Genetic Influences on Muscle Strength, Lean Body Mass, and Bone Mineral Density: A Twin Study


  • N. K. Arden,

    Corresponding author
    1. Twin Research Unit, St. Thomas' Hospital, London, United Kingdom
    • Address reprint requests to: Nigel Arden, Department of Rheumatology, Southampton General Hospital, Tremona Road, Southampton SO16 6YD, U.K.
    Search for more papers by this author
  • T. D. Spector

    1. Twin Research Unit, St. Thomas' Hospital, London, United Kingdom
    Search for more papers by this author

  • Presented at the 18th Annual meeting of the American Society of Bone and Mineral Reseach, September 1996, Seattle, WA, U.S.A.


Lean body mass and muscle strength are both associated with bone mineral density (BMD), which is known to be under strong genetic control. In this classical twin study, we examine the size of the genetic component of both muscle strength and lean body mass and to what degree they account for the genetic component of BMD. In all, 706 postmenopausal women were examined; 227 pairs of monozygous (MZ) twins and 126 pairs of dizygous (DZ) twins. Grip strength was measured using a hand-help grip bulb and leg strength using a dynamic leg extensor power rig. Lean body mass and BMD at multiple sites were measured by dual-energy X-ray absorptiometry. BMD correlated with both leg extensor strength (r = 0.16–0.26) and grip strength (r = 0.12–0.21). Lean mass was significantly correlated with BMD at all sites (r = 0.20–0.39). All three muscle variables have a moderate genetic component with heritability estimates of 0.52 for lean body mass, 0.46 for leg extensor strength, and 0.30 for grip strength (all p < 0.05). The genetic component of BMD was not significantly reduced after adjusting for lean mass and muscle strength, with less than 20% of the genetic variance of BMD explained by the muscle variables. In conclusion, these data suggest that the three muscle variables have a modest genetic component, suggesting the potential for clinical intervention and lifestyle modifications. The genetic component to muscle bulk and strength accounts for little of the genetic component to BMD, confirming the rationale for research into bone-specific genes.


TWIN STUDIES HAVE demonstrated a major genetic component to peak bone mass in both genders and also to postmenopausal bone mineral density (BMD) at all sites measured.1–4 Body weight has long been recognized to be strongly associated with BMD, but only relatively recently has the availability of dual-photon absorptiometry (DPA) and dual-energy X-ray absorptiometry (DXA) technology allowed us to simply and reliably discriminate fat from lean mass. Numerous studies have confirmed a strong association between lean body mass and BMD,4–7 although there is still some debate as to the relative importance of the association of fat mass and lean mass with BMD.8,9 Muscle strength measured at different sites has shown consistent associations with BMD both at sites local to the action of the muscle and at distant anatomical sites.10–12

Family studies have demonstrated a significant familial resemblance for lean mass and to a lesser extent fat mass, suggesting that they are under genetic control.7 These studies, however, did not directly quantitate the size of the genetic component, or heritability (h)2 of lean and fat mass. An Australian twin study of mixed pre- and postmenopausal women estimated the genetic component of lean body mass to explain 80% of its total variance.13 Studies in humans and animals have demonstrated a significant genetic component to muscle fiber composition14,15 but not to muscle enzyme activities.14 Grip strength has been shown to have a moderate genetic component in a previous study of elderly male twins, but information was lacking on other muscle or bone indices.16

The aims of the current study were first to examine the heritability of lean muscle mass and muscle power in normal postmenopausal women, and second to estimate the proportion of the genetic variance of BMD that was attributable to the muscle traits.



Three hundred and fifty-three pairs of female postmenopausal twins aged 45–70 years, 227 monozygous (MZ) and 126 dizygous (DZ), were recruited from a national media campaign asking for normal female twins to take part in a variety of research projects including those on bone and joint problems.

The zygosity of the twins was determined using a validated questionnaire and confirmed using multiplex DNA fingerprinting using variable tandem repeats.17 All subjects completed a nurse administered osteoporosis risk factor questionnaire, and those with serious medical illnesses affecting mobility or bone (two cases of cancer, one case of multiple sclerosis, and one case of morbid obesity) were excluded from the study.

Physical activity was assessed using a questionnaire based on the Allied Dunbar National Fitness Survey, which has been used in the Chingford population cohort to determine units of weight- bearing exercise. The questionnaire found a correlation with bone density levels.18 Height was measured using a standard wall stadiometer and weight using a digital scale.

BMD was measured using a Hologic QDR–2000 DXA scanner (Hologic Inc., Waltham, MA, U.S.A.). The sites measured included the lumbar spine (L1–L4), femoral neck, ultradistal radius, and whole body. Reproducibility was assessed by performing duplicate scans of 40 normal volunteers and elderly patients. The coefficients of variation (CVs) were between 0.6% and 1.6%, depending on the site scanned. The software also produced an estimate of total body lean muscle mass which has a CV of 0.8%.

Muscle strength was assessed by two techniques. Leg extensor power was measured on an unselected group of 160 MZ pairs and 90 DZ pairs of twins using a dynamic leg extensor power rig (NUMAS, Nottingham, U.K.). The subject sits, arms folded, on a rig with an adjustable, armless, low-backed seat and gives a maximal push to a large foot pedal. The pedal is connected through a chain to a flywheel. An opto-switch detects the flywheel speed, and a recorder displays the outcome in watts. The final (maximum) speed of the flywheel is used to calculate the average power throughout the push.19 The seat is adjusted on runners for each individual to allow the leg to reach maximum extension exactly at the end of the pedal travel, which is 0.165 m. The time taken for the push varies from 0.25–0.75 s. After at least one practice push, the subject performs three maximal pushes with strong verbal encouragement, and the best of the three pushes is recorded. This technique of assessing leg extensor power has been validated against jumping, stair climbing, and walking in both the young and elderly populations.20,21 The test was repeated in 19 subjects with at least 1 h between tests with a CV of 8.7%. Isometric grip strength was assessed using a standard grip bulb on the dominant arm. Reproducibility was assessed by repeated measurement on 24 individuals, giving a CV of 11.4%.

Statistical analysis

To compare the clinical characteristics of the two zygosities, the Chi square test was used for categorical variables, analysis of variance (ANOVA) for normally distributed variables, and the Mann–Whitney U-test for nonparametric variables. The classical twin study compares the within-pair similarities of MZ and DZ twins. The rationale for this design is that MZ twins share 100% of their genes and therefore differences between them are solely due to environmental factors, whereas DZ twins share on average only 50% of their genes and therefore differences between them are due to a combination of environmental and genetic factors. By comparing the within pair similarities of the two zygosities, one can obtain an estimate of the genetic component or heritability of a trait. Within-pair similarity was estimated by calculating intraclass correlations (r) for each zygosity. For MZ twins:

equation image

where AMZ is the among mean squares and WMZ is the within mean squares for the MZ twins.

The Falconer estimate of heritability (h)2 was calculated from the formula

equation image

This estimate relies on three assumptions: (1) that any genetic variance is additive, i.e., there are no gene interactions; (2) that the environmental covariance of MZ twins is equal to that of DZ twins; and (3) that total MZ and DZ variances are equal. Deviations from these assumptions can bias the estimate, for example if there is a gene dominance effect, if DZ variance is greater than MZ variance, or if MZ environmental covariance is greater than DZ then this will tend to exaggerate the heritability estimate. To adjust for the effect of height and weight, multiple linear regression modeling was performed, and intraclass correlation coefficients and heritability estimates were calculated as previously described using the residuals produced from the regression. To estimate the proportion of genetic variance of BMD attributable to the variance of the three muscle variables, the genetic variance was estimated using maximum likelihood estimation of parameters based on covariance matrices for the two zygosities before and after adjusting the BMD data lean body mass, grip strength, and leg extensor strength. Twin analysis was performed using the FORTRAN based TWINAN90 software package.23


Correlations of muscle parameters

The two zygosities were well matched for age, weight, years since menopause, hormone replacement therapy (HRT) duration, smoking, alcohol use, and physical activity. The DZ twins were, however, on average 1.2 cm taller than the MZ twins (Table 1). There were significant negative correlations (p < 0.01) of age with grip strength (r = −0.18) and also of leg extensor strength and lean mass (r = −0.26 and −0.09, respectively). Weight was significantly correlated with all three variables: r = 0.18, 0.13, and 0.58, respectively. Height was also significantly correlated with all three variables: r = 0.24, 0.22, and 0.57, respectively. Not smoking, regular alcohol consumption, the use of HRT, nor taking the oral contraceptive pill were significantly associated with any of the three variables.

original image

Lean mass correlated with grip strength and leg extensor strength: r = 0.33 (p < 0.01) and 0.30 (p < 0.01), respectively. There was a weak correlation between grip strength and leg extensor strength: r = 0.19 (p < 0.01). Lean mass correlated with BMD at all sites measured with coefficients of 0.20–0.39. There was a weak but consistent correlation between grip strength and BMD and a stronger association between leg extensor strength and BMD at all sites measured (Table 2). Together the three muscle variables explained between 8% and 18.5% of the variance of BMD depending on the site measured.

original image

Heritability of muscle parameters

The assumption of equal MZ and DZ variances was met, and, although not directly tested, there was no clear evidence from the rMZ:rDZ ratios of gene interactions. To assess the assumption of equal environmental covariance, the concordance of twins for current smoking, regular alcohol intake, HRT, oral contraceptive pill (OCP) usage, and the intrapair differences in usual physical activity were compared between zygosities. More MZ twin pairs than DZ twin pairs were concordant for smoking (Table 3), but there was no significant difference in concordance between zygosities for regular alcohol intake, OCP usage, or usual physical activity.

original image

In all analyses, the MZ correlations for the muscle bulk and strength (rMZ) were significantly greater than those for the DZ pairs (rDZ) (Table 4). The heritibility estimates obtained demonstrated a modest but significant genetic component to grip strength and larger genetic components to lean body mass and leg extensor strength. The heritability estimates were not significantly altered by adjusting for age, height, weight, or physical activity. For all sites of BMD, the rMZ was significantly greater than the rDZ (h2 = 0.57–0.88) both before and after adjusting for the muscle parameters. On adjusting for the muscle variables, the total additive genetic variance of BMD was reduced by 6.8% at the lumbar spine, 10.5% for the total body, 15.2% at the ultradistal radius, and 18.6% at the femoral neck. The environmental variance, however, was reduced by 24.2–50.0% (Table 5).

original image
original image


Our data confirm a significant correlation between lean body mass and BMD, which was consistent across the different sites of BMD measurements, with lean body mass explaining between 6% and 16% of the variance of BMD depending on the site measured. The correlation coefficients were of the same order of magnitude as those of other studies using either DXA or DPA to determine lean body mass.5,7 The correlations between measurements of muscle strength and BMD were fairly consistent across the different anatomical sites at which BMD was measured. Other studies have shown that muscle strength is often associated as strongly with BMD at sites anatomically distant from the site of action of the muscle as it is at sites close to the muscle group measured.6,10–12 This suggests that the association is not simply due to local biomechanical factors but to factors involved in the determination of general muscle strength. Whether these factors acting jointly on BMD and muscle strength are environmental (such as weight-bearing exercise), endocrine and growth factors, or genetic factors is still uncertain.

Genetic factors explain about half of the total variance of lean body mass in this cohort of female postmenopausal, Caucasian twins, which is consistent with previous family studies.7,24,25 These studies, which used DXA or underwater weighing to estimate lean body mass, have shown significant familial correlations for lean mass, suggesting a genetic component but without directly estimating its size. Our estimate is, however, smaller than that obtained by the authors of an Australian twin study, also using DXA, which estimated that the genetic component accounts for 80% of total variance.13 There were some differences between the studies which could explain this; their cohort of 112 pairs of twins were significantly younger than ours (mean age 45) and were highly selected from an initial cohort of 2000 twin pairs on the basis of discordancy in lifestyle habits, which were not adjusted for in the analysis. Lean body mass is strongly correlated with height and weight, which are both under strong genetic control,1,26 and it is possible that the genetic component to lean body mass may just reflect a genetic component to body size. However, adjusting for height and weight did not significantly diminish the heritability estimate for lean body mass, implying that there is an independent genetic component to leanness or muscle mass distinct from body size.

Leg extensor strength had a moderate genetic component and grip strength only a small genetic component. These estimates were not decreased by adjusting for height and weight, again suggesting that the genetic component is not solely explained by the genetic component to body size. Karlsson previously found a major genetic component to muscle fiber composition but no significant genetic component to muscle strength assessed by isometric knee extension, but a major genetic component to maximum muscle power assessed by stair running.27 This study was in young adults of both genders and only included 31 sets of twins, limiting its statistical power to detect all but a major genetic component to muscle strength. Muscle strength is determined by a combination of number and area of fibers, fiber tension, and percentage activation. Although muscle fiber composition apparently has a major genetic component, muscle fiber area is related to training at all ages,28 and muscle strength will therefore be defined by a combination of genetic and environmental factors. Our heritability estimate for grip strength of 0.30 is consistent with results of smaller twin studies in adult men and in children.16,29

We have confirmed a major genetic component to BMD at all sites measured as previously demonstrated in an earlier smaller cohort of twins from this study.1 Only 6.8– 18.6% of the genetic variance of BMD could be explained by the variance of the three muscle variables with a statistically significant independent component to BMD. This confirms the findings of an Australian study of elderly female twins who also found a significant genetic component to lumbar spine, hip, and total body BMD, but not to total forearm BMD after adjusting for lean body mass.4 The variance of the three muscle variables explained up to 50% of the environmental variance of BMD, suggesting the presence of important common environmental risk factors.

The only problematic assumption of the twin model was the greater environmental covariance for current smoking and any HRT use in MZ twins than in DZ twins, which could potentially have biased our estimate. However, because neither HRT usage nor smoking were associated with lean mass or muscle strength in our study, the heritabilty estimates are unlikely to be significantly inflated.

In conclusion, we have shown that lean muscle mass, leg extensor strength, and grip strength are all associated with BMD in normal Caucasian women. Muscle bulk and strength have a moderate genetic component explaining 30–50% of their total variance, leaving the majority of the variance to be explained by environmental factors. This indicates a major potential for lifestyle modifications and other clinical interventions such as physical exercise regimes in modifying muscle mass and strength and ultimately BMD. Nevertheless, there is a major genetic component to BMD at all sites which is independent of lean mass and muscle strength, confirming the importance of focusing on bone-specific genes in the search for genes responsible for osteoporosis.


We would like to thank the staff of the twin research unit for help in the running of the study and especially the twins themselves. This study was partly funded by grants from the Wellcome Trust, The Arthritis and Rheumatism Council, The Chronic Disease Research Foundation, the National Osteoporosis Society of the U.K., and Gemini Research Ltd.