- Email: [email protected]
Densitometer-Specific Differences in the Correlation Between Body Mass Index and Lumbar Spine Trabecular Bone Score
ARTICLE IN PRESS Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health, vol. ■, no. ■, 1–6, 2016 © Copyright 2016 by The International Society for Clinical Densitometry 1094-6950/■:1–6/$36.00 http://dx.doi.org/10.1016/j.jocd.2016.11.003
Densitometer-Specific Differences in the Correlation Between Body Mass Index and Lumbar Spine Trabecular Bone Score Gillian Mazzetti,1 Claudie Berger,2 William D. Leslie,3 Didier Hans,4 Lisa Langsetmo,5 David A. Hanley,6 Christopher S. Kovacs,7 Jerrilyn C. Prior,8 Stephanie M. Kaiser,9 K. Shawn Davison,10 Robert Josse,11 Alexandra Papaioannou,12 Jonathan R. Adachi,12 David Goltzman,1,2 Suzanne N. Morin*,1 for The CaMos Research Group 1
Department of Medicine, McGill University, Montréal, Québec, Canada; 2CaMos National Coordinating Centre, McGill University, Montréal, Québec, Canada; 3Department of Medicine, University of Manitoba, Winnipeg, Manitoba, Canada; 4 Centre of Bone Diseases, Lausanne University Hospital, Lausanne, Switzerland; 5Division of Epidemiology and Community Health, University of Minnesota, Minneapolis, MN, USA; 6Department of Medicine, University of Calgary, Calgary, Alberta, Canada; 7Faculty of Medicine, Memorial University, St. John’s, Newfoundland, Canada; 8Department of Medicine, University of British Columbia, Vancouver, BC, Canada; 9Department of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada; 10Saskatoon Osteoporosis and CaMos Centre, Saskatoon, Saskatchewan, Canada; 11 Department of Medicine, University of Toronto, Toronto, Ontario, Canada; and 12Department of Medicine, McMaster University, Hamilton, Ontario, Canada
Abstract Trabecular bone score (TBS) is a gray-level texture measure derived from lumbar spine dual-energy X-ray absorptiometry (DXA) images that predicts fractures independent of bone mineral density (BMD). Increased abdominal soft tissue in individuals with elevated body mass index (BMI) absorbs more X-rays during image acquisition for BMD measurement and must be accommodated by the TBS algorithm. We aimed to determine if the relationship between BMI and TBS varied between 2 major manufacturers’ densitometers, because different densitometers accommodate soft tissues differently.We identified 1919 women and 811 men, participants of the Canadian Multicentre Osteoporosis Study, aged ≥40 yr with lumbar spine DXA scans acquired on GE Lunar (4 centers) or Hologic (3 centers) densitometers at year 10 of follow-up. TBS was calculated for L1–L4 (TBS iNsight® software, version 2.1). A significant negative correlation between TBS and BMI was observed
Received 09/6/16; Revised 11/8/16; Accepted 11/9/16. Contributions: All authors substantially contributed to the conception (GM, CB, WDL, and SNM), design (GM, CB, WDL, and SNM), analysis (GM, CB, and SNM), and interpretation of data (all); drafting (GM, CB, WDL, and SNM) or revision of the article critically for important intellectual content (all); and final approval of the version to be published (all). SNM accepts full responsibility for the work and the conduct of the study, had full access to all the data, and controlled the decision to publish. Disclosures: D Hans: co-owner of the TBS patent and has corresponding ownership share and position at Medimaps Group. DA Hanley: research grants from Amgen and Eli Lilly Canada; speaking honoraria and advisory board for Amgen Canada. CS Kovacs: honoraria from Amgen and advisory board for Eli Lilly Canada. SM Kaiser: honoraria, advisory board member, and speaker for Amgen. JD Adachi: honoraria from Actavis, Amgen, Eli Lilly Canada, and Merck; advisory board membership: Actavis, Amgen, Eli Lilly Canada, and Merck; participation in studies: Amgen, Eli Lilly Canada, and Merck. D Goltzman: research grants paid to institution: Amgen and Lilly; honoraria: Amgen advisory board membership. SN Morin: research grants paid to institution: Amgen and Merck; honoraria: Amgen Advisory Board membership. *Address correspondence to: Suzanne N. Morin, MD MSc, McGill University Health Centre- Montreal General Hospital, 1650 Cedar Ave, Room B2-118, Montreal, Québec H3G 1A4, Canada. E-mail: [email protected]
1
ARTICLE IN PRESS 2
Mazzetti et al. when TBS measurements were performed on Hologic densitometers in men (Pearson r = −0.36, p < 0.0001) and in women (Pearson r = −0.33, p < 0.0001); significant correlations were not seen when TBS was measured on GE Lunar densitometers (Pearson r = 0.00 in men, Pearson r = −0.02 in women).Age-adjusted linear regression models confirmed significant interactions between BMI and densitometer manufacturer for both men and women (p < 0.0001). In contrast, comparable positive correlations were observed between BMD and BMI on both Hologic and GE Lunar densitometers in men and women. In conclusion, BMI significantly affects TBS values in men and women when measured on Hologic but not GE Lunar densitometers.This finding has implications for clinical and research applications of TBS, especially when TBS is measured sequentially on DXA densitometers from different manufacturers or when results from different machines are pooled for analysis. Key Words: Body mass index; bone mineral density; cohort study; densitometer; trabecular bone score.
Introduction Trabecular bone score (TBS) is a gray-level texture measure derived from lumbar spine dual-energy X-ray absorptiometry (DXA) images (1,2). It measures the rate of local variation in gray-scale levels within lumbar spine DXA images such that a high TBS value reflects a bone with many gray-level variations of small amplitude and potentially stronger bone, whereas a low TBS value suggests fewer gray-level variations reflecting lower trabecular mass and connectivity (2–4). TBS has been shown to be lower in individuals with prevalent fragility fractures than in those without fractures and to predict fracture risk independent of FRAX clinical risk factors and of bone mineral density (BMD) (2,5,6).A recent meta-analysis demonstrated that the predictive value of the FRAX score for hip and major osteoporotic fractures is improved with the addition of TBS (7). Body mass index (BMI) has been shown to be a reliable surrogate for waist circumference, for global adiposity, and also for visceral adiposity in both men and women (8,9). Increased amounts of abdominal soft tissue, such as central adiposity in individuals with elevated BMI, absorb more X-rays during DXA acquisition and will result in lower “raw” TBS values (6). To compensate for this effect that are in part BMI based, adjustments have been incorporated in the TBS algorithm (version 2.1) (10), and TBS measurement has been optimized for individuals with a BMI in the range of 15–37 kg/m2 (6). We hypothesized that bone densitometers from different manufacturers may not provide comparable TBS measurements, particularly in individuals with greater abdominal adiposity, because DXA images may be degraded differentially across densitometers because of intrinsic differences in how soft tissue effects are accommodated during DXA scan acquisition and processing. Using data from the Canadian Multicentre Osteoporosis Study (CaMos), we aimed to determine if the relationship between BMI and TBS varied across the 2 major densitometer manufacturers, namely, GE and Hologic.
Methods Participants and Study Design CaMos is an ongoing Canada wide population-based prospective cohort study that enrolled 9423 community-
dwelling participants (6539 women and 2884 men) aged 25 yr and older in 1995–1997. Subjects were recruited from a list of randomly selected residential phone numbers within a 50-km radius of centers in 9 cities (CaMos research centers) across the country. The participants were randomly selected from eligible household members using an age- and sex-stratified design, which oversampled women and older groups to ensure sufficient sample size among those at the highest risk of fracture. The sample frame consisted of 40% of the Canadian population (11). Data collection at baseline and every 5 yr included intervieweradministered questionnaires, clinical assessments, and BMD measurements by DXA (total hip, femoral neck, and lumbar spine, L1–L4) for all centers. Ethics approval was granted through McGill University and the research ethics board for each participating center. Signed informed consent was obtained from all study participants in accordance with the Helsinki Declaration. The current study represents a cross-sectional analysis of 1919 women and 811 men aged ≥40 yr who had TBS calculated from lumbar spine DXA images for BMD measurements collected at year 10 of follow-up (2005–2007). Seven centers (Vancouver, Calgary, Saskatoon, Toronto, Quebec City, Halifax, and St. John’s) were part of the analysis, because BMD images obtained from DXA machines in 2 centers (Hamilton and Kingston) at year 10 were not compatible with the TBS algorithm.
Measurements Height was measured using either a ruler on the wall, a measure on a weight scale, or a wall-mounted stadiometer, and weight was measured in light clothing using a beam balance or electronic scale (12). Standard protocols are in place at each center for the measurement of height (in centimeter) and weight (in kilogram), performed at the time of DXA scanning; however, there were no cross-calibration across sites. BMI was calculated using weight in kilogram divided by height in square meter. Four centers (Vancouver,Toronto, Halifax, and St. John’s) used GE Lunar densitometers (Prodigy; GE Healthcare, Madison, WI) and 3 centers (Calgary, Saskatoon, and Quebec City) used Hologic densitometers (QDR 4500, Discovery, or Delphi; Hologic Inc., Bedford, MA). BMD measurements from GE Lunar densitometers were converted to equivalent Hologic values using published methods (13).
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume ■, 2016
ARTICLE IN PRESS Differences in the Correlation Between BMI and Lumbar Spine TBS Details on BMD cross-calibration and quality control have been previously published (14). TBS calculations for L1– L4 were blindly calculated in the Bone Disease Centre at the University of Lausanne in Switzerland using TBS iNsight® (version 2.1; Medimaps, Merignac, France). TBS was evaluated from the same region of interest used to calculate the lumbar spine BMD. Statistical cross-calibration was conducted based on the determination of constants (firstly for intra-DXA device calibration [e.g., Prodigy vs Prodigy] and secondly for inter-DXA device calibration [e.g., GE Lunar vs Hologic]), which were then applied to the TBS scores to compensate for differences between centers and device types.
Statistical Analysis All analyses were stratified by sex. Continuous variables were reported as means with standard deviations and counts were reported with percentages. t-Tests (continuous variables) and chi-squared tests (categorical variables) were used to compare baseline characteristics in women and men. Bivariate Pearson correlation coefficients of BMI vs TBS and BMD for GE Lunar and Hologic densitometers were calculated to determine the strengths of the relationships between these variables. Multiple linear regression analyses adjusted for age (as a continuous variable) with TBS as the dependent variable were performed to test for an interaction between BMI (as a continuous variable in the first model or categorized in tertiles in the second model) and densitometer manufacturer. The residuals of both sets of models were normally distributed as per visual inspection of the plotted residuals. A p value <0.05 was considered to be statistically significant. Statistical analyses were performed using SAS version 9.4 (Cary, NC).
Results Characteristics of the study population are described in Table 1. Compared with women, men were significantly younger and had a significantly higher BMI, lumbar spine BMD, and TBS. BMI was higher than 37 kg/m2 in 0.3% of men (n = 2) and 1.0% of women (n = 19). We observed a significant negative correlation between TBS and BMI in men (Pearson r = −0.36, p < 0.0001) and in women (Pearson r = −0.33, p < 0.0001) when TBS measurements were derived from Hologic densitometer images (Table 2). No significant correlations were observed between TBS and BMI when TBS measurements were derived from GE Lunar densitometers (Pearson r = 0.00 in men, Pearson r = −0.02 in women). In contrast, significant positive correlations were observed between BMD and BMI that were similar for the GE Lunar and Hologic densitometers in both men (Pearson r = 0.26 for Hologic, Pearson r = 0.24 for GE Lunar; p < 0.0001) and women (Pearson r = 0.26 for Hologic, Pearson r = 0.25 for GE Lunar; p < 0.0001). Scatter plots of TBS vs BMI stratified by sex and densitometer manu-
3
Table 1 Characteristics of the Study Population by Sex Characteristics DXA manufacturer, n (%) GE Lunar Hologic Age (yr)* Weight (kg)* Height (cm)* BMI (kg/m2)* L1–L4 BMD (g/cm2)* TBS*
Women (N = 1919)
Men (N = 811)
1092 (57) 827 (43) 69.6 ± 9.8 67.2 ± 11.8 159.3 ± 6.3 26.5 ± 4.3 0.931 ± 0.144 1.273 ± 0.108
480 (59) 331 (41) 66.3 ± 11.4 82.5 ± 12.6 173.9 ± 7.0 27.2 ± 3.6 1.069 ± 0.173 1.297 ± 0.107
Notes: Data are presented as means and standard deviations or counts and percentage. Abbr: BMD, bone mineral density; BMI, body mass index; DXA, dual-energy X-ray absorptiometry; TBS, trabecular bone score (unitless). *p < 0.05.
facturer along with age-adjusted linear regressions are shown in Fig. 1. In age-adjusted linear regression, with TBS as the dependent variable, there was a significant interaction between BMI and machine manufacturer (p < 0.0001). This interaction consisted of a negative association between TBS and BMI when TBS was measured on Hologic densitometers (with GE Lunar densitometers as the reference). Specifically, the participants’ TBS scores measured on Hologic densitometers were lower by 0.011 (95% confidence interval: 0.007–0.015) for each increase in 1 kg/m2 of BMI in men and by 0.006 (95% confidence interval: 0.004–0.008) in women compared with those measured on GE Lunar machines. Similar results were obtained between TBS and BMI on Hologic or GE Lunar densitometers when BMI was categorized in tertiles (data not shown).
Discussion We report the presence of a negative correlation between TBS and BMI in both men and women when TBS is measured on Hologic densitometers when using version 2.1 of the TBS algorithm. This TBS-BMI relationship was not observed when TBS is measured on GE Lunar densitometers. This is the first report on manufacturer-specific differences in the BMI-TBS relationship. In contrast, we also observed the well-recognized positive correlation between lumbar spine BMD and BMI, which was similar across the 2 manufacturers of bone densitometers (15–17). Negative correlations between TBS and BMI have previously been reported. In American adults (N = 7682, 20 yr and older) surveyed in 2 consecutive cycles of the National Health and Nutrition Examination Survey (NHANES) (2005–2008), lumbar spine BMD, measured on
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume ■, 2016
ARTICLE IN PRESS 4
Mazzetti et al. Table 2 Pearson Correlation Coefficients of BMI vs Lumbar Spine TBS and BMD by DXA Manufacturer Hologica
Lumbar spine TBS Lumbar spine BMD
GE Lunar b
BMI, Men N = 331
BMI, Women N = 827
BMI, Men N = 480
BMI, Women N = 1092
−0.36* 0.26*
−0.33* 0.25*
0.00 0.24*
−0.02 0.25*
Abbr: BMD, bone mineral density; BMI, body mass index; DXA, dual-energy X-ray absorptiometry; TBS, trabecular bone score. a QDR 4500, Discovery, or Delphi. b Prodigy. *p < 0.0001.
Hologic QDR 4500A densitometers, was found to be positively correlated in both men and women with all body size variables measured including BMI, waist circumference, total body fat mass, trunk lean mass, and trunk fat mass, whereas TBS (algorithm version 2.1) was negatively correlated with these variables (18). Mean TBS values among those with BMI values of over 37 kg/m2 were approximately 26% lower in men and 15% lower in women than in those with BMI
values between 15 and 37 kg/m2 (p < 0.001 for both sexes). The negative correlations between TBS and BMI seen in the NHANES cohort (women Pearson r = −0.33 and men Pearson r = −0.53) are close to those documented in the current CaMos study for individuals whose TBS was measured on Hologic densitometers (Pearson r = −0.33 in women and Pearson r = −0.36 in men). Using data from the Osteoporotic Fractures in Men study (MrOs), Langsetmo
Fig. 1. Scatter plots and age-adjusted linear regression lines of TBS by BMI stratified by sex and manufacturer. BMI, body mass index; TBS, trabecular bone score. Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume ■, 2016
ARTICLE IN PRESS Differences in the Correlation Between BMI and Lumbar Spine TBS et al noted that TBS (measured on Hologic QDR 4500 densitometers, TBS algorithm version 2.1) varied with BMI, trunk lean mass, and trunk fat mass such that men in the highest quintile of body size or body composition had the lowest median TBS. Increasing BMI quintile was associated with lower TBS and, in contrast, with higher volumetric BMD (19). It is noteworthy that an earlier version of the TBS algorithm (version 1.8), optimized for women of average body size, provided lower TBS scores in men and individuals with elevated BMI. The current TBS algorithm (version 2.1) is the result of modifications made to address technical limitations recognized to be associated with degraded DXA image texture in the presence of greater abdominal adiposity. Indeed, using the Manitoba (Canada) clinical BMD registry (47,736 women and 4348 men; GE Lunar densitometers), Leslie et al compared correlations between TBS and BMI using the 2 versions of the TBS algorithms. Leslie et al reported negative correlations between TBS and BMI (Pearson r = −0.18 in women, Pearson r = −0.40 in men) using the early version of the TBS algorithm (version 1.8). When the updated version of the TBS algorithm (version 2.1) was applied in the same population, the correlations between TBS and BMI changed to Pearson r = −0.01 in women and Pearson r = 0.01 in men (10,20). From the above-reported data, it appears that the updated TBS algorithm (version 2.1) is less affected by the effects of abdominal soft tissue and BMI when applied on GE Lunar densitometers than when used on Hologic densitometers. Our findings have clinical implications given that TBS is incorporated into the FRAX score calculation for fracture prediction (7). Although the predictive value of the FRAX score for osteoporotic fractures is improved with the addition of TBS, our study demonstrates that TBS measured on a Hologic densitometer will be lower in patients with a higher BMI compared with those measured on a GE Lunar machine. This finding may lead to overestimation of a given individual’s risk of fracture when TBS is incorporated in the risk calculation and subsequent overtreatment of patients. Further confusion could arise if individuals have serial TBS measurements on machines from different manufacturers which might be affected by the manufacturer-specific relationships between BMI and TBS, or if an individual has a significant change in BMI and TBS is measured serially on a Hologic densitometer. Finally, implications for multicenter research studies that include measurements of TBS on different densitometers should also be recognized. The strengths of our study include its large sample size, the inclusion of both men and women, and a diversity of DXA instruments. In addition, all TBS measurements were calculated at the same time using the most recent software TBS algorithm available (version 2.1), allowing for a direct comparison between the TBS values. Our study is population based, allowing for the results to be generalizable to the Canadian population, although the effect of different race and ethnicity cannot be examined because
5
the majority of this cohort is white. Other limitations must be mentioned. Firstly, we did not have TBS measurements in the same participants measured on both densitometers, making our inference about the relationship between BMI and TBS and densitometer manufacturer an indirect one. Secondly, we did not take into consideration comorbidities or anti-osteoporosis medication use that might have affected the relationships between TBS and BMI independent of densitometer manufacturer. Finally, our findings only apply to the algorithm (version 2.1) that was used for TBS measurement in this cohort. In conclusion, this descriptive study demonstrates that a manufacturer-specific negative correlation exists between BMI and TBS when TBS is measured on Hologic densitometers as compared to GE Lunar densitometers. This finding has implications for clinical and research applications of TBS, especially when TBS is measured sequentially on DXA densitometers from different manufacturers or when results from different machines are pooled for analysis.
Acknowledgments The Canadian Multicentre Osteoporosis Study is currently funded by the Canadian Institutes of Health Research (CIHR, MOP 111103), Amgen, and Eli Lilly Canada. These funding sources had no role in the conception of this analysis, statistical methods, or interpretation of the data. SN Morin is a research scholar of the Fonds de Recherche du Québec en Santé.
References 1. Silva BC, Leslie WD, Resch H, et al. 2014 Trabecular bone score: a noninvasive analytical method based upon the DXA image. J Bone Miner Res 29:518–530. 2. Binkley N, Leslie WD. 2016 Clinical application of spine trabecular bone score (TBS). Clin Rev Bone Miner Metab 14:14–25. 3. Silva BC, Bilezikian JP. 2014 Trabecular bone score: perspectives of an imaging technology coming of age. Arq Bras Endocrinol Metabol 58:493–503. 4. Pothuaud L, Carcellar P, Hans D. 2008 Correlations between grey-level variations in 2D projection images (TBS) and 3D microarchitecture: applications in the study of human trabecular bone microarchitecture. Bone 42:775–787. 5. McCloskey EV, Odén A, Harvey NC, et al. 2015 Adjusting fracture probability by trabecular bone score. Calcif Tissue Int 96:500–509. 6. Silva BC, Broy SB, Boutroy S, et al. 2015 Fracture risk prediction by non-BMD DXA measures: the 2015 ISCD official positions part 2: trabecular bone score. J Clin Densitom 18:309–330. 7. McCloskey EV, Odén A, Harvey NC, et al. 2016 A metaanalysis of trabecular bone score in fracture risk prediction and its relationship to FRAX. J Bone Miner Res 31:940– 948. 8. Borruel S, Moltó JF, Alpañés M, et al. 2014 Surrogate markers of visceral adiposity in young adults: waist circumference and body mass index are more accurate than waist hip ratio, model
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume ■, 2016
ARTICLE IN PRESS 6
9.
10.
11. 12.
13.
14.
Mazzetti et al. of adipose distribution and visceral adiposity index. PLoS ONE 5:e114112. Sunmer AE, Micklesfield LK, Ricks M, et al. 2011 Waist circumference, BMI, and visceral adipose tissue in white women and women of African descent. Obesity (Silver Spring) 19:671– 674. Leslie WD, Winzenrieth R, Majumdar SR, et al. 2014 Clinical performance of an updated version of trabecular bone score (TBS) in men and women: the Manitoba BMD Cohort. J Bone Miner Res 29(Suppl 1). Kreiger N, Tenenhouse A, Joseph L, et al. 1999 The Canadian Multicentre Osteoporosis Study (CaMos): background, rationale, methods. Can J Aging 18:376–387. Hopman WM, Leroux C, Berger C, et al. 2007 Changes in body mass index in Canadians over a five-year period: results of a prospective, population-based study. BMC Public Health 7:150. Genant HK, Grampp S, Glüer CC, et al. 1994 Universal standardization for dual X-ray absorptiometry: patient and phantom cross-calibration results. J Bone Miner Res 9:1503– 1514. Berger C, Goltzman D, Langsetmo L, et al. 2010 Peak bone mass from longitudinal data: implications for the preva-
15.
16. 17. 18.
19.
20.
lence, pathophysiology and diagnosis of osteoporosis. J Bone Miner Res 25:1948–1957. Felson DT, Zhang Y, Hannan MT, Anderson JJ. 1993 Effects of weight and body mass index on bone mineral density in men and women: the Framingham study. J Bone Miner Res 8:567–573. Reid IR. 2002 Relationships among body mass, its components and bone. Bone 31:547–555. Morin SN, Leslie WD. 2009 High bone mineral density is associated with high body mass index. Osteoporos Int 20:1267– 1271. Looker AC, Sarafrazi Isfahani N, Fan B, Shepherd JA. 2016 Trabecular bone scores and lumbar spine bone mineral density of US adults: comparison of relationships with demographic and body size variables. Osteoporos Int 27:2467– 2475. Langsetmo L, Vo TN, Ensrud KE, et al. 2016 The association between trabecular bone score and lumbar spine volumetric BMD is attenuated among older men with high body mass index. J Bone Miner Res 10:1820–1826. Leslie WD, Krieg MA, Hans D. 2013 Clinical factors associated with trabecular bone score. J Clin Densitom 16:374– 379.
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume ■, 2016
Densitometer-Specific Differences in the Correlation Between Body Mass Index and Lumbar Spine Trabecular Bone Score Gillian Mazzetti,1 Claudie Berger,2 William D. Leslie,3 Didier Hans,4 Lisa Langsetmo,5 David A. Hanley,6 Christopher S. Kovacs,7 Jerrilyn C. Prior,8 Stephanie M. Kaiser,9 K. Shawn Davison,10 Robert Josse,11 Alexandra Papaioannou,12 Jonathan R. Adachi,12 David Goltzman,1,2 Suzanne N. Morin*,1 for The CaMos Research Group 1
Department of Medicine, McGill University, Montréal, Québec, Canada; 2CaMos National Coordinating Centre, McGill University, Montréal, Québec, Canada; 3Department of Medicine, University of Manitoba, Winnipeg, Manitoba, Canada; 4 Centre of Bone Diseases, Lausanne University Hospital, Lausanne, Switzerland; 5Division of Epidemiology and Community Health, University of Minnesota, Minneapolis, MN, USA; 6Department of Medicine, University of Calgary, Calgary, Alberta, Canada; 7Faculty of Medicine, Memorial University, St. John’s, Newfoundland, Canada; 8Department of Medicine, University of British Columbia, Vancouver, BC, Canada; 9Department of Medicine, Dalhousie University, Halifax, Nova Scotia, Canada; 10Saskatoon Osteoporosis and CaMos Centre, Saskatoon, Saskatchewan, Canada; 11 Department of Medicine, University of Toronto, Toronto, Ontario, Canada; and 12Department of Medicine, McMaster University, Hamilton, Ontario, Canada
Abstract Trabecular bone score (TBS) is a gray-level texture measure derived from lumbar spine dual-energy X-ray absorptiometry (DXA) images that predicts fractures independent of bone mineral density (BMD). Increased abdominal soft tissue in individuals with elevated body mass index (BMI) absorbs more X-rays during image acquisition for BMD measurement and must be accommodated by the TBS algorithm. We aimed to determine if the relationship between BMI and TBS varied between 2 major manufacturers’ densitometers, because different densitometers accommodate soft tissues differently.We identified 1919 women and 811 men, participants of the Canadian Multicentre Osteoporosis Study, aged ≥40 yr with lumbar spine DXA scans acquired on GE Lunar (4 centers) or Hologic (3 centers) densitometers at year 10 of follow-up. TBS was calculated for L1–L4 (TBS iNsight® software, version 2.1). A significant negative correlation between TBS and BMI was observed
Received 09/6/16; Revised 11/8/16; Accepted 11/9/16. Contributions: All authors substantially contributed to the conception (GM, CB, WDL, and SNM), design (GM, CB, WDL, and SNM), analysis (GM, CB, and SNM), and interpretation of data (all); drafting (GM, CB, WDL, and SNM) or revision of the article critically for important intellectual content (all); and final approval of the version to be published (all). SNM accepts full responsibility for the work and the conduct of the study, had full access to all the data, and controlled the decision to publish. Disclosures: D Hans: co-owner of the TBS patent and has corresponding ownership share and position at Medimaps Group. DA Hanley: research grants from Amgen and Eli Lilly Canada; speaking honoraria and advisory board for Amgen Canada. CS Kovacs: honoraria from Amgen and advisory board for Eli Lilly Canada. SM Kaiser: honoraria, advisory board member, and speaker for Amgen. JD Adachi: honoraria from Actavis, Amgen, Eli Lilly Canada, and Merck; advisory board membership: Actavis, Amgen, Eli Lilly Canada, and Merck; participation in studies: Amgen, Eli Lilly Canada, and Merck. D Goltzman: research grants paid to institution: Amgen and Lilly; honoraria: Amgen advisory board membership. SN Morin: research grants paid to institution: Amgen and Merck; honoraria: Amgen Advisory Board membership. *Address correspondence to: Suzanne N. Morin, MD MSc, McGill University Health Centre- Montreal General Hospital, 1650 Cedar Ave, Room B2-118, Montreal, Québec H3G 1A4, Canada. E-mail: [email protected]
1
ARTICLE IN PRESS 2
Mazzetti et al. when TBS measurements were performed on Hologic densitometers in men (Pearson r = −0.36, p < 0.0001) and in women (Pearson r = −0.33, p < 0.0001); significant correlations were not seen when TBS was measured on GE Lunar densitometers (Pearson r = 0.00 in men, Pearson r = −0.02 in women).Age-adjusted linear regression models confirmed significant interactions between BMI and densitometer manufacturer for both men and women (p < 0.0001). In contrast, comparable positive correlations were observed between BMD and BMI on both Hologic and GE Lunar densitometers in men and women. In conclusion, BMI significantly affects TBS values in men and women when measured on Hologic but not GE Lunar densitometers.This finding has implications for clinical and research applications of TBS, especially when TBS is measured sequentially on DXA densitometers from different manufacturers or when results from different machines are pooled for analysis. Key Words: Body mass index; bone mineral density; cohort study; densitometer; trabecular bone score.
Introduction Trabecular bone score (TBS) is a gray-level texture measure derived from lumbar spine dual-energy X-ray absorptiometry (DXA) images (1,2). It measures the rate of local variation in gray-scale levels within lumbar spine DXA images such that a high TBS value reflects a bone with many gray-level variations of small amplitude and potentially stronger bone, whereas a low TBS value suggests fewer gray-level variations reflecting lower trabecular mass and connectivity (2–4). TBS has been shown to be lower in individuals with prevalent fragility fractures than in those without fractures and to predict fracture risk independent of FRAX clinical risk factors and of bone mineral density (BMD) (2,5,6).A recent meta-analysis demonstrated that the predictive value of the FRAX score for hip and major osteoporotic fractures is improved with the addition of TBS (7). Body mass index (BMI) has been shown to be a reliable surrogate for waist circumference, for global adiposity, and also for visceral adiposity in both men and women (8,9). Increased amounts of abdominal soft tissue, such as central adiposity in individuals with elevated BMI, absorb more X-rays during DXA acquisition and will result in lower “raw” TBS values (6). To compensate for this effect that are in part BMI based, adjustments have been incorporated in the TBS algorithm (version 2.1) (10), and TBS measurement has been optimized for individuals with a BMI in the range of 15–37 kg/m2 (6). We hypothesized that bone densitometers from different manufacturers may not provide comparable TBS measurements, particularly in individuals with greater abdominal adiposity, because DXA images may be degraded differentially across densitometers because of intrinsic differences in how soft tissue effects are accommodated during DXA scan acquisition and processing. Using data from the Canadian Multicentre Osteoporosis Study (CaMos), we aimed to determine if the relationship between BMI and TBS varied across the 2 major densitometer manufacturers, namely, GE and Hologic.
Methods Participants and Study Design CaMos is an ongoing Canada wide population-based prospective cohort study that enrolled 9423 community-
dwelling participants (6539 women and 2884 men) aged 25 yr and older in 1995–1997. Subjects were recruited from a list of randomly selected residential phone numbers within a 50-km radius of centers in 9 cities (CaMos research centers) across the country. The participants were randomly selected from eligible household members using an age- and sex-stratified design, which oversampled women and older groups to ensure sufficient sample size among those at the highest risk of fracture. The sample frame consisted of 40% of the Canadian population (11). Data collection at baseline and every 5 yr included intervieweradministered questionnaires, clinical assessments, and BMD measurements by DXA (total hip, femoral neck, and lumbar spine, L1–L4) for all centers. Ethics approval was granted through McGill University and the research ethics board for each participating center. Signed informed consent was obtained from all study participants in accordance with the Helsinki Declaration. The current study represents a cross-sectional analysis of 1919 women and 811 men aged ≥40 yr who had TBS calculated from lumbar spine DXA images for BMD measurements collected at year 10 of follow-up (2005–2007). Seven centers (Vancouver, Calgary, Saskatoon, Toronto, Quebec City, Halifax, and St. John’s) were part of the analysis, because BMD images obtained from DXA machines in 2 centers (Hamilton and Kingston) at year 10 were not compatible with the TBS algorithm.
Measurements Height was measured using either a ruler on the wall, a measure on a weight scale, or a wall-mounted stadiometer, and weight was measured in light clothing using a beam balance or electronic scale (12). Standard protocols are in place at each center for the measurement of height (in centimeter) and weight (in kilogram), performed at the time of DXA scanning; however, there were no cross-calibration across sites. BMI was calculated using weight in kilogram divided by height in square meter. Four centers (Vancouver,Toronto, Halifax, and St. John’s) used GE Lunar densitometers (Prodigy; GE Healthcare, Madison, WI) and 3 centers (Calgary, Saskatoon, and Quebec City) used Hologic densitometers (QDR 4500, Discovery, or Delphi; Hologic Inc., Bedford, MA). BMD measurements from GE Lunar densitometers were converted to equivalent Hologic values using published methods (13).
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume ■, 2016
ARTICLE IN PRESS Differences in the Correlation Between BMI and Lumbar Spine TBS Details on BMD cross-calibration and quality control have been previously published (14). TBS calculations for L1– L4 were blindly calculated in the Bone Disease Centre at the University of Lausanne in Switzerland using TBS iNsight® (version 2.1; Medimaps, Merignac, France). TBS was evaluated from the same region of interest used to calculate the lumbar spine BMD. Statistical cross-calibration was conducted based on the determination of constants (firstly for intra-DXA device calibration [e.g., Prodigy vs Prodigy] and secondly for inter-DXA device calibration [e.g., GE Lunar vs Hologic]), which were then applied to the TBS scores to compensate for differences between centers and device types.
Statistical Analysis All analyses were stratified by sex. Continuous variables were reported as means with standard deviations and counts were reported with percentages. t-Tests (continuous variables) and chi-squared tests (categorical variables) were used to compare baseline characteristics in women and men. Bivariate Pearson correlation coefficients of BMI vs TBS and BMD for GE Lunar and Hologic densitometers were calculated to determine the strengths of the relationships between these variables. Multiple linear regression analyses adjusted for age (as a continuous variable) with TBS as the dependent variable were performed to test for an interaction between BMI (as a continuous variable in the first model or categorized in tertiles in the second model) and densitometer manufacturer. The residuals of both sets of models were normally distributed as per visual inspection of the plotted residuals. A p value <0.05 was considered to be statistically significant. Statistical analyses were performed using SAS version 9.4 (Cary, NC).
Results Characteristics of the study population are described in Table 1. Compared with women, men were significantly younger and had a significantly higher BMI, lumbar spine BMD, and TBS. BMI was higher than 37 kg/m2 in 0.3% of men (n = 2) and 1.0% of women (n = 19). We observed a significant negative correlation between TBS and BMI in men (Pearson r = −0.36, p < 0.0001) and in women (Pearson r = −0.33, p < 0.0001) when TBS measurements were derived from Hologic densitometer images (Table 2). No significant correlations were observed between TBS and BMI when TBS measurements were derived from GE Lunar densitometers (Pearson r = 0.00 in men, Pearson r = −0.02 in women). In contrast, significant positive correlations were observed between BMD and BMI that were similar for the GE Lunar and Hologic densitometers in both men (Pearson r = 0.26 for Hologic, Pearson r = 0.24 for GE Lunar; p < 0.0001) and women (Pearson r = 0.26 for Hologic, Pearson r = 0.25 for GE Lunar; p < 0.0001). Scatter plots of TBS vs BMI stratified by sex and densitometer manu-
3
Table 1 Characteristics of the Study Population by Sex Characteristics DXA manufacturer, n (%) GE Lunar Hologic Age (yr)* Weight (kg)* Height (cm)* BMI (kg/m2)* L1–L4 BMD (g/cm2)* TBS*
Women (N = 1919)
Men (N = 811)
1092 (57) 827 (43) 69.6 ± 9.8 67.2 ± 11.8 159.3 ± 6.3 26.5 ± 4.3 0.931 ± 0.144 1.273 ± 0.108
480 (59) 331 (41) 66.3 ± 11.4 82.5 ± 12.6 173.9 ± 7.0 27.2 ± 3.6 1.069 ± 0.173 1.297 ± 0.107
Notes: Data are presented as means and standard deviations or counts and percentage. Abbr: BMD, bone mineral density; BMI, body mass index; DXA, dual-energy X-ray absorptiometry; TBS, trabecular bone score (unitless). *p < 0.05.
facturer along with age-adjusted linear regressions are shown in Fig. 1. In age-adjusted linear regression, with TBS as the dependent variable, there was a significant interaction between BMI and machine manufacturer (p < 0.0001). This interaction consisted of a negative association between TBS and BMI when TBS was measured on Hologic densitometers (with GE Lunar densitometers as the reference). Specifically, the participants’ TBS scores measured on Hologic densitometers were lower by 0.011 (95% confidence interval: 0.007–0.015) for each increase in 1 kg/m2 of BMI in men and by 0.006 (95% confidence interval: 0.004–0.008) in women compared with those measured on GE Lunar machines. Similar results were obtained between TBS and BMI on Hologic or GE Lunar densitometers when BMI was categorized in tertiles (data not shown).
Discussion We report the presence of a negative correlation between TBS and BMI in both men and women when TBS is measured on Hologic densitometers when using version 2.1 of the TBS algorithm. This TBS-BMI relationship was not observed when TBS is measured on GE Lunar densitometers. This is the first report on manufacturer-specific differences in the BMI-TBS relationship. In contrast, we also observed the well-recognized positive correlation between lumbar spine BMD and BMI, which was similar across the 2 manufacturers of bone densitometers (15–17). Negative correlations between TBS and BMI have previously been reported. In American adults (N = 7682, 20 yr and older) surveyed in 2 consecutive cycles of the National Health and Nutrition Examination Survey (NHANES) (2005–2008), lumbar spine BMD, measured on
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume ■, 2016
ARTICLE IN PRESS 4
Mazzetti et al. Table 2 Pearson Correlation Coefficients of BMI vs Lumbar Spine TBS and BMD by DXA Manufacturer Hologica
Lumbar spine TBS Lumbar spine BMD
GE Lunar b
BMI, Men N = 331
BMI, Women N = 827
BMI, Men N = 480
BMI, Women N = 1092
−0.36* 0.26*
−0.33* 0.25*
0.00 0.24*
−0.02 0.25*
Abbr: BMD, bone mineral density; BMI, body mass index; DXA, dual-energy X-ray absorptiometry; TBS, trabecular bone score. a QDR 4500, Discovery, or Delphi. b Prodigy. *p < 0.0001.
Hologic QDR 4500A densitometers, was found to be positively correlated in both men and women with all body size variables measured including BMI, waist circumference, total body fat mass, trunk lean mass, and trunk fat mass, whereas TBS (algorithm version 2.1) was negatively correlated with these variables (18). Mean TBS values among those with BMI values of over 37 kg/m2 were approximately 26% lower in men and 15% lower in women than in those with BMI
values between 15 and 37 kg/m2 (p < 0.001 for both sexes). The negative correlations between TBS and BMI seen in the NHANES cohort (women Pearson r = −0.33 and men Pearson r = −0.53) are close to those documented in the current CaMos study for individuals whose TBS was measured on Hologic densitometers (Pearson r = −0.33 in women and Pearson r = −0.36 in men). Using data from the Osteoporotic Fractures in Men study (MrOs), Langsetmo
Fig. 1. Scatter plots and age-adjusted linear regression lines of TBS by BMI stratified by sex and manufacturer. BMI, body mass index; TBS, trabecular bone score. Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume ■, 2016
ARTICLE IN PRESS Differences in the Correlation Between BMI and Lumbar Spine TBS et al noted that TBS (measured on Hologic QDR 4500 densitometers, TBS algorithm version 2.1) varied with BMI, trunk lean mass, and trunk fat mass such that men in the highest quintile of body size or body composition had the lowest median TBS. Increasing BMI quintile was associated with lower TBS and, in contrast, with higher volumetric BMD (19). It is noteworthy that an earlier version of the TBS algorithm (version 1.8), optimized for women of average body size, provided lower TBS scores in men and individuals with elevated BMI. The current TBS algorithm (version 2.1) is the result of modifications made to address technical limitations recognized to be associated with degraded DXA image texture in the presence of greater abdominal adiposity. Indeed, using the Manitoba (Canada) clinical BMD registry (47,736 women and 4348 men; GE Lunar densitometers), Leslie et al compared correlations between TBS and BMI using the 2 versions of the TBS algorithms. Leslie et al reported negative correlations between TBS and BMI (Pearson r = −0.18 in women, Pearson r = −0.40 in men) using the early version of the TBS algorithm (version 1.8). When the updated version of the TBS algorithm (version 2.1) was applied in the same population, the correlations between TBS and BMI changed to Pearson r = −0.01 in women and Pearson r = 0.01 in men (10,20). From the above-reported data, it appears that the updated TBS algorithm (version 2.1) is less affected by the effects of abdominal soft tissue and BMI when applied on GE Lunar densitometers than when used on Hologic densitometers. Our findings have clinical implications given that TBS is incorporated into the FRAX score calculation for fracture prediction (7). Although the predictive value of the FRAX score for osteoporotic fractures is improved with the addition of TBS, our study demonstrates that TBS measured on a Hologic densitometer will be lower in patients with a higher BMI compared with those measured on a GE Lunar machine. This finding may lead to overestimation of a given individual’s risk of fracture when TBS is incorporated in the risk calculation and subsequent overtreatment of patients. Further confusion could arise if individuals have serial TBS measurements on machines from different manufacturers which might be affected by the manufacturer-specific relationships between BMI and TBS, or if an individual has a significant change in BMI and TBS is measured serially on a Hologic densitometer. Finally, implications for multicenter research studies that include measurements of TBS on different densitometers should also be recognized. The strengths of our study include its large sample size, the inclusion of both men and women, and a diversity of DXA instruments. In addition, all TBS measurements were calculated at the same time using the most recent software TBS algorithm available (version 2.1), allowing for a direct comparison between the TBS values. Our study is population based, allowing for the results to be generalizable to the Canadian population, although the effect of different race and ethnicity cannot be examined because
5
the majority of this cohort is white. Other limitations must be mentioned. Firstly, we did not have TBS measurements in the same participants measured on both densitometers, making our inference about the relationship between BMI and TBS and densitometer manufacturer an indirect one. Secondly, we did not take into consideration comorbidities or anti-osteoporosis medication use that might have affected the relationships between TBS and BMI independent of densitometer manufacturer. Finally, our findings only apply to the algorithm (version 2.1) that was used for TBS measurement in this cohort. In conclusion, this descriptive study demonstrates that a manufacturer-specific negative correlation exists between BMI and TBS when TBS is measured on Hologic densitometers as compared to GE Lunar densitometers. This finding has implications for clinical and research applications of TBS, especially when TBS is measured sequentially on DXA densitometers from different manufacturers or when results from different machines are pooled for analysis.
Acknowledgments The Canadian Multicentre Osteoporosis Study is currently funded by the Canadian Institutes of Health Research (CIHR, MOP 111103), Amgen, and Eli Lilly Canada. These funding sources had no role in the conception of this analysis, statistical methods, or interpretation of the data. SN Morin is a research scholar of the Fonds de Recherche du Québec en Santé.
References 1. Silva BC, Leslie WD, Resch H, et al. 2014 Trabecular bone score: a noninvasive analytical method based upon the DXA image. J Bone Miner Res 29:518–530. 2. Binkley N, Leslie WD. 2016 Clinical application of spine trabecular bone score (TBS). Clin Rev Bone Miner Metab 14:14–25. 3. Silva BC, Bilezikian JP. 2014 Trabecular bone score: perspectives of an imaging technology coming of age. Arq Bras Endocrinol Metabol 58:493–503. 4. Pothuaud L, Carcellar P, Hans D. 2008 Correlations between grey-level variations in 2D projection images (TBS) and 3D microarchitecture: applications in the study of human trabecular bone microarchitecture. Bone 42:775–787. 5. McCloskey EV, Odén A, Harvey NC, et al. 2015 Adjusting fracture probability by trabecular bone score. Calcif Tissue Int 96:500–509. 6. Silva BC, Broy SB, Boutroy S, et al. 2015 Fracture risk prediction by non-BMD DXA measures: the 2015 ISCD official positions part 2: trabecular bone score. J Clin Densitom 18:309–330. 7. McCloskey EV, Odén A, Harvey NC, et al. 2016 A metaanalysis of trabecular bone score in fracture risk prediction and its relationship to FRAX. J Bone Miner Res 31:940– 948. 8. Borruel S, Moltó JF, Alpañés M, et al. 2014 Surrogate markers of visceral adiposity in young adults: waist circumference and body mass index are more accurate than waist hip ratio, model
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume ■, 2016
ARTICLE IN PRESS 6
9.
10.
11. 12.
13.
14.
Mazzetti et al. of adipose distribution and visceral adiposity index. PLoS ONE 5:e114112. Sunmer AE, Micklesfield LK, Ricks M, et al. 2011 Waist circumference, BMI, and visceral adipose tissue in white women and women of African descent. Obesity (Silver Spring) 19:671– 674. Leslie WD, Winzenrieth R, Majumdar SR, et al. 2014 Clinical performance of an updated version of trabecular bone score (TBS) in men and women: the Manitoba BMD Cohort. J Bone Miner Res 29(Suppl 1). Kreiger N, Tenenhouse A, Joseph L, et al. 1999 The Canadian Multicentre Osteoporosis Study (CaMos): background, rationale, methods. Can J Aging 18:376–387. Hopman WM, Leroux C, Berger C, et al. 2007 Changes in body mass index in Canadians over a five-year period: results of a prospective, population-based study. BMC Public Health 7:150. Genant HK, Grampp S, Glüer CC, et al. 1994 Universal standardization for dual X-ray absorptiometry: patient and phantom cross-calibration results. J Bone Miner Res 9:1503– 1514. Berger C, Goltzman D, Langsetmo L, et al. 2010 Peak bone mass from longitudinal data: implications for the preva-
15.
16. 17. 18.
19.
20.
lence, pathophysiology and diagnosis of osteoporosis. J Bone Miner Res 25:1948–1957. Felson DT, Zhang Y, Hannan MT, Anderson JJ. 1993 Effects of weight and body mass index on bone mineral density in men and women: the Framingham study. J Bone Miner Res 8:567–573. Reid IR. 2002 Relationships among body mass, its components and bone. Bone 31:547–555. Morin SN, Leslie WD. 2009 High bone mineral density is associated with high body mass index. Osteoporos Int 20:1267– 1271. Looker AC, Sarafrazi Isfahani N, Fan B, Shepherd JA. 2016 Trabecular bone scores and lumbar spine bone mineral density of US adults: comparison of relationships with demographic and body size variables. Osteoporos Int 27:2467– 2475. Langsetmo L, Vo TN, Ensrud KE, et al. 2016 The association between trabecular bone score and lumbar spine volumetric BMD is attenuated among older men with high body mass index. J Bone Miner Res 10:1820–1826. Leslie WD, Krieg MA, Hans D. 2013 Clinical factors associated with trabecular bone score. J Clin Densitom 16:374– 379.
Journal of Clinical Densitometry: Assessment & Management of Musculoskeletal Health
Volume ■, 2016