Search tips
Search criteria 


Logo of wtpaEurope PMCEurope PMC Funders GroupSubmit a Manuscript
J Bone Miner Res. Author manuscript; available in PMC 2010 May 24.
Published in final edited form as:
PMCID: PMC2875165



Fat mass may be a causal determinant of bone mass, but the evidence is conflicting, possibly reflecting the influence of confounding factors. The recent identification of common genetic variants related to obesity in children provides an opportunity to implement a Mendelian randomization study of obesity and bone outcomes, which is less subject to confounding and several biases than conventional approaches. Genotyping was retrieved for variants of two loci reliably associated with adiposity (the fat mass and obesity related gene FTO and that upstream of the MC4R locus) within 7470 children from the Avon Longitudinal Study of Parents and Children (ALSPAC) who have undergone total body DXA scans at mean 9.9 years. Relationships between both fat mass/genotypes and bone measures were assessed in efforts to determine evidence of causality between adiposity and bone mass. In conventional tests of association, both with and without height adjustment, total fat mass was strongly related to total body, spinal, upper and lower limb bone mineral content (BMC) (ratio of geometric means [RGM] 1.118 [1.112, 1.123], 1.110 [1.102, 1.119], 1.101 [1.093, 1.108], 1.146 [1.143, 1.155]; p=<10e-10 (adjusted for gender, height and sitting height)). Equivalent or larger effects were obtained from instrumental variable (IV) regression including the same covariates (1.139 [1.064, 1.220], 1.090 [1.010, 1.177], 1.142 [1.049, 1.243], 1.176 [1.099, 1.257]; p=0.0002, 0.03, 0.002, 2.3e-06 respectively). Similar results were obtained after adjusting for puberty, when trunkal fat mass was used in place of total fat, and when bone area was used instead of bone mass. In analyses where total body BMC adjusted for bone area (BA) was the outcome (reflecting volumetric bone density), linear regression with fat mass showed evidence for association (1.004 [1.002, 1.007], p=0.0001). IV regression also revealed a positive effect (1.031 [1.000, 1.062], p=0.05). When MC4R and FTO markers were used as instruments for fat mass, similar associations with BMC were seen to those with fat mass as measured by DXA. This suggests that fat mass is on the causal pathway for bone mass in children. In addition, both directly assessed and IV assessed relationships between fat mass and volumetric density showed evidence for positive effects supporting a hypothesis that fat effects may be important past simple co-association with overall bone size.


There is considerable interest in understanding the factors that influence skeletal development, since this may enable population-based interventions aimed at reducing the burden of osteoporotic fractures in later life by optimising bone mass accrual in childhood. Weight, body composition and particularly lean mass are amongst the strongest determinants of bone mass throughout life, largely reflecting adaptation of skeletal modelling to loading. However, whether fat mass affects skeletal development independently of lean mass remains controversial. For example, obese children have been reported to have a lower bone mass for a given weight in several previous studies (1-5). In contrast, in a study in which indices of proximal femur geometry were derived from hip DXA scans in overweight adolescents, fat mass was not found to influence any skeletal parameter independently of lean mass (6). In a study of 18 obese and 30 non-obese children, bone age in the former group was more advanced, but BMD was similar (7).

Cohort studies of the relationship between fat and bone mass in children and young adults have likewise yielded conflicting results. In 1068 men aged ~19 years, fat mass was positively correlated with tibial cross sectional area as assessed by pQCT, whereas a negative association was observed at the radius, suggesting that adipose tissue acts to stimulate growth of weight-bearing bones only (8). On the other hand, in the Avon Longitudinal Study of Parents and Children (ALSPAC), fat mass was positively related to bone mass as reflected by bone mineral content (BMC) in 3032 children at age ~10 years as measured at the total body, upper and lower limbs (9). In contrast, in a study based on 300 subjects aged 13-21, fat mass was not found to be related to either leg or lumbar spine bone mass or femoral or spinal cross sectional area as measured by pQCT (10). These apparently conflicting reports into the relationship between fat and bone mass in childhood are mirrored by findings in adults (11).

A possible explanation for these conflicting results is that the relationship between fat and bone mass is subject to confounding, which distinct studies may adjust for to differing degrees. Diet, physical activity, socio-economic factors, puberty, lean mass and illness are among many factors which may influence both fat and bone mass, and may therefore act as possible confounding factors. In terms of the true nature of any functional relationship between fat and bone mass in childhood, theoretically, fat might exert both positive and negative effects. For example, adipose tissue is known to express aromatase enzymes that convert steroid precursors to estrogen, which has variously been reported to stimulate (12) and suppress periosteal (13) bone growth in childhood. Furthermore, increased leptin levels secondary to higher fat mass have been suggested to mediate the negative association between fat mass and periosteal growth observed at non-weight bearing sites (8). Conversely, fat mass may stimulate bone growth via a direct mechanical action of increased load (14), by association with increased lean mass that occurs in obese subjects (14), by association with the secretion of bone active hormones from pancreatic beta cells (15), or by an indirect action on timing of pubertal events (16).

Mendelian randomisation, whereby genetic variation associated with the risk exposure of interest is used as a non-confounded proxy for that exposure (17,18), represents a potential approach for non-confounded assessment of the relationships which may exist between variables such as fat and bone mass in childhood. The application of this approach has been made possible by the recent discovery of common markers related to the FTO and MC4R genes, which can act as independent genetic associates of a small, but detectable, portion of the variance in human adiposity (19,20). In the present study, we aim to use a Mendelian randomisation approach to explore the relationship between fat and bone mass in childhood, by utilising FTO and MC4R markers as instrumental variable(s) (IV) for fat mass.


Study Population

The Avon Longitudinal Study of Parents and Children (ALSPAC) is a prospective study, which recruited pregnant women with expected delivery dates between April 1991 and December 1992 from Bristol, UK. The cohort is population-based and broadly representative at the point of recruitment (21). Individuals of known non-white ethnic origin were excluded from all analyses. DNA was collected from mothers and children as described previously (22). Of these births, 13988 were alive at 12 months. The present study is based on results for total body DXA scans obtained at a research clinic to which the whole cohort was invited at mean age of 9.8 years. Ethical approval was obtained from the ALSPAC Law & Ethics Committee, and local research ethics committees. Parental consent and child’s assent were obtained for all measurements made.

Measurement of Height, Weight and DXA-Derived Parameters

Of the 7725 children who attended the research clinic at age 9, 7470 agreed to undergo a whole body DXA scan, which was performed using a Lunar Prodigy with paediatric scanning software. At the same time, sitting and standing height were measured using a Harpenden Stadiometer, as was weight using a Tanita Body Fat Analyser and leg length. After exclusion of scans with anomalies (e.g. movement artefacts, artefacts caused by jewellery), complete scans were available for 7336 children. These were evaluated and re-analysed as necessary, to ensure that borders between adjacent sub-regions were optimally placed. Fat mass was expressed as total and trunkal fat mass. Bone variables comprised whole body minus head (hereafter referred to as ‘total body’), upper limb and lower limb BMC, bone area, and ‘areal’ BMD (obtained by dividing BMC by bone area). A more rigorous method for adjusting for skeletal size, area-adjusted BMC (ABMC), was also derived to provide a measure of ‘volumetric’ BMD, by using linear regression to adjust BMC for bone area. The coefficient of variation for total body BMD was 0.8%, based on analysis of results from 122 children who had two scans performed on the same day. DXA variables for the spine were derived from spine sub-regional analysis of whole body scans, excluding scans showing any spinal curvature, as previously described (23).


Genotyping of the FTO SNP rs9939609 was undertaken in 8,480 children (19). Genotyping was performed by KBiosciences (Hoddesdon, UK) using their own system of fluorescence-based competitive allele-specific PCR (KASPar). Details of assay design are available from the KBiosciences website ( Genotyping of the MC4R SNP rs17782313 was undertaken in the same manner within the ALSPAC cohort. For this variant 9024 individuals had recorded genotype information. Together with DXA data, for both rs9939609 and rs17782313, this yielded an approximate working sample size of n=5,300 individuals for total body measures, and 3,000 for spine values.

Other Variables

Gender was obtained from birth notifications. A puberty questionnaire was completed by the child’s carer (usually the child’s mother), which included questions on pubertal stage, tanner staging subsequently being based on pubic hair distribution (24). Maternal education was used as a proxy measure for social class as previously described (25).

Statistical Analysis

Relationships with the FTO and MC4R loci, including adjustments for sitting height, height and sex, were assessed via linear regression. In efforts to account for non-gaussian traits, log transformations were undertaken and results shown as geometric means and ratios of change between groups. Analyses assumed an additive genetic model (19) and included analysis of the variants rs9939609 (FTO) and rs17782313 (MC4R) with both basic anthropometric traits and measures of both fat mass and bone composition/area. In all basic analyses, twins and non-Europeans were excluded and covariates were centred for the generation of adjusted means by groups defined by either quantile of fat mass or genotype. Hardy Weinberg equilibrium was tested via exact test using the command “genhwi” (part of the “genassoc” suite of commands, All analyses were undertaken using the data analysis software STATA, version 10.

In a Mendelian Randomization framework, we used IV methods to obtain estimates of the association between fat mass and bone related continuous outcomes (26). We used two-stage least squares to fit the IV models in the main analyses (27). Within this model, we used both of the independent, fat mass associated loci (rs9939609 and rs17782313) to explain the largest, non-confounded and non-reverse caused, portion of variance in fat mass possible. Within these models, sex, sitting height and height were included in order to generate adjusted observational estimates of fat mass effect and equivalent estimates derived from IV analysis. This aimed to provide a comparison of best observational relationships between fat mass and bone measures and then to compare these against those known to be uninfluenced by confounding or reverse causation.

We compared the IV estimates to those from ordinary linear regression using the Durbin form of the Durbin-Wu-Hausman statistic, but checked results using the Hansen J statistics (limited information maximum likelihood) and the Anderson-Rubin statistics (generalized method of moments). We examined F-statistics from the first-stage regressions to evaluate the strength of the instruments. Values greater than ten are often taken to indicate sufficient strength to ensure the validity of IV methods (28,29).


Linear regression revealed an expected, positive association between total fat mass tertile, as measured by DXA, and total body, spinal, upper and lower limb BMC after adjustment for sitting height, height and sex (Tables 1A and 1B). For total body, spine, upper and lower limb measurements respectively, the ratios of geometric means (RGMs) between tertiles total fat mass were 1.118 [1.112, 1.123], 1.110 [1.102, 1.119], 1.101 [1.093, 1.108], 1.146 [1.143, 1.155]; p=<10-10. At all sites, fat mass had a similar association with bone area to that seen for BMC. These effects are similar to those reported previously in this cohort (9). In addition, fat mass was positively related to both spinal and total body ABMC, again after adjustment for sex, sitting height and height (RGM 1.004 [1.002. 1.007], p=0.0001 and 1.041 [1.035, 1047], p=<10-10).

Table 1A
Instrumental variable analysis of the effect of fat mass tertile on total body bone variables within the ALSPAC cohort (adjusted)
Table 1B
Instrumental variable analysis of the effect of fat mass tertile on spine, upper and lower limb bone variables within the ALSPAC cohort (adjusted)

Comparison of anthropometric and DXA data between those with and without genetic data showed no substantive differences between these groups in all bar fat mass data which show slight elevations of fat in those not genotyped (Table 2). Allele frequencies for the loci rs99439609 and rs17782313 (FTO and MC4R respectively) have been reported for this cohort elsewhere (19) and both loci were seen to adhere to Hardy Weinberg equilibrium within this population. As previously reported (19,30), variation at the FTO and MC4R was consistently associated with fat mass as assessed through a series of body region specific, DXA derived measurements (Table 3).

Table 2
Descriptive characteristics for key variables from the ALSPAC cohort
Table 3
Relationship between genotypes known to be associated with fat mass and DXA scan results in the ALSAC cohort

In the analysis of measures of stature, which are strongly associated with bone mass, both sitting height and height showed strong relationships with fat mass (RGM 1.023 [1.022, 1.024], p=<10e-10 and 1.024 [1.022, 1.025], p=<10-10). Similarly, lean mass was strongly related to fat mass (RGM 1.074 [1.070, 1.078], p=<10-10). Socio-economic status, reflected by level of maternal education, which we previously found to be related to bone mass (25), was also related to fat mass (p=0.003) (Table 4). In contrast to these results, the FTO and MC4R loci used in this study as proxies for fat mass measurement were un-related to any of these factors (Table 5), with the exception of weak associations between the FTO marker, sitting height and lean mass

Table 4
Relationship between fat mass and basic anthropometry in the ALSAC cohort
Table 5
Relationship between genotypes known to be associated with fat mass and basic anthropometry in the ALSAC cohort

Variation at rs9939609 (FTO) and rs17782313 (MC4R) was positively related to total body BMC (Table 6A), as found previously (30). These associations partly reflected genetic effects on bone size. For example, total body bone area showed a broadly similar relationship to total body BMC with variation at rs9939609 and rs17782313 (RGM rs9939609/bone area 1.008 [1.002, 1.013], p=0.006 and rs9939609/BMC 1.013 [1.005, 1.020], p=0.002; rs17782313/bone area 1.007 [1.000, 1.013], p=0.04 and rs17782313/BMC 1.011 [1.002, 1.020], p=0.01). Although there was some evidence that these genotypes were also related to total body ABMC (measure of ‘volumetric’ BMD), these associations were relatively weak (RGM rs9939609/ABMC 1.002 [1.000, 1.003], p=0.07 and rs17782313/ABMC 1.002 [1.000, 1.004], p=0.07).

Table 6A
Relationship between total body bone variables and FTO and MC4R genotypes within the ALSPAC cohort

Equivalent associations were observed with respect to regional BMC as measured at the spine, upper and lower limbs, with the exception that relationships between rs9939609 variation and spinal BMC, and rs17782313 variation and lower limb BMC, were relatively weak (RGM 1.011 [1.000, 1.0211], p=0.05 and 1.011 [1.001, 1.020], p=0.03 respectively) (Table 6B). For spinal ABMC, effects were stronger for rs17782313 than for rs9939609 (RGM 1.008 [1.003, 1.014], p=0.002 and 1.002 [0.998, 1.007], p=0.4 respectively).

Table 6B
Relationship between spine, upper and lower limb bone variables and FTO and MC4R genotypes within the ALSPAC cohort

In tests for association between FTO and MC4R genotypes and bone mass and other skeletal parameters, adjusting for total fat mass largely attenuated associations between these genotypes and total body, spinal, upper and lower limb BMC and bone area (Tables 7A and 7B). The relatively weak associations between FTO and MC4R genotypes and total body ABMC described above remained so after adjusting for fat mass (RGM rs9939609/ABMC 1.002 [1.000, 10.03], p=0.05 and rs17782313/ABMC 1.002 [1.000, 1.004], p=0.06).

Table 7A
Relationship between total body bone variables and FTO and MC4R genotypes adjusted for total fat mass within the ALSPAC cohort
Table 7B
Relationship between spine, upper and lower limb bone variables and FTO and MC4R genotypes adjusted for total fat mass within the ALSPAC cohort

IV regressions based on FTO and MC4R genotypes revealed similar associations with skeletal parameters to those found for measured fat mass. For total body, spinal, upper and lower limb BMC, effects of a tertile increase in total fat mass as predicted by IV regression, shown as RGM, were 1.139 [1.064, 1.220], 1.090 [1.010, 1.177], 1.142 [1.049, 1.243], 1.176 [1.099, 1.257]; p=0.0002, 0.03, 0.002, 2.3 × 10-6 respectively (Tables 1A and 1B). Overall, tests for difference between those estimates derived from adjusted observational associations and IV estimates did not yield consistent evidence for shifts in effect. If anything, IV estimates of fat mass effects on total and limb BMC, BMD and ABMC were equal to or larger than those derived from observational analysis, suggesting possible negative confounding or over-adjustment. The opposite was the case for spinal measurements. These differences were not statistically robust. Relationships between fat mass and bone parameters, and IV regressions, were also examined without height adjustment, as shown in Tables 8A and 8B.

Table 8A
Instrumental variable analysis of the effect of fat mass tertile on total body bone variables within the ALSPAC cohort (unadjusted)
Table 8B
Instrumental variable analysis of the effect of fat mass tertile on spine, upper and lower limb bone variables within the ALSPAC cohort (unadjusted)


Previous studies of the relationship between fat and bone mass in childhood have yielded conflicting results, possibly reflecting an influence of confounding factors. In the present investigation, we have confirmed that FTO and MC4R genotypes, which are known to be independent determinants of fat mass (19,30), are also related to bone mass (30). Moreover, the relationship between these genetic markers and bone mass appear to be directed through their effects on fat mass, since the associations between FTO and MC4R genotypes and bone mass were largely attenuated by adjustment for fat mass. In subsequent IV analyses, in which markers of common variation in these loci were used as instruments for fat mass, a similar positive relationship with bone mass was observed to that between fat mass as measured by DXA and bone mass. Mendelian randomisation approaches using IV analyses as described here has previously been used to make causal inferences in a range of epidemiological studies (18). However, a noteworthy aspect of the present investigation is that IV analyses were based on two independent genetic markers located on separate chromosomes.

Compared to measured fat mass, FTO and MC4R polymorphisms can be thought of as (and are shown to be) largely independent of factors which might confound the relationship with bone mass. As such, the present findings suggest that at least part of the observed variance in fat mass is likely to bear a causal relationship with accrual of bone mass in childhood. However, genetic determinants of obesity may still be related to other variables, such as those derived from the possible pleiotrophic action of FTO or MC4R. Although associations between fat and bone mass based on IV analyses are less likely to be explained by confounding factors, this possibility cannot be excluded completely. Nevertheless, to the extent that a causal relationship exists between fat and bone mass, results from IV analysis would appear to exclude reverse causality, whereby bone mass might influence fat mass, since it would seem unlikely that bone mass influences the risk of having genetic markers of obesity. This conclusion would appear to go against recent observations from animal studies suggesting that bone mass may be a determinant of fat mass, based on findings that osteoblast-derived osteocalcin stimulates adipocyte gene expression (31).

The suggestion from our results that a causal relationship exists between fat and bone mass in childhood is consistent with our previous observations in the ALSPAC cohort of a strong association between fat and bone mass, despite adjusting for potential confounding factors such as height and lean mass (9). Moreover, in longitudinal analyses, fat mass was found to predict gain in bone mass over the following two years (9). The magnitude of this fat mass effect, as confirmed by our Mendelian randomisation study, is large enough to have significant implications for public health. For example, there was approximately a one SD difference in BMD between children whose fat mass was in the lower versus upper tertile, which if translated to an adult population equates with a two-fold increase in fracture risk (32). Therefore, population-based strategies intended to reduce fat mass in childhood would appear to run the risk of increasing fracture risk in later life, unless these can be combined with strategies to counteract any negative effect on skeletal development, for example through increased exposure to weight bearing exercise.

Individually, FTO and MC4R polymorphisms also showed broadly similar associations with bone mass to that observed for fat and bone mass. There were nominal indications of fat mass difference between those with and without genotype data, but owing to the essentially random allocation of alleles with respect to this, any systematic effect this may have had will be uniformly distributed. We are therefore confident in our observations that, like fat mass, FTO and MC4R polymorphisms appear positively related to BMC at equivalent regions i.e. total body, spine, upper and lower limb.

As seen for fat mass, these genetic markers exerted their effects on bone mass largely by influencing bone size, as reflected by associations with bone area which were broadly similar to those with bone mass. Although neither FTO or MC4R polymorphisms were found to affect height in the present study, a small positive association between the MC4R locus studied here and height was reported in 90,000 individuals (30). Nevertheless, this association with height was considerably weaker than that with bone size, suggesting these genetic markers predominantly affect periosteal rather than longitudinal bone growth, leading to an increase in bone cross section rather than length, equivalent to effects of fat mass as reported previously (9). The fact that the present study provides further support for the hypothesis that fat mass is a positive influence on periosteal growth is significant, in light of the fact that bone cross sectional area is a major determinant of bone strength and hence fracture risk in later life (33).

Although size effects accounted for the major part of the relationship between fat and bone mass, both IV analyses and observational analysis revealed positive relationships with total body and spinal ABMC, suggesting that the positive influence of fat mass on bone mass also involves effects on volumetric bone density. A tendency for fat mass to enhance volumetric density could result from an increase in cortical/trabecular ratio secondary to suppression of endosteal expansion, or by a gain in trabecular bone mass. Further studies are planned to examine these possibilities, based on analysis of pQCT scan data which we are in the process of acquiring in this cohort.

Whilst the findings from IV analyses largely supported those of sex and anthropometrically adjusted observational analysis, there were trends which may offer insight into the interpretation of previous studies. For example, for total and limb based measures, effects of fat mass on BMC, bone area, BMD and ABMC appeared consistently lower in observational analyses. Although these differences were not robust, it seems likely that these may be the result of a combination of over-adjustment (i.e. with sex, sitting height and height in the model) and the greater ability of genotype effects to provide estimates of lifecourse fat mass effects.

Pleiotrophism, whereby genetic markers affect the outcome of interest independently of the exposure variable (in this instance fat mass), is one of several factors which can complicate interpretation of MR analyses (18). For example, MC4R is known to be a major regulator of cocaine- and amphetamine-regulated transcript (CART) in the hypothalamus, absence of which in mice leads to a low bone mass phenotype secondary to increased bone resorption (34). Moreover, children heterozygous for an MC4R loss of function mutation were found to have increased levels of CART, and reduced bone resorption, as compared with control subjects matched for weight and fat mass (35). To the extent that the rs17782313 rare allele is associated with reduced function of MC4R, alterations in bone resorption could conceivably contribute to the association between this marker and skeletal phenotypes reported here. Nevertheless, the finding that associations between the MC4R marker and skeletal measures were largely attenuated by adjusting for fat mass suggests that any direct effect of this marker on skeletal phenotype independently of fat mass is likely to be relatively weak.

In summary, we have used two independent genetic markers of obesity, related to the FTO and MC4R genes, as instrumental variables to explore the relationship between fat and bone mass in approximately 5000 nine-year-old children. We found that the relationship between these instrumental variables and bone mass mirrored that between bone mass and fat mass as measured by DXA, suggesting that fat mass is on the causal pathway for bone mass. This relationship between FTO and MC4R genetic markers, fat mass and bone mass largely involved effects on bone size, which we assume represents a stimulatory effect of fat mass on periosteal bone formation. In addition, FTO and MC4R polymorphisms and fat mass were both associated with total body and spinal volumetric bone density, suggesting that fat mass may also act to influence bone remodelling.


We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses. The UK Medical Research Council, the Wellcome Trust and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors who serve as guarantors for the contents of this paper.


1. Goulding A, Taylor RW, Jones IE, McAuley KA, Manning PJ, Williams SM. Overweight and obese children have low bone mass and area for their weight. Int J Obes Relat Metab Disord. 2000;24:627–632. [PubMed]
2. Lazcano-Ponce E, Tamayo J, Cruz-Valdez A, Diaz R, B H, Del Cueto R, Hernandez-Avila M. Peak bone mineral area density and determinants among females aged 9 to 24 years in Mexico. Osteoporos Int. 2003;14:539–547. [PubMed]
3. Mobley SL, Ha E, Landoll JD, Badenhop-Stevens NE, Clairmont A, HGoel P, Matkovic V. Children with bone fragility fractures have reduced bone mineral areal density at the forearm and hip and higher percent body fat. J Bone Miner Res. 2005;20:S34.
4. Skaggs DL, Loro ML, Pitukcheewanont P, Tolo V, Gilsanz V. Increased body weight and decreased radial cross-sectional dimensions in girls with forearm fractures. J Bone Miner Res. 2001;16:1337–1342. [PubMed]
5. Weiler HA, Janzen L, Green K, Grabowski J, Seshia MM, Yuen KC. Percent body fat and bone mass in healthy Canadian females 10 to 19 years of age. Bone. 2000;27:203–207. [PubMed]
6. Petit MA, Beck TJ, Shults J, Zemel BS, Foster BJ, Leonard MB. Proximal femur bone geometry is appropriately adapted to lean mass in overweight children and adolescents. Bone. 2005;36(3):568–76. [PubMed]
7. Klein KO, Larmore KA, de Lancey E, Brown JM, Considine RV, Hassink SG. Effect of obesity on estradiol level, and its relationship to leptin, bone maturation, and bone mineral density in children. J Clin Endocrinol Metab. 1998;83(10):3469–75. [PubMed]
8. Lorentzon M, Landin K, Mellstrom D, Ohlsson C. Leptin is a negative independent predictor of areal BMD and cortical bone size in young adult Swedish men. J Bone Miner Res. 2006;21(12):1871–8. [PubMed]
9. Clark EM, Ness AR, Tobias JH. Adipose tissue stimulate bone growth in prepubertal children. J Clin Endocrinol Metab. 2006;91:2534–2541. team As. [PMC free article] [PubMed]
10. Janicka A, Wren TA, Sanchez MM, Dorey F, Kim PS, Mittelman SD, Gilsanz V. Fat mass is not beneficial to bone in adolescents and young adults. J Clin Endocrinol Metab. 2007;92(1):143–7. [PubMed]
11. Zhao LJ, Jiang H, Papasian CJ, Maulik D, Drees B, Hamilton J, Deng HW. Correlation of obesity and osteoporosis: effect of fat mass on the determination of osteoporosis. J Bone Miner Res. 2008;23(1):17–29. [PubMed]
12. Bouillon R, Bex M, Vanderschueren D, Boonen S. Estrogens are essential for male pubertal periosteal bone expansion. J Clin Endocrinol Metab. 2004;89:6025–6029. [PubMed]
13. Lorentzon M, Swanson C, Andersson N, Mellstrom D, Ohlsson C. Free Testosterone Is a Positive, Whereas Free Estradiol Is a Negative, Predictor of Cortical Bone Size in Young Swedish Men:The GOOD Study. J Bone Miner Res. 2005;20:1334–41. [PubMed]
14. Turner CH, Forwood MR, Rho JY, Yoshikawa T. Mechanical loading thresholds for lamellar and woven bone formation. J Bone Miner Res. 1994;9(1):87–97. [PubMed]
15. Reid IR. Relationships between fat and bone. Osteoporos Int. 2008;19(5):595–606. [PubMed]
16. Fisch RC, Revelle R. Height and weight at menarche and a hypothesis of menarche. Arch Dis Child. 1971;46:695–701. [PMC free article] [PubMed]
17. Davey Smith G, Ebrahim S. ‘Mendelian randomization’: can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol. 2003;32(1):1–22. [PubMed]
18. Lawlor DA, Harbord RM, Sterne JA, Timpson N, Davey Smith G. Mendelian randomization: Using genes as instruments for making causal inferences in epidemiology. Stat Med. 2008;27(8):1133–63. [PubMed]
19. Frayling TM, Timpson NJ, Weedon MN, Zeggini E, Freathy RM, Lindgren CM, Perry JR, Elliott KS, Lango H, Rayner NW, Shields B, Harries LW, Barrett JC, Ellard S, Groves CJ, Knight B, Patch AM, Ness AR, Ebrahim S, Lawlor DA, Ring SM, Ben-Shlomo Y, Jarvelin MR, Sovio U, Bennett AJ, Melzer D, Ferrucci L, Loos RJ, Barroso I, Wareham NJ, Karpe F, Owen KR, Cardon LR, Walker M, Hitman GA, Palmer CN, Doney AS, Morris AD, Smith GD, Hattersley AT, McCarthy MI. A common variant in the FTO gene is associated with body mass index and predisposes to childhood and adult obesity. Science. 2007;316(5826):889–94. [PMC free article] [PubMed]
20. Loos RJ, Lindgren CM, Li S, Wheeler E, Zhao JH, Prokopenko I, Inouye M, Freathy RM, Attwood AP, Beckmann JS, Berndt SI, Jacobs KB, Chanock SJ, Hayes RB, Bergmann S, Bennett AJ, Bingham SA, Bochud M, Brown M, Cauchi S, Connell JM, Cooper C, Smith GD, Day I, Dina C, De S, Dermitzakis ET, Doney AS, Elliott KS, Elliott P, Evans DM, Sadaf Farooqi I, Froguel P, Ghori J, Groves CJ, Gwilliam R, Hadley D, Hall AS, Hattersley AT, Hebebrand J, Heid IM, Lamina C, Gieger C, Illig T, Meitinger T, Wichmann HE, Herrera B, Hinney A, Hunt SE, Jarvelin MR, Johnson T, Jolley JD, Karpe F, Keniry A, Khaw KT, Luben RN, Mangino M, Marchini J, McArdle WL, McGinnis R, Meyre D, Munroe PB, Morris AD, Ness AR, Neville MJ, Nica AC, Ong KK, O’Rahilly S, Owen KR, Palmer CN, Papadakis K, Potter S, Pouta A, Qi L, Randall JC, Rayner NW, Ring SM, Sandhu MS, Scherag A, Sims MA, Song K, Soranzo N, Speliotes EK, Syddall HE, Teichmann SA, Timpson NJ, Tobias JH, Uda M, Vogel CI, Wallace C, Waterworth DM, Weedon MN, Willer CJ, Wraight, Yuan X, Zeggini E, Hirschhorn JN, Strachan DP, Ouwehand WH, Caulfield MJ, et al. Common variants near MC4R are associated with fat mass, weight and risk of obesity. Nat Genet. 2008;40(6):768–75. [PMC free article] [PubMed]
21. Golding J, Pembrey M, Jones R. ALSPAC- The Avon Longitudinal Study of Parents and Children 1. Study methodology. Paediatric and Perinatal Epidemiology. 2001;15:74–87. [PubMed]
22. Jones RW, Ring S, Tyfield L, Hamvas R, Simmons H, Pembrey M, Golding J. A new human genetic resource: a DNA bank established as part of the Avon longitudinal study of pregnancy and childhood (ALSPAC) Eur J Hum Genet. 2000;8(9):653–60. [PubMed]
23. Tobias JH, Steer CD, Emmett PM, Tonkin R, Cooper C, Ness A. Bone mass is related to maternal diet in pregnancy. Osteoporosis Int. 2005;16:1731–1741. [PubMed]
24. Tanner JM. Normal growth and techniques of growth assessment. J Clin Endocrinol Metab. 1986;15:411–451. [PubMed]
25. Clark EM, Ness A, Tobias JH. Social position affects bone mass in childhood through opposing actions on height and weight. J Bone Miner Res. 2005;20:2082–2089. [PubMed]
26. Greenland S. An introduction to instrumental variables for epidemiologists. Int J Epidemiol. 2000;29:722–29. [PubMed]
27. Timpson NJ, Lawlor DA, Harbord RM, Gaunt TR, Day IN, Palmer LJ, Hattersley AT, Ebrahim S, Lowe GD, Rumley A, Davey Smith G. C-reactive protein and its role in metabolic syndrome: mendelian randomisation study. Lancet. 2005;366(9501):1954–9. [PubMed]
28. Staiger D, Stock JH. Instrumental variables regression with weak instruments. Econometrica: Journal of the Econometric society. 1997;65:557–586.
29. Stock JH, Wright JH, Yogo M. A survey of weak instruments and weak identification in generalized method of moments. J Business and Economic Statistics. 2002;20:518–529.
30. Loos RJ, Lindgren CM, Shengxu L. Association studies involvong over 90,000 samples demonstrate that common variants near to MC4R influence fat mass, weight and risk of obesity. Nature genetics. 2008 In Press.
31. Ferron M, Hinoi E, Karsenty G, Ducy P. Osteocalcin differentially regulates beta cell and adipocyte gene expression and affects the development of metabolic diseases in wild-type mice. Proc Natl Acad Sci U S A. 2008;105(13):5266–70. [PubMed]
32. Cummings SR, Black DM, Nevitt MC, Browner W, Cauley J, Ensrud K, Genant HK, Palermo L, Scott J, Vogt TM. Bone density at various sites for prediction of hip fractures. The Lancet. 1993;341:72–75. [PubMed]
33. Seeman E. Periosteal bone formation - a neglected determinant of bone strength. New Engl. J. Med. 2003;349:320–323. [PubMed]
34. Elefteriou F, Ahn JD, Takeda S, Starbuck M, Yang X, Liu X, Kondo H, Richards WG, Bannon TW, Noda M, Clement K, Vaisse C, Karsenty G. Leptin regulation of bone resorption by the sympathetic nervous system and CART. Nature. 2005;434(7032):514–20. [PubMed]
35. Ahn JD, Dubern B, Lubrano-Berthelier C, Clement K, Karsenty G. Cart overexpression is the only identifiable cause of high bone mass in melanocortin 4 receptor deficiency. Endocrinology. 2006;147(7):3196–202. [PubMed]