Size-Corrected BMD Decreases During Peak Linear Growth: Implications for Fracture Incidence During Adolescence


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


Peak adolescent fracture incidence at the distal end of the radius coincides with a decline in size-corrected BMD in both boys and girls. Peak gains in bone area preceded peak gains in BMC in a longitudinal sample of boys and girls, supporting the theory that the dissociation between skeletal expansion and skeletal mineralization results in a period of relative bone weakness.

Introduction: The high incidence of fracture in adolescence may be related to a period of relative skeletal fragility resulting from dissociation between bone expansion and bone mineralization during the growing years. The aim of this study was to examine the relationship between changes in size-corrected BMD (BMDsc) and peak distal radius fracture incidence in boys and girls.

Materials and Methods: Subjects were 41 boys and 46 girls measured annually (DXA; Hologic 2000) over the adolescent growth period and again in young adulthood. Ages of peak height velocity (PHV), peak BMC velocity (PBMCV), and peak bone area (BA) velocity (PBAV) were determined for each child. To control for maturational differences, subjects were aligned on PHV. BMDsc was calculated by first regressing the natural logarithms of BMC and BA. The power coefficient (pc) values from this analysis were used as follows: BMDsc =BMC/BApc.

Results: BMDsc decreased significantly before the age of PHV and then increased until 4 years after PHV. The peak rates in radial fractures (reported from previous work) in both boys and girls coincided with the age of negative velocity in BMDsc; the age of peak BA velocity (PBAV) preceded the age of peak BMC velocity (PBMCV) by 0.5 years in both boys and girls.

Conclusions: There is a clear dissociation between PBMCV and PBAV in boys and girls. BMDsc declines before age of PHV before rebounding after PHV. The timing of these events coincides directly with reported fracture rates of the distal end of the radius. Thus, the results support the theory that there is a period of relative skeletal weakness during the adolescent growth period caused, in part, by a draw on cortical bone to meet the mineral demands of the expanding skeleton resulting in a temporary increased fracture risk.


Skeletal fracture in childhood and adolescence is a major clinical concern.(1 The risk of sustaining some form of fracture from birth to 16 years of age has been reported to be 42% in boys and 27% in girls.(2 A recent study from the United Kingdom showed that about one third of boys and girls sustained at least one fracture before 17 years of age, and at peak occurrence, the incidence of fractures was surpassed only at 85 years of age in women and never in men.(1 In a Canadian study, the peak incidence of fracture to the distal end of the radius during adolescence was equivalent to the fracture incidence at age 54.(3 The physiological rationale for this high rate of fracture during growth is not clear; however, there has been speculation that one of the factors may be the result of a dissociation of bone expansion and bone mineralization during peak growth at adolescence. The theory that there may be dissociation between linear skeletal growth and bone mineralization during adolescence leading to increased bone fragility is not novel, because the hypothesis dates back >40 years(4–6; however, there is little prospective data available to test the premise. In a previous paper, we showed that the age at peak incidence of fracture to the distal radius directly matched the age at which peak velocity of growth in height occurred in both boys and girls.(3 We postulated that this relationship may have resulted from a differential growth rate between linear growth of bone and bone mineralization, causing a transient period of skeletal fragility during the period of most rapid linear growth. Others have shown that the peak age of fracture occurrence in boys and girls is close to the age at which the dissociation between height and volumetric BMC is most pronounced.(7

There is still much to know about true bone mineralization patterns during the growing years. Our knowledge is limited because of the fact that most studies on this age group have been cross-sectional in nature or have failed to control for the wide range in maturational development for children of the same chronological age. In addition, areal BMD (aBMD), the most widely used marker for bone mineralization, is size dependent, which leads to major problems in interpretation in growing children. It is now well documented that DXA-derived aBMD values are inappropriate in assessing bone status in growing children.(8–11 The difficulty with DXA-derived aBMD is that this measure does not correct sufficiently for size during the active growing years. Projectional methods used to measure BMD, such as photon absorption techniques like DXA, measure an area density: larger bones resulting from growth will have higher measurements than previously smaller bones of equal volumetric density. From a clinical perspective, misinterpretation of BMD values from DXA can lead to an overdiagnosis of osteoporosis in children(12 and can cloud our understanding of true bone mineralization during the growing years.(13,14 Much of the existing data on bone mineral acquisition are based on aBMD derived from DXA.(15–18 In addition, these data are also primarily based on chronological age, and thus do not account for maturational differences among children. Maturational, or biological, age can only be controlled in longitudinal designs that prospectively measure subjects over their entire growth period. Thus, although these existing data have provided valuable information, they also have led to some confusion as to the pattern, degree, and timing of bone mineral accrual over the growing years.

The purpose of this study was to test the hypothesis that there is a critical time in adolescence when there is a transient period of relative skeletal fragility accompanying the rapid skeletal growth around the age of PHV. For this to be true, the timing of bone expansion would necessarily have to be in advance of bone mineral accrual, which in turn would be reflected by a temporary decline in size-adjusted BMD. To test this hypothesis, we tracked the size-adjusted total body BMD changes over the growing years from 41 boys and 46 girls followed longitudinally across the adolescent years, and we compared these curves to historical distal radius fracture data.



Participants were from the University of Saskatchewan Bone Mineral Accrual Study that was initiated in 1991 in Saskatoon, Canada.(19 Of the 375 eligible children (ages 8–15) attending two elementary schools, we received written parental consent for 113 boys and 115 girls to be enrolled in the study, and 220 of these children were scanned with DXA. From 1992 to 1993, an additional 31 subjects were recruited. Of this total of 251 subjects, 109 males and 121 females were measured on DXA on at least two or more occasions (median of six scans) over the next 7 years. Five years after completion of the original study, 161 of these original subjects returned for DXA scans as young adults and were measured again for up to 3 additional years. This paper included 41 males and 46 who had been measured annually for up to 7 consecutive years across the adolescent growth spurt and again for up to 3 years as young adults. All study procedures were approved by appropriate university and hospital ethics committees.(19


Height and weight were measured every 6 months during the growth period by trained personnel following standard anthropometric techniques.(20

Biological age—aligned data

Age at peak height velocity (PHV) was used as an indicator of somatic maturity reflecting the age of maximum growth in stature during adolescence.(21 PHV was determined by fitting a cubic spline curve to the whole year height velocities for each individual subject (GraphPad Prism version 3.00 for Windows; GraphPad Software, San Diego, CA, USA). We aligned each individual's bone measurements around years from PHV and fitted a sigmoidal curve to this data. Data for individual's maturational time points (from −2 years PHV to +4 years from PHV) were interpolated from this curve. We were not able to define values at —2 years PHV for 18 girls and 10 boys; thus, the BMDsc values for this time-point came from 28 girls and 31 boys only. In addition, we identified the age of attainment of an adult plateau from the sigmoidal curve and identified a value at a chronological age of 20. This data point represented an average biological age of about +6.5 years from PHV for boys and +8 years for girls.

BMC and bone area

BMC for the total body (TB) and bone area (BA) were measured annually for 7 years using DXA and again when the subjects were adults. (Hologic 2000; Hologic, Waltham, MA, USA). From 1991 to 1997, the mean age increment between measurement occasions was 0.998 ± 0.048 years. Data were collected using array mode, and analyses were done with enhanced global software version 7.1. Short-term precision in vivo for BMC expressed as the CV (%) was 0.60.

Size-adjusted BMD

To determine the more appropriate method for size-adjusting BMD in this age group, we compared the effectiveness two size-correction methods.(10,22 In the former(10 method, BA and BMC were converted to their natural logarithms to obtain a more linear relationship between variables and a linear regression analysis was done to determine the relationship of BA to BMC.(10 The resulting coefficients were used to adjust BMD for size as follows: size-corrected BMD (BMDsc) =BMC/BApc. The second method(22) applies a correction factor that assumes a constant relationship between the variables as bone mineral apparent density (BMAD) =BMC/BA2/height.

The effectiveness of the size-correction techniques was tested by correlation analysis (Spearman) between BA, body mass, and height with BMDsc and aBMD. As shown in Table 1, aBMD and BMAD(22) were significantly correlated with the size variables, whereas BMDsc(10 was not; thus, we used the latter method(10 to size correct BMD in this study. Using this procedure resulted in power coefficients of 1.555 ± 0.015 for males and 1.445 ± 0.015 for females. The subsequent BMCsc units were BMCsc in (g/cm2)1.555 =g/cm3.11 for boys and (g/cm2)1.455 =g/cm2.91 for girls.

Table Table 1.. Correlation Values (R2) Between aBMD and BMDsc With Height, Body Mass, and BA
  inline image

As noted previously, the sample size at age −2 years from PHV consisted of 28 females and 31 males; thus, comparison in BMDsc between this age group and −1 year from PHV were done using a dependent t-test. For all other analyses, the full sample set of 41 boys and 46 girls was used. An ANOVA for repeated measures was used to test for differences in BMDsc across the remaining biological age groups, and posthoc pairwise comparisons were done with a Bonferroni test. All analyses were tested at the 0.05 level.


Physical characteristics of subjects by maturity age are shown in Table 2. Figure 1 shows the distance curve of aBMD with increasing biological age. In comparison, when the appropriate size correction is done, the distance curve looks quite different (Fig. 2). As shown in Fig. 2, BMDsc followed a U curve, actually declining before age at PHV, and thereafter showing a rapid increase. There was a significant difference between ages −2 and −1 years from PHV (for boys, T30 =13.2; for girls, T27 =10.5; p < 0.05). For the full sample (from ages −1 years on), results of the ANOVA for repeated measures showed a significant age effect (F6,35 =35.9 for boys and F6,35 =14.5 for girls; p < 0.05). Succeeding pairwise comparisons revealed a significant decline in BMDsc from −1 years to PHV in both boys and girls. After PHV, BMDsc increased significantly until +4 years from PHV. There was no significant increase from this time-point until chronological age of 20 years (+6.5 years PHV for boys and +8 for girls). In Fig. 3, BMDsc from this study is plotted against age-specific fracture incidence of the distal end of the radius from our previous study on Canadian children.(3 The data for the fracture incidence were historical, collected between 1970 and 1985. However, these data were collected from the same general geographical area (Saskatchewan) and represented similar racial distribution (primarily white) as the subjects in this study. For this comparison, we aligned the fracture data on chronological ages of 12 years for girls and 14 years for boys. As shown, the incidence of fracture increased with decreasing BMDsc and decreased as BMDsc rebounded after PHV.

Table Table 2.. Descriptive Data (Mean, SD, Median, Range) for Age, Height, and Weight by Biological Age Groups and Sex
  inline image
Figure Figure 1.

Areal BMD by biological age for males and females.

Figure Figure 2.

Size-corrected BMD (BMDsc) for males and females by biological age. Units for BMDsc are g/cm3.11 for boys and g/cm2.94 for girls. aSignificant difference from —2 to —1 years and to PHV (n =32 for boys and 27 for girls); bsignificant difference from —1 to PHV for full sample; csignificant difference in BMDsc from PHV to +3 and +4 years from PHV; nsno difference between +4 years from PHV and young adult value. All values at p < .05.

Figure Figure 3.

Distal radius fracture incidence for (A) boys and (B) girls compared with the distance values for total body BMDsc aligned by biological age. Data for fracture incidence is total fractures reported over a 14-year period from hospital admissions.(3

As shown in Fig. 4, the age of peak BA velocity preceded significantly the age of peak BMC velocity in both boys and girls (p < 0.05). In boys, peak BA occurred at 13.69 ± 0.96 years and peak BMC occurred at 14.14 ± 1.05 years; in girls, peak BA occurred at 12.19 ± 0.89 years and peak BMC occurred at 12.67 ± 0.99 years.

Figure Figure 4.

Comparison of ages of peak BMC velocity at peak BA velocity for (A) boys and (B) girls.


Unlike other size-correction methods for aBMD, the statistical approach of Prentice allows the most appropriate relationship between BMC and BA to be determined by the actual data and is not determined from an imposed assumed relationship.(8 Our data show that BMDsc was independent from bone size and therefore should be a more accurate measure of bone mineral accrual relative to size changes over the growing years. Thus, although the method may not be feasible for development of normative databases, nor can comparisons be made between data sets, we feel it is the best method for size adjustment for DXA-derived bone mineral data in understanding BMD changes over the growing years in a longitudinal data set. As noted above, we applied total group (but sex specific) coefficients to correct for BA, and these data are the basis for our comparative analyses. However, to see if the relationship between BA and BMC changed over the active growing years, we also ran regression analyses at each maturity age (Table 3). Using resulting size-corrected BMD values from the different coefficients results in different units of area; thus, a direct comparison across age groups is not valid using these data. Regardless of this limitation, however, we found size-corrected BMD based on these maturity-specific coefficients also decreased between −2 years from PHV to PHV and then leveled off after PHV. The changing relationship between BA and BMC over the active growing years may reflect the changing dimensions of bone during this time period. That is, its possible that the increases in BA results more from gains in linear growth of the skeleton and not periosteal expansion. As maturity proceeds, there is a relative decrease in longitudinal bone growth and a concomitant increase in periosteal expansion. From a biomechanical perspective, the combination of greater linear expansion of bone, combined with relatively lower periosteal expansion, would lead to a weaker bone (and more susceptible to fracture) during the active growing years.

Table Table 3.. Results of Linear Regression of LogBA and LogBMC at Maturational Age Groups (−2 Years PHV to +2 Years PHV) Showing the Resulting Power Coefficients, SE, and CI
  inline image

Previous studies have suggested that DXA-derived aBMD increases steadily from childhood to adolescence,(15 and this is well shown in our data (Fig. 1). However, others have noted that increases in aBMD over the growing years can be attributed to increasing bone size and not BMD.(8,9,23,24 Our results support this view. We actually found a decrease in bone mass relative to bone size before the age at PHV (11.9 years in girls and 13.6 years in boys) followed by a rebound after the age of PHV (Fig. 2). Others have reported that the major proportion of bone mass and size are acquired by late adolescence, followed by small increases thereafter, and the mechanism of bone mass consolidation may be caused by continued increases in bone size at some sites, but could also be caused by increases in volumetric BMD (vBMD) at other sites.(14

We found a clear dissociation between the timing of peak velocity in BA (PBAV) and peak BMC accrual (PBMCV). This confirms previous work showing that growth in size precedes increases in bone mass,(23 and peak velocity in BMC occurs −6 months later than PHV(25 and peak calcium accrual.(26 Our results support the theory that there is a period of relative bone weakness resulting from a dissociation of bone accrual and bone expansion around the time of peak linear growth in both boys and girls. Others have reported that the peak age of fracture occurrence in boys and girls is close to the age at which dissociation between height gain and volumetric BMC is most pronounced,(7 and children with distal forearm fractures have lower aBMD and vBMD than age- and sex-matched controls.(27 The observation that there is actually a decline in BMDsc before PHV is physiologically consistent with the observations of the dissociation of BA and BMC; thus, the results provide strong support for the validity of the size-correction technique of Prentice.(10,17 The lag between bone expansion and bone mineralization makes physiological sense. Bone turnover and remodeling are dramatically increased during growth, and markers of bone turnover peak in boys around age 14 and in girls around age 12(28; during this time of high turnover, there is inevitably relatively more undermineralized bone than in time of low turnover.(29 In the basic multicellular unit, a packet of bone is removed relatively quickly; however, it takes up to 3–6 months for complete mineralization of the osteoid after being laid down by osteoblasts. It also has been speculated that a temporary increase in cortical porosity may occur through remodeling to provide calcium required for the rapidly growing metaphyses of the long bones during the adolescent growth spurt.(30

In a pQCT study, cross-sectional area (CSA) and BMC of the distal radius were found to peak at 16 and 9 months, respectively, before the age of menarche, showing growth asynchrony, but this was not shown at the tibial shaft.(31 Others have reported the timing of peak total body BMC and menarche to be coincident,(32 and there are site-specific differences in the tempo and pattern of skeletal development during the growing years.(23 Thus, the growth patterns at sites that have relatively more cortical bone (tibial shaft) may be different than at sites with relatively greater trabecular bone (distal radius). This difference may in part reflect the different fracture incidence between the two sites. In the study of Cooper et al.,(1 the fracture incidence for girls at the distal radius was about five times greater than at the tibia and fibular sites.

Our results are limited for several reasons. We do not have site-specific data on major fracture sites such as the distal radius; however, we chose to use total body data because the relative contribution of cortical to trabecular bone in the TB is relatively close to that at the distal radius (typical values are −20% trabecular bone at the TB and −25% at the distal radius). The fracture data that we compared our results to are historical and not based on fracture incidence in this sample; however, subjects in our bone data set are representative of the same geographical area and ethnic mix as the subjects in the fracture data base. Finally, it has been suggested that the peak incidence of fracture during adolescence is the result of increased levels of physical activity, but participation in most youth sport programs declines considerably during the second decade of life.(33 During childhood and adolescence, both boys and girls reduce their level of physical activity as they grow older.(34 If activity were the sole reason, why is there a 2-year differential in peak incidence between boys and girls?

In summary, our results provide strong evidence to support the conjecture that high incidence of fractures in adolescence is related to a period of relative skeletal weakness resulting from dissociation between bone expansion and bone mineralization during this period of rapid growth. The data also show the problems of using aBMD in describing bone mineral accrual during the growing years.


This study was supported in part by grants from The Canadian National Health and Research Development Program (NHRDP), the Canadian Institute of Health Research (CIHR), and the Saskatchewan Health Research Foundation (SHRF). PBMAS group members include DA Bailey, ADG Baxter-Jones, PE Crocker, KS Davison, DT Drinkwater, E Dudzic, RA Faulkner, K Kowalski, HA McKay, RL Mirwald, WM Wallace, and SJ Whiting.