Childhood Fractures Are Associated With Decreased Bone Mass Gain During Puberty: An Early Marker of Persistent Bone Fragility?


  • Serge L Ferrari MD,

    Corresponding author
    1. Service of Bone Diseases, WHO Collaborating Center for Osteoporosis Prevention, Department of Rehabilitation and Geriatrics, Geneva University Hospital, Geneva, Switzerland
    • Service of Bone Diseases, Geneva University Hospital (HUG) 21, rue Micheli-du-Crest 1211, Geneva 14, Switzerland
    Search for more papers by this author
  • Thierry Chevalley,

    1. Service of Bone Diseases, WHO Collaborating Center for Osteoporosis Prevention, Department of Rehabilitation and Geriatrics, Geneva University Hospital, Geneva, Switzerland
    Search for more papers by this author
  • Jean-Philippe Bonjour,

    1. Service of Bone Diseases, WHO Collaborating Center for Osteoporosis Prevention, Department of Rehabilitation and Geriatrics, Geneva University Hospital, Geneva, Switzerland
    Search for more papers by this author
  • René Rizzoli

    1. Service of Bone Diseases, WHO Collaborating Center for Osteoporosis Prevention, Department of Rehabilitation and Geriatrics, Geneva University Hospital, Geneva, Switzerland
    Search for more papers by this author

  • The authors state that they have no conflicts of interest.


Whether peak bone mass is low among children with fractures remains uncertain. In a cohort of 125 girls followed over 8.5 years, 42 subjects reported 58 fractures. Among those, BMC gain at multiple sites and vertebral bone size at pubertal maturity were significantly decreased. Hence, childhood fractures may be markers of low peak bone mass acquisition and persistent skeletal fragility.

Introduction: Fractures in childhood may result from a deficit in bone mass accrual during rapid longitudinal growth. Whether low bone mass persists beyond this period however remains unknown.

Materials and Methods: BMC at the spine, radius, hip, and femur diaphysis was prospectively measured over 8.5 years in 125 girls using DXA. Differences in bone mass and size between girls with and without fractures were analyzed using nonparametric tests. The contribution of genetic factors was evaluated by mother-daughter correlations and that of calcium intake by Cox proportional hazard models.

Results: Fifty-eight fractures occurred in 42 among 125 girls (cumulative incidence, 46.4%), one-half of all fractures affecting the forearm and wrist. Girls with and without fractures had similar age, height, weight. and calcium intake at all time-points. Before and during early puberty, BMC and width of the radius diaphysis was lower in the fracture compared with no-fracture group (p < 0.05), whereas aBMD and BMAD were similar in the two groups. At pubertal maturity (Tanner's stage 5, mean age ± SD, 16.4 ± 0.5 years), BMC at the ultradistal radius (UD Rad.), femur trochanter, and lumbar spine (LS), and LS projected bone area were all significantly lower in girls with fractures. Throughout puberty, BMC gain at these sites was also decreased in the fracture group (LS, −8.0%, p = 0.015; UD Rad., −12.0%, p = 0.004; trochanter, −8.4%, p = 0.05 versus no fractures). BMC was highly correlated between prepuberty and pubertal maturity (R = 0.54–0.81) and between mature daughters and their mothers (R = 0.32–0.46). Calcium intake was not related to fracture risk.

Conclusions: Girls with fractures have decreased bone mass gain in the axial and appendicular skeleton and reduced vertebral bone size when reaching pubertal maturity. Taken together with the evidence of tracking and heritability for BMC, these observations indicate that childhood fractures may be markers for low peak bone mass and persistent bone fragility.


Fractures constitute 10–25% of all pediatric trauma. Large epidemiological studies have found an annual incidence of fracture of 103–165/10,000 girls and 162–257/10,000 boys, with 27–40% of girls and 42–51% of boys sustaining at least one fracture during growth.(1–4) Moreover, a first fracture is associated with an increased risk of multiple fractures during growth.(3,5) Twenty-five percent to 35% of all fractures affect the distal forearm (37–39/10,000 children/year(4,6)), followed by the fingers, carpal-metacarpal bones, and clavicle.(2) In contrast, the incidence of femoral fractures in childhood is very low (2–3/10,000(4,7)).

It has been hypothesized that the high incidence of fractures in childhood could mainly result from a transient deficit in bone mass relative to longitudinal growth.(8) Indeed, the peak incidence of fractures in girls occurs between 11 and 12 years of age and in boys between 13 and 14 years of age.(1,4) This period corresponds to the age of peak height velocity (PHV) in both sexes and precedes by nearly 1 year the time of peak BMC velocity (PBMCV).(9–11) Two early studies using single-photon absorptiometry showed that BMC at the forearm was reduced about 10% among children with fractures compared with controls.(12,13) Two case-control studies in, respectively, 100 girls (3–15 years) and 100 boys (3–19 years) with distal forearm fractures further indicated that BMC and/or BMD measured by DXA was 5–10%, significantly lower at the ultradistal radius, the lumbar spine, and trochanter compared with age-matched controls.(14,15) Furthermore, after 4 years, BMC in these fractured girls remained lower compared with controls.(3) Taken together with the notion of a bone mass “tracking” during growth,(16) these data suggest that fractures in childhood might be associated with decreased peak bone mass and increased risk of osteoporosis later in life.(17) Whether girls with fractures have low bone mass before puberty, at the time of PHV, and/or after they reach pubertal maturity remains unclear.

In 1993, we enrolled 149 prepubertal girls in a 1-year, randomized, double-blind, placebo-controlled trial to assess the effects of calcium supplementation on bone mass gain.(18) This cohort was followed for up to 8.5 years (i.e., until they reached pubertal maturity; Tanner's stage 5).(19,20) We now report the relationship between BMC and the occurrence of fractures in these girls and study the influence of dietary calcium and heredity therein.



Healthy, prepubertal, white girls were recruited within the Geneva district in 1993. The exclusion criteria were ratio weight/height <3rd or >97th percentile, physical signs of puberty, chronic disease, malabsorption, bone disease, and regular use of medication. One hundred forty-nine subjects (Tanner's stage 1; age range: 6.6–9.4 years) were included in a 1-year, randomized, double-blind, placebo-controlled trial on the effects of calcium-enriched foods (800 mg/day) on bone mass gain.(18) Follow-up visits occurred 1, 3.5, and 7.5 years after the end of the intervention study, and the results from DXA measurements performed at these times have been reported in detail elsewhere.(19,20) The mothers of these girls were also invited once for DXA measurements, and correlations for bone mass between the mothers and their prepubertal daughters have been published.(16) The ethics committee of the Department of Pediatrics of the University Hospitals of Geneva approved the protocol. Informed consent was obtained from parents and children.

Clinical assessment

Weight, stadiometer-derived standing height, and body mass index (BMI, kg/m2) were measured in all subjects. Tanner's pubertal stage was determined by a pediatrician at baseline and the end of the intervention study and at each follow-up visit by self-assessment based on drawings and written description of Tanner's breast and pubic hair classification. Menarcheal age was determined by questionnaire at the last follow-up visit. Three among the 125 girls who completed follow-up were unable to provide precise information about their first menstruation. Fracture history, including skeletal site, year of event, and type of intervention, was recorded at each follow-up visit from the children and their parents. No other disorder susceptible to affect the skeleton was found during this time. Dietary calcium intake was evaluated by frequency questionnaires at baseline, 6 and 12 months, and at each follow-up visit as previously described.(19,20)

Bone variables

BMC (g) was determined by DXA (QDR-2000 instrument; Hologic, Waltham, MA, USA) at six skeletal sites: radius ultradistal (UD Rad.) and diaphysis, femoral neck (FN), trochanter and diaphysis (midshaft), and L2–L4 vertebrae (LS) in antero-posterior view. The CV of repeated measurements at these sites as determined in young healthy adults varied between 1.0% and 1.6%. LS areal BMD (aBMD, g/cm2) was provided by DXA, and bone mineral apparent density (BMAD, g/cm3) was calculated as LS BMC/projected bone area3/2.(21)


At baseline, all girls were Tanner's pubertal stage 1, whereas at their last visit, they were all Tanner's stage 5. Moreover, age, weight, and height did not differ between the fracture and no-fracture groups (Table 2). Accordingly, crude BMC data were used for comparisons within prepubertal and pubertally mature girls. To compare BMC at visit 3, when early pubertal girls were at Tanner's stages 1 or 2, BMC means and SD were derived from the whole cohort for each pubertal stage and 1-year strata therein, and individual Z scores were calculated. BMC gain was calculated over 7.5 years as the difference between pubertal maturity and the end of the calcium intervention.

BMC data in the fracture and/or no-fracture groups were often abnormally distributed, as evaluated by the Kolmogorov-Smirnov (KS) test. Accordingly, nonparametric tests were used first for comparisons, including the KS test for equality of distribution between these groups, the Mann-Whitney U-test for two-group medians, and the Kruskal-Wallis rank test for multiple groups (to compare BMC in girls with upper limb fractures, multiple fractures, and no fractures). However, for large data sets (N > 40), the central limit theorem suggests that the t-test will produce valid results even in the face of non-normally distributed data. Hence, unpaired Student's t-tests were also used secondarily to compare means. Because for most sites, BMC gain over 7.5 years was positively and significantly correlated to BMC measured in prepubertal years, logistic regression was further used to assess whether BMC gain was associated with fractures after adjusting for prepubertal BMC. Pearson's correlation coefficients (R) for crude BMC values in prepubertal and pubertally mature girls and between the latter and their mothers (n = 93 pairs) were assessed by simple linear regression, as previously described.(16) Association of dietary calcium intake with BMC gain was evaluated by simple and multiple regression analyses, the latter also including age of menarche, height, and weight gain. Fracture distribution between girls who took calcium supplements during year 1 (active treatment group, n = 49) and those who interrupted calcium supplements or received placebo during this year (n = 76) was evaluated by χ2 analysis, whereas the proportion of girls in each group that remained fracture-free during follow-up was evaluated by Kaplan-Meier analyses. The relative risk of incidental fractures associated with calcium supplements, dietary calcium intake, menarcheal age, gains in weight and height, and BMC was evaluated by Cox proportional hazard models. This model does not require assumptions about the nature or shape of the hazard function and can therefore be considered a nonparametric method.(22) All tests were computed by Statview 5.0 version of SAS.


Fracture incidence and distribution

Of the 149 prepubertal girls originally enrolled, 125 completed 8.5 years of follow-up (retention rate, 84%). Among those, 42 girls reported 58 fractures (cumulative incidence, 46.4%), with a peak incidence between 9 and 12 years of age. There were 5 fracture cases at baseline (cohort mean age ± SD, 7.9 ± 0.5 years), 3 new cases during the first year (by 8.9 ± 0.5 years), 9 during the second year (by 9.9 ± 0.5 years), 12 over the next 2 years (by 12.5 ± 0.5 years), and 13 during the last 4 years of follow-up (by 16.4 ± 0.5 years). Fractures most commonly affected the forearm (n = 14) and wrist (n = 14), together representing 48% of all fractures, followed by the hand/fingers (n = 6) and the arm/shoulder (n = 5). Twenty-four percent of fractures (n = 14) affected the lower limb, including the foot, but femur fractures did not occur. Only five nonlimb fractures were reported (three for vertebrae, one for the sacrum, and one for skull). Thirteen girls reported multiple(2–4) fractures throughout puberty, which accounted for one-half of all fractures, in which case the upper limb was always affected. Only one case of refracture occurred at the same site (forearm) within 1 year, thereby excluding nonunion as a common mechanism of multiple fractures in this cohort.


At baseline, age, height, and weight were similar in the fracture and no-fracture groups. In contrast, BMC at the radius diaphysis was significantly lower in the former (Table 1). After 2 years, 36% of subjects had reached Tanner's pubertal stage 2. At this early pubertal stage, BMC (Z scores, adjusted and standardized for pubertal stage and age) was significantly lower at radius diaphysis and LS in the fracture compared with the no-fracture group (Fig. 1). As all girls reached pubertal maturity (Tanner's stage 5), BMC at the ultradistal radius (UD Rad.), femoral trochanter, and LS were all significantly lower in girls with fractures, whereas height and weight remained similar in both groups (Table 1). The age of menarche was also similar in girls with and without fractures (13.1 ± 0.15 and 12.9 ± 0.13 years, respectively; not significant). In the subgroup of mature girls who reported upper limb fractures, either as a single fracture (n = 19) or as part of a multiple fractures history (n = 13), BMC was lower at both UD Rad. (p = 0.021) and LS (p = 0.011 versus no-fracture group by Kruskall-Wallis test).

Table Table 1.. Clinical Characteristics and BMC in Girls With and Without Fractures
original image
Figure Figure 1.

BMC in early pubertal girls with and without fractures. BMC measured by the mean age ± SD of 9.9 ± 0.5 years was adjusted and standardized for pubertal stage (Tanner's stages 1 and 2) and for age and expressed as Z scores. *p = 0.05 by both unpaired Student's t-test and Mann-Whitney U-test, and p = 0.067 by unpaired Student's t-test and p = 0.033 by Mann-Whitney U-test compared with no-fracture group. Bars represent means ± SE.

Bone size and BMD

To evaluate whether BMC differences between fractured and nonfractured girls reflected bone size (geometry) and/or BMD at these sites, the projected bone area (BA), bone width, aBMD, and BMAD, an estimate of volumetric bone density, were further evaluated. In prepubertal girls, the radius diaphysis BA and width were significantly lower in those with fractures compared with no fractures (BA: 2.15 ± 0.02 versus 2.23 ± 0.02 cm2; width:1.07 ± 0.01 versus 1.11 ± 0.01 cm, p = 0.028 by nonparametric test and p = 0.019 by t-test for both), whereas similar values were found for radius aBMD (0.432 ± 0.005 versus 0.436 ± 0.003, p > 0.1 in fracture and no-fracture groups, respectively) and BMAD (0.513 ± 0.005 versus 0.501 ± 0.04 g/cm3, p > 0.1 in fracture and no-fracture groups, respectively). In pubertally mature girls, LS BA was significantly lower in the fracture compared with the no-fracture group (42.39 ± 0.68 versus 43.96 ± 0.44 cm2, p = 0.039 by nonparametric test and p = 0.047 by t-test), whereas aBMD was not significantly decreased (1.006 ± 0.015 versus 1.025 ± 0.012 g/cm2, p = 0.09 by nonparametric test), and BMAD was identical in both groups (0.155 ± 0.002 g/cm3 in both). Furthermore, at those sites where BMC did not differ between fractured and nonfractured girls, such as the femur neck and diaphysis, we found no significant differences in bone diameter between groups (data not shown).

BMC gain, tracking, and heritability

Next, we examined 7.5-year changes in BMC from prepuberty (Tanner's stage 1) to pubertal maturity (Tanner's stage 5). Compared with girls without fractures, BMC gain in the fracture group was significantly lower at UD Rad. (−12.0%), LS (−8.0%), and trochanter (−8.4%) and tended to be lower at the radius diaphysis (−6.0%; Table 1). Although BMC gain was positively and significantly correlated with prepubertal BMC values at most sites, adjusting for prepubertal BMC did not alter these results (data not shown). In contrast, girls with and without fractures had similar weight gain and height gain (Table 1). To evaluate a potential deficit in bone mass (and size) accrual relative to longitudinal growth in girls with incidental fractures, we calculated the ratio of BMC-over-height gain during 7.5 years. Compared with those without fractures, BMC/height gain was significantly lower in these girls for both LS and UD Rad. and marginally lower (p = 0.09) at the trochanter (Fig. 2).

Figure Figure 2.

Ratio of BMC-over-height gain in girls with and without fractures. BMC/height (g/cm) changes over 7.5 years were calculated between the mean ages of 8.9 (Tanner's pubertal stage 1) and 16.5 years (Tanner's pubertal stage 5) in girls with incidental fractures during this period (n = 37) and those without incidental fractures (n = 88). **p < 0.01 by Mann-Whitney U-test. Bars represent means ± SE.

In this cohort, we previously showed bone mass tracking and heritability during early puberty.(16) To confirm and extend these findings, we studied BMC correlations between prepuberty and pubertal maturity and between mature girls and their premenopausal mothers. BMC was highly correlated within these girls and also between the latter and their mothers (Table 2), leading to heritability estimates (h2) in the range of 60–90%, except for UD Rad. BMC.

Table Table 2.. BMC Correlations in Girls and Mother-Daughter Pairs
original image

Calcium intake and predictors of fracture risk

Mean calcium intake was similar in the fracture and no-fracture groups both at baseline (899.7 ± 47.8 and 877.7 ± 35.0 mg/day, respectively; not significant) and during follow-up (968.4 ± 57.1 and 910.8 ± 31.9 mg/day, respectively; not significant). Moreover, fracture distribution was similar among subjects who took calcium supplements consistently during the first year (n = 49) and those who did not (n = 76; χ2p value = 0.46); 70% of subjects in each group remained fracture-free during 7.5 years of follow-up (Kaplan-Meier p value = 0.9). Subgroup analysis restricted to girls with incidental upper limb and multiple fractures led to similar results (data not shown).

We further examined the independent relative risk of an incidental fracture associated with calcium supplements (versus placebo), mean calcium intake, age of menarche, gain in weight and height, and BMC at baseline or BMC gain during puberty (by Cox proportional hazards model). Consistent with our previous analyses, this model identified four predictors of fracture risk in these girls: prepubertal BMC at radius diaphysis and BMC gain at UD Rad., LS, and trochanter (Table 3).

Table Table 3.. Predictors of Childhood Fractures
original image


By prospectively studying a cohort of girls with a high incidence of fractures throughout puberty, we made three major findings. First, before and during early puberty, BMC and bone size were mostly decreased at the radius diaphysis in the fracture group. In turn, prepubertal BMC at the radius diaphysis was an independent predictor of the relative risk of incidental fractures. Although the vast majority of childhood fractures results from a high-energy trauma and their predominant forearm localization may be explained by the dynamics of falls and protective responses at that age, these data suggest that DXA of the forearm in early childhood may be an indicator of fracture risk during growth. Second, we found that girls with fractures had decreased 7.5-year BMC gain not only at the UD Rad., which is a common site of fractures, but also at the spine and trochanter, where fractures rarely occur. This deficit in bone mass gain persisted after adjustment for height gain, supporting the hypothesis that a delay in bone widening during rapid longitudinal growth may be a major mechanism for childhood fractures.(8) Third, and perhaps most importantly, we showed that girls with fractures have lower bone mass at the radius, trochanter, and spine when reaching pubertal maturity. This widespread deficit of bone mass, and particularly the smaller vertebral size, is unlikely to result from the fracture itself and/or its related immobilization. Rather, we propose that a history of fracture in childhood may point toward low peak bone mass acquisition and persistent skeletal fragility. This interpretation is further supported by our observations of a high degree of correlation for BMC across puberty and between mature daughters and their mothers (Table 2), which confirms an important tracking and heritability for bone mass.(16,23,24) Noteworthy, we recently reported that, in these girls, mean aBMD at the hip and LS were equal to +0.21 and −0.36 T scores, respectively, when they had reached Tanner's pubertal stage 5,(20) further indicating that peak bone mass was virtually achieved at these sites. Whether females with a history of fractures in childhood have a higher prevalence of osteoporosis later in life remains to be directly studied.

The fracture incidence rate in our cohort is comparable with several large-scale observational studies indicating that traumatic fractures affect nearly one in three female children and adolescents.(1–4,6) In this regard, our sample is likely to be representative of broader populations sharing similar ethnic and environmental backgrounds. In keeping with our own data, childhood fractures have previously been associated with decreased BMC at the forearm,(12,13) the major site of fractures at this age, and also with BMC at fracture-independent sites, such as the trochanter and LS.(14,15) A single follow-up study reported that BMC gain over 4 years was lower in girls with a forearm fracture, whereas 50% of subjects had not yet reached pubertal maturity at the time of their second evaluation.(25) In contrast, in prepubertal children, some investigators failed to find an association between BMD and fracture risk.(26) This discrepancy could be explained both by the low prevalence of fractures in prepubertal children (i.e., before the period of peak height velocity) and by partial correction for bone size using aBMD. Consistent with this proposal, we show here that radius and spine aBMD and furthermore BMAD, an estimate of volumetric BMD, were not significantly decreased in the fracture group, contrary to BMC and projected bone area. Indeed, most of the skeletal fragility in childhood could be explained by a deficit in bone widening concomitant to rapid longitudinal growth.(9–11,27)

It has been proposed that poor calcium intake might be responsible for the significant increase in the incidence of fractures in children and adolescents observed in recent years(6) and that improving bone mass gain with calcium supplements could have a favorable impact on fracture risk in childhood. Our study is the first to address this issue directly. Observational studies on the relationship between calcium intake and BMD gain have led equivocal results,(28–30) but several calcium intervention trials in childhood and adolescence have shown that increasing calcium intake improves bone mass gain, both before and during puberty.(19,31–35) In contrast, we found no association between calcium intake or supplements and childhood fractures. One possible explanation would be that calcium influences the accrual of bone mineral mass without major effects on bone geometry, particularly on bone cross-sectional area. Another possible explanation would be that the mean dietary calcium intake in our cohort was sufficient to prevent calcium supplements effects on fractures. Young children avoiding milk are prone to fracture before puberty, particularly at the forearm.(36) However, poor milk intake not only implies low calcium intake but also decreased phosphate and protein intake, which also play an essential role in bone strength.(37) Moreover, younger children with fractures not only have low BMD but also high body weight,(36) hence a heavier load imparted to their bones on falling. New prospective studies are therefore needed to investigate the effects of calcium intake on childhood fractures.

Our study has some limitations. First, fractures were self-reported rather than systematically ascertained through X-rays and medical records. However, the accuracy of self-report for fractures was recently shown to be 80% or greater for forearm/wrist and hip in postmenopausal women from Women's Health Initiative (WHI).(38) Moreover, this study concluded that some of the unconfirmed self-reports may be caused by poor medical record systems. The average educational and social-economic status of our study population was high, and by conducting yearly interviews of both children and their parents, we were able to collect fracture history within each year of a new fracture event. We are confident therefore that self-report of fractures in this cohort was accurate. Second, in absence of pQCT evaluation, we were unable to directly evaluate the contribution of cortical bone size versus cancellous BMD on fracture risk.(39,40) Another study in girls with forearm fractures has shown that cross-sectional area at the distal radius was 8% lower compared with controls, whereas cancellous BMD was similar.(27) Future studies using high-resolution pQCT should be able to examine this question in more details. Third, based on the observation that calcium supplements (without vitamin D) decrease fracture risk by ∼20% in postmenopausal women,(41) our study had a limited power to detect a significant reduction in fractures in these children. Despite the high incidence of fractures in this cohort, calcium should decrease fracture risk by 75% to have 80% power to detect significant differences in fractures over 7 years in our groups. Prospective studies including large numbers of subjects may therefore be necessary to show whether calcium supplements could decrease the incidence of fractures in childhood.

In conclusion, we show that BMC gain is decreased in girls with fractures. This results in lower bone mass at multiple sites and smaller vertebral bone size when these females reach pubertal maturity. Hence, a history of fractures in childhood could be a marker of increased osteoporosis risk later in life. The genetic and environmental factors implicated in this phenomenon remain to be identified.


We thank Dr F Hermann from the Epidemiology Unit of the Department of Rehabilitation and Geriatrics for advice with statistical methods; Dr D Hans and the technicians from the Service of Nuclear Medicine for performing DXA measurements; MA Schaad, H Clavien, S Gardiol, and M Fueg-Clarisse for taking care of the study subjects and their families; and Prof S Suter and the staff of the Pediatrics Department for their diligent support to our studies.