Whole body BMC was assessed in 104 children and young adults with CD and 233 healthy controls. CD was associated with significant deficits in BMC and lean mass, relative to height. Adjustment for lean mass eliminated the bone deficit in CD. Steroid exposure was associated with short stature but not bone deficits relative to height.
Introduction: Children with Crohn disease (CD) have multiple risk factors for impaired bone accrual. The confounding effects of poor growth and delayed maturation limit the interpretation of prior studies of bone health in CD. The objective of this study was to assess BMC relative to growth, body composition, and maturation in CD compared with controls.
Materials and Methods: Whole body BMC and lean mass were assessed by DXA in 104 CD subjects and 233 healthy controls, 4–26 years of age. Multivariable linear regression models were developed to sequentially adjust for differences in skeletal size, pubertal maturation, and muscle mass. BMC-for-height z scores were derived to determine CD-specific covariates associated with bone deficits.
Results: Subjects with CD had significantly lower height z score, body mass index z score, and lean mass relative to height compared with controls (all p < 0.0001). After adjustment for group differences in age, height, and race, the ratio of BMC in CD relative to controls was significantly reduced in males (0.86; 95% CI, 0.83, 0.94) and females (0.91; 95% CI, 0.85, 0.98) with CD. Adjustment for pubertal maturation did not alter the estimate; however, addition of lean mass to the model eliminated the bone deficit. Steroid exposure was associated with short stature but not bone deficits.
Conclusion: This study shows the importance of considering differences in body size and composition when interpreting DXA data in children with chronic inflammatory conditions and shows an association between deficits in muscle mass and bone in pediatric CD.
Crohn disease (CD) is characterized by gastrointestinal tract inflammation with malabsorption, malnutrition, anemia, pubertal delay, growth failure, and impaired bone mineral accretion in children and adolescents. The pathogenesis of osteopenia in children with CD is multifactorial, including increased production of bone-resorptive inflammatory cytokines, delayed pubertal maturation, decreased physical activity, vitamin D deficiency, and glucocorticoid exposure.(1–4) Osteopenia in children and adolescents with CD poses an immediate fracture risk and compromises peak bone mass in early adulthood, resulting in skeletal fragility later in life.(5,6)
Numerous studies have shown decreased areal BMD (aBMD) using DXA in children and adults with CD.(2,3,7–10) While decreased aBMD is the accepted measure of osteoporosis in adults,(11) it is well recognized that aBMD is confounded by bone size in children and adolescents.(12–14) The use of aBMD in children with CD may lead to spurious associations with other body size-dependent measures, such as corticosteroid exposure, muscle mass, or dietary intake.(15)
Alternative strategies have been proposed for the analysis and interpretation of DXA results in children with abnormal growth and body composition.(16–18) A recent study showed that DXA estimates of whole body BMC, adjusted for height, correlated with cortical BMC, dimensions, and strength, as measured by pQCT.(17) In contrast, adjustment of whole body BMC for bone area was not correlated with bone strength. Schoenau et al.(18) proposed that the functional assessment of bone deficits should determine whether muscle mass is reduced relative to body size (height) and whether bone mass is normally adapted to muscle mass. These approaches have been described in small series of children with varied chronic diseases.(16,18,19)
The objective of this study was to assess whole body BMC relative to height and muscle mass in children and young adults with CD compared with healthy controls and to determine the independent effects of growth, maturation, body composition, and disease characteristics on BMC deficits.
MATERIALS AND METHODS
Individuals with CD, 4–25 years of age, treated at the Children's Hospital of Philadelphia or the University of Pennsylvania Medical Center, were eligible for the study. Diagnosis was confirmed by radiographic, histological, and clinical information. Individuals with CD were excluded for other medical conditions unrelated to CD that could potentially affect growth or bone health. Lumbar spine aBMD, vitamin D levels, growth, and maturation have been reported in these subjects.(1–4)
Healthy control subjects were recruited from general pediatric clinics in the surrounding community and through newspaper advertisements. Control subjects were excluded for any co-existing conditions known to affect growth, nutritional status, dietary intake, pubertal development, or bone mineralization. The protocol was approved by the Institutional Review Board at the Children's Hospital of Philadelphia. Informed consent was obtained from the young adult participants and the parents or guardians of those <18 years of age. Assent was obtained from those <18 years old.
Medical records were reviewed for age at disease onset, disease characteristics, medical, nutritional, and surgical interventions, and hematologic and biochemical abnormalities since diagnosis. Site of disease was coded as upper gastrointestinal tract only, colon only, or both. Laboratory abnormalities were defined as one or more values below the normal reference range (based on age and gender) after the diagnosis of CD had been established. Among subjects <18 years of age, CD severity was assessed using the Pediatric Crohn's Disease Activity Index (PCDAI), which is based on history (30%), physical examination (30%), laboratory data (20%), and height velocity (20%).(20,21) Scores are categorized as follows: no disease activity (0–10) mild disease activity (11–30), and moderate to severe disease activity (>30).
Use of the following medications was documented: 6-mercaptopurine, azulfidine, mesalamine (Pentasa or Asacol), metronidazole, calcium supplementation, and corticosteroid enemas. All doses of enteral and parenteral corticosteroids were noted and were converted to prednisone equivalents. Corticosteroid exposure was summarized as lifetime cumulative prednisone dose (g) and cumulative milligram per kilogram body weight at the time of the dose (mg/kg). Average doses during intervals of corticosteroid therapy were summarized as milligrams per day and milligrams per kilograms per day. This study was conducted before the use of TNF-α inhibitors in pediatric CD.
Anthropometry and pubertal development
Weight and height were measured using a digital scale to the nearest 0.1 kg (Scaltronix, White Plains, NY, USA) and a stadiometer to the nearest 0.1 cm (Holtain, Croswell, Crymych, UK), respectively. Age- and gender-specific SD scores (z scores) for weight, height, and body mass index (BMI) were calculated using the National Center for Health Statistics (NCHS) 2000 Center for Disease Control growth data using the LMS method.(22) Pubertal stage was assessed according to the method of Tanner et al.(23)
Whole body DXA scans were performed using a Hologic QDR 2000 bone densitometer (Hologic, Bedford, MA, USA) with a fan beam in the array mode in all subjects. The measurements were performed using standard supine positioning techniques and were analyzed to generate estimates of BMC (g), lean mass (kg), and fat mass (kg). All whole body DXA measures excluded the skull (post-cranial). The instrument was calibrated daily with a hydroxyapatite phantom, and the in vitro CV at our institution is <0.6%. The in vivo CV is <1% when measured in adults.
Blood samples were obtained from individuals with CD at the time of the study visit. Serum albumin (g/dl), total protein (g/dl), hemoglobin (g/dl), and erythrocyte sedimentation rate (mm/h) were measured in the Clinical Laboratory at the Children's Hospital of Philadelphia using standard techniques. Serum 25(OH)vitamin D levels were measured by radioimmunoassay with a radioiodinated tracer.(24)
Descriptive analyses included means, SDs, median, and ranges of continuous variables and distributions of categorical variables. Differences of means were assessed using Student's t-test if the variable was normally distributed. The Wilcoxon rank-sum test was used if these data were not normally distributed. Group differences in categorical variables were assessed using the χ2 or Fisher's exact test, where appropriate. Analyses were conducted using Stata 8.2 (Stata Corp., College Station, TX, USA). Two-sided tests of hypotheses were used, and a p value <0.05 was considered to be statistically significant.
The primary outcome was whole body BMC. DXA results are traditionally expressed relative to age and gender. However, assessment of BMC in children and adolescents requires adjustment for bone size. Natural log transformation of BMC and height results in a linear relationship between the two measures that does not exist in nontransformed models. A prediction model using this approach has been described in children(16) and validated with QCT.(17)
A series of log-transformed models for whole body BMC relative to height were sequentially adjusted for known determinants of BMC that may confound the comparison of individuals with CD and healthy controls. Given the known gender differences in bone mineral accretion during growth,(25) all results are presented stratified on gender. All models were adjusted for race (black versus all others). Models were subsequently adjusted for Tanner stage of pubertal maturation (stage 1 as the referent group, with indicator variables for Tanner stages 2–5) and lean mass. To compare the relationship between lean mass and BMC in subjects with CD and in healthy controls, a CD-lean mass interaction term was added to a multivariable regression model of BMC on CD, height, age, and race. The fit of each model was assessed through the adjusted R2 value. The assumptions of the regression models were assessed through graphical checks, the Shapiro-Wilk test of normality of the residuals, the Ramsey omitted variable test, and the Cook-Weisburg test for heteroscedasticity.
Because the outcome variable was log-transformed, the independent effect of CD in each multivariate model was summarized as the adjusted ratio of the outcome measure in the subjects with CD divided by the outcome measure in the control subjects; we note that the ratios have no units. The adjusted ratio and 95% CI for each covariate were calculated as the exponentiated estimate of the regression parameter and the 95% confidence limits of the regression parameter.
To aid the clinical interpretation of the magnitude of the deficits in CD in the series of multivariable models and to assess the impact of CD-specific effects on bone within the subjects with CD, the whole body BMC results were converted to gender-specific BMC z scores. Data from the control subjects were used to derive the predicted value of BMC relative to either age or height using linear regression. The residuals from each of the models were not heteroscedastic; therefore, the root mean square error served as the SD at all levels of the predictor variable. The z scores were calculated as follows: ([Observed − Predicted]/SD). For example, a BMC-for-height z score of −1.0 indicates a whole body BMC that was 1 SD below that of control subjects of the same height and gender. Lean mass-for-height z score, a secondary outcome, was derived in a similar manner. The z scores were compared between subjects with CD versus controls in regression models that also included other explanatory variables.
Bivariate analyses of CD-specific factors and anthropometric measures associated with low BMC-for-height z scores within the subjects with CD were identified using simple logistic regression. BMC-for-height z score was considered abnormal if < −1.0 (equivalent to the 16th percentile). The correlations between glucocorticoid exposure, BMC-for-height z score, and height-for-age z score were assessed using Pearson product-moment estimates.
Subject characteristics and body composition
A total of 104 subjects with CD and 233 healthy controls completed the study. Subject characteristics are summarized in Table 1. Controls were significantly younger and less mature than the subjects with CD. The white predominance among the subjects with CD was consistent with the demographics of the disease. Comparison of age distributions within Tanner stages suggested that CD was associated with delayed pubertal maturation: within Tanner stages 2 and 4, individuals with CD were an average of 1.4 and 1.5 years older than controls (p < 0.05 and p < 0.01, respectively), adjusted for gender and race. There were no gender differences in age, Tanner stage, height, or BMI z score within the participants with CD.
Table Table 1. Subject Characteristics in Crohn Disease and Control Participants
Individuals with CD had significantly lower height, weight, BMI-for-age, and lean mass-for-height z scores than healthy controls (all p < 0.0001). The BMI z score distribution within the control subjects was consistent with a recent report in the U.S. population.(26) The mean lean mass-for-height z score, adjusted for age and race, was significantly decreased in the subjects with CD (p < 0.001).
Disease characteristics, medications, and laboratory values were available in 76–88% of individuals with CD and are detailed in Table 2. Among the 88% with complete medication data, 90% were treated with glucocorticoids. Treatment and disease characteristics were compared between the male and female subjects with CD to determine possible differences in disease severity. Males were exposed to glucocorticoids for a significantly greater duration than females (median duration, 15.2 months; [range, 0–128.3 months] versus 8.4 months [0–79.5 months]; p = 0.01), receiving a greater total dose over the course of their disease (median dose, 7.9 g [range, 0–74.0 g] versus 6.2 g [range, 0–41.7 g]; p = 0.01). However, during the exposure to glucocorticoids, the doses (mg/day and mg/kg/day) were similar. Males were more frequently treated with mesalamine (Pentasa) than females (32.7% versus 16.2%, p = 0.01) and were more likely to have had hypoalbuminemia (27.3% versus 8.1%, p = 0.02).
Table Table 2. Disease Characteristics, Medication Exposure, and Laboratory Assessment in Participants With Crohn Disease
Whole body BMC
A series of models were developed to assess BMC in CD compared with control subjects. All models were adjusted for the confounding effects of age and race, given the differences between the subjects with CD and the controls. Because delayed pubertal maturation is associated with decreased bone mass,(27) adjusting for pubertal stage may mask clinically important differences in whole body BMC between CD and controls. Accordingly, gender-specific statistical models are presented with and without adjustment for Tanner stage of pubertal maturation. Table 3 summarizes four sequential models in males and females. The ratio in each model represents the expected BMC in a subject with CD divided by the BMC in a control subject with the same values for the stated covariates (e.g., the ratio of BMC in CD subjects versus controls of the same age and race).
Table Table 3. Sequential Linear Regression Models of the Independent Effect of Crohn Disease on Whole Body BMC
The least adjusted models assessed whole body BMC in CD compared with controls, adjusted for age and race, and revealed substantial deficits. The ratio of BMC in males with CD compared with controls was 0.74 (95% CI, 0.68, 0.82), representing a 26% reduction in BMC on average. Assessment of BMC without consideration of the decreased skeletal size for age (decreased height z scores) in subjects with CD group may overestimate bone deficits. Accordingly, the second model was also adjusted for height (Table 3). Figure 1 shows the distribution of bone measures expressed relative to age and relative to height. Log transformation improved model fit. Adjustment for height attenuated the CD effect; however, significant BMC deficits persisted in males and females with CD compared with controls. To determine if delayed pubertal maturation for age contributed to the decreased BMC observed relative to height and age in subjects with CD, the third model adjusted for Tanner stage. Adjustment for delayed pubertal maturation did not appreciably change the estimate of BMC deficits in CD. Tanner stage was not independently associated with BMC in males after adjustment for age and height (p ≥ 0.16 for Tanner stages 2–5); however, Tanner stage was independently associated with BMC in females, adjusted for age and height (p < 0.05 for Tanner stages 4 and 5, p ≥ 0.08 for Tanner stages 2 and 3).
The fourth and final model was also adjusted for whole body lean mass to determine if the significant deficits in BMC in CD were associated with the lean mass deficits. Lean mass was independently associated with BMC in subjects with CD and healthy control subjects (both p < 0.001). This relationship between lean mass and BMC was conserved in subjects with CD (interaction p > 0.2). No significant differences in BMC were detected between subjects with CD and controls when lean mass was included in the model. In males, results were unchanged in a model adjusted for height, age, and lean mass (0.96; 95% CI, 0.91, 1.01; p = 0.09), but omitting Tanner stage, consistent with the insignificant relationship between Tanner stage and BMC in males in model 3.
BMC z score models produced similar results (Table 4). There were significant deficits in BMC in males and females relative to age and relative to height. These deficits were attenuated by adjustment of the BMC-for-height z score for age, race, Tanner stage, and lean mass.
Table Table 4. Multiple Linear Regression Models Summarizing the Independent Effect of CD on Whole Body BMC Z Scores
The gender-specific BMC-for-height z scores are shown in Fig. 2. The mean unadjusted BMC-for-height z scores were −0.65 ± 0.85 in males and −0.45 ± 0.95 in females. The z scores were less than −1.0 in 34% of individuals with CD. By definition, the z scores were 0.0 ± 1.0 in controls, with 16% less than −1.0.
CD-specific factors and bone health
Having established group differences in whole body BMC between individuals with CD and controls, models were designed to examine the association of BMC-for-height z scores with anthropometric measures and the CD-specific variables summarized in Table 2. Factors that were found to be significantly associated with a BMC-for-height z score less than −1.0 in bivariate analyses are listed in Table 5. Serum vitamin D levels were not correlated with bone deficits. Many of these covariates were co-linear and indicative of severe disease. For example, use of mesalamine (Pentasa) was predictably associated with hypoalbuminemia (p = 0.05) and site of disease (p = 0.03), because it is generally used to target inflammation in the upper gastrointestinal tract. Individuals with a history of hypoalbuminemia were more likely to have received nasogastric feeding (p = 0.02). A 1 SD decrease in the BMI z score was associated with an increased odds of exposure to parenteral nutrition (OR, 1.6; 95% CI, 1.03, 2.54; p = 0.04).
Table Table 5. Bivariate Analyses of Factors Associated with Decreased BMC-for-Height Z Score in CD
Corticosteroid exposure and bone health
The relationships between corticosteroid exposure, growth, and BMC-for-height z scores were examined. None of the corticosteroid measures summarized in Table 2 were significantly correlated with BMC-for-height z scores. However, height z score was negatively and significantly associated with duration of exposure to corticosteroids (r = −0.24, p = 0.02), cumulative corticosteroid dose (g; r = −0.23, p = 0.03), and cumulative mg/kg (r = −0.36, p < 0.001). Height z score was not associated with the duration of disease or average corticosteroid exposure in milligrams per kilograms per day. The correlation of cumulative PCDAI and height z score approached statistical significance (r = −0.27; p = 0.06).
This examination of 104 individuals with CD showed significant deficits in whole body BMC compared with a concurrent group of 233 healthy controls. Adjustment for the confounding effects of stature, race, and delayed pubertal maturation attenuated, but did not fully explain, the bone deficits. However, adjustment for the decreased lean mass in CD eliminated the BMC deficits in males and females. These analyses show the importance of a concurrent comparison group of healthy subjects to adjust for differences in growth and body composition across the broad age range of subjects.
Cortical bone comprises 80% of the skeletal bone mass; therefore, whole body DXA BMC reflects predominantly cortical bone mass and dimensions. The primary function of cortical bone is mechanical strength. Our recent study showed that measures of DXA whole body BMC relative to height correlated with cortical bone mass, dimensions, and strength, as validated by pQCT.(17) Lean mass was not included in our recent study because children with disease states may have abnormal body composition, which may independently impact bone strength. The BMC deficits in CD reported here likely represent structurally significant deficits in cortical bone, which are associated with deficits in lean mass and consistent with increased fracture risk in CD.(6)
Cowan et al.(28) reported whole body DXA results in 32 subjects with CD compared with 58 healthy children. The BMC results were adjusted for bone area, age, height, weight, pubertal stage, and gender and showed a 4% decrement compared with controls. However, a recent study suggested that whole body BMC adjusted for bone area is not correlated with bone strength.(17) Furthermore, the final adjusted model may have masked clinically significant deficits in bone strength, as detailed below. In contrast, our multistaged approach revealed the contributions of each confounding variable to structurally significant bone deficits.
It is critical to note that the absence of a bone deficit after adjustment for lean mass does not imply that the bones are normal or adequate. As muscles increase in size and strength during growth, bones adapt by increasing mass, dimensions, and strength.(29) Growth, in the absence of normal loading, results in bones that are adapted to their diminished functional requirement, with decreased mass, size, and strength. These bones may be inadequate to withstand even minor trauma. This is evident is disorders such as cerebral palsy that are characterized by decreased muscle mass, decreased bone dimensions, and propensity to fracture with minimal trauma.(30) Moreover, a recent study showed significantly increased risk of hip fracture (OR, 1.86; 95% CI, 1.08, 3.21) in adults with CD.(6) Disease severity, assessed by the number of symptoms, predicted fracture even after adjusting for corticosteroid use.
While the bone and muscle deficits are highly correlated, this does not prove a causal relationship. This close association may be mediated by nutritional, hormonal, or inflammatory factors that influence muscle and bone. For example, the cytokines TNF-α and interleukin (IL)-1 were independently associated with cachexia in adults with inflammatory disorders.(31,32) Inflammatory cytokines are also known to promote bone resorption. Recent developments in the understanding of bone loss in inflammatory conditions have highlighted the critical role of TNF-α, IL-1, and IL-6, which stimulate RANKL expression by osteoblasts.(33)
Alternatively, physical activity may be decreased in those with CD, resulting in decreased muscle forces, with a consequent reduction in bone mass. Some observational studies in healthy athletes and randomized trials with modest interventions in healthy children (e.g., jumping activities three times per week) have shown significant beneficial effects on bone mass and size.(34–36) In addition, a recent weight-bearing intervention in non-ambulatory children with cerebral palsy resulted in increased bone mass.(37) We did not measure physical activity in this study, which precludes an assessment of the relationship between lean mass in CD and physical activity levels. However, the independent relationship between lean mass and BMC was highly significant and was comparable in subjects with CD and in healthy controls. That is, the magnitude of the bone deficit associated with a given reduction in muscle mass in CD was the same as in healthy controls.
Studies have yielded differing conclusions concerning the role of corticosteroids in the pathogenesis of osteopenia in children with CD. Most,(3,7,10) but not all,(28) have shown a relationship between cumulative corticosteroid dose and bone outcomes. Cowan et al.(28) showed a difference in BMC between corticosteroid-treated and corticosteroid-naïve subjects with pediatric CD. Reports of correlations between bone deficits and corticosteroid exposure in prior studies may be explained by two types of confounding: confounding by indication and confounding by short stature. First, subjects with severe inflammation are more likely to receive corticosteroids and were more likely to suffer bone deficits; therefore, corticosteroids may be a marker of disease severity. Second, corticosteroids are associated with decreased stature, and DXA measures relative to age are systematically underestimated in shorter subjects.(38) DXA outcomes expressed relative to age are subject to confounding by skeletal size and are poorly suited to assess the impact of corticosteroids on bone mass.
Our data did not show differences in BMC-for-height z scores between corticosteroid-treated and corticosteroid-naïve individuals with CD. However, there were few individuals with no steroid exposure (10%). BMC-for-height z scores were not associated with any of the corticosteroid measures. In contrast, these data showed markedly decreased stature for age in the subjects with CD, consistent with the disease process and with corticosteroid effects. While retrospective collection of lifetime corticosteroid exposure may result in misclassification of exposure status, the significant correlation between height deficits and glucocorticoids supported the validity of these estimates.
This study aimed to address cortical bone deficits independently associated with CD. Prior studies of lumbar spine aBMD, a trabecular site, in pediatric CD have noted significant deficits in age-adjusted z scores, which were, in part, attributable to delayed maturation.(2,7,10) However, these analyses did not fully account for the confounding effect of decreased stature on aBMD results. Therefore, reports of an association between decreased lumbar spine aBMD and glucocorticoids may be confounded by glucocorticoid-induced short stature.
The clinical significance of poor bone health in CD lies in the occurrence of fractures during childhood and beyond.(5,6) The data presented here highlight the relationship between bone and muscle deficits in CD. Prospective studies are needed to evaluate the relationship between decreased muscle mass, muscle strength, physical activity level, and bone mass in CD. Weight-bearing physical activity may represent a powerful and safe intervention to improve bone health and lean mass in CD and should be explored in future clinical trials.
We thank David Piccoli, Robert Baldassano, and the Center for Pediatric Inflammatory Bowel Disease at the Children's Hospital of Philadelphia for support. We greatly appreciate the dedication and enthusiasm of the children and their families who participated in this study. This work was supported by the American College of Rheumatology Research and Education Foundation (JMB), the General Clinical Research Center (M01RR00240), and the Nutrition and Growth Center at the Children's Hospital of Philadelphia.