INTRODUCTION

Celiac disease (CD) is an immune-mediated enteropathy that occurs in genetically susceptible individuals after exposure to gliadin peptides present in wheat, rye and barley. Metabolic bone disease manifested as bone pain, pathologic fractures, osteomalacia, osteopenia and osteoporosis are among the most common complications of untreated CD ⁽¹⁾. Using the technique of dual-energy x-ray absorptiometry (DXA), several studies show that low bone mineral density (BMD) is almost universal in newly diagnosed adult and pediatric CD patients ^(2-15). The effect of gluten-free diet (GFD) on BMD in adults with CD is controversial and appears to be related to age at diagnosis and to age at institution of GFD ^(2-7). In some, but not all studies of children and adolescents with CD, implementation of GFD has been shown to promote a rapid increase of BMD and lead to a complete recovery of bone mineralization by 1 year ^(8-16). An increased risk of fractures is the main consequence of osteoporosis. Although previous studies found a sevenfold increase in risk of fracture in patients with CD, recent population-based cohort studies have found only a slight increase in fracture risk in these patients ^(17-21).

Quantitative ultrasound (QUS) is a relatively new modality for assessment of bone. The United States Food and Drug Administration has recently approved the use of some QUS devices for the routine diagnosis of bone mineral status and determination of fracture risk ⁽²²⁾. It has the advantage of being radiation-free, noninvasive, mobile and friendly to both user and patient, making it ideal for use in children. Studies using QUS have found varying but generally good correlation with DXA, and QUS technique has been shown to be a good predictor of fracture risk in postmenopausal women, independent of DXA ⁽²³⁾. Using QUS we evaluated bone status in several pediatric disorders and showed that the usefulness of the QUS technique and the correlation with DXA varies among different disorders ^(24-26). Only one study has used QUS technique to examine bone status in celiac patients ⁽²⁷⁾.

In the current study we examined the application of the new technique of bone QUS in a group of children diagnosed with CD and compared QUS with the conventional DXA technique.

PATIENTS AND METHODS

This cross-sectional study included 41 children and adolescents (13 girls, 28 boys) diagnosed with CD, aged 11.2 ± 3.6 years (median, 11 years; range, 5-18 years). All patients were evaluated at the Pediatric Gastroenterology and Nutrition Unit, Meyer Children's Hospital of Haifa, Rambam Medical Center, Israel. The study was approved by the Ethics Committee of Rambam Medical Center, and informed consent was obtained from children' parents and guardians before the performance of DXA and QUS examinations.

The diagnosis of CD was made in all patients according to the criteria established by the European Society of Pediatric Gastroenterology, Hepatology and Nutrition ⁽²⁸⁾. Briefly, the diagnosis was based on the combination of suggestive clinical features, positive celiac-related serology (anti-endomysial antibodies [EMA]), small bowel biopsy showing characteristic mucosal lesion and clinical improvement after the initiation of GFD. None of the patients showed clinical or biochemical evidence of alterations in the liver, kidney or thyroid function or of other conditions known to affect bone mass and metabolism. No patient received calcium or vitamin D supplementation.

Dietary assessment was performed, based on a 3-day diet record of all food consumed. The total daily energy, protein and calcium intake was derived from diet analysis using DIET 4 statistical software and reference data from the Israeli Food Composition Tables ⁽²⁹⁾. Daily energy, protein, calcium and vitamin D intakes were expressed as percentage of recommended dietary allowance (RDA) for age and sex ⁽³⁰⁾. Compliance with the GFD was ascertained by clinical evaluation; that included a thorough inquiry of the parents and the children's own view of GFD observance and measurement of serum EMA and anti-tissue transglutaminase antibodies (tTGA) during the follow-up period.

All children were measured for height and weight and had their body mass index (BMI) calculated. Z-scores for height, weight and BMI were calculated using a database and statistics computer program (Epi Info 2000, Version 1.1.2; The Division of Surveillance and Epidemiology Program Office, Centers for Disease Control, Atlanta, GA). Pubertal staging was performed using the criteria of Tanner and Whitehouse ⁽³¹⁾.

The following biochemical studies were performed in all patients: serum albumin, calcium, phosphate, total alkaline phosphatase, immunoglobulin A, EMA and tTGA. Vitamin D metabolites (25 (OH) cholecalciferol and 1,25(OH)₂ cholecalciferol) and intact parathyroid hormone (iPTH) were determined in 11 of 41 patients. Bone assessment was performed by the two methods on the same day:

Bone Mineral Density

BMD of the lumbar spine was measured using the dual-energy x-ray absorptiometry (DXA) method with a commercial DPX instrument (Lunar, Madison, WI). The result of the second, third and fourth lumbar vertebrae bone mineral content (g) divided by the projected area of these vertebrae (cm²) expressed the bone mineral density (BMD) (g/cm²). The results were compared to age and sex matched controls and expressed as z-scores. The control group was selected from the database at our Nuclear Medicine Institute that had BMD measurements of more than 300 healthy Israeli children who represented the reference population for the examinations performed. Areal BMD z-scores were furthermore height-adjusted as described by Levine et al ⁽²⁴⁾. The in vitro precision and long-term stability of the instrument used were evaluated during the study by measuring a spine phantom before each session, and a mean coefficient of variation of 0.45% was obtained. The in vivo precision was assessed by repeated measurements in five healthy volunteers of the lumbar spine, each time with repositioning. The in vivo mean coefficients of variation were 1.1%.

Quantitative Bone Ultrasound

Measurements of the velocity of ultrasound wave, expressed as speed of sound (SOS) in m/s, were performed using the Omnisesnse 7000P, ultrasound bone sonometer device (Sunlight Ltd., Tel Aviv, Israel) at two skeletal sites. The distal radius examination site corresponded to the point halfway between the edge of the olecranon and the tip of the distal phalanx of the outstretched third digit of the left hand. Mid-tibia SOS measurements were performed with the patient seated and the knee flexed 90°. The site of examination was the point halfway between the edge of the heel and the proximal edge of the knee. A specialized pediatric transducer was placed on the marked site of measurement and rotated without lifting the transducer from the skin. SOS measurements were repeatedly performed. When the SOS score was reproducible three times in a row at the premarked location, that measurement was used. Z-scores (difference between the patient's value and the age-specific mean value divided by the reference group's standard deviation) were calculated for each site in all patients. The reference group was represented by more than 1000 healthy Israeli children and adults (database gathered by Sunlight Company) ⁽³²⁾. The non-dominant side was uniformly used for examinations; this was usually the left side unless a history of fracture was present. The device was calibrated before each examination against a control block supplied by the manufacturer. The same operator performed all the examinations. In this study the maximal accepted intra-operator variability was less than 0.6% at either site.

Statistical Analysis

Statistical analyses were performed using the SPSS statistical program, version 10.0 (SPSS, Chicago, IL). The results for continuous variables are given as mean ± SD and range. The results for noncontinuous variables are given as a frequency and a percentage. Multiple comparisons for differences among compliant and noncompliant subjects were tested by one-way analysis of variance or by the nonparametric Kruskal-Wallis method. The correlations of BMD/SOS with demographic, anthropometric, nutritional and serologic data were examined by linear regression analysis. A multiple linear regression analysis was used to assess the influence of nutritional and non-nutritional factors on BMD and SOS after adjusting for confounding variables. P value <0.05 was considered statistically significant.

RESULTS

The mean age at diagnosis of CD was 5.8 ± 4 years (median, 5 years; range, 1-14.5 years). At study enrollment, all children have been diagnosed with CD for at least 1 year, with a mean duration of 5.7 ± 4.3 years (median, 4 years; range, 1-17 years) on GFD (whether compliant or non-compliant with GFD). The presenting manifestations at the diagnosis of CD were anemia (16 of 41), failure to thrive (10 of 41), diarrhea (9 of 41), short stature (9 of 41), abdominal pain (5 of 41) and abdominal distention (5 of 41) as single symptoms or combinations. The presence of fractures was reported in two children (5%) (one wrist fracture and one fracture of tibia); both occurred after minor trauma before the diagnosis of CD was made.

The daily caloric intakes of children expressed as %RDA were relatively low, mean 85% ± 22% (median, 88%; range, 38%-133%). Calcium and vitamin D mean daily intakes were 44% ± 20% of RDA and 22% ± 12% of RDA, respectively. Evaluation of adherence to GFD showed a satisfactory and complete compliance in 19 of 41 patients (46%). The remaining 22 patients (54%) admitted to occasional lapses from GFD (15 of 22 infrequently and 7 of 22 repeatedly), although all patients denied the presence of any symptoms. Based on the results of these analyses, patients were divided into two groups: patients fully compliant with the GFD (n = 19) and patients on a gluten-containing diet (n = 22).

Table 1 presents the demographic, clinical, anthropometric, nutritional and blood biochemistry data of the study participants segregated according to their compliance to GFD. No differences were present between the two groups with regard to their demographic (age, sex, pubertal status) and clinical characteristics (age at diagnosis and duration of GFD and presenting symptoms). There were no differences in their dietary intakes and serum albumin, calcium, phosphate and total alkaline phosphatase. Vitamin D metabolites and intact parathyroid hormone examined in a small number of patients were found to be within the normal limits. However, as expected, children non-compliant to GFD were shorter than compliant children (height z-score −0.9 ± 1.1 versus −0.3 ± 1.4, t = 2, P = 0.05). In addition, there was a significant correlation between adherence to GFD and EMA status (χ² = 18.2, df = 2, P < 0.001).

Spinal DXA examinations showed that 18 of 41 children (44%) had BMD z-scores less than −1 SD; among them, 4 of 41 (10%) had BMD z-scores less than −2 SD and 14 of 41 (34%) had z-scores between −1 SD and −2 SD. Bone QUS examinations showed that bone SOS was differentially affected at the two examination sites. SOS z-scores at radius were less than −1 SD in 8 of 41 children (19%); in 1 of 41 (2%) SOS z-score was less than −2 SD and 7 of 41 (17%) had SOS z-scores between −1 SD and −2 SD. In contrast, SOS z-scores at tibia were less than −1 SD in 27 of 41 children (66%), less than −2 SD in 20 of 41 children (49%) and between −1 SD and −2 SD in 7 of 41 children (17%). The difference in the prevalence of poor bone status (BMD < −2 SD and SOS < −2 SD) as evaluated by spinal DXA and tibia QUS was statistically significant (χ² = 13.3, df = 1, P = 0.0002).

Figures 1 and 2 present the results of bone examinations by the two techniques in compliant (n = 19) and noncompliant (n = 22) children. There was a statistically significant difference in the prevalence of tibia SOS z-scores lower than −2 SD between non-compliant and compliant children (15 of 22 versus 5 of 19, χ² = 6.2, df = 1 P = 0.01) (Fig. 1). Moreover, non-compliant children had statistically significant lower tibia SOS z-scores compared with compliant children (−2.3 ± 1.8 versus −1.1 ± 1.5, t =2, P = 0.04) (Fig. 2). There were no statistically significant differences in lumbar spine DXA and radius SOS z-scores (either prevalence or mean) between compliant and non-compliant children.

Using linear regression analysis BMD z-scores were found to be significantly correlated to height (r = 0.485, P = 0.001), whereas no correlation was found between tibia or radius SOS and anthropometric data. Additionally, there was a significant correlation between low BMD z-scores (less than −1 SD) and positive EMA tests. SOS radius z-scores were negatively correlated with age at diagnosis (r = −0.35, P = 0.03) and positively with duration of GFD (r = 0.38, P = 0.01). SOS tibia z-scores were significantly correlated with female sex (r = 32, P = 0.034) and %RDA of protein intake (r = 0.33, P = 0.03). There was no significant correlation between DXA and SOS at any site and a trend toward correlation between SOS z scores at tibia and radius (r = 0.3, P = 0.06). Using stepwise regression analysis two variables were identified as independently correlated with SOS z-scores: radius SOS z scores correlated significantly with duration of GFD (Fig. 3) in compliant children, and tibia SOS z scores correlated significantly with %RDA of protein intake in female only (Fig. 4).

DISCUSSION

This study investigated the application of QUS technique in the evaluation of bone status in a group of children with CD and documented that the majority of children with CD have poor bone status when examined by either DXA or QUS. Furthermore, children with CD and poor compliance with GFD were shorter, had a significantly higher prevalence of low tibia SOS z-scores and significantly lower tibia SOS z-scores compared with children strictly adherent to GFD. In contrast, BMD as assessed by DXA technique was correlated with patient weight and height and showed no difference regardless of patients' adherence with GFD, as previously described ^(5,6).

All children with CD examined in this study had at least 1 year on GFD, as this time period has been alleged to be sufficient for bone recovery in children ^(8-11). However, a significantly large number of children were found to have a poor bone status when judged by the results of tibia QUS examination, and this finding was significantly more prominent in children non-compliant to GFD. In vitro studies showed that ultrasound velocity depends on the material properties of the medium through which the ultrasound wave travels, and there is general agreement that ultrasound velocity is influenced by bone density, architecture and elasticity, whereas BMD gives information limited to bone mineral content ⁽³³⁾.

Although decreased bone mass is clearly documented in CD patients, the underlying pathologic mechanisms are still a matter of debate ⁽¹⁾. In patients with CD the presence of malabsorption and small bowel inflammation are associated with other recognized risk factors for osteoporosis such as low BMI, low dietary calcium intake, early menopause, old age at diagnosis, smoking and insufficient physical activity ^(34,35). Each of these factors affects differently axial or appendicular skeleton, trabecular or cortical bone, bone formation or resorption during different stages of growth and pubertal development. Bone densitometry studies showed, for example, that in patients with hyperparathyroidism appendicular skeleton is more severely affected than axial skeleton ⁽³⁶⁾. In this study, intact parathyroid hormone levels were within normal limits in the children in whom they were determined. In addition, the two techniques evaluate different bone characteristics: mineral quantity by DXA and mineralization together with other bone structural properties by QUS. Therefore, the lack of correlation that we found between the two techniques is not surprising, nor is the slight correlation between QUS measurements at the two examination sites. These data suggest that peripheral bone QUS does not identify the same patients with reduced BMD status assessed by other techniques, probably because different methods do not measure identical properties of bone tissue. We suggest that examination of skeleton using QUS mode might be able to recognize additional bone abnormalities and possibly, the reason for the very poor bone status found at the tibia site.

Strict compliance to GFD is a prerequisite for bone healing and this study showed, as previous studies, that years after diagnosis of CD only half of children still maintain strict adherence to GFD ⁽³⁷⁾. In our study, the compliance to GFD was attested by the significant difference in serum EMA status between compliant and non-compliant children. Children non-compliant to GFD were shorter than the strictly compliant children. The adherence to GFD had a strong influence on QUS examination at the tibia site, whereas radius SOS and spine BMD were not able to discriminate between compliant and non-compliant children. These findings are in contrast to the results reported by Pedrera et al. in the single study that examined quantitative phalangeal ultrasound in children with CD after long-term GFD ⁽²⁷⁾.

Bone quality examined by radius SOS showed a significant correlation with duration of GFD. After elimination of confounding variables, radius SOS showed a continuous, time-related improvement in bone quality in patients strictly adherent to GFD. Studies in children have indicated that 1 year of GFD is usually sufficient to normalize BMD; however, our study showed that continuous adherence to GFD brought further improvement of radius SOS.

Low BMD is frequent at the diagnosis of CD; however, recent studies showed fracture risk to be only minimally increased in celiac patients, and BMD did not correlate with fractures in the Vazquez et al. study ^(17-21). These new findings together with the evident improvement in BMD after the institution of GFD have recently questioned the necessity of screening or follow-up for osteoporosis of all celiac patients as currently recommended ⁽³⁸⁾. Most studies, however, had been performed in adults and there is no data regarding the long-term follow-up of children or the ability of children with CD to achieve appropriate peak bone mass and maintain normal bone mass through adult life. Furthermore, normalization of BMD in children with CD is conditioned by a strict adherence to GFD, which is frequently difficult for children. Periodic evaluation of bone status in children at least until growth conclusion seems advisable and QUS technique might be the right tool.

Our study suffered from the lack of cut-off or threshold criteria for the definition of osteoporosis and osteopenia in children for DXA and QUS. Although we found several factors significantly correlated with bone status as assessed by QUS, cross-sectional study associations do not necessarily imply a causative relationship. Furthermore, we had no complementary parameters to support our hypothesis that QUS identified additional bone abnormalities not detected by DXA. The confirmation and the significance of the high prevalence of abnormal QUS results in patients with CD, therefore, need to be further demonstrated.

In summary, in this study we showed that QUS identified severe bone abnormalities in children with CD. Tibia SOS was significantly lower and more frequently abnormal in children non-adherent to GFD compared with compliant children. Furthermore, radius SOS showed a positive time-related correlation with GFD adherence. Thus, we suggest that in this small cohort of children with CD, QUS provided more information relevant to the disease status than DXA. However, the relevance of this information is unclear at the moment and further research is needed to establish the application of peripheral QUS in the study of bone disease and evaluation of fracture risk in patients with CD.

Acknowledgment: We would like to acknowledge our clinical dietician, Mrs. Shlomit Itzkowitz, who performed the diet records analyses.