Osteoporos Int DOI 10.1007/s00198-014-2856-5
ORIGINAL ARTICLE
Age-related normative values of trabecular bone score (TBS) for Japanese women: the Japanese Population-based Osteoporosis (JPOS) study M. Iki & J. Tamaki & Y. Sato & R. Winzenrieth & S. Kagamimori & Y. Kagawa & H. Yoneshima
Received: 12 May 2014 / Accepted: 12 August 2014 # International Osteoporosis Foundation and National Osteoporosis Foundation 2014
Abstract Summary Trabecular bone score (TBS), a surrogate measure of bone microarchitecture, represents fracture risk independently of bone density. We present normative TBS values from a representative population study of Japanese women. This database would enhance our understanding of trabecular bone microarchitecture and improve osteoporosis management. Introduction TBS is a texture parameter that quantifies local variation in gray level distribution within dual-energy X-ray absorptiometry (DXA) images of the lumbar spine. While TBS is associated with fracture risk independently of areal bone mineral density (aBMD), normative TBS values have M. Iki (*) Department of Public Health, Kinki University Faculty of Medicine, 377-2 Oono-higashi, Osaka-Sayama, Osaka 589-8511, Japan e-mail: [email protected] J. Tamaki Department of Hygiene and Public Health, Osaka Medical College, 2-7 Daigaku-machi, Takatsuki, Osaka 569-8686, Japan Y. Sato Department of Human Life, Jin-ai University, 3-1-1 Otemachi, Echizen, Fukui 915-8586, Japan R. Winzenrieth Med-Imaps, Hôpital Xavier Arnozan, Avenue du Haut Lévèque, Pessac, 33600 Bordeaux, France S. Kagamimori University of Toyama, 2630 Sugitani, Toyama, Toyama 930-0194, Japan
only been reported for Caucasian women. This study provides age-specific normative values of TBS from a representative sample of Japanese women. Methods We randomly selected 4,550 women aged 15– 79 years from 7 areas throughout Japan. Women younger than 20 years and those with any medical history which might affect bone metabolism were excluded, and the remaining 3,069 with at least two assessable vertebrae from the first to the fourth vertebrae were subjected to analysis. TBS values were calculated from spine DXA images using TBS iNsight software (Med-Imaps, France). Age-related models of TBS were constructed using piecewise linear regression analysis. Results Participant age, body mass index (BMI), spine aBMD, and TBS (mean±SD) were 48.7±16.8 years, 22.9± 3.4, 0.888±0.169 g/cm2, and 1.187±0.137, respectively. A three-piece linear regression model of TBS on age explained 70.7 % of the total variance in TBS and comprised very small age-related changes in the youngest segment of the regression line, rapid loss in the middle segment, and small loss in the oldest segment. TBS was lower in Japanese women than in Caucasian women across all age ranges, with the difference increasing with age up through 65 years. Conclusions The normative values of TBS for Japanese women presented here would enhance our understanding of trabecular bone microarchitecture and help improve the management of osteoporosis. Keywords Bone microarchitecture . DXA . Japanese women . Normative values . Osteoporosis . Trabecular bone score
Y. Kagawa Kagawa Nutrition University, 3-9-21 Chiyoda, Sakado, Saitama 350-0288, Japan
Introduction
H. Yoneshima Shuwa General Hospital, 1200 Tanihara-shinden, Kasukabe, Saitama 344-0035, Japan
Trabecular bone score (TBS) is a texture parameter that quantifies local variation in the gray level distribution of
Osteoporos Int
dual-energy X-ray absorptiometry (DXA) images [1, 2]. TBS is not a direct physical measurement of bone microarchitecture but is significantly correlated with three-dimensional parameters of bone microarchitecture independently of areal bone mineral density (aBMD) in human cadavers [2]. It also correlates with compression stiffness of the human vertebra [3], whereby a higher TBS indicates stronger and more fracture-resistant microarchitecture. Several retrospective case-control studies [4–6] have found that vertebral fracture identification is significantly enhanced by the combination of TBS and aBMD, compared with aBMD alone, in an evaluation of the area under the receiver operating characteristic curve. A recent article from the Manitoba study, a cohort study of postmenopausal Canadian women, reported that lower TBS of the spine at baseline was associated with an increased risk of clinical spine, hip, and any osteoporotic fracture, independently of spine aBMD, during a mean follow-up period of 4.7 years [7]. Another cohort study in France showed that TBS and aBMD predicted incidence of fragility fractures equally well [8]. We recently observed that TBS increases the predictive validity for incident morphometry-confirmed vertebral fractures over 10 years relative to that achieved by aBMD and clinical risk factors in a Japanese female population [9]. Accumulating evidence thus suggests that TBS, when used in conjunction with aBMD and clinical risk factors, can improve fracture risk assessment and thus should be utilized and considered in osteoporosis screening and management practices. For TBS to be truly useful in clinical settings, physicians must be able to compare TBS values with reference data, as is routinely done in other types of clinical examinations, including aBMD testing. However, normative values of TBS have only been reported for Caucasian women in France [10] and the USA [11]. Furthermore, epidemiologic features of TBS such as its age-related change or correlation with body size are not known for Asian women, who have different physical and fracture risk profiles from those of Caucasian women [12, 13]. The Japanese Population-based Osteoporosis (JPOS) study is a representative study which targets Japanese women spanning a broad age range, focusing on bone health [14]. As such, it would be the most suitable study from which a normative database can be constructed. The present study aimed to accomplish the following: 1. Provide age-specific normative values of TBS for the Japanese female population, 2. Determine the relationship between TBS and basic characteristics including age and body size, and 3. Compare data from Japanese women to those from Caucasian women with regard to age-related changes in TBS.
Subjects and methods Subjects The source population comprised participants in the JPOS baseline study, which was previously described [14, 15]. Briefly, 50 women were randomly selected from each 5-year age group (age range, 15–79 years) in seven areas of Japan. The sample size was determined to ensure that the observed mean aBMD would remain within 2.5 % from the true value with a probability of 0.95 in each of the 5-year age groups. Since the standard deviation (SD) relative to the mean for TBS is smaller than that for aBMD, the sampling error for TBS would be smaller than that for aBMD in the present study. Of the 4,550 women selected, 3,985 (87.6 %) completed the baseline study. Of the 3,985 women, participants who were younger than 20 years; who had a history or present condition affecting bone metabolism including glucocorticoid administration, amenorrhea, oligomenorrhea, bilateral oophorectomy, parathyroid gland disease, hyperthyroidism, rheumatoid arthritis, gastrectomy due to gastric cancer, myasthenia gravis, or ossification of the posterior longitudinal ligament; who had severe deformity in the lumbar spine; who had extreme values of TBS; or whose body mass index (BMI) exceeded 40 were excluded. All participants provided written informed consent prior to participating in the baseline study. The study protocol was approved by the Ethics Committee of the Kinki University Faculty of Medicine. Interviews Detailed interviews were conducted by trained nurses using a structured questionnaire. Items included questions about menstrual history, current and past gynecological conditions, fracture history, and other conditions or medications that could affect bone metabolism. History of drug treatment for osteoporosis was recorded from interviews. Most treatments included activated vitamin D and/or calcium, as bisphosphonates and selective estrogen receptor modulators were not commercially available. If participants had a history of such drug therapy, they were excluded from the analyses. Bone mass measurement Measurements were obtained by certified radiological technologists using a single DXA scanner (QDR4500A, Hologic Inc., Bedford, MA, USA) installed in a mobile test room. Participants lay in a supine position in the middle of the imaging table of the densitometer parallel to the longitudinal table axis. The participants’ legs were raised by a support cushion for the lower legs such that the thighs were as vertical as possible. The entire lumbar spine was scanned in
Osteoporos Int
posteroanterior projection in the fast array mode. When the spine was markedly off-centered in the scan field, participants were repositioned and rescanned. Then, the right proximal femur was scanned with the leg straightened in a 30° adduction at the hip joint, fixed by using a specially designed fixture. If a participant had any deformity in the right hip, the other side was scanned. aBMD was calculated for the first to fourth vertebrae (LS) using densitometric software (Apex software version 2.3, Hologic Inc., Bedford, MA, USA). We excluded vertebrae with fractures or degenerative changes causing more than a 1 SD greater aBMD from the immediately adjacent vertebrae in accordance with the International Society for Clinical Densitometry rules for individual vertebrae exclusion [16]. Femoral neck (FN) and total hip aBMD values were also calculated with the same software. The in vivo precision of aBMD, represented by the coefficient of variation (CV), was 1.04 % for the spine, 1.10 % for the femoral neck, and 0.66 % for the total hip, all of which were calculated from five measurements in five postmenopausal female volunteers (age range, 53–61 years) [14]. A lumbar spine phantom was scanned daily before and after study measurements for quality control purposes (0.34 %, in vitro CV). No remarkable drift in aBMD values of the phantom was observed during the study period [15].
TBS calculations TBS details were previously published [7, 8]. Lumbar spine TBS was obtained using DXA images archived in the JPOS baseline study. TBS was calculated at the same regions of interest used for aBMD measurements using TBS iNsight software (Version 1.9.2, Med-Imaps, Bordeaux, France) by one of the authors (RW) who was blinded to the clinical data of participants. TBS values were calibrated to standard values using the TBS calibration phantom (17-cm thickness and 25 % fat mass equivalent) to eliminate inter-manufacturer difference between GE Lunar and Hologic in average. TBS values were adjusted for BMI to 24.6592, as was used for Caucasian women [10]. Short-term precision of TBS was calculated as 1.75 % (CV), with individual CVs ranging from 0.54 to 2.76 %, from the same set of DXA scans used to evaluate the precision of aBMD measurements [14].
Body size measurements Participant height (cm) and weight (kg) were measured with an automatic scale (TK-11868h, Takei Kagaku Co., Tokyo, Japan). BMI was calculated as weight divided by height squared (kg/m2).
Statistics All statistical analyses were conducted using the SAS System (Version 9.3). Prior to calculating normative values of TBS, we rejected extreme values of TBS from the analysis. This rejection process was conducted in participants younger than 40 years of age and those aged 40 years and older separately, as age-related changes in TBS seemed to differ across these delineations, according to a preliminary inspection of a scattergram of TBS against age. The best regression model for TBS by age for each group was selected from polynomial models using Akaike’s information criteria [17]. A linear model and a quadratic model were selected for the younger group and the older group, respectively. Next, we rejected a TBS value when the difference in the observed TBS value from the expected value by the regression model exceeded the 99 % tolerance limit, ensuring that this TBS value belonged to the same population. The piecewise linear regression model was developed to represent an age-related change in TBS. The number of segments of this model was set to three, in accordance with the scattergram of TBS against age and the value used in the US study [11]. We determined the least-squares estimates of a three-piece linear regression model by a convergence method introduced by Marquardt [18] by using the SAS procedure NLIN [19]. For comparisons of TBS between Japanese women and French Caucasian women [10], we used the second and third segments of the same regression model. TBS and aBMD values were expressed as mean±SD. Associations between TBS and demographic indices were evaluated by Pearson’s correlation coefficients.
Results Of the 3,985 participants, we excluded 260 women who were younger than 20 years, 520 who had a history or present condition affecting bone metabolism, 35 who were under osteoporosis treatment at baseline, 4 who had BMI values exceeding 40, 61 who showed deformity in the lumbar spine, and 36 through the extreme value rejection process. Of the remaining 3,069 women (mean age 48.7±16.8 years), data from 2,571 women with four assessable vertebrae, 449 women with three, and 49 women with two were included in the analyses. Their height, weight, and BMI are shown in 5-year age groupings in Table 1. Scatter plots of TBS against age are shown in Fig. 1 with the three-piece regression line. This regression model explained 70.7 % of the total variance in TBS. Age-specific mean values of TBS and aBMD are shown in Table 2. Peak TBS values were observed in the youngest group and appeared 15 years earlier than that for spine aBMD.
Osteoporos Int Table 1 Demographic characteristics of participants in the Japanese Population-based Osteoporosis (JPOS) baseline study Age (years)
Number
Height (cm)
Weight (kg)
BMI
Mean
Mean
Mean
SD
SD
SD
20–24
255
158.3
5.4
51.8
7.5
20.7
2.8
25–29 30–34 35–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 Total
284 267 277 283 290 236 252 253 255 222 195 3,069
158.3 157.5 156.9 154.9 154.1 152.0 151.0 148.7 147.9 147.0 144.6 153.0
5.1 5.2 4.8 5.5 5.2 4.7 5.4 5.1 5.1 5.3 5.4 6.8
52.1 53.4 54.9 54.6 55.8 54.9 54.4 53.6 52.3 52.4 49.1 53.4
7.1 8.2 8.1 7.9 8.6 7.9 8.1 8.2 7.8 8.6 7.6 8.2
20.8 21.5 22.3 22.7 23.5 23.7 23.9 24.2 23.9 24.2 23.4 22.9
2.6 3.1 3.1 3.0 3.5 3.2 3.3 3.4 3.2 3.5 3.3 3.4
BMI body mass index, SD standard deviation
A subgroup of young adults defined as premenopausal women aged 20–39 years included a total of 1,074 women; we used this subgroup as a young adult reference to calculate T-score from TBS data. Age-specific mean T-scores of TBS, LS-aBMD, and FN-aBMD are shown in Fig. 2. T-scores of TBS in older groups appeared to be lower than those of aBMD. The mean T-scores of TBS, LS-aBMD, and FNaBMD in the oldest group were −4.12, −2.92, and −2.15, respectively, and these values corresponded to 78.1, 68.9, and 71.3 % of the young adult mean values, respectively. According to this young adult reference group, a cutoff value of aBMD may be set for osteoporosis diagnosis according to the World Health Organization (WHO) criteria [20], i.e., more than 2.5 SD below the young adult mean, which was
Trabecular Bone Score
1.6
-0.0013/year
-0.0112/year
-0.0022/year
1.4
1.2
1.0
0.8 41.5
63.1
0.6 20
30
40
50
60
70
80
Age (years)
Fig. 1 Scatter plots of trabecular bone score (TBS) against age with a piecewise linear regression line from the Japanese Population-based Osteoporosis (JPOS) baseline study
calculated to be 0.728 g/cm2. The prevalence rate of osteoporosis according to this cutoff was 20.0 % of the entire study population. When we applied the same approach to TBS, the cutoff value of TBS corresponding to 2.5 SD below the mean was 1.148 and the prevalence rate of this TBS range was 37.1 %. This was 85 % higher than the prevalence rate of osteoporosis. We also determined a tentative TBS cutoff value, so that the prevalence rate of participants with TBS lower than this cutoff would be 20 %, i.e., the same as the osteoporosis prevalence rate according to aBMD. This value was calculated to be 1.049, which corresponded to a T-score of −3.7. This cutoff value was much lower than −2.5, and only 62.3 % of this low-TBS population had osteoporotic aBMD at LS. We evaluated the effects of vertebrae exclusion due to degenerative changes on TBS and aBMD. TBS and aBMD values were both significantly greater before exclusion compared with after, but the magnitude of the effects of exclusion was smaller in TBS than in aBMD (% difference before exclusion compared with after, 0.20 vs. 0.74 % for overall participants and 0.64 vs. 3.22 % in participants aged 75 to 79 years, which showed the greatest effect across all age groups). Correlation coefficients between TBS and demographic indices are provided in Table 3. TBS showed a significant inverse correlation with age and a significant positive correlation with aBMD. Age adjustment reduced the correlations between TBS and aBMD to 0.527 at the spine and 0.343 at the femoral neck for all participants, and both correlations remained statistically significant. In addition, these correlation coefficients were reduced in subgroups divided by menopausal status or age. To compare patterns of age-related changes in TBS in the present study with those reported for Caucasian women [10, 11], we superimposed the piecewise linear regression line on the TBS scattergram of Caucasian women in the USA [11], as shown in Fig. 3. Mean TBS in the present study was lower than that reported in the US study, regardless of age range groupings. The mean TBS was 5.9 % lower at age 35 years and 9.6 % lower at age 45 years in Japanese women, relative to that in Caucasian women. In Japanese women, TBS decreased by 16.2 % at age 63 and 19 % at age 80, relative to that at age 45. Corresponding decreases in Caucasian women were 6.1 and 13.3 %, respectively. The differences in TBS between Japanese and Caucasian populations increased with age in middle-aged women but decreased in older women. Given the correlation between TBS and height, we adjusted TBS values for height according to age-specific mean height reported in the US study [11]. As a result, TBS for Japanese women aged 50–59, 60–69, and 70–79 years increased by 6.3, 12.5, and 14.7 %, respectively. The differences in TBS between Caucasian and Japanese women decreased after this adjustment, but adjusted TBS for Japanese women in these
Osteoporos Int Table 2 Age-specific values of TBS and aBMD of participants in the Japanese Population-based Osteoporosis (JPOS) baseline study Age (years)
TBS
20–24 25–29 30–34 35–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 Young adults Total
LS-aBMD (g/cm2)
FN-aBMD (g/cm2)
Mean
SD
Q1
Median
Q3
Mean
SD
Mean
SD
1.313 1.311 1.299 1.300 1.268 1.232 1.164 1.104 1.056 1.044 1.024 1.019 1.306 1.187
0.069 0.071 0.069 0.069 0.071 0.079 0.081 0.073 0.086 0.073 0.075 0.079 0.070 0.137
1.263 1.266 1.248 1.260 1.226 1.182 1.107 1.057 0.998 1.004 0.972 0.970 1.260 1.077
1.310 1.315 1.300 1.302 1.273 1.234 1.165 1.100 1.057 1.041 1.023 1.019 1.308 1.206
1.353 1.355 1.351 1.349 1.317 1.284 1.223 1.153 1.111 1.100 1.080 1.075 1.352 1.298
0.966 0.986 0.999 1.015 0.998 0.982 0.876 0.816 0.764 0.731 0.712 0.686 0.992 0.888
0.108 0.099 0.099 0.111 0.110 0.123 0.129 0.124 0.124 0.115 0.131 0.130 0.106 0.167
0.826 0.798 0.779 0.805 0.801 0.796 0.743 0.709 0.665 0.638 0.612 0.573 0.802 0.736
0.121 0.103 0.095 0.103 0.100 0.111 0.096 0.097 0.084 0.093 0.091 0.082 0.107 0.127
Young adults comprised 1,073 premenopausal women aged 20–39 years TBS trabecular bone score, aBMD areal bone mineral density, LS lumbar spine (L1-4), FN femoral neck, SD standard deviation, Q1 lower quartile, Q3 upper quartile
age groups were still 10.2, 9.0, and 5.9 % lower, respectively, than TBS for US women in the corresponding age groups.
Discussion Accumulating evidence suggests that the addition of TBS increases the prediction accuracy of osteoporotic fractures
Age (years) 1.0
20 25 30 35 40 45 50 55 60 65 70 75
0.0
T-score
-1.0 -2.0 -3.0 TBS
-4.0
LS-aBMD FN-aBMD
-5.0 Fig. 2 Age-specific T-scores of trabecular bone score (TBS) and lumbar spine (LS) and femoral neck (FN) areal bone mineral density (aBMD) from the Japanese Population-based Osteoporosis (JPOS) baseline study
based on aBMD [4–9]. TBS is highly advantageous over other bone microarchitecture assessment methods such as high resolution computed tomography and magnetic resonance imaging; that is, clinicians can evaluate the status of trabecular bone microarchitecture as well as patient bone mass from one DXA scan without any additional X-ray exposure, time, or financial cost. In order for TBS to be used in clinical practice, a reference database representative of the population is desired. As such, the present study first provided agespecific normative TBS data for Asian women which will enhance physician understanding of bone microarchitecture status, thereby improving osteoporosis management in clinical settings. In terms of accuracy, the present TBS data were derived from randomly selected subjects from municipalities throughout Japan with a sufficiently high participation rate. Only one DXA machine was used to obtain all data, and the precision of the measurement was well controlled. A potential bias may have resulted from the exclusion of vertebrae due to degenerative changes. This exclusion process was developed in order to obtain assessable aBMD [16] but not TBS. The differences in T-scores before and after exclusion were significantly smaller in TBS than in aBMD. Previous studies reported that TBS was not affected by degenerative changes due to osteoarthritis in the spine whereas spine aBMD is affected substantially [10, 21]. Therefore, a bias due to this method of vertebrae exclusion would be small if it exists. To date, two reference databases for TBS have been reported: one for French Caucasian women and one for US non-
Osteoporos Int Table 3 Correlation coefficients between TBS and demographic variables or aBMD in participants in the Japanese Population-based Osteoporosis (JPOS) baseline study
TBS in all participants p value TBS in premenopausal women p value
Number
Age
Height
Weight
BMI
LS-aBMD
FN-aBMD
3,069
−0.813
ORIGINAL ARTICLE
Age-related normative values of trabecular bone score (TBS) for Japanese women: the Japanese Population-based Osteoporosis (JPOS) study M. Iki & J. Tamaki & Y. Sato & R. Winzenrieth & S. Kagamimori & Y. Kagawa & H. Yoneshima
Received: 12 May 2014 / Accepted: 12 August 2014 # International Osteoporosis Foundation and National Osteoporosis Foundation 2014
Abstract Summary Trabecular bone score (TBS), a surrogate measure of bone microarchitecture, represents fracture risk independently of bone density. We present normative TBS values from a representative population study of Japanese women. This database would enhance our understanding of trabecular bone microarchitecture and improve osteoporosis management. Introduction TBS is a texture parameter that quantifies local variation in gray level distribution within dual-energy X-ray absorptiometry (DXA) images of the lumbar spine. While TBS is associated with fracture risk independently of areal bone mineral density (aBMD), normative TBS values have M. Iki (*) Department of Public Health, Kinki University Faculty of Medicine, 377-2 Oono-higashi, Osaka-Sayama, Osaka 589-8511, Japan e-mail: [email protected] J. Tamaki Department of Hygiene and Public Health, Osaka Medical College, 2-7 Daigaku-machi, Takatsuki, Osaka 569-8686, Japan Y. Sato Department of Human Life, Jin-ai University, 3-1-1 Otemachi, Echizen, Fukui 915-8586, Japan R. Winzenrieth Med-Imaps, Hôpital Xavier Arnozan, Avenue du Haut Lévèque, Pessac, 33600 Bordeaux, France S. Kagamimori University of Toyama, 2630 Sugitani, Toyama, Toyama 930-0194, Japan
only been reported for Caucasian women. This study provides age-specific normative values of TBS from a representative sample of Japanese women. Methods We randomly selected 4,550 women aged 15– 79 years from 7 areas throughout Japan. Women younger than 20 years and those with any medical history which might affect bone metabolism were excluded, and the remaining 3,069 with at least two assessable vertebrae from the first to the fourth vertebrae were subjected to analysis. TBS values were calculated from spine DXA images using TBS iNsight software (Med-Imaps, France). Age-related models of TBS were constructed using piecewise linear regression analysis. Results Participant age, body mass index (BMI), spine aBMD, and TBS (mean±SD) were 48.7±16.8 years, 22.9± 3.4, 0.888±0.169 g/cm2, and 1.187±0.137, respectively. A three-piece linear regression model of TBS on age explained 70.7 % of the total variance in TBS and comprised very small age-related changes in the youngest segment of the regression line, rapid loss in the middle segment, and small loss in the oldest segment. TBS was lower in Japanese women than in Caucasian women across all age ranges, with the difference increasing with age up through 65 years. Conclusions The normative values of TBS for Japanese women presented here would enhance our understanding of trabecular bone microarchitecture and help improve the management of osteoporosis. Keywords Bone microarchitecture . DXA . Japanese women . Normative values . Osteoporosis . Trabecular bone score
Y. Kagawa Kagawa Nutrition University, 3-9-21 Chiyoda, Sakado, Saitama 350-0288, Japan
Introduction
H. Yoneshima Shuwa General Hospital, 1200 Tanihara-shinden, Kasukabe, Saitama 344-0035, Japan
Trabecular bone score (TBS) is a texture parameter that quantifies local variation in the gray level distribution of
Osteoporos Int
dual-energy X-ray absorptiometry (DXA) images [1, 2]. TBS is not a direct physical measurement of bone microarchitecture but is significantly correlated with three-dimensional parameters of bone microarchitecture independently of areal bone mineral density (aBMD) in human cadavers [2]. It also correlates with compression stiffness of the human vertebra [3], whereby a higher TBS indicates stronger and more fracture-resistant microarchitecture. Several retrospective case-control studies [4–6] have found that vertebral fracture identification is significantly enhanced by the combination of TBS and aBMD, compared with aBMD alone, in an evaluation of the area under the receiver operating characteristic curve. A recent article from the Manitoba study, a cohort study of postmenopausal Canadian women, reported that lower TBS of the spine at baseline was associated with an increased risk of clinical spine, hip, and any osteoporotic fracture, independently of spine aBMD, during a mean follow-up period of 4.7 years [7]. Another cohort study in France showed that TBS and aBMD predicted incidence of fragility fractures equally well [8]. We recently observed that TBS increases the predictive validity for incident morphometry-confirmed vertebral fractures over 10 years relative to that achieved by aBMD and clinical risk factors in a Japanese female population [9]. Accumulating evidence thus suggests that TBS, when used in conjunction with aBMD and clinical risk factors, can improve fracture risk assessment and thus should be utilized and considered in osteoporosis screening and management practices. For TBS to be truly useful in clinical settings, physicians must be able to compare TBS values with reference data, as is routinely done in other types of clinical examinations, including aBMD testing. However, normative values of TBS have only been reported for Caucasian women in France [10] and the USA [11]. Furthermore, epidemiologic features of TBS such as its age-related change or correlation with body size are not known for Asian women, who have different physical and fracture risk profiles from those of Caucasian women [12, 13]. The Japanese Population-based Osteoporosis (JPOS) study is a representative study which targets Japanese women spanning a broad age range, focusing on bone health [14]. As such, it would be the most suitable study from which a normative database can be constructed. The present study aimed to accomplish the following: 1. Provide age-specific normative values of TBS for the Japanese female population, 2. Determine the relationship between TBS and basic characteristics including age and body size, and 3. Compare data from Japanese women to those from Caucasian women with regard to age-related changes in TBS.
Subjects and methods Subjects The source population comprised participants in the JPOS baseline study, which was previously described [14, 15]. Briefly, 50 women were randomly selected from each 5-year age group (age range, 15–79 years) in seven areas of Japan. The sample size was determined to ensure that the observed mean aBMD would remain within 2.5 % from the true value with a probability of 0.95 in each of the 5-year age groups. Since the standard deviation (SD) relative to the mean for TBS is smaller than that for aBMD, the sampling error for TBS would be smaller than that for aBMD in the present study. Of the 4,550 women selected, 3,985 (87.6 %) completed the baseline study. Of the 3,985 women, participants who were younger than 20 years; who had a history or present condition affecting bone metabolism including glucocorticoid administration, amenorrhea, oligomenorrhea, bilateral oophorectomy, parathyroid gland disease, hyperthyroidism, rheumatoid arthritis, gastrectomy due to gastric cancer, myasthenia gravis, or ossification of the posterior longitudinal ligament; who had severe deformity in the lumbar spine; who had extreme values of TBS; or whose body mass index (BMI) exceeded 40 were excluded. All participants provided written informed consent prior to participating in the baseline study. The study protocol was approved by the Ethics Committee of the Kinki University Faculty of Medicine. Interviews Detailed interviews were conducted by trained nurses using a structured questionnaire. Items included questions about menstrual history, current and past gynecological conditions, fracture history, and other conditions or medications that could affect bone metabolism. History of drug treatment for osteoporosis was recorded from interviews. Most treatments included activated vitamin D and/or calcium, as bisphosphonates and selective estrogen receptor modulators were not commercially available. If participants had a history of such drug therapy, they were excluded from the analyses. Bone mass measurement Measurements were obtained by certified radiological technologists using a single DXA scanner (QDR4500A, Hologic Inc., Bedford, MA, USA) installed in a mobile test room. Participants lay in a supine position in the middle of the imaging table of the densitometer parallel to the longitudinal table axis. The participants’ legs were raised by a support cushion for the lower legs such that the thighs were as vertical as possible. The entire lumbar spine was scanned in
Osteoporos Int
posteroanterior projection in the fast array mode. When the spine was markedly off-centered in the scan field, participants were repositioned and rescanned. Then, the right proximal femur was scanned with the leg straightened in a 30° adduction at the hip joint, fixed by using a specially designed fixture. If a participant had any deformity in the right hip, the other side was scanned. aBMD was calculated for the first to fourth vertebrae (LS) using densitometric software (Apex software version 2.3, Hologic Inc., Bedford, MA, USA). We excluded vertebrae with fractures or degenerative changes causing more than a 1 SD greater aBMD from the immediately adjacent vertebrae in accordance with the International Society for Clinical Densitometry rules for individual vertebrae exclusion [16]. Femoral neck (FN) and total hip aBMD values were also calculated with the same software. The in vivo precision of aBMD, represented by the coefficient of variation (CV), was 1.04 % for the spine, 1.10 % for the femoral neck, and 0.66 % for the total hip, all of which were calculated from five measurements in five postmenopausal female volunteers (age range, 53–61 years) [14]. A lumbar spine phantom was scanned daily before and after study measurements for quality control purposes (0.34 %, in vitro CV). No remarkable drift in aBMD values of the phantom was observed during the study period [15].
TBS calculations TBS details were previously published [7, 8]. Lumbar spine TBS was obtained using DXA images archived in the JPOS baseline study. TBS was calculated at the same regions of interest used for aBMD measurements using TBS iNsight software (Version 1.9.2, Med-Imaps, Bordeaux, France) by one of the authors (RW) who was blinded to the clinical data of participants. TBS values were calibrated to standard values using the TBS calibration phantom (17-cm thickness and 25 % fat mass equivalent) to eliminate inter-manufacturer difference between GE Lunar and Hologic in average. TBS values were adjusted for BMI to 24.6592, as was used for Caucasian women [10]. Short-term precision of TBS was calculated as 1.75 % (CV), with individual CVs ranging from 0.54 to 2.76 %, from the same set of DXA scans used to evaluate the precision of aBMD measurements [14].
Body size measurements Participant height (cm) and weight (kg) were measured with an automatic scale (TK-11868h, Takei Kagaku Co., Tokyo, Japan). BMI was calculated as weight divided by height squared (kg/m2).
Statistics All statistical analyses were conducted using the SAS System (Version 9.3). Prior to calculating normative values of TBS, we rejected extreme values of TBS from the analysis. This rejection process was conducted in participants younger than 40 years of age and those aged 40 years and older separately, as age-related changes in TBS seemed to differ across these delineations, according to a preliminary inspection of a scattergram of TBS against age. The best regression model for TBS by age for each group was selected from polynomial models using Akaike’s information criteria [17]. A linear model and a quadratic model were selected for the younger group and the older group, respectively. Next, we rejected a TBS value when the difference in the observed TBS value from the expected value by the regression model exceeded the 99 % tolerance limit, ensuring that this TBS value belonged to the same population. The piecewise linear regression model was developed to represent an age-related change in TBS. The number of segments of this model was set to three, in accordance with the scattergram of TBS against age and the value used in the US study [11]. We determined the least-squares estimates of a three-piece linear regression model by a convergence method introduced by Marquardt [18] by using the SAS procedure NLIN [19]. For comparisons of TBS between Japanese women and French Caucasian women [10], we used the second and third segments of the same regression model. TBS and aBMD values were expressed as mean±SD. Associations between TBS and demographic indices were evaluated by Pearson’s correlation coefficients.
Results Of the 3,985 participants, we excluded 260 women who were younger than 20 years, 520 who had a history or present condition affecting bone metabolism, 35 who were under osteoporosis treatment at baseline, 4 who had BMI values exceeding 40, 61 who showed deformity in the lumbar spine, and 36 through the extreme value rejection process. Of the remaining 3,069 women (mean age 48.7±16.8 years), data from 2,571 women with four assessable vertebrae, 449 women with three, and 49 women with two were included in the analyses. Their height, weight, and BMI are shown in 5-year age groupings in Table 1. Scatter plots of TBS against age are shown in Fig. 1 with the three-piece regression line. This regression model explained 70.7 % of the total variance in TBS. Age-specific mean values of TBS and aBMD are shown in Table 2. Peak TBS values were observed in the youngest group and appeared 15 years earlier than that for spine aBMD.
Osteoporos Int Table 1 Demographic characteristics of participants in the Japanese Population-based Osteoporosis (JPOS) baseline study Age (years)
Number
Height (cm)
Weight (kg)
BMI
Mean
Mean
Mean
SD
SD
SD
20–24
255
158.3
5.4
51.8
7.5
20.7
2.8
25–29 30–34 35–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 Total
284 267 277 283 290 236 252 253 255 222 195 3,069
158.3 157.5 156.9 154.9 154.1 152.0 151.0 148.7 147.9 147.0 144.6 153.0
5.1 5.2 4.8 5.5 5.2 4.7 5.4 5.1 5.1 5.3 5.4 6.8
52.1 53.4 54.9 54.6 55.8 54.9 54.4 53.6 52.3 52.4 49.1 53.4
7.1 8.2 8.1 7.9 8.6 7.9 8.1 8.2 7.8 8.6 7.6 8.2
20.8 21.5 22.3 22.7 23.5 23.7 23.9 24.2 23.9 24.2 23.4 22.9
2.6 3.1 3.1 3.0 3.5 3.2 3.3 3.4 3.2 3.5 3.3 3.4
BMI body mass index, SD standard deviation
A subgroup of young adults defined as premenopausal women aged 20–39 years included a total of 1,074 women; we used this subgroup as a young adult reference to calculate T-score from TBS data. Age-specific mean T-scores of TBS, LS-aBMD, and FN-aBMD are shown in Fig. 2. T-scores of TBS in older groups appeared to be lower than those of aBMD. The mean T-scores of TBS, LS-aBMD, and FNaBMD in the oldest group were −4.12, −2.92, and −2.15, respectively, and these values corresponded to 78.1, 68.9, and 71.3 % of the young adult mean values, respectively. According to this young adult reference group, a cutoff value of aBMD may be set for osteoporosis diagnosis according to the World Health Organization (WHO) criteria [20], i.e., more than 2.5 SD below the young adult mean, which was
Trabecular Bone Score
1.6
-0.0013/year
-0.0112/year
-0.0022/year
1.4
1.2
1.0
0.8 41.5
63.1
0.6 20
30
40
50
60
70
80
Age (years)
Fig. 1 Scatter plots of trabecular bone score (TBS) against age with a piecewise linear regression line from the Japanese Population-based Osteoporosis (JPOS) baseline study
calculated to be 0.728 g/cm2. The prevalence rate of osteoporosis according to this cutoff was 20.0 % of the entire study population. When we applied the same approach to TBS, the cutoff value of TBS corresponding to 2.5 SD below the mean was 1.148 and the prevalence rate of this TBS range was 37.1 %. This was 85 % higher than the prevalence rate of osteoporosis. We also determined a tentative TBS cutoff value, so that the prevalence rate of participants with TBS lower than this cutoff would be 20 %, i.e., the same as the osteoporosis prevalence rate according to aBMD. This value was calculated to be 1.049, which corresponded to a T-score of −3.7. This cutoff value was much lower than −2.5, and only 62.3 % of this low-TBS population had osteoporotic aBMD at LS. We evaluated the effects of vertebrae exclusion due to degenerative changes on TBS and aBMD. TBS and aBMD values were both significantly greater before exclusion compared with after, but the magnitude of the effects of exclusion was smaller in TBS than in aBMD (% difference before exclusion compared with after, 0.20 vs. 0.74 % for overall participants and 0.64 vs. 3.22 % in participants aged 75 to 79 years, which showed the greatest effect across all age groups). Correlation coefficients between TBS and demographic indices are provided in Table 3. TBS showed a significant inverse correlation with age and a significant positive correlation with aBMD. Age adjustment reduced the correlations between TBS and aBMD to 0.527 at the spine and 0.343 at the femoral neck for all participants, and both correlations remained statistically significant. In addition, these correlation coefficients were reduced in subgroups divided by menopausal status or age. To compare patterns of age-related changes in TBS in the present study with those reported for Caucasian women [10, 11], we superimposed the piecewise linear regression line on the TBS scattergram of Caucasian women in the USA [11], as shown in Fig. 3. Mean TBS in the present study was lower than that reported in the US study, regardless of age range groupings. The mean TBS was 5.9 % lower at age 35 years and 9.6 % lower at age 45 years in Japanese women, relative to that in Caucasian women. In Japanese women, TBS decreased by 16.2 % at age 63 and 19 % at age 80, relative to that at age 45. Corresponding decreases in Caucasian women were 6.1 and 13.3 %, respectively. The differences in TBS between Japanese and Caucasian populations increased with age in middle-aged women but decreased in older women. Given the correlation between TBS and height, we adjusted TBS values for height according to age-specific mean height reported in the US study [11]. As a result, TBS for Japanese women aged 50–59, 60–69, and 70–79 years increased by 6.3, 12.5, and 14.7 %, respectively. The differences in TBS between Caucasian and Japanese women decreased after this adjustment, but adjusted TBS for Japanese women in these
Osteoporos Int Table 2 Age-specific values of TBS and aBMD of participants in the Japanese Population-based Osteoporosis (JPOS) baseline study Age (years)
TBS
20–24 25–29 30–34 35–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75–79 Young adults Total
LS-aBMD (g/cm2)
FN-aBMD (g/cm2)
Mean
SD
Q1
Median
Q3
Mean
SD
Mean
SD
1.313 1.311 1.299 1.300 1.268 1.232 1.164 1.104 1.056 1.044 1.024 1.019 1.306 1.187
0.069 0.071 0.069 0.069 0.071 0.079 0.081 0.073 0.086 0.073 0.075 0.079 0.070 0.137
1.263 1.266 1.248 1.260 1.226 1.182 1.107 1.057 0.998 1.004 0.972 0.970 1.260 1.077
1.310 1.315 1.300 1.302 1.273 1.234 1.165 1.100 1.057 1.041 1.023 1.019 1.308 1.206
1.353 1.355 1.351 1.349 1.317 1.284 1.223 1.153 1.111 1.100 1.080 1.075 1.352 1.298
0.966 0.986 0.999 1.015 0.998 0.982 0.876 0.816 0.764 0.731 0.712 0.686 0.992 0.888
0.108 0.099 0.099 0.111 0.110 0.123 0.129 0.124 0.124 0.115 0.131 0.130 0.106 0.167
0.826 0.798 0.779 0.805 0.801 0.796 0.743 0.709 0.665 0.638 0.612 0.573 0.802 0.736
0.121 0.103 0.095 0.103 0.100 0.111 0.096 0.097 0.084 0.093 0.091 0.082 0.107 0.127
Young adults comprised 1,073 premenopausal women aged 20–39 years TBS trabecular bone score, aBMD areal bone mineral density, LS lumbar spine (L1-4), FN femoral neck, SD standard deviation, Q1 lower quartile, Q3 upper quartile
age groups were still 10.2, 9.0, and 5.9 % lower, respectively, than TBS for US women in the corresponding age groups.
Discussion Accumulating evidence suggests that the addition of TBS increases the prediction accuracy of osteoporotic fractures
Age (years) 1.0
20 25 30 35 40 45 50 55 60 65 70 75
0.0
T-score
-1.0 -2.0 -3.0 TBS
-4.0
LS-aBMD FN-aBMD
-5.0 Fig. 2 Age-specific T-scores of trabecular bone score (TBS) and lumbar spine (LS) and femoral neck (FN) areal bone mineral density (aBMD) from the Japanese Population-based Osteoporosis (JPOS) baseline study
based on aBMD [4–9]. TBS is highly advantageous over other bone microarchitecture assessment methods such as high resolution computed tomography and magnetic resonance imaging; that is, clinicians can evaluate the status of trabecular bone microarchitecture as well as patient bone mass from one DXA scan without any additional X-ray exposure, time, or financial cost. In order for TBS to be used in clinical practice, a reference database representative of the population is desired. As such, the present study first provided agespecific normative TBS data for Asian women which will enhance physician understanding of bone microarchitecture status, thereby improving osteoporosis management in clinical settings. In terms of accuracy, the present TBS data were derived from randomly selected subjects from municipalities throughout Japan with a sufficiently high participation rate. Only one DXA machine was used to obtain all data, and the precision of the measurement was well controlled. A potential bias may have resulted from the exclusion of vertebrae due to degenerative changes. This exclusion process was developed in order to obtain assessable aBMD [16] but not TBS. The differences in T-scores before and after exclusion were significantly smaller in TBS than in aBMD. Previous studies reported that TBS was not affected by degenerative changes due to osteoarthritis in the spine whereas spine aBMD is affected substantially [10, 21]. Therefore, a bias due to this method of vertebrae exclusion would be small if it exists. To date, two reference databases for TBS have been reported: one for French Caucasian women and one for US non-
Osteoporos Int Table 3 Correlation coefficients between TBS and demographic variables or aBMD in participants in the Japanese Population-based Osteoporosis (JPOS) baseline study
TBS in all participants p value TBS in premenopausal women p value
Number
Age
Height
Weight
BMI
LS-aBMD
FN-aBMD
3,069
−0.813
