Bone Geometry and Microarchitecture Deficits in Children with Alagille Syndrome

Abstract

Alagille syndrome (ALGS) is an autosomal dominant disorder attributed to mutations in the Notch signaling pathway. Children with ALGS are at increased risk for fragility fracture of unknown etiology. Our objective was to characterize bone mass, geometry, and microarchitecture in children with ALGS. This was a cross-sectional study of 10 children (9 females), ages 8-18 years, with a clinical diagnosis of ALGS. Bone density was assessed via DXA (Hologic Discovery A) at several skeletal regions. Tibia trabecular and cortical bone was assessed via pQCT (Stratec XCT 2000) at the distal 3% and 38% sites, respectively. Tibia bone microarchitecture was assessed via HR-pQCT (Scanco XtremeCT II) at an ultradistal site located at 4% of tibia length and a cortical site at 30% of tibia length. Z-scores were calculated for DXA and pQCT measures. In the absence of XtremeCT II HR-pQCT reference data, these outcome measures were descriptively compared to a sample of healthy children ages 5-20 years (n=247). Anthropometrics and labs were also collected. Based on one-sample t-tests, mean Z-scores for height and weight (both p<0.05), were significantly less than zero. DXA bone Z-scores were not significantly different from zero, but were highly variable. For pQCT bone measures, Z-scores for total bone mineral content at the distal 3% site and cortical bone mineral content, cortical area, and cortical thickness at the distal 38% site were significantly less than zero (all p<0.05). There was good correspondence between pQCT measures of cortical thickness Z-scores and DXA Z-scores for aBMD at the whole body less head, 1/3 radius, and femoral neck (all p<0.05). Compared to healthy children, those with ALGS generally had lower trabecular number and greater trabecular separation despite having greater trabecular thickness (measured via HR-pQCT). Bilirubin and bile acids, markers of hepatic cholestasis, were associated with poorer bone measures. For example, greater bilirubin was associated with lower trabecular number (Spearman’s rho [ρ]=−0.82, p=0.023) and greater trabecular separation (ρ=0.82, p=0.023) measured via HR-pQCT, and greater bile acids were associated with lower cortical area measured via pQCT (ρ=−0.78, p=0.041) and lower serum insulin-like growth factor-1 (ρ=−0.86, p=0.002). In summary, deficits in cortical bone size and trabecular bone microarchitecture were evident in children with ALGS. Further investigation is needed to understand the factors contributing to these skeletal inadequacies, and the manner in which these deficits contribute to increased fracture risk.

1. INTRODUCTION

Alagille syndrome (ALGS) is an autosomal dominant disorder attributed to mutations in the Notch signaling pathway, occurring in approximately 1 in 70,000 live births.^1–3 Most ALGS cases are attributed to mutations in the JAG1 gene, with only a few involving mutations in NOTCH2. Multiple organ systems are affected in ALGS, but the phenotypic manifestations can vary greatly across individuals. Cholestasis resulting from bile duct paucity, congenital heart defects, dysmorphic facial features, ocular defects, and abnormal vertebrae (e.g., butterfly vertebrae) are seminal characteristics of ALGS.^3,4 The “classical criteria” for diagnosis of ALGS requires the presence of three of these five clinical features in addition to bile duct paucity.³ Although not considered classical manifestations of ALGS, shorter stature,⁵ lower bone mass,^6,7 and increased risk for lower-limb pathologic fractures⁸ also have been reported.

The fracture pattern among children with ALGS is unusual because of the greater incidence of fragility fractures in the lower limbs⁸. The mechanism responsible for the compromised bone health in ALGS is likely multifaceted, potentially involving aberrant Notch-mediated bone metabolism,⁹ osteoblast dysfunction attributed to elevated bilirubin,¹⁰ impaired nutrient absorption as a consequence of cholestasis and steatorrhea,¹¹ and low dietary energy and calcium consumption.¹¹ In two separate studies, lower areal bone mineral density (aBMD) and bone mineral content (BMC) of the total body and lumbar spine measured by dual energy x-ray absorptiometry (DXA) were reported in children with ALGS.^6,7 Since DXA measurements of bone density are confounded by height, and ALGS is associated with shorter stature, it is important to adjust for height to determine the extent to which lower bone mass or density is attributed to shorter stature.^12,13 Olsen et al. reported that controlling for stature in children with ALGS nullified bone mass deficits at the lumbar spine, but not at the total body.⁷ The lumbar spine and total body are comprised of predominantly trabecular and cortical bone, respectively, and mouse studies indicate that perturbed Notch signaling differentially impacts these bone compartments.¹⁴ To date, clinical studies involving bone health in children with ALGS have exclusively utilized DXA, which is unable to separately assess cortical and trabecular bone compartments.

The objective of this study was to characterize bone health in children with ALGS using measures of cortical and trabecular bone geometry, volumetric density, and microarchitecture to gain added insight regarding the discrete skeletal features that might contribute to the increased risk for fragility fractures in this population. To compare our findings against prior studies and to draw clinically meaningful conclusions, aBMD and BMC were also assessed via DXA at several skeletal sites. Furthermore, we examined the relationships between markers of hepatic cholestasis, bilirubin and bile acids, and bone health outcomes.

2. METHODS

2.1. Study Participants

Children with ALGS receiving care at the Children’s Hospital of Philadelphia were invited to participate in this study. Patients were identified through personal physician contact and referral. Males and females, ages 8-21 years, with a clinical diagnosis of ALGS were eligible to participate. Diagnosis of ALGS was based on the presence of a) bile duct paucity and b) 3 of the 5 typically occurring clinical features including cholestasis, ophthalmologic abnormalities, characteristic facial features, congenital heart disease, and butterfly vertebrae.³ Individuals who did not meet these clinical criteria but had JAG1 or NOTCH2 mutations, which are known to be associated with ALGS, were also included in this study. Clinical features and genotype for included participants are provided in Supplemental Table 1. Potential participants were excluded based on the following criteria: history of liver or kidney transplantation, unrepaired congenital heart disease, use of medications known to affect growth, or currently pregnant or lactating (females). For those under the age of 18, written consent was obtained from the participant’s legal guardian and written assent was obtained from the participant. For the patients 18 years of age, written consent was obtained. All study protocols and procedures were approved by the Institutional Review Board at the Children’s Hospital of Philadelphia.

2.2. Anthropometrics and Maturation

Height was measured using a wall-mounted stadiometer (Holtain, Crymych, UK), weight was measured using an electronic scale (Scaletronix, White Plains, NY), and body mass index was calculated. Z-scores for height, weight, and BMI were calculated using the CDC 2000 growth charts.¹⁵ Breast and genital development for females and males, respectively, was assessed by using a validated self-assessment questionnaire containing pictograms and written descriptions,¹⁶ and categorized into maturity stages according to Tanner.^17,18

2.3. Fracture History

A fracture history questionnaire assessed whether a fracture had occurred at any point throughout the participant’s life. Participants/parental guardians were asked to provide details regarding the skeletal site and circumstance of each fracture.

2.4. Biochemistries

Non-fasting blood draws were performed, and samples were immediately processed and stored at the Children’s Hospital of Philadelphia Clinical Laboratory. Analyses were performed using standard laboratory techniques for several measures including bilirubin (Ortho Clinical Diagnostics, Raritan, NJ) and bile acids, intact parathyroid hormone (PTH; Ortho Clinical Diagnostics, Raritan, NJ), and 25(OH)D. Bile acids were analyzed at Cincinnati Children’s Hospital Medical Center. Vitamin D was analyzed using Liquid Chromotography-Tandem Mass Spectoscopy. Insulin-like growth factor-1 (IGF-1; Siemens, Terrytown, NY) was measured in the CHOP Center for Human Phenomic Sciences (CHPS) Core Laboratory using a commercially available ELISA kit. C-terminal telopeptide (CTX; Roche, Indianapolis, IN) and bone-specific alkaline phosphatase (BSAP; Quidel Corp., San Diego, CA) were measured in the CHPS Core Laboratory using an immunoassay kit. To perform descriptive comparisons, assay-specific pediatric reference data for IGF-1 and unpublished data for BSAP, CTX, PTH, and 25(OH)D from the Bone Mineral Density in Childhood Study were used.

2.5. Dual-Energy X-Ray Absorptiometry

Dual-energy X-ray absorptiometry scans were obtained at the total body, lumbar spine, left hip, and non-dominant forearm using a Discovery A densitometer (Hologic, Bedford, MS). Scans were analyzed using software version 13.4. DXA outcomes were converted to age, sex, and ancestry-specific Z-scores using the Bone Mineral Density in Childhood Study reference ranges.¹⁹ Bone Z-scores were further adjusted for height-for-age Z-scores.¹²

2.6. Peripheral Quantitative Computed Tomography

We have reported pQCT scan acquisition and analysis protocols in greater detail previously.^20,21 Tibia length was measured from the medial malleolus to the medial tibial plateau using a sliding caliper (Rosscroft, Surrey, BC, Canada). Scans were acquired on the left tibia at relative distances of 3, 38, and 66% sites from the distal region using the Stratec XCT 2000 scanner (Orthometrics, Inc.; White Plains, NY) with a voxel size of 0.4 mm, slice thickness of 2.3 mm, and scan speed of 25 mm/second (software version 5.5). A scout view scan was acquired and the reference line was placed at the medial proximal border of the distal growth plate for those with an open growth plate. For those with a fused growth plate, the reference line was placed at the medial proximal border of the endplate. Trabecular volumetric BMD (mg/cm³) and total BMC (mg) were assessed at the 3% site. Cortical volumetric BMD (mg/mm³), cortical BMC (mg), cortical area (mm²), cortical thickness (mm), periosteal circumference (mm), endosteal circumference (mm), and section modulus (mm³) were assessed at the 38% site. Muscle cross-sectional area and subcutaneous fat cross-sectional area were assessed at the 66% site.

To account for the age, sex, and ancestry-specific differences in bone volumetric density and structure,²² pQCT outcomes were converted to age-, sex-, and ancestry-specific Z-scores based on our reference database of >650 healthy children and adolescents ages 10-18 years.²² Z-scores for cortical geometry measures were further adjusted for tibia length-for-age Z-scores using a method similar to that proposed by Zemel et al.¹² Trabecular volumetric BMD (3% site), total BMC (3% site), and cortical volumetric BMD (38% site) were not adjusted for leg length due to the null association between leg length and these pQCT bone outcomes.²¹

2.7. High Resolution Peripheral Quantitative Computed Tomography

HR-pQCT scans of the tibia were obtained using the Scanco XtremeCT II device (SCANCO Medical AG, Bruettisellen, Switzerland). This 2^nd generation HR-pQCT device has an isotropic voxel size of 61μm and direct voxel-based measures of skeletal features, providing discrete measures of trabecular and cortical bone structure. The quality of each scan based on movement artifacts was rated on a scale of 1 to 5, with ‘1’ indicating excellent scan quality. Scans that received a score of 4 or 5 were not included in our analyses.

A scout view was performed to place the reference line at the proximal edge of the distal growth plate. Measurements were acquired on the non-dominant limb. Two “blocks” of 168 slices per block were obtained at the metaphysis to capture a region of interest that included a fixed distance of 8 mm proximal to the reference line and a relative distance of 4% of tibia length. Measures of total volumetric BMD (mg hydroxyapatite [HA]/cm³). trabecular area (mm²), trabecular number (1/mm), trabecular thickness (mm), trabecular separation (mm), and bone volume to total volume fraction (1) were acquired at the 4% site. The diaphyseal scan of 168 slices was acquired at the distal 30% site of the tibia. Cortical and trabecular compartments were segmented using a fully automatic contouring procedure. Measures of cortical volumetric BMC (mg HA/mm³), total area (mm²), cortical thickness (mm), cortical porosity (1), and cortical pore diameter (mm) were acquired at this region.

Pediatric reference data for the for the Scanco XtremeCT II device are not yet available. For this reason, HR-pQCT values of trabecular and cortical bone microarchitecture and geometry were descriptively compared to values for healthy non-African American children ages 5-20 years (n=247) acquired at the Stanford Assessment of Bone and Muscle Across the Ages (SAMBA) Center at the Stanford University School of Medicine following identical scan procedures. Scans of ALGS study participants were analyzed at Stanford using the same analysis algorithms. These data were plotted against age to facilitate descriptive comparisons.

2.8. Statistical Analyses

Descriptive characteristics are presented as mean ± standard deviation (SD). Anthropometric, DXA, and pQCT measures were expressed as Z-scores and compared to the expected population value, which is a mean of 0 and SD of 1, using one-sample t-tests. Spearman’s correlations were performed to examine the relationships between pQCT and DXA bone Z-scores (adjusted for leg length and height Z-scores, respective), as well as relationships between bilirubin and bile acids with DXA and pQCT derived measures of bone and body composition, and blood biochemistries. All statistical analyses were performed using STATA (version 15.1). RStudio (version 1.1.463)²³ and the ‘ggplot2’ package²⁴ were used for the generation of select figures. P-values <0.05 were considered statistically significant.

3. RESULTS

3.1. Participant characteristics

Descriptive characteristics of the ten study participants enrolled in our study are presented in Table 1. Three patients reported a previous fracture. One patient reported three fractures (foot, clavicle, and finger), one patient reported two fractures (carpal/finger and heel), and one patient reported one fracture (toe). Five of the six reported fractures were considered ‘low impact.’

Table 1.

Patient characteristics

	Mean	SD	p*	Z-score < −2.0 (%)
Age (years)	12.6	3.69
Tanner Stage (n, stage 1/2/3/4/5)	2/2/2/3/1
Sex (n, female/males)	9/1
Fractures (low impact/total)	5/6
Anthropometries
Height (Z-scores)	−0.84	1.17	0.050	20%
Weight (Z-scores)	−0.92	1.19	0.037	20%
BMI (Z-scores)	−0.45	0.72	0.081	0%
Tibia length (Z-scores)	−0.44	1.00	0.193	0%
DXA Z-scores
Age-adjusted
Total body (less head) BMC	−0.29	4.88	0.854	50%
Total body (less head) aBMD	−0.08	4.43	0.957	40%
Total body BMC	−0.34	4.58	0.818	50%
Total body aBMD	−0.05	4.46	0.975	30%
Lumbar spine BMC	−0.19	4.57	0.897	50%
Lumbar spine aBMD	0.19	3.86	0.878	30%
Total hip BMC	−1.44	4.86	0.375	40%
Total hip aBMD	−0.40	2.81	0.667	30%
Femoral neck BMC	−0.86	3.76	0.486	50%
Femoral neck aBMD	−0.90	2.07	0.202	30%
1/3 radius BMC	0.19	4.46	0.895	40%
1/3 radius aBMD	−1.58	4.15	0.259	30%
Age and height-adjusted
Total body (less head) BMC	0.35	1.44	0.464	0%
Total body (less head) aBMD	0.19	1.61	0.718	10%
Total body BMC	0.28	1.56	0.587	0%
Total body aBMD	0.17	2.39	0.828	10%
Lumbar spine BMC	0.40	1.50	0.424	0%
Lumbar spine aBMD	0.66	1.73	0.258	0%
Total hip BMC	−0.87	2.30	0.260	20%
Total hip aBMD	0.11	1.55	0.829	10%
Femoral neck BMC	−0.39	1.27	0.360	0%
Femoral neck aBMD	−0.64	1.13	0.109	10%
1/3 radius BMC	0.64	2.38	0.418	10%
1/3 radius aBMD	−1.45	2.62	0.115	30%
pQCT Z-scores
Age-adjusted
Metaphysis (3% site)
Trabecular vBMD	−0.22	1.40	0.625	20%
Total BMC	−1.81	1.55	0.005	60%
Diaphysis (38% site)
Cortical vBMD	0.55	1.57	0.300	0%
Cortical BMC	−1.51	1.92	0.034	30%
Cortical Area	−1.71	1.99	0.024	60%
Cortical Thickness	−2.16	1.69	0.003	60%
Periosteal Circumference	−0.58	1.81	0.336	30%
Endosteal Circumference	0.96	1.39	0.057	20%a
Section modulus	−0.95	1.94	0.158	30%
Soft tissue (66% site)
Muscle Cross-Sectional Area	−2.16	1.39	0.006	43%b
Fat Cross-Sectional Area	−0.89	1.22	0.102	29% b
Age and tibia length-adjusted
Diaphysis (38% site)
Cortical BMC	−1.26	1.46	0.024	30%
Cortical Area	−1.44	1.51	0.014	30%
Cortical Thickness	−2.00	1.41	0.002	50%
Periosteal Circumference	−0.31	1.41	0.511	10%
Endosteal Circumference	1.12	1.32	0.025	30%a
Section modulus	−0.66	1.48	0.191	20%
Soft tissue (66% site)
Muscle Cross-Sectional Area	−2.00	1.21	0.005	43%
Fat Cross-Sectional Area	−0.77	1.19	0.140	10%
Blood Biochemistries
Bilirubin (mg/dL)	3.32	3.36
Bile Acids (μmol/L)	119.98	111.14
IGF-1 (ng/mL)	190.84	151.57
CTX (ng/mL)	1.05	0.57
BSAP (U/L)	104.68	37.43
25(OH)D (ng/mL)	44.24	30.57
PTH (ng/L)	29.17	16.27

^*P-value for a one-samples t-test

^aPercent of patients with a Z-score > 2.0, a potential cause of “thinner” bones

^bPercent based on n=7 (3 scans were excluded due to excessive movement)

BMI, body mass index; DXA, dual-energy X-ray absorptiometry; BMC, bone mineral content; aBMD, areal bone mineral density; pQCT, peripheral quantitative computed tomography; vBMD, volumetric bone mineral density; IGF-1, insulin-like growth factor-1; CTX, collagen type-1 cross-linked C-telopeptide; BSAP, bone-specific alkaline phosphatase; 25(OH)D, 25-hydroxyvitamin D; PTH, parathyroid hormone

Height and weight Z-scores were negative and significantly different from zero (both p<0.05). Mean Z-scores for nearly all DXA measures were negative with large standard deviations, yet none were significantly different from zero. DXA Z-scores that were adjusted for height-for-age Z-scores were predominantly positive, but none differed significantly from zero. Nevertheless, the percentage of the sample that had height-for-age adjusted bone Z-scores less than −2 SD was 10% for total body and femoral neck aBMD, and 30% for the distal 1/3 radius aBMD.

In contrast, mean pQCT Z-scores were mostly negative, and several measures (total BMC at the 3% site; cortical BMC, cortical area, and cortical thickness at the 38% site; and muscle cross-sectional area at the 66% site) were significantly different from zero (p<0.05; Table 1 and Figure 1). Mean endosteal circumference Z-score was positive, but was not significantly different from zero (p=0.06). When adjusted for tibia length, children with ALGS had significantly lower cortical BMC, cortical area, cortical thickness, and muscle cross-sectional area, and significantly increased endosteal circumference (all p<0.05), consistent with thinner bones. Indeed, 50% of children had cortical thickness Z-scores (adjusted for tibia length) below −2 SD.

Figure 1.

Tibia pQCT Z-scores in children with ALGS. White boxes represent age-adjusted Z-scores and gray boxes represent age and tibia length-adjusted Z-scores. *Z-score significantly different from zero via one-sample t-test (P<0.05). Tb.vBMD, trabecular volumetric bone mineral density; Tot.BMC, total bone mineral content; Ct.vBMD, cortical volumetric bone mineral density; Ct.BMC, cortical bone mineral content; Ct.Ar, cortical area; Ct.Th, cortical thickness; Peri.Circ, periosteal circumference; Endo.Circ, endosteal circumference.

In general, there was good correspondence between pQCT measures of cortical thickness Z-scores and DXA Z-scores for aBMD at the whole body less head (Spearman’s rho [ρ]=0.76, p=0.011), 1/3 radius (ρ=0.81, p=0.005), and femoral neck (ρ=0.67, p=0.033), but not total hip (ρ=0.60, p=0.067) or lumbar spine (ρ=0.48, p=0.162). Trabecular vBMD Z-score was significantly associated with total hip aBMD (ρ=0.70, p=0.025) and femoral neck aBMD (ρ=0.68, p=0.029), but not aBMD at the whole body less head (ρ=0.26, p=0.467), 1/3 radius (ρ=0.32, p=0.366), or lumbar spine (ρ=0.01, p=0.987).

3.2. Visual comparison of HR-pQCT-derived bone measures between ALGS and healthy children

Several HR-pQCT scans were excluded due to excessive movement, including three scans from the 4% site (one male and two females) and 3 scans from the 30% site (three females). Therefore, seven patients had available data at the 4% (seven females) and 30% (six females, one male) sites.

Overall, children with ALGS appeared to have inferior bone density and microarchitecture at the 4% site compared to the healthy individuals (Figure 2). At least 5 of the 7 participants with usable scans had values below the median for ultradistal tibia total and trabecular volumetric BMD, trabecular area, and trabecular number, and above the median for trabecular separation and trabecular thickness. At the mid-shaft (Figure 3), at least 5 of 7 participants had below median values for total area, cortical thickness, and cortical pore diameter.

Figure 2.

Ultradistal tibia (3% site) trabecular bone microarchitecture (measured via HR-pQCT) in children with ALGS (black dots) and healthy controls (gray dots). A loess regression line was fitted to the data from the healthy controls. A) Total vBMD, total volumetric bone mineral density; B) Trabecular vBMD, trabecular vBMD; C) Trabecular Area; D) Tb.N, trabecular number; E) Tb.Th, trabecular thickness; F) Tb.Sp, trabecular separation; and G) BV/TV, bone volume to total volume fraction.

Figure 3.

Distal tibia (30% site) cortical bone density, geometry, and porosity (measured via HR-pQCT) in females (left) and males (right) with ALGS (black dots) and healthy controls (gray dots). A loess regression line was fitted to the data from the healthy controls. A) Ct.vBMD, cortical volumetric bone mineral density; B) Tt.Ar, total area; C) Ct.Ar, cortical area; D) Ct.Th, cortical thickness; E) Ct.Po, cortical pore number; and F) Ct.Po.Dm, cortical pore diameter.

3.3. Blood biochemistries

Serum IGF-1 concentration relative to age and sex is graphically displayed for the children with ALGS compared to assay-specific reference ranges (Figure 4). Although IGF-1 was somewhat variable, half of our patients had IGF-1 values below the 2^nd percentile and only three had values above the 50^th percentile. Supplemental Figure 1 displays values for BSAP, CTX, PTH, and 25(OH)D in children with ALGS compared to healthy children from the Bone Mineral Density in Childhood Study. Overall, children with ALGS consistently had higher BSAP, but CTX, PTH, and 25(OH)D were highly variable. Only one patient had a 25(OH)D less than 20 ng/mL.

Figure 4.

Serum IGF-1 (ng/mL) relative to age in females (left) and males (right) with ALGS compared to assay-specific reference values. The top, middle, and bottom gray lines represent the 98^th, 50^th, and 2^nd percentiles, respectively. Data for the ALGS patients are displayed in black, and the accompanyingnumber corresponds to the “Participant” number in Supplemental Table 1.

3.4. Relationships between bilirubin, bile acids, and bone measures

Bilirubin was negatively associated with DXA Z-scores for BMC and aBMD at all skeletal sites, as well as height-adjusted aBMD Z-scores for total body less head, total body, and distal 1/3 radius (all P<0.05; Table 2). Bile acids were significantly negatively associated with Z-scores for femoral neck BMC, as well as height-adjusted Z-scores for total body aBMD and femoral neck BMC (all P<0.05). Bilirubin was not associated with any pQCT measure, but bile acids were negatively associated with tibia length-adjusted cortical area Z-scores (all P<0.05; Table 3). With respect to HR-pQCT measures, bilirubin was associated negatively with trabecular number and positively with trabecular separation at the 4% tibia (both P<0.05; Table 4). Bilirubin and bile acids did not correlate with CTX, BSAP, 25(OH)D, or PTH, but both were negatively associated with IGF-1 (P<0.001; Table 5). Relationships between serum bilirubin and bile acids and were also performed while excluding the one male patient. These results were generally consistent with those reported herein (results not shown).

Table 2.

Spearman’s correlations between bilirubin and bile acids with DXA bone Z-scores in children with ALGS

	Bilirubin		Bile Acids
	Spearman’s Correlation	p value	Spearman’s Correlation	p value
Age-adjusted Z-scores
Total body (less head) BMC	−0.72	0.019	−0.55	0.102
Total body (less head) aBMD	−0.76	0.011	−0.60	0.066
Total body BMC	−0.72	0.019	−0.55	0.102
Total body aBMD	−0.76	0.011	−0.60	0.066
Lumbar spine BMC	−0.71	0.022	−0.58	0.077
Lumbar spine aBMD	−0.71	0.022	−0.58	0.077
Total hip BMC	−0.73	0.016	−0.60	0.069
Total hip aBMD	−0.72	0.019	−0.58	0.077
Femoral neck BMC	−0.72	0.019	−0.63	0.050
Femoral neck aBMD	−0.76	0.011	−0.60	0.069
1/3 radius BMC	−0.70	0.025	−0.60	0.066
1/3 radius aBMD	−0.72	0.019	−0.55	0.102
Age and height-adjusted Z-scores
Total body (less head) BMC	−0.56	0.090	−0.43	0.213
Total body (less head) aBMD	−0.72	0.019	−0.58	0.077
Total body BMC	−0.67	0.033	−0.57	0.089
Total body aBMD	−0.84	0.002	−0.72	0.020
Lumbar spine BMC	−0.61	0.060	−0.44	0.199
Lumbar spine aBMD	−0.53	0.117	−0.38	0.275
Total hip BMC	−0.43	0.215	−0.57	0.084
Total hip aBMD	−0.25	0.489	−0.29	0.413
Femoral neck BMC	−0.54	0.108	−0.66	0.037
Femoral neck aBMD	−0.32	0.366	−0.35	0.318
1/3 radius BMC	−0.71	0.022	−0.61	0.062
1/3 radius aBMD	−0.85	0.002	−0.58	0.077

DXA, dual-energy X-ray absorptiometry; ALGS, Alagille syndrome; BMC, bone mineral content; aBMD, areal bone mineral density

Table 3.

Spearman’s correlations between bilirubin and bile acids with pQCT Z-scores in children with ALGS

	Bilirubin		Bile Acids
	Spearman’s Correlation	p value	Spearman’s Correlation	p value
Age-adjusted Z-scores
Metaphysis (3% site)
Tb.vBMD	−0.39	0.383	0.02	0.969
Tot.BMC	−0.14	0.760	−0.36	0.427
Diaphysis (38% site)
Ct.vBMD	−0.54	0.215	−0.31	0.504
Ct.BMC	−0.39	0.383	−0.74	0.058
Ct.Ar	−0.07	0.879	−0.50	0.248
Ct.Th	−0.57	0.180	−0.67	0.102
Peri.Circ	0.14	0.760	−0.54	0.210
Endo.Circ	0.46	0.294	−0.31	0.504
Section modulus	0.14	0.760	−0.54	0.210
Soft tissue (66% site)
M.CSA	0.07	0.879	−0.18	0.699
SF.CSA	−0.68	0.094	−0.18	0.699
Age and tibia length-adjusted Z-scores
Diaphysis (38% site)
Ct.BMC	−0.39	0.383	−0.74	0.058
Ct.Ar	−0.18	0.702	−0.77	0.041
Ct.Th	−0.57	0.180	−0.67	0.102
Peri.Circ	0.14	0.760	−0.54	0.210
Endo.Circ	0.43	0.337	−0.18	0.699
Section modulus	0.14	0.760	−0.54	0.210
Soft tissue (66% site)
M.CSA	0.07	0.879	−0.18	0.699
SF.CSA	−0.68	0.094	−0.18	0.699

pQCT, peripheral quantitative computed tomography; ALGS, Alagille syndrome; Tb.vBMD, trabecular volumetric bone mineral density; Tot.BMC, total bone mineral content; Ct.vBMD, cortical volumetric bone mineral density; Ct.BMC, cortical bone mineral content; Ct.Ar, cortical area; Ct.Th, cortical thickness; Peri.Circ, periosteal circumference; Endo.Circ, endosteal circumference; M.CSA, muscle cross-sectional area; SF.CSA, subcutaneous fat cross-sectional area

Table 4.

Spearman’s correlations between bilirubin and bile acids with HR-pQCT bone outcomes in children with ALGS

	Bilirubin		Bile Acids
	Spearman’s Correlation	p value	Spearman’s Correlation	p value
Ultradistal (4% site)
Tot.vBMD	−0.75	0.052	−0.71	0.071
Tb.vBMD	−0.68	0.090	−0.56	0.193
Tb.Ar	−0.11	0.819	−0.21	0.645
Tb.N	−0.82	0.023	−0.43	0.337
Tb.Th	0.00	1.000	−0.32	0.482
Tb.Sp	0.82	0.023	0.43	0.337
BV/TV	−0.68	0.090	−0.67	0.102
Distal (30% site)
Ct.vBMD	−0.50	0.253	0.04	0.939
Tot.Ar	−0.21	0.645	−0.50	0.253
Ct.Th	−0.75	0.052	−0.75	0.052
Ct.Po	0.56	0.193	0.14	0.758
Ct.Po.Dm	−0.04	0.939	−0.04	0.939

HR-pQCT, high resolution peripheral quantitative computed tomography; ALGS, Alagille syndrome; Tot.vBMD, total volumetric bone mineral density; Tb.vBMD, trabecular volumetric bone mineral density; Tb.Ar, trabecular area; Tb.N, trabecular number; Tb.Th, trabecular thickness; Tb.Sp, trabecular separation; BV/TV, bone volume to total volume fraction; Tb.1/N.SD, inhomogeneity of trabecular network; Ct.vBMD, cortical volumetric bone mineral density; Tot.Ar, total area; Ct.Ar, cortical area; Ct.Th, cortical thickness; Ct.Po, cortical porosity; Ct.Po.Dm, cortical pore diameter

Table 5.

Spearman’s correlations between bilirubin and bile acids with serum biochemistries in children with ALGS

	Bilirubin		Bile Acids
	Spearman’s Correlation	p value	Spearman’s Correlation	p value
IGF-1	−0.76	0.011	−0.86	0.002
CTX	−0.09	0.803	−0.59	0.073
BSAP	0.30	0.405	0.07	0.841
25(OH)D	−0.58	0.082	−0.27	0.444
PTH	−0.04	0.907	−0.31	0.383

ALGS, Alagille syndrome; IGF-1, insulin-like growth factor-1; CTX, collagen type-1 cross-linked C-telopeptide; BSAP, bone-specific alkaline phosphatase; 25(OH)D, 25-hydroxyvitamin D; PTH, parathyroid hormone

4. DISCUSSION

We performed a comprehensive evaluation of bone health in children with ALGS. Our main finding was that children with ALGS had smaller and thinner tibia cortical bone diaphyses, even after adjustment for shorter limb length. Additionally, deficits in trabecular bone microarchitecture were evident in the children with ALGS. Specifically, ALGS was associated with fewer and more separated trabeculae, and greater bilirubin was inversely associated with bone density from DXA and bone microarchitecture from HR-pQCT. This small study provides novel insight into the skeletal abnormalities in ALGS that might contribute to the increased fracture risk in this population.⁸

This study identified compromised cortical and trabecular bone geometry and microarchitecture in children with ALGS using quantitative CT techniques that were not readily apparent by DXA measures, the preferred method for clinical bone health assessment in children.²⁵ DXA Z-scores were highly variable among children with ALGS and not significantly different from zero overall. In contrast, at the tibia diaphysis, cortical bone deficits were apparent, as characterized by a smaller and thinner diaphysis with lower bone mass. Size adjusted DXA Z-scores of the whole body less head, hip, and forearm were significantly associated with cortical bone deficits identified by pQCT. The high proportion (30%) of children with low Z-scores (i.e., < −2) for the distal 1/3 radius suggests that this may be a good clinical measure to detect these deficits. DXA size adjusted Z-scores at the total hip and femoral neck were associated with trabecular bone deficits by pQCT. Importantly, although the lumbar spine is a recommended scan site for pediatric bone health assessment,²⁵ size adjusted DXA spine aBMD Z-scores were not associated with pQCT outcomes for trabecular density or cortical thickness.

We did not observe deficits in periosteal circumference by pQCT, suggesting that the thinner and lighter cortex was predominantly attributed to an expanded endosteal border. In a Jag1 deletion mouse model, Youngstrom et al. demonstrated that progenitor cell Jag1 deletion resulted in a greater endosteal perimeter, but also a larger and stronger cortex due to accompanying periosteal apposition.¹⁴ Alternatively, osteoblast Jag1 deletion had a small effect on cortical bone size, but appeared to be detrimental to trabecular bone microarchitecture. We also found potentially compromised trabecular bone microarchitecture in our ALGS patients. On average, these individuals had lower trabecular bone mass and density, consistent with a trabecular bone network that was comprised of fewer, more separated trabeculae.

Bile duct paucity and cholestasis are seminal features of ALGS,⁴ and markers of cholestasis were indicative of inferior bone health. One previous study in childhood ALGS reported consistent inverse relationships between bilirubin, but not bile acids, and DXA bone measures.⁶ In the current study, we also observed that bilirubin was associated with lower bone mass at several skeletal sites, including the total body, lumbar spine, femoral neck, and forearm. Interestingly, bile acids were associated with a smaller cortical bone area but did not correlate with other bone geometry mesures, so this relationship should be interpreted conservatively. Cholestasis in ALGS is accompanied by nutritional inadequacies, such as low dietary energy and micronutrient consumption and nutrient malabsorption.⁵ Unfortunately, we did not assess typical dietary patterns or nutrient absorption, namely that of fat soluble vitamins (vitamins A, D, E, and K) or calcium. The majority of our patients (90%), however, reported typical vitamin D supplementation ranging from 1,000 to 50,000 IU cholecalciferol per day and only one had 25(OH)D < 20 ng/mL.

Insulin-like growth factor-1 is an important hormone in childhood bone accrual, notably with respect to periosteal bone mineral apposition.²⁶ Systemic IGF-1 is regulated by growth hormone, and growth hormone insensitivity can accompany ALGS.²⁷ Furthermore, IGF-1 concentrations can be lowered by compromised nutrition.²⁸ Serum IGF-1 tended to be lower in our cohort, specifically those with greater bilirubin and bile acid concentrations. Others have similarly reported lower IGF-1 in children (ages 1.9 to 16 years) with hepatic cholestasis.²⁹ In addition to the bone geometry and microarchitecture findings, the lower muscle area also might have involved IGF-1-mediated mechanisms since this hormone plays a prominent role in muscle mass accrual.³⁰ Interestingly, average periosteal circumference Z-scores, although low, were not significantly from zero despite the lower muscle cross-sectional area, suggesting that deficits in cortical area and thickness were not attributed to skeletal muscle.

Growth failure in ALGS is evident at birth,³¹ yet only 20% of our cohort had “short stature,” defined as height-for-age Z-score < −2.0. This figure is substantially lower than reported previously in a different cohort of children with ALGS where approximately 90% (12 of 13 children) of pre-pubertal children had short stature.⁵ Compared to the patients from Wasserman and colleagues,⁵ our study participants were substantially older, were at a later stage of maturation, and presented with fewer ALGS-related clinical features, suggesting a less severe phenotype. Perhaps the extent to which longitudinal growth is impacted in children with ALGS depends on the severity of the condition. Nevertheless, cortical bone Z-scores remained lower even after adjusting for tibia length, indicating that these findings were not attributed to shorter stature. Height adjustment reduced the wide variability in DXA bone Z-scores, but similar to the age-adjusted measures, did not differ significantly from zero. This is in opposition to earlier findings where lower bone mass at the lumbar spine, but not the total body, in children with ALGS appeared to be attributed to shorter stature.⁷ However, this earlier study used a method for adjusting for short stature that did not account for age-specific effects. “Low bone mass for age” is defined as a BMC Z-score < −2.0,³² and the authors from this earlier study reported that 40% of children with ALGS had a height-specific lumbar spine BMC Z-score < −2.0. In the current study, we used the preferred method of adjusting bone outcomes for height-for-age Z-score;³² but, none of our patients surpassed this threshold for lumbar spine bone measures. It is still unknown whether stature-adjusted Z-scores are more closely linked to fracture risk in comparison to age-adjusted measures.

With respect to bone health, fracture is the most important clinical outcome. Albeit in a small sample size, thirty percent of our patients reported sustaining at least one broken bone in their lifetime. Six total fractures were reported and all but one were attributed to a “low impact” event. These figures are in line with those published by Bales and colleagues who reported that 28% of children with ALGS sustained at least one fracture, and that the vast majority of fractures were experienced with little or no trauma.⁸ Fracture etiology in ALGS is still unknown, but others⁶ showed lower bone density in those with a history of fracture. Our small sample size limited our ability to compare bone measures based on fracture history. Adequately powered studies are needed to determine whether bone geometry or microarchitecture predict fracture risk in ALGS.

As discussed above, our sample size was small, so we had limited statistical power to observe relationships between ALGS and bone measures, or to perform formal statistical tests on relationships between fracture and maturation stage with bone outcomes. We used robust age, sex, and ancestry-specific reference data for DXA and pQCT Z-score calculations, which represents a major strength of this study. Nevertheless, because our cohort was predominantly female (90%), the generalizability of these findings to male patients with ALGS is unknown. Pediatric reference data for the XtremeCT II HR-pQCT scanner have yet to be developed, but a dataset comprised of healthy boys and girls collected using identical scan parameters and acquisition procedures was available to help facilitate visual comparisons against our ALGS cohort. A similar approach also was utilized for the blood biochemical markers, as strong pediatric reference data were not available for all of these measures. Although formal statistical tests were not performed, these hypothesis-generating visual comparisons are suggestive of bone microarchitectural deficits and a compromised growth hormone-IGF-1 axis in children with ALGS. Furthermore, as discussed above, our patients had lower muscle cross-sectional area at the lower leg, which might have contributed to the lower cortical bone Z-scores. Additional insight regarding the role of skeletal muscle in ALGS-related bone deficits might be gained through functional testing, as muscle mass and area are used as surrogate measures of the forces exerted upon the bone during muscle contraction.

In summary, we report novel evidence linking ALGS with inferior cortical and trabecular bone geometry and microarchitecture. Decreased cortical thickness of the tibia, associated with increased endosteal circumference, and distal 1/3 radius aBMD by DXA were particularly pronounced. In addition, ALGS was associated with shorter stature and lower muscle area, and markers of cholestasis were associated with inferior trabecular bone microarchitecture and lower IGF-1. Collectively, these results expand on earlier studies implicating ALGS in lower bone mass^6,7,14 and increased fracture risk.⁸ Additional work is needed to more thoroughly define the mechanisms for these skeletal deficits in ALGS. Notably, prospective studies including a larger patient cohort while utilizing clinically relevant and advanced bone imaging technologies will help progress this area of research and the translatability of study findings to the clinical arena. Such work will be integral to developing efficacious preventative and treatment strategies aimed at reducing fracture risk in children with ALGS.

Supplementary Material

2

Supplemental Figure 1. Blood biochemical markers in females (left) and males (right) with ALGS (black dots) and healthy controls (gray dots). BSAP (U/L), bone-specific alkaline phosphatase; CTX (ng/mL), collagen type-1 cross-linked C-telopeptide; PTH (ng/L), parathyroid hormone; and 25(OH)D (ng/mL), 25-hydroxyvitamin D. The horizontal dotted lines in the 25(OH)D plots represent the Institute of Medicine’s cutoff for vitamin D insufficiency (20 ng/mL).

Highlights.

1. Children with Alagille syndrome had deficits in cortical bone size and trabecular bone microarchitecture.

2. Cortical bone deficits in children with Alagille syndrome were maintained after accounting for shorter limb length.

3. Greater bilirubin was associated with inferior trabecular bone microarchitecture in children with Alagille syndrome.