Previous studies have shown that habitual physical activity during the growing years increases bone mineral accrual.1–3 Furthermore, there is recent evidence to support a continued benefit of childhood habitual physical activity on bone parameters into young adulthood.4 It appears that the skeleton responds to systematic impact-loading activity in childhood and adolescence by increasing bone mineral accrual.5, 6 It has also been found that sports training during childhood and adolescence increases bone mineral content and density.7, 8 However, it is unclear if the benefits are maintained after retirement from sport and removal of the osteogenic stimulus. Currently, the best evidence linking childhood exercise to bone health in adulthood arises from cross-sectional and short-term prospective studies in retired competitive athletes.9–14 Some studies suggest that retired athletes who started training in childhood have bone mass benefits that are maintained into adulthood; however, other studies report no influence of previous sport participation on bone health.15–17 Prospective studies of the influence of impact loading during childhood on adult bone health are currently lacking and could potentially help clarify this relationship.
Retirement from artistic gymnastics provides a unique model to examine the impact of systematic childhood weight-bearing training on adult bone health. Competitive adolescent female gymnasts have greater areal bone mineral density (aBMD, g/cm2) and bone mineral content (BMC, g) when compared with other athletic and nonathletic populations.9, 18–22 Retired artistic female gymnasts also have significantly higher aBMD values when compared with nongymnasts, with differences ranging from 5% to 22%.9, 11, 12, 23 These effects suggest that potential bone gains from participation in high-impact sport in childhood and adolescence persist into adulthood after removal of the stimulus. However, to date, few studies have examined high-impact or weight-bearing training in childhood and followed the same individuals into adulthood after cessation of sport participation. Scerpella and colleagues24 recently reported that premenarcheal gymnastics training was associated with greater bone parameters in the forearm and that this benefit was maintained 4 years after the cessation of activity. The question that arises, therefore, is: How much of the greater premenarcheal bone mineral observed is maintained in young adulthood after long-term withdrawal of the osteogenic stimulus at other clinically relevant sites? The purpose of the current study was to assess whether the previously reported higher bone mineral content in premenarcheal gymnasts' lumbar spine (LS), femoral neck (FN), and total body (TB)25 were still apparent 10 years after the cessation of participation and withdrawal of the gymnastics loading stimulus. We hypothesized that retired female gymnasts would have greater TB, LS, and FN size-adjusted BMC and aBMD compared with nongymnasts.
Materials and Methods
In 1995, 30 elite premenarcheal female gymnasts were recruited into a study investigating the role of high-impact physical activity on bone mass in childhood.25 Gymnasts were recruited from two gymnastics clubs in Saskatoon, Saskatchewan, Canada. Gymnasts were considered elite if they were competing at the provincial level or higher and training a minimum of 15 hours per week. They also had to have been involved in competitive gymnastics training for at least 2 years before study initiation. In 1995, the gymnasts were 8 to 15 years of age (11.7 ± 1.9 years) and were all premenarcheal. Gymnasts were age-matched to a nongymnast premenarcheal female (n = 30) drawn from the University of Saskatchewan's Pediatric Bone Mineral Accrual Study (PBMAS). PBMAS has been described in detail elsewhere.1 In brief, PBMAS utilized a mixed longitudinal design to examine bone development throughout childhood, adolescence, and into young adulthood. A total of 253 PBMAS children, aged 8 to 15 years, were recruited at study entry (1991 to 1993) and were followed from 1991 to 2011; 169 are still currently enrolled in the study, aged 28 to 35 years. All gymnasts and nongymnasts were white.
In 2009 to 2010, the participants (n = 60) from the 1995 study25 were recontacted. Twenty-seven of the original 30 gymnasts were traced, and 25 agreed to participate in the current study (83%). Of the 27 traced, one was pregnant and one declined to participate. All of the gymnasts had retired from gymnastics training and competition in the preceding years. Gymnasts had been retired for between 6 and 14 years in 2009 to 2010. Twenty-seven of the original 30 PBMAS nongymnasts from 1995 were traced, and 22 agreed to participate (73%). Of the original 27 who were traced, three had withdrawn from PBMAS and two were pregnant. Written informed consent was obtained from all participants, and the study was approved by the University of Saskatchewan's Biomedical Research Ethics Board (Bio # 88-102).
Chronological age, biological age, and anthropometrics
The chronological age of each participant was recorded to the nearest 0.1 year by subtracting the decimal year of the participant's date of birth from the decimal year of the day of testing. Height was recorded to the nearest millimeter using a wall-mounted stadiometer (Holtain Ltd., Crosswell, UK) and body mass to the nearest 0.5 kg using a digital scale (Model 1631, Tanita Corp., Tokyo, Japan). All measures were performed twice; if the difference was >0.4 cm/kg, a third measure was recorded. The mean or median was then reported depending on whether two or three measures were recorded.26 Age at menarche was retrospectively obtained from the retired female gymnasts and prospectively obtained in the PBMAS nongymnasts. Retrospective recall of age at the occurrence of menarche has a high degree of accuracy within a 4-month window of the event.27 Age at the attainment of menarche was then used to create a biological age. Chronological age was subtracted from age at menarche to create years from menarche (such that −1 represents one year before age at menarche and +1 represents one year after the event).
Physical activity, dietary, and health assessment
Physical activity was assessed in 1995 using the Physical Activity Questionnaire for Children (PAQ-C) and in 2009 to 2010 using the Physical Activity Questionnaire for Adults (PAC-AD). The PAQ-C/AD are self-administered 7-day recall questionnaires created to assess general levels of physical activity.28–30 The total activity score on the PAQ-C/AD is calculated as the mean of seven items (each scored on a five-point scale) with five representing high activity and one representing low activity. Calcium and vitamin D intakes were assessed through the use of a 24-hour recall questionnaire at both measurement points. Dietary data were analyzed using the Food Processor and Nutritional Software version 8.5 (ESHA Research, Salem, OR, USA).
In 2009 to 2010, data on menstrual history and use of oral contraceptives were assessed by questionnaire. In gymnasts, further questions included age of onset of gymnastics activity, intensity and duration of training (number of sessions/hours of training per week and level of competition), and age at retirement as well as reason for retirement from gymnastics activity.
Dual-energy X-ray absorptiometry (DXA)
Body composition measurements were performed in 1995 using a Hologic 2000 QDR DXA scanner (Hologic, Inc., Waltham, MA, USA) and in 2009 to 2010 using a Hologic Discovery Wi DXA scanner (Hologic, Inc.). A different scanner was utilized at baseline and follow-up; however, because data were not being compared between time points, the authors did not adjust the values obtained. Three different scans were performed: total body (TB), lumbar spine (LS), and femoral neck (FN). Bone mineral content (g), areal bone mineral density (g/cm2), lean mass (kg), and fat mass (kg) were derived from the scans. All scans were administered and analyzed by a certified radiology technologist. Quality-control phantom scans were performed daily. The coefficients of variation (CV%) for these measures from our laboratory, based on duplicate measures in young, healthy female university students (aged 20 to 30 years), are 0.5% for total body BMC, 0.7% for lumbar spine BMC, and 1.0% for the proximal femur BMC. Fat and lean tissue mass were assessed from the total body scans. Our laboratory has determined coefficients of variation for these measures to be 3.0% and 0.5% respectively.
Variables are presented as means and standard deviation (SD). Group differences (gymnasts versus nongymnasts) for age, age at menarche, biological age, height, weight, lean mass, percent body fat, physical activity levels, vitamin D and calcium intake, and absolute bone values were assessed by independent sample t tests. Multivariate analysis of covariance (MANCOVA) was used to assess differences between groups in TB, LS, and FN BMC and aBMD, while accounting for differences in body size, age, and maturity, all of which have been shown to impact bone parameters (covariates: age, height, weight, and years from menarche  or age at menarche [2009 to 2010]). Correlations between years of retirement and bone parameters were assessed using bivariate Pearson correlations. All analyses were performed using SPSS version 18.0 (SPSS, Inc., Chicago, IL, USA). Alpha was set as p < 0.05.
Anthropometric, body composition, and lifestyle characteristics as well as absolute bone parameters for gymnasts and nongymnasts are presented in Tables 1 and 2. Table 1 contains the 1995 premenarcheal data, and Table 2 contains the young adulthood data. Retired gymnasts trained, on average, 20 hours per week at the peak of their training (range 16 to 30 hours) and had been retired for approximately 10 years (range 6 to 14 years) at follow-up. Gymnasts were significantly shorter and lighter in childhood (p < 0.05; Table 1); however, there was no significant difference in height in adulthood (p > 0.05; Table 2). Gymnasts also had a lower percentage body fat at both measurement occasions (Tables 1 and 2) and reported an older age of menarche at follow-up (p < 0.05; Table 2). Gymnasts were approximately 2 years from age at menarche when tested in childhood, whereas nongymnasts were 1 year premenarcheal, suggesting the gymnasts were later maturers than the nongymnasts.
Table 1. Premenarcheal Anthropometric, Body Composition, and Lifestyle Data for Gymnasts and Nongymnasts (Mean ± SD)
Gymnasts (n = 25)
Nongymnasts (n = 22)
Biological age = years from age at menarche; % fat = total body percent fat; TB = total body; BMC = bone mineral content; aBMD = areal bone mineral density; LS = lumbar spine; FN = femoral neck; PA score = Physical Activity Questionnaire for Children score.
Gymnasts significantly different from nongymnasts (p < 0.05).
Table 2. Adulthood Anthropometric, Body Composition, and Lifestyle Data for Retired Gymnasts and Nongymnasts (Mean ± SD)
Retired gymnasts (n = 25)
Nongymnasts (n = 22)
Biological age = years from menarche; TB = total body; % fat = total body percent fat; BMC = bone mineral content; aBMD = areal bone mineral density; LS = lumbar spine; FN = femoral neck; PA score = Physical Activity Questionnaire for Adults score.
Gymnasts significantly different from nongymnasts (p < 0.05).
Gymnasts had significantly greater unadjusted aBMD at the femoral neck premenarchealy and significantly greater unadjusted total body and lumbar spine aBMD as well as femoral neck BMC (p < 0.05) 14 years later (Tables 1 and 2, respectively). As shown in Figure 1, gymnasts had significantly greater size-adjusted BMC and aBMD compared with nongymnasts at all sites (p < 0.05), with the exception of FN aBMD in adulthood (p > 0.05). There was no significant correlation between the number of years the gymnasts had been retired and any adult bone parameter measured (p > 0.05).
There was no significant difference between groups in history of oral contraceptive use (p > 0.05). In adulthood, the retired female gymnasts had been on oral contraceptives for 5.3 ± 3.7 years, whereas the nongymnasts had been on for 4.7 ± 4.4 years. Including history of oral contraceptive use as a covariate did not alter the significance of any bone parameter (data not shown). The most common reasons cited for retirement from gymnastics training and competition were social life and lack of interest, followed by school demands, injury, and the perception of being too old.
To our knowledge, this is the first study to prospectively examine the long-term (>5 years) skeletal benefits of gymnastics training in premenarcheal girls. The study is unique because we followed the same individuals from childhood to young adulthood, measuring the clinically relevant sites of the LS and FN. The aim was to assess whether the previously reported greater BMC in these elite premenarcheal gymnasts25 was still apparent 10 years after the removal of the gymnastics loading stimulus. The main finding was that gymnasts had greater size-adjusted BMC and aBMD than nongymnasts both premenarchealy and in young adulthood after retirement from gymnastics training and competition. Although the results suggest that gymnastics training in premenarcheal girls results in benefits that are maintained after long-term retirement, it is also possible that females with higher bone mass in adolescence succeeded in gymnastics and that these natural genetic differences were maintained into young adulthood.
The decreased mechanical loading experienced upon retirement from impact-loading sports should, in theory, result in a decrease in BMC and aBMD.15, 17, 31 Kirchner and colleagues23, 32 examined current and retired collegiate-level gymnasts and found that both groups had significantly higher aBMD compared with nongymnasts but that the relation was more pronounced in the active college-level gymnasts compared with retired gymnasts. For example, lumbar spine aBMD was 18% greater in collegiate gymnasts and 16% in retired gymnasts and femoral neck aBMD was 22% greater in collegiate gymnasts compared with 18% greater in retired gymnasts.31, 32 The authors speculated that although some advantages may be lost, these findings suggest that there is a residual benefit of gymnastics participation on bone mass that carries on years after gymnastics participation of the retired gymnasts had ended.23 However, the active collegiate gymnasts and retired gymnasts in the previous studies were not the same individuals; therefore, the observed differences between groups could also be related to differences in genetics or gymnastics exposure (i.e., years or level of training).
Kudlac and colleagues31 examined the influence of gymnastics detraining on aBMD. They measured collegiate female gymnasts at the beginning of their final competitive year and approximately 4 years later. The gymnasts had significantly greater BMC and aBMD at the total body, femoral neck, trochanter, and total hip in their final year of gymnastics competition as well as 4 years after retirement compared with nongymnasts.31 They also reported that aBMD declined at a similar rate in both gymnasts and nongymnasts at the hip (approximately 0.72% to 1.9% a year).31 They proposed that the similar rate of bone loss at the hip between gymnasts and nongymnasts suggests that the gymnasts would maintain an advantage in bone mass as age progresses.31 If so, this advantage may reduce fracture risk at this site. However, it should be noted that the gymnasts in the Kudlac and colleagues31 study were approximately 3 years younger than nongymnasts, and the researchers only measured 10 gymnasts and 9 nongymnasts, which may have influenced the findings. More recently, Scerpella and colleagues24 prospectively examined the influence of retirement from gymnastics training on radius bone parameters in young female gymnasts. They found retired female gymnasts had significant skeletal benefits at the radius throughout growth and in early adulthood, despite cessation of gymnastics training around the time of menarche.24 However, the study sample was small, and the study did not assess the clinically relevant femoral neck and lumbar spine sites.
The gymnasts in the current study were significantly shorter and lighter and had significantly greater size-adjusted bone mass premenarchealy.25 The high-intensity training associated with competitive gymnastics participation during childhood and adolescence has been suggested to negatively impact growth and maturation, resulting in compromised adult stature.33, 34 However, our results do not support this conjecture. Retired gymnasts and nongymnasts did not differ in adult height; this would suggest that premenarcheal gymnastics training did not compromise the attainment of adult stature in this group. Observed differences in stature during childhood25 reflected the fact that the gymnasts were maturing at a later age. The current findings are consistent with the previous literature that gymnasts, in general, report an older age of attainment of menarche compared with nongymnasts.35–37 The attainment of menarche at an older age in the retired gymnasts is of interest because it has been suggested that non-gymnastic late-maturing females possibly have lower bone mass in young adulthood.38, 39 It is argued that the increase in alpha estrogen receptors during puberty alters the stress-strain set point on the inner bone surface, increasing the sensitivity of bone to mechanical stimulation.40, 41 Thus, when estrogen levels are increased, a similar mechanical stimulus results in greater bone adaptation compared with when estrogen levels are low. It is postulated that earlier-maturing individuals experience this rise in estrogen at a younger age compared with those who mature later and thus have a prolonged period of increased sensitivity. As such, it has been suggested that individuals who mature earlier will emerge from adolescence with greater bone mass compared with those who mature later.38, 39 However, in the current study, retired female gymnasts had greater bone mineral content despite having a later age at the attainment of menarche than nongymnasts. This finding suggests a potential for the gymnastics loading stimulus to be overriding the natural tendency for later-maturing individuals to have less bone mass in young adulthood. However, it should also be noted that other studies suggest bone may be most responsive to mechanical loading before puberty,42, 43 which was when the current cohort was actively training.
In the current cohort, retired gymnasts were found to have greater size-adjusted total body, lumbar spine, and femoral neck BMC compared with nongymnasts (13%, 19%, and 13%, respectively). Retired gymnasts also had 8% greater TB and 13% greater LS adjusted aBMD. Furthermore, the observed difference between groups in adulthood, an average of 10 years after retirement from sport and removal of the gymnastics stimulus, was similar to the difference found between premenarcheal gymnasts and nongymnasts. This finding would suggest that the benefit of gymnastics training in premenarcheal girls relative to nongymnasts was maintained 14 years later. Of note, retired female gymnasts were also found to have some significantly higher unadjusted bone outcomes. Retired gymnasts had 5% greater absolute TB aBMD, 7% greater absolute LS aBMD, and 11% greater absolute FN BMC despite being approximately 3 cm shorter and having a mean weight approximately 7 kg less than the nongymnasts. Although retired gymnasts were found to weigh significantly less than the nongymnasts, there was no difference in fat-free mass between the groups; therefore, it would appear the difference between the groups was related not only to their shorter size but also to their relative lean and fat mass. We have previously shown44 that stature predicts approximately 70% of BMC, with lean mass a further 23% and fat mass predicting only about 1% of an individual's BMC. However, because fat mass has been found to be a significant positive independent predictor of bone mass,44 a greater fat mass should result in greater BMC. Although the nongymnasts had greater fat mass and were slightly taller, this advantage did not appear to have influenced bone mass, further emphasizing the possible effects of gymnastics training stimuli on relative lean mass and bone accrual.
After retirement from gymnastics, individuals participated in other competitive sports at the high school and collegiate level, such as track and field, basketball, and soccer, and were found to participate in average levels of physical activity in adulthood. There were no differences between groups in the reported levels of physical activity either before menarche25 or as an adult; however, it may be that the average levels of activity that the retired gymnasts participated in was sufficient to maintain the benefits of previous gymnastics participation. It is unknown if the same retention would be observed if the retired gymnasts in the current study were sedentary. Zanker and colleagues12 examined sedentary retired gymnasts, defined as not meeting the UK requirement for participation in 30 minutes of physical activity on at least 5 days of the week, and found that the retired gymnasts still had 6% to 11% greater aBMD compared with nongymnasts. Duration of retirement had no effect on bone parameters, further lending support to the suggestion that the positive effect of premenarcheal participation does not diminish over time.12 Pollock and colleagues11 found retired female gymnasts maintained higher aBMD values an average 24 years after retirement and showed a similar pattern of bone loss as nongymnasts.
This study has several strengths. To our knowledge, this is first long-term prospective study examining the benefits of gymnastics training in the premenarcheal years on adult skeletal health. The previous studies examining retirement from sport and bone parameters in gymnasts have generally utilized former collegiate-level athletes; however, the majority of gymnasts retire in their teenage years. The gymnasts in the current study retired before the collegiate level, some as young as 12 years of age. Therefore, the current cohort may be a better representation of a composite group of former gymnasts than the studies utilizing solely collegiate-level athletes.
A limitation of this study is that although gymnasts were measured in childhood and found to have significantly greater premenarcheal size-adjusted BMC,25 there were no measures before the onset of gymnastics training. Therefore, it is possible that these observed benefits in bone mass were present before the onset of training and not the result of gymnastics participation. Previous researchers, however, have found that young competitive gymnasts not only have greater aBMD as assessed by cross-sectional comparison but also accrue more bone over time, suggesting that gymnastics training during growth is at least partially responsible for the greater BMC and aBMD observed in young gymnasts.19, 45, 46 Nonetheless, the potential role of genetics and self selection into the sport should also be considered whenever assessing the impact of gymnastics training on bone parameters. Physical activity and diet were assessed using self-report questionnaires. There is always the possibility for reporting errors when self-report is utilized; however, both questionnaires have been previously validated for use in such populations.28–30 Finally, assessment of bone mineral content with exercise studies is limited because important changes in the structural properties of bone may occur and go undetected.47 Both young and retired female gymnasts have been found to have site-specific cortical and trabecular structural advantages, resulting in estimated strength benefits of 14% to 38% in the upper extremity and 5% to 14% in the lower extremity compared with nongymnasts.48–50 Therefore, the current investigation may be underestimating the impact of gymnastics training in the premenarcheal years on adulthood bone health. Longitudinal investigations of bone structural adaptation are required to better understand the effect of premenarcheal gymnastics training on future bone strength and subsequent fracture risk.
In summary, the skeletal benefits of gymnastics training in premenarcheal girls were maintained 10 years after retirement from the sport. The observed differences between gymnasts and nongymnasts were similar in childhood, when gymnasts were actively training, and in adulthood after retirement. The findings support the assertion that structured physical activity during growth is an effective means to increase bone mass and potentially prevent or delay the risk of osteoporosis and related fracture. However, long-term studies are required that follow retired female gymnasts as they approach menopause and bone loss accelerates, to better understand the impact of premenarcheal gymnastics participation on osteoporosis and related fracture risk.
All the authors state that they have no conflicts of interest.
The authors thank and acknowledge the study participants for their commitment to the project. The authors also thank Dr Jay Mafukidze, who collected the 1995 gymnast data as part of her MSc project, and Drs Robert Mirwald and Donald Bailey for their contributions to the University of Saskatchewan's Pediatric Bone Mineral Accrual Study. This study was supported in part by funding from the Canadian Institutes of Health Research (CIHR), the Saskatchewan Health Research Foundation (SHRF), and the CIHR doctoral regional partnership program.
Authors' roles: MCE assisted with study design, collected the data, performed the data analysis, wrote the first draft of the manuscript, and incorporated reviews by all coauthors. SAK, PDC, CMA, and RAF offered important suggestions regarding study design, provided guidance during the data collection process, and reviewed and approved multiple drafts of the manuscript. ADGBJ coordinated the overall concept and design of the study, provided guidance during the data collection and analysis process, and reviewed and provided feedback on all manuscript drafts.