Associations Between Lean Mass, Muscle Strength and Power, and Skeletal Size, Density and Strength in Older Men

Abstract

Studies examining the relationship between muscle parameters and bone strength have not included multiple muscle measurements and/or both central and peripheral skeletal parameters. The purpose of this study was to explore the relationship between lean mass, muscle strength and power, and skeletal size, bone density, and bone strength. We studied the association between appendicular lean mass (AL M), grip strength, and leg power, and central quantitative computed tomography (QCT) parameters in 2857 men aged 65 years or older; peripheral QCT was available on a subset (n = 786). ALM, grip strength, and leg power were measured by dual-energy X-ray absorptiometry (DXA), Jamar dynamometer, and the Nottingham Power Rig, respectively. Multivariable models adjusting for potential confounders including age, race, study site, BMI, and muscle measurements were developed and least squares means were generated from linear regression models. For the multivariable model, percent differences of bone parameters between lowest (Q1) and highest quartiles (Q4) of ALM, grip strength, and leg power were reported. ALM was significantly associated with central and peripheral QCT parameters: percent higher values (Q4 versus Q1) ranging from 3.3% (cortical volumetric bone mineral density [vBMD] of the femoral neck) to 31% (vertebral strength index of the spine). Grip strength was only significantly associated with radial parameters: percent higher values (Q4 versus Q1) ranging from 2.5% (periosteal circumference) to 7.5% (33% axial strength index [SSIx]). Leg power was associated with vertebral strength and lower cross-sectional area with percent lower values (Q4 versus Q1) of –11.9% and –2.7%, respectively. In older men, stronger associations were observed for ALM compared to muscle strength and power. Longitudinal studies are needed to examine the relationship between independent changes in muscle measurements and skeletal size, density and strength.

Introduction

Muscles and bones are neighboring tissues with close ties. They are both derived from a common mesenchymal precursor.⁽¹⁾ Throughout life, the loss of muscle mass and bone mass has been shown to be coupled and has been hypothesized to be part of the same functional unit.(2) The positive association of muscle mass or strength with dual-energy X-ray absorptiometry (DXA) measures of areal bone mineral density (BMD) is well established.^(3–8) Nonetheless, areal BMD is an integrated measure of trabecular and cortical bone mass that does not provide compartment specific information on how bone mass is distributed. Because many non-osteoporotic individuals experience fractures, bone geometry not captured by DXA may also play a role in fracture risk.^(9,10) Previous studies have shown that peripheral and central quantitative computed tomography (QCT) measures of bone geometry are associated with fracture risk.^(11–13) According to the mechanostat theory, the physical attachment through tendons and muscle contractions leads to strain on bone, potentially affecting its geometry.⁽¹⁴⁾ Therefore, the purpose of this study is to examine the relationship between muscle measurements and central and peripheral QCT bone parameters.

Few human studies have connected whole-body DXA images to muscle function measurements and QCT scans to examine the association between skeletal muscle mass, muscle strength and power, and bone geometry and strength properties. Available literature is limited to animal studies, one or two muscle measurements, to specific skeletal sites, to younger individuals, to women only, or to a single technology; ie, DXA or QCT.^(15–18) Of importance, these studies adjusted for potential confounders but not for other muscle mass or muscle function measurements in their analyses. Hence, it is unknown if these associations are independent of each other.

In the current analysis, we sought to determine if each of the muscle mass, strength and power measurements are independently associated with measures of skeletal size, density, and strength. We hypothesized that older men with larger muscle mass, higher grip strength, and stronger leg power have greater skeletal size, density and strength.

Subjects and Methods

The Osteoporotic Fractures in Men (MrOS) study is a multicenter prospective cohort study designed to identify risk factors for osteoporosis and osteoporotic fractures. This study enrolled 5994 older men recruited from six sites (Birmingham, AL; Minneapolis, MN; Palo Alto, CA; Pittsburgh, PA; Portland, OR; and San Diego, CA) across the United States from March 2000 to April 2002.^(19,20) To be eligible, men needed to be age 65 years or older, be able to walk without assistance of another person, and to have reported no bilateral hip replacement. Human subjects’ approval was obtained at all sites with written informed consent obtained from all participants.

Central QCT analytical cohort

A subset of the first 650 men and all nonwhite men enrolled at each clinical site were referred for QCT scans of the hip and lumbar spine as part of their baseline visit for a total of 3786 men (63% of the baseline cohort). Of these participants, 588 were excluded because of incomplete central QCT measurements. From the remaining participants, 2857 had complete wholebody DXA, grip strength, and leg power measurements (Fig. 1). Except for a higher proportion of minorities (12.9% versus 10.5%), the characteristics of men in the QCT subset were similar to the overall sample of men in the cohort.

Fig. 1.

Flowchart of participants included in analyses.

Scanner models used at the sites were GE Prospeed (Birmingham, AL, USA), GE Hispeed Advantage (Minneapolis, MN, USA), Phillips MX-8000 (Palo Alto, CA, USA), Siemens Somatom +4 (Pittsburgh, PA, USA), Phillips CT-Twin (April–July, 2000, 190 participants, Portland, OR, USA), Toshiba Acquilion (December 2000–March 2002, 467 participants; Portland, OR, USA), and Picker PQ-5000 (San Diego, CA, USA). Calibration standards containing known hydroxyapatite concentrations were included with the participant in every scan.

The pelvic region from the femoral head to 3.5 cm below the lessertrochanter was scanned at settings of 80kVp, 280 mA, 3-mm slice thickness, and 512 × 512 matrix size in spiral reconstruction mode. At the femoral neck, the cross-sectional area (cm²) was computed as the area within the periosteal boundary at the minimum cross-section, the cortical thickness was measured 3 cm below the lesser trochanter, and the trabecular and cortical volumetric BMD (vBMD; g/cm³) were computed as the concentration of calcium hydroxyapatite averaged over all voxels in the trabecular and cortical volumes respectively.⁽²¹⁾

Lumbar spine scans were obtained from 5 mm above the L₁ superior endplate to 5 mm below the L₂ inferior endplate at settings of 120 kVp, 150 mA, 1-mm slice thickness, and 512 512 matrix size in spiral reconstruction mode. The lumbar spine region of interest was defined as the 10-mm slice in the midvertebra section for each vertebra. The spine parameters consisted of the total vertebral vBMD, cross-sectional area, and vertebral strength. Total vertebral vBMD was computed over all voxels within this region. The cross-sectional area (CSA) was determined by reconstructing a 10-mm-thick cross-section through the mid-vertebral plane. The CSA was taken as the periosteal area of the vertebra excluding the posterior elements and spinous processes. The vertebral strength index was derived as a linear combination of the vBMD and vertebral CSA from both the L₁ and L₂ vertebrae.^(22–24) DXA Hologic QDR 4500 (Bedford, MA) was used to measure areal femoral neck (FN) and spine BMD as described.⁽²⁵⁾

Peripheral QCT analytical cohort

Men with a second in-person exam an average of 4.7 ± 0.3 years later (from March 2005 to May 2006) were invited to participate in an ancillary study involving peripheral QCT (pQCT) at the Minneapolis and Pittsburgh clinical centers. A total of 657 men were deceased or had terminated before being contacted for the second visit, and 109 declined to participate. Of the 1550 men at the Pittsburgh and Minneapolis sites, 1172 (76%) completed the second clinic visit and agreed to participate in a pQCT ancillary study. Out of these participants, 88 had incomplete pQCT measurements, and 16 non-white men were excluded because the number is small to stratify by race. After excluding 282 participants with missing whole-body DXA, grip strength, and leg power, this analysis included 786 subjects (Fig. 1).

Slices were obtained (2.3 ± 0.2 mm) at the 4%, 33% and 66% sites of the left tibia and at the 33% site of the non-dominant forearm (radius) as described.⁽¹⁵⁾ Slices are taken as a percentage of limb length from the distal end of the relevant bone. The XCT 2000 and the XCT-3000 (Stratec Inc., Pforzheim, Germany) were used to obtain the scans in Pittsburgh and Minneapolis, respectively. For the Minneapolis site, precision with repositioning was determined as a coefficient of variation (CV, %) and varied from 0.28 (TotBMD) to 1.20 (TrabArea) at the distal tibia and from 0.31 (CortBMD) to 0.41 (TotArea) at the shaft.⁽¹⁵⁾ Similar precision values were reported at the Pittsburgh site. An anthropomorphic phantom was scanned daily for quality assurance at both sites. The same acquisition and analysis software (Stratec XCT version 5.5) was used to analyze scans at both sites. The details of our image analysis protocol (contour algorithms, filters, and density thresholds) were previously reported.^(13,15,24)

The pQCT parameters that we selected for this analysis were previously shown to be associated with non-spine fractures.⁽¹³⁾ We included parameters that estimate the bone strength in bending and torsion based on distribution of density-weighted bone voxels (axial strength index [SSIx], mm³),⁽²⁶⁾ the ability of the bone to resist torsion (polar moment of inertia [PMI], mm⁴),⁽¹³⁾ the cortical thickness (mm), the resistance of bone to bending and torsional force (cross sectional moment of inertia [CSMI], mm⁴),⁽²⁷⁾ and cross-sectional area (CSA, mm²).⁽²⁸⁾

Skeletal muscle mass, strength, and power measurements

Lean mass of the extremities and total body fat were obtained using the Hologic QDR 4500. Appendicular lean mass was calculated as the sum of lean mass in the arms and legs. Bone mineral content was removed from the lean mass calculation. Grip strength (kg) was measured twice using a Jamar dynamometer (Jackson, MI, USA) in both the right and left arms.⁽²⁹⁾ The maximum grip strength from all tries was used in our analysis. The Nottingham Power Rig was used to measure leg power extension in watts (W).^(30,31) Participants were seated with their arms crossed over their chest and instructed to push down on a pedal with one foot as hard and as fast as possible through a full range of motion. The maximum power output in watts from five trials was used.⁽³²⁾ All clinic staff performing DXA, leg power, and grip strength measures completed formal, centralized training. ALM, muscle power, and grip strength were used from visit 2 for the pQCT analysis.

Other measurements

Participants completed a questionnaire that collected information on demographics at baseline for QCT cohort and at visit 2 for pQCT cohort. Height was measured on a Harpenden stadiometer (Holtain Ltd., Crosswell, Wales, UK). Body weight was measured on balance beam scales (except for one site which used a digital scale). Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. Physical activity was assessed with the Physical Activity Scale for the Elderly (PASE) with higher scores indicating a greater level of activity.⁽³³⁾ Information on previous fractures was obtained.

Statistical analysis

The baseline characteristics of the central QCT cohort (years 2000–2002) and the visit 2 characteristics of the peripheral QCT cohort (years 2005–2006) were compared across appendicular lean mass quartiles using ANOVA for continuous variables and chi-square for categorical ones. Central and peripheral QCT parameters were compared across lean mass, muscle strength and muscle power quartiles (Tables 2 and 3).

Table 2.

Quartiles of Appendicular Lean Mass, Grip Strength, and Muscle Power by Central QCT Parameters: Age, Race, Weight, Height, Site Adjusted

	Appendicular lean mass (kg)
QCT parameters (unit)	Q1 20.04 ± 1.41 (n = 714)	Q2 22.99±0.60 (n = 714)	Q3 25.08 ± 0.64 (n = 714)	Q4 28.75 ± 2.21 (n = 715)	p trend
Femoral neck
Cross-sectional area (cm2)	12.09 ± 1.44	12.41 ± 1.53	12.60 ± 1.53	13.06 ± 1.50	<0.0001
Cortical vBMD (g/cm3)	0.519 ± 0.052	0.521 ± 0.052	0.530 ± 0.056	0.533 ± 0.064	0.0047
Trabecular vBMD (g/cm3)	0.077 ± 0.040	0.080 ± 0.042	0.085 ± 0.041	0.085 ± 0.041	0.0512
Areal BMD (g/cm2)	0.787 ± 0.108	0.798 ± 0.113	0.819 ± 0.113	0.825 ± 0.119	<0.0001
Spine
Total vertebral vBMD (g/    cm3)	0.111 ± 0.035	0.116 ± 0.032	0.122 ± 0.034	0.128 ± 0.035	<0.0001
Cross-sectional area (cm2)	11.63 ± 1.68	11.97 ± 1.60	12.05 ± 1.53	12.43 ± 1.78	<0.0001
Areal spine (g/cm2)	1.039 ± 0.183	1.063 ± 0.175	1.090 ± 0.178	1.117 ± 0.182	<0.0001
Vertebral strength index L1–    L2 (N)	0.27 ± 0.17	0.29 ± 0.16	0.32 ± 0.19	0.35 ± 0.20	<0.0001
	Grip strength (kg)
QCT parameters (unit)	Q1 30.41 ± 3.93 (n = 573)	Q2 38.27 ± 1.64 (n = 767)	Q3 43.85 ± 1.63 (n = 746)	Q4 52.65 ± 4.38 (n = 771)	p trend
Femoral neck
Cross-sectional area (cm2)	12.37 ± 1.48	12.44 ± 1.44	12.63 ± 1.64	12.74 ± 1.49	0.0001
Cortical vBMD (g/cm3)	0.525 ± 0.055	0.526 ± 0.055	0.525 ± 0.058	0.529 ± 0.057	0.6317
Trabecular vBMD (g/cm3)	0.079 ± 0.044	0.079 ± 0.039	0.082 ± 0.040	0.087 ± 0.041	0.0070
Areal BMD (g/cm2)	0.801 ± 0.117	0.800 ± 0.110	0.809 ± 0.111	0.822 ± 0.118	0.0045
Spine
Total vertebral vBMD    (g/cm3)	0.116 ± 0.036	0.119 ± 0.034	0.121 ± 0.033	0.123 ± 0.034	0.0215
Cross sectional area (cm2)	12.05 ± 1.75	11.93 ± 1.65	12.11 ± 1.67	12.07 ± 1.58	0.1679
Areal spine (g/cm2)	1.065 ± 0.208	1.065 ± 0.174	1.084 ± 0.173	1.096 ± 0.171	0.0073
Vertebral strength index L1–    L2 (N)	0.306 ± 0.219	0.296 ± 0.151	0.310 ± 0.161	0.320 ± 0.186	0.0726
	Leg power (W)
QCT parameters (unit)	Q1 130.3 ± 25.6 (n = 712)	Q2 183.9 ± 11.9 (n = 716)	Q3 225.3 ± 11.3 (n = 713)	Q4 289.3 ± 38.7 (n = 716)	p trend
Femoral neck
Cross-sectional area (cm2)	12.32 ± 1.57	12.52 ± 1.52	12.66 ± 1.47	12.75 ± 1.49	0.0001
Cortical vBMD (g/cm3)	0.525 ± 0.054	0.526 ± 0.058	0.529 ± 0.054	0.527 ± 0.060	0.5103
Trabecular vBMD (g/cm3)	0.080 ± 0.043	0.080 ± 0.041	0.083 ± 0.040	0.085 ± 0.040	0.1625
Areal BMD (g/cm2)	0.799 ± 0.115	0.803 ± 0.109	0.816 ± 0.118	0.817 ± 0.113	0.0219
Spine
Total vertebral vBMD (g/    cm3)	0.119 ± 0.036	0.118 ± 0.035	0.121 ± 0.034	0.122 ± 0.031	0.2192
Cross-sectional area (cm2)	12.08 ± 1.79	12.07 ± 1.72	12.02 ± 1.58	11.97 ± 1.54	0.6960
Areal spine (g/cm2)	1.079 ± 0.201	1.077 ± 0.175	1.082 ± 0.178	1.081 ± 0.166	0.9464
Vertebral strength index L1–    L2 (N)	0.320 ± 0.201	0.305 ± 0.180	0.312 ± 0.180	0.307 ± 0.150	0.5106

Statistically significant trend across quartiles with p < 0.006 adjusted for multiple comparisons.

vBMD = volumetric bone mineral density.

Table 3.

Quartiles of Appendicular Lean Mass, Grip Strength, and Muscle Power by Peripheral QCT Parameters: Age, Weight, Height, Site Adjusted

	Appendicular lean mass (kg)
QCT parameters	Q1 20.14 ± 1.27 (n = 197)	Q2 22.85±0.56 (n = 196	Q3 24.82 ± 0.62 (n = 196	Q4 28.52 ± 2.01 (n = 197	p-trend
FN aBMD Tibia 4%	0.776 ± 0.103	0.799 ± 0.128	0.791 ± 0.118	0.803 ± 0.126	0.3033
SSIx (mm3)	1177 ± 323	1283 ± 358	1297 ± 374	1394 ± 430	0.0023
Tibia 33%
PMI (mm4)	29357 ± 5565	32093 ± 5720	33333 ± 5815	37461 ± 6867	<0.0001
SSIx (mm3)	1183 ± 183	1261 ± 178	1276 ± 179	1382 ± 215	<0.0001
Cortical thickness   (mm)	5.35 ± 0.71	5.54 ± 0.71	5.63 ± 0.65	5.69 ± 0.71	0.0085
Tibia 66%
SSIx (mm3)	2142 ± 316	2296 ± 326	2333 ± 338	2528 ± 361	<0.0001
Cortical thickness   (mm)	3.85 ± 0.69	4.00 ± 0.71	4.01 ± 0.72	4.03 ± 0.79	0.2027
Radius 33%
Total CSA (mm2)	134 ± 16	142 ± 18	147 ± 18	155 ± 19	<0.0001
PC (mm)a	40.89 ± 2.43	42.12 ± 2.64	42.87 ± 2.56	44.04 ± 2.73	<0.0001
CSMI (mm4)	1136 ± 248	1276 ± 287	1349 ± 312	1503 ± 344	<0.0001
SSIx (mm3)	189 ± 32	206 ± 37	214 ± 37	230 ± 40	<0.0001
Cortical thickness   (mm)	3.18 ± 0.47	3.29 ± 0.53	3.33 ± 0.46	3.39 ± 0.51	0.035
	Grip strength (kg)
QCT parameters	Q1 31.11 ± 3.18 (n = 175)	Q2 38.17 ± 1.61 (n = 218)	Q3 42.88 ± 1.01 (n = 159)	Q4 50.56 ± 4.50 (n = 234)	p-trend
FN aBMD	0.803 ± 0.122	0.776 ± 0.112	0.802 ± 0.115	0.794 ± 0.126	0.0755
Tibia 4%
SSIx (mm3)	1286 ± 375	1245 ± 353	1301 ± 370	1321 ± 399	0.2102
Tibia 33%
PMI (mm4)	32916 ± 6399	32688 ± 5619	32414 ± 6311	34050 ± 6674	0.0703
SSIx (mm3)	1263 ± 207	1270 ± 174	1260 ± 194	1303 ± 201	0.1411
Cortical thickness   (mm)	5.47 ± 0.75	5.47 ± 0.73	5.60 ± 0.60	5.66 ± 0.68	0.0252
Tibia 66%
SSIx (mm3)	2273 ± 323	2306 ± 330	2301 ± 334	2404 ± 377	0.0045
Cortical thickness   (mm)	3.97 ± 0.77	3.96 ± 0.74	3.94 ± 0.72	4.00 ± 0.70	0.8968
Radius 33%
Total CSA (mm2)	140 ± 19	142 ± 17	144 ± 18	150 ± 18	<0.0001
PC (mm)a	41.83 ± 2.77	42.17 ± 2.52	42.40 ± 2.69	43.37 ± 2.63	<0.0001
CSMI (mm4)	1248 ± 310	1272 ± 282	1314 ± 326	1413 ± 310	<0.0001
SSIx (mm3)	200 ± 38	205 ± 35	211 ± 39	222 ± 37	<0.0001
Cortical thickness   (mm)	3.21 ± 0.50	3.25 ± 0.48	3.32 ± 0.53	3.40 ± 0.47	0.0033
	Leg power (W)
QCT parameters	Q1 108.4 ± 20.2 (n = 190)	Q2 152.7 ± 11.3 (n = 200)	Q3 192.7 ± 11.8 (n = 195)	Q4 254.8 ± 35.3 (n = 201)	p-trend
FN aBMD	0.789 ± 0.122	0.792 ± 0.120	0.797 ± 0.117	0.790 ± 0.119	0.9173
Tibia 4%
SSIx (mm3)	1262 ± 367	1276 ± 370	1292 ± 384	1323 ± 383	0.5651
Tibia 33%
PMI (mm4)	32517 ± 6572	32586 ± 6445	33405 ± 5375	33856 ± 6601	0.2058
SSIx (mm3)	1263 ± 203	1258 ± 197	1289 ± 175	1296 ± 200	0.2178
Cortical thickness   (mm)	5.42 ± 0.75	5.55 ± 0.71	5.58 ± 0.69	5.66 ± 0.62	0.0390
Tibia 66%
SSIx (mm3)	2265 ± 341	2308 ± 368	2351 ± 315	2381 ± 352	0.0301
Cortical thickness   (mm)	3.91 ± 0.79	3.98 ± 0.71	4.02 ± 0.72	3.98 ± 0.70	0.5767
Radius 33%
Total CSA (mm2)	141 ± 18	144 ± 18	145 ± 18	147 ± 20	0.0797
PC (mm)a	42.04 ± 2.71	42.48 ± 2.58	42.55 ± 2.58	42.91 ± 2.85	0.0685
CSMI (mm4)	1270 ± 302	1299 ± 294	1336 ± 306	1363 ± 335	0.0676
SSIx (mm3)	203 ± 37	209 ± 36	213 ± 37	215 ± 40	0.0410
Cortical thickness   (mm)	3.27 ± 0.54	3.27 ± 0.46	3.32 ± 0.51	3.34 ± 0.48	0.5241

Multiple regression analysis was used to determine the association between bone strength, geometry, and volumetric density across quartiles of ALM, grip strength, and leg power. All analyses were initially adjusted for age, weight, height, and study site, which constitutes the minimally adjusted model. For the central QCT cohort, the minimally adjusted model also included race because previous findings have shown race and ethnic variation in proximal femur structure.⁽³⁴⁾ The multivariable model additionally adjusted for grip strength and leg power (for ALM), ALM and leg power (for grip strength), and ALM and grip strength (for leg power). Collinearity was assessed using the variance inflation factor (VIF). No VIF exceeded five. Results were presented as least squares (LS) means and respective standard deviations. A trend test was performed to compare the LS means of the QCT parameters across muscle mass, grip strength, and leg power quartiles. For the multivariate model, percent differences between the lowest and highest quartiles (Q4 versus Q1) were reported. A Bonferroni adjustment was performed to account for multiple testing. Statistical significance was set at p < 0.006 and p < 0004 for central and peripheral cohorts, respectively. All statistical analyses were conducted using SAS version 9.3 (SAS Institute, Cary, NC, USA).

Results

Baseline characteristics

In the central QCT cohort, men with higher ALM were younger and more physically active compared to men with lower ALM. They also had higher BMI and total body fat. A greater percentage of men with high ALM had a history of fractures. Men with higher ALM also had a stronger grip strength and leg power (Table 1). The distribution and frequency of baseline characteristics were similar between the central QCT (Table 1) and pQCT samples. Baseline characteristics of central QCT men across muscle quartiles were similar to visit 2 characteristics of peripheral QCT men (results not shown).

Table 1.

Baseline Characteristics of Quantitative Computed Tomography Cohort Participants Across Appendicular Lean Mass Quartiles

Characteristics	Q1 (13.7–21.9 kg) (n = 714)	Q2 (21.9–24.0 kg) (n = 714)	Q3 (24.0–26.3 kg) (n = 714)	Q4 (26.3–37.9 kg) (n = 715)	p trend
Age (years), mean ± SD	76.24 ± 6.40	73.82 ± 5.53	72.46 ± 5.21	70.95 ± 4.70	<0.0001
White, n (%)	599 (83.89)	650 (91.04)	652 (91.32)	631 (88.25)	<0.0001
Height (m), mean ± SD	1.69 ± 0.06	1.73 ± 0.05	1.75 ± 0.06	1.79 ± 0.06	<0.0001
Weight (kg), mean ± SD	69.87 ± 7.37	78.45 ± 7.43	85.91 ± 7.93	95.70 ± 11.02	<0.0001
BMI (kg/m2), mean ± SD	24.54 ± 2.65	26.32 ± 2.72	27.99 ± 3.16	29.86 ± 3.73	<0.0001
Total body fat (kg), mean ± SD	17.61 ± 5.24	19.97 ± 5.89	22.74 ± 6.45	24.88 ± 7.45	<0.0001
PASE score (0–400), mean ± SD	141.09 ± 68.37	148.74 ± 66.41	153.37 ± 65.65	155.28 ± 68.39	<0.0001
Nottingham leg power (W),  mean ± SD	163.60 ± 45.67	196.17 ± 49.54	218.52 ± 54.33	250.76 ± 66.37	<0.0001
Appendicular lean mass (kg),  mean ± SD	20.04 ± 1.41	22.99 ± 0.60	25.08 ± 0.64	28.75 ± 2.21	<0.0001
Grip strength (kg), mean ± SD	36.57 ± 6.88	40.90 ± 7.14	43.35 ± 7.27	47.29 ± 8.73	<0.0001
Previous fracture, n (%)	350 (49.02)	394 (55.18)	395 (55.32)	407 (56.92)	0.015

Bold values are statistically significant at p < 0.05.

FN

In the minimally adjusted models, statistically significant positive associations in all QCT parameters (except trabecular vBMD) were observed across ALM quartiles (Table 2). In the multivariable model, percent differences (Q4 versus Q1) for statistically significant QCT parameters between the lowest and highest ALM quartiles ranged between 3.3% (cortical vBMD) and 6.7% (cross-sectional area). FN areal BMD was positively associated with ALM (Fig. 2). In the minimally adjusted models, FN areal BMD and cross-sectional area were positively associated with grip strength. Only cross-sectional area was positively associated with the muscle power quartiles (Table 2). In the multivariable model, these associations with muscle function were attenuated to non-significance (Fig. 2).

Fig. 2.

Percent difference in means of central and peripheral bone parameters between 1st and 4th quartiles of muscle mass and function. Statistically significant with p<0.006 (for central QCT parameters) and p<0.004 (for peripheral QCT parameters); adjusted for site, age, race, height, weight; grip strength and leg power (for ALM); ALM and leg power (for grip strength); and ALM and grip strength (for leg power). BMD = bone mineral density; CSA = cross-sectional area; CSMI = cross sectional moment of inertia; FN = femoral neck; PC = periosteal circumference; PMI = polar moment of inertia; SSIx = axial strength index; vBMD = volumetric BMD.

Lumbar spine

In the minimally adjusted model, statistically significant positive associations in all spine QCT parameters were observed across ALM quartiles (Table 2). In the multivariate model, percent differences for statistically significant QCT parameters between the lowest and highest ALM quartiles were 7.3% (cross-sectional area) and 31.0% (vertebral strength index L₁–L₂). Grip strength and leg power were not associated with spine QCT parameters in the minimally adjusted model. In the multivariable model, percent differences (Q4 versus Q1) between the lowest and highest leg power quartiles were –2.7% and –11.9% for CSA and vertebral strength, respectively (Fig. 2).

Radius

In the minimally adjusted model, statistically significant positive associations in radial QCT parameters were observed across quartiles of ALM (except cortical thickness) and grip strength. For muscle power, there were no statistically significant associations (Table 3). In the multivariable model, percent differences for statistically significant pQCT parameters between the lowest and highest quartiles ranged between 7.0% (periosteal circumference) and 28.6% (CSMI) for ALM, and between 2.5% (periosteal circumference) and 7.5% (SSIx) for grip strength. Leg power differences were attenuated to non-significance after multivariate adjustment radius pQCT parameters (Fig. 2).

Tibia

In the minimally adjusted model, statistically significant positive associations in all tibia QCT parameters were observed across quartiles of ALM (except cortical thickness at site 4%). There was no association with grip strength and leg power (Table 3). In the multivariable model, percent differences for statistically significant pQCT parameters between the lowest and highest quartiles ranged between 15.5% (SSIx at the 33% site) and 26% (PMI at the 33% site) for ALM. No significant association of grip strength and leg power existed for tibia pQCT parameters after multivariate adjustment (Fig. 2).

Discussion

Findings of this analysis showed that ALM was positively associated with central and peripheral QCT parameters. These associations were independent of grip strength and muscle power. The relationship between ALM and all QCT parameters was stronger compared to that of grip strength and leg power. This positive association remained robust even after adjusting for the other muscle measurements which suggests that appendicular lean mass may contribute more to the size, density, and strength of bones than direct strength and power measures of select muscle groups. This finding is consistent with a previous study that showed that skeletal mass relative to body size was associated with cortical and trabecular bone geometric parameters at multiple sites.⁽¹⁶⁾ In contrast, the mechanostat theory of Frost suggests that peak muscle forces, mainly through the tendon-bone junction, result in a positive skeletal response.⁽¹⁴⁾ Nonetheless, risk factors such as lack of exercise, depleted hormone levels, and malnutrition have been found to challenge the integrity of the mechanostat theory by shifting upward the physiological set points at which bone modeling and remodeling occur.⁽³⁵⁾

Muscles and bones have common embryological origins and experience an orchestrated organogenesis. Postnatally, load bearing bones in vertebrates adapt their strength to the mechanical loads of the muscles on them.⁽¹⁴⁾ With age, the loss of muscle strength is greater than that of muscle mass.⁽³⁶⁾ Thus, the strong association with ALM may suggest that even in the elderly, the associations built in younger age persist and determine the links between bone and muscle. Some of these life-long determinants include genetic factors, poor nutrition, lack of exercise, and sex steroids during growth and adulthood.⁽³⁷⁾ In men and women, estradiol is the main sex steroid responsible for bone resorption, and androgens affect the muscle mass as well as the trabecular bone formation.^(38,39) Our data suggests that crude adjustments for age, race, height, and body mass do not fully account for variations in body phenotype with respect to the close relationship between these tissues. Within the musculoskeletal system, common mechanotransduction, pleiotropy, and secretory/endocrine factors have been attributed to the coupled age-related deterioration in muscles and bones.⁽⁴⁰⁾ Greater muscle mass in older participants stretches collagen fibers and periosteum at the interface, resulting in the stimulation of osteocytes and local bone growth. Pleiotropic genes such as IGF-1 have been linked to maintaining bone mass and strength, as well as muscle hypertrophy.⁽⁴¹⁾ Furthermore, muscles and bones act as endocrine and target organs to each other. Myostatin, a muscle-derived factor that suppresses skeletal muscle growth, was shown to stimulate the differentiation of osteoclasts.⁽⁴²⁾ Osteocalcin, a bone-derived factor, was shown to maintain muscle mass in older mice.⁽⁴³⁾ These factors influencing muscle and bone in parallel may shed some light on why the older men with low ALM also had low bone size, density, and strength. On the other hand, measures of muscle strength and power have properties such as excitatory drive from supraspinal centers, motor unit coordination, recruitment, and discharge rates.^(44,45) These properties vary with pain, cognitive function, and age, and may not be directly linked to bone parameters.

At the FN, ALM was associated with most hip QCT parameters. In a previous study, the positive association between relative weight and proximal femur strength has been accounted for by lean mass.⁽⁴⁶⁾ In accordance with our findings, Lebrasseur and colleagues⁽¹⁶⁾ found statistically significant correlations between skeletal mass and the bone area, but not trabecular vBMD. In our cohort, the FN CSA and cortical vBMD were related to ALM. A 1-SD decrease in these parameters with and without aBMD adjustment has been previously linked to a higher risk of hip fracture.⁽¹¹⁾ At the spine, the ALM interquartile percent differences were the highest for vertebral strength. This is an important finding because in older men, vertebral strength is highly related to vertebral fracture risk in comparison to areal BMD.⁽¹²⁾ Therefore, the central QCT parameters may act as potential mediators between low skeletal mass and fractures. However, this association is yet to be established, independently of falls. Older MrOS men with larger skeletal mass had favorable radius and tibia bone parameters including strength and size. Previous studies have reported that muscle mass is associated with bone size,⁽¹⁸⁾ bone content, and estimated strength of the forearm.⁽⁴⁷⁾ In a younger group of men, lean mass was phenotypically and genetically correlated with radius and tibia density and size parameters,⁽⁴⁸⁾ suggesting that pleiotropy may partly explain the strong association between ALM and pQCT parameters in our cohort. To our knowledge, no studies have reported an association between ALM and tibia pQCT parameters in older men, making our findings novel. However, Frank and colleagues⁽⁴⁷⁾ have shown a site-specific association between lower leg muscle CSA and estimated strength and bone content of the tibia.

Grip strength was not associated with central QCT parameters. However, we found an association between grip strength and the geometry and strength parameters at the radius, but not at the tibia, suggesting site-specific effects. Muscles produce osteocyte viability factors including irisin and fibroblast growth factor (FGF)-2, especially upon muscle contraction.⁽⁴⁹⁾ After adjusting for variance in the lean mass that produces these factors, older men with weak grip strength still had low bone mass and poor structure at the radius, suggesting that local torques acting on and across the bones of forearm and hand are important for bone strength. There is a large body of evidence on the muscle-bone relationship that supports the association observed between grip and radius structural and geometric parameters.^(24,25) Concordantly, Kaji and colleagues⁽⁵⁰⁾ demonstrated in a Japanese population that grip strength is correlated with bone parameters of the forearm including the polar strength strain index and cortical thickness. The STRAMBO study also showed that low grip strength was associated with poor cortical and trabecular microarchitecture of the radius in older men.⁽¹⁸⁾ In a younger cohort of healthy adults, bone strength was significantly associated with isometric, concentric, and eccentric hand grip strength after adjusting for body size and gender; however, there was only a trend (p < 0.07) for compressive bone strength at the distal radius.⁽⁵¹⁾ Muscles of the forearm such as the flexors originate from the epicondyle of the humerus (proximal to radius) and insert at the wrist (distal to radius); their contractions exert bending moments across the radius and ulna. We utilized grip strength as a measure of overall muscle strength, and a high correlation with radius parameters was expected. True to the importance of local torques acting on bone, we did not observe similar percent differences between the lowest and highest grip strength quartiles for bone parameters at the hip, spine, and tibia.

No relationship was found between leg power and hip and tibia bone parameters. Very few muscles directly span the FN, with most muscles attaching more distally on the trochanter of the femur. It is possible that the Nottingham Power Rig does not adequately simulate habitual loading on the neck of the femur, particularly the static and locomotor loads at that site.^(52,53) These results differ from a previous MrOS paper in which we found positive associations between leg power and pQCTderived bone parameters (CSA and the polar strength index [SSIp] of the tibia).⁽¹⁵⁾ However, in that analysis the multivariable model did not include lean mass and grip strength as potential confounders. Examining the independent association of leg power is important because it is an influential determinant of physical performance,⁽⁵⁴⁾ which declines with age.

Negative associations were observed between leg power and spine parameters (CSA and vertebral strength). We did not anticipate negative associations between muscle power and vertebral parameters. We have previously reported significantly decreased Nottingham Power Rig performance in men with kyphosis of the spine in the MrOS cohort.⁽⁵⁵⁾ Kyphosis of the spine most commonly affects both the mid-thoracic and lumbar junction regions, and is highly associated with degenerative disc disease.⁽⁵⁶⁾ We postulate that the consequences of kyphosis are responsible for reducing leg power and concomitantly presenting the paradox of greater bone strength at the L₁–L₂ level in our analysis. Future research is needed to confirm this possibility.

There are several strengths to our study. MrOS is a multicenter prospective study examining potential risk factors for fractures in a large population of older men. We examined the association of lean mass, grip strength, and leg power with both central and peripheral QCT parameters. In our multivariable model, we adjusted for the other muscle parameters to test whether the associations were independent.

However, there are also several limitations to our study. Foremost, our analysis includes functional outcomes (grip strength, leg power) that are susceptible to the inhibitory effects of osteoarthritic pain. Arthritis of the hands and hip was not assessed at baseline, preventing us from adjusting our final results for this potential confounder. Peripheral QCT measurements were available for a smaller number of MrOS men, and the peripheral and central parameters cohorts were not the same. Another limitation is that the Nottingham Power Rig is limited as a measure of muscle power because it does not separate velocity from force. Therefore, we could not examine the distinct relationship between muscle power components and bone parameters. Also, because this is a cross-sectional study, we were unable to establish a temporal relationship between the muscle measurements and QCT parameters. Finally, our strength measures were limited to grip strength. Although grip strength is common and favored for its clinical utility and as a surrogate for whole-body strength, future studies could incorporate other clinically useful hand-held dynamometers for direct measures of lower extremity muscle strength.⁽⁵⁷⁾

In older men, ALM was positively associated with both central and peripheral QCT parameters, independently of grip strength and muscle power. Grip strength was associated with peripheral strength and geometric parameters of the radius that have been previously related to non-spine fractures.⁽¹³⁾ Compared to muscle power or strength, the stronger relationships observed for ALM were unexpected and contradict the mechanostat theory of Frost. Longitudinal studies are needed to examine the relationship between independent changes in muscle measurements and skeletal size, density, and strength.