Relationship between body composition and bone mineral content in young and elderly women.

Primary objective : To study the relationship between bone mineral content (BMC), lean tissue mass (LTM) and fat mass (FM) in a large sample of young and elderly women. Research design : Cross-sectional. Methods and procedures : BMC, LTM and FM were measured by dual-energy X-ray absorptiometry in 2009 free-dwelling Caucasian women aged 63 - 7 years (mean - SD; range: 37-88 years). The majority of women were postmenopausal (96%). Results : LTM explained 13% more variance of BMC than FM ( R 2 adj = 0.39 vs 0.26, p < 0.0001) but weight (Wt) explained 5% more variance of BMC than LTM ( R 2 adj = 0.44, p < 0.0001). The prediction of BMC obtained from LTM and FM ( R 2 adj = 0.46, p < 0.0001) was only slightly better than that obtained from Wt. After the effects of age, Wt and height (Ht) on BMC were taken into account by multiple regression, the contribution of LTM and FM to BMC was just one-fifth of that of Wt ( R 2 adj for full models r 0.56, p < 0.0001). After a further correction for bone area (BA), the contribution of LTM and FM to BMC was just one-tenth of that of BA and not different from that of Wt and Ht on practical grounds ( R 2 adj for full models = 0.84, p < 0.0001). Thus, after inter-individual differences in age, Wt, Ht (and bone size) are taken into account, the relationship between body composition and BMC is substantially weakened. Conclusions : In Caucasian women, (1) LTM is a stronger predictor of BMC than FM, but (2) Wt is a better predictor of BMC than body composition for practical purposes, and (3) Wt and body composition are not able to explain more than 46% of BMC variance.


Introduction
Ageing is accompanied by a progressive decline in bone mineral content (BMC) and density (BMD). The identi®cation of the factors responsible for this decline may help prevent its pathological manifestations, i.e. osteopenia and osteoporosis (Christiansen 1995).
A direct relationship exists between body weight (Wt), BMC and BMD, with overweight subjects experiencing the lowest prevalence of osteoporosis and incidence of fractures (Wardlaw 1996). However, the contribution of fat and lean tissues to this protective e ect of Wt is controversial (Taa e, Villa, Holloway et al. 2000). While some studies have shown that lean tissues are stronger predictors of BMC and BMD than fat tissues (Aloia, Vaswani, Ma et al. 1995, Chen, Lohman, Stini et al. 1997, the opposite was shown by others (Compston, Bhambhani, Laskey et al. 1992, Reid, Ames, Evans et al. 1992, Reid, Plank and Evans 1992, Taa e et al. 2000. Decrease of BMC is however just one of the changes in body composition that occur with ageing (Van Loan 1996). A decrease in fat-free components other than BMC and an increase in fat mass (FM) were observed in elderly as compared with young women (Mazariegos, Wang, Gallagher et al. 1994). Apart from these mod-i®cations, the relationship between body compartment s may change with ageing and this may have important implications for the prevention of osteoporosis. In order to answer this question, a study sample consisting of both young and elderly women should be employed.
The present study aimed therefore at assessing the contribution of lean and fat tissues to BMC in a large sample of young and elderly women.

Subjects
The study hypothesis was tested on a large series of free-dwelling women evaluated at our Centre during a study on nutritional status and osteoporosis (Bedogni, Simonini, Viaggi et al. 1999). All the women recruited as of September 2001 who had never made use of oestrogens, diphosphonates or vitamin D were selected for this study (n = 2009). The study protocol was approved by the local Ethical Committee and all subjects gave their informed consent.

Dual-energy X-ray absorptiometr y (DXA)
FM, lean tissue mass (LTM) and BMC were measured using a Lunar DPX-L densitometer (Lunar Corporation, Cary, NC, USA, software version 3.6). Percent fat mass (FM:Wt), percent lean tissue mass (LTM:Wt) and percent bone mineral content (BMC:Wt) were obtained by dividing FM, LTM and BMC, respectively, by Wt. The di erence between body mass measured by DXA and Wt measured by scale was ¡ 0.5 § 1.1 kg (mean § SD; n = 2009). Although this di erence is statistically signi®cant (p < 0.0001, paired t-test), it amounts to only ¡ 1 § 2% (mean § SD) of Wt and is therefore negligible on practical grounds.

Statistical analysis
Statistical analysis was performed on a MacOS computer using the Statview 5.0.1 (SAS, Chicago, IL, USA) and SPSS 10.0 (SPSS, USA) software packages. BMC was log-transforme d to better approach the normal distribution. Between-group comparisons were performed by unpaired t-tests. The adjusted determination coe cient (R 2 adj ) and the root mean square error (RMSE) obtained from simple and multiple regressions of BMC versus anthropometric dimensions, body compartments and age were used to quantify the contribution of these variables to BMC (Guo, Chumlea and Cockram 1996). To control the e ects of age, Wt and Ht on the relationship between BMC and body composition, a multiple regression model was employed using age, Wt, Ht and LTM:Wt or FM:Wt as predictors. Another model added bone area (BA) to the above predictors to control the confounding e ect of bone size on BMC (Prentice, Parsons and Cole 1994). All regressions were performed on logtransformed values to ensure homoscedasticity of residuals. Statistical signi®cance was set to a value of p < 0.05 for all tests.

Results
The measurements of the study subjects are given in table 1. The 2009 studied women were aged 63 § 7 years (mean § SD; range: 37±88 years). A total of 786 women were aged 65 years or higher and were classi®ed as`elderly' while the remaining 1223 women were classi®ed as`young'. This classi®cation was made for descriptive purposes only because age is a continuous variable whose association with body composition is better controlled for by regression analysis. As in our previous report (Bedogni et al. 1999), the majority of women were post-menopausal (96%).
As expected, age was higher (p < 0.0001) and Ht lower (p < 0.0001) in elderly than young women. However, Wt, BMI, FM and FM:Wt were not di erent between groups (p = NS). LTM was lower in elderly than young women (p = 0.01) but no di erence was seen for LTM:Wt (p = NS). As expected, BMC and BMC:Wt were signi®cantly lower in elderly than young women (p < 0.0001). The percentage of osteoporotic and osteopenic women was virtually the same observed in our previous report (Bedogni et al. 1999): 8% and 37% respectively.
The variance of BMC explained by age, BA, Wt, Ht, LTM and FM and selected combinations of them is given table 2. As expected (Prentice et al. 1994), BA was the strongest predictor of BMC (R 2 adj = 0.80). Even if LTM explained 13% more variance of BMC than FM (R 2 adj = 0.39 vs 0.26), Wt explained 5% more variance of BMC than LTM (R 2 adj = 0.44). Moreover, the prediction of BMC obtained from LTM and FM (R 2 adj = 0.46) was only slightly better than that obtained from Wt. Age explained only 8% of BMC variance. However, its use as a predictor with Wt increased the explained variance of BMC by 7% as compared to Wt alone (R 2 adj = 0.51). A lower increase in the explained variance of BMC (4%) was seen when age was used as a predictor with LTM and FM (R 2 adj = 0.50). Taken together, these data suggest that even if LTM is a better determinant of BMC than FM, it is not superior to Wt. To test this hypothesis more thoroughly, we evaluated the contribution of LTM:Wt and FM:Wt to BMC after correction for age, Wt, Ht and BA (tables 3 and 4). (Using LTM instead of LTM:Wt or FM instead of FM:Wt in the same models was prone to multicollinearityÐwith variance in¯ation factors as high as 9 for the FM modelsÐand was thus avoided.) The contribution of LTM:Wt to BMC after correction for age, Wt and Ht was 18% of that of Wt (model A1 of table 3; standardized regression coe cient or ˆ0:12 vs 0.67) and removing LTM:Wt from the model did not change the accuracy of the estimate. When BA was added to the predictors, the contribution of

Discussion
The contribution of LTM and FM to the protective e ect of Wt on BMC is controversia l (Taa e et al. 2000). The present study suggests that in postmenopausa l women LTM is a better predictor of BMC than FM. Our results are thus in agreement with those of Chen et al. (1997) (n = 50) and Aloia et al. (1995) (n = 164), showing that LTM measured by DXA or fat-free mass measured by a variety of methods is the major determinant of BMC or total body calcium in postmenopausa l women. An advantage of this study as compared to the others available in the literature is its very high number of subjects (n = 2009), which allows a greater degree of con®dence in the results. However, a transversal study cannot by its very nature test any cause±e ect relationship and only longitudinal studies should establish whether LTM is a more useful predictor of BMC than FM. Since there is preliminary evidence that LTM is inversely associated with the occurrence of bone fractures (Takada, Washino and Iwata 1997), the ®nding of LTM as the body compartment most strongly associated with BMC, suggests the opportunity of trials aimed at testing whether interventions targeted at increasing LTM can reduce the risk of osteoporosis. These longitudinal trials would also o er the possibility of separating more thoroughly the e ect of LTM on BMC from that of FM.
Even if LTM emerges from this study as a better predictor of BMC than FM, it was not superior to Wt. In fact, after inter-individual di erences in age, Wt and Ht were taken into account, the relationship between LTM and BMC was substantially weakened (this may be partly due to the higher precision with which Wt is measured as compared to LTM). This implies that Wt is to be preferred to LTM on practical grounds for selecting patients with low BMC. Wt is, in fact, simpler to measure and o ers a better discrimination of normal, osteopenic and osteoporotic women than many other anthropometric indicators (Bedogni et al. 1999). This does not modify, however, the pathophysiologica l relevance of being able to separate the e ects of LTM on BMC from those of FM by means of longitudinal studies.
Use of LTM and FM as predictors did increase the accuracy of the estimate of BMC as compared with LTM alone but this estimate was only slightly better than Abbreviations: coe . = regression coe cient; std. coe . = standardized regression coe cient; p-coe . = p-value for the regression coe cient; R that based on Wt. It is nonetheless of interest that the inclusion of age among the predictors did increase the power of the estimate. However, the most relevant ®nding of this study is that age and body compartments leave a large portion of BMC variability unexplained. A value of 50% for the unexplained variance of BMC does indeed suggest that factors other than age and body composition in¯uence BMC. Among these factors, genetics may play a role, as shown by twin studies, but environmental factors, especially nutrition and physical activity, may be involved too (Seeman, Hopper, Young et al. 1996, Nguyen, Howard, Kelly et al. 1998. We conclude that in Caucasian women: (1) LTM is a stronger predictor of BMC than FM, but (2) Wt is a better predictor of BMC than body composition for practical purposes, and (3) Wt and body composition are not able to explain more than 46% of BMC variance.