Risk Factors for Low Bone Mineral Density among Females: The Effect of Lean Body Mass

26, 633–638 (1997) PM970170

PREVENTIVE MEDICINE ARTICLE NO.

Risk Factors for Low Bone Mineral Density among Females: The Effect of Lean Body Mass Haruko Takada, M.D.,1 Kaei Washino, M.S., and Hirotoshi Iwata, M.D. Department of Hygiene, Gifu University School of Medicine, 40, Tsukasa-machi, Gifu City 500, Japan

Background. This study determines if body composition and lifestyle are risk factors for low radius-bone mineral density (R-BMD) and evaluates the role of body composition in the age-related decline of R-BMD. Methods. Data on age, menopausal status, fat mass, lean body mass (LBM), drinking, smoking, and occupation were collected from 3,867 females ages 37–69, whose R-BMD was also measured. Multiple logistic regression analyses examined the predictive accuracy of these factors for low (20th percentile) R-BMD. Results. Age, LBM, and menopausal status were risk factors for the 37–55 age group, while age, LBM, and lifestyle (alcohol consumption 3 or fewer days per week and currently smoking) were risk factors for the 56–69 age group. The odds ratio (OR) for LBM was nearly reciprocal to the ORs for age and for menopausal status. Our model has low sensitivity, high specificity, low positive predictive value, and high negative predictive value. Conclusions. Consumption of alcohol 3 or fewer days per week and being a current smoker have a negative effect on R-BMD among older (56–69) women. The positive effect of LBM on R-BMD continues from age 37 on. LBM has an effect almost equal but opposite to those of aging and menopause on the risk of low R-BMD. High LBM after age 37 predicts normal R-BMD. © 1997 Academic Press

Key Words: body composition; bone mineral density; drinking; female; smoking.

INTRODUCTION

It is well known that bone mineral density (BMD) decreases with age and is also influenced by menopause and body weight. Recent studies have documented the positive correlation between BMD and body weight among children and young people [1,2], middle-aged people [3–5], and people over 70 [6]. With regard to the influence of weight loss on BMD, Avenell et al. carried out a study in which 16 obese postmenopausal women lost 20% of their excess body weight over a 6-month period and then returned to their original weights over the 1

To whom reprint requests should be addressed. Fax: 058-267-2958.

next 6 months. They reported a significant decrease of spinal BMD, which did not return even after the weight gain in the second 6-month period [7]. Cohn et al. speculatively suggested a positive correlation between the BMD of the radius (R-BMD) and muscle mass on the basis of a study that showed that black women had greater R-BMD and greater muscle mass than white women [8]. A recent study concerning the relationship between BMD and body composition among 139 men has documented that lean body mass (LBM) is significantly related to BMD/height [9], while similar studies with female subjects have given mixed results. Three studies on women have suggested the correlation is between BMD and LBM, rather than fat mass (FM) [10–12]. One investigator has produced two studies indicating the correlation is between BMD and FM, as opposed to LBM [13,14], and a single study maintains both correlations (LBM and FM) are valid [15]. It is likely that the result is complicated by the various regions used to measure BMD and because both BMD and body composition change with age. These studies used small sample sizes and subjects in a limited age range [9–15]. Our study is different in that we use (a) a large population size, (b) a single, consistent method of measuring R-BMD—the dual energy X-ray absorptiometry (DXA) method, (c) a newly developed version of the tetrapolar bioelectrical impedance method, one with greater reliability, to estimate body composition, and (d) subjects with a broad range of ages. These four characteristics allow us to clearly determine which component of body composition (LBM or FM) affects R-BMD. The three purposes of this cross-sectional study are (1) to determine if body composition and lifestyle are risk factors for low R-BMD, (2) to measure the present mean R-BMD for Japanese women as a function of age, and (3) to evaluate the role of body composition in the age-related decline of R-BMD. METHODS

Subjects Data for this study come from the records of physical examinations performed at The Gifu Prefecture Health Center in Gifu City from April 1995 to May 1996.

633 0091-7435/97 $25.00 Copyright © 1997 by Academic Press All rights of reproduction in any form reserved.

634

TAKADA, WASHINO, AND IWATA

The few subjects who were below 30 and above 70 were eliminated from the data. This left 3,867 females, with a mean age of 48.6 years and a standard deviation (SD) of 8.0 years, in the data. The data included body height (Ht), body weight (Wt), body mass index (BMI), percentage of body fat (%Fat), R-BMD, drinking status, smoking status, and job type. The use of data for research purposes was approved by the Director of the Health Center. Anthropometry Percentage of fat was estimated using tetrapolar bioelectrical impedance analysis. The tetrapolar impedance method has been found to be both valid and reliable [16–18]. In this study, measurements were made after more than 8 hr of fasting and within 30 min after voiding. The subjects stood barefooted on the scale of a tetrapolar impedance analyzer (TBF-202, Tanita, Co. Tokyo; constant current: 50 kHz, 0.8 mA). Currentinjector electrodes were placed just below the balls of the feet, and detector electrodes were placed just below the heels. Then bioelectrical impedance was measured. The correlation between %Fat as measured by TBF202 and hydrodensitometry methods is r 4 0.841, P < 0.001, SEE 4 2.9 for women. The coefficients of variation (CV%) for estimated [19] %Fat is 0.38% [19]. Sakamoto’s formula was used to calculate the estimated body density (Db) from Ht, Wt, and bioimpedance (Z). %Fat, FM, and LBM were derived from Db using the formulas below [20]: Db = 1.0907 1 0.112 2 Wt (kg)/[Ht (cm)]2 2 Z (ohms) + 0.000134 × Z (ohms), %Fat = (4.570/Db 1 4.142) 2 100, FM = Wt 2 %Fat/100, and LBM 4 Wt − FM. The impedance analyzer (TBF-202) includes scales for measuring Wt and Ht, so that these were automatically measured at the same time. BMI was calculated from the formula BMI 4 Wt (kg)/[Ht (m)]2. The Measurement of R-BMD DCS-600EX (ALOKA, Co. Tokyo) was used to determine BMD at the sites of distal radius 1/3. Dual energy X-ray absorptiometry has been shown to be both valid and reliable [21,22]. The correlation between bone mineral content (BMC) as measured by DCS-600EX and the standard BMC (K2HPO4) is r 4 0.999, P < 0.001. CV% for BMC is 0.72–1.30% [23].

The Question Format The question format consists of the categories used in the Health Center’s physical examination. Postmenopause (yes/no) was defined as no menstruation for a period of 1 year or more. Drinking status (yes/no) consisted of alcohol consumption more (yes) or fewer (no) than 3 days a week. Smoking status was treated simply as a dichotomous category (yes/no), while job type contained three categories (mostly standing/ mostly sitting/no job). Statistical Analyses (1) Ht, Wt, BMI, %Fat, FM, LBM, and R-BMD measurements were compared between four groups—ages 30–36 (all subjects premenopausal), ages 37–55 (premenopausal), ages 37–55 (postmenopausal), and ages 56–69 (all subjects postmenopausal)—using a one-way ANOVA. (2) To indicate the effect of aging on R-BMD, a cubic formula for R-BMD with age as the independent variable was determined from the mean R-BMD values for each age (taken at 1-year intervals). The age of peak R-BMD was then calculated from this formula. (3) One-way ANOVAs were conducted to compare the mean R-BMD between post- and premenopausal groups for subjects from ages 37 to 55 at 1-year intervals and for the groups as a whole. (4) To determine risk factors for low R-BMD (defined as the lowest 20th percentile: 0.621 g/cm2 in the 30–36 age group, 0.585 g/cm2 in the 37–55 age group, and 0.459 g/cm2 in the 56–69 age group), multiple logistic regression analyses (performed stepwise procedure) were conducted for three groups: (a) ages 30–36, (b) ages 37–55 (both pre- and postmenopausal groups included), and (c) ages 56–69. For the 30–36 and 56–69 age groups there are five independent variables: three continuous variables— age, FM, and LBM—and two discrete variables—a two-dimensional (drinking/smoking) composite variable and job type. The reference categories set for the two discrete variables were no drinking/no smoking and mostly standing. For the 37–55 age group, there is the additional independent variable of menopausal status. In addition, the FM and LBM variables have been made discrete because of the nonlinear relationships between each of the variables taken separately and the corresponding log odds ratios of low BMD. These discrete variables have then been combined into a single two-dimensional composite variable so as to give the model a better fit as determined using Akaike’s information criterion. A dichotomous FM variable (<15/ù15) was used as one dimension and a dichotomous LBM variable (<35/LBM ù35) as the other dimension. The reference category set for this variable is FM <15 and LBM ø 35. All

635

THE EFFECT OF LEAN BODY MASS

TABLE 1 Descriptive Data of Subjects Premenopause group Ages 30–36 (n 4 228)

Age (years) Body height (m) Body weight (kg) BMI %Fat (%) FM (kg) LBM (kg) R-BMD (g/cm2)

Postmenopause group

Ages 37–55 (n 4 2,083)

Ages 37–55 (n 4 752)

Ages 56–69 (n 4 804)

Mean

(SD)

Mean

(SD)

Mean

(SD)

Mean

(SD)

34.0 1.59 53.1 21.0 26.3 14.4 38.8 0.65

(1.9) (0.05) (8.3) (3.0) (5.7) (5.7) (3.5) (0.05)

44.7 1.56 53.9 22.1 27.0 14.8 39.0 0.63

(4.3)** (0.05)** (7.7) (2.9)** (5.4) (4.9) (3.8) (0.05)**

51.3 1.54 52.8 22.2 27.4 14.7 38.1 0.59

(3.6)** (0.05)** (7.1) (2.7)* (5.0) (4.4) (3.5) (0.07)**,†

60.0 1.52 52.6 22.8 28.3 15.2 37.4 0.52

(3.5)** (0.05)** (7.3) (3.0)** (5.5)** (4.8) (3.5)** (0.07)**

Note. BMI, body mass index; %Fat, percentage of fat; FM, fat mass; LBM, lean body mass; R-BMD, bone mineral density of radius. * P < 0.05 in reference to the 30–36 age group. ** P < 0.01 in reference to the 30–36 age group. † P < 0.01 in reference to the premenopausal 37–55 age group.

statistical analyses were computed using SAS statistical package (Version 6.04). The probability of low R-BMD was calculated for each subject using the appropriate logistic model for the subject’s group, and predicted prevalence of low R-BMD was used to confirm the goodness of fit for each model. (5) Multiple linear regression analysis was performed using a stepwise procedure. R-BMD was the dependent variable, and there were six independent variables: age, FM, LBM, menopausal status, drinking status, and smoking status. RESULTS

(1) The descriptive data for all subjects are shown in Table 1. When mean values of all three other groups were compared with the means for the 30–36 age group, the mean BMIs and R-BMDs were all found to be significantly higher. The mean %Fat was signifi-

FIG. 1. R-BMD as a function of age (n 4 3,867). Mean R-BMD values with SD for each age (taken at 1-year intervals).

cantly higher only for the 56–69 age group (P < 0.01). Only the mean LBM for the 56–69 age group was significantly lower (P < 0.01). No significant differences appeared with respect to Wt and FM. The mean R-BMD for the postmenopausal 37–55 age group was significantly lower than that for premenopausal 37–55 age group. (2) The R-BMD curve shows a continuous decrease with respect to age after reaching a peak mean R-BMD at age 36–37 (Fig. 1). A cubic equation was the simplest formula that could be found with an appropriately high regression coefficient (indicating the level of significance for the test of goodness of fit). The formula for the curve is Y 4 6.90 × 10−6 × X 3 − 1.14 × 10−3 × X2 + 5.55 × 10−2 × X − 0.18,

FIG. 2. R-BMD with respect to age for the premenopause (n 4 2,311) and postmenopause (n 4 1556) groups. Mean R-BMD values with SD for each age (taken at 1-year intervals). *P < 0.05, **P < 0.01; significantly low R-BMD in reference to the premenopause group for each year.

636

TAKADA, WASHINO, AND IWATA

TABLE 2-1 Multiple Logistic Model of Low BMD (Bottom 20th Percentile, 0.585 g/cm2 or Less) in the 37–55 Age Group (Model 1) Variable

Regression coefficient

Standard error

F value

P value

Age (years) Menopausal status LBM ù35/FM <15 LBM ù35/FM ù15 Const.

0.1383 0.5445 −0.5177 −0.5014 −8.1682

0.0126 0.1185 0.1378 0.1394 0.5570

120.03 21.11 14.12 12.93 215.02

<0.001 <0.001 <0.001 <0.001 <0.001

Odds ratio (95% confidence interval) 1.15 (1.12–1.18) 1.72a (1.37–2.18) 0.60b (0.46–0.78) 0.61b (0.46–0.80)

Note. All subjects: n 4 2,835; low-BMD subjects: n 4 614. Conducted by multiple logistic regression analysis. Odds ratio in reference to premenopausal group. b Odds ratio in reference to LBM <35/FM <15 group. a

where X is age in years and Y is R-BMD in g/cm2. The model R2 is 0.971, with a P < 0.001. The R-BMD curve shows a rate of decrease of 0.3% per year between ages 40 and 50, so that the mean R-BMD has been reduced by age 50 to 90% of the peak mean value (at age 36–37). This rate of decrease increases to 2.1% between ages 50 and 55, then drops back down to 1.5% for the next 5-year period, so that the mean R-BMD has diminished to 80% of the peak mean value by age 60. The rate then drops again to the initial rate of 0.3% for ages 60–69, bringing the mean R-BMD down by age 69 to 75% of the peak mean value. (3) The mean R-BMD for the subgroups (taken at 1-year intervals) was always lower and often significantly lower for the postmenopausal group, ages 45 (P < 0.05), 47 (P < 0.05), 49 (P < 0.05), 51 (P < 0.01), and 52 (P < 0.01), than for the premenopausal group (Fig. 2). The curve for the postmenopausal group is steeper than that for the premenopausal group. (4-1) The multiple logistic regression analysis for the 30–36 age group indicated no significant risk factors for low R-BMD. (4-2) The multiple logistic regression analysis for the 37–55 age group (Logistic Model 1) indicated four significant factors for low R-BMD: two positive factors— age and postmenopausal status—and two negative factors—LBM ù 35/FM < 15 and LBM ù 35/FM ù 15. The analysis shows an odds ratio (OR) for low BMD to be 1.15 per year increase, 1.72 for the difference in menopausal status, 0.60 for LBM ù 35/FM < 15, and 0.61 for LBM ù 35/FM ù 15 (Table 2-1).

Logistic Model 1 (for ages 37–55) was able to predict low R-BMD in only 109 of the 614 subjects measured by the DXA method as having low R-BMD (sensitivity 0.176); 2,157 of the 2,221 subjects with normal R-BMD were predicted (specificity 0.971). Thus, the positive predictive value was 0.63 (109 subjects in low R-BMD among 173 subjects in predictive low R-BMD: 109/173), the negative predictive value was 0.81 (2,157 subjects in normal R-BMD among 2,662 subjects in predictive normal R-BMD: 2,157/2,662), and predictive accuracy was 0.80 (2,266/2,835). (4-3) The multiple logistic regression analysis for the 56–69 age group (Logistic Model 2) indicated three significant risk factors for low R-BMD: two positive factors—age and no drinking/yes smoking—and one negative—LBM. The analysis shows an OR for low R-BMD of 1.16 per year increase and 0.84 per kilogram increase in LBM and 2.19 for no drinking/yes smoking (Table 2-2). Logistic Model 2 (for ages 56–69) was able to predict low R-BMD in only 27 of 174 subjects measured by the DXA method as having low R-BMD (sensitivity 0.155) and 608 subjects of 630 as having normal R-BMD (specificity 0.965). The positive predictive value was 0.55 (27/49), the negative predictive value was 0.81 (608/755), and predictive accuracy was 0.79 (635/804). (5) The positive effect of LBM was also supported by multiple linear regression analysis. Increased age, postmenopausal status, and increased LBM were all correlated with increased R-BMD (partial correlation coefficients being −0.3629, −0.1440, and 0.2142, respec-

TABLE 2-2 Multiple Logistic Model of Low BMD (Bottom 20th Percentile, 0.459 g/cm2 or Less) in the 56–69 Age Group (Model 2) Variable

Regression coefficient

Standard error

F value

P value

Age (years) LBM (kg) No drinking/yes smoking Const.

0.1455 −0.1736 0.7845 −3.7365

0.0248 0.0289 0.3625 1.8395

34.31 36.15 4.69 4.13

<0.001 <0.001 <0.05 <0.05

Note. All subjects: n 4 804; low-BMD subjects: n 4 175. Conducted by multiple logistic regression analysis. Odds ratio in reference to no drinking/no smoking group.

a

Odds ratio (95% confidence interval) 1.16 (1.10–1.21) 0.84a (0.79–0.89) 2.19a (1.08–4.46)

637

THE EFFECT OF LEAN BODY MASS

tively, each of them significant (P < 0.01). R2 4 0.6354; P < 0.01). DISCUSSION

Four risk factors for low R-BMD were found in this study. As for lifestyle, ‘‘alcohol consumption 3 or fewer days per week and currently smoking’’ appeared as a risk factor for low BMD in the late 50s and after. This result is supported by previous studies that showed that tobacco consumption is associated with reduction in BMD among subjects over 60 [24], BMD among smokers was lower than among nonsmokers [25], and alcohol intake has a positive effect on BMD [26]. The effect of drinking and smoking on the age-related decline in BMD is clearly shown in our analyses, because we employed a two-dimensional composite variable of drinking and smoking. From the cubic formula of BMD determined in this study, the age of peak R-BMD was found to occur at age 36–37. The mean R-BMD has decreased by age 50 to 90% of the peak mean value (at age 36–37) and to 80% of the peak mean value by age 60, bringing the mean R-BMD down by age 69 to 75% of the peak mean value. Thus we can clearly see the shape of age-related decline in R-BMD. As for the effects of body composition on BMD, researchers have tried to separate the individual effects of FM and LBM from their joint effect on the agerelated decline in BMD [9–15]. The results have been mixed. Through metaanalyses, this study has succeeded in clearly separating the effects of LBM and FM. We found that there was no risk factor for low R-BMD in the 30–36 age group. Neither age nor LBM had any correlation with R-BMD in the early 30s. In the 37–55 and 56–69 age groups, on the other hand, high LBM was a negative risk factor for low R-BMD and FM had no independent effects. The positive effect of LBM on R-BMD continues from the late 30s on. ORs for low R-BMD were compared to evaluate the role of LBM in two different age groups. For the 37–55 age group, the OR for a 1-year increase in age was 1.15. The OR, with respect to LBM < 35 with FM < 15, for LBM ù 35 for either FM value was 0.60–0.61, while the postmenopausal group had the reciprocal OR of 1.72. For the 56–69 age group, the OR for a 1-kg increase in LBM was 0.84, while the OR for a 1-year increase in age was 1.16, a nearly reciprocal relationship. ORs show two positive risk factors (age from the late 30s and menopausal status) and one negative (high LBM) to be of about equal strength. The goodness of fit for the two groups was analyzed to determine the model’s limits as a predictor of BMD level. The results—low sensitivity, high specificity, low positive predictive value, and high negative predictive value—suggest that our model is much better at predicting prevalence of normal or high R-BMD than of low R-BMD. From these

results, we can draw four conclusions: (1) alcohol consumption 3 or fewer days per week and currently smoking have a negative effect on R-BMD after the late 50s, (2) the positive effect of LBM on R-BMD continues from the late 30s on, (3) the effect of LBM on the risk of low R-BMD is almost equal but opposite to that of aging and menopause, and (4) high LBM after the late 30s predicts normal R-BMD corresponding to the subjects’ ages. The implications for middle-aged women and above are clear: exercise is effective in preventing loss of BMD. This effect is for the most part an indirect effect, which is mediated through body composition, specifically LBM. FM has no effect on BMD. So an increase in weight as a woman ages, unless it represents an increase of LBM, is no guarantee against the loss of BMD. ACKNOWLEDGMENTS The authors thank the staff at The Gifu Prefecture Health Care Center and Drs. Yoshihiro Matsuda and Hidekatsu Takahashi for their cooperation in this study. REFERENCES 1. Lloyd T, Rollings N, Andon MB, Demers LM, Eggli DF, Kieselhorst K, et al. Determinants of density in young women. J Clin Endoclinol Metab 1992;75:383–7. 2. Kreipe RE. Bone mineral density in adolescents. Pediatr Ann 1995;24:308–15. 3. Tuppurainen M, Kroger H, Saarikoshi S, Honkanen R, Alhava E. The effect of gynecological risk factors on lumbar and femoral bone mineral density in peri- and postmenopausal women. Maturates 1995;21:137–45. 4. Nguyen TV, Kelly PJ, Sambrook PN, Gilbert C, Pocock NA, Eisman JA. Lifestyle factors and bone density in the elderly: implications for osteoporosis prevention. J Bone Miner Res 1994;9: 1339–46. 5. Alekel L, Clasey JL, Fehling PC, Weigel RM, Boileau RA, Erdman JW, et al. Contributions of exercise, body composition, and age to bone mineral density in premenopausal women. Med Sci Sports Exerc 1995;27:1477–85. 6. Ooms ME, Lips P, Van Lingen A, Valkenburg HA. Determinants of bone mineral density and risk factors for osteoporosis in healthy elderly women. J Bone Miner Res 1993;8:669–75. 7. Avenell A, Richmond PR, Lean ME, Reid DM. Bone loss associated with a high fibre weight reduction diet in postmenopausal women. Eur J Clin Nutr 1994;48:561–6. 8. Cohn SH, Abesamis C, Yasumura S, Aloia JF, Zanzi I, Ellis KJ. Comparative skeletal mass and radial bone mineral content in black and white women. Metabolism 1977;26:171–8. 9. Nuti R, Martini G, Gennari C. Age-related changes of whole skeleton and body composition in healthy men. Calcif Tissue Int 1995;57:336–9. 10. Gordon CL, Webber CE. Body composition and bone mineral distribution during growth in females. Can Assoc Radiol J 1993; 44:112–6. 11. Sowers MF, Kshirsager A, Crutchfield MM, Updike S. Joint influence of fat and lean body composition compartments on femo-

638

12.

13.

14.

15.

16.

17.

18.

TAKADA, WASHINO, AND IWATA

ral bone mineral density in premenopausal women. Am J Epidemiol 1992;136:257–65. Compston JE, Bhambhani M, Laskey MA, Murphy S, Khaw KT. Body composition and bone mass in post-menopausal women. Clin Endocrinol 1992;37:426–31. Reid IR, Ames R, Evans MC, Sharpe S, Gamble G, France JT, et al. Determinants of total body and regional bone mineral density in normal postmenopausal women—a key role for fat mass. J Clin Endoclinol Metab 1992;75:45–51. Reid IR, Evans MC, Ames R. Volumetric bone density of the lumbar spine is related to fat mass but not lean mass in normal postmenopausal women. Osteoporosis Int 1994;4:362–7. Bevier WC, Wiswell RA, Pyka G, Kozak KC, Newhall KM, Marcus R. Relationship of body composition, muscle strength, and aerobic capacity to bone mineral density in older men and women. J Bone Miner Res 1989;4:421–32. Segal KR, Gutin B, Presta E, Wang J, Itallie TBV. Estimation of human body composition by electrical impedance methods: a comparative study. J Appl Physiol 1985;58:1565–71. Lukaski HC, Bolonchuk WW, Hall CB, Siders WA. Validation of tetrapolar impedance method to assess human body composition. J Appl Physiol 1986;60:1327–32. Hoffer EC, Meadow CK, Simpson DC. Correlation of whole body impedance with total body water volume. J Appl Physiol 1969; 27:531–4.

19. Sakamoto Y, Satou T, Aikyou M. Evaluation of body fat by bioimpedance analysis [in Japanese]. Jpn J Obes Soc 1992;Suppl: 279–80. 20. Brozek J, Grande F, Anderson JT, Keys A. Densitometric analysis of body composition: revision of some quantitative assumptions. Ann N Y Acad Sci 1963;110:113–40. 21. Svendsen O, Hassager C, Skodt V, Christiansen C. Impact of soft tissue on in vivo accuracy of bone mineral measurements in the spine, hip, and forearm: a human cadaver study. J Bone Miner Res 1995;10:868–73. 22. Minaguchi H, Gorai I, Taga M, Nakayama M, Taguchi Y. Measurement of bone mineral density of radius by DTX-200 [in Japanese]. J Kanagawa Jpn Gynecol Soc 1995;32:12–6. 23. Fukunaga M, Tomomitsu T, Morita R. Bone mineral determination using dual X-ray absorptiometry (DEXA) [in Japanese]. Jpn J Clin Radiol 1990;35:41–8. 24. Nguyen TV, Kelly PJ, Sambrook PN, Gilbert C, Pocock NA, Eisman JA. Lifestyle factors and bone density in the elderly: implications for osteoporosis prevention. J Bone Miner Res 1994;9: 1339–46. 25. Broulik PD, Kapitola J. Interrelations between body weight, cigarette smoking and spine mineral density in osteoporotic Czech women. Endocr Regul 1993;27:57–60. 26. Latinen K, Vaimaki M. Alcohol and bone. Calcif Tissue Int 1991;49 Suppl:S70–3.