doi:10.1111/j.1440-1746.2006.04679.

GASTROENTEROLOGY

Tumor necrosis factor-a haplotype is strongly associated with bone mineral density in patients with Crohns disease
Naomi Lee,* Elizabeth Fowler, Susan Mason, Douglas Lincoln, Dennis R Taaffe* and Graham Radford-Smith,
*School of Human Movement Studies, University of Queensland, Inammatory Bowel Disease Research Unit, Royal Brisbane and Womens Hospital Research Foundation Clinical Research Center, Queensland Institute of Medical Research, Department of Gastroenterology, Royal Brisbane and Womens Hospital and Cancer and Population Studies Group, Queensland Institute of Medical Research, Queensland, Australia

Key words BMD, BMI, Crohns, osteoporosis, TNF-a. Accepted for publication 14 June 2006. Correspondence Dr Graham Radford-Smith, Department of Gastroenterology, Level 9A, Ned Hanlon Building, Royal Brisbane and Womens Hospital, PO Herston, Qld 4029, Australia. Email: graham_radford-smith@health.qld.gov.au

Abstract
Background and Aim: There is limited consensus on the major variables that determine bone integrity and bone loss in patients with Crohns disease. Twin and family studies in the general population indicate that up to 85% of variance in bone mineral density is inherited. The aim was to determine the prevalence of bone loss and both molecular and clinical risk factors for bone loss in a large Crohns disease population. Methods: This was a cross-sectional study of 304 patients with Crohns disease attending the Inammatory Bowel Disease unit at Royal Brisbane and Womens Hospital, Queensland. The results of bone density testing were ascertained directly and by a mailed questionnaire. Bone mineral density data were combined with clinical information and correlated with single nucleotide polymorphisms within the tumor necrosis factor-a (TNFa), interleukin-10, and NOD2/CARD15 genes. Results: Of 304 Crohns disease patients, 101 had undergone previous bone density testing. Forty-ve patients (45%) had been diagnosed with osteopenia and 18 (18%) were osteoporotic. After multivariate analysis, both the TNF-a GT haplotype and the -857 CC genotype showed strong associations with bone mineral density overall (P = 0.003 and P = 0.002, respectively). Body mass index (P = 0.01) and previous bowel resection in female patients (P = 0.03) were predictive of a higher spine bone density, while body mass index (P = 0.003) and the effect of years since rst bowel resection (P = 0.02) remained independent predictors of proximal femur bone mineral density. There were no other signicant associations observed. Conclusions: This study has identied a novel protective association between a TNF-a haplotype and bone mineral density in Crohns disease. It conrms the important inuence of body mass index and intestinal resection on bone loss in this population.

Introduction
Low bone mass is a frequent complication of Crohns disease (CD). The pathogenesis of osteoporosis in this chronic inammatory disorder is multifactorial. Contributing factors may include disease severity, systemic inammation (and cytokines), genetic susceptibility, corticosteroid treatment, bowel resection (and associated malabsorption), low bodyweight and lean mass, reduced physical activity levels, hormonal factors, comorbid conditions and inadequate nutrition.110 The independent effects of these risk factors on bone mineral density (BMD) are not well established and the magnitude of their impact may vary in different CD populations. Osteoporosis and subsequent fragility fractures are major causes of morbidity in the population, and are associated with high rates of morbidity and mortality, reduced quality of life, and occur

at a large nancial cost. Osteopenia and osteoporosis have been reported to occur in up to 80% and 30% of CD sufferers, respectively, although these values depend largely on the geographic location of the study population, as well as the BMD measurement technique and diagnostic criteria (T-score cut-offs) used.19,11 Despite careful management, some of these contributors may be largely unavoidable. However, if the presence of osteopenia and osteoporosis can be detected early and the appropriate therapy instituted, the probability of fracture is greatly diminished. In addition to clinical variables that are sometimes difcult to quantify and compare from study to study, the molecular basis of complex disorders such as CD and osteoporosis has also been extensively investigated. Major advances have been made in CD including the discovery of the NOD2/CARD15 susceptibility gene, while potential susceptibility genes for osteoporosis include
913

Journal of Gastroenterology and Hepatology 22 (2007) 913919 2006 The Authors Journal compilation 2006 Journal of Gastroenterology and Hepatology Foundation and Blackwell Publishing Asia Pty Ltd

Crohns disease and bone

N Lee et al.

the CL1A1 gene, osteoprotegerin, and the tumor necrosis factor receptor 2 (TNFR2) genes.1215 Limited data are available for bone loss in CD. The interleukin (IL)-1b -511 single nucleotide polymorphism (SNP; associated with hypersecretion of IL-1b) predicted lower BMD in a combined study of 36 patients with CD and 39 patients with ulcerative colitis.16 This has not been replicated as yet in the published literature. Two studies of the IL-6 gene in similar-sized inammatory bowel disease (IBD) cohorts have provided conicting results.17,18 The purpose of the present study was to assess the prevalence of osteopenia and osteoporosis in an Australian CD population and to examine the relationship between BMD and a number of patient and disease characteristics, including both known and potential susceptibility genes such as TNF-a, NOD2/CARD15, and IL-10. Potential roles for TNF-a and IL-10 in bone metabolism have been highlighted by both human and animal studies. Several studies have indicated a role for TNF-a as a skeletal catabolic agent, while the IL-10 knockout mouse has recently been investigated as a model of metabolic bone disease.1921 To date, there has been no research published pertaining to low bone mass in adult CD patients in Australia.

and need for immunosuppressive therapy. The latter was dened as use of azathioprine, 6-mercaptopurine, or methotrexate for a minimum of 6 consecutive months at currently accepted therapeutic doses.

Bone mineral densitometry

Bone densitometry scans were performed on a Lunar Prodigy densitometer (GE, Madison, WI, USA) by a single licensed operator. The in vivo coefcients of variation for our densitometer were 0.6% for L2-L4 BMD and 0.4% for total proximal hip BMD. The most recent lumbar spine (L2L4) and proximal femur (total hip) BMD results of subjects who had previously undergone osteoporosis screening on the aforementioned machine were used for analysis. A bone mineral density (BMD) measurement of 1 and <2.5 SD below the young normal mean (T-score) and 2.5 SD below the young normal mean (T-score) are clinically diagnostic of osteopenia and osteoporosis, respectively. A diagnosis of osteopenia or osteoporosis at either site was the criterion for a subject to be classied as osteopenic or osteoporotic, regardless of the higher BMD at the other site. Weight and height were taken from the densitometer printouts provided by subjects to reect weight and height at the time of measurement.

Methods
The study protocol was approved by the Human Research Ethics Committee of the Royal Brisbane and Womens Hospital, and the University of Queensland.

Genotyping
The role of genetic predisposition to bone loss in this population was investigated with genotyping of patients for functional promoter polymorphisms in the TNF-a (-308G/A and -857C/T), IL-10 (-592C/A and -1082G/A) and NOD2/CARD15 (SNP8, SNP12, and SNP13) genes. The 304 CD patients have previously been assessed in both a casecontrol association study and a genotypephenotype analysis for the aforementioned SNP.23 Genomic DNA were isolated from peripheral blood mononuclear cells using a standard proteinase-K/high salt extraction method.24 The CD patients were genotyped for all six SNP by polymerase chain reactionrestriction fragment length polymporphism as described previously.23 Investigators performing genotyping were blinded to patients clinical characteristics, those carrying out the phenotyping were blinded to genotypes, and the majority of the statistical analysis was carried out by individuals working outside the IBD research program to minimize any potential bias (DL).

Patients
Individuals with a primary diagnosis of CD were identied from our IBD database within the Department of Gastroenterology at the Royal Brisbane and Womens Hospital (RBWH). This hospital provides the only public gastroenterology service for the population of north Brisbane (approx. 500 000), and data on IBD patients have been prospectively collected since 1995. Of these, 304 CD patients between the ages of 20 and 80 years (175 female, 112 male) with complete clinical records pertaining to their disease including ongoing follow up with the IBD unit, were identied from the list and were invited to participate. A diagnosis of CD was conrmed using a combination of radiological, histological and endoscopic assessment, and based upon currently accepted criteria.22 Patients were excluded if they had any medical condition that may alter bone metabolism including renal or hepatic disease, thyrotoxicosis, hyperparathyroidism and hypogonadism.

Statistics
All continuous data are expressed as mean SD. Correlations between continuous variables were assessed using Pearsons correlation coefcients together with a 95% condence interval. The associations between BMD measurements and continuous covariates were tested using linear regression models (and t-tests). The differences between group mean BMD scores for the categorical covariates were tested using analysis of variance (anova) models (and F-tests). The anova model was also used to test for differences in mean BMD scores of groups dened by number of resections. Logistic regression analysis was conducted to assess the signicance of genetic or clinical associations or both, after adjusting for variables that attained P 0.2 at the univariate level. Haplotypes were constructed from the population genotype data using haplotype and regression analysis packages (brlr, hap-

Procedure
Previous bone density testing was determined by direct patient interview, chart review, and interrogation of the IBD database. Patients were also sent a brief questionnaire by mail asking if they had undergone previous bone density testing. Subjects were invited to complete the questionnaire with as much detail as possible and then return it by post with a copy of their most recent bone density test results if applicable. All the information collected was combined with other relevant demographic and disease variables contained in the IBD database. Clinical variables included age at diagnosis, disease duration, location, and behavior, smoking history, body mass index (BMI), family history, surgery for CD,
914

Journal of Gastroenterology and Hepatology 22 (2007) 913919 2006 The Authors Journal compilation 2006 Journal of Gastroenterology and Hepatology Foundation and Blackwell Publishing Asia Pty Ltd

N Lee et al.

Crohns disease and bone

Table 1

Clinical characteristics of 101 patients with Crohns disease

Clinical characteristics n = 101 Age (years) Sex (M/F) BMI (kg/m2) Disease duration (years) Age at diagnosis (years) Previous bowel resection (%) Duration since rst bowel resection (years) Disease site (%) Ileal Ileocolonic Colonic Perianal component Pack years of cigarette smoking Disease Phenotype (%) Stricturing Penetrating Inammatory Use of immunosuppression (%) BMI, body mass index. 42.9 37/64 25.3 13.4 29.5 66 11.3 44 38 18 24 17.5 29 42 29 73 13.0 7.9 7.0 12.9 7.3

more likely to inherit the high-TNF CC genotype (P = 0.002) and had a strong negative association with the low-TNF GT haplotype (P = 0.002; Table 2). None of the patients with normal BMD carried this haplotype, compared to 12.5% of those with osteopenia and 33.3% with osteoporosis. The gures for spinal BMD alone were very similar, with signicant associations between normal spine BMD and the CC genotype (P = 0.003), and the strong negative association with a GT haplotype (P = 0.005, Table 2). These associations remained highly signicant after multivariate analysis both for overall BMD versus the GT haplotype (P = 0.008, odds ratio [OR] 6.58) and spinal BMD versus GT haplotype (P = 0.003, OR 7.70, Table 3). There were no signicant associations between IL-10 or NOD2/CARD15 variants and BMD in this population.

15.3

Clinical variables and overall bone mineral density

Low BMD (T-score < -1 SD) was reported in 63 patients (63%), with 45 patients having osteopenia (45%) and 18 having osteoporosis (18%). Osteopenia was equally common at the spine (68%) and hip (70%), whereas osteoporosis was more prevalent at the spine (89%) compared to the hip (28%) in this population. The relationships between overall BMD and clinical characteristics are displayed in Table 4. Comparisons were made between those subjects with osteoporosis and normal BMD, and between those with osteopenia versus osteoporosis. Univariate analysis demonstrated signicant differences in BMI with osteoporotic patients having a lower BMI compared to those with normal BMD (P = 0.003), and osteopenic patient groups (P = 0.08). The osteoporotic group also displayed a signicantly longer disease duration compared to both normal BMD (P = 0.003) and osteopenic groups (P = 0.04). Osteoporotic patients also had a younger age at diagnosis, a longer time interval since rst bowel resection, and a lower take up rate for immunosuppression compared to the normal group (Table 4), but these differences failed to reach statistical signicance. Multivariate analysis of these clinical variables conrmed the strong association between bone mass and low BMI (P = 0.007) and a weaker but signicant association with longer disease duration (P = 0.05).

lo.stats) in the R statistical program.25 A proportional odds logistic regression model was used to test for an association between BMD group and TNF haplotype, after adjusting for stated clinical covariates. The clinical covariates were added to the model stepwise and retained if either the model t was signicantly improved (assessed by Akaike information criterion) or there was evidence of substantial confounding of the TNF haplotype effect estimate. The latter was deemed substantial if the percent change in the TNF haplotype was >10% with the addition/exclusion of the potential confounder. A signicance level of P < 0.05 was used unless otherwise stated.

Results
Overall, 101 patients were found to have undergone previous BMD testing (33%). The clinical characteristics of patients with both previous BMD data and complete phenotype are detailed in Table 1. Patient demographics and clinical variables were compared between patients with previous BMD testing and those without previous testing. The only signicant difference observed was that patients who had not undergone previous bone densitometry had a shorter disease duration (9.3 6.7 years) compared to those who had a previous BMD assessment (13.4 7.0 years; P < 0.001).

Hip and spine BMD

Hip and spine BMD scores are summarized in Table 5. In view of differences in the distribution of bone mass between hip and spine, further analysis was carried out to determine independent predictors of osteoporosis at these two sites. There were no signicant differences in bone density t-scores at the spine (P = 0.2) and hip (P = 0.3) between male and female patients, although men tended to have a higher t-score than women at both sites. Spine T-score (-0.95 1.3, range -4.4 to 2.0) and hip-T score (-0.89 1.0, range -4.0 to 1.8) were signicantly correlated (r = 0.66, P < 0.001) and both were positively associated with BMI (r = 0.245, P = 0.017 and r = 0.305, P = 0.003, respectively). Spine BMD had a weak negative correlation with years since rst bowel resection (r = -0.272, P = 0.026), while hip T-score had a similar but stronger negative correlation with years since rst resection (r = -0.454, P < 0.001) and a weaker negative association with disease duration (r = -0.271, P = 0.006). In view of the
915

Effect of genotype and haplotype

Associations between functional variants in the TNF-a, IL-10 and NOD2/CARD15 genes in this CD population have been published.23 In the present study and for the rst time, their relationships with BMD were explored by both genotype and haplotype analysis. The results for both the TNFa -857 variant alone, and the TNF-a haplotype (-857 and -308) were signicant. Individuals with normal BMD both at the hip and spine were signicantly

Journal of Gastroenterology and Hepatology 22 (2007) 913919 2006 The Authors Journal compilation 2006 Journal of Gastroenterology and Hepatology Foundation and Blackwell Publishing Asia Pty Ltd

Crohns disease and bone

N Lee et al.

Table 2

Association of BMD with TNF-a -857 genotypes and with the combined TNFa -308/-857 GT haplotype, in Crohns disease (n = 86) -857 CC genotype n (%) -857 CT/TT genotype n (%) Non-carriers of GT haplotype n (%) Carriers of GT haplotype n (%)

BMD groups

Overall BMD Normal Osteopenia Osteoporosis Spinal BMD Normal Osteopenia Osteoporosis

31 (100) 34 (85) 10 (67)* 44 (98) 23 (79) 8 (67)***

0 (0) 6 (15) 5 (33) 1 (2) 6 (21) 4 (33)

31 (100) 35 (88) 10 (67)** 44 (98) 24 (83) 8 (67)****

0 (0) 5 (12) 5 (33) 1 (2) 5 (17) 4 (33)

BMD, bone mineral density. *P = 0.002; **P = 0.002; ***P = 0.003; ****P = 0.005.

Table 3

Signicant independent associations with BMD in Crohns disease Variable Unadjusted OR Univariate 95%CI Lower Upper 32.32 1.13 0.99 26.82 1.21 Unadjusted P . 0.002 0.042 0.018 0.005 0.007 Adj. OR Multivariate 95%CI Lower 6.58 1.06 0.95 7.70 1.11 1.64 1.00 0.89 2.01 1.02 Upper 26.45 1.14 1.01 29.49 1.21 0.008 0.065 0.099 0.003 0.015 Adjusted P

Overall BMD (Spine + Hip) Spinal BMD

TNF haplotype BMI Disease duration TNF haplotype BMI

8.34 1.07 0.93 7.25 1.12

2.15 1.00 0.87 1.96 1.03

BMD, bone mineral density; BMI, body mass index; CI, condence interval; OR, odds ratio; TNF, tumor necrosis factor. Univariate and multivariate results for overall BMD and spinal BMD alone.

negative association between increasing time from rst resection and BMD, we compared BMD data between patients with and without surgical recurrence, but there was no signicant difference between these groups (P = 0.2). When adjusting for age, sex and BMI, the effect of years since rst resection on spine BMD diminished (P = 0.21), while previous bowel resection in female patients (P = 0.03) and higher BMI (P = 0.01) remained predictive of a higher spine BMD. When previous bowel resection was further explored by stratifying for sex, the effect remained moderately signicant (P = 0.05). After controlling for age, sex and BMI at the hip, only the effect of years since rst bowel resection (P = 0.02) and BMI (P = 0.003) remained signicant.

Discussion
This study demonstrates a high prevalence of bone loss in Australian CD patients. Frequencies of 45% for osteopenia and 18% for osteoporosis are similar to those reported in CD populations of the Northern hemisphere, where the prevalence of bone loss overall ranges from 50 to 75%.4,7,26 The prevalence of osteopenia and osteoporosis in the present CD population suggests that approximately 60% of Australian CD patients may be at greater risk of fracture compared to healthy controls. In the present study, osteopenia was equally common at the hip and the spine whereas osteoporosis was noticeably more common at the spine (16/18, 89%) compared to the hip (5/18, 28%). These results conrm those published recently by Sifledeen et al.27 but
916

differ from some earlier studies.7,28 The difference in bone density between the spine and hip in this study is similar to that found in postmenopausal women, and in tobacco related bone disease.2931 This is the rst study of bone loss in CD to investigate multiple genetic markers that may be linked to the underlying disease as well as the presence of bone loss in this population. We wanted to test the hypothesis that patients with CD who carried the highproducing TNF-a alleles for the -308 (A) and -857 (C) polymorphisms would be at greater risk of osteoporosis. This is based on the known association between this gene and both CD and bone resorption.3234 We also wanted to establish whether IL-10 had a role in human bone disease given the recent ndings in the murine IL-10 knockout,21 and our published ndings for the role of variants in this gene in the CD population.23 The genotype and haplotype results were highly signicant for TNF-a and are supported by recent data from studies of bone loss in the general population. Specically, normal overall BMD was strongly associated with the high-producing -857 CC genotype, and had a strong negative association with the low-TNF GT haplotype. None of the normal BMD patients carried the TNF-a GT haplotype, compared with 12% of osteopenic patients and 33% of osteoporotic subjects (Table 2). The multivariate analysis underlined the signicance of these genetic associations over clinical variables including BMI and disease duration (Table 3). Our ndings are supported by those of a Japanese study of BMD in 177 postmenopausal women.34 Specically, women carrying the -857 TT (low-TNF) genotype had signicantly lower BMD compared with those with either the CT or CC genotypes. It is impor-

Journal of Gastroenterology and Hepatology 22 (2007) 913919 2006 The Authors Journal compilation 2006 Journal of Gastroenterology and Hepatology Foundation and Blackwell Publishing Asia Pty Ltd

N Lee et al.

Crohns disease and bone

Table 4

Clinical features of those patients with normal bone density compared to those with either osteopenia or osteoporosis Bone mineral density Normal Osteopenic 45 (45) 60 43.8 12.9 24.7 8.2 29.5 12.7 14.3 6.8 62 7.2 7.3 44.4 33.3 22.3 28.6 26.7 44.4 28.9 17.1 73 -1.22 -1.16 Osteoporotic 18 (18) 56 44.0 16.4 21.8 4.0 25.6 15.2 18.4 7.4, 72 11.9 10.9 33.3 50.0 16.7 23.5 27.8 44.4 27.8 17.3 61 -2.81 -2.07

Clinical characteristics

n (%) Female (%) Age (years) BMI (kg/m2) Age at diagnosis (years) Disease duration (years) Previous bowel resection (%) Duration since rst bowel resection (years) Disease site (%) Ileal Ileo-colonic Colonic Perianal component Disease phenotype (%) Stricturing Penetrating Inammatory Pack years of cigarette smoking Immunosuppression use (%) BMD T-score Spine L2L4 Total hip

38 (37) 71 43.6 11.5 27.6 8.4 31.2 11.2 12.4 6.3 68 5.8 6.6 47.4 36.8 15.8 19.5 32.4 37.8 29.8 18.1 79

18.4

14.2

8.0

0.10 0.89 -0.10 0.68

0.76 0.68

0.92 0.85

BMI, body mass index; BMD, bone mineral density. Statistical signicance relative to normal BMD group (P < 0.05). Statistical signicance relative to osteopenic group (P < 0.05).

Table 5 Bone mineral density results in 101 patients with Crohns disease Mean Spine T-score (young-adult matched) Spine Z-score (age- and weight-matched) Hip T-score (young-adult matched) Hip Z-score (age- and weight-matched) -0.95 -0.57 -0.89 -0.55 SD 1.33 1.26 1.01 0.92 Range -4.4 to 2.0 -3.0 to 2.6 -4.0 to 1.8 -2.5 to 1.9

tant to note that the genotype frequencies for TNF-a -857 in that Japanese population differ from the present CD population overall and from the present controls due to differences in race. However, the fact that Japanese patients with CD alone show signicant associations with certain TNF-a polymorphisms (TNFa -1031, -863 and -857) as do Caucasian CD populations (-308 and -857), suggests that TNF-a may inuence BMD and intestinal inammation in both these populations despite their racial differences.23,33,35 There are no other studies of the -857 variant in bone disease, while the TNF-a -863 variant has been studied in the aforementioned Japanese population, and in a cohort of adolescent female subjects.36 The results of these studies are discordant.

The effect of the TNF-a -308 variant on bone strength has recently been analyzed in a large Caucasian population of 4306 older women who were prospectively assessed for bone strength phenotype and risk of fracture.37 In that population, the TNF-a AA genotype was associated with signicantly better femoral neck strength and a 63% lower age-adjusted risk of hip fracture compared to women with the TNF-a GG genotype. However, the study did not show signicant differences in BMD between these genotypes, indicating that TNF-a may also inuence other determinants of bone strength including the overall size and geometry of bone. This would be consistent with previous studies demonstrating that up to 80% of variance in measures of femoral neck cross-sectional area may be under genetic regulation. A number of other studies, including measurement of circulating TNF-a in postmenopausal women, have suggested an important role for TNF-a as a skeletal catabolic agent.19,20,38,39 Aging may see a cytokine-driven increase in osteoclastic bone resorption at the endosteal surface. This may be opposed by age-related periosteal bone formation. The majority of observations in this area have been based upon in vitro work using cultured cells or cell lines, or in animal models.32 There are no human functional data that clearly implicate TNF-a as a major resorptive agent of bone, and therefore its role in metabolic bone disease remains unclear. This may be related to the difculties in studying circulating and tissue cytokines in view of their very short half-life. In view of this, genetically determined variation in the concentration of TNF-a may be a more reliable method to determine whether this cytokine plays a role in bone metabolism in the long term.
917

Journal of Gastroenterology and Hepatology 22 (2007) 913919 2006 The Authors Journal compilation 2006 Journal of Gastroenterology and Hepatology Foundation and Blackwell Publishing Asia Pty Ltd

Crohns disease and bone

N Lee et al.

Several proposed functional SNP have been identied in the TNF-a promoter region, including the two SNP used in the present study. However, the functional studies do show conicting results and none of them have measured TNF-a levels in bone. However, a more comprehensive approach was adopted in the present study, in determining a genetic association between the TNF-a gene and bone loss by incorporating both genotype and haplotype analysis. There is strong correlation between these analyses both for BMD overall and spinal BMD alone, underlining the signicance of the ndings. An alternative explanation for our ndings is that the true association in the present study is between BMD and another gene that is in linkage disequilibrium with the reported TNF-a genotypes and haplotype at 6p21. There are several possibilities here including the regulatory protein, bone morphogenetic protein-6 (BMP-6), located at 6p24, which regulates growth and differentiation of mesodermally derived tissues as well as having a protective effect on intestinal tissue in animal models of IBD.40 A detailed analysis of associations between BMD and a range of clinical variables identied a positive association with BMI and a negative association with disease duration. Osteoporotic patients maintained a signicantly lower BMI compared to those with normal BMD, and osteopenic patients, and conrms previous observations from several recent studies.2,4,41 However, unlike previous reports, we found that osteoporotic patients had a signicantly longer disease duration compared to the normal and osteopenic groups. These results raise the important issue of case ascertainment in such series, and the impact that this may have on patient clinical characteristics and the concordance of results between studies. The mean disease duration in the present study was 13.4 7.0 years, compared to a range of 4.09.6 years in previous studies of a similar size.2,4,5,7 Recent longitudinal studies on CD have underlined the importance of disease evolution, indicating that a signicant percentage of less complex or inammatory CD will progress to either the penetrating or stricturing phenotype or both.42,43 One study estimates this period of evolution to be at least 8 years.43 Evolution of the disease into a more aggressive form may have signicant effects on bone metabolism, related to disease activity and treatment modalities. In contrast to disease duration, age at diagnosis and current age did not have any signicant associations with BMD. Age at diagnosis became even less signicant in a multivariate model, indicating that any impact on BMD was through disease duration. The results for current age and BMD are consistent with the disease aficting a young, otherwise t population. A previous study that demonstrated age as a risk factor for osteoporosis had a signicantly younger mean age (33 years) and shorter disease duration (8.0 years) compared to the present study.4 These factors are likely to impact on disease evolution as discussed here. The signicant difference in frequency of osteoporosis at the spine (89% of all osteoporotic patients) compared to the hip (28%) in this population indicated that separate analysis of risk factors for bone loss at these two sites may be worthwhile. After multivariate analysis, we conrmed the protective effect of higher BMI at both hip and spine, while previous bowel resection was also protective for spine BMD in women only. Years since rst bowel resection had a negative correlation with hip BMD. Several recent studies have discussed the steroid-sparing effect of bowel resection in CD. de Jong et al. reported BMI and a history of bowel resection to be
918

predictive of bone density,44 while Vestergaard et al. demonstrated a signicantly reduced fracture risk in patients who had undergone bowel surgery.45 Removal of the inamed segment may allow temporary relief from pro-inammatory cytokines such as IL-6 that play a key role in bone resorption, as well as reducing the need for oral corticosteroids for a period of time. Although de Jong et al. reported a greater steroid burden in osteoporotic patients compared to those with a normal BMD, neither cumulative corticosteroid dose nor duration of steroid use were found to be predictive of BMD due to the effect of potential confounders.44 The protective effect of intestinal resection is likely to wear off, as indicated by the negative association between hip BMD and years since rst surgery in the present study. Although the prevalence of abnormal bone status may be overestimated in the present study due to subject selection, it is evident that osteoporosis is a frequent complication of CD in Australia. Importantly, disease duration was the only variable that differed signicantly between those patients who either had or had not undergone BMD assessment. We have identied a novel, highly signicant and independent association between both TNF-a genotype and haplotype and BMD in this population. This observation is supported by very recent independent studies in postmenopausal women.34,37 The temporary reduction in intestinal inammation and associated reduction in steroid usage seen with intestinal resection may temporarily prevent the abnormal bone metabolism observed in patients without previous resection and in those with greater duration since resection.

Acknowledgments
Ms Naomi Lee is supported by an Australian National PhD Scholarship; Dr Elizabeth Fowler is supported by the Reginald Ferguson Fellowship, University of Queensland; and Dr Graham RadfordSmith is supported by a Practitioner Fellowship from the Queensland Institute of Medical Research and the Royal Brisbane and Womens Hospital Research Foundation.

References
1 Schoon EJ, Geerling BG, Van Dooren IM et al. Abnormal bone turnover in long-standing Crohns disease in remission. Aliment. Pharmacol. Ther. 2001; 15: 78392. 2 Andreassen H, Hylander E, Rix M. Gender, age, and body weight are the major predictive factors for bone mineral density in Crohns disease: a case-control cross-sectional study of 113 patients. Am. J. Gastroenterol. 1999; 94: 8248. 3 Ardizzone S, Bollani S, Bettica P, Bevilacqua M, Molteni P, Bianchi Porro G. Altered bone metabolism in inammatory bowel disease: there is a difference between Crohns disease and ulcerative colitis. J. Intern. Med. 2000; 247: 6370. 4 Habtezion A, Silverberg MS, Parkes R, Mikolainis S, Steinhart AH. Risk factors for low bone density in Crohns disease. Inamm. Bowel Dis. 2002; 8: 8792. 5 Robinson RJ, al-Azzawi F, Iqbal SJ et al. Osteoporosis and determinants of bone density in patients with Crohns disease. Dig. Dis. Sci. 1998; 43: 25006. 6 Jahnsen J, Falch JA, Aadland E, Mowinckel P. Bone mineral density is reduced in patients with Crohns disease but not in patients with ulcerative colitis: a population based study. Gut 1997; 40: 313 19.

Journal of Gastroenterology and Hepatology 22 (2007) 913919 2006 The Authors Journal compilation 2006 Journal of Gastroenterology and Hepatology Foundation and Blackwell Publishing Asia Pty Ltd

N Lee et al.

Crohns disease and bone

7 Bjarnason I, Macpherson A, Mackintosh C, Buxton-Thomas M, Forgacs I, Moniz C. Reduced bone density in patients with inammatory bowel disease. Gut 1997; 40: 22833. 8 Staun M, Tjellesen L, Thale M, Schaadt O, Jarnum S. Bone mineral content in patients with Crohns disease: a longitudinal study in patients with bowel resections. Scand. J. Gastroenterol. 1997; 32: 22632. 9 Haugeberg G, Vetvik K, Stallemo A, Bitter H, Mikkelsen B, Stokkeland M. Bone density reduction in patients with Crohn disease and associations with demographic and disease variables: cross-sectional data from a population-based study. Scand. J. Gastroenterol. 2001; 36: 75965. 10 Bernstein CN, Leslie WD. Osteoporosis and inammatory bowel disease. Aliment. Pharmacol. Ther. 2004; 19: 94152. 11 Cino M, Greenberg GR. Bone mineral density in Crohns disease: a longitudinal study of budesonide, prednisone, and nonsteroid therapy. Am. J. Gastroenterol. 2002; 97: 91521. 12 Hugot JP, Chamaillard M, Zouali H et al. Association of NOD2 leucine-rich repeat variants with susceptibility to Crohns disease. Nature 2001; 411: 599603. 13 Langdahl BL, Carstens M, Stenkjaer L, Eriksen EF. Polymorphisms in the osteoprotegerin gene are associated with osteoporotic fractures. J. Bone Miner. Res. 2002; 17: 124555. 14 Mann V, Hobson EE, Li B et al. A COL1A1 Sp1 binding site polymorphism predisposes to osteoporotic fracture by affecting bone density and quality. J. Clin. Invest. 2001; 107: 899907. 15 Spotila LD, Rodriguez H, Koch M et al. Association of a polymorphism in the TNFR2 gene with low bone mineral density. J. Bone Miner. Res. 2000; 15: 137683. 16 Nemetz A, Toth M, Garcia Gonzalez MA et al. Allelic variation at the interleukin 1beta gene is associated with decreased bone mass in patients with inammatory bowel diseases. Gut 2001; 49: 6449. 17 Schulte CM, Dignass AU, Goebell H, Roher HD, Schulte KM. Genetic factors determine extent of bone loss in inammatory bowel disease. Gastroenterology 2000; 119: 90920. 18 Schulte C, Goebell H, Roher HD, Schulte KM. Genetic determinants of IL-6 expression levels do not inuence bone loss in inammatory bowel disease. Dig. Dis. Sci. 2001; 46: 25218. 19 Roggia C, Gao Y, Cenci S et al. Up-regulation of TNF-producing T cells in the bone marrow: a key mechanism by which estrogen deciency induces bone loss in vivo. Proc. Natl Acad. Sci. USA 2001; 98: 13 9605. 20 Kimble RB, Matayoshi AB, Vannice JL, Kung VT, Williams C, Pacici R. Simultaneous block of interleukin-1 and tumor necrosis factor is required to completely prevent bone loss in the early postovariectomy period. Endocrinology 1995; 136: 305461. 21 Cohen SL, Moore AM, Ward WE. Interleukin-10 knockout mouse: a model for studying bone metabolism during intestinal inammation. Inamm. Bowel Dis. 2004; 10: 55763. 22 Lennard-Jones JE. Classication of inammatory bowel disease. Scand. J. Gastroenterol. Suppl. 1989; 170: 26;discussion 1619. 23 Fowler EER, Hume G, Johnstone S et al. TNF and IL10 SNPs act together to predict disease behaviour in Crohns disease. J. Med. Genet. 2005; 42: 5238. 24 Miller SA, Dykes DD, Polesky HF. A simple salting out procedure for extracting DNA from human nucleated cells. Nucleic Acids Res. 1988; 16: 1215. 25 Team RDC. A Language and Environment for Statistical Computing (3-900051-00-3).Vienna: Foundation for Statistical Computing, 2004. Available from: http://www.R-project.org. 26 Dinca M, Fries W, Luisetto G et al. Evolution of osteopenia in inammatory bowel disease. Am. J. Gastroenterol. 1999; 94: 12927.

27 Sifledeen JS, Fedorak RN, Siminoski K et al. Bones and Crohns: risk factors associated with low bone mineral density in patients with Crohns disease. Inamm. Bowel Dis. 2004; 10: 2208. 28 Schoon EJ, van Nunen AB, Wouters RS, Stockbrugger RW, Russel MG. Osteopenia and osteoporosis in Crohns disease: prevalence in a Dutch population-based cohort. Scand. J. Gastroenterol. Suppl. 2000; 232: 437. 29 Clements D, Compston JE, Evans WD, Rhodes J. Hormone replacement therapy prevents bone loss in patients with inammatory bowel disease. Gut 1993; 34: 15436. 30 Slemenda CW, Hui SL, Longcope C, Evans AA, Kirsner JB. Cigarette smoking, obesity, and bone mass. J. Bone Miner. Res. 1989; 4: 73741. 31 Hopper JL, Seeman E. The bone density of female twins discordant for tobacco use. N. Engl. J. Med. 1994; 330: 38792. 32 Nanes MS. Tumor necrosis factor-alpha: molecular and cellular mechanisms in skeletal pathology. Gene 2003; 321: 115. 33 van Heel DA, Udalova IA, De Silva AP et al. Inammatory bowel disease is associated with a TNF polymorphism that affects an interaction between the OCT1 and NF (-kappa) B transcription factors. Hum. Mol. Genet. 2002; 11: 12819. 34 Furuta I, Kobayashi N, Fujino T et al. Bone mineral density of the lumbar spine is associated with TNF gene polymorphisms in early postmenopausal Japanese women. Calcif. Tissue Int. 2004; 74: 50915. 35 Negoro K, Kinouchi Y, Hiwatashi N et al. Crohns disease is associated with novel polymorphisms in the 5-anking region of the tumor necrosis factor gene. Gastroenterology 1999; 117: 10628. 36 Wennberg P, Nordstrom P, Lorentzon R, Lerner UH, Lorentzon M. TNF-alpha gene polymorphism and plasma TNF-alpha levels are related to lumbar spine bone area in healthy female Caucasian adolescents. Eur. J. Endocrinol. 2002; 146: 62934. 37 Moffett SP, Zmuda JM, Oakley JI et al. Tumour necrosis factor alpha polymorphism, bone strength phenotypes, and the risk of fracture in older women. J. Clin. Endocrinol. Metab. 2005; 90: 34917. 38 Pacici R. Estrogen, cytokines, and pathogenesis of postmenopausal osteoporosis. J. Bone Miner. Res. 1996; 11: 104351. 39 Bertolini DR, Nedwin GE, Bringman TS, Smith DD, Mundy GR. Stimulation of bone resorption and inhibition of bone formation in vitro by human tumour necrosis factors. Nature 1986; 319: 51618. 40 Maric I, Poljak L, Zoricic S et al. Bone morphogenetic protein-7 reduces the severity of colon tissue damage and accelerates the healing of inammatory bowel disease in rats. J. Cell Physiol. 2003; 196: 25864. 41 Tjellesen L, Nielsen PK, Staun M. Body composition by dual-energy X-ray absorptiometry in patients with Crohns disease. Scand. J. Gastroenterol. 1998; 33: 95660. 42 Louis E, Collard A, Oger AF et al. Behaviour of Crohns disease according to the Vienna classication: changing pattern over the course of the disease. Gut 2001; 49: 77782. 43 Picco MF, Bayless TM. Tobacco consumption and disease duration are associated with stulizing and stricturing behaviors in the rst 8 years of Crohns disease. Am. J. Gastroenterol. 2003; 98: 3638. 44 de Jong DJ, Corstens FH, Mannaerts L, van Rossum LG, Naber AH. Corticosteroid-induced osteoporosis: does it occur in patients with Crohns disease? Am. J. Gastroenterol. 2002; 97: 201115. 45 Vestergaard P, Krogh K, Rejnmark L, Laurberg S, Mosekilde L. Fracture risk is increased in Crohns disease, but not in ulcerative colitis. Gut 2000; 46: 17681.

Journal of Gastroenterology and Hepatology 22 (2007) 913919 2006 The Authors Journal compilation 2006 Journal of Gastroenterology and Hepatology Foundation and Blackwell Publishing Asia Pty Ltd

919