Prevalent fractures are related to cortical bone geometry in young healthy men at age of peak bone mass



Low areal bone mass is a risk factor for fractures in men. Limited data are available on fractures and bone geometry in men, and the relation with sex steroids is incompletely understood. We investigated prevalent fractures in relation to peak bone mass, bone geometry, and sex steroids in healthy young men. Healthy male siblings (n = 677) at the age of peak bone mass (25 to 45 years) were recruited in a cross-sectional population-based study. Trabecular and cortical bone parameters of the radius and cortical bone parameters of the tibia were assessed using peripheral quantitative computed tomography (pQCT). Areal bone mineral density (aBMD) was determined using dual-energy X-ray absorptiometry (DXA). Sex steroids were determined using immunoassays, and fracture prevalence was assessed using questionnaires. Fractures in young men were associated with a longer limb length, shorter trunk, lower trabecular BMD, smaller cortical bone area, and smaller cortical thickness (p < .005) but not with bone-size-adjusted volumetic BMD (vBMD). With decreasing cortical thickness [odds ratio (OR) 1.4/SD, p ≤ .001] and decreasing cortical area (OR 1.5/SD, p ≤ .001), fracture odds ratios increased. No association between sex steroid concentrations and prevalent fractures was observed. Childhood fractures (≤15 years) were associated with a thinner bone cortex (−5%, p ≤ .005) and smaller periosteal size (−3%, p ≤ .005). Fractures occurring later than 15 years of age were associated with a thinner bone cortex (−3%, p ≤ .05) and larger endosteal circumference (+3%, p ≤ .05) without differences in periosteal bone size. In conclusion, prevalent fractures in healthy young men are associated with unfavorable bone geometry and not with cortical vBMD when adjusting for bone size. Moreover, the data suggest different mechanisms of childhood fractures and fractures during adult life. © 2010 American Society for Bone and Mineral Research


Bone mass at any stage in adult life is determined by the growth and mineralization of the skeleton during the first two decades of life leading to the peak bone mass (PBM) and by subsequent net changes in bone mass, usually bone loss, starting in the middle of the third decade.1 Although fracture risk in old age is lower in men than in women, fracture risk is higher in boys and young men than in girls or young women.2, 3 In both young and older subjects, low areal bone mineral density (aBMD) has been associated with increased fracture prevalence, although most evidence has been obtained in women4, 5 or elderly men,6 and only limited data are available in boys or young men.7–9

An important limitation of dual-energy X-ray absorptiometry (DXA) is that it measures projected aBMD and not volumetric BMD (vBMD). In determining PBM and future facture risk, this has important clinical implications because bone strength depends not only on the amount of bone tissue but also on the diameter, shape, and volumetric bone density, as well as the distribution of cortical and trabecular bone.10 Using peripheral quantitative computed tomography (pQCT), more detailed information on bone mass, volumetric bone density, and geometry can be obtained.

Recently, we demonstrated lower trabecular and cortical bone density as well as a thinner bone cortex owing to a larger endosteal circumference in a cohort of men with idiopathic osteoporosis, with a similar phenotype in their adult sons, providing evidence for a deficient acquisition of PBM in these men.11, 12 These studies suggested a deficient estrogen exposure as a contributing pathophysiologic mechanism. Several reports in men have illustrated the importance of sex steroids and specifically estradiol for the determination of bone mass and subsequent fracture risk,13–17 whereas data on fracture risk in young men are scarce. Based on these findings, we hypothesized that less favorable bone geometry, possibly related to lower estrogen action during growth and maturation, might be associated with fracture risk in young men.

We investigated bone fracture prevalence in relation to anthropometry, volumetric bone density, and bone geometry using pQCT together with gonadal steroids in young healthy male subjects.

Materials and Methods

Study design and population

Seven-hundred and sixty-seven young men were recruited from three semirural to suburban communities around Ghent, Belgium. Male brothers aged 25 to 45 years were contacted by mail in 2002–2004. The study design has been described previously.12, 16, 18 After exclusion, 296 pairs of brothers were included in the study. Sixty-four men were included as single participants when their brother could not participate in the study, and 19 men were included as third brother in a family and 2 as fourth brother. The maximal age difference between brother pairs was set at 12 years. All participants were in good health and completed questionnaires about previous illness, lifestyle, physical activity, education, profession, nutrition, and smoking.

Exclusion criteria were defined as illnesses or medication use affecting body composition or hormone or bone metabolism: current use (<3 months) of glucocorticosteroids, antiandrogens, vitamin D supplements, insulin, thyroxine, previous or current use of antiepileptic drugs, hypogonadism, hyperthyroidism, cystic fibrosis, malabsorption or eating disorders, disorders of collagen metabolism or bone development, chronic renal failure, alcohol abuse, and autoimmune rheumatoid disease. The study protocol was approved by the Ethical Committee of the Ghent University Hospital, and informed consent was obtained from all participants.

Current calcium intake was estimated by a food questionnaire evaluating weekly averages of dairy products. Smoking habits were registered as current and previous smoking. Standing height was measured using a wall-mounted Harpenden stadiometer (Holtain, Ltd., Crymuch, UK). Body weight was measured in light indoor clothing without shoes. Anthropometric measurements were performed as described previously12 according to the Antropometric Standardization Reference Manual.19 Standing and truncal heights (in the sitting position using a standard chair: distance from sitting surface to jugular notch) were measured to the nearest 0.1 cm using a wall-mounted Harpenden stadiometer. Arm span was defined as the distance between the tips of the middle (longest) fingers, when both arms are extended laterally to the level of the shoulder.19 Tibia length was determined as the distance between the knee joint line and the tip of the medial malleolus.19 Questionnaires were used to collect data on previous fractures. Fracture data consisted of the prevalence of previous fractures on different sites and age at fracture.


A pQCT device (XCT-2000, Stratec Medizintechnik, Pforzheim, Germany) was used to scan the dominant leg (tibia) and forearm (radius). The dominant side was selected to allow assessment of the relationship between muscle area and bone parameters. The cortical volumetric bone mineral density (vBMD, mg/cm3), cortical cross-sectional area (mm2), endosteal and periosteal circumferences (mm), and cortical thickness (mm) were measured at the midradius (66% of bone length from the distal end) and tibia (66%). Trabecular vBMD (mg/cm3) was measured using a scan through the metaphysis (at 4% of bone length) at the nondominant radius. Adjusted cortical density to bone size (partial volume correction) was calculated according to previously published formulas.20

Biochemical determinations

Venous blood samples were obtained between 8 and 10 a.m. after overnight fasting. All samples were stored at −80°C until batch analysis. Commercial immunoassays were used to determine serum concentrations of testosterone and sex hormone–binding globulin (SHBG, Orion Diagnostica, Espoo, Finland), and estradiol (Clinical Assay, DiaSorin s.r.l., Saluggia, Italy) according to a modified protocol that doubles the serum amount.16, 18 Serum free and bioavailable fractions of testosterone and estradiol were calculated from serum total testosterone, total estradiol, SHBG, and albumin concentrations using a previously validated equation derived from the mass-action law.21, 22 The intra- and interassay coefficients of variation (CVs, %) were below 10% and 15% for all measurements, respectively.


Descriptives are expressed as mean ± SD or median (1st to 3rd quartile) when criteria for normality were not fulfilled (Kolmogorov-Smirnov), and variables (ie, bone parameters, steroid concentrations) were log-transformed in subsequent linear models. Linear mixed-effects modeling with random intercepts and a simple residual correlation structure for random effects was used to evaluate cross-sectional relationships in our study population, taking the interdependence of measurements within families into account. Logistic regression, taking into account the family structure, was modeled in S-Plus using the glme function. Parameters of fixed effects were estimated via restricted maximum-likelihood estimation and reported as estimates of effect size (β) with their respective standard error. Associations were considered significant at p values less than .05. Statistical analyses were performed using S-Plus 7.0 (Insightful, Seattle, WA, USA) and SAS 9.1.3 Software (SAS Institute, Inc., Cary, NC, USA).


General characteristics

Six-hundred and seventy-seven subjects with a mean age of 34 ± 6 years were included in the study. Mean height was 179 ± 6 cm, and mean weight was 81 ± 12 kg, with a body mass index (BMI) of 25 ± 4 kg/m2.

Finger, toe, and cranial fractures were excluded from the analysis. One or more clinically significant bone fractures were reported in 33% (n = 219) of the subjects, with 16% (n = 108) sustaining a wrist fracture. Other fractures included shoulder (n = 35, 5%), humerus (n = 23, 4%), and lower limb (n = 32, 7%). Four participants had a hip fracture (0.6%) after major trauma. Most fractures were associated with occupational or sport activities, although the intensity of trauma could not be quantified. The peak incidence of fractures occurred in this cohort between 14 and 16 years of age. One-hundred and five subjects reported a first fracture prior to age 15 years. Table 1 summarizes the general characteristics of our study population according to prevalent fractures. Smoking was found to be associated with fracture prevalence, as described previously, in this cohort.18 The odds ratio (OR) for fractures associated with current smoking was 2.13 [95% confidence interval (CI) 1.43–3.18].18

Table 1. Descriptive Anthropometric and Hormonal Parameters According to Previous Fractures
 No previous fracturePrevious fracture (n = 219)
  • Descriptives are expressed as mean ± SD or median (interquartile range).

  • a

    p ≤ .005;

  • b

    p ≤ .01;

    cp ≤ .05 (all analyses adjusted for age, height, weight);

  • d

    p = .008 (chi-square).

Age (years)34.3 ± 5.534.9 ± 5.5
Height (cm)179 ± 7179 ± 6
Weight (kg)81 ± 1282 ± 11
BMI (kg/m2)25.2 ± 3.625.6 ± 3.4
Arm span (cm)182 ± 7.6183 ± 6.5a
Truncal height (cm)61.6 ± 2.761.3 ± 2.6b
Tibia length (cm)38.3 ± 2.438.6 ± 1.9b
Current smoking (%)19.7%29.6%d
Daily calcium intake (mg)572 (425–755)551 (424–770)
Testosterone (ng/dL)570 ± 146559 ± 144
Free testosterone (ng/dL)13.7 ± 3.213.8 ± 3.3
Estradiol (pg/mL)20.2 ± 4.820.3 ± 4.5
Free estradiol (pg/mL)0.40 ± 0.100.41 ± 0.10

Anthropometrics and sex steroids in relation to prevalent fractures

A slightly larger arm span and limb length and shorter truncal height were observed in subjects with fractures compared with subjects with no previous fractures (Table 1, adjusted for age, height, and weight). Subjects with childhood fractures exhibited a shorter trunk (61.1 ± 2.6 cm, p = .013) and larger tibial length (38.7 ± 1.9 cm, p = .018) than subjects without fracture. Fracture ORs increased with higher limb length or arm span but were inversely related to truncal height (Table 2).

Table 2. Logistic Regression Models Predicting Self-Reported Fractures in Normal Healthy Males According to Anthropometrics
 All fractures (prevalence OR/1 SD increase)Childhood fractures (prevalence OR/1 SD increase)
  1. Note: Results from mixed-effects models to account for family structure and adjusted for age, height, and weight (except models for height and weight: adjusted for age).

Height1.01 (0.85–1.20)1.02 (0.85–1.20)
 p = .89p = .88
Weight1.08 (0.92–1.28)1.02 (0.92–1.28)
 p = .33P = .82
Arm span (cm)1.47 (1.07–2.02)1.07 (0.72–1.60)
 p = .018p = .72
Tibia length (cm)1.54 (1.15–2.05)1.61 (1.14–2.28)
 p = .015p = .007
Truncal height (cm)0.68 (0.49–0.92)0.63 (0.43–0.92)
 p = .014p = .018

Total (free) serum testosterone and estradiol concentrations were comparable in subjects with or without fractures, considering all fractures (Table 1) or childhood fractures (data not shown). Fracture ORs were unrelated to sex steroid concentrations (data not shown).

Prevalent fractures and areal bone parameters

The areal bone mass indices according to fracture status are given in Table 3. In general, a lower aBMD is observed in spine, hip, and whole-body DXA measurements in subjects with previous or childhood fractures. The bone areas as determined by DXA were not different, except for a lower bone area on whole-body DXA measurements in subjects with a childhood fracture.

Table 3. Areal Bone Mass Indices According to Reported Prevalent Fractures
DXA measurementsNo previous fracturePrevious fracture (n = 219)#Childhood fracture (≤15 years) (n = 105)Fracture (>15 years) (n = 107)
  • a

    p ≤ .005;

  • b

    p ≤ .01;

  • c

    p ≤ .05, adjusted for age, height and weight; no previous fracture as reference.

  • No accurate time of fracture in 7 subjects.

Whole body
 Total area (cm2)2358 ± 1632343 ± 1382322 ± 134a2360 ± 138
 BMC (g)2904 ± 3782822 ± 354c2791 ± 332a2851 ± 380
 BMD (g/cm2)1.228 ± 0.0971.201 ± 0.101b1.199 ± 0.093c1.204 ± 0.11
Lumbar spine
 Total area (cm2)71.7 ± 6.471.4 ± 6.370.9 ± 6.171.8 ± 6.5
 BMC (g)77.0 ± 12.574.0 ± 12.5c73.7 ± 12.0c74.2 ± 13.2
 BMD (g/cm2)1.071 ± 0.1211.033 ± 0.123a1.038 ± 0.122c1.029 ± 0.128c
Total hip
 Total area (cm2)45.2 ± 4.545.5 ± 4.045.1 ± 3.845.8 ± 4.2
 BMC (g)49.2 ± 8.148.2 ± 7.847.5 ± 7.049.0 ± 8.4
 BMD (g/cm2)1.089 ± 0.1321.061 ± 0.1431.053 ± 0.131c1.071 ± 0.155

Volumetric bone parameters in relation to prevalent fractures

Table 4 gives the descriptive pQCT data of the distal (4%) and midshaft (66%) radius and tibia (66%). All values are within the expected normal range for young male subjects.

At the radius, periosteal bone circumference was comparable between subjects with or without previous fractures, whereas fracture was associated with a lower cortical thickness and lower cortical bone area (Table 4). In subjects with childhood fractures, we observed a smaller periosteal bone size and cortical bone area and a lower cortical thickness. Adjusting for periosteal bone size, the relative endosteal circumference was found to be larger in subjects with childhood fractures (β = 0.53 ± 0.15, p = .003). Fractures during young adulthood (>15 years of age) were not associated with periosteal bone size but with a smaller cortical thickness and a larger endosteal circumference.

Significantly lower trabecular and cortical vBMD values were observed in subjects who had sustained a previous fracture, although the difference in cortical vBMD disappeared after adjustment for bone size (partial-volume correction). At the lower limb, no differences in bone size or cortical thickness were observed, whereas cortical density was lower in subjects with previous fracture, although significance was lost after partial-volume correction (Table 4). The ORs for prevalent fractures, based on pQCT measurements, are given in Fig. 1. With decreasing cortical thickness and cortical bone area, we observed a marked and gradual increase in fracture ORs. At the radius, increasing fracture ORs were found with decreasing trabecular density (Fig. 1). Lower cortical vBMD at both the radius and tibia were associated with increased fracture risk (radius: OR 1.2/1 SD decrease, p = .02; tibia: OR 1.2/1 SD decrease, p = .02). After adjustment for bone size (partial-volume correction), these associations were no longer significant. Periosteal or endosteal circumference was not predictive for all fractures. However, a smaller periosteal circumference was associated with increased fractures during childhood (Tables 4 and 5), whereas a larger endosteal circumference was associated with fractures occurring later than 15 years of age.

Figure 1.

Fracture prevalence odds ratios (ORs) according to trabecular and cortical bone parameters at both radius and tibia. ORs are presented by quartiles of bone parameters, with the highest quartile (Q4) as reference and adjusted for age, height, and weight.

Table 4. Volumetric Bone Mass Indices According to Reported Prevalent Fractures and Smoking Habits at Radius and Tibia
pQCT measurementsNo previous fracturesPrevious fracture (n = 219)*Childhood fracture (≤ 15 years) (n = 105)Fracture (>15 years) (n = 107)
  • a

    p ≤ .005;

  • b

    p ≤ .01;

  • c

    p ≤ .05, adjusted for age, height and weight; no previous fracture as reference.

  • *

    No accurate time of fracture in 7 subjects.

  • **

    Adjusted for bone size (partial-volume correction).

 Trabecular bone area (mm2)188 ± 26186 ± 26182 ± 26188 ± 25
 Trabecular bone density (mg/cm3)232 ± 40222 ± 38a223 ± 36c222 ± 41
 Cortical bone density (mg/cm3)1103 ± 351096 ± 35c1097 ± 341094 ± 35
 Corrected cortical bone density (mg/cm3)**1319 ± 221320 ± 201322 ± 201317 ± 21
 Cortical bone area (mm2)103 ± 1499 ± 12a96 ± 12a101 ± 12
 Periosteal circumference (mm)48.4 ± 3.848.0 ± 3.747.1 ± 3.6a48.9 ± 3.6
 Endosteal circumference (mm)32.3 ± 4.432.6 ± 4.531.7 ± 4.133.4 ± 4.7a
 Cortical thickness (mm)2.55 ± 0.322.46 ± 0.31a2.44 ± 0.28a2.48 ± 0.34c
 Cortical bone density (mg/cm3)1114 ± 241108 ± 35c1112 ± 231106 ± 25b
 Corrected cortical bone density (mg/cm3)**1244 ± 211242 ± 221243 ± 201240 ± 23
 Cortical bone area (mm2)366 ± 46359 ± 49357 ± 45362 ± 54
 Periosteal circumference (mm)94.8 ± 6.294.8 ± 5.593.9 ± 5.495.5 ± 5.6
 Endosteal circumference (mm)66.1 ± 6.966.8 ± 6.465.7 ± 6.267.5 ± 6.5
 Cortical thickness (mm)4.56 ± 0.544.46 ± 0.614.49 ± 0.554.45 ± 0.66
Table 5. Logistic Regression Models Predicting Self-Reported Fractures During Childhood and Young Adulthood in Normal Healthy Males According to Radial Bone Geometry (Adjusted for Age, Height, and Weight) as Determined by pQCT
 Childhood fractures (<15 years; prevalence OR/1 SD decrease)Fracture > 15 years; prevalence OR/1 SD decrease
Cortical thickness (mm)1.66 (1.31–2.10)1.40 (1.12–1.74)
 p < .0001p = .003
Cortical bone area (mm2)1.93 (1.52–2.45)1.28 (1.01–1.62)
 p < .0001p = .04
Periosteal circumference (mm)1.46 (1.14–1.85)0.88 (0.70–1.01)
 p = .002p = .28
Endosteal circumference (mm)1.1 (0.86–1.35)0.78 (0.62–0.97)
 p = .50p = .025

Although no associations between sex steroid concentrations and fractures at any of the reported sites were found (Table 1; specific site data not shown), sex steroids were important determinants of bone geometry and density, as described previously.16 Free estradiol was positively associated with cortical vBMD and cortical thickness and inversely associated with endosteal circumference at the radius.16


In this population of healthy men at age of peak bone mass, we demonstrated that prevalent fractures were associated with a lower cortical area, thinner bone cortex, and lower trabecular density. In particular, childhood fractures were associated with a thinner bone cortex owing to a smaller bone size (periosteal circumference), whereas in subjects with a first fracture occurring later than age 15 years we observed a thinner bone cortex but mainly owing to a wider endosteal circumference (schematic representation in Fig. 2), providing evidence for a different mechanism of fractures during childhood and in adult life.

Figure 2.

Schematic representation of our findings relating bone parameters to fracture risk. All fractures were associated with a marked smaller cortical thickness. In childhood fractures, we observed a decreased periosteal bone circumference, whereas in subjects with a first fracture in young adulthood, a larger endosteal circumference was observed.

As reported previously,18 the peak fracture incidence in our cohort is observed around 14 years of age. This observation is in line with literature data because 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.2, 3, 23–25 Underlying this observation is a transient increase in bone fragility and cortical porosity leading to a deficit in bone mass relative to longitudinal growth.26, 27 In this regard, the peak height velocity in both boys and girls precedes the time of peak increase in bone mineral content by 1 year.28, 29

Several studies have demonstrated low areal bone mass in children with fractures compared with healthy controls. Two earlier studies using single-photon absorptiometry demonstrated a reduction of about 10% in bone mineral content (BMC) at the forearm in children with fractures.30, 31 Larger case-control studies in both girls and boys (3 to 19 years of age) indicated that prevalent distal forearm fractures are associated with significantly lower BMC and aBMD values at the radius but also at the lumbar spine and hip compared with age-matched controls.6, 32 In prospective cohorts of young girls, low areal bone mass was a risk for incident fracture, and the observed low bone mass was found to persist after puberty.33, 34 Moreover, in girls with childhood fractures, a lower BMC, smaller bone area, and smaller bone width as assessed by DXA were observed after puberty (16 years of age) but also prior to menarche (9 years of age). These findings provide evidence that childhood fractures are associated with low peak bone mass and could be a marker for persistent bone fragility in women.35 In men and boys, information on aBMD and fractures is more limited, but the available evidence suggests that low aBMD is a risk for fracture.6, 7 Using pQCT, a recent study36 in young men (18 to 19 years of age) demonstrated a marked lower trabecular density and only minor differences in cortical thickness in men with a previous fracture. However, our data indicate that a smaller periosteal circumference and thinner cortex at the radius are predominantly associated with fractures during childhood, whereas cortical vBMD, adjusted for bone size, is not. At the tibia, trends toward a thinner cortex in subjects with fractures were observed, although these were not significant. Previously, we have reported differences between weight-bearing sites and non-weight-bearing sites.16, 18 Moreover, the cortex at the tibia is much thicker and is not a predisposition site for fracture during growth or adult life. Furthermore, whether the delay in bone maturation relative to growth as observed at the radius during early puberty28, 29 also occurs at the tibia has not been studied in detail.

In contrast to the findings in subjects with childhood fractures, in those with a first fracture occurring after 15 years of age, we observed a smaller cortical thickness owing to a larger endosteal circumference and not a smaller periosteal circumference. Based on our data, we hypothesize that disturbances in the action of sex steroids during growth give rise to a thinner bone cortex and suboptimal acquisition of peak bone mass that increases the risk of fracture in adulthood. In line with our reports suggesting a disturbed estrogen action in male osteoporosis patients,11, 12 we hypothesized that low estradiol levels could be a risk factor for fracture in our cohort of healthy young men. Indeed, low estradiol concentrations were found to be associated with unfavorable bone geometry and vBMD in both healthy men16, 37 and men with idiopathic osteoporosis.11, 12

Moreover, the pQCT-derived bone characteristics (ie, wider endosteal cavity, thinner bone cortex, and lower trabecular density) associated with fractures are determined by sex steroids, more specifically by estradiol, as demonstrated previously in this cohort.16 In addition, in elderly men, both free testosterone and estradiol were negatively associated with fracture risk.14, 15, 38, 39 Although these arguments strongly suggest a role for estradiol in the acquisition of peak bone mass, we could not demonstrate a relationship between prevalent fractures and estradiol concentrations at adult age. A possible explanation could be the low fracture incidence in men of this age6 or other undetermined factors. In addition, our data suggest a different mechanism for childhood fractures and fractures occurring after age 15 years. A (genetic) predisposition toward smaller bones could increase the risk of childhood fractures during the critical period in which longitudinal growth precedes full bone mineralization. Fractures occurring later in life (after age 15 years) were unrelated to periosteal bone circumference. Indeed, Pye and colleagues40 previously demonstrated that childhood fractures do not predispose to fracture later in life.

Considering the association we have observed between anthropometric proportions and fracture risk, Kindblom and colleagues demonstrated in the Gothenburg Osteoporosis and Obesity Determinants (GOOD) cohort that late pubertal growth is associated with prevalent fractures.41 An increased arm span and longer limb length could be indicators for a late puberty, which is indeed a risk factor for low bone mass and fractures.41, 42 In this regard, sex steroids and especially estradiol again could play a role in pubertal timing, limb growth, and bone expansion. Based on the observations in girls with late puberty, showing that a deficit in aBMD could be observed prior to menarche, Chevalley and colleagues proposed an alternative hypothesis. A shorter estrogen exposure from prepuberty to PBM would not be the main factor for a deficient bone mass acquisition. Underlying shared genes, determining both pubertal timing and bone mass, could be responsible for the deficient acquisition of bone mass in girls with late puberty.43 In our cohort, the altered body proportions at adult age could be a marker of disturbed estrogen action during puberty and persisting bone fragility throughout life. On the other hand, during adult life, a defective estradiol-mediated bone modeling and remodeling could result in unfavorable bone geometry and increased fracture risk in older men. The contrasting data between older men, in whom low estradiol is a risk factor for fracture, and the absence of association in our cohort of young men can be explained by the relatively stable bone mass in our young men and the resulting low incidence of fractures.

Some limitations of our study should be discussed. First, fractures were self-reported in our cohort and were not ascertained by systematic radiographs. Under- and overreporting can occur owing to poor recall.44 However, our fracture incidence corresponds to published data,2 and the accuracy for self-report of fractures was found to be good in postmenopausal women for forearm or hip fractures.45 Moreover, imprecise data would be expected to weaken the observed associations and in general underestimate associations. Second, the determination of estradiol by immunoassay has a lower precision at low estradiol concentrations. In this study, we modified our protocol, using a double amount of serum, to increase precision at low concentrations of estradiol, and all estradiol values were in the normal range for men of this age. Moreover, the associations between bone parameters and estradiol, as assessed by both immunoassay or mass spectrometry, are comparable.46 Finally, our study design is cross-sectional, and we do not have prospective data on fracture incidence and thus cannot comment on the ability of pQCT measurement to predict future fractures.

In conclusion, we have shown a clear association between fracture prevalence and low bone mass on DXA measurements. Using pQCT measurements, fractures are associated with bone geometry, with the strongest associations observed with cortical thickness and cortical bone area and not with cortical vBMD after adjustment for bone size. Childhood fractures are associated mainly with a smaller periosteal bone size, whereas fractures during young adulthood are associated with a wider endosteal circumference, both leading to a smaller cortical thickness.


YT and BL contributed equally to this study. All the authors state that they have no conflicts of interest.


This study was supported by a grant from the Fund for Scientific Research, Flanders (FWO Vlaanderen Grant G.0662.07). GVB and YT are holders of a PhD fellowship and postdoctoral fellowship, respectively, from the Research Foundation, Flanders (FWO).