DEXA using a Hologic QDR450 instrument (Hologic Inc, Waltham, Mass, USA). The same machine was used on all patients.

QUS of the left calcaneus was performed on the same day using a CUBA Mark II ultrasound instrument (McCue Ultrasonics, London, UK). The two most commonly used QUS parameters were measured: broadband ultrasound attenuation (BUA), which is thought to reflect bone mass and architecture, and velocity of sound (VOS), which reflects mass and elasticity of bone.4 

Analyses Each BMD and QUS value was converted to a T score (number of standard deviations from the population mean for young, healthy, sex-matched adults). This population mean was estimated from measurements in 50 women aged 20-30 years who also took part in the twin study. T scores were used to categorise BMD values as normal (T > -1) or indicating osteopenia (T, -2.5 to -1) or osteoporosis (T < -2.5), as proposed by a working party of the World Health Organization.5 QUS values were classified in the same way. Although no consensus has been reached on what T-score cut-offs and diagnostic categories to use with QUS, the instrument used commonly in Australian pharmacies uses the WHO criteria and cut-off values.

Subjects were classified as having osteoporosis if at least one of the three BMD measurements (lumbar spine, neck of left femur or total left hip) indicated osteoporosis, and as having osteopenia if at least one measurement indicated osteopenia but none indicated osteoporosis.

BUA and VOS results were combined as a cQUS category: this was defined as normal if both results were normal, as osteopenic if either indicated osteopenia but neither indicated osteoporosis, and as osteoporotic if either indicated osteoporosis. Some QUS scanners calculate a stiffness parameter (unrelated to mechanical stiffness) from a linear combination of normalised BUA and VOS values. We calculated stiffness in an analogous manner,6 and categorised it as normal, osteopenic or osteoporotic based on T scores in the same way as other QUS values.

We calculated the sensitivity and specificity of QUS parameters in predicting BMD-defined osteoporosis and osteopenia. We also calculated the likelihood ratio (LR) for each QUS result (sensitivity/1 - specificity), defined as the ratio of the probability of the particular QUS result (normal, osteopenic or osteoporotic) in women with BMD-defined osteoporosis or osteopenia to the probability of the same result in women with normal BMD.7

As there is no consensus on what QUS cut-off values should be used to diagnose osteoporosis, the statistical analyses were repeated using a range of QUS T-score cut-off values for osteoporosis between -2.5 and -1.0.

Subjects and osteoporosis
There were 326 subjects, with mean age 58.5 years; 255 (78%) were postmenopausal. Of the 326, 47 (14%) had a BMD measurement indicating osteoporosis at one or more of the three sites where BMD was measured, and a further 160 (49%) had a measurement indicating osteopenia.

QUS results are compared with BMD results in Box 1. The percentage of women with values in the osteoporotic range varied between QUS parameters (1% for BUA, 17% for VOS and the combined BUA-VOS category, and 14% for stiffness).

Sensitivity and specificity of QUS
Sensitivity and specificity of QUS for predicting BMD diagnostic categories are shown in Box 2. Sensitivity and specificity varied between QUS parameters. A BUA result in the osteoporotic range (T < -2.5) had very low sensitivity for predicting BMD-defined osteoporosis (9%), but high specificity (100%). In contrast, VOS, cQUS and stiffness results in the osteoporotic range had sensitivities of almost 50%, and specificities that were again high. For predicting either osteoporosis or osteopenia, stiffness had the best combination of sensitivity (77%) and specificity (81%). Positive and negative predictive values are also shown in Box 2. Negative predictive values were high (87%-91%) for QUS as a predictor of BMD-defined osteoporosis versus osteopenia/normal BMD. This indicated that a woman with BMD-defined osteoporosis was unlikely to have a QUS result in the normal-osteopenic range.

Likelihood ratios of QUS
LRs for different QUS results to predict BMD-defined osteoporosis are summarised in Box 2 and interpreted in Box 3.

Normal QUS result: A BUA, VOS, cQUS or stiffness result in the normal range had a low LR (0-0.2) (ie, a normal result significantly lowered the odds or probability of the woman's having BMD-defined osteoporosis).

Osteoporotic QUS result: A BUA result in the osteoporotic range had an LR approaching infinity and so was highly predictive of BMD-defined osteoporosis. In contrast, a VOS, cQUS or stiffness result in the osteoporotic range had a much lower LR (4.0-5.2), increasing the odds of BMD-defined osteoporosis, but to a lesser extent than a BUA result in the osteoporotic range.

Osteopenic QUS result: QUS results in the osteopenic range were less predictive, as LR ranged from 1.0 to 2.4. Between 37% and 50% of subjects (depending on the QUS parameter) had results in this range (Box 1).

LRs for predicting low BMD (ie, osteoporosis or osteopenia; BMD T score < -1) are also shown in Box 2, and followed a similar pattern to LRs for predicting BMD-defined osteoporosis.

When QUS T-score cut-off values for osteoporosis were increased from -2.5 to -1.0, sensitivity increased, but at the expense of decreasing specificity and LR (data not shown). For example, a BUA cut-off of -1.0 gave 83% sensitivity, 69% specificity and LR, 2.6. Corresponding values for a VOS cut-off of -1.0 were 96%, 41% and 1.6.

Discussion We found that QUS had variable usefulness in predicting BMD categories. Specificity for predicting BMD-defined osteoporosis was high for all QUS parameters (88%-100%), but sensitivity was low and variable (9%-47%). A BUA result in the osteoporotic range was highly predictive of BMD-defined osteoporosis (LR, Infinity), but had low sensitivity (9%). Results in the osteoporotic range for other QUS parameters and in the osteopenic range for all QUS parameters were less predictive of BMD-defined osteoporosis (LRs, 1.0-5.2).

In the light of our study, how can we interpret QUS results?

  • If a BUA result is in the osteoporotic range (LR, Infinity), then BMD-defined osteoporosis is almost certain (predictive value, 100%). However, the low sensitivity of BUA (9%) means that many women with osteoporosis would be missed if BUA alone were used.
  • If results are
  • normal for both BUA and VOS (cQUS normal; LR, 0), we can confidently rule out BMD-defined osteoporosis. If LR is 0, then, no matter what the pre-test probability of osteoporosis, the post-test probability will be < 5% (Box 3).
  • All results in the osteopenic range, and VOS and stiffness results in the osteoporotic range, are less predictive of BMD category.

Therefore, if QUS were performed on a population similar to ours (14% prevalence of osteoporosis), then (from Box 1) 1% would have a BUA result in the osteoporotic range (likely to have osteoporosis) and 33% would have both BUA and VOS results in the normal range (osteoporosis could fairly confidently be ruled out, with a post-test probability < 5%). However, there would be a degree of uncertainty about the remaining 66%, who would then need a DEXA scan to identify those with osteoporosis.

Previous studies of QUS as a predictor of BMD have generally used conventional sensitivity and specificity analyses only, not LRs, and have not used the WHO BMD definitions. For example, two community-based cross-sectional studies on 700 postmenopausal10 and 1000 perimenopausal women,11 respectively, found that there was a 40%-50% overlap in the number of women in the lowest quartile of both DEXA and QUS measurements. Two other studies found QUS parameters to have a sensitivity of 65%-70% for BMD in the lowest quartile.6,12 Only one study other than ours has evaluated QUS in terms of WHO BMD definitions. It found BUA and VOS to have higher sensitivities, of 77% and 69%, respectively, for diagnosing osteoporosis in 100 women aged 60-69 years.13These higher sensitivities may have been due to use of higher BUA and VOS cut-off values. As expected, specificities were lower than in our study.

The ultrasound instrument used in this study, the McCue Cuba Mark II, is not identical to the Achilles ultrasound instrument (Lunar, Madison, Wis, USA) used in Australian pharmacies. Nevertheless, a comparison of the two machines found that BUA measurements on a Cuba Mark II instrument were highly correlated with "stiffness" measurements on a Lunar Achilles instrument (r = 0.906; 95% CI, 0.873-0.931).14 Another study compared measurements of 30 women between the Cuba Mark II used in our study and an Achilles, finding an r value of 0.8.15 Therefore, it is unlikely that the Achilles instrument would be a significantly better predictor of BMD than our Cuba Mark II.

There is no consensus on what cut-off values to use with QUS to diagnose osteoporosis. We found that changing the cut-off could achieve higher sensitivity, but only by accepting higher rates of false positives (lower specificity) and less discriminating LRs.

Although there is enough evidence to support the use of QUS as an independent predictor of fracture risk,1,2 our study shows that QUS should not been seen as a substitute for BMD measurement. The results of our study suggest that, when women in the community are "screened" for osteoporosis using QUS, a few women will be confidently identified with BMD-defined osteoporosis. Another small group will be able to be reassured that they are unlikely to have osteoporosis. However, for the great majority, the presence or absence of osteoporosis will remain uncertain.

  1. Hans D, Dargent MP, Schott AM, et al. Ultrasonographic heel measurements to predict hip fracture in elderly women: the EPIDOS prospective study. Lancet 1996; 348 (9026): 511-514.
  2. Bauer DC, Gluer CC, Cauley JA, et al. Broadband ultrasound attenuation predicts fractures strongly and independently of densitometry in older women. A prospective study. Study of Osteoporotic Fractures Research Group. Arch Intern Med 1997; 157: 629-634.
  3. Sackett DL, Haynes BR, Guyatt GH, et al. Clinical epidemiology. A basic science for clinical medicine. 2nd ed. Boston: Little, Brown and Company, 1991.
  4. Gluer CC, Wu CY, Jergas M, et al. Three quantitative ultrasound parameters reflect bone structure. Calcif Tissue Int 1994; 55: 46-52.
  5. World Health Organization Study Group. Assessment of fracture risk and its application to screening for prostmenopausal osteoporosis. World Health Organ Tech Rep Ser 1994; 843: 1-129.
  6. Herd RJ, Blake GM, Miller CG, et al. The ultrasonic assessment of osteopenia as defined by dual X-ray absorptiometry. Br J Radiol 1994; 67: 631-635.
  7. Fletcher RH, Fletcher SW, Wagner EH. Clinical epidemiology: the essentials. 3rd ed. Baltimore: Williams & Wilkins, 1988: 65.
  8. Fagan TJ. Normogram for Bayes theorem [letter]. N Engl J Med 1975: 293; 257.
  9. Australian National Consensus Conference 1996. The prevention and management of osteoporosis. Consensus statement. Med J Aust 1997; 167 Suppl: S1-S15.
  10. van Daele Burger H, Algra D, et al. Age-associated changes in ultrasound measurements of the calcaneus in men and women: the Rotterdam Study. J Bone Miner Res 1994; 9: 1751-1757.
  11. Massie A, Reid DM, Porter RW. Screening for osteoporosis: comparison between dual energy X-ray absorptiometry and broadband ultrasound attenuation in 1000 perimenopausal women. Osteoporos Int 1993; 3: 107-110.
  12. Young H, Howey S, Purdie DW. Broadband ultrasound attenuation compared with dual-energy X-ray absorptiometry in screening for postmenopausal low bone density. Osteoporos Int 1993; 3: 160-164.
  13. Langton CM, Ballard PA, Bennett DK, Purdie DW. A comparison of the sensitivity and specificity of calcaneal ultrasound measurements with clinical criteria for bone densitometry (DEXA) referral. Clin Rheumatol 1997; 16: 117-118.
  14. Greenspan SL, Bouxsein ML, Melton ME, et al. Precision and discriminatory ability of calcaneal bone assessment technologies. J Bone Miner Res 1997; 12: 1303-1313.
  15. Harris ND, Griffiths MR, Nguyen TV, et al. Quantitative ultrasound of the heel: a comparison of Lunar and McCue instruments [abstract]. Proceedings of the Australian and New Zealand Bone and Mineral Society 7th Annual Scientific Meeting. 1997. 29 Sept-1 Oct; Canberra, ACT: 62.

(Received 24 Dec 1998, accepted 9 Jul 1999)

Authors' details Department of Rheumatology, Royal North Shore Hospital, Sydney, NSW.
Vasi Naganathan, FRACP, Research Scholar;
Lyn March, FRACP, FAFPHM, Staff Specialist;
David Hunter, MB BS, Advanced Physician Trainee in Rheumatology;
Joanna Markovey, MSc, Research Bone Densitometry Technician;
Philip N Sambrook, FRACP, MD, Head of Department.

Department of Nuclear Medicine, St Vincent's Hospital, Sydney, NSW.
Nick A Pocock, FRACP, MD, Senior Staff Specialist.

Reprints will not be available from the authors.
Correspondence: Dr V Naganathan, Department of Rheumatology, Royal North Shore Hospital, St Leonards, NSW 2065.

1: Association between quantitative heel ultrasound (QUS) results and bone mineral density* in 326 women aged 45-80 years

Bone mineral density
QUS resultNormal (n = 119)Osteopenia (n = 160)Osteoporosis (n = 47)Total (n = 326)

Broadband ultrasound (BUA)
 Normal103918202 (62%)
 Osteoporosis0044 (1%)
Velocity of sound (VOS)
 Normal81332116 (36%)
 Osteopenia369623155 (47%)
 Osteoporosis2312255 (17%)
Combined category (cQUS)
 Normal79280107 (33%)
 Osteopenia3810125164 (50%)
 Osteoporosis2312255 (17%)
 Normal96444144 (44%)
 Osteopenia209421135 (41%)
 Osteoporosis3222247 (14%)

* Measured by dual energy x-ray absorptiometry. Combined result for broadband ultrasound and velocity of sound: normal if both normal; osteopenic if either osteopenic and neither osteoporotic; and osteoporotic if either osteoporotic.
Back to text

2: Use of ultrasound parameters to predict osteoporosis or osteopenia defined by dual energy x-ray absorptiometry (DEXA) measurement of bone mineral density

Predictive valuesLikelihood ratio for ultrasound measurement (95% confidence interval)
Sensitivity SpecificityPositiveNegativeNormalOsteopeniaOsteoporosis

To predict osteoporosis
Broadband ultrasound (BUA)9%100%100%87%0.2 (0.11-0.38)2.4 (1.9-3.1)Infinity
Velocity of sound (VOS)46%88%40%91%0.1 (0.03-0.4)1.0 (0.7-1.4)4.0 (2.6-6.2)
Combined category* (cQUS)47%88%40%91%0(1.1 (0.8-1.5)4.0 (2.6-6.2)
Stiffness47%91%46%91%0.2 (0.08-0.5)1.1 (0.8-1.6)5.2 (3.2-8.4)
To predict osteoporosis or osteopenia
Broadband ultrasound (BUA)52%87%87%51%0.6 (0.5-0.7)3.7 (2.3-6.0)Infinity
Velocity of sound (VOS)83%68%82%70%0.25 (0.18-0.35)1.9 (1.4-2.6)15.0 (3.7-60.5)
Combined category* (cQUS)86%66%82%74%0.20 (0.14-0.29)1.9 (1.4-2.5)15.2 (3.8-61.3)
Stiffness77%81%87%67%0.3 (0.2-0.4)3.3 (2.2-5.0)8.4 (2.7-26.5)

DEXA = dual energy x-ray absorptiometry. * Combined broadband ultrasound attenuation and velocity of sound category.
Back to text

3: Interpretation of likelihood ratios

The likelihood ratio (LR) indicates how much a test result raises or lowers the probability of an individual's having "disease". It can be used to determine post-test probability of a disease from the estimated pre-test probability using a normogram (Figure). Thus, if an early postmenopausal woman (aged 60-64 years) had a pre-test probability of osteoporosis of 15%,9 then, from the normogram, a cQUS result in the osteoporot ic range (LR, 4.0) would increase her probability to 40%. If she had a very high pre-test probability of osteoporosis, for example 50%, because of multiple risk factors (eg, including family history of osteoporosis and recent wrist fracture after a fall), then a cQUS result in the osteoporotic range would increase her probability to 80%.

A BUA result in the osteoporotic range (LR, Infinity) would make BMD-defined osteoporosis highly likely, no matter what the pre-test risk. However, a cQUS result in the normal range (LR, 0) would make BMD-defined osteoporosis unlikely. QUS results in the osteopenic range, with LRs closer to unity (LR, 1.0-2.4), would make BMD category far less certain.

Normogram for applying likelihood ratios, adapted from Fagan.8 Lines show post-test probabilities when LR = 4.0 and pre-test probabilities are 15% and 50%, respectively.
Back to text

Received 23 September 2018, accepted 23 September 2018

  • Vasi Naganathan
  • Lyn March
  • David Hunter
  • Nick A Pocock
  • Joanna Markovey
  • Philip N Sambrook



remove_circle_outline Delete Author
add_circle_outline Add Author

Do you have any competing interests to declare? *

I/we agree to assign copyright to the Medical Journal of Australia and agree to the Conditions of publication *
I/we agree to the Terms of use of the Medical Journal of Australia *
Email me when people comment on this article

Responses are now closed for this article.