Trabecular volumetric bone mineral density is associated with previous fracture during childhood and adolescence in males: The GOOD study

Authors


Abstract

Areal bone mineral density (aBMD) measured with dual-energy X-ray absorptiometry (DXA) has been associated with fracture risk in children and adolescents, but it remains unclear whether this association is due to volumetric BMD (vBMD) of the cortical and/or trabecular bone compartments or bone size. The aim of this study was to determine whether vBMD or bone size was associated with X-ray-verified fractures in men during growth. In total, 1068 men (aged 18.9 ± 0.6 years) were included in the population-based Gothenburg Osteoporosis and Obesity Determinants (GOOD) Study. Areal BMD was measured by DXA, whereas cortical and trabecular vBMD and bone size were measured by peripheral quantitative computerized tomography (pQCT). X-ray records were searched for fractures. Self-reported fractures in 77 men could not be confirmed in these records. These men were excluded, resulting in 991 included men, of which 304 men had an X-ray-verified fracture and 687 were nonfracture subjects. Growth charts were used to establish the age of peak height velocity (PHV, n = 600). Men with prevalent fractures had lower aBMD (lumbar spine 2.3%, p = .005; total femur 2.6%, p = .004, radius 2.1%, p < .001) at all measured sites than men without fracture. Using pQCT measurements, we found that men with a prevalent fracture had markedly lower trabecular vBMD (radius 6.6%, p = 7.5 × 10−8; tibia 4.5%, p = 1.7 × 10−7) as well as a slightly lower cortical vBMD (radius 0.4%, p = .0012; tibia 0.3%, p = .015) but not reduced cortical cross-sectional area than men without fracture. Every SD decrease in trabecular vBMD of the radius and tibia was associated with 1.46 [radius 95% confidence interval (CI) 1.26–1.69; tibia 95% CI 1.26–1.68] times increased fracture prevalence. The peak fracture incidence coincided with the timing of PHV (±1 year). In conclusion, trabecular vBMD but not aBMD was independently associated with prevalent X-ray-verified fractures in young men. Further studies are needed to determine if assessment of trabecular vBMD could enhance prediction of fractures during growth in males. © 2010 American Society for Bone and Mineral Research

Introduction

Osteoporosis and its related fractures are a major public health issue.1 It has been demonstrated that low areal bone mineral density (BMD) is associated with fracture risk in adults.2, 3 Evidence has been presented suggesting that there is also an association between bone mass and fractures in childhood.4 Approximately one in three children suffers a fracture during growth,5 with boys having a higher risk than girls of experiencing a fracture in the first 16 years of life.6 The peak incidence of fractures has been shown to be at the age of 13 to 14 years among boys and at 11 years among girls, with a sharp decline in rate thereafter.5, 7, 8 The age of the peak forearm fracture incidence has been reported to coincide with the age of peak height velocity (PHV),9, 10 but fracture incidence in boys and girls has not been investigated previously directly in relation to PHV. The distal forearm has been identified as the most common fracture site in children, constituting 23% to 32% of all fractures.9, 11, 12 Over the past 30 years, there has been a significant increase in distal forearm fractures in children and adolescents.13 Increased knowledge concerning bone mass in children may help to prevent fractures later in life14 because studies suggest that childhood fractures are associated with low peak bone acquisition and persistent skeletal fragility, and low BMD in a young patient could be considered a potential risk factor for future fracture.15, 16 Both case-control and prospective studies have shown that areal BMD (aBMD) measured with dual-energy X-ray absorptiometry (DXA) can predict fracture risk in boys and girls.8, 17–19 In several studies, it is uncertain how many self-reported fractures could be confirmed in the radiology records.17 A study investigating the accuracy of self-report of fractures in elderly women showed that 11% of self-reports of fracture were false-positive (radiographs were negative), and a total of 20% could not be confirmed (radiographs were negative, uncertain, or not available).20 We have not found a similar study of adolescents, but confirming a fracture with radiology records must be preferred, when possible, in young individuals as well as in older people. Most of the previous studies concerning the association between bone mass and fractures among healthy children that we found used either DXA or quantitative ultrasound. One small study (n = 100) in overweight 10-year-old girls reported an association between cross-sectional area of the radius and fracture of the distal forearm.21 Owing to the limitations of the DXA technique, it remains unclear whether it is the bone size or the volumetric BMD (vBMD) of the different bone compartments, that is, trabecular and cortical bone, that is the most important predictor of fracture risk.

The aim of this study was to determine if it is the bone size or the vBMD that is most strongly associated with X-ray-verified prevalent fractures during childhood and adolescence using a large population-based cohort of young Swedish men.

Subjects and Methods

Subjects

The Gothenburg Osteoporosis and Obesity Determinants (GOOD) Study was initiated with the aim to determine both environmental and genetic factors involved in the regulation of bone and fat mass. Study subjects were randomly identified using national population registers, contacted by telephone, and asked to participate in the study. A total of 1068 men aged 18.9 ± 0.6 years (mean ± SD) from the greater Gothenburg area were included. To be included in the GOOD Study, subjects had to be between 18 and 20 years of age, male, and willing to participate in the study. There were no other exclusion criteria. A standardized questionnaire was used to collect information about fracture history, present physical activity [amount (hours) of all types of regular sport activity was summarized to give the total amount of weekly physical activity (hours/week)],22 nutritional intake (dairy products), and smoking. Calcium intake was estimated from dairy product intake, assuming an average calcium content of 240 mg/glass of milk or yoghurt and of 74 mg/serving of cheese. The GOOD Study was approved by the Ethics Committee at Gothenburg University. Written and oral informed consent was obtained from all study participants.

Anthropometric measurements

Height and weight were measured using standardized equipment. The coefficient of variation (CV) values were below 1% for these measurements.

Peak height velocity (PHV)

PHV was calculated as described previously23 and was available in 600 of the 991 subjects in the total cohort. In summary, detailed growth and weight charts from birth until age 19 years were obtained from a central archive (city of Gothenburg) containing public school health records. These growth charts were used to calculate PHV according to the infancy-childhood-puberty (ICP) model.24 The age of maximum growth velocity (PHV) during puberty was calculated using the algorithm. On average, 23 measurements between birth and 19 years of age were used. PHV generally is believed to be reached within 2 years of pubertal onset.

X-ray-verified fractures

In total, 303 subjects reported that they had sustained any kind of previous fracture. In order to verify these fractures, we collected X-ray records from hospitals and clinics in the greater Gothenburg area during 2007 and 2008. Social security numbers of the study participants were used to search the centralized computerized X-ray registers containing X-ray reports from the three largest public hospitals (Sahlgrenska, Mölndal, and Östra) in Gothenburg. This register contains all X-ray reports from the year of 1991 onward. A central archive containing X-ray reports from the entire Västra Götalandsregionen (a region encompassing a large part of southwestern Sweden with approximately 1.5 million inhabitants) of earlier dates was searched manually using social security numbers. Computerized registers also were searched in the small private hospitals (Lundby and Carlanderska) not included in the preceding central register. Another 78 subjects had not reported fracture but were found to have had a fracture based on X-ray records. They were included in the prevalent fracture group. In contrast, 77 subjects reported fractures that we could not verify in the X-ray records. They were excluded from further analysis, resulting in 991 remaining included men, of which 304 men had an x-ray-verified fracture and 687 were defined as nonfracture subjects.

Dual-energy X-ray absorptiometry (DXA)

Areal bone mineral density (aBMD; g/cm2) of the whole body, femoral neck (of the left leg), lumbar spine, and left and right radii were assessed using the Lunar Prodigy DXA (GE Lunar Corp., Madison, WI, USA). The CVs for the aBMD measurements ranged from 0.5% to 3% depending on application.

Peripheral quantitative computed tomography (pQCT)

A pQCT device (XCT-2000, Stratec Medizintechnik, GmbH, Pforzheim, Germany) was used to scan the distal leg (tibia) and the distal arm (radius) of the nondominant leg and arm, respectively. A 2 mm thick single tomographic slice was scanned with a voxel size of 0.50 mm. The cortical vBMD (not including the bone marrow; mg/cm3), cortical bone mineral content (BMC; mg/mm), cortical cross-sectional area (CSA; mm2), endosteal and periosteal circumference (EC and PC), and cortical thickness (mm) were measured using a scan through the diaphysis (at 25% of the bone length in the proximal direction of the distal end of the bone) of the radius and tibia. Trabecular vBMD (mg/cm3) was measured using a scan through the metaphysis (at 4% of the bone length in the proximal direction of the distal end of the bone) of these bones. Tibia length was measured from the medial malleolus to the medial condyle of the tibia, and length of the forearm was defined as the distance from the olecranon to the ulna styloid process. The CVs were less than 1% for all pQCT measurements.

Statistical analysis

Differences in anthropometric characteristics and bone parameters were investigated using an independent-samples t test. Chi-square tests were used to determine whether or not the distribution of smokers differed between the fracture and nonfracture groups. Chi-square tests also were used to analyze the distribution of side of radius fracture according to dexterity. To evaluate the association between bone variables and previous fracture, binary logistic regression analysis was performed. Weight was not normally distributed and therefore was log-transformed before being entered into the regression analysis. A p value of less than .05 was considered significant. All statistical analyses were performed using SPSS (Version 16.0, SPSS, Inc., Chicago, IL, USA).

Results

Anthropometric characteristics and fracture incidence

The 304 men with prevalent fracture did not differ from the 687 men without fracture in age, smoking habits, height, body weight, calcium intake, or hours per week of physical activity (Table 1). In total, the 304 men had experienced 385 fractures. Of the study participants, 240 had sustained a single fracture, 49 had had two fractures, 13 subjects three fractures, and two subjects four fractures. The peak incidence of fractures in our cohort occurred at the age of 14 years (Fig. 1A). The most common fracture site was the distal forearm, which constituted 33.5% (129/385) of all fractures, followed by fractures of the phalanges of the hand, 16.9% (65/385); fractures of the carpal and metacarpal region, 11.2% (43/385); fractures of the foot (including toes, tarsal/metatarsal region, and calcaneus) 7.5% (29/385); and fractures of the clavicle, 6.5% (25/385). Left-handed subjects were more likely to have had a fracture on the right radius (9/12), whereas right-handed men were more likely to have fractured their left radius (64/98; p = .002).

Table 1. Anthropometrics of the Total Cohort and of the Cohort Divided According to Fracture History
 All subjects (n = 991)Subjects with fracture (n = 304)Subjects without fracture (n = 687)p value
  1. Values are given as mean ± SD. Differences in bone parameters and covariates between the two groups were investigated using independent-samples t test, except for smoking, where the chi-squared test was used.

Age (years)18.9 ± 0.618.9 ± 0.518.9 ± 0.6.453
Smoking (%)8.58.88.3.764
Physical activity (hours/week)4.3 ± 5.24.6 ± 6.04.2 ± 4.8.380
Calcium intake (mg/day)1088 ± 7121069 ± 6531096 ± 737.570
Height (cm)181.2 ± 6.7181.8 ± 6.4181.0 ± 6.8.080
Weight (kg)73.7 ± 11.874.5 ± 12.373.3 ± 11.5.139
Figure 1.

(A) Fracture incidence. The number of prevalent X-ray-verified fractures according to age (n = 991). (B) Fracture incidence according to PHV. Fracture incidence increased with advancing age and reached its peak at the time of PHV, followed by a sharp decline in incidence thereafter (n = 600).

Fracture incidence according to PHV

In the subsample of subjects in whom PHV was available (n = 600), fracture incidence increased with advancing age and reached its peak at PHV, followed by a sharp decline in incidence thereafter (see Fig. 1B). The mean age of PHV was 13.6 ± 1.0 years.

Previous fracture was associated with areal BMD

Areal BMD of the radius, lumbar spine, total body, and total femur was significantly lower in subjects with prevalent fracture than in subjects without fracture (Table 2).

Table 2. Volumetric BMD (pQCT) and Areal BMD (DXA) Are Reduced in Young Men With Prevalent Fracture
 Men with prevalent fracture (n = 304)Men without prevalent fracture (n = 687)p value
  1. Values are given as mean ± SD. Differences in bone parameters between the two groups were investigated using independent-samples t test.

DXA
 Total body aBMD (g/cm2)1.23 ± 0.101.25 ± 0.10<.01
 Lumbar spine aBMD (g/cm2)1.22 ± 0.151.24 ± 0.15<.01
 Total femur aBMD (g/cm2)1.14 ± 0.161.17 ± 0.16<.01
 Radius nondominant aBMD (g/cm2)0.57 ± 0.050.59 ± 0.06<.001
pQCT
 Radius cortical vBMD (mg/cm3)1161 ± 23.41166 ± 21.9<.01
 Radius trabecular vBMD (mg/cm3)209.3 ± 38.3224.0 ± 41.3<.001
 Radius cortical CSA (mm2)95.4 ± 11.596.5 ± 11.8.19
 Radius cortical thickness (mm)2.88 ± 0.282.94 ± 0.26<.01
 Radius periosteal circumference (mm)42.3 ± 3.042.0 ± 2.9.32
 Radius endosteal circumference (mm)24.2 ± 3.323.6 ± 3.0<.01
 Tibia cortical vBMD (mg/cm3)1153 ± 21.01156 ± 19.7<.05
 Tibia trabecular vBMD (mg/cm3)257.3 ± 33.0269.6 ± 33.7<.001
 Tibia cortical CSA (mm2)269.5 ± 34.2270.1 ± 34.1.79
 Tibia cortical thickness (mm)4.40 ± 0.514.44 ± 0.51.20
 Tibia periosteal circumference (mm)75.2 ± 5.074.9 ± 4.8.29
 Tibia endosteal circumference (mm)47.6 ± 5.746.9 ± 5.4.10

Previous fracture was associated with vBMD

We next used pQCT to investigate whether the difference in aBMD between men with and without prevalent fracture was due to alterations in bone size or vBMD of the cortical and/or trabecular bone compartments. Men with prevalent fractures had 6.6% lower trabecular vBMD of the radius and 4.5% lower trabecular vBMD of the tibia than men without fracture. They also had a slightly lower cortical vBMD of the radius and the tibia, 0.4% and 0.3%, respectively (see Table 2). There was no significant difference in cortical CSA or in periosteal circumference between subjects with and without prevalent fracture. Subjects with prevalent fracture had a slightly larger endosteal circumference and reduced cortical thickness of the radius but not of the tibia than subjects without fracture (see Table 2).

We assessed the association among aBMD, vBMD, bone geometry, and prevalent fractures with binary logistic regression, both unadjusted and adjusted for covariates (i.e., present age, height, weight, calcium intake, smoking, and physical activity). Low aBMD was associated with previous fracture [crude odds ratio (OR) 1.22–1.26 and adjusted OR 1.35–1.50, depending on assessed bone site] (Table 3).

Table 3. Volumetric BMD (pQCT) and Areal BMD (DXA) Are Associated With Prevalent Fracture in Young Adult Men
 Crude odds ratio (95% CI)Adjusted odds ratio (95% CI)
  1. Odds ratio (OR) and 95% CI/SD decrease in each bone parameter are presented. ORs are calculated using binary logistic regression analysis. Adjusted ORs were obtained using age, adult physical activity, smoking, calcium intake, height, and weight as covariates. In total, 304 men with prevalent fracture and 687 men without previous fracture were included in the analysis.

DXA
 Total body aBMD (g/cm2)/SD decrease1.22 (1.06–1.40)1.50 (1.26–1.79)
 Lumbar spine aBMD (g/cm2)/SD decrease1.22 (1.06–1.40)1.35 (1.15–1.57)
 Total femur aBMD (g/cm2)/SD decrease1.22 (1.06–1.40)1.35 (1.16–1.59)
 Radius nondominant aBMD (g/cm2)/SD decrease1.26 (1.10–1.45)1.40 (1.20–1.64)
pQCT
 Radius cortical vBMD (mg/cm3)/SD decrease1.25 (1.10–1.43)1.27 (1.11–1.47)
 Radius trabecular vBMD (mg/cm3)/SD decrease1.46 (1.26–1.69)1.51 (1.30–1.76)
 Radius cortical CSA (mm2)/SD decrease1.10 (0.96–1.25)1.22 (1.04–1.42)
 Radius cortical thickness (mm)/SD decrease1.27 (1.10–1.46)1.34 (1.16–1.55)
 Radius periosteal circumference (mm)/SD decrease0.93 (0.82–1.07)1.00 (0.85–1.16)
 Radius endosteal circumference (mm)/SD decrease0.82 (0.72–0.94)0.84 (0.73–0.98)
 Tibia cortical vBMD (mg/cm3)/SD decrease1.19 (1.04–1.36)1.17 (1.02–1.35)
 Tibia trabecular vBMD (mg/cm3)/SD decrease1.46 (1.26–1.68)1.55 (1.33–1.82)
 Tibia cortical CSA (mm2)/SD decrease1.02 (0.89–1.17)1.16 (0.98–1.37)
 Tibia cortical thickness (mm)/SD decrease1.09 (0.95–1.25)1.14 (0.99–1.33)
 Tibia periosteal circumference (mm)/SD decrease0.93 (0.81–1.06)1.03 (0.86–1.23)
 Tibia endosteal circumference (mm)/SD decrease0.89 (0.78–1.02)0.93 (0.80–1.09)

Especially trabecular vBMD but also cortical vBMD of the radius and tibia were associated with previous fracture. Every SD decrease in trabecular vBMD of the radius and tibia was associated with 1.46 (unadjusted) and 1.51 and 1.55 (adjusted) times increased fracture prevalence, respectively (see Table 3). CSA of the radius but not of the tibia was associated with previous fracture, but only after adjustments for covariates was made (see Table 3). Cortical thickness of the radius, but not of the tibia, was associated with prevalent fracture both before and after adjustments for covariates (see Table 3).

Trabecular vBMD but not aBMD is independently associated with previous fracture

In order to determine whether aBMD (measured with DXA) or trabecular vBMD (measured with pQCT) of the radius independently predicted the risk of any fracture, a binary logistic regression analysis was used, including both these bone variables, together with present age, height, weight, calcium intake, smoking, and physical activity as covariates. In this analysis, trabecular vBMD was significantly associated with prevalent fracture per SD decrease (OR = 1.42; 95% CI 1.17–1.72), whereas aBMD (OR = 1.11; 95% CI 0.91–1.36) was not. Furthermore, we also analyzed the associations between these bone variables and prevalent fracture of the radius, a bone measured with both techniques. In total, 110 subjects had a verified fracture of the distal radius. We assessed the association among aBMD, vBMD, bone geometry, and prevalent fracture for subjects with distal radius fracture with binary logistic regression. Low aBMD and reduced cortical thickness of the radius were associated with previous fracture of the distal radius, but the association was stronger for vBMD and particularly trabecular vBMD, adjusted OR = 1.86 per SD decrease (Table 4). An additional regression analysis also was performed using both aBMD and trabecular vBMD of the radius, together with all covariates as earlier, simultaneously in the logistic regression analysis, aiming to determine which of these bone parameters was most strongly associated with prevalent fracture. In this analysis, only trabecular vBMD (OR = 2.00 per SD decrease; 95% CI 1.49–2.69; p = 3.93 × 10−6) but not aBMD (OR = 0.88 per SD decrease; 95% CI 0.65–1.19; p = 0.39) was associated with prevalent distal radius fracture.

Table 4. Volumetric BMD (pQCT) Is Associated With Previous Fracture of the Radius
 Crude odds ratio (95% CI)Adjusted odds ratio (95% CI)
  1. Odds ratio (OR) and 95% CI/SD decrease in each bone parameter are presented. Unadjusted (crude) ORs and ORs adjusted for age, adult physical activity, smoking, calcium intake, height, and weight are shown. In total, 110 men with prevalent fracture of the radius and 687 men without previous fracture were included in the analysis.

DXA
 Radius non-dominant aBMD (g/cm2)/SD decrease1.26 (1.02–1.55)1.38 (1.09–1.74)
pQCT
 Radius cortical vBMD (mg/cm3)/SD decrease1.54 (1.27–1.87)1.61 (1.31–1.97)
 Radius trabecular vBMD (mg/cm3)/SD decrease1.78 (1.42–2.23)1.86 (1.47–2.35)
 Radius cortical thickness (mm)/SD decrease1.33 (1.08–1.63)1.37 (1.11–1.71)
 Radius cortical CSA (mm2)/SD decrease1.11 (0.91–1.36)1.19 (0.94–1.49)
 Radius periosteal circumference (mm)/SD decrease0.94 (0.77–1.15)0.96 (0.77–1.21)
 Radius endosteal circumference (mm)/SD decrease0.81 (0.67–0.99)0.81 (0.65–1.00)

In order to eliminate the risk of assessing vBMD at a previously fractured site, which could confound our found associations, we divided the 110 subjects with distal radius fracture into two groups, one with fracture of the measured arm (n = 80) and one with fracture of the nonmeasured arm (n = 30). Binary logistic regression was performed as earlier to evaluate the association between bone variables of the radius and previous fracture of the nonmeasured arm. The association between trabecular vBMD and prevalent fracture remained (crude OR = 1.90 per SD decrease; 95% CI 1.26–2.86; p = .0023; adjusted OR = 2.20; 95% CI 1.41–3.46; p = .00057).

Trabecular vBMD was associated with fractures prior to and at the age of PHV

Binary logistic regression was performed, adjusted as earlier, in the 600 subjects in which PHV was available to determine if vBMD was associated most strongly with prevalent fractures before, around, or after the age of PHV. Trabecular vBMD of the radius and tibia was associated with fractures more than 2 years prior to PHV (95 fractures; radius OR = 1.80; 95% CI 1.38–2.34) and tibia OR = 1.79; 95% CI 1.37–2.35 per SD decrease] and at the time of PHV (PHV ± 2 years; 84 fractures; radius OR = 1.63; 95% CI 1.24–2.13; tibia OR = 1.54; 95% CI 1.16–2.03) but not with fractures that occurred more than 2 years after PHV (20 fractures; radius OR = 1.04; 95% CI 0.64–1.67; tibia OR = 1.27; 95% CI 0.77–2.10). Cortical vBMD was associated with fractures more than 2 years prior to PHV (radius OR = 1.30; 95% CI 1.03–1.65; tibia OR = 1.36; 95% CI 1.08–1.70 per SD decrease) but not at the time of PHV (PHV ± 2 years; radius OR = 1.23; 95% CI 0.96–1.56; and tibia OR = 1.21; 95% CI 0.96–1.53), or after PHV (radius OR = 0.89; 95% CI 0.64–1.67; tibia OR = 1.12; 95% CI 0.71–1.75).

Discussion

In this large, population-based cohort study we investigated the relationship between X-ray-verified prevalent fractures in childhood and adolescence, aBMD, and vBMD, as well as cortical bone geometry, measured using both DXA and pQCT. We found that low trabecular vBMD was independently associated with previous fracture. Subjects with a prevalent fracture had 6.6% and 4.5% lower trabecular vBMD of the radius and tibia, respectively, than subjects without fracture. There was an increased fracture prevalence of 51% (radius) and 55% (tibia) per SD lower trabecular vBMD. Furthermore, we demonstrated that the peak fracture incidence coincides with the age of PHV and that primarily trabecular vBMD is associated with fractures occurring at the time of PHV.

We also found that previous fracture was associated with low aBMD, results that are consistent with previous studies in adults2, 3, 25 and in children, as discussed in the meta-analysis by Clark and colleagues from 2006.4 In our study, BMD assessed by pQCT allowed us to evaluate which component of aBMD was most strongly associated with previous fracture. In order to optimize this evaluation, we studied the radius, a bone measured by both techniques (DXA and pQCT). When analyzing what could best predict previous fracture of the radius, trabecular vBMD was the strongest predictor. When we included both trabecular vBMD and aBMD of the radius in the same regression analysis, only trabecular vBMD was associated with prevalent fracture, suggesting that it is trabecular vBMD that mediates the association seen with aBMD. One previous study using CT (but not pQCT) by Skaggs and colleagues21 reported smaller cross-sectional area at the distal radius but similar cancellous, integral, and cortical bone densities when comparing a group of 50 girls with distal forearm fracture with control subjects. Although we observed in our cohort that men with prevalent radius fracture had lower cortical thickness, we found no difference in the cortical cross-sectional area between the groups. There are several possible reasons why these results differ from the ones we present in this study. First, CT measurements were obtained with a CT scanner and not with a pQCT device. Second, both study subjects and control subjects were overweight (on average 12 kg), and perhaps the results are restricted to overweight children. Third, the previous study was small and consisting of younger (9.6 ± 3.04 years of age) girls, suggesting that possible gender and age differences could have influenced the discrepant associations seen in the previous study21 and our study.

Reduced trabecular vBMD among subjects with fracture was seen in studies of postmenopausal women with distal forearm fracture and vertebral fracture by Melton and colleagues,26, 27 although in these studies there were other, even stronger predictors of fracture such as total radius vBMD, cortical thickness, and axial rigidity for radius and total vertebral vBMD, cortical thickness, and applied-load-to-bone-strength ratio for the spine. These differences imply that for younger individuals, such as the cohort of our study, trabecular vBMD is relatively more important, whereas in older individuals, such as those studied by Melton and colleagues, cortical bone contributes to the bone's ability to resist fracture to a greater extent. The studies by Melton and colleagues used high-resolution pQCT, allowing measurement of micro- and macrostructural variables of the measured bone. It was reported recently that in boys, cortical thickness at the distal radius was low and maintained between pre- to midpuberty before increasing substantially toward the end of puberty.28 In our cohort, cortical thickness was associated with prevalent radius fracture, suggesting that both reduced cortical thickness and lower trabecular vBMD could be responsible for the peak in fracture incidence at the time of PHV. The reduced cortical thickness was a result of increased endosteal circumference, whereas the outer bone dimensions, as reflected by periosteal circumference, were not lower in men with prevalent fractures. To learn more about which components of bone structure are associated with fractures in adolescents and young adults, prospective studies including high-resolution pQCT are needed.

Interestingly, there was a large discrepancy between self-reported fractures and fractures that could be verified in radiology records in our cohort. A previous study of elderly women showed that 20% of self-reported fractures could not be confirmed20; in this study, that number was even greater, namely, 25% (77/304). Possible explanations could be injury while on holiday or having moved during childhood (thus no radiology records being available in clinics in the Gothenburg area), or being informed by a doctor that there might be a fracture even though the X-ray was negative or uncertain, or simply not remembering correctly. There were also 78 study participants who had not reported fracture who were found to have had a fracture in the radiology records. A corresponding group of false-negative reports was not found in the previously mentioned study by Nevitt and colleagues.20 Since we questioned only the young men and not their parents/guardians, failure to recollect fractures that occurred early in life probably was the main reason for this underreporting. For our analysis, we used only X-ray-verified fractures, which adds strength to our results.

The importance of dexterity for radius fracture risk has not, to our knowledge, been investigated. When analyzing the distribution of radius fractures, we found that prevalent fractures of the nondominant radius (mostly the left side) were clearly overrepresented in our young men. We have demonstrated previously that aBMD is greater in the dominant than in the nondominant arm in our cohort,22 indicating that lower aBMD could have contributed to this fracture distribution.

Previous studies have reported that the age of the peak fracture incidence coincides with the age of PHV, but to the best of our knowledge, fracture incidence during growth has not been investigated directly in relation to PHV.9, 10 In our cohort, the peak fracture incidence coincided with the age of PHV, indicating that the skeleton is most susceptible to fracture when longitudinal growth is at its peak. Body size–corrected aBMD has been reported to decrease to its minimum, in both males and females, during PHV,29 offering a plausible explanation for the peak in fracture incidence at this age. In the subsample of men in whom PHV was available in our cohort, we found that trabecular vBMD, but not cortical vBMD, was associated with fractures during PHV, indicating that reduced trabecular vBMD is important for fracture risk during maximum longitudinal growth. We speculate that a reduction in trabecular vBMD, rather than cortical vBMD, could be responsible for the previously reported reduction in aBMD at the time of PHV.

There are some limitations to this study. We could not distinguish which fractures were caused by slight, moderate, or severe trauma. Previously, it was thought that only low-trauma fractures were related to underlying skeletal fragility. However, recent studies by Mackey and colleagues showed that low BMD is associated with high-trauma fractures in older women and men,30 and Clark and colleagues showed that underlying skeletal fragility contributes to fracture risk in children, even after moderate/severe trauma (although greater skeletal fragility was seen in children who fracture because of slight trauma).31 Furthermore, in the material used by Clark and colleagues, only 2.9% of fractures were caused by severe trauma, suggesting that this group tends to be small in children. Even though we performed a very thorough search of all available X-ray registers and archives, it is possible that some fractures were not detected using our methodology. In order to minimize the risk of false associations, we excluded the 77 men with self-reported but not X-ray-verified fractures from the present study. Another limitation is that our measurements of bone mass were performed after the subjects had fractured, suggesting that we cannot exclude that the previous fracture has caused a reduced bone mass. For most fractured subjects in our material, several years had passed between the fracture and the measurement, which increases the likelihood that our findings represent long-term bone mass rather than a reduced bone mass caused by the fracture itself. More important, we found when comparing subjects with radius fracture of the nonmeasured arm with subjects without fracture that the association between trabecular vBMD and prevalent fracture remained, suggesting that there is a true difference in vBMD between the fracture and nonfracture groups that is not dependent on the previous fracture. Furthermore, trabecular vBMD of the tibia, a bone not commonly fractured, also was independently associated with prevalent fractures, arguing that the association between trabecular vBMD and prevalent fractures was not due to impaired bone growth owing to previous fracture. To minimize the risk of previous fracture having an impact on bone variables, prospective studies are needed.

In summary, the results of this study demonstrate that trabecular vBMD but not aBMD is independently associated with prevalent X-ray-verified fracture. Further studies are needed to determine if assessment of trabecular vBMD could enhance prediction of fractures during growth in males.

Disclosures

AD and CO contributed equally to this study. The authors state that they have no conflicts of interest.

Acknowledgements

This study was supported by the Swedish Research Council, the Lundberg Foundation, and ALF/LUA grants from the Sahlgrenska University Hospital.

Ancillary