Significant bone microarchitecture impairment in premenopausal women with active celiac disease

Bone 76 (2015) 149–157

Contents lists available at ScienceDirect

Bone

j ourna l homepage: www.e lsev ie r .com/ locate /bone

Original Full Length Article

Significant bone microarchitecture impairment in premenopausalwomen with active celiac disease

María Belén Zanchetta a,c,⁎, Florencia Costa b, Vanesa Longobardi a, Gabriela Longarini b,Roberto Martín Mazure b, María Laura Moreno b, Horacio Vázquez b, Fernando Silveira a, Sonia Niveloni b,Edgardo Smecuol b, María de la Paz Temprano b, Hui Jer Hwang b, Andrea González b, Eduardo César Mauriño b,Cesar Bogado a,c, Jose R. Zanchetta a,c, Julio César Bai b,d

a IDIM, Instituto de Diagnóstico e Investigaciones Metabólicas, Buenos Aires, Argentinab Sección Intestino Delgado, Departamento de Medicina, Hospital de Gastroenterología “Dr. C. Bonorino Udaondo”, Buenos Aires, Argentinac Cátedra de Osteología y Metabolismo Mineral, Universidad del Salvador, Buenos Aires, Argentinad Cátedra de Gastroenterología Facultad de Medicina, Universidad del Salvador, Buenos Aires, Argentina

⁎ Corresponding author at: Libertad 836, 1012 BuenosE-mail address: [email protected] (M.B. Zanc

http://dx.doi.org/10.1016/j.bone.2015.03.0058756-3282/© 2015 Elsevier Inc. All rights reserved.

a b s t r a c t

Article history:Received 17 October 2014Revised 23 January 2015Accepted 5 March 2015Available online 14 March 2015

Edited by Nuria Guanabens

Keywords:Bone microarchitectureCeliac diseaseHRp-QCTOsteoporosisFractures

Patients with active celiac disease (CD) aremore likely to have osteoporosis and increased risk of fractures. High-resolution peripheral quantitative computed tomography (HR-pQCT) permits three-dimensional exploration ofbonemicroarchitectural characteristicsmeasuring separately cortical and trabecular compartments, and giving amore profound insight into bone disease pathophysiology and fracture. We aimed to determine the volumetricand microarchitectural characteristics of peripheral bones—distal radius and tibia—in an adult premenopausalcohort with active CD assessed at diagnosis. We prospectively enrolled 31 consecutive premenopausal womenwith newly diagnosed CD (median age 29 years, range: 18–49) and 22 healthy women of similar age (medianage 30 years, range 21–41) and body mass index. Compared with controls, peripheral bones of CD patientswere significantly lower in terms of total volumetric density mg/cm3 (mean ± SD: 274.7 ± 51.7 vs. 324.7 ±45.8, p 0.0006 at the radius; 264.4 ± 48.7 vs. 307 ± 40.7, p 0.002 at the tibia), trabecular density mg/cm3

(118.6 ± 31.5 vs. 161.9 ± 33.6, p b 0.0001 at the radius; 127.9 ± 28.7 vs. 157.6 ± 15.6, p b 0.0001 at thetibia); bone volume/trabecular volume ratio % (9.9 ± 2.6 vs. 13.5 ± 2.8, p b 0.0001 at the radius; 10.6 ± 2.4vs. 13.1 ± 1.3, p b 0.0001 at the tibia); number of trabeculae 1/mm (1.69 ± 0.27 vs. 1.89 ± 0.26, p 0.009 at theradius; 1.53± 0.32 vs. 1.80± 0.26, p 0.002 at the tibia); and trabecular thicknessmm(0.058±0.010 vs. 0.071±0.008, p b 0.0001 at the radius with no significant difference at the tibia). Cortical density was significantly lowerin both regions (D compmg/cm3 860±57.2 vs. 893.9±43, p 0.02; 902.7±48.7 vs. 932.6±32.6, p 0.01 in radiusand tibia respectively). Although cortical thickness was lower in CD patients, it failed to show any significantinter-group difference (a—8% decay with p 0.11 in both bones). Patients with symptomatic CD (n = 22) had agreater bonemicroarchitectural deficit than thosewith subclinical CD. HR-pQCTwas used to successfully identifysignificant deterioration in the microarchitecture of trabecular and cortical compartments of peripheral bones.Impairment was characterized by lower trabecular number and thickness—which increased trabecular networkheterogeneity—and lower cortical density and thickness. In theprospective follow-up of this group of patientsweexpect to be able to assess whether bone microarchitecture recovers and to what extend after gluten-free diet.

© 2015 Elsevier Inc. All rights reserved.

1. Introduction

Celiac disease (CD) is a gluten-dependent systemic disorder charac-terized by an autoimmune damage initiated in the small bowel ofpredisposed individuals [1]. Previous studies have shown thatosteopenia and osteoporosis are well-recognized complications in CDpatients [2–9]. Bone disease constitutes a major problem through theassociation with an increased risk of bone fractures, predominantly in

Aires, Argentina.hetta).

the peripheral skeleton and often produced by minimal or moderatetrauma [10]. In such context, former studies have shown that the distalradius is the most common fractured site corresponding to more than50% of events in CD cases [10–14]. Although controversial, some studiessuggest that the increased risk can be reverted by strict adherence to aspecific treatment, the gluten-free diet (GFD) [11,12,15].

Bone health, characterized by its mass, density, and micro-architectural and material properties, is maintained by a balanced sys-tem of remodeling [9,16,17]. The deterioration of any of thoseparameters or an uncoupling of the remodeling process, leads to bonefragility and an increased risk of fractures. Osteoporosis is characterized

150 M.B. Zanchetta et al. / Bone 76 (2015) 149–157

by low bone mass, thin porous cortices and decreased trabecular num-ber and connectivity [17]. It may be diagnosed before fractures occur bymeasuring areal bone mineral density (aBMD) using dual energy X-rayabsorptiometry (DXA) [18]. Thus, DXA has been considered the goldstandard for predicting osteoporotic fractures. Despite the ability ofDXA to assess bone mass (BMC, BMD), the inherent planar nature ofthe measurements makes geometric assessment of a bone impossibleand bone strength estimation very limited. Furthermore, the methodis unable to discriminate between trabecular and cortical bone tissues,which have been shown to be differentially affected by many condi-tions [19,20]. Furthermore, DXA cannot assess the microarchitecturalstructure of bones and, therefore, explore the intimate pathogenesis ofosteoporotic fractures. High resolution peripheral quantitative comput-ed tomography (HR-pQCT), which is a non-invasive method forvolumetric three-dimensional characterization of peripheral skeletalsites, allows quantification of bone microarchitecture parametersand permits the assessment of bone mechanical properties by FEA(Finite element analysis) [21–23]. For this purpose, the resolutionof HR-pQCT (82 μm) permits the direct and reliable assessment ofmicroarchitectural parameters which are relevant to bone strength.Hence, HR-pQCT has shown to provide better prediction of bonestrength when compared with DXA [24].

Some years ago, our group conducted a longitudinal analysis of bonestructure and strength in a series of newly diagnosed CD patients usingperipheral and axial quantitative computerized tomography (QCT) [25].That seminal study allowed us to speculate about the pathogenesis ofboneweakening in CD patients. There, we suggested that secondary hy-perparathyroidism and the inflammatory process of CD would enhancebone remodeling inducing trabecular thinning, cortical–subcorticalbone mass loss, increase of intracortical porosity and impairment ofbone mechanical quality.

To our knowledge, no studies have assessed the microarchitecturalquality of bones in CD patients. Therefore, our aims in this study were:1—to determine the microarchitectural characteristics of peripheralbones in a consecutive series of adult premenopausal women with ac-tive CD assessed at the time of diagnosis using HR-pQCT and 2—to com-pare these results with those of healthy women of similar age and bodymass index (BMI). Secondary aims of the study were to assess the asso-ciation of the clinical phenotype at diagnosis with the severity of boneimpairment.

2. Materials and methods

2.1. Patients and controls

Female patients with newly diagnosed CDwere recruited at a singlecenter (Sección Intestino Delgado, Hospital de Gastroenterología “Dr. C.Bonorino Udaondo”). Between May 2011 and November 2012, 67 con-secutive female patients with recent diagnosis of CD were screened forinclusion and exclusion criteria. We defined premenopausal status on aclinical basis: less than one year since lastmenstrual cycle. Patients witha concomitant known disorder affecting bone metabolism (thyroiddisease, pregnant at the time of diagnosis, breast feeding, etc.) orconsuming medications potentially affecting bones were excludedfrom the study. Patients having former diagnosis of CD and those havingperformed any kind of dietary gluten restriction were also excluded.During the enrolment period, 31 consecutive premenopausal patientsfulfilled these criteria and were finally enrolled in the study. Twentytwo healthy premenopausal women of comparable age and BMIwere used as a control group. Diagnosis of CD was based on well-established criteria as are the concurrent evidence of positive CD serologytests and abnormal duodenal histology (IIIa or greater damage accordingto Marsh's modified criteria) [26]. Exclusion of CD in healthy controlfemaleswas exclusively based on anegative CD-specific serology. The cat-egorization of the clinical picture at the time of diagnosis was establishedaccording to the Oslo's nomenclature [1]. Patients were categorized as

having symptomatic CD (gastrointestinal and/or extra-intestinal symp-toms) or subclinical CD. At the enrolment interview, patients and controlswere requested to report known risk factors for osteoporosis (smoking,alcohol consumption, parent hip fracture history and menstrual cycles).Excessive alcohol consumption was defined according to FRAX: 3 ormore units of alcohol daily. History of fractures, if present and intensityof trauma were also recorded.

2.2. Study design

The present study reports cross-sectional results obtained at thetime of diagnosis of CD compared with a control group. The overallstudy was designed as a prospective assessment of patients before andafter one year of treatment (gluten free diet).

2.3. HR-pQCT: bone microarchitecture assessment

Bonemicroarchitectural parameters for patients and healthy controlswere determined at the distal non-dominant radius and tibia using HR-pQCT (XtremeCT; Scanco Medical AG, Bassersdorf, Switzerland). Scanswere performed by a specialized technologist. The arm and leg of thesubject were positioned in the scanner and were immobilized duringthe examination in an anatomical carbonfiber shell. Scanswere repeatedin the event of significant motion. The region of interest was defined bymanual placement of a reference line at the endplate of the radius ortibia, with the first slice 9.5 and 22.5 mm proximal to the reference lineat the radius and tibia, respectively. A stack of 110 parallel CT sliceswas obtained at the distal end of both sites with a nominal voxel sizeof 82 μm. This provided a three-dimensional image of approximately9 mm in the axial direction. Image processing and calculation of numer-ical values were performed using Scanco software. The analysis methodshad been previously described, validated ex vivo against the gold stan-dard μ-CT at the radius and tibia and applied in various clinical studies[27–30]. Briefly, the volume of interest was automatically separatedinto cortical (Ct.BMD) and trabecular (Tb.BMD) regions. Cortical Thick-ness (Ct.Th) was defined as the mean cortical volume divided by theouter bone surface. The high isotropic resolution of HR-pQCT permitteddirect 3D assessment of trabecular number (Tb.N) but all other trabecu-lar microarchitecture parameters required calculation. Trabecular bonevolume to tissue volume ratio (BV/TV) (%) was derived from trabeculardensity (Tb. BMD), assuming the density of fully mineralized bone tobe 1.2 g HA/cm3 (BV/TV=100× Tb. BMD/1200mgHA/cm3). Trabecularthickness (Tb. Th.), spacing (Tb. Sp.) and the intra-individual distributionof separation (Tb. Sp.SD), a parameter that reflects the heterogeneity ofthe trabecular network, were derived from BV/TV and Tb. N. using for-mulas from traditional quantitative histomorphometry.

Quality control of the scanner was performed on a daily basis beforethe measurement of the first patient and using a phantom provided bythe manufacturer. The coefficient of variation (CV) of the method hadbeen previously published by our group [31], performing two consecu-tive measurements with full repositioning in 56 women. Reproducibili-ty of volumetric BMD measurements ranged from 0.5 to 0.8% inpremenopausal women. Reproducibility of structural parameters wasslightly lower ranging from 0.4% to 3.1%, which was similar to whatothers had published for this method [30].

HR-pQCT scans were performed in the period of time between diag-nosis and up to one month after Gluten Free Diet initiation.

2.4. Areal bone density

Areal BMD at the lumbar spine (L1 to L4), femoral neck and distal ra-dius was measured in all patients and controls by DXA (Lunar ProdigyAdvance-Soft 13.6. GE Healthcare; USA). Bone mineral density was re-ported as g/cm2 and z-score as recommend by the International Societyfor Clinical Densitometry (ISCD) for premenopausal women [32]. Anegative z-score below 2 or lower was defined as ¨ below the expected

Table 1Demography, clinical and biochemical data of patients and healthy control.

Findings Celiac disease patients Healthy controls

Number of subjects 31 22Age. Median (range). yr 29 (18–49) 30 (21–41)Body weight. Mean (SD). kg 57.7 (13.5) 58.7 (7.7)Height. Mean (SD). m 1.59 (0.05) 1.61 (0.07)Body mass index. Mean (SD) kg/m2 22.9 (5.8) 22.6 (2.3)

Clinical categorization at diagnosis Number of cases (%)

Symptomatic celiac disease 22 (71)Subclinical celiac disease 9 (29)

Celiac disease serology Number of positive tests (%)

IgA tissue transglutaminase 29 (93.5)DGP/tTG Screen 31 (100)Biochemical parameters Mean (SD)Serum calciummg/dL (reference values 8.4–10.2) 8.7 (0.8)

Parathormonepg/mL (reference values 10–65) 64.6 (34.7)

25-OH Vitamin D3ng/mL (normal values N30) 20.1 (10.4)

C-telopeptidepg/mL (reference values 75–550) 614 (738)

Table 2Demography, clinical and biochemical data of CD symptomatic and subclinical patients.

Findings Symptomatic CD Subclinical CD

Number of subjects 22 9Age. Median (range). years. 27 (19–49) 34 (18–43)Body weight. Mean (SD). kg 54.8 (11.8) 62.5 (11.1)Height. Mean (SD). m. 1.59 (0.05) 1.59 (0.06)Body mass index. Mean (SD) kg/m2 21.7 (5.7) 24.8 (4.6)

Celiac disease serology Number of positive tests (%)

IgA tissue transglutaminase 20 (91) 9/9 (100)DGP/tTG Screen 22 (100) 9/9 (100)

Biochemical parameters Mean (SD)

Serum calciummg/dL (reference values 8.4–10.2) 8.7 (0.9) 8.9 (0.4)

Parathormone pg/mL(reference values 10–65) 66.6 (40.1) 61.9 (22.5)

25-OH Vitamin D3ng/mL (reference values N30) 19.4 (11.6) 21.7 (8.2)

C-telopeptideng/mL (reference values 75–550) 751 (863) 327 (167)

151M.B. Zanchetta et al. / Bone 76 (2015) 149–157

range for age¨. DXA scanswere performed in the period of timebetweendiagnosis and up to one month after GFD initiation.

2.5. Biochemical measurements

Blood samples after overnight fast were obtained from all patients atthe enrolment visit. All patients were tested for serum calcium (normalrange: 8.8–10.5 mg%), 25 OH Vitamin D (Method: Chemiluminescencenormal values N30 ng/mL), parathormone (PTH) (immunoradiometricassay, normal range 10–65 pg/mL) and C-Telopeptide (enzyme immu-noassay, normal range: 75–550 pg/ml). Serological tests used for diag-nosis of CD were: anti-tissue transglutaminase IgA and IgA and IgGantibodies against synthetic deamidated gliadin-derived peptides,both detected by ELISA assays (Inova Diagnostic Inc., San Diego, CA,USA) as reported in previous studies [1].

2.6. Ethical issues

The protocol was approved by the Research and Ethical Committeesof the Hospital de Gastroenterología “Dr. C. Bonorino Udaondo”. Awritten consent was obtained after diagnosis and previous to enroll-ment. Patients who rejected participation in the study receivedstandard care.

2.7. Statistics

Based on distribution, descriptive data are reported either as meanand standard deviation (SD) or median and range. Comparisons be-tween groups were performed using unpaired T test or Mann–Whitneyaccording to distribution of data. In order to assess associations, thePearson product–moment correlation coefficientwas used for paramet-ric distributions and the Spearman's rank correlation coefficient wasused for non-parametric variable measurements. The statistical levelof significance was 0.05. Data were analyzed using MedCalc® version11.2.1.0. (MedCalc Software bvba; Mariakerke, Belgium).

We were not aware of any previous study describing bone microarchitecture status in CD patients. Therefore, for power calculation weassumed that CD patients would have, at least, microarchitecture dete-rioration similar to subjects with severe osteopenia. BV/TV values of agroup of patients with severe osteopenia [27] and a cohort of premeno-pausal, healthywomenwith normal BMD (n= 24) were used to deter-mine the effect size. Based on these figures we needed a sample size of13 patients per group to achieve a 95% power to detect differences inBV/TVwith a level of significance of 0.05 (two tails). Power calculationswere performed using G*Power (version 3.1.7, Kiel University,Germany).

3. Results

3.1. Clinical features

Untreated CD patients and healthy controls were comparable interms of height, body weight and BMI (p NS) (Table 1). Only Six CDpatients had low weight, with a BMI less than 18, 5. While 22 patientshad a symptomatic clinical presentation, nine cases were consideredas having subclinical CD (two diagnosed by case finding among 1st de-gree relatives, three in the investigation of microcytic anemia and fourin the endoscopic investigation because of gastroesophageal refluxsymptoms). All of these subclinical patients were considered as clinical-ly silent at diagnosis. Most patients categorized as having symptomaticCD had gastrointestinal symptoms (diarrhea was present in 63% ofthem). All patients enrolled had abnormally increased concentrationsfor at least one of the CD specific serologic tests. Twenty-nine CD cases(93.5%) had a positive IgA tTG test and all had positive tTG/DGP screentest. Twenty-five, four and two patients had IIIc, IIIb and IIIa histologicstages as the most severe mucosal damage in the diagnostic duodenal

biopsy, respectively. No patient enrolled in the study reported a fractureevent prior to diagnosis. One CD patient was a current smoker and an-other reported excessive alcohol consumption, six had menstrual cycledisorders (oligomenorrhea and/or amenorrhea) and five were takingoral contraceptives. In the control group, one patient reported excessivealcohol consumption and two patients were taking oral contraceptives.No other risk factors were reported.

Compared with patients reporting a symptomatic CD at diagnosis,thosewith a subclinical course (n= 9) had similar demography and se-rological and biochemical parameters (Table 2). Despite the lack of sta-tistical significance, mean values for body weight and BMI were higherin the subclinical population.

3.2. HR-pQCT: bone microarchitecture assessment

Compared with healthy controls, all parameters of the trabecularand cortical compartments of the distal radius were statistically signifi-cant lower with the exception of cortical thickness (Table 3 and Fig. 2).Most significant changes detected in the distal radiuswere found in tra-becular parameters. Compared to mean values of controls, BV/TV andtrabecular density were 26.4% lower, trabecular thickness 18% lower,

Table 3HR-pQCT measurements of distal radius and distal tibia in untreated celiac disease patients and healthy controls.

Distal radius Distal tibia

HR-pQCTMean (SD)

Celiac patients Healthy controls p value Celiac patients Healthy controls p value

No. of patients 31 22 31 22D 100 (mg/cm3) 274.7 (51.7) 324.7 (45.8) 0.0006 264.4 (48.7) 305.0 (40.7) 0.002D comp (mg/cm3) 860.2 (57.2) 893.9 (43.0) 0.02 * 902.7 (48.7) 932.6 (32.6) 0.01 *Ct.Th (mm) 0.69 (0.15) 0.75 (0.13) 0.11 1.07 (0.22) 1.16 (0.16) 0.11 *D Trab (mg/cm3) 118.6 (31.5) 161.9 (33.6) b0.0001 * 127.9 (28.7) 157.6 (15.6) b0.0001 *BV/TV (%) 9.9 (2.6) 13.5 (2.8) b0.0001 * 10.6 (2.4) 13.1 (1.3) b0.0001 *Tb.N (1/mm) 1.69 (0.27) 1.89 (0.26) 0.009 1.53 (0.32) 1.80 (0.26) 0.002Tb.Th (mm) 0.058 (0.010) 0.071 (0.008) b0.0001 0.069 (0.011) 0.073 (0.010) 0.16Tb.Sp (mm) 0.539 (0.097) 0.468 (0.075) 0.004 0.581 (0.110) 0.482 (0.065) 0.0002 *Tb.1/N.SD (1) 0.228 (0.047) 0.186 (0.032) 0.0005 * 0.289 (0.072) 0.216 (0.032) 0.0001 *

Parameters reported in the table are: Total volumetric density (D100), cortical density (D comp), cortical thickness (Ct Th), trabecular bone density (D Trab), trabecular bone volume frac-tion (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th), trabecular spacing (Tb.Sp), trabecular heterogeneity (Tb.1/N.SD). Data are reported as mean and standard deviation(SD) irrespective of data distribution. According to data distribution, statistical comparisons were performed with t-Test or Mann–Whitney test (for independent samples) (*).

152 M.B. Zanchetta et al. / Bone 76 (2015) 149–157

trabecular number 10.5% lower, while trabecular spacing was 15%higher and heterogeneity of trabecular network was 23% higher. Asregards cortical parameters, cortical densitywas 3.6% lower and corticalthickness 8% lower; this last variable did not reach statisticalsignificance.

Similarly, at the distal tibia most parameters were significantlylower except for cortical thickness and trabecular thickness (Table 3).Compared to mean values of controls, BV/TV and trabecular densitywere 19.2% lower, trabecular thickness 5.5% lower, trabecular number15% lower, while trabecular spacing was 20.5% higher and heterogene-ity of trabecular network was 34% higher. As regards cortical parame-ters, cortical density was 3.2% lower and cortical thickness 8% lower;again this last variable did not reach statistical significance. Fig. 1shows aHRp-QCT three dimensional images clearly illustrating thefind-ings described.

Patients categorized as presenting symptomatic CD had greatermean microarchitectural deficit compared with those categorized assubclinical CD (Table 4). However, probably due to the small numberof patients, these values were only statistically significant at BV/TVand D. trab (−16.7% and −17,6% respectively in symptomatic CD pa-tients) in distal radius. Deterioration of cortical densitywasmuch small-er for the subclinical than for the symptomatic group of patients(subclinical: −1.4%; symptomatic: −4.7% for the radius; and −1.1%and −4.1% for the tibia, respectively). Importantly, the differences ob-served between symptomatic and healthy individuals were largerthan those observed between the whole group of patients and the con-trol group (3.8% and 3.2%, respectively).

3.3. Areal bone density

Table 5 shows comparative results of DXA parameters in CD patientsand healthy controls. Compared with healthy controls, BMD z-scoreswere significantly lower at lumbar spine, femoral neck and distal radius(p b 0.009; p b 0.02 and p b 0.002, respectively). Although significantlylower compared with controls, mean aBMD z-scores of patients at lum-bar spine and femoral neck were within the range of normality. In con-trast, their mean aBMD zscore at the distal radius was in the borderlinelower level of normality (−1.9). Densitometry impairmentwas notablyhigher in the distal radius compared with other sites of interest. Thus,while only two patients had z-score at the lumbar spine or femoralneck below−2, at distal radius 14 had z-scores below−2 (below ade-quate for age).

Compared with DXA measurements in symptomatic patients,those with subclinical CD evidenced a significant lower aBMD com-promise at the radius (−2.3 vs. −1.0, respectively; p b 0.02).No significant differences were detected at the lumbar spine andfemoral neck.

3.4. Biochemical measurements

Mean serum calcium concentration was in the normal range but 7out of 31 patients (22.6%) had concentrations below the lower end ofthe reference range (Table 1). Mean serum parathormone was in theupper end of normal range and 11/31 patients (35.5%) had abnormallyincreased values. Serum Vitamin D levels were below 10 ng/mL in fourpatients (13%), between 10 and 20 ng/mL in 14 patients (45%), between20 and 30 ng/ml in 10 patients (32%) and only 3 patients (10%)had concentrations above 30 ng/mL. Nine out of 28 patients (29%) hadC-telopeptide serum concentrations above the normal range forpremenopausal women and therefore, the median value for theoverall study populationwas high. Although not reaching statistical sig-nificance, C-telopeptide values shifted from normal (327 ng/ml) in sub-clinical patients to pathologically enhanced levels (751 ng/ml) insymptomatic cases. Compared with symptomatic cases, the subclinicalcohort did not show significant differences in other bone metabolismbiochemical parameters like calcium, PTH or 25-OH-D3 (Table 2).

3.5. Correlations

We observed significant correlations between HR-pQCT parametersand age, BMI, Ctelopeptide and 25 OH Vitamin D. An inverse correlationwas found between C-telopeptide and distal radius cortical density(r = −0.58, p = 0.002). In contrast, there was a direct correlation be-tween vitamin D and trabecular density, BV/TV and trabecular numberat the radius (r = 0.50, p = 0.006; r = 0.50, p = 0.006 and r = 0.49,p=0.007 respectively). At the distal tibia, therewas a direct correlationbetween number of trabeculae and BMI (r = 0.62, p = 0.0003) butan inverse correlation between trabecular parameters and BMI(r = −0.39, p = 0.03 for trabecular thickness; r = −0.67, p =0.0001 for trabecular spicing; r = −0.53, p = 0.003 for intra-individual distribution of separation—trabecular heterogeneity. Therewere also inverse correlations comparing trabecular thickness withage (r =−0.39, p = 0.03). Patients with concentrations of IgA tTG an-tibodies above 200 AU/mL (n = 21) had borderline significant higherPTH concentrations than those having serologic values below suchcut-off (n = 10) (PTH mean ± SD: 72.7 ± 38.8 vs. 47.5 ± 13.2, respec-tively; p= 0.0567). Other biochemical, densitometric and HR-pQCT pa-rameters were not significantly different according to the concentrationlevel of tTG antibodies.

4. Discussion

A series of recent studies and a meta-analysis have shown that CDpatients have an increased incidence of bone fractures,mostly in the pe-ripheral skeleton [10–16]. Such risk is associated with the symptomatic

Fig. 1.HRp-QCT threedimensional images of the radius and tibia of one of the celiac diseasewomen at the time of diagnosis (left) and a control subject (right). Note the clear differences inthe trabecular network.

153M.B. Zanchetta et al. / Bone 76 (2015) 149–157

clinical phenotype at the time of diagnosis, is related to low trauma(osteoporotic fractures), is mainly produced before diagnosis and,finally, may be recurrent in a subset of patients [9–15]. In this context,studies have shown that the most common site for fractures (N50% ofevents) is the distal radius [9].

Trabecular bone is the area closest to the bonemarrowand intimate-ly related to blood vessels, therefore resulting in themost metabolicallyactive bone compartment [17]. Interestingly, trabecular bone remodelsup to an order of magnitude faster than cortical bone. Considering thatone of themain functions of our skeleton is to act as a calcium reservoir,this compartment is specially designed to quickly liberate this ion help-ing to maintain serum calciumwithin normal levels, an essential condi-tion for vital intracellular functions [33]. Consequently, any disease,disorder or even physiological menopause may affect trabecular bonefirstly. In CD, where there is a chronic negative balance of calcium, weexpected to find, predominantly, trabecular impairmentwith decreasedtrabecular thickness and number. However, as themost common avail-able technique—DXA—does not permit the evaluation of bone compart-ments separately, it has not been possible to confirm this hypothesis sofar [17]. The new imagingmethod employed in the present study allowsus to more accurately appreciate main features of bone loss. The in-creased bone resorption in these patients produces thinned trabeculaewith a subsequent decrease in its number, both phenomena leading tomore distant trabeculae of different thicknesses which finally result ina more heterogeneous network.

On the other hand, cortical bone is more compact and hard; it con-forms the bone shell with less proximity to the vessels and, therefore,lower metabolic activity than trabecular bone. When bone turnover in-creases, the cortical compartment participates increasing its porosity as

well [32]. The cortical compartment, generally, is not as easily andquickly affected as the trabecular bone, and therefore is more preservedin case of disease. This natural preservation of cortical bone tissue isprobably due to its crucial role in the mechanical prevention of frac-tures. As amatter of comparison, cortical bone tissue has a natural stressranking no less than one degree of magnitude higher than that of tra-becular structures. Pistoia et al. evaluated which structural parameter(cortical thickness, trabecular thickness, number of trabeculae or overallthinning of structures)was the best predictor for mechanical propertiesof the human radius using microcomputer tomography scanner andmirofinite element simulations [34]. This analysis revealed that mostof the loadwas carried by the cortical bone and that little changes in cor-tical thickness could cause dramatic bone strength reductions.

HR-pQCT permits three-dimensional exploration of bone micro-architectural characteristics measuring separately cortical and trabecu-lar compartments, and giving amore profound insight into bonediseasepathophysiology and fracture risk. The present study employed, for thefirst time, HR-pQCT to assess bonemicroarchitecture in premenopausaluntreated CD patients. We compared microarchitectural parametersassessed in our cohort of patients at the time of diagnosis with thoseof a cohort of healthy women of similar age and body mass index.Thus, we detected a significant deterioration in total, cortical and tra-becular volumetric density.

Trabecular density was the most impaired, being 26.4% and 19.2%(both p b 0.0001) lower in the radius and the tibia respectively. Thisfinding was in total agreement with the disminished aBMD Z-scoresfound in the distal radius. At the distal radius, both BV/TV and trabecularnumber and thickness were significantly impaired leading to anincrease in the distance among trabeculae and showing a more

264,4(-18,6)

299,8(-7,7)

324.7

261,3(-14,3)

271,9(-10,9)

305.0

0

50

100

150

200

250

300

350

Symptomatic CD

Subclinical CD Control Symptomatic CD

Subclinical CD Control

Distal radius Distal tibia

D10

0 (m

g/cm

3 )

D100

111,6(-31,1)

135,8(-16,1)*

161.9

125,9(-20,1)

132,7(-15,8)

157.6

0

20

40

60

80

100

120

140

160

180

Symptomatic CD

Subclinical CD Control Symptomatic CD

Subclinical CD Control

Distal radius Distal tibia

Dtr

ab (

mg/

cm3 )

Dtrab *p<0,05 (Symptomatic CD

9,3(-31,1)

11,3(-16,3)*

13.5

10,5(-19,8)

11,0(-15,7)

13.1

0

2

4

6

8

10

12

14

16

Symptomatic CD

Subclinical CD Control Symptomatic CD

Subclinical CD Control

Distal radius Distal tibia

BV

/TV

(%

)

BV/TV *p<0,05 (Symptomatic CD

851,5(-4,7)

881,5(-1,4) 893.9

894,6(-4,1)

922,4(-1,1) 932.6

0

100

200

300

400

500

600

700

800

900

1000

Symptomatic CD

Subclinical CD Control Symptomatic CD

Subclinical CD Control

Distal radius Distal tibia

Dco

mp

(mg/

cm3 )

Dcomp

0,67(-10,7)

0,73(-2,7) 0.75

1,10(-5,2)

1,11(-4,3)

1.16

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Symptomatic CD

Subclinical CD Control Symptomatic CD

Subclinical CD Control

Distal radius Distal tibia

Ct.T

h (m

m)

Ct.Th

1,68(-11,1)

1,72(-9,0)

1.89

1,54(-14,4)

1,51(-16,1)

1.80

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

Symptomatic CD

Subclinical CD Control Symptomatic CD

Subclinical CD Control

Distal radius Distal tibia

Tb.

N (

1/m

m)

Tb.N

0,05(-28,6)

0,06(-14,3)*

0.070,07(0,0)

0,07(0,0) 0.07

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

Symptomatic CD

Subclinical CD Control Symptomatic CD

Subclinical CD Control

Distal radius Distal tibia

Tb.

Th

(mm

)

Tb.Th *p<0,05 (Symptomatic CDvs Subclinical CD)

Fig. 2. Figure shows HR-pQCT parameters's absolute values and percentages differences between controls and clinical and subclinical CD patients.

154 M.B. Zanchetta et al. / Bone 76 (2015) 149–157

0,54(14,9)

0,53(12,8)

0.47

0,61(27,1)

0,62(29,2)

0.48

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Symptomatic CD

Subclinical CD Control Symptomatic CD

Subclinical CD Control

Distal radius Distal tibia

Tb.

Sp

(mm

)Tb.Sp

0,26(36,8) 0,23

(21,1)

0.19

0,33(50,0)

0,36(63,6)

0.22

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Symptomatic CD

Subclinical CD Control Symptomatic CD

Subclinical CD Control

Distal radius Distal tibia

Tb.

1/N

.Sd

(1)

Tb.1/N.SD

Fig. 2 (continued).

Table 5Bonemineral density (aBMD)determined byDEXA at lumbar spine, femoral neck and dis-tal radius sites in celiac disease patients and healthy control females expressed in terms ofz-score. Comparisons were performed using the unpaired t-test.

Bone mineral density z-scoreMean value (SD)

Celiacpatients

Healthy controls pvalue

Lumbar spine (L1–L4). −0.50 (0.9) 0.20 (0.7) 0.003Femoral neck. −0.20 (0.8) 0.30 (0.8) 0.03Distal radius. −1.9 (1.3) −0.50 (1.1) 0.001

155M.B. Zanchetta et al. / Bone 76 (2015) 149–157

heterogeneous trabecular network. At the tibia, most parameters weresignificantly lower compared with those of healthy controls. However,the relative % deficit in the distal tibia was lower than that detected inthe distal radius (e.g. D.trab: −26.4% vs. −19.2%, in radius and tibiarespectively).

Cortical density was also lower than in the control group (3.6% in ra-dius and 3.2% in tibia), probably due to increased cortical porosity andintra-cortical remodeling. Despite a slight trend to become also lowerin CD patients than in controls, cortical thickness failed to show any sig-nificant inter-group difference (a—8% decay with p 0.11 in both bones).These findings in cortical bone, despite seeming small, can be very im-portant biologically, due to their mechanical correlate, in terms of

Table 4HR-pQCT measurements of distal radius and distal tibia in celiac disease patients categorized a

Distal radius

HR-pQCT Mean (SD) Symptomatic CD Subclinical CD

No. of patients 22 9D 100 (mg/cm3) 264.4 (41.5) 299.8 (67.0)D.comp (mg/cm3) 851.5 (60.4) 881.5 (44.5)Ct.Th (mm) 0.67 (0.15) 0.73 (0.11)D. Trab (mg/cm3) 111.6 (22.7) 135.8 (43.4)BV/TV (%) 9.3 (1.9) 11.3 (3.6)Tb.N (1/mm) 1.68 (0.27) 1.72 (0.3)Tb.Th (mm) 0.05 (0.01) 0.06 (0.01)Tb. Sp (mm) 0.54 (0.10) 0.53 (0.10)Tb.Sp.SD (1) 0.26 (0.11) 0.23 (0.06)

Parameters reported in the table are: Total volumetric density (D100), cortical density (D.comp)tion (BV/TV), trabecular number (Tb.N), trabecular thickness (Tb.Th), trabecular spacing (Tb.Sp(SD) irrespective of data distribution. According to data distribution, statistical comparisons w

bone strength. Previously, we had studied other parameters in CD pa-tients employing standard peripheral quantitative computerizedtomography—pQCT [25]. This technique can assess mechanical vari-ables as the moment of inertia of the cortical bone cross sectional area(CSMI)—an indicator of the architectural efficiency of the bone design-and bone strength index (BSI)—an estimation of themechanical quality(stiffness and strength) of the whole bone-, calculated as the product ofthe vBMD of the cortical bone and the CSMI. We found a significant im-pairment of the CSMI and BSI in the radius, suggesting that bone weak-ening in CD is related to both themass and the spatial distribution of thebone tissue.

The analysis of bone microarchitecture in CD patients according tothe clinical phenotype also provided us with relevant information.Interestingly, patients having symptomatic CD (mostly gastrointestinalsymptoms) had a more profound bone compromise compared withthose considered subclinical. This last group was also affected in theirbone microarchitecture, although to a lesser degree. Trabecular densityshowed some homogeneous trend to progress with the course of thedisease, as judged by the successive differences shownby the subclinicaland symptomatic patients with respect to their controls (−16.1% and−31.1% for the radius; −15.8% and −20.1% for the tibia). It seemsthat themore severe the disease, the greater the compromise in trabec-ular density. However, in general terms, the evolution of the disease(as judged by the symptomatic vs. subclinical comparisons) seemednot to have affected Tb.N, Tb.1/N.SD (the heterogeneity of the trabecularnetwork) and Tb.Sp in a progressive fashion.

In contrast with that observed for D Trab, the deterioration of DComp during the course of the disease looked far delayed, as judgedby the quite smaller differences observed between subclinical patientsand controls vs symptomatic patients and controls (subclinical,−1.4%; symptomatic, −4.7% for the radius; and −1.1% and −4.1% forthe tibia, respectively). The analysis of differences between the twosymptomatic phenotypes described suggests a different evolution ofCD effects on trabecular (smooth, progressive impairment) and corticalbone (impairment somewhat delayed with respect to the trabecular af-fectation). This may be explained, at least in part, by the differences ob-served in bone turnover, whichwasmuch higher in clinical CD patients.

Women enrolled in this study were in their thirties and only veryfew of them reported additional osteoporotic risk factors: one reported

ccording to the clinical presentation: symptomatic vs. subclinical celiac disease.

Distal tibia

p value Symptomatic CD Subclinical CD p value

22 90.0810 261.3 (51.7) 271.9 (42.2) 0.59140.1905 894.6 (53.5) 922.4 (27.7) 0.15300.2578 1.1 (0.26) 1.11 (0.1) 0.57550.0497 125.9 (23.3) 132.7 (40.3) 0.55680.0489 10.5 (1.9) 11.04 (3.3) 0.56930.7186 1.54 (0.3) 1.51 (0.3) 0.81420.0072 0.07 (0.02) 0.07 (0.01) 0.68100.7241 0.61 (0.2) 0.62 (0.2) 0.92250.4710 0.33 (0.21) 0.36 (0.2) 0.7263

, cortical thickness (Ct Th), trabecular bone density (D. Trab), trabecular bone volume frac-), trabecular heterogeneity (Tb.Sp.SD). Data are reported as mean and standard deviationere performed with t-Test or Mann–Whitney test (for independent samples) (*).

156 M.B. Zanchetta et al. / Bone 76 (2015) 149–157

current smoking, one excessive alcohol consumption and six patientshad menstrual cycle disorders (oligomenorrhea and/or amenorrhea).No patient in this cohort had experienced bone fractures.

In the present study, DXAvalueswere significantly lower in all regionsassessed compared with healthy controls. Although lumbar spine andfemoral neck z-scores were within the normal range, distal radius z-scores were borderline what is considered normal for age. This observa-tion coincides with microarchitectural findings and it is likely associatedwith the fact that the distal radius is the site where most fractures occurin the young CD population. Our group and several others had previouslyshown a lower aBMD in all areas of the skeleton assessed by DXA in CDpatients [2–8]. Global bone mineral density assessment, including arealBMDDXAmeasurements in the radius and total vBMD (D100)measuredby HR-pQCT of both radius and distal tibia, showed a comparable deteri-oration with a similar statistical significance.

As regards biochemical parameters of bone metabolism, the presentstudy confirms former observations about the presence of secondary hy-perparathyroidism, considered one of the pathophysiological mecha-nisms involved in bone damage [5,7,9,23]. High serum C-telopeptidevalues suggest that a higher than normal bone turnover (presumablywith a negativemass balance) should have affected both trabecular (thin-ning, elimination) and cortical (intra-cortical porosity—not measured)behavior in mechanical terms, perhaps keeping a progressive pattern ofaffectation as shown by the higher c-telopeptide levels in symptomaticthan in subclinical patients. On the other hand, local and distant immuno-logical factors, including autoimmunity, should also be considered for thepathogenesis of bone impairment [35,38]. The association betweenproin-flammatory cytokines and bone mineral deterioration present in CD, andthe relation between tTG antibody concentrations and bone impairmentreportedby former studies have opened thediscussionof the potential in-volvement of CD immunological disturbances and autoimmunity in themultifactorial pathogenesis of bone loss [9,36–38].

Limitations of our study were the small sample of patients and thatHR-pQCT can only assess microarchitecture of the peripheral skeleton.The first risk of bias seems to be ruled out by the statistical significanceof the present results. The second suggested limitation has beenrecently addressed by a study performed in the general osteoporoticpopulation where authors suggested that HR-pQCT determinations inperipheral skeleton may also reflect the mechanical competence andrisk of fractures in the central skeletonwheremany of themost clinical-ly serious fractures can occur [23]. Finally, in this studywe could not as-sess cortical porosity, Finite Element Analysis (FEA), inter trabecularconnectivity or total body composition. The mechanical correlate ofour results is surely important, but due to the former limitations isquite difficult to assess and define. The lack of determinations like theinter-trabecular connectivity and FEA-derived or other suitable stressindicators precludes extrapolating these data to predict bone behaviorwhen different kinds of stress are induced on the trabecular structureduring the mechanism of production of peripheral fractures. Those canbe areas for future studies.

5. Conclusion

In conclusion, the present study provides insight into the pathogen-esis of bonemass loss in CDpatients. In this group of youngwomenwithnewly diagnosed celiac disease, HRpQCT allowed us to detect a severecompromise of both, volumetric density and bone microarchitecture.To our knowledge, this is the first time that HR-pQCT has describedcharacteristics of bone microarchitecture deterioration in CD patients.In the prospective follow-up of this group of patients we expect to beable to assess whether bone microarchitecture recovers and to whatextend after gluten-free diet.

Disclosures

All authors state that they have no conflicts of interest.

Acknowledgments

We thank Susana Carballo for editing the English language of themanuscript.

The study was partially funded by: 1—Consejo de Investigación enSalud; Ministerio de Salud; Gobierno de la Ciudad de Buenos Aires.

2—FIM, Fundación de Investigaciones Metabólicas.

References

[1] Ludvigsson J, Leffler D, Bai J, Biagi F, Fasano A, Green P, et al. The Oslo definitions forcoeliac disease and related terms. Gut 2013;62:43–52.

[2] Caraceni M, Molteni N, Bardella M, Ortolani S, Nogara A, Bianchi P. Bone and mineralmetabolism in adult celiac disease. Am J Gastroenterol 1988;83:274–7.

[3] Bodé S, Hassager C, Gudmand-Høyer E, Christiansen C. Body composition and calci-um metabolism in adult treated coeliac disease. Gut 1991;32:1342–5.

[4] Mora S, Weber G, Barera G, Bellini A, Pasolini D, Prinster C, et al. Effect of gluten-freediet on bone mineral content in growing patients with celiac disease. Am J Clin Nutr1993;57:224–8.

[5] González D, Mazure R, Mautalen C, Vazquez H, Bai J. Body composition and bonemineral density in untreated and treated patients with celiac disease. Bone 1995;16:231–4.

[6] Mazure R, Vazquez H, Gonzalez D, Mautalen C, Pedreira S, Boerr L, et al. Bone min-eral affection in asymptomatic adult patients with celiac disease. Am J Gastroenterol1994;89:2130–4.

[7] Corazza G, Di Sario A, Cecchetti L, Tarozzi C, Corrao G, Bernardi M, et al. Bone massand metabolism in patients with celiac disease. Gastroenterology 1995;109:122–8.

[8] McFarlane X, Bhalla A, Reeves D, Morgan L, Robertson D. Osteoporosis in treatedadult coeliac disease. Gut 1995;36:710–4.

[9] Corazza G, Di Stefano M, Mauriño E, Bai J. Bones in coeliac disease: diagnosis andtreatment. Best Pract Res Clin Gastroenterol 2005;19:453–65.

[10] Vázquez H, Mazure R, Gonzalez D, Flores D, Pedreira S, Niveloni S, et al. Risk offractures in celiac disease patients: a cross-sectional, case-control study. Am JGastroenterol 2000;95:183–9.

[11] Moreno M, Vazquez H, Mazure R, Smecuol E, Niveloni S, Pedreira S. et. al, Stratifica-tion of bone fracture risk in patients with celiac disease. Clin Gastroenterol Hepatol2004;2:127–34.

[12] Ludvigsson J, Michaelsson K, EkbomA,Montgomery S. Coeliac disease and the risk offractures - a general population-based cohort study. Aliment Pharmacol Ther 2007;25:273–85.

[13] West J, Logan R, Card T, Smith C, Hubbard R. Fracture risk in people with celiacdisease: a populationbased cohort study. Gastroenterology 2003;125:429–36.

[14] Olmos M, Antelo M, Vazquez H, Smecuol E, Mauriño E, Bai J. Systematic review andmeta-analysis of observational studies on the prevalence of fractures in coeliacdisease. Dig Liver Dis 2008;40:46–53.

[15] Pinto Sánchez M, Mohaidle A, Baistrocchi A, Matoso D, Vazquez H, Gonzalez A. Riskof fractures in celiac disease: gender, dietary compliance or both? World JGastroenterol 2011;17:3035–42.

[16] Fickling W, McFarlane X, Bhalla A, Robertson D. The clinical impact of metabolicbone disease in coeliac disease. Postgrad Med J 2001;77:33–6.

[17] Ferretti J. Biomechanical properties of bone. In: Genant HK, Guglielmi G, JergasM, editors. Bone densitometry and osteoporosis. Berlin: Springer; 1998.p. 143–61.

[18] Mazess R, Barden H, Bisek J, Hanson J. Dual-energy X-ray absorptiometry for totalbody fat and regional bone-mineral and soft-tissue composition. Am J Clin Nutr1990;51:1006–12.

[19] Osteoporosis prevention, diagnosis, and therapy NIH consensus Statement. 2000;17:1–45.

[20] Marshall D, Johnell O,Wedel H.Meta-analysis of howwell measures of bonemineraldensity predict occurrence of osteoporotic fractures. BMJ 1996;312:1254–9.

[21] Laib A, Hauselmann H, Ruegsegger P. In vivo high resolution 3DQCT of the humanforearm. Technol Health Care 1998;6:329–37.

[22] Boutroy S, Bouxsein M, Munoz F, Delmas P. In vivo assessment of trabecular bonemicroarchitecture by high-resolution peripheral quantitative computed tomogra-phy. J Clin Endocrinol Metab 2005;90:6508–15.

[23] Liu X, Cohen A, Shane E, Yin P, Stein E, Rogers H, et al. Bone density, geometry, mi-crostructure and stiffness: relationships between peripheral and central skeletalsites assessed by DXA HR-pQCT, and cQCT in premenopausal women. J BoneMiner Res 2010;25:2229–38.

[24] Dall´Ara E, Pahr D, Varga P, Kainberger F, Zysset P. QCT-based finite element modelspredict human vertebral strength in vitro significantly better than simulated DEXA.Osteoporos Int 2012;23:563–72.

[25] Ferretti J, Mazure R, Tanouem P, Marino A, Cointry G, Vazquez H. Analysis of thestructure and strength of bones in celiac disease patients. Am J Gastroenterol2003;98:382–90.

[26] Bai J, Fried M, Corazza G, Schuppan D, FarthingM, Catassi C. World gastroenterologyorganization Global Guidelines on Celiac Disease. J Clin Gastroenterol 2013;47(2):121–6.

[27] ZanchettaMB, Silveira F, SestaM,Massari F, Zanchetta JR, Bogado C. Volumetric BMDand bone microarchitecture are no different between postmenopausal women withosteoporosis or osteopenia with lumbar spine T-score values below – 2. J BoneMiner Res 2009;24(Suppl. 1):s448.

157M.B. Zanchetta et al. / Bone 76 (2015) 149–157

[28] Liu X, Zhang X, Sekhon K, Adams M, McMahon D, Bilezikian J, et al. High-resolutionperipheral quantitative computed tomography can assess microstructural and me-chanical properties of human distal tibial bone. J Bone Miner Res 2010;25:746–56.

[29] Melton L, Riggs B. G.van Lenthe, S. Achenbach, R. Müller, M. Bouxsein, et al. Contri-bution of in vivo structural measurements and load/strength ratios to the determi-nation of forearm fracture risk in postmenopausal women. J Bone Miner Res 2007;22:1442–8.

[30] Walker M, McMahon D, Udesky J, Liu G, Bilezikian J. Application of high-resolutionskeletal imaging to measurements of volumetric BMD and skeletalmicroarchitecture in ChineseAmerican and white women: explanation of a paradox.J Bone Miner Res 2009;24:1953–9.

[31] Silveira F, Monteverde C, Ripero V. MB. Zanchetta, F. Massari F, JR. Zanchetta, et al.Reproducibility of HRp-QCT measurements in women with normal BMD and post-menopausal women with osteoporosis. Osteoporos Int 2011;22(Suppl.1):S43–4.

[32] Schousboe J, Shepherd J, Bilezikian J, Baim S, et al. 5th ISCD Position DevelopmentConference. J Clin Densitom; 2013 [http://www.iscd.org/official-positions/5th-iscd-position-development-conference-adult,].

[33] Frost H. Defining osteopenias and osteoporosis. Another view (with insights from anew paradigm). Bone 1997;20(5):385–91.

[34] Pistoia W, van Rietbergen B, Rüegsegger P. Mechanical consequences of differentscenarios for simulated bone atrophy and recovery in the distal radius. Bone Dec2003;33(6):937–45.

[35] Fornari M, Pedreira S, Niveloni S, Gonzalez D, Diez R, Vazquez H, et al. Pre- and post-treatment serum levels of cytokines IL-1b, IL-6 and IL-1 receptor antagonist in coe-liac disease. Are they related to the associated osteopenia? Am J Gastroenterol 1998;93:413–8.

[36] Malyankar U, Scatena M, Suchland K, Yun T, Clark E, Giachelli C. Osteoprotegerin isan v b3-induced NFkB-dependent survival factor for endothelial cells. J Biol Chem2000;275:2959–62.

[37] Sugai E, Chernavsky A, Pedreira S, Smecuol E, Vazquez H, Niveloni S, et al. Bone-spe-cific antibodies in sera from patients with coeliac disease: characterization and im-plications in osteoporosis. J Clin Immunol 2002;22:353–62.

[38] Duerksen D, Leslie W. Positive celiac disease serology and reduced bone mineraldensity in adult women. Can J Gastroenterol 2010;24:103–17.