Abstract. The aim of this cross-sectional study was to use a novel method of data analysis to demonstrate that patients with osteoporosis have significantly lower ...
Osteoporos Int (1998) 8:141-146 © 1998 European Foundation for Osteoporosis and the National Osteoporosis Foundation
osteoporosls International
Original Article The Relationship Between Bone Mineral Density and Ultrasound in Postmenopausal and Osteoporotic Women S. S. Y e a p 1, D. P e a r s o n 2, S. A. C a w t e 2 a n d D. J. H o s k i n g Departments of iMedicine and 2Nuclear Medicine, City Hospital, Nottingham, UK
Abstract. The aim of this cross-sectional study was to use a novel method of data analysis to demonstrate that patients with osteoporosis have significantly lower ultrasound results in the heel after correcting for the effect of bone mineral density (BMD) measured in the spine or hip. Three groups of patients were studied: healthy early postmenopausal women, within 3 years of the menopause (n = 104, 50%), healthy late postmenopausal women, more than 10 years from the menopause (n = 75, 36%), and a group of women with osteoporosis as defined by WHO criteria (n = 30, 14%). Broadband ultrasound attenuation (BUA), speed of sound (SOS) and Stiffness were measured using a Lunar Achilles heel machine, and BMD of the lumbar spine and left hip was measured using dual-energy X-ray absorptiometry (DXA). SOS, BUA and Stiffness were regressed against lumbar spine BMD and femoral BMD for all three groups combined. The correlation coefficients were in the range 0.52-0.58, in agreement with previously published work. Using a calculated ratio R, analysis of variance demonstrated that the ratio was significantly higher in the osteoporotic group compared with the other two groups. This implied that heel ultrasound values are proportionately lower in the osteoporotic group compared with the other two groups for an equivalent value of lumbar spine and femoral neck BMD. We conclude that postmenopausal bone loss is not associated with different ultasound values once lumbar spine or femoral neck BMD is taken into account. Ultrasound does not give additional information about patterns of bone loss in postmenopausal patients but is important in those patients with osteoporosis and fractures. Correspondence and offprint requests to: Dr S. S. Yeap, Rheumatology Unit, City Hospital, Hucknall Road, Nottingham NG5 1PB, UK. Tel: +44 (0)115 985 8898; Fax: +44 (0)115 962 7900.
Keywords: Bone mineral density; DXA; Osteoporosis; Postmenopausal women; Reproducibility; Ultrasound
Introduction Osteoporotic fractures are becoming an increasing health care problem in the Western world. Measurement of bone mass can be used to predict the risk of future fractures [1] but not all individuals with low bone mineral density (BMD) suffer fractures. Therefore, other factors such as bone strength, fragility and microarchitecture must have a role to play, but this is difficult to measure in vivo. Quantitative ultrasound is a method that may reflect BMD as well as other characteristics of bone such as structure [2] and strength [3]. Recent prospective studies have shown that heel ultrasound measurements are an independent predictor of future fracture risk [4,5] and previous studies have shown that ultrasound values are lower in subjects who have had vertebral [6,7] or hip [8] fractures. However, ultrasound measurements are correlated with BMD [7,9] and decline with age [10,11], and since patients with fractures generally have a lower BMD compared with healthy subjects, the difficulty lies in isolating the extra factor(s) that ultrasound measures apart from BMD. In addition, postmenopausal bone loss itself leads to a decrease in BMD and could thus affect ultrasound measurements even without the presence of fractures. Previous studies have corrected for the effect of lower BMD on ultrasound measurements using statistical techniques such as multivariate analysis [5] or logistic regression [8]. This cross-sectional study used a novel method of data analysis to correct heel ultrasound measurements for varying levels of BMD measured in
142
S. S. Yeap et aI.
the spine or hip. In addition, we also examined differences in ultrasound measurements between early ( < 3 years) and late ( > 1 0 years) postmenopausal women and compared these values with those obtained in women with osteoporosis to determine whether postmenopausal bone loss was associated with different ultrasound values.
Materials a n d M e t h o d s
Subjects Three groups of subjects were studied: two groups of healthy postmenopausal women, one group within 3 years of the menopause ( < 3 YSM; n = 104), the other more than 10 years past the menopause ( > 10 YSM; n =75), and a group of women who had osteoporosis (n=30). The healthy postmenopausal women were recruited as part of a population-based study on the effect of a bisphosphonate on postmenopausal bone loss in women aged between 45 and 59 years. The poslmenopausal women with osteoporosis all fulfilled the World Health Organization (WHO) definition of osteoporosis (a BMD more than 2.5 SD below the young normal mean) [12]; 20 (67%) had also suffered an osleoporotic vertebral fracture, diagnosed on plain radiographs of the thoraco-lumbar spine. In the early postmenopausal group ( < 3 YSM) there were 7 of 104 (7%) women who were osteoporotic and 37 of 104 (36%) with osteopenia. In the late postmenopausal group ( > l0 YSM), 6 of 75 (8%) were osteoporotic and 29 of 75 (39%) had osteopenia ( W H O definitions [12]).
then positioned between the transducers, with the ultrasound beam propagating laterally through the centre of the os calcis [8]. Broadband ultrasound attenuation (BUA) and speed of sound (SOS) of the right os calcis were measured. Stiffness was calculated by the computer based on a combination of BUA and SOS measurements established by Lunar: Stiffness = (0.67 x B U A + 0.28 x SOS) - 420 (R. B. Mazess, personal communication). B M D (g/cm 2) of the lumbar spine (L1-4) and the left femoral neck was measured using a Hologic QDR 2000 dual-energy X-ray absorptiometry (DXA) machine (Hologic, Waltham, MA).
Reproducibility The short term in vivo reproducibility of the DXA measurements was measured in 280 women who attended for screening during recruitment for a drug trial and were measured twice within a median time of 20 days. Within the population there were 92 women who also had duplicate ultrasound measurements. The reproducibility was assessed using the intraclass correlation coefficient. Good reproducibility is indicated by an intraclass correlation coefficient of close to 1, suggesting that the wtriability within patients is small compared with the variability within the whole dataset. For comparison purposes the coefficient of variation and standardized coefficient of variation [13] are also given. The standardized coefficient of variation is the standard deviation of measurements divided by the clinical range of measured results. The results are given in Table 1. The short-term coefficient of variation of the European Spine Phantom was 0.5% for the Hologic machine [14].
Measlt r e n l e n t s The Achilles ultrasound device (Lunar, Madison, WI) was used. The system consists of two 2.54 cm diameter, unfocused transducers, mounted coaxially approximately 9.5 cm apart. Acoustic coupling is accomplished by submerging the transducer pair and the heel in a water bath maintained at 35 °C, with surfactant to make the foot wet. One transducer acts as the transmitter and the other as the receiver. The subject's heel is placed on the tk)ot support plate and strapped into place. The heel is
Statistical Analysis Correlation coefficients were obtained between each ultrasound measurement and B M D of the lumbar spine and femoral neck. The data were stratified for lumbar spine and femoral neck B M D and each ultrasound measurement compared among the three groups of women using an analysis of variance (ANOVA). Where a significant difference was demonstrated, a
Table 1. Reproducibility of DXA and ultrasound measurements
DXA AP spine Femoral neck
Intraclass correlation coefficient
Coefficientof variation %
Standardized coefficient of variation %
No. of patients
0.989 0.981
1.40 1.90
2.00 2.38
280 279
0.820 0.898 0.914
2.20 3.25 0.37
6.50 5.00 4.40
92 92 92
Ultrasound BUA Stiffness SOS
AP, anteroposterior; BUA, broadband ultrasound attenuation; SOS, speed of sound.
143
The Relationship Between BMD and Ultrasound Scheff6 range test was carried out. However, because of the low numbers within the osteoporosis group, and the lack of data with low values o f B M D within the other two groups, the regression o f each ultrasound measurement on B M D was calculated for all the patient groups combined, having tested that there was no significant difference between the regressions in the groups separately. The following ratio (R1) was calculated for each B M D and ultrasound measure: R1 = BMD/(Ultrasound -
intercept o f regression)
This measure will be independent of BMD. In patients with a proportionately lower ultrasound result c o m p a r e d with their B M D (i.e. the osteoporotic group), the value o f R1 will be raised and vice versa. R1 was c o m p a r e d a m o n g groups using A N O V A . The SPSS for W i n d o w s p r o g r a m m e (SPSS, Chicago, IL) was used for these statistical tests.
Results The baseline characteristics of the subjects are shown in Table 2. As would be expected, the late ( > 10 Y S M ) postmenopausal group were significantly older than the early ( < 3 Y S M ) postmenopausal group. The osteoporotic group was older than both groups of healthy postmenopausal women. The postmenopausal w o m e n (n = 280) used to measure B M D precision had a mean age of 53.3 years (range 4 5 - 5 9 years). Those used to measure ultrasound precision (n = 92) had a mean age o f 52.9 years (range 4 5 - 5 9 years). B o d y mass index (BMI) was similar in all three groups. All ultrasound values (BUA, SOS and Stiffness) as well as lumbar spine and femoral neck B M D were lower in the group with osteoporosis (Table 2). The correlation coefficients between B M D and ultrasound measurements for all three groups are shown in Table 3. There was a
reasonable correlation between lumbar spine and femoral neck B M D (r = 0.73), and the correlation between B M D and ultrasound measurements was similar for B U A , SOS and Stiffness at lumbar spine and femoral neck sites (r = 0.52-0.58). The absolute values o f the ultrasound measurements at each stratum o f B M D for the lumbar spine are shown in Table 4 and femoral neck in Table 5. There was no significant difference between the early and late postmenopausal w o m e n at any level o f B M D . However, ultrasound values were generally lower in the group with osteoporosis for any given level o f B M D and this reached significance in the 0.7-0.8 g / c m 2 groups at the lumbar spine and < 0.6 g/cm 2 groups at the femoral neck (Tables 4, 5). The Ra ratio for lumbar spine B M D and Stiffness (AP lumbar spine BMD/(Stiffness -- intercept)) was plotted against lumbar spine B M D (Fig. 1). The ratio was not correlated with B M D and the regression line not significantly different f r o m zero, as expected. Similar plots are obtained for the other B M D regions and ultrasound measurements. However, analysis o f variance demonstrates that the ratio is significantly higher in the osteoporosis group c o m p a r e d with the other two groups (Table 6).
Table 3. Pearson's correlation coefficients between BMD and ultrasound measurements (pooled data)
Femoral neck BUA SOS Stiffness
Lumbar spine
Femoral neck
BUA
SOS
0.73 0.55 0.55 0.58
0.56 0.53 0.56
0.82 0.94
0.96
All significant at p < 0.001.
Table 2. Baseline characteristics of study groups Postmenopausal women
Number Age (years) BMI (kg/m2) BUA (dB/MHz) SOS (m/s2) Stiffness Lumbar spine BMD (g/cm2) Femoral neck BMD (g/cm2)
Osteoporosis group
Early (YSM < 3)
Late (YSM > 10)
104 (50%) 51.1 _+3.2* 25.4 _ 3.1 116.2 + 10.2 1536 _+32 87.7 _+14.7 0.97 + 0.14 0.77 +_0.12
75 (36%) 54.5 _+3.5* 25.5 + 4.1 114.3 ± 9.3 1533 _+31 85.3 _+14.0 0.95 --t-_0.13 0.74 + 0.10
30 (14%) 65.9 _+10.6" 25.5 + 4.6 99.8 _ 7.4** 1492 _+21 ** 64.4_+ 10.3"* 0.72 + 0.11 ** 0.57 _ 0.10"*
All values mean + SD. YSM, years since menopause; BMI, body mass index; BUA, broadband ultrasound attenuation; SOS, speed of sound. *Significantly different p < 0.0001. **Significantly lower p
