JOURNAL OF BONE AND MINERAL RESEARCH Volume 5, Number 11, 1990 Mary Ann Liebert, Inc., Publishers
Sex Differences in Peak Adult Bone Mineral Density PAUL J. KELLY,' LANCE TWOMEY,' PHILIP N. SAMBROOK,' and JOHN A. EISMAN'
ABSTRACT Osteoporotic fractures are more common in women than men. Although accelerated bone loss following the menopause i s recognized as of major importance, it i s generally considered that a lower peak adult bone mass in females also contributes to their increased risk of osteoporosis in later life. To examine potential sex differences in peak adult bone mass we studied 29 pairs of dizygotic twins of differing within-pair sex in whom the female twin was premenopausal (mean age 37 years, range 21-55). Bone mineral density (BMD, g/cm2) was measured at the lumbar spine and femoral neck by dual-photon absorptiometry; 22 pairs also had BMD measured in the distal and 21 pairs in the ultradistal radius by single-photon absorptiometry. There was no significant difference in usual dietary calcium intake or tobacco consumption between the twin pairs. Consistent with accepted dogma, BMD at both radial sites were higher (+27%) in the males than their female cotwins. I n contrast, there was no sex difference (male versus female) in BMD (mean f SEM) in the femoral neck (0.96 f 0.02 versus 0.97 f 0.03), and surprisingly, the females had a greater lumbar spine BMD than their male cotwins (1.19 f 0.03 versus 1.26 f 0.03, p < 0.05). This difference was observed despite the fact that the males were taller ( p = 0.033). I f the femoral neck BMD values in the females were corrected for this difference in BMI, their values (0.99 f 0.03 g/cm2) were significantly higher than those in their male cotwin (p < 0.05). Alternatively i f the lumbar spine BMD in the female twins was corrected for the sex difference in anterior-posterior vertebral dimension (measured in cadaveric vertebrae) the female values were then 15% greater than those of their male cotwin. Therefore, a sex difference in peak adult bone density at the lumbar spine or femoral neck does not contribute to the higher incidence of osteoporotic fracture at these sites in aging women. Rather, the data suggest that before the menopause the female sex i s associated with a bone density greater than expected on the basis of muscle strength or body size or weight.
INTRODUCTION STEOPOROSIS is a common condition in postmenopausal women, affecting up to half of the elderly female population, but it affects a smaller number of elderly men.''' The effect of bone loss following the menopause contributes significantly to the sex difference in the incidence of osteoporosis, but it is unknown whether females have a lower peak bone mass that may also contribute to their risk of osteoporosis in later life. Studies that have specifically assessed a sex effect on bone density and have found greater bone density in males than females have
0
examined peripheral bone sites, such as the forearm." The possibility of a major sex effect on peak bone density at important axial bone sites has implications for the prevention of this disease. Bone density in adults is under strong and influence. Twin pairs of differing within-pair sex provide a useful model for examining the effect of sex on bone density as they are controlled in part for the genetic effect. We studied bone density at the hip, spine, and forearm in twins pairs of differing within-pair sex in whom the female twin was premenopausal to assess the effect of sex on peak bone density.
'Garvan Institute of Medical Research, St. Vincent's Hospital, Sydney, NSW 2010 Australia. 'School of Physiotherapy, Curtin University, F'erth, WA, Australia.
1169
KELLY ET AL.
1170
SUBJECTS AND METHODS A total of 29 pairs of twins of differing within-pair sex were included in the study. In each pair the female twin was premenopausal, as determined by the presence of regular menses. The subjects were obtained through the Australian National Health and Medical Research Council Twin Registry. The age of the subjects was 36.9 f 1.8 years (mean + SEM, range 21-55). Subjects with a previous history of bone disease, illness, or drug use that could affect bone mass were excluded from the study. All subjects included in the study had normal renal function as assessed by serum creatinine. Dietary calcium was estimated by questionnaire as described elsewhere.'') Grip strength was determined in 19 twin pairs on the nondominant arm, as used for single-photon absorptiometry, using a Haski grip strength meter (Wallace Instruments, Bayswater, Victoria, Australia). The maximal strength of three attempts was recorded as the forearm grip strength. Bone mineral content (BMC, grams) and BMC of the axial skeleton were measured using a Lunar DP3 dual-photon absorptiometer (Lunar Radiation, Madison, WI) as previously described."Ol The lumbar spine (L2-4) and three regions of the proximal femur (Ward's triangle, the trochanteric region, and the femoral neck) were measured. The coefficient of variation for BMD was 1.4% with cadaveric vertebrae and 2.6% in normal volunteers.['") The coefficients of variation for BMD measurements at the proximal femur in normal volunteers (aged 24-76, three measurements over 12 months, n = 18) are femoral neck, 3.8%; Ward's triangle, 4.1 %; and trochanteric region, 2.9%. Lumbar spine radiographs obtained in all subjects older than 40 years excluded the presence of fractures or severe osteoarthritis, which may spuriously elevate calculated lumbar BMD. BMD was also measured at the distal (21 pairs) and ultradistal radius (22 pairs) of the nondominant forearm by single-photon absorptiometry using a Lunar SP-2 densitometer (Lunar Radiation, Madison, WI). The precision of this method is 1.6% at both sites. Vertebral dimensions for normal subjects were determined by direct measurement of the first to fourth lumbar vertebrae in 48 cadavers (age range 20-39, 24 males and 24 females). The specimens were a subset from 204 specimens covering a wide age range obtained to assess movement and dimensions of lumbar vertebrae. The specimens were obtained from the public health mortuary of the Queen Elizabeth I 1 Medical Center in Western Australia, between the years 1976 and 1980. Care was taken to exclude from the study any subjects in whom there was evidence of spinal trauma, neoplasm, infection, or gross deformity. The vertebral height, width (lateral dimension), and depth (anterior-posterior dimension) were measured at the midvertebral level on all specimens by one operator (Twomey) using Mitotoyo dial vernier stainless steel calipers, accurate to 0.05 mm. The width of each vertebral body was measured at the waist of the vertebra before sagittal sectioning. The vertebral height and depth were measured on the vertebrae following midline sectioning. The diameter of the femoral neck for each subject was estimated from the femoral neck area as determined by
dual-photon absorptiometry. As the length of the sampling area along femoral neck BMD is constant (1.5 cm), the difference in femoral neck scanned area between subjects is due to the difference in the mean diameter of the femoral neck. Blood samples were collected in the early morning following an overnight fast. Osteocalcin was measured in serum from 28 twin pairs by an in-house radioimmunoassay using rabbit antisheep osteocalcin antisera with intraassay and interassay coefficients of variation of 8 and lo%, respectively. Urinary creatinine and hydroxyproline were measured on samples from 28 twin pairs collected over 2 h after an overnight fast to determine the hydroxyproline to creatinine (OHP/Crt, mmol/mmol) ratio."') Analysis of the difference between twin pairs was performed by two-sided paired Student's [-test using Statview 512' statistical package (Abacus Concepts, Inc.). Predictors of differences at any site were analyzed by backward stepwise multiple regression analysis, including height, weight, body mass index (BMI, kg/mz), and grip strength as independent variables. Daily alcohol consumption and tobacco use and data for serum osteocalcin and urinary OHP/Crt were not normally distributed; therefore, the Wilcoxon signed-rank test was used to analyze the difference in these parameters between twin pairs. The effect of body mass index on BMD at the lumbar spine and proximal femur were corrected for the sex difference in body size using partial regression coefficients previously published."") T o determine the best predictor of BMC at the lumbar spine and femoral neck we examined the relationship between body weight, height, and body mass index, and BMC at these sites in 1 1 1 normal premenopausal females recruited in epidemiologic studies of bone density. Body weight was the best independent predictor, and the regression equations were Lumbar spine BMC = 34.8 + 0.22 weight r = 0.36, p = 0.0001 Femoral neck BMC = 3.0 + 0.02 weight r = 0.36, p = 0.002
RESULTS The within-pair comparison of anthropometric features, calcium, tobacco and alcohol intake, and indices of bone turnover are given in Table I . There was no significant difference in dietary calcium intake or tobacco consumption, although the males consumed significantly more alcohol than their female cotwins. There were no sex differences in the indices of bone turnover. The osteocalcin level was slightly higher in the males than females, but this did not achieve significance and there was no difference in urinary OHP/Crt between males and females (Table 1). Grip strength was significantly greater in the males than the females (mean f SEM, 903.3 + 26.5 and 714.1 f 26.6 mm Hg, p = 0.OOOl). The BMD at both radial sites was significantly higher (+ 27%) in the males than the females (distal radius, 0.46 + 0.02 versus 0.36 f 0.01, p = O.OOO1; ultra0.01, p = distal radius, 0.42 + 0.02 versus 0.33
*
1171
SEX DIFFERENCES IN BONE DENSITY 0.0o0l). This sex difference was predicted at the distal site by the sex difference in weight ( r = 0.45, p < 0.05) and at the ultradistal site by the difference in BMI ( r = 0.52, p < 0.02). The lumbar vertebral dimensions of the cadaveric vertebrae are shown in Table 2. The male cadaveric lumbar vertebrae had a significantly greater width and depth than the female vertebrae, but there was no significant difference in lumbar vertebral height between the sexes. Bone mineral density (g/cmL, mean f SEM) at the lumbar spine was significantly greater in the females ( I . 19 f 0.03 versus 1.26 * 0.03, p < 0.05) compared to their male cotwins (Fig. 1). That the males did not have greater lumbar spine BMD than their female cotwin was surprising, as the male twins were significantly taller and heavier than their female cotwins and would have been expected to have higher BMD in the lumbar spine and femoral neck based on our and other previous The male twins BMI (range 18.5-35.1 kg/m') was comparable to the range of BMI (17.36-33.79 kg/m') of the normal females from
which the published BMI partial regression coefficients had been calculated.'"11We therefore calculated the female lumbar spine BMD values to correspond to the BMI of their male cotwin and found that the sex difference in lumbar spine BMD was further accentuated (corrected female lumbar spine BMD 1.29 g/cm', p < 0.003; Fig. I ) . A potential confounding factor in the analysis of sex differences in lumbar spine BMD is a sex difference in vertebral dimensions as shown in Table 1. We therefore compared BMC, that is, the mass of bone in the three vertebrae uncorrected for their dimensions, between the twin pairs. Bone mineral content (g, mean f SEM) tended to be greater at the lumbar spine (56.9 f 1.9 versus 52.6 f 1.6; Fig. 1) in the male cotwins, but this difference was not significant. BMD (i.e., bone mineral content corrected for vertebral area, BMC per cm') does correct for any differences in vertebral width and height, but it cannot account for any difference in vertebral depth. As shown in Table I , male vertebrae have significantly greater anterior-posterior dimensions (depth) than those of females. In an effort to
TABLE 1. COMPARISON OF INTRAPAIR DIFFERENCES IN ANTHROPOMETRIC FEATURES, DIE'TARY CALCIUM INTAKE, ALCOHOL, A N D TOBACCO CONSUMPTION, SERUMOSTEOCALCIN, AND URINARY HYDROXYPROLINE:CREATININE EXCRETION (URINARY OHP:CRT) BETWEEN TWINS OF DIFFERING WITHINPAIRSEX
Weight, kgb Height, cmb Body mass index, kg/mzb Dietary calcium, mg/dayb Alcohol, g/dayc Tobacco, pack-yearsa Osteocalcin, nglliterc Urinary 0HP:Crtc
P
Male
Female
77.7 f 2.4 176.8 f 1.3 24.9 f 0.7 819 80 10 (0-30) 6.8 (0-45) 1 1 .O (1.2-27) 0.012 (0.006-0.019)
62.0 f 1.7 164.6 f 1.4 23.0 f 0.7 663 f 68 1.25 (0-34) 0.4 (0-34) 8.9 (0.5-35) 0.01 1 (0.005-0.032)
o.oO01
0.0o01
< 0.05 0.13 0.02 0.12 0.13 0.98
aOnly twin pairs in which at least one twin was a tobacco smoker ( n = I5 pairs) were included in this analysis.
hMean f SEM, significance of difference determined by two-tailed paired I-test.
Sex Differences in Peak Adult Bone Mineral Density PAUL J. KELLY,' LANCE TWOMEY,' PHILIP N. SAMBROOK,' and JOHN A. EISMAN'
ABSTRACT Osteoporotic fractures are more common in women than men. Although accelerated bone loss following the menopause i s recognized as of major importance, it i s generally considered that a lower peak adult bone mass in females also contributes to their increased risk of osteoporosis in later life. To examine potential sex differences in peak adult bone mass we studied 29 pairs of dizygotic twins of differing within-pair sex in whom the female twin was premenopausal (mean age 37 years, range 21-55). Bone mineral density (BMD, g/cm2) was measured at the lumbar spine and femoral neck by dual-photon absorptiometry; 22 pairs also had BMD measured in the distal and 21 pairs in the ultradistal radius by single-photon absorptiometry. There was no significant difference in usual dietary calcium intake or tobacco consumption between the twin pairs. Consistent with accepted dogma, BMD at both radial sites were higher (+27%) in the males than their female cotwins. I n contrast, there was no sex difference (male versus female) in BMD (mean f SEM) in the femoral neck (0.96 f 0.02 versus 0.97 f 0.03), and surprisingly, the females had a greater lumbar spine BMD than their male cotwins (1.19 f 0.03 versus 1.26 f 0.03, p < 0.05). This difference was observed despite the fact that the males were taller ( p = 0.033). I f the femoral neck BMD values in the females were corrected for this difference in BMI, their values (0.99 f 0.03 g/cm2) were significantly higher than those in their male cotwin (p < 0.05). Alternatively i f the lumbar spine BMD in the female twins was corrected for the sex difference in anterior-posterior vertebral dimension (measured in cadaveric vertebrae) the female values were then 15% greater than those of their male cotwin. Therefore, a sex difference in peak adult bone density at the lumbar spine or femoral neck does not contribute to the higher incidence of osteoporotic fracture at these sites in aging women. Rather, the data suggest that before the menopause the female sex i s associated with a bone density greater than expected on the basis of muscle strength or body size or weight.
INTRODUCTION STEOPOROSIS is a common condition in postmenopausal women, affecting up to half of the elderly female population, but it affects a smaller number of elderly men.''' The effect of bone loss following the menopause contributes significantly to the sex difference in the incidence of osteoporosis, but it is unknown whether females have a lower peak bone mass that may also contribute to their risk of osteoporosis in later life. Studies that have specifically assessed a sex effect on bone density and have found greater bone density in males than females have
0
examined peripheral bone sites, such as the forearm." The possibility of a major sex effect on peak bone density at important axial bone sites has implications for the prevention of this disease. Bone density in adults is under strong and influence. Twin pairs of differing within-pair sex provide a useful model for examining the effect of sex on bone density as they are controlled in part for the genetic effect. We studied bone density at the hip, spine, and forearm in twins pairs of differing within-pair sex in whom the female twin was premenopausal to assess the effect of sex on peak bone density.
'Garvan Institute of Medical Research, St. Vincent's Hospital, Sydney, NSW 2010 Australia. 'School of Physiotherapy, Curtin University, F'erth, WA, Australia.
1169
KELLY ET AL.
1170
SUBJECTS AND METHODS A total of 29 pairs of twins of differing within-pair sex were included in the study. In each pair the female twin was premenopausal, as determined by the presence of regular menses. The subjects were obtained through the Australian National Health and Medical Research Council Twin Registry. The age of the subjects was 36.9 f 1.8 years (mean + SEM, range 21-55). Subjects with a previous history of bone disease, illness, or drug use that could affect bone mass were excluded from the study. All subjects included in the study had normal renal function as assessed by serum creatinine. Dietary calcium was estimated by questionnaire as described elsewhere.'') Grip strength was determined in 19 twin pairs on the nondominant arm, as used for single-photon absorptiometry, using a Haski grip strength meter (Wallace Instruments, Bayswater, Victoria, Australia). The maximal strength of three attempts was recorded as the forearm grip strength. Bone mineral content (BMC, grams) and BMC of the axial skeleton were measured using a Lunar DP3 dual-photon absorptiometer (Lunar Radiation, Madison, WI) as previously described."Ol The lumbar spine (L2-4) and three regions of the proximal femur (Ward's triangle, the trochanteric region, and the femoral neck) were measured. The coefficient of variation for BMD was 1.4% with cadaveric vertebrae and 2.6% in normal volunteers.['") The coefficients of variation for BMD measurements at the proximal femur in normal volunteers (aged 24-76, three measurements over 12 months, n = 18) are femoral neck, 3.8%; Ward's triangle, 4.1 %; and trochanteric region, 2.9%. Lumbar spine radiographs obtained in all subjects older than 40 years excluded the presence of fractures or severe osteoarthritis, which may spuriously elevate calculated lumbar BMD. BMD was also measured at the distal (21 pairs) and ultradistal radius (22 pairs) of the nondominant forearm by single-photon absorptiometry using a Lunar SP-2 densitometer (Lunar Radiation, Madison, WI). The precision of this method is 1.6% at both sites. Vertebral dimensions for normal subjects were determined by direct measurement of the first to fourth lumbar vertebrae in 48 cadavers (age range 20-39, 24 males and 24 females). The specimens were a subset from 204 specimens covering a wide age range obtained to assess movement and dimensions of lumbar vertebrae. The specimens were obtained from the public health mortuary of the Queen Elizabeth I 1 Medical Center in Western Australia, between the years 1976 and 1980. Care was taken to exclude from the study any subjects in whom there was evidence of spinal trauma, neoplasm, infection, or gross deformity. The vertebral height, width (lateral dimension), and depth (anterior-posterior dimension) were measured at the midvertebral level on all specimens by one operator (Twomey) using Mitotoyo dial vernier stainless steel calipers, accurate to 0.05 mm. The width of each vertebral body was measured at the waist of the vertebra before sagittal sectioning. The vertebral height and depth were measured on the vertebrae following midline sectioning. The diameter of the femoral neck for each subject was estimated from the femoral neck area as determined by
dual-photon absorptiometry. As the length of the sampling area along femoral neck BMD is constant (1.5 cm), the difference in femoral neck scanned area between subjects is due to the difference in the mean diameter of the femoral neck. Blood samples were collected in the early morning following an overnight fast. Osteocalcin was measured in serum from 28 twin pairs by an in-house radioimmunoassay using rabbit antisheep osteocalcin antisera with intraassay and interassay coefficients of variation of 8 and lo%, respectively. Urinary creatinine and hydroxyproline were measured on samples from 28 twin pairs collected over 2 h after an overnight fast to determine the hydroxyproline to creatinine (OHP/Crt, mmol/mmol) ratio."') Analysis of the difference between twin pairs was performed by two-sided paired Student's [-test using Statview 512' statistical package (Abacus Concepts, Inc.). Predictors of differences at any site were analyzed by backward stepwise multiple regression analysis, including height, weight, body mass index (BMI, kg/mz), and grip strength as independent variables. Daily alcohol consumption and tobacco use and data for serum osteocalcin and urinary OHP/Crt were not normally distributed; therefore, the Wilcoxon signed-rank test was used to analyze the difference in these parameters between twin pairs. The effect of body mass index on BMD at the lumbar spine and proximal femur were corrected for the sex difference in body size using partial regression coefficients previously published."") T o determine the best predictor of BMC at the lumbar spine and femoral neck we examined the relationship between body weight, height, and body mass index, and BMC at these sites in 1 1 1 normal premenopausal females recruited in epidemiologic studies of bone density. Body weight was the best independent predictor, and the regression equations were Lumbar spine BMC = 34.8 + 0.22 weight r = 0.36, p = 0.0001 Femoral neck BMC = 3.0 + 0.02 weight r = 0.36, p = 0.002
RESULTS The within-pair comparison of anthropometric features, calcium, tobacco and alcohol intake, and indices of bone turnover are given in Table I . There was no significant difference in dietary calcium intake or tobacco consumption, although the males consumed significantly more alcohol than their female cotwins. There were no sex differences in the indices of bone turnover. The osteocalcin level was slightly higher in the males than females, but this did not achieve significance and there was no difference in urinary OHP/Crt between males and females (Table 1). Grip strength was significantly greater in the males than the females (mean f SEM, 903.3 + 26.5 and 714.1 f 26.6 mm Hg, p = 0.OOOl). The BMD at both radial sites was significantly higher (+ 27%) in the males than the females (distal radius, 0.46 + 0.02 versus 0.36 f 0.01, p = O.OOO1; ultra0.01, p = distal radius, 0.42 + 0.02 versus 0.33
*
1171
SEX DIFFERENCES IN BONE DENSITY 0.0o0l). This sex difference was predicted at the distal site by the sex difference in weight ( r = 0.45, p < 0.05) and at the ultradistal site by the difference in BMI ( r = 0.52, p < 0.02). The lumbar vertebral dimensions of the cadaveric vertebrae are shown in Table 2. The male cadaveric lumbar vertebrae had a significantly greater width and depth than the female vertebrae, but there was no significant difference in lumbar vertebral height between the sexes. Bone mineral density (g/cmL, mean f SEM) at the lumbar spine was significantly greater in the females ( I . 19 f 0.03 versus 1.26 * 0.03, p < 0.05) compared to their male cotwins (Fig. 1). That the males did not have greater lumbar spine BMD than their female cotwin was surprising, as the male twins were significantly taller and heavier than their female cotwins and would have been expected to have higher BMD in the lumbar spine and femoral neck based on our and other previous The male twins BMI (range 18.5-35.1 kg/m') was comparable to the range of BMI (17.36-33.79 kg/m') of the normal females from
which the published BMI partial regression coefficients had been calculated.'"11We therefore calculated the female lumbar spine BMD values to correspond to the BMI of their male cotwin and found that the sex difference in lumbar spine BMD was further accentuated (corrected female lumbar spine BMD 1.29 g/cm', p < 0.003; Fig. I ) . A potential confounding factor in the analysis of sex differences in lumbar spine BMD is a sex difference in vertebral dimensions as shown in Table 1. We therefore compared BMC, that is, the mass of bone in the three vertebrae uncorrected for their dimensions, between the twin pairs. Bone mineral content (g, mean f SEM) tended to be greater at the lumbar spine (56.9 f 1.9 versus 52.6 f 1.6; Fig. 1) in the male cotwins, but this difference was not significant. BMD (i.e., bone mineral content corrected for vertebral area, BMC per cm') does correct for any differences in vertebral width and height, but it cannot account for any difference in vertebral depth. As shown in Table I , male vertebrae have significantly greater anterior-posterior dimensions (depth) than those of females. In an effort to
TABLE 1. COMPARISON OF INTRAPAIR DIFFERENCES IN ANTHROPOMETRIC FEATURES, DIE'TARY CALCIUM INTAKE, ALCOHOL, A N D TOBACCO CONSUMPTION, SERUMOSTEOCALCIN, AND URINARY HYDROXYPROLINE:CREATININE EXCRETION (URINARY OHP:CRT) BETWEEN TWINS OF DIFFERING WITHINPAIRSEX
Weight, kgb Height, cmb Body mass index, kg/mzb Dietary calcium, mg/dayb Alcohol, g/dayc Tobacco, pack-yearsa Osteocalcin, nglliterc Urinary 0HP:Crtc
P
Male
Female
77.7 f 2.4 176.8 f 1.3 24.9 f 0.7 819 80 10 (0-30) 6.8 (0-45) 1 1 .O (1.2-27) 0.012 (0.006-0.019)
62.0 f 1.7 164.6 f 1.4 23.0 f 0.7 663 f 68 1.25 (0-34) 0.4 (0-34) 8.9 (0.5-35) 0.01 1 (0.005-0.032)
o.oO01
0.0o01
< 0.05 0.13 0.02 0.12 0.13 0.98
aOnly twin pairs in which at least one twin was a tobacco smoker ( n = I5 pairs) were included in this analysis.
hMean f SEM, significance of difference determined by two-tailed paired I-test.
