Bone Structure at the Distal Radius During Adolescent Growth


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

  • Published online on December 29, 2008


The incidence of distal forearm fractures peaks during the adolescent growth spurt, but the structural basis for this is unclear. Thus, we studied healthy 6- to 21-yr-old girls (n = 66) and boys (n = 61) using high-resolution pQCT (voxel size, 82 μm) at the distal radius. Subjects were classified into five groups by bone-age: group I (prepuberty, 6–8 yr), group II (early puberty, 9–11 yr), group III (midpuberty, 12–14 yr), group IV (late puberty, 15–17 yr), and group V (postpuberty, 18–21 yr). Compared with group I, trabecular parameters (bone volume fraction, trabecular number, and thickness) did not change in girls but increased in boys from late puberty onward. Cortical thickness and density decreased from pre- to midpuberty in girls but were unchanged in boys, before rising to higher levels at the end of puberty in both sexes. Total bone strength, assessed using microfinite element models, increased linearly across bone age groups in both sexes, with boys showing greater bone strength than girls after midpuberty. The proportion of load borne by cortical bone, and the ratio of cortical to trabecular bone volume, decreased transiently during mid- to late puberty in both sexes, with apparent cortical porosity peaking during this time. This mirrors the incidence of distal forearm fractures in prior studies. We conclude that regional deficits in cortical bone may underlie the adolescent peak in forearm fractures. Whether these deficits are more severe in children who sustain forearm fractures or persist into later life warrants further study.


The most common site of fracture during adolescence is the distal forearm, and the incidence peaks at the time of the pubertal growth spurt.(1–3) The incidence of forearm fractures in childhood is rising,(1–4) and our previous study in Rochester, MN, showed that forearm fractures have increased by 32% in boys and 56% in girls over the past 30 yr.(5) Changes in recreational activities and participation in sports seem to account for only part of the increased incidence of these fractures,(5) raising the possibility that the acquisition of bone mass (and thus bone strength) during adolescence is being impaired. Whether this is caused by changing dietary habits or other lifestyle factors is not clear. Nonetheless, because 25–50% of the peak bone mass of adulthood is accumulated during the pubertal growth spurt,(6) it is possible that adolescents today will be at increased risk for osteoporotic fractures later in life.(7)

Previous studies using DXA have found significant increases in bone mass through puberty,(8,9) but size-corrected DXA measurements suggest decreases in areal BMD (aBMD) during peak linear growth.(10) Studies using pQCT at the distal radius found no changes in trabecular volumetric BMD (vBMD), but increases in cortical vBMD toward the end of puberty.(11,12) Decreased aBMD(13–15) and bone cross-sectional area(11,16,17) have also been found at the ultradistal radius in children with forearm fractures compared with control children. However, these studies are limited by the imaging techniques used: DXA measurements are confounded by bone size and are unable to differentiate cortical from trabecular bone. The standard pQCT used in previous studies in children has a resolution of ∼400 μm and thus lacks the ability to accurately assess bone microarchitecture or to evaluate bone strength. This is now possible using high-resolution pQCT (HRpQCT), which has a voxel size of 82 μm, can define trabecular and cortical microstructure, shows excellent correlation with ex vivo μCT imaging (resolution of 20 μm or better),(18,19) and can be used to construct microfinite element (μFE) models of bone strength.(20) We recently used this technique in adults to obtain an in vivo assessment of bone microarchitecture across life in women and men, making it possible to obtain “noninvasive bone biopsy”(21) data in population studies.(22) Because of its high resolution, this technique allows for the study of microarchitectural changes in bone during growth in humans. We thus conducted a cross-sectional study to quantify changes in bone microarchitecture and strength at the distal radius in girls and boys through puberty, specifically testing the hypothesis that changes in bone microarchitecture during this period may provide an explanation for the observed peak in the incidence of forearm fractures in adolescence.(1–3) We further measured a number of key biochemical and hormonal parameters and related these to the observed microarchitectural and strength variables to define potential determinants of these variables during growth.


Study subjects

The study was approved by the Mayo Clinic Institutional Review Board. Informed written consent was obtained from all subjects >12 yr of age and from a parent for all subjects <18 yr of age. We recruited 140 healthy girls and boys (n = 70 for each sex) age 6–21 yr without prior history of fracture. Reflecting the ethnic composition of the population of Rochester, MN, all but six subjects were white: four girls were black, one girl was white/Asian, and one boy was Asian. None of the subjects had a chronic illness, dietary restrictions, or intake of calcium supplements >1000 mg/d or vitamin D supplements >200 IU/d. No subject had ever used sodium fluoride, calcitonin, bisphosphonates, antiepileptic drugs, or had a history of oral steroid use for >7 days. None of the girls had a history of oral contraceptive use. Skeletal maturity was assessed using plain hand and wrist X-rays, and subjects were divided into groups based on bone-age using the Tanner-Whitehouse III method,(23) although subjects who had completed skeletal maturation (bone-age >15 yr for girls and >16.5 yr for boys) were classified according to chronological age. We divided the subjects into five groups: group I, prepubertal (bone-age 6–8 yr); group II, early pubertal (bone-age 9–11 yr); group III, midpubertal (bone-age 12–14 yr); group IV, late pubertal (bone-age 15–17 yr); and group V, postpubertal (age 18–21 yr). Anthropometric data were collected on all subjects. Calcium and vitamin D intake was assessed using a 3-day food diary, as previously described.(24,25) All subjects underwent a physical examination for pubertal Tanner stage assessment. Data on 13 subjects were excluded because of motion artifact in the HRpQCT scans; thus, 127 subjects (66 girls and 61 boys) were analyzed.

HRpQCT measurements

Measurements were obtained from the nondominant wrist on all subjects using the Xtreme CT (Scanco Medical, Brüttisellen, Switzerland). Using a scout view, a reference line was set at the proximal limit of the epiphyseal growth plate. For subjects whose epiphyseal plates had fused, the remnant of the plate was still visible to set the reference line. The scan was performed on a segment spanning 9.02 mm, starting at a distance 1 mm proximal to the reference line, thereby ensuring that, despite differences in arm length, all subjects had the scans performed as close to the identical anatomic site as possible; this is also the site where adolescent fractures most commonly occur.(1–3) Data were obtained using a 3D stack of 110 high-resolution CT slices with an isotropic voxel size and slice thickness of 82 μm, using an effective energy of 40 keV, field of view of 125.9 mm, and image matrix of 1536 × 1536 pixels. The radiation exposure to the subjects was minimal (local absorbed dose, 0.065 cGy; total radiation exposure, <0.01 mSv).

Trabecular parameters

Bone volume/total volume (BV/TV, %) was derived from trabecular vBMD, assuming mineral density of fully mineralized bone of 1.2 g hydroxyappatite/cm3. Recognizing that individual trabeculae would not be resolved at their correct thickness (∼100 μm) because of partial volume effects, a thickness-independent structure extraction was used to identify 3D ridges (center points of the trabeculae)(26); trabecular number (Tb.N, mm−1) was taken as the inverse of the mean spacing of the ridges.(27) Analogous with standard histomorphometry,(28) trabecular thickness (Tb.Th, μm) was calculated using the formula, Tb.Th = BV/TV ÷ Tb.N, and trabecular spacing (Tb.Sp, μm) was calculated as Tb.Sp = (1 − BV/TV) ÷ Tb.N. Validation studies show excellent correlation (R ≥ 0.96) for these parameters compared with the gold standard ex-vivo μCT technique.(18)

Cortical parameters

The cortex was segmented from the grayscale image with a Gaussian filter and threshold.(27) Cortical vBMD and area were measured directly, and the periosteal circumference was calculated from the contour. Cortical thickness (Ct.Th, μm) was derived using the following formula: Ct.Th = area/circumference. Again, excellent correlation (R = 0.98) has been shown for Ct.Th measurements with HRpQCT versus μCT.(19) Endocortical circumference was calculated assuming that the trabecular compartment was circular. The automatically generated cortical mask was applied to the whole bone structure to obtain the cortical bone, followed by an inversion resulting in a negative image, including only the “cortical pores.” The cortical porosity index was defined as the ratio of cortical pore volume to cortical bone volume.

μFE analysis

Bone strength at the ultra-distal radius was calculated directly from μFE models.(29) Conceptually, a complicated object is divided into a finite number of small and manageable pieces (i.e., elements), using a direct conversion of voxels to hexahedral elements. Frictionless plate compression testing of the region as assessed by HRpQCT was simulated, with the material assumed to be isotropic and linear-elastic. A Young's modulus of 10 GPa and a Poisson's ratio of 0.3(20,30) was applied. To solve these large problems with up to 16 million degrees of freedom, a PCG based parallel solver(31) was used on a CRAY XT4 system. Bone strength was estimated by scaling the resulting load from a test simulating 1% compression, such that 2% of all elements had an effective strain >7000 microstrain. Failure loads calculated from such μFE models correlated highly (R = 0.87) with compressive loads producing Colles' fractures in 54 cadaveric forearms.(20) Additionally, the relative load supported by cortical versus trabecular bone was assessed by calculating the strain energy dissipated in the cortex as a fraction of the total strain energy.

Applied loads and factor-of-risk

For these estimates, we used the loading conditions for a forward fall. The load applied to the wrist was estimated from predicted impact forces on the upper extremity during a fall on the outstretched hand.(32) We assessed the ratio of fall load to overall bone strength, as determined by μFE, as an estimate of the load-to-strength ratio, or factor of risk (Φ).

Areal BMD, BMC, and bone area measurements

aBMD, BMC, and bone area measurements were made from DXA scans performed on the nondominant radius (total radius) (Lunar Prodigy System; GE Healthcare, Madison, WI, USA).

Hormone and bone turnover measurements

A serum bone formation marker, amino-terminal propeptide of type I collagen (PINP), and a serum bone resorption marker, C-terminal telopeptide of type I collagen (CTX), were measured using ELISAs (Immunodiagnostic Systems, Fountain Hills, AZ, USA, for PINP [intra-assay CV < 10%] and Nordic Biosciences, Herlev, Denmark, for CTX [CV < 8%]). IGF-1 and IGFBP-2 were measured by radioimmunoassay (Diagnostic Systems Laboratories, Webster, TX, USA; CVs < 7%). IGFBP-3 was measured by immunoradiometric assay (Diagnostic Systems Laboratories; CV = 3%). PTH was measured by automated immunometric assay (Diagnostic Products, Los Angeles, CA, USA; CV = 7%), and 25-hydroxyvitamin-D [25(OH)D], total estradiol (E2), and testosterone (T) were measured using tandem mass spectroscopy (API 5000; Applied Biosystems-MDS Sciex, Foster City, CA, USA). Details regarding the mass spectroscopy measurements have been previously described.(33)

Statistical analysis

Data are expressed as means ± SE. For all parameters, the mean values of each bone-age group were compared using one-way ANOVA and compared with the prepubertal bone-age group (group I, 6–8 yr) using Dunnet's test, accounting for multiple comparisons. This analysis was performed using the entire cohort as well as excluding the six nonwhite subjects (see above); however, because the pattern of differences in all variables across pubertal groups was virtually identical with or without inclusion of the nonwhite subjects, the data presented are those including all subjects. Associations between imaging parameters and serum biochemical/hormonal variables were assessed using multivariable linear regression models. On these models, both bone age and bone age2 were used, because some of the observed relationships were nonlinear. A stepwise variable selection process was used to build the models.


Table 1 shows the clinical, anthropometric, and hormonal/biochemical parameters in the study subjects. As is evident, in this healthy cohort of children, bone age corresponded closely to chronological age. Tanner stages advanced as expected with bone age. Boys were taller than girls in late puberty, but weight and BMI were similar. Calcium and vitamin D intake were also similar in girls and boys and did not differ significantly across bone age groups. Serum PINP levels were higher in boys than girls by early puberty (group II) and remained significantly higher in the boys thereafter. Both serum PINP and CTX levels tended to be lower in girls and boys in the late pubertal groups. Serum 25(OH)D levels were higher in all groups in boys, reaching statistical significance in groups II, III, and IV. Serum PTH levels were similar across all groups and did not differ between sexes. Serum sex steroid and IGF-I levels showed the expected changes during adolescence. Serum IGFBP-2 levels tended to be lower through puberty and be higher in boys compared with girls, whereas serum IGFBP-3 levels changed in parallel with serum IGF-I levels, reflecting increases in growth hormone secretion.

Table Table 1.. Clinical, Anthropometric, and Biochemical Parameters in the Study Subjects
original image

Figure 1 shows the trabecular and cortical structural parameters from the HRpQCT scans. Compared with the prepubertal group (group I), girls showed no significant difference in BV/TV, Tb.N, Tb.Th, or Tb.Sp through puberty, whereas boys showed significant increases in BV/TV, and Tb.Th in late to postpuberty. Compared with girls, boys had significantly higher values for BV/TV and Tb.Th, but not Tb.N, from late puberty onward. Girls showed significant decreases in Ct.Th and cortical vBMD in midpuberty before increasing sharply at the end of puberty. Conversely, boys maintained Ct.Th and cortical vBMD from pre- to midpuberty, before showing marked increases toward the end of puberty. There was no difference in Ct.Th or cortical vBMD between boys and girls at the end of puberty. In both sexes, periosteal and endosteal circumference by HRpQCT rose from pre- to midpuberty, coming to a plateau toward the end of puberty. In contrast to these structural changes defined by HRpQCT, aBMD, BMC, and bone area by DXA at the total radius increased throughout puberty, with higher values in boys compared with girls at the end of puberty (Fig. 2).

Figure Figure 1.

Trabecular and cortical bone parameters in bone-age groups I through V. (A) BV/TV, bone volume/total volume; (B) Tb.N, trabecular number; (C) Tb.Th, trabecular thickness; (D) Tb.Sp, trabecular spacing; (E) Ct.Th, cortical thickness; (F) cortical vBMD; (G) periosteal circumference; (H) endosteal circumference. *p < 0.05, **p < 0.01, and ***p < 0.001 vs. group I; p < 0.05, ††p < 0.01, and †††p < 0.001 for comparison with the respective group of girls.

Figure Figure 2.

DXA parameters at the radius in bone-age groups I through V. (A) aBMD, (B) BMC, and (C) bone area. *p < 0.05, **p < 0.01, and ***p < 0.001 vs. group I; p < 0.05, ††p < 0.01, and †††p < 0.001 for comparison with the respective group of girls.

Total bone strength, assessed using μFE models, was greater in both sexes from pre- to late puberty, reaching a plateau at the end of puberty in girls (Fig. 3A) but continuing to be higher even at the end of puberty in boys (Fig. 3B). Compared with girls, boys showed greater total bone strength after midpuberty. Because of the increase in height during growth, estimated fall force increased linearly in both sexes (Figs. 3C and 3D), but the load:strength ratio (factor-of-risk, Φ) decreased progressively during puberty in both girls and boys (Figs. 3E and 3F). Thus, overall differences in bone strength, fall loads, or Φ could not provide a clear explanation for the midpubertal peak in distal forearm fractures. Further analysis of the μFE models showed, however, that during growth, the percent of load carried by cortical bone (Figs. 3G and 3H) and the ratio of cortical to trabecular bone volume (Figs. 3I and 3J) decreased sharply in midpuberty in both sexes, with these changes occurring around the ages of peak distal forearm fracture in adolescents.(1–3) Concurrently, the cortical porosity index also increased sharply in both sexes at these ages (Figs. 4A and 4B).

Figure Figure 3.

(A and B) Total bone strength; (C and D) estimated fall loads; (E and F) factor of risk (Φ, ratio of fall loads to bone strength); (G and H) percent of load carried by cortical bone; and (I and J) ratio of cortical to trabecular bone volume in girls (left column) and boys (right column). Shaded regions represent the approximate chronological age ranges when the incidence of adolescent forearm fractures peaks based on previous data from Rochester, MN,(5) and elsewhere.(1–3) *p < 0.05, **p < 0.01, and ***p < 0.001 vs. group I; p < 0.05, ††p < 0.01, and †††p < 0.001 for comparison with the respective group of girls.

Figure Figure 4.

Cortical porosity index in girls (A) and boys (B). Shaded regions represent the approximate chronological age ranges when the incidence of adolescent forearm fractures peaks based on previous data from Rochester, MN,(5) and elsewhere.(1–3) ***p < 0.001 vs. group I; ††p < 0.01 and †††p < 0.001 for comparison with the respective group of girls.

Figure 5 shows a visual representation of the changes in cortical and trabecular bone from subjects in the pre- (group I), mid- (group III), and postpubertal (group V) groups. The deficits in the cortical compartment in the midpubertal group are clearly evident in both sexes. In addition, Fig. 6 shows reconstructions of cortical bone from subjects in these groups, visually depicting the increase in cortical porosity during midpuberty.

Figure Figure 5.

Representative 3D reconstructions of trabecular and cortical bone of the measured region of the ultradistal radius in girls and boys from bone-age groups I, III, and V. The color coding is for the strain energy density (SED) from the μFE models, with low SED values indicating low strains (relatively strong areas) and high SED values indicating high strains (relatively weak areas). Note the thinning of the cortex (indicated by arrows) relative to the amount of trabecular bone present in the subjects in group III.

Figure Figure 6.

Representative 3D reconstructions of cortical bone and the very proximal trabecular bone in girls and boys from bone-age groups I, III, and V. The areas in red indicate cortical pores used to calculate the cortical porosity index.

Table 2 shows the results of the multivariate models for predicting the key bone structural and strength variables. None of the biochemical or hormonal variables were significantly associated with BV/TV, Tb.N, Tb.Th, or Tb.Sp in the girls, whereas greater BV/TV and Tb.Th were related to higher serum T levels and Tb.N and Tb.Sp were most closely associated with higher IGF-I levels in the boys. Ct.Th and cortical vBMD were inversely associated with the bone turnover markers in the girls and with PTH in the boys. Periosteal circumference was most closely associated with IGF-I levels in the girls and T levels in the boys; endosteal circumference was associated with IGF-I levels in the girls and T levels in the boys. Overall bone strength was associated with serum E2 levels in girls and T levels in boys. The percent load carried by cortical bone was negatively associated with CTX and IGF-I levels in the girls and negatively associated with IGF-I and PTH levels in the boys. The cortical porosity index, on the other hand, was positively associated with CTX and IGF-I levels in the girls and positively associated with IGF-I levels in the boys.

Table Table 2.. Multivariable Models for the Key Bone Structural and Strength Variables in the Girls and Boys
original image


Combining HRpQCT with μFE analysis, we describe for the first time the changes that occur in bone microarchitecture and strength in healthy children during the course of puberty. We observed significant differences in trabecular and cortical bone development at the ultradistal radius between girls and boys during the pubertal growth spurt. Trabecular parameters did not change significantly during puberty in girls, suggesting that trabecular bone volume and structure at this site may be programmed early in life in girls and do not change significantly through growth. In contrast, trabecular BV/TV and Tb.Th were higher throughout puberty in boys, and these changes were driven mainly by T and IGF-1 levels. Changes in cortical bone also differed, with girls showing lower Ct.Th and cortical vBMD during midpuberty that was associated with elevated markers of bone turnover. Boys maintained Ct.Th and cortical vBMD through puberty; however, both parameters were negatively associated with PTH, suggesting that an increased demand for calcium during the more pronounced growth spurt in boys may be met by increasing PTH levels.(34) This may, in turn, compromise Ct.Th and cortical vBMD. The larger bone size, reflected by periosteal circumference, was driven mainly by T in boys and IGF-1 in girls, with boys showing a greater bone size from late puberty onward. Thus, whereas studies with estrogen receptor and aromatase-deficient men have clearly shown an important role for E in the development of the male skeleton,(35–39) our data suggest that T may regulate the development of trabecular structure and bone size in boys during puberty, although it remains possible that some of these effects of T may be mediated by local aromatization of T to E in bone.

Our findings at the forearm are somewhat different from studies by Gilsanz et al.(40,41) using QCT at the spine and femur in growing children. Because of the lower resolution, those investigators could not assess bone structure, but they did observe increases in trabecular vBMD in late puberty in girls and boys. In addition, estimates of cortical vBMD at the midshaft of the femur showed little or no change during growth in girls or boys.(42) Whether differences in trabecular and cortical bone changes between these central sites (vertebrae and femur) and the ultradistal radius site assessed in our study are caused by site-specific changes during growth, the particular techniques used (conventional QCT versus HRpQCT), or both is unclear and requires further study. Of note, previous studies using conventional pQCT at the forearm, while not able to assess bone structure, found similar changes in trabecular and cortical vBMD as described here.(11,12)

To try and explain the peak in forearm fracture risk during midpuberty,(1–3,5) we used μFE models to calculate bone strength during this key phase of growth. Total bone strength increased in both sexes through puberty, being driven mainly by T in boys and E2 in girls, with a higher body mass index in girls being associated with greater bone strength. Boys showed greater bone strength than girls from late puberty onward, which continued to increase even at the end of puberty when growth was completed. However, changes either in bone strength or in the factor of risk failed to provide an explanation for the midpubertal increase in forearm fractures. Thus, it was of interest that the percent load borne by cortical bone (which reflects the relative strength of cortical to trabecular bone) and the cortical to trabecular bone volume ratio were at their lowest at midpuberty in girls and late puberty in boys, before rebounding to the original values seen at the start of puberty. This was accompanied by an increase in the cortical porosity index. These changes in cortical bone mirror the incidence of adolescent forearm fractures in a very similar population from Rochester(5) as well as other studies,(1–3) suggesting that these transient deficits in cortical bone during midpuberty may underlie the adolescent peak in forearm fractures. Of note, cadaveric studies have shown the importance of cortical bone as a critical determinant of failure load at the wrist,(43) and a similar reduction in the percent load borne by cortical bone has recently been reported in elderly women with wrist fractures.(44) Our in vivo estimates of cortical porosity are also consistent with the hypothesis proposed by Parfitt that transient increases in cortical porosity may develop in response to the increased calcium demand during rapid longitudinal growth.(34)

We should note that, in addition to analyzing our data using bone-age classifications, we also performed analyses using Tanner staging to classify the subjects and found that the pattern was essentially the same as was seen using bone-age assessments (data not shown). We specifically chose to classify subjects according to bone-age rather than Tanner stage for a number of reasons. First, we believe that bone-age gives a more accurate assessment of the effects of sex steroids on skeletal maturity than Tanner staging. Bone-age assessment is the more objective of the two measures, thus making our classification more accurate. Also, if we restricted ourselves to Tanner stages, we would not be able to appreciate the changes that occur at the distal radius between late pubertal (age 15–17 yr) and postpubertal subjects (age 18–21 yr). As our data shows, there are significant differences in a number of measures during this time, and if all subjects between the ages of 15–21 yr were classified as one group, these differences would not be appreciated.

Our study has significant strengths and some potential limitations. Because of the high resolution possible with HRpQCT (voxel size, 82 μm), our data are perhaps at the limit of feasibility in terms of assessing bone microstructure in vivo in humans using a technology that has been validated for both trabecular and cortical parameters against the current gold standard, μCT (which can only be used in excised bones).(18,19) Nonetheless, we recognize that even with this high resolution, the parameters we assessed using HRpQCT are likely estimates of the true measures. In addition, our cross-sectional findings need to be confirmed by longitudinal studies, but it is important to note that these estimates seem to change in a predictable manner through puberty and reflect differences that have been observed in adults using bone biopsies. For example, our demonstration of a higher BV/TV and Tb.Th at the wrist toward the end of puberty in boys compared with girls is exactly what has been observed for differences in trabecular structure by histomorphometry in men versus women using iliac crest bone biopsies.(45) Thus, given limitations on radiation exposure, HRpQCT is perhaps the best validated tool to assess bone microstructure in humans, especially children.

In conclusion, our findings provide the first description of changes in bone structure during growth as well as a possible structural basis for the adolescent peak in forearm fractures. Further studies are needed to test whether the deficits in cortical bone we describe here are more severe in adolescents who sustain forearm fractures compared with control, nonfracture subjects. In addition, whether adults who sustained forearm fractures during adolescence have persistent skeletal deficits that may predispose them to osteoporotic fractures later in life also warrants further study.


The authors thank Elizabeth Atkinson and Sara Achenbach for help with the statistical analyses, Kelley Hoey for sample processing, James Peterson for data management, Margaret Holets for performing the HRpQCT scans, and the Mayo Immunochemical Core Laboratory for performance of the biochemical and hormonal assays. We also thank the Data and Visualization group at the Swiss National Super Computing Centre CSCS for support. This work was supported in part by the Thrasher Research Fund, NIH Grants AR027065 and UL1-RR24150 (Center for Translational Science Activities), U.S. Public Health Service. Supercomputer time was granted by the Swiss National Supercomputing Centre (CSCS).