Because it affects the material and structural properties of bone, physical activity is an important strategy to optimize bone strength during the growing years (1). Moreover, physical activity can improve skeletal muscle mass (2), which is a strong determinant of bone strength (3). Improvements in bone properties associated with physical activity and skeletal muscle are due to mechanical loading generated from gravitational load and muscular contractions (4,5). This mechanism is described in the Frost’s mechanostat theory. It proposes that bone strains, caused by mechanical loading that exceed a modeling threshold range, stimulates bone to increase its strength by changing its mass, density, and structure (6,7). In addition, muscle may also regulate bone modeling via paracrine and endocrine signaling pathways (8).
The variation of lean mass (a surrogate of muscle mass) in boys during puberty seems to explain a significant percentage (~45%) of the bone mass accrual at the proximal femur (PF) in this period (9). Compared with other femoral regions, the association between lean mass and bone mass (expressed as bone mineral content, [BMC] or bone mineral density [aBMD]) does not seem to be so obvious (9) or may not even exist at the femoral neck (10). At this clinically significant region, there is also evidence that lean mass (or fat-free mass) may more determinant for bone mass differences among girls and that physical activity may be more important in boys accounting between 10% and 20% for these variations (11,12).
Despite the relevant role that muscle plays in the skeletal response to mechanical loading during physical activity, most of the studies analyzing the associations/effects of physical activity on bone outcomes have not considered muscle and bone together as a functional unit. Too often, a single analysis of each of these factors (i.e., physical activity or muscle) is examined. In fact, a recent systematic review on youth, physical activity, and bone strength reported that only 46% of the reviewed studies considered the contribution of muscle to bone outcomes within studies using physical activity as the exposure (13).
Much remains to be learned concerning the indirect effect (through muscle) of physical activity on bone outcomes. Very few studies examined associations between physical activity and bone health across childhood and adolescence using multiple data points of objective measure of physical activity and bone outcomes (14,15), and to the best of our knowledge, there are no studies that have analyzed the effect of muscle on this relationship, using mediation analysis. The mediation analysis allows quantifying the indirect effect of exposure on a given outcome through a specified intermediate variable providing a better understanding of the mechanism or process by which one variable influences another variable (16). Thus, the main purpose of this study was to analyze the indirect (muscle-mediated) effects of distinct trajectories of physical activity across childhood and adolescence on late adolescent bone parameters. More specifically, we conceptualized physical activity as a behavioral process where levels of physical activity were likely to change across the growing years; we captured this concept using group-based trajectory modeling which enables the data themselves to divide a study population into subgroups and to fit trajectory models for each of those subgroups. We then examined the associations of the physical activity trajectories at age 17 yr 1) aBMD of the total PF and its regions (trochanter [TR], intertrochanter [IT], and neck [N], including its inferomedial [IM] and the superolateral [SL] subregions), 2) aBMD distribution, and 3) specific PF geometric measures. Mediation analysis was used to explore the underlying mechanism of how physical activity indirectly influences bone parameters via muscle.
METHODS
Sample design and data collection
The Iowa Bone Development Study (IBDS) is an ongoing longitudinal study of bone health during childhood, adolescence, and young adulthood. Participants are a subset of the Iowa Fluoride Study birth cohort; 1882 families from eight Iowa hospital postpartum wards who were recruited between 1993 and 1997 (17). The IBDS cohort was recruited between 1998 and 2002 when Iowa Fluoride Study participants were approximately 5 yr of (child) age. The IBDS continues to use a rolling admission to allow Iowa Fluoride Study members to participate in any follow-up examinations. This report focused on data collected from 1998 to 2013. We did not exclude participants based on medical conditions, mostly we have healthy participants. The following conditions that could be related to bone development were reported for analysis sample in concurrent health questionnaire: musculoskeletal diseases (scoliosis, 2; kyphosis, 1), diabetes (3), and arthritis (2). Current asthma was reported by a relatively high number of participants (22). Adjusting for body size and sex, PF bone outcomes were slightly lower for participants with current asthma, but differences were not statistically significant. The study was approved by the University of Iowa Institutional Review Board (Human Subjects).
At approximately ages 5, 8, 11, 13, 15, and 17 yr, accelerometer measures of physical activity were obtained, and a clinical examination that included anthropometry and dual-energy X-ray absorptiometry (DXA) imaging was conducted. Each measurement wave was conducted over a 3-yr period, and the age standard deviation for each wave was ~0.4 yr resulting in 4 and 6 yr olds within the 5-yr-old wave, 7 and 9 yr olds within the 8-yr-old wave, and so on. Because the anatomical landmarks for a proper identification of the specific regions of the PF (such as the greater and lesser TR) are fully formed by late adolescence (18), only age 17 yr was used to derive bone parameters from DXA.
Accelerometry
At each clinical examination, participants and their parents were given instructions on accelerometer wear. ActGraph accelerometers (Pensacola, FL) were mailed to participants during the autumn season (September to November). Due to availability, model 7164 was used for ages 5, 8, 11, 13 yr; GT1M for age 15 yr; and GT3X for age 17 yr. At age 5 and 8 yr, participants were asked to wear the monitor during all waking hours for four consecutive days, including one weekend day. At the other examination ages, they were asked to wear for five consecutive days, including both weekend days. Accelerometry movement counts were collected in a 1-min epoch at ages 5, 8, 11, and 13 yr. Accelerometry data at age 15 yr were collected in a 5-s epoch, and raw acceleration data were collected at age 17 yr. Age 15 yr and age 17 yr accelerometry data were reintegrated to 1-min epochs.
Accelerometers were considered as having not been worn if a period of 60 consecutive minutes of zero accelerometry counts (with allowance for two nonzero interruptions) was encountered in the accelerometry data array. Accelerometry data were only used from participants who wore an accelerometer for a minimum of 10 h·d−1 and 3 d at each examination. Moderate- and vigorous-intensity physical activity (MVPA) was defined as 2296 or greater accelerometry counts per minute (19,20).
Dual-energy X-ray absorptiometry
At age 17 yr, DXA measures were conducted using the Hologic QDR 4500A DXA (Delphi upgrade) with software version 12.3, in the fan-beam mode. A spine phantom was scanned daily to maintain quality assurance. To minimize operator-related variability, all measurements were conducted by one of two experienced technicians. The precision error for BMC measurements was low in our laboratory (coefficient of variation of <1% for quality control scans performed daily using the Hologic phantom).
aBMD and bone mass distribution
aBMD measures used in this study included aBMD of the total PF and its regions: IT, TR, and N. The analysis of the scans was performed by software-specific Global regions of interest (ROI), which designate the general boundaries of the hip images. A review of the bone within the ROI box was confirmed by the operator and edited to ensure appropriate bone-edge detection. A specific femoral neck analysis was also performed to differentiate the aBMD of the N between the SL and IM subregions. This approach was published in a previous work (21). The aBMD ratios among PF regions were used as surrogates of bone mass distribution and calculated as follows:
;)
This approach provides measures of intra-individual relative difference of mineralization at the PF regions (21).
PF geometry
Linear geometric measures were derived from hip DXA images using the Hip Structure Analysis program (Hologic Apex 3.0 software). The geometric measures used in this study included the hip axis length (HAL, cm), the narrow neck width (NNW, cm), and the neck-shaft angle (NSA, degree). The HAL is the linear distance from the base of the greater TR to the edge of the pelvic inlet. The NNW is the narrowest distance across the femoral neck. The NSA is the angle between derived axes of neck and shaft. The HAL is calculated by the system from the lateral bone map edge of the greater TR to the edge of the pelvic inlet or to the Global ROI edge, whichever point is intersected first. Therefore, a strict protocol was followed for positioning the Global ROI edge in the same point of the edge of the pelvic inlet.
Leg lean soft tissue
Leg lean soft tissue (LST) (kg) was estimated from a total body scan by DXA and used as a surrogate of leg muscle mass. National Health and Nutrition Examination Survey body composition method was used to analyze age 17 yr whole body scans obtained on Hologic QDR 4500 DXA (Delphi upgrade with software version 13.4).
Other measures
Research nurses trained in anthropometry measured the participant’s height (cm) and weight (kg). Participant’s height (cm) and weight (kg) were measured by a Healthometer physician’s scale (Continental, Bridgeview, IL) and Harpenden stadiometer (Holtain, UK). Both devices were routinely checked for accuracy and precision. Children were measured while wearing indoor clothes but without shoes. Leg length (cm) was calculated as standing height minus sitting height. Peak height velocity (PHV) age was estimated using predictive equations established by Mirwald et al. (22) where height, weight, age, sex, sitting height, leg length, and weight by height ratio were predictors of somatic maturity. Because we had estimates at age 11, 13, and 15 yr, as advised by Malina and colleagues (23), the PHV age estimates closest to a chronological age when measurements were done was used as the best estimate.
Statistics analysis
Data were analyzed using SPSS statistical software package (version 24.0 for Windows; SPSS, Chicago, IL). Distribution properties were examined using the Kolmogorov-Smirnov test. Differences between groups (females and males) for age, PHV age, body size, and composition, aBMD, aBMD distribution, and specific geometric measures of PF were analyzed by independent-samples t test when normally distributed with equal variance and by independent-samples t test with Welch correction when there was normality, but variances were not equal. Mann–Whitney nonparametric test was used when there was no normality.
To characterize patterns of physical activity over the study period, each participant’s MVPA (approximate ages: 5, 8, 11, 13, 15, and 17 yr) was grouped within a pattern of conditional probabilities based on structural equation modeling theory that assumes individuals differ qualitatively as members of homogeneous (latent) subgroups (24). Individual-specific probabilities of belonging to each subgroup allowed assignment to a subgroup based on the highest probability. The relationship between MVPA and age was fitted up to a third-degree polynomial model that included the latent subgroup variable. All available data for physical activity were used. Under the assumption that data were missing at random, individuals with incomplete data were included with the requirements of at least three measurements per participant over the study period. The best fitting polynomial model was determined by comparing the Bayesian Information Criterion for models with different numbers of subgroups. SAS procedure TRAJ developed by Jones and Nagin was used for analysis (24).
ANCOVA were used for comparison of bone variables (aBMD, aBMD distribution, and specific geometric measures) between groups defined by MVPA trajectories for females and males. To reduce the confounding effects of body size and maturity on bone variables, the lower limb length, body weight, and PHV age were included as covariates. Normality and homoscedasticity of the residuals were checked, and there were no severe departures from these assumptions.
To examine whether leg LST mediated the relationship between MVPA trajectories and bone variables for females and males, a model for each bone outcome included MVPA trajectory group as an independent variable, and with leg LST as mediator. Lower-limb length and body weight were added as covariates. Figure 1 depicts the overall mediation model used in the analysis. Mediation analyses were performed according to the MacKinnon et al. approach (25), which advises that mediation can occur even with a lack of a significant relationship between the independent (X) and the dependent variable (Y). Briefly, these authors postulate that statistical test of the overall relation between X and Y can have less power than the test of the links in the mediation model in several circumstances. Before testing mediation, the path a (the effect of independent variable on the mediator) and path b (the effect of mediator on outcomes partialling out the effect of the independent variable) were quantified with unstandardized regression coefficients. Indirect effects through the mediator were analyzed only when both paths were significant. SPSS macro developed by Preacher and Hayes were used to test mediation (16). This tool tests the significance of total and specific indirect effects using the product-of-coefficients approach or by bootstrap procedures. Bootstrap procedures are favored since they do not require assumptions of normality of the sampling, improve power and reduce type I error rates when compared with the product-of-coefficients approach (16,26). We report results for both, product-of-coefficients approach and bootstrap procedures, with a resampling of 5000 bootstrap samples (95% confidence intervals [95% CI]). Finally, the mediated effects were described by calculating the effect ratios which express the amount of the total effect that is explained by the indirect effect via the mediator (26).
FIGURE 1: Overall mediation model. The model includes one independent variable (objectively measured MVPA trajectories across childhood and adolescence), one mediating variable (leg LST), and dependent variables (aBMD, aBMD distribution, and specific geometric measures of PF).
RESULTS
Accelerometry data from ages 5 to 17 yr were available for 524 participants (264 females) with at least three accelerometry assessments over time. The number of accelerometry assessments at each wave was 396, 498, 506, 460, 380, and 358 for waves 1 to 6, respectively. Combining PF DXA data with physical activity latent groups, we have 349 participants (191 females).
Participant characteristics
Table 1 describes the maturity status, body size, and bone outcomes of the participants by sex at age 17 yr.
TABLE 1: Description of maturity, body size, and bone outcomes of participants at age17 yr.
Physical activity trajectories
The MVPA trajectories are presented in Figure 2A (females) and Figure 2B (males). For females, two latent group trajectories were identified: 1) relatively low MVPA over all study period with some decrease over time and 2) relatively high MVPA with some decreasing activity over time. The proportion of individuals in each group was 66.7% and 33.3%, respectively. On average, group 1 females participated in approximately 38 min·d−1 of MVPA at age 5 yr, which steadily declined to 20 min by age 17 yr. Group 2 females participated in 55 min·d−1 of MVPA at age 5 yr, 61 min·d−1 of MVPA at age 8 yr, which declined to 31 min·d−1 by age 17 yr.
FIGURE 2: Trajectories of MVPA (min·d−1) across childhood and adolescence females (A), and males (B), based on latent group membership. Each line represents a percentage of the Iowa Bone Development Cohort which clustered within a discrete physical activity pattern.
For males, three latent group trajectories were identified: 1) relatively low MVPA with decreasing MVPA levels over time; 2) moderately active with increasing MVPA levels during middle childhood followed by decreasing levels over time and 3) active with increasing MVPA levels during middle childhood followed by decreasing levels over time. The proportion of individuals in each group was 45.2%, 44.1%, and 10.7%, respectively. On average, group 1 males participated in approximately 46 min·d−1 of MVPA at age 5 yr, which declined steadily to 33 min·d−1 by age 17 yr. Group 2 males participated in 61 min·d−1 of MVPA at age 5 yr and increased at age 8 yr to 76 min·d−1, before steadily declining to 35 min·d−1 by age 17 yr. Finally, group 3 males participated in 79 min·d−1 of MVPA at age 5 yr and increased at age 8 yr to 105 min·d−1, before declining to 50 min·d−1 by age 17 yr. Due to a small sample size for males’ group 3, groups 2 and 3 were combined for further analyses and are referred as group 2 in tables and Discussion.
Analysis of covariance
Leg LST and bone outcomes were compared for MVPA trajectory groups 1 (less active) and 2 (more active) by sex, adjusting for lower limb length, body weight, and PHV age (Table 2) as estimated marginal means from ANCOVA models. Females and males in group 2 showed greater leg LST than group 1. Females in group 2 had significantly greater aBMD at the total, and all PF regions than females in the group 1. Females in the group 2 had lower IM|SL neck aBMD (i.e., a more homogeneous aBMD distribution between the SL and IM regions of the femoral neck) than females in the group 1. Similarly to females, males in the group 2 had significantly greater aBMD at the total and all PF regions than males in the group 1. Males in the group 2 had longer HAL than males in group 1.
TABLE 2: EMM and SE for leg lean soft tissue and bone outcomes (adjusted for lower limb length, body weight and PHV age) by MVPA trajectory group membership.
Mediation analyses
The mediation analyses (depicted in Fig. 1) are summarized in Table 3 for females and in Table 4 for males. There was a significant indirect effect of MVPA trajectory groups on aBMD of total and all PF regions via leg LST in both females and males. The directions of the path a (see Table, Supplemental Digital Content, path a—females and males unstandardized regression coefficients of physical activity to leg LST, https://links.lww.com/MSS/B386) and path b (see Table, Supplemental Digital Content, path b—females and males unstandardized regression coefficients of leg LST to bone outcomes, https://links.lww.com/MSS/B386) were positive. The effect ratios show that leg LST explained 43% to 49% of the total effects of MVPA on PF aBMD in females and 27% to 32% in males. Significant direct effects of MVPA trajectory groups on aBMD of total and all PF regions were also observed, but only in males.
TABLE 3: Females mediation analysis.a
TABLE 4: Males mediation analysis.a
For aBMD distribution, an indirect effect of MVPA trajectory groups on TR|PF aBMD through leg LST was observed in females, although the total effect was not significant (P = 0.073). The directions of the path a and path b were positive. In addition, an indirect effect of MVPA trajectory groups on IM|SL through leg LST was observed in both females and males; however, in males, the total effect was not significant (P = 0.105). The direction of the path a was positive, whereas the direction of the path b was negative.
Finally, for specific geometric measures, we found an indirect effect of MVPA trajectory groups on HAL via leg LST in both females and males and on NNW only in males. The directions of the path a and path b were positive. However, for females HAL and males NNW, the mediation model was inconsistent as the direct effect of MVPA on the geometric measures was negative but not significant (P = 0.797 and P = 0.823, respectively). Mediation analyses were not performed for N|PF aBMD, IT|PF aBMD, NNW, and NSA in females and for N|PF aBMD, TR|PF aBMD, IT|PF aBMD, and NSA in males, because path b was not significant for these variables.
DISCUSSION
The purpose of this study was to examine the effect of MVPA trajectories on adolescent bone parameters including an analysis of the indirect (i.e., muscle-mediated) effects.
aBMD
Females and males who accumulated higher levels of MVPA across childhood and adolescence had greater aBMD at the total PF and its regions at age 17 yr when compared with less active peers. A significant indirect effect of MVPA on aBMD of the PF through leg LST was observed in both females and males. The causal paths suggest that high levels of MVPA across childhood and adolescence increase muscle mass, which in turn increases the mechanical loading imposed on bone resulting in positive adaptations. However, muscles may also influence healthy adaptations by secreting local growth factors that can stimulate osteogenesis (27). The leg LST explained a higher percentage of the total effects of MVPA on PF aBMD in females than in males (43%–49% vs 27%–32%). Moreover, a significant direct effect of MVPA on aBMD was observed in males but not in females. These findings emphasize the importance of muscle loading for bone health, particularly in females, because muscle appears to mediate a greater proportion of physical activity and bone relationship in females than males, despite their lower lean mass. This may be a possible explanation for girls having a more bone per unit of lean body mass compared to boys (11,28). We do not know to what extent this better adaptation of the bone to the muscular forces in female sex with extra storage of calcium is a factor preventing the need for gestation and lactation (29). In practice, these findings provide an additional intervention avenue, focus on physical activities that increase muscle mass. Frost, in its Mechanostat theory, already highlighted the bone adaptation to muscle forces (6,7). In accordance, the increase of momentary muscle forces as it occurs in physical activities with strong (des)accelerations of the body or with maximal-force muscle contractions are the ones with greater potential to augment muscle mass and bone strength (30). The results also suggest that in males, the muscle may not matter as much for health because they are getting more vigorous physical activity (i.e., more impact forces) than females (31). So muscle loading/strength training may be particularly important for youth who do not engage in impact activities.
aBMD distribution
In relation to aBMD distribution, an indirect effect of MVPA on the ratio between the IM and the SL N regions through leg LST was observed in both females and males. Thus, the effect of physical activity via muscle may also contribute to more homogenous bone mass distribution between the IM and the SL N regions. This finding is noteworthy because a greater imbalance between these two regions can compromise the femoral neck structural integrity and, with aging, increases the risk for local buckling, particularly at the SL N region (32). In fact, this latter region normally has a lower BMD and/or cortical thickness when compared with the IM N region. To sum up, our findings indicate that a more uniform mineral distribution between the IM and SL N associated with higher levels of physical activity may contribute to the femoral neck resistance to fracture. Furthermore, the BMD at TR region has a critical role in the risk of intertrochanteric fractures later in life (33). The proportion of intertrochanteric fractures increases with age in women but not in men (34,35) and appear to reflect a bigger bone loss at the trochanteric region with age in women (33). Therefore, the indirect effect of physical activity on the ratio between the TR and the total PF through leg LST which we report for females may be important to prevent this specific type of fracture in older age.
Geometric measures
We report an indirect effect of physical activity on HAL via leg LST in females and males and on NNW only in males. The causal paths suggest that high levels of physical activity across childhood and adolescence increase muscle mass which in turn increase HAL and NNW. However, in female HAL and male NNW, the mediation models were inconsistent because the direct effect of physical activity on these geometric measures was negative, although insignificant. This finding suggests that in determining HAL for females and NNW for males, the absolute amount of muscle may be more important than its use (via physical activity). Nevertheless, physical activity appears to be determinant to increase muscle mass. Associations between leg LST and geometric measures were reported previously (36) but the information is scarce. Future studies should use computational approaches (e.g., finite element models) to understand better how mechanic loading from physical activity and muscle could explain the variation of HAL and NNW.
Limitations and strengths
This study has some limitations that should be acknowledged. First, our study includes cross-sectional (age, 17 yr) DXA measures; it is possible that participants in the most active groups had greater bone parameters due to genetics, hormonal factors, or other factors that were not measured. Also, the participants of this study were drawn from a regional sample of children born in Iowa. Thus, caution should be used to the generalization of study results. Second, we used leg LST as a surrogate of muscle mass with the assumption that muscle mass (size) indicates muscle strength (function). Although studies indicate positive associations between muscle mass and muscle strength (37), it is not clear whether increases in muscle mass necessarily parallel increases in muscle strength or vice versa. Third, the evaluation of the physical activity through accelerometry does not allow inferring about the muscular activity requested. However, the greater the volume or intensity of physical activity, the greater the muscle activity involved (38). There is evidence that more than the volume may be the pattern of accumulation of physical activity that presents greater osteogenic potential (39). Future studies should investigate the total volume versus bouts of physical activity in muscle health, particularly in girls, since muscle–bone interaction appears to be more determinant in girls’ bone health than in boys. Considering the negative associations between sedentary behavior and bone health (40), it is also important to analyze whether this association is mediated by muscle mass/strength. Finally, although the point of our study was not to precisely estimate time spent in different physical activity categories, the necessary use of a 1-min epoch, due to the long (12 yr) follow-up, would suggest an underestimate of MVPA particularly when our participants were children. Future research will limit analysis to ages 15 and older to exploit the advancement of accelerometry technology including the use of g as the unit (raw data) of analysis. An additional limitation associated with our long follow-up included the necessary use of three models of Actigraph.
Study strengths include our 12-yr comprehensive assessment of physical activity using accelerometry and group-based trajectory modeling that provided an analytic method to effectively utilize multiple points of data assessed longitudinally and avoid predefined subgroups; for example, increase/decrease/no change groups. An additional strength was our focus on the indirect effects of physical activity through muscle rather than inferring the muscle–bone unit. Our chosen bootstrapping approach tends to have good statistical power and is seen as a useful tool to avoid Type I error (26). In addition, bone mass distribution was estimated through aBMD ratios among PF regions, rather than trying to estimate the bone mass distribution of complex three-dimensional bone cross section from two-dimensional DXA images. Our approach provided measures of intra-individual relative difference of mineralization at the PF in the same plane of 2D-DXA images. Finally, the distinction between the IM and the SL aspects of neck aBMD supplied additional insight.
CONCLUSIONS
To the best of our knowledge, this is the first study addressing the indirect effect (through muscle) of physical activity during growth on PF bone outcomes. Our study highlights the importance to consider muscle and bone together as a functional unit to better understand the mechanisms of bone adaptation. More studies need to be conducted to clarify the mediator role of muscle in the relationship between physical activity and bone outcomes, not only as a mechanical factor but as an endocrine organ.
In conclusion, this study indicates that across childhood and adolescence a higher level of physical activity is associated with greater bone health of the clinically significant PF. We also show that a significant portion of the effects of physical activity on the PF is mediated by muscle, specifically, a higher aBMD and a better aBMD distribution in both, females and males and a longer HAL in males. These results suggest that interventions of physical activity to improve bone health during childhood and adolescence should include activities that improve muscle mass and that these types of interventions may be particularly efficacious in females.
The authors are grateful to the parents, children, and staff of the Iowa Bone Development Study and the Iowa Fluoride Study. This work was supported by the Portuguese Science and Technology Foundation (FCT) (SFRH/BD/79828/2011 and PTDC/DES/115607/2009). It is also supported by the National Institute of Dental and Craniofacial Research grants R01-DE12101, R01-DE09551, and P30-DE10126, and the General Clinical Research Centers Program from the National Center for Research Resources, M01-RR00059 and UL1-RR024979.
No potential conflict of interest was reported by the authors. The authors have no conflicts of interest to disclose in relation to this study. The results of the present study do not constitute endorsement by American College of Sports Medicine. The results of this study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation.
REFERENCES
1. Weaver CM, Gordon CM, Janz KF, et al. The National Osteoporosis Foundation’s position statement on peak bone mass development and lifestyle factors: a systematic review and implementation recommendations.
Osteoporos Int. 2016;27(4):1281–386.
2. Baxter-Jones AD, Eisenmann JC, Mirwald RL, Faulkner RA, Bailey DA. The influence of physical activity on lean mass accrual during adolescence: a longitudinal analysis.
J Appl Physiol (1985). 2008;105(2):734–41.
3. Sioen I, Lust E, De Henauw S, Moreno LA, Jimenez-Pavon D. Associations between body composition and bone health in children and adolescents: a systematic review.
Calcif Tissue Int. 2016;99(6):557–77.
4. Judex S, Carlson KJ. Is bone’s response to mechanical signals dominated by gravitational loading?
Med Sci Sports Exerc. 2009;41(11):2037–43.
5. Ireland A, Maden-Wilkinson T, McPhee J, et al. Upper limb muscle–bone asymmetries and bone adaptation in elite youth tennis players.
Med Sci Sports Exerc. 2013;45(9):1749–58.
6. Frost HM, Schonau E. The “muscle–bone unit” in children and adolescents: a 2000 overview.
J Pediatr Endocrinol Metab. 2000;13(6):571–90.
7. Frost HM. Bone’s mechanostat: a 2003 update.
Anat Rec A Discov Mol Cell Evol Biol. 2003;275(2):1081–101.
8. Lombardi G, Sanchis-Gomar F, Perego S, Sansoni V, Banfi G. Implications of
exercise-induced adipo-myokines in bone metabolism.
Endocrine. 2016;54(2):284–305.
9. Vicente-Rodriguez G, Ara I, Perez-Gomez J, Dorado C, Calbet JA. Muscular development and physical activity as major determinants of femoral bone mass acquisition during growth.
Br J Sports Med. 2005;39(9):611–6.
10. Wilkinson K, Vlachopoulos D, Klentrou P, et al. Soft tissues, areal
bone mineral density and hip geometry estimates in active young boys: the PRO-BONE study.
Eur J Appl Physiol. 2017;117(4):833–42.
11. Baptista F, Barrigas C, Vieira F, et al. The role of lean body mass and physical activity in bone health in children.
J Bone Miner Metab. 2012;30(1):100–8.
12. Sardinha LB, Baptista F, Ekelund U. Objectively measured physical activity and bone strength in 9-year-old boys and girls.
Pediatrics. 2008;122(3):e728–36.
13. Tan VP, Macdonald HM, Kim S, et al. Influence of physical activity on bone strength in children and adolescents: a systematic review and narrative synthesis.
J Bone Miner Res. 2014;29(10):2161–81.
14. Janz KF, Letuchy EM, Burns TL, Eichenberger Gilmore JM, Torner JC, Levy SM. Objectively measured physical activity trajectories predict adolescent bone strength: Iowa bone development study.
Br J Sports Med. 2014;48(13):1032–6.
15. Gabel L, Macdonald HM, Nettlefold L, McKay HA. Physical activity, sedentary time, and bone strength from childhood to early adulthood: a mixed longitudinal HR-pQCT study.
J Bone Miner Res. 2017;32(7):1525–36.
16. Preacher KJ, Hayes AF. SPSS and SAS procedures for estimating indirect effects in simple mediation models.
Behav Res Methods Instrum Comput. 2004;36(4):717–31.
17. Levy SM, Warren JJ, Davis CS, Kirchner HL, Kanellis MJ, Wefel JS. Patterns of fluoride intake from birth to 36 months.
J Public Health Dent. 2001;61(2):70–7.
18. Crabtree NJ, Arabi A, Bachrach LK, et al. Dual-energy X-ray absorptiometry interpretation and reporting in children and adolescents: the revised 2013 ISCD pediatric official positions.
J Clin Densitom. 2014;17(2):225–42.
19. Evenson KR, Catellier DJ, Gill K, Ondrak KS, McMurray RG. Calibration of two objective measures of physical activity for children.
J Sports Sci. 2008;26(14):1557–65.
20. Trost SG, Loprinzi PD, Moore R, Pfeiffer KA. Comparison of accelerometer cut points for predicting activity intensity in youth.
Med Sci Sports Exerc. 2011;43(7):1360–8.
21. Cardadeiro G, Baptista F, Rosati N, Zymbal V, Janz KF, Sardinha LB. Influence of physical activity and skeleton geometry on bone mass at the proximal
femur in 10- to 12-year-old children—a longitudinal study.
Osteoporos Int. 2014;25(8):2035–45.
22. Mirwald RL, Baxter-Jones AD, Bailey DA, Beunen GP. An assessment of maturity from anthropometric measurements.
Med Sci Sports Exerc. 2002;34(4):689–94.
23. Malina RM, Choh AC, Czerwinski SA, Chumlea WC. Validation of maturity offset in the Fels longitudinal study.
Pediatr Exerc Sci. 2016;28(3):439–55.
24. Jones BL, Nagin DS. Advances in group-based trajectory modeling and an SAS procedure for estimating them.
Sociol Methods Res. 2007;35(4):542–71.
25. Mackinnon DP, Fairchild AJ. Current directions in mediation analysis.
Curr Dir Psychol Sci. 2009;18(1):16.
26. Hayes AF. Beyond baron and Kenny: statistical mediation analysis in the new millennium.
Commun Monogr. 2009;76(4):408–20.
27. Hamrick MW, McNeil PL, Patterson SL. Role of muscle-derived growth factors in bone formation.
J Musculoskelet Neuronal Interact. 2010;10(1):64–70.
28. Crabtree NJ, Kibirige MS, Fordham JN, et al. The relationship between lean body mass and bone mineral content in paediatric health and disease.
Bone. 2004;35(4):965–72.
29. Fricke O, Beccard R, Semler O, Schoenau E. Analyses of muscular mass and function: the impact on
bone mineral density and peak muscle mass.
Pediatr Nephrol. 2010;25(12):2393–400.
30. Frost HM. Why do bone strength and “mass” in aging adults become unresponsive to vigorous
exercise? Insights of the Utah paradigm.
J Bone Miner Metab. 1999;17(2):90–7.
31. Janz KF, Letuchy EM, Francis SL, Metcalf KM, Burns TL, Levy SM. Objectively measured physical activity predicts hip and spine bone mineral content in children and adolescents ages 5–15 years: Iowa bone development study.
Front Endocrinol (Lausanne). 2014;5:112.
32. Poole KE, Treece GM, Mayhew PM, et al. Cortical thickness mapping to identify focal osteoporosis in patients with hip fracture.
PLoS One. 2012;7(6):e38466.
33. Mautalen CA, Vega EM, Einhorn TA. Are the etiologies of cervical and trochanteric hip fractures different?
Bone. 1996;18(3 Suppl):133S–7S.
34. Tanner DA, Kloseck M, Crilly RG, Chesworth B, Gilliland J. Hip fracture types in men and women change differently with age.
BMC Geriatr. 2010;10(1):1–4.
35. Crilly RG, Kloseck M, Mequanint S. Hip fracture types in Canadian men and women change differently with age: a population-level analysis.
Clin Med Insights Arthritis Musculoskelet Disord. 2016;9:75–9.
36. Zymbal V, Janz KF, Baptista F. Sexual dimorphism in bone–muscle relationship in young adults.
J Sports Sci. 2017;35(24):2433–8.
37. Kim SE, Hong J, Cha JY, et al. Relative appendicular skeletal muscle mass is associated with isokinetic muscle strength and balance in healthy collegiate men.
J Sports Sci. 2016;34(21):2114–20.
38. Pouliot-Laforte A, Veilleux LN, Rauch F, Lemay M. Validity of an accelerometer as a vertical ground reaction force measuring device in healthy children and adolescents and in children and adolescents with osteogenesis imperfecta type I.
J Musculoskelet Neuronal Interact. 2014;14(2):155–61.
39. Gabel L, Macdonald HM, Nettlefold L, McKay HA. Bouts of vigorous physical activity and bone strength accrual during adolescence.
Pediatr Exerc Sci. 2017;29(4):465–75.
40. Koedijk JB, van Rijswijk J, Oranje WA, et al. Sedentary behaviour and bone health in children, adolescents and young adults: a systematic review-supplementary presentation.
Osteoporos Int. 2017;28(10):3075–6.
