Bone density and structure in long-term survivors of pediatric allogeneic hematopoietic stem cell transplantation



Children requiring allogeneic hematopoietic stem cell transplantation (alloHSCT) have multiple risk factors for impaired bone accrual. The impact of alloHSCT on volumetric bone mineral density (vBMD) and cortical structure has not been addressed. Tibia peripheral quantitative computed tomography (pQCT) scans were obtained in 55 alloHSCT recipients, ages 5 to 26 years, a median of 7 (range, 3–16) years after alloHSCT. pQCT outcomes were converted to sex- and race- specific Z-scores relative to age based on reference data in >700 concurrent healthy participants. Cortical section modulus (Zp; a summary measure of cortical bone structure and strength), and muscle and fat area Z-scores were further adjusted for tibia length for age Z-scores. AlloHSCT survivors had lower height Z-scores (−1.21 ± 1.25 versus 0.23 ± 0.92; p < 0.001), versus reference participants; BMI Z-scores did not differ. AlloHSCT survivors had lower trabecular vBMD (−1.05; 95% confidence interval [CI], −1.33 to −0.78; p < 0.001), cortical Zp (−0.63; 95% CI, −0.91 to −0.35; p < 0.001), and muscle (−1.01; 95% CI, −1.30 to −0.72; p < 0.001) Z-scores and greater fat (0.82; 95% CI, 0.54–1.11; p < 0.001) Z-scores, versus reference participants. Adjustment for muscle deficits eliminated Zp deficits in alloHSCT. Total body irradiation (TBI) was associated with lower trabecular vBMD (−1.30 ± 1.40 versus −0.49 ± 0.88; p = 0.01) and muscle (−1.34 ± 1.42 versus −0.34 ± 0.87; p < 0.01) Z-scores. Growth hormone deficiency (GHD) was associated with lower Zp Z-scores (−1.64 ± 2.47 versus −0.28 ± 1.24; p = 0.05); however, muscle differences were not significant (−1.69 ± 1.84 versus −0.78 ± 1.01; p = 0.09). History of graft versus host disease was not associated with pQCT outcomes. In summary, alloHSCT was associated with significant deficits in trabecular vBMD, cortical geometry, and muscle area years after transplantation. TBI and GHD were significant risk factors for musculoskeletal deficits. Future studies are needed to determine the metabolic and fracture implications of these deficits, and to identify therapies to improve bone accrual following alloHSCT during childhood. © 2012 American Society for Bone and Mineral Research.


Allogeneic hematopoietic stem-cell transplantation (alloHSCT) is an established therapy for hematologic malignancies in children and adults.1 Five-year cure rates for childhood alloHSCT currently exceed 60%.2 AlloHSCT survivors have numerous risk factors for poor bone health, including malnutrition,3 vitamin D deficiency,4 reduced muscle strength,5 sex hormone deficiencies,6 chemotherapy,7 total body irradiation (TBI),8 and immune suppressive therapies. Graft-versus-host-disease (GVHD) and dysregulation of the immune system impose significant additional threats to bone health, potentially resulting in osteoclast activation and decreased number and function of osteoblasts.9 These risk factors may have lasting effects, long after alloHSCT treatment is complete.

During childhood, skeletal development is characterized by sex, maturation, and race-specific increases in trabecular and cortical bone mineral density (BMD) and cortical dimensions.10 This rapid accumulation of bone mass requires the coordinated actions of growth factors and sex steroids in the setting of adequate biomechanical loading and nutrition. The growing skeleton may be especially vulnerable to the effects of alloHSCT therapies and complications that suppress bone formation and promote bone resorption, such as inflammatory cytokines, GVHD, calcineurin inhibitors. and glucocorticoid therapy.11, 12

Dual energy X-ray absorptiometry (DXA)-based studies have demonstrated substantial bone deficits in children and adults within the first year following alloHSCT.13, 14 Bone biomarker studies have confirmed elevated markers of bone resorption and low markers of bone formation15 with vertebral compression fractures in 10% to 20% of pediatric and adult alloHSCT survivors.13, 15 Although these studies indicate significant abnormalities in bone metabolism following alloHSCT, the structural underpinnings of bone deficits in the growing and mature skeleton following alloHSCT have not been characterized. DXA is a two-dimensional technique that combines trabecular and cortical bone mass within the projected bone area and does not provide discrete measures of trabecular and cortical volumetric BMD (vBMD) or cortical dimensions.

Quantitative computed tomography (QCT), on the other hand, is a three-dimensional technique that distinguishes between cortical and trabecular bone. It measures vBMD and estimates cortical bone dimensions that are highly correlated with bone strength.16 Kaste and colleagues previously reported significant deficits in QCT measures of spine trabecular vBMD in alloHSCT survivors; however, this study did not include measures of cortical vBMD, cortical dimensions, or muscle area, and the reference data were limited.17

The objectives of this prospective study were (1) to assess trabecular and cortical vBMD, cortical dimensions, and muscle and fat area using tibia peripheral QCT (pQCT) in children and adolescents in long-term alloHSCT survivors, compared to a robust, healthy reference group, and (2) to identify correlates of bone deficits in survivors of childhood alloHSCT, such as delayed maturation, muscle deficits, disease, and treatment characteristics.

Patients and Methods

Study participants

The study population included children and young adults treated with alloHSCT and followed at the Children's Hospital of Philadelphia (CHOP). Inclusion criteria included age greater than 5 years, with at least a 3-year interval since alloHSCT. Subjects were excluded if they had a history of diseases known to affect bone health, including neuromuscular disease, inflammatory bowel disease, sickle cell anemia, active malignancy, or renal dysfunction (estimated glomerular filtration rate [eGFR] < 60 mL/min/1.73 m2). A total of 73 eligible subjects were identified and 55 enrolled. Nonparticipants did not differ from participants in age at disease diagnosis, age at alloHSCT, conditioning regimen, type of marrow donor, frequency of GVHD, or endocrinopathies after alloHSCT.

AlloHSCT subjects were compared to 985 healthy reference participants, ages 5 to 30 years. These participants were recruited from general pediatric and internal medicine practices in the greater Philadelphia area and through community advertisements to characterize bone health and body composition in healthy subjects, as previously described.10, 18, 19 Healthy reference participants were ineligible if they had a history of illnesses or medications that may affect growth, nutritional status, or bone health. Because of changes in the scanning protocols over the study interval and exclusion of scans with movement or other artifacts, pQCT measures of muscle and fat were available for comparison with alloHSCT subjects in 952 reference participants, whereas pQCT measures of trabecular and cortical bone outcomes were available in 735 and 759 reference participants, respectively. The distributions of age, sex, race, height, and body mass index (BMI, kg/m2) did not differ across the healthy reference groups contributing data for each measurement.

DXA whole-body fat and lean mass results were recently described in these alloHSCT recipients, demonstrating significantly lower lean mass and greater fat mass, compared with the reference participants. DXA lean mass and fat mass Z-scores were not significantly associated with alloHSCT complications or treatment characteristics.20

The study protocol was approved by the Institutional Review Board at CHOP. Informed consent was obtained directly from study participants older than 18 years, and assent along with parental consent from participants less than 18 years of age.

Anthropometry, physical maturity, and race

Height was measured with a stadiometer (Holtain, Crymych, UK) and weight with a digital scale (Scaletronix, White Plains, NY, USA). Pubertal development stage in the alloHSCT participants was determined according to the method of Tanner by a pediatric endocrinologist (S.M-M.). For the reference group, Tanner stage was determined using a validated self-assessment questionnaire.21 Tibia length was measured with a segmometer from the distal margin of the medial malleolus to the proximal border of the medial tibia condyle. Study participants and their parents were asked to categorize the participant's race according to the National Institute of Health categories.

Disease characteristics and medications

All 55 subjects underwent alloHSCT for leukemia or bone marrow failure syndrome. The medical records were reviewed for alloHSCT disease and treatment characteristics including date and type of primary diagnosis, date of alloHSCT, donor type (matched related or unrelated), conditioning regimen (including use and dose of total body irradiation), and history of acute or chronic GVHD. Myeloablative conditioning regimens consisted of cyclophosphamide, thiotepa, and fractionated TBI (range, 1200–1320 cGy) or busulfan and cyclophosphamide. Nonmyeloablative regimens included busulfan and cytoxan ± melphalan or fludarabine. All recipients received cyclosporine A infusions, starting 2 days before transplantation, and the dosage was adjusted to maintain plasma levels of 300 to 400 ng/mL. Prior to discharge, cyclosporine A was transitioned to tacrolimus and discontinued by 6 months after alloHSCT in the absence of GVHD.

For subjects with GVHD, onset, duration, and type of immunosuppressive regimen were determined. GVHD classification was based on a four-point scale (I indicating mild disease, and IV indicating severe disease).22 Cumulative glucocorticoid exposure from time of transplantation to the date of study visit was assessed through review of the medical record and summarized as cumulative milligrams (mg), cumulative milligrams per kilogram (mg/kg), average milligrams per kilogram per day (mg/kg/d), and days since last dose.

Subjects and parents were interviewed at the study visit to review the medical history, including fracture history, presence or absence of gonadal insufficiency, growth hormone deficiency (GHD), and thyroid disease. Endocrinopathy characteristics and hormone therapy were further delineated with detailed medical chart review. For subjects reporting the use of dietary supplements, the calcium and vitamin D contents from supplement sources were recorded during the study interview, and when necessary, confirmed by telephone after the study visit.

Bone assessment by pQCT

Bone measures in the left tibia were obtained by pQCT using a Stratec XCT2000 device (Orthometrix, White Plains, NY, USA) with a 12-detector unit, voxel size of 0.4 mm, slice thickness of 2.3 mm, and scan speed of 25 mm/s. All scans were analyzed with Stratec software version 5.50 at CHOP. A scout view was obtained to place the reference line at the proximal border of the distal tibia growth plate in participants with open growth plates and at the distal endplate in participants with fused growth plates. Bone measures were obtained 3% and 38% proximal to the reference line. At the 3% metaphyseal site, scans were analyzed for trabecular vBMD (mg/cm3). At the 38% diaphyseal site, scans were analyzed for cortical vBMD (mg/cm3), periosteal circumference (mm), endosteal circumference (mm), cortical cross-sectional area (CSA, mm2), and polar section modulus (Zp, mm3). Zp is a function of the cortical periosteal, and endosteal dimensions, and is strongly associated with bone failure load.16 Muscle and fat CSA (mm2) were evaluated at the 66% site. The manufacturer's hydroxyapatite phantom was scanned daily for quality assurance. In our laboratory, the coefficient of variation (CV) for short-term precision ranged from 0.5% to 1.6% for pQCT outcomes in children and young adults.

Laboratory studies

Nonfasting blood samples were collected during the study visit in all alloHSCT subjects and a subset of the healthy reference participants. Serum 25 hydroxyvitamin D [25(OH)D] concentration was analyzed using 125I-labeled radioimmunoassay (Diasorin Corporation, Stillwater, MN, USA).23 The intraassay and interassay CVs were 6% and 15%, respectively. Serum bone-specific alkaline phosphatase (BSAP, µg/L) was measured as a marker of bone formation (Quest Diagnostics Laboratories, San Juan Capistrano, CA, USA) using a two-site immunoradiometric assay (CV 8%). Serum β-C-terminal telopeptide (β-CTX, pg/mL) was measured as a marker of bone resorption (Quest Diagnostic Laboratories) using the Roche Cobas E170 electrochemiluminescent assay (CV 5%). Serum 25(OH)D and bone biomarkers in the reference group were assayed in the same laboratory using the same methods as the alloHSCT samples. Bone biomarkers were available only in reference participants less than 21 years of age. Serum 25(OH)D specimens were obtained in 208 and bone biomarkers in 187 and 409 (β CTX and BSAP) reference subjects respectively—based on participation in varied concurrent protocols.24

Statistical analysis

Stata 10.0 (Stata Corp., College Station, TX, USA) was used for all statistical analyses. A p value <0.05 was considered statistically significant, and two-sided tests of hypotheses were used throughout. Group differences between alloHSCT recipients and healthy reference subjects were tested using the t test, with adjustment for unequal variance as needed, or using the rank sum test if indicated. Differences in proportions were assessed using the chi-square test.

Age- and sex-specific Z-scores (standard deviation scores) for height and BMI were calculated in participants less than 20 years of age using National Center for Health Statistics 2000 Center for Disease Control growth data.25 The pQCT outcomes were converted to Z-scores relative to age using the LMS method, as described.18, 19 All of the pQCT Z-scores were sex- and race-specific (black versus all others) and were generated using the LMS Chartmaker Program, version 2.3 (Medical Research Council, UK) in the healthy reference participants.26 The LMS method fits three parameters (LMS) as cubic splines by nonlinear regression. The three parameters represent the median (M), standard deviation (S), and power in the Box-Cox transformation (L) that vary as a function of age. The pQCT cortical geometry and muscle and fat outcomes were highly correlated with tibia length (all p < 0.0001). The alloHSCT subjects were older, on average, than reference subjects of a comparable tibia length (p < 0.001) due to growth failure. Therefore, the cortical geometry and muscle and fat CSA Z-scores that were generated relative to age were subsequently adjusted for tibia length-per-age Z-score using linear regression analyses. This approach was used to capture the differences in the joint distributions of age and tibia length in children with alloHSCT compared with the healthy reference group. Of note, the investigators previously used a similar approach to adjust DXA Z-scores for height Z-score, demonstrating that this approach eliminated the bias introduced by lower height Z-scores.27 Multivariate regression models were used to examine pQCT outcomes in alloHSCT subjects compared with the healthy reference participants. Models were further adjusted for Tanner stage to determine if delayed maturation contributed to bone and body composition deficits. The models for cortical geometry were subsequently adjusted for muscle CSA Z-scores in order to determine if adjustment for muscle deficits attenuated the bone deficits in alloHSCT recipients, compared with the healthy reference participants.28

Values for serum biomarkers of bone turnover (BSAP and β-CTX), cumulative glucocorticoid exposure (mg/kg), and average glucocorticoid (mg/kg/d) during treatment interval were natural log transformed to achieve normal distributions using the lnskew0 function in Stata 10.0. Tanner- and sex-specific Z-scores for bone biomarker levels were calculated using reference participant data.24 Linear regression models were used to compare bone turnover biomarkers in alloHSCT subjects, compared with reference participants, adjusted for sex, Tanner stage, and the significant sex-by-Tanner interaction, as described.24 Additional linear regression models were used to assess associations between the laboratory parameters and pQCT Z-scores. Vitamin D deficiency was defined as 25(OH)D level <20 ng/mL, consistent with a recent Institute of Medicine Report.29 Multivariate logistic regression was used to examine the odds of vitamin D deficiency in alloHSCT subjects, compared with healthy reference participants, adjusted for age, race, and winter season, as described.30 Given the association of decreased 25(OH)D levels in children and adults based on BMI-based categorization of obesity in previous studies,31 we also examined the odds of vitamin D deficiency in alloHSCT subjects adjusted for BMI Z-scores.

Additional multivariable linear regression models limited to alloHSCT subjects were used to identify potential determinants of bone outcomes such as disease characteristics, conditioning regimen, exposure to TBI, history of GVHD, glucocorticoid exposure, and endocrinopathies after alloHSCT. In order to minimize the heterogeneity of the subject population, we also examined the pQCT results limited to the 35 participants in the two largest alloHSCT subgroups: acute myelogenous leukemia (AML, n = 23) and acute lymphoblastic leukemia (ALL, n = 12).


Participant and disease characteristics

A total of 55 alloHSCT subjects, median age 15 (range, 5–26) years, 69% male, and 7% black, were enrolled. AlloHSCT subjects showed significantly delayed pubertal maturation: within Tanner stages 2, 3, and 4, alloHSCT subjects were an average of 2.1, 2.5, and 2.8 years older than the reference participants (p < 0.01 for all), adjusted for sex and race. AlloHSCT subjects had significantly lower median (range) height Z-scores compared to healthy reference participants (−1.21 [range, −4.19 to 1.95] versus 0.23 [range, −2.59 to 3.19]; p < 0.001]. BMI Z-scores did not differ significantly (0.28 [range, −4.79 to 2.68] versus 0.40 [range, −3.09 to 2.99]; p = 0.26). Height and BMI Z-scores did not differ between male and female alloHSCT subjects.

AlloHSCT disease characteristics are summarized in Table 1. The most common diagnosis necessitating alloHSCT was AML (23 subjects; 42%). The majority received a matched-related donor source (29 subjects; 52%). Three participants (5%) required a second alloHSCT due to recurrence of disease. The median interval between alloHSCT and study visit was 7 (range, 3–16) years. Thirty eight (69%) received TBI as part of the alloHSCT conditioning regimen and 29 subjects (53%) did not experience GVHD. Among the 26 subjects treated with glucocorticoids for GVHD, only one was still on glucocorticoids at the time of study visit and the remainder had discontinued glucocorticoids a median of 6 (range, 2–16) years prior. Disease and treatment characteristics did not differ between male or female alloHSCT subjects.

Table 1. Disease and Treatment Characteristics in AlloHSCT Recipients
  • Data presented as median (range) or n (%).

  • alloHSCT = allogeneic hematopoietic stem cell transplantation; HSCT = hematopoietic stem cell transplantation; TBI = total body irradiation.

  • a

    Ovarian failure present in 5 of 10 female alloHSCT participants >12 years of age.

Age at study enrollment, years15.1 (5.1–25.5)
Age at diagnosis, years5.7 (0.0–20.0)
Age at transplant, years6.7 (0.1–21.1)
Interval since transplant, years6.8 (3.0–16.4)
Diagnosis, n (%)
 Acute myeloid leukemia23 (42)
 Acute lymphoblastic leukemia12 (22)
 Chronic myeloid leukemia6 (11)
 Myelodysplastic syndrome5 (9)
 Juvenile myelomonocytic leukemia5 (9)
 Aplastic anemia2 (4)
 Bone marrow failure syndrome2 (4)
Donor source, n (%)
 Related29 (53)
 Unrelated25 (45)
 Cord1 (2)
TBI conditioning regimen, n (%)38 (69)
Graft versus host disease, n (%)
 None29 (53)
 Acute22 (85)
 Chronic16 (62)
 Active at time of study visit1 (4)
Treatment with glucocorticoids following HSCT, n (%)
 None29 (53)
 >2 years prior25 (45)
 Current1 (2)
Endocrine abnormalities, n (%)
 Hypothyroid20 (36)
 Growth hormone deficient16 (29)
 Hypogonad (males)6 (16)
 Ovarian failure (females)a5 (29)

A total of 49 alloHSCT subjects (89%) were diagnosed and treated for an endocrinopathy. Hypothyroidism was the most common endocrine diagnosis with 20 participants (36%) requiring treatment with thyroid replacement. Sixteen participants (27%) were diagnosed with GHD and eight of these participants (50%) received treatment with growth hormone (GH). Among the 10 alloHSCT girls greater than 12 years of age, seven were menarcheal, only two of these seven reported regular menses, and five required hormone replacement for treatment-related ovarian failure. Six males (16%) required testosterone replacement for hypogonadism, all older than 16 years of age at the time of testosterone initiation. Only three subjects (5%) had a previous diagnosis of precocious puberty requiring treatment with Lupron during concomitant treatment with growth hormone therapy. None had adrenal insufficiency. Nine alloHSCT subjects (16%) were diagnosed with two hormonal deficits (hypothyroidism and GHD), while four subjects (7%), exhibited multiple hormone deficits (hypothyroidism, GHD, and primary gonadal failure). All alloHSCT subjects with a diagnosis of endocrinopathy were on appropriate hormone replacement at the time of this study.

pQCT outcomes

The Z-scores for the pQCT outcomes in alloHSCT subjects are summarized in Table 2, with p values representing comparison with the healthy reference participants. There were no significant sex differences in any of the bone parameter Z-scores presented in Table 2. Trabecular vBMD Z-scores were significantly lower (p < 0.001) in alloHSCT subjects compared with the healthy reference participants. Cortical vBMD Z-scores did not differ. AlloHSCT survivors had significantly lower cortical section modulus, cortical CSA, and periosteal circumference Z-scores, compared with healthy reference participants (p < 0.001 for all); endosteal circumference Z-scores were not significantly different. Further adjustment for delayed maturation in alloHSCT survivors did not attenuate the group differences for bone deficits. AlloHSCT survivors had significantly higher fat CSA and lower muscle CSA Z-scores (both p < 0.001), compared with healthy reference participants, consistent with the prior whole body DXA results.20

Table 2. pQCT Z-Scores in alloHSCT Subjects, Compared With Reference Participants
Z-scoreAlloHSCT, β (95% CI)p
  1. The sex- and race-specific Z-scores are presented as β coefficient (95% confidence intervals) in alloHSCT recipients compared with reference participants. The Z-scores for the pQCT vBMD outcomes were generated relative to age. The pQCT bone geometry, muscle, and fat CSA Z-scores were generated relative to age and subsequently adjusted for tibia length for age Z-scores.

  2. pQCT = peripheral quantitative computed tomography; alloHSCT = allogeneic hematopoietic stem cell transplantation; vBMD = volumetric bone mineral density; NS = not significant; CSA = cross-sectional area.

Trabecular vBMD−1.05 (−1.33 to −0.78)<0.001
Cortical vBMD−0.20 (−0.48 to 0.08)NS
Section modulus−0.63 (−0.91 to −0.35)<0.001
Cortical CSA−0.71 (−0.99 to −0.43)<0.001
Periosteal circumference−0.53 (−0.78 to −0.27)<0.001
Endosteal circumference−0.16 (−0.44 to −0.12)NS
Muscle CSA−1.01 (−1.30 to −0.72)<0.001
Fat CSA0.82 (0.54 to 1.11)<0.001

Muscle CSA Z-scores were highly correlated with Zp (r = 0.67, p < 0.001), cortical CSA (r = 0.74, p < 0.001), and periosteal circumference (r = 0.67, p < 0.001) Z-scores within the alloHSCT recipients. Comparable correlations (r = 0.63, r = 0.65, and r = 0.63, p < 0.001 for all) were observed within the reference participants. Muscle CSA Z-score was moderately correlated with endosteal circumference Z-score in both alloHSCT and reference participants (r = 0.29, p = 0.03 and r = 0.29, p < 0.001, respectively). When the Zp, cortical CSA, and periosteal circumference Z-score models were adjusted for the lower muscle CSA Z-scores in alloHSCT survivors, the cortical bone deficits in alloHSCT subjects were no longer significant, compared with healthy reference participants. Zp (β coefficient: −0.04; 95% CI, −0.28 to 0.19; p = 0.72), cortical CSA (−0.08; 95% CI, −0.31 to 0.15; p = 0.50), and periosteal circumference (0.02; 95% CI, −0.20 to 0.24; p = 0.83) Z-scores.

Biomarkers of bone turnover

When compared to the healthy reference participants, BSAP and β-CTX levels were comparable in alloHSCT subjects: log-transformed BSAP (0.06; 95% CI, −0.01 to 0.13; p = 0.11) and β CTX (−0.06; 95% CI, −0.15 to 0.02; p = 0.13), adjusted for sex, Tanner stage, and sex-by-Tanner stage interaction.19 BSAP Z-scores in alloHSCT subjects were lower but not significantly different from healthy reference participants (median, −0.32; interquartile range, −1.74 to 2.71; p = 0.85), while β-CTX Z-scores were significantly lower (median, −0.46; interquartile range, −1.96 to 2.84; p = 0.04). Bone biomarker Z-scores were not significantly associated with pQCT bone results, TBI, or GHD.

Vitamin D levels and nutritional supplementation

The proportion of alloHSCT subjects with vitamin D deficiency (29%) was not significantly different compared with the reference participants (38%). The odds of vitamin D deficiency in alloHSCT subjects was not significantly different compared with the reference group when adjusted for age, black race, and season (odds ratio, 1.02; 95% CI, 0.49–2.11; p = 0.96). Further adjustment for BMI Z-scores did not change the results. Serum 25(OH)D levels were not associated with bone or body composition Z-scores in the healthy reference participants or alloHSCT subjects.

Fourteen alloHSCT subjects (25%) were taking a multivitamin that contained vitamin D or were prescribed vitamin D supplementation. The vitamin D dose varied from 400 to 1200 IU/d, with a median dose of 400 IU/d. Vitamin D supplementation was associated with higher levels (mean 37 versus 26 ng/mL, p = 0.003). Thirteen (24%) of alloHSCT subjects were taking a multivitamin or supplement that contained calcium ranging from 100 to 1266 mg/d. Of these, eight (15% overall) were taking 100 mg/d, with three (5%) subjects taking 1000 mg/d. None of the pQCT or laboratory measures differed between alloHSCT subjects taking versus not taking calcium supplementation.

Associations between AlloHSCT disease characteristics, growth, and pQCT outcomes

AlloHSCT subjects had significantly lower height Z-scores when compared to the reference participants, with 18% of alloHSCT subjects demonstrating a height Z-score < −2 SD (p < 0.001). The median height Z-score limited to alloHSCT recipients less than 16 years of age at transplantation (ie, those with likely remaining growth potential) was −1.28 (range, −4.19 to 1.95). Greater height deficits were associated with younger age at time of alloHSCT (r = 0.33, p = 0.01), exposure to TBI (p = 0.01), and diagnosis of GHD (p < 0.001) (Table 3). Furthermore, a history of GVHD (p < 0.01), greater cumulative glucocorticoid (p < 0.001), and longer duration of glucocorticoid exposure (p < 0.001) after alloHSCT were additional risk factors for height deficits in survivors.

Table 3. AlloHSCT Z-Scores According to Disease and Treatment Characteristics
Z-ScoreHeightTrabecular vBMDSection ModulusPeriosteal CircumferenceEndosteal CircumferenceMuscle Cross-Sectional AreaFat Cross-Sectional Area
  1. Cortical vBMD Z-scores did not differ according to treatment and disease characteristics.

  2. vBMD = volumetric bone mineral density; TBI = total body irradiation; GVHD = graft versus host disease; GHD = growth hormone deficiency.

TBIp = 0.01p = 0.01p = 0.04p = 0.07p > 0.2p < 0.01p > 0.2
 Yes−1.38 ± 1.00−1.30 ± 1.40−0.95 ± 2.02−0.73 ± 1.26−0.16 ± 1.11−1.34 ± 1.420.79 ± 1.19
 No−0.40 ± 1.29−0.49 ± 0.88−0.09 ± 0.88−0.16 ± 0.84−0.21 ± 0.95−0.34 ± 0.870.82 ± 0.98
GVHDp < 0.01p > 0.2p > 0.2p > 0.2p > 0.2p > 0.2p > 0.2
 Yes−1.53 ± 1.10−1.24 ± 1.51−0.90 ± 2.35−0.67 ± 1.36−0.17 ± 1.06−1.32 ± 1.480.69 ± 1.28
 No−0.67 ± 1.11−0.88 ± 1.11−0.51 ± 1.10−0.47 ± 0.99−0.18 ± 1.08−0.78 ± 1.190.91 ± 0.97
GHDp < 0.001p = 0.07p = 0.05p = 0.03p = 0.10p = 0.09p = 0.04
 Yes−2.02 ± 0.97−1.56 ± 1.62−1.64 ± 2.47−1.15 ± 1.35−0.57 ± 1.19−1.69 ± 1.840.24 ± 1.27
 No−0.69 ± 1.04−0.84 ± 1.12−0.28 ± 1.24−0.30 ± 1.000.00 ± 0.95−0.78 ± 1.011.03 ± 0.99

Age at time of alloHSCT and interval since alloHSCT were not associated with pQCT vBMD, cortical geometry, or muscle Z-scores (all p > 0.42). Table 3 summarizes alloHSCT disease and treatment characteristics and the associations with pQCT outcomes. The Z-scores for pQCT vBMD outcomes were generated relative to age, whereas pQCT bone geometry, muscle, and fat CSA Z-scores were generated relative to age and subsequently adjusted for tibia length for age Z-score (body size). TBI was associated with significantly lower trabecular vBMD and muscle CSA but not cortical geometry Z-scores. GVHD after alloHSCT was not associated with bone outcome or muscle deficits. In addition, pQCT bone Z-scores were not associated with prior cumulative glucocorticoid exposure (mg) (β coefficient: −0.08; 95% CI, −0.73 to 0.57; p = 0.80), average glucocorticoid exposure (mg/kg/d) (−0.29; 95% CI, −1.59 to 1.01; p = 0.65), or days since last glucocorticoid dose (−0.11; 95% CI, −0.76 to 0.54; p = 0.74) for treatment of GVHD after alloHSCT. GHD was associated with lower section modulus and periosteal circumference Z-scores. Figures 1 and 2 illustrate the associations between pQCT outcomes and TBI and GHD. Repeat analysis limited to combined ALL and AML subgroups, compared with the healthy reference participants showed similar results (data not shown).

Figure 1.

Total body irradiation (TBI) and pQCT outcomes: trabecular bone mineral density (BMD), section modulus, and muscle cross sectional area (CSA) Z-scores in alloHSCT subjects. Darker boxes represent TBI exposure.

Figure 2.

Growth hormone deficiency (GHD) and pQCT outcomes: trabecular bone mineral density (BMD), section modulus, and muscle cross sectional area (CSA) Z-scores in alloHSCT subjects. Darker boxes represent GHD.

Fracture summary

A total of 15 alloHSCT subjects (27%) reported fractures with seven (47%) subjects experiencing fracture prior to alloHSCT and eight (53%) subjects reporting fracture post-alloHSCT over a total of 478 patient-years. The fracture sites following transplantation included wrist (n = 2), ankle (n = 2), phalanges (n = 2), metacarpal (n = 1), and vertebral bone (n = 1).


To our knowledge, this is the first study to assess trabecular and cortical vBMD, cortical structure, and the functional muscle-bone unit in survivors of pediatric alloHSCT. Although the vast majority of alloHSCT recipients had not been treated with glucocorticoids or other immunosuppressive medications for many years, this cohort demonstrated substantial growth failure, low trabecular vBMD, smaller cortical CSA, Zp, and periosteal circumference (adjusted for shorter tibia length), as well as cachexia. The magnitude of these deficits exceeded those observed in children with active Crohn's disease,18 juvenile rheumatoid arthritis,32 and chronic kidney disease,33 highlighting the lasting impact of alloHSCT and its therapies. Furthermore, this study delineated the discrete associations between TBI, GHD, and trabecular and cortical deficits. Importantly, the robust reference participants in this study constitute the largest reference population used in the assessment of skeletal outcome in pediatric alloHSCT survivors, allowing adjustment for age, sex, race, bone length, pubertal maturation, and body composition.

Skeletal deficits in survivors of childhood alloHSCT have not been adequately characterized. There are few studies to date specifically assessing BMD in pediatric populations after alloHSCT.13, 17, 34 In the earliest study, Bhatia and colleagues reported DXA total body BMD deficits (median Z-score −0.5) in a small sample of 10 pediatric patients, ages 3 to 18 years, a median of 2 years following alloHSCT for hematologic malignancies.13 The study was limited by lack of data on the effect of growth failure on DXA BMD Z-scores relative to age.27, 35 In contrast, Nysom and colleagues reported no difference in height-adjusted DXA whole-body BMD and bone mineral content in 25 survivors of childhood alloHSCT, 4 to 13 years after treatment, compared to healthy controls.34 However, as noted by Zemel and colleagues, DXA methods adjusting BMD Z-scores for height without consideration of age results in an underestimation of bone deficits27; this may explain the absence of bone deficits reported in the Nysom and colleagues study.34 In the only QCT study to date in alloHSCT survivors, Kaste and colleagues reported significant spine vBMD deficits (median Z-score, −0.88) in 48 participants a median of 5 years after alloHSCT.17 These deficits were not associated with gender, age at alloHSCT, time since transplantation, conditioning regimen, or endocrine dysfunction in survivors. Our current study extends these findings to include measures of cortical vBMD, cortical structure, and body composition.

We recently demonstrated significant excess adiposity and muscle deficits using whole-body DXA in these alloHSCT recipients, despite normal BMI.20 A potential explanation for bone deficits and excess adiposity in alloHSCT survivors stems from the inverse relationship between bone and fat formation within the marrow cavity,36 where complex local stimuli regulate mesenchymal stem-cell differentiation into osteoblasts or adipocytes.37 Aging and osteoporosis have been associated with increased marrow adiposity and decreased osteogenesis.37, 38 Prior studies in adolescents with anorexia nervosa and bone deficits have documented hypoplasia of the hematopoietic marrow with an increase in marrow fat.39 Longitudinal QCT studies of bone accrual in healthy young females have shown an inverse correlation with marrow adiposity.40 Similarly, in a recent pQCT study, Farr and colleagues demonstrated that greater intermuscular fat was associated with lower trabecular vBMD and bone strength index in healthy girls,41 underscoring the important interactions between bone and fat.

It is well-established that glucocorticoids result in preferential trabecular bone loss with reductions in bone volume fraction, trabecular thickness, and trabecular connectivity.42 Prior studies have demonstrated that the degree of trabecular network disruption correlates with cumulative exposure to oral glucocorticoids; however, this disruption is reversible shortly after cessation of glucocorticoids, particularly with longitudinal growth.43 While we cannot exclude the possibility that prior glucocorticoid exposure contributed to trabecular bone deficits in this cross-sectional study of alloHSCT recipients, it is unlikely given that the median interval since glucocorticoid therapy was 6 years. Furthermore, the bone biomarkers were not consistent with our recent observation that glucocorticoid therapy was associated with lower BSAP levels in children.24 At the same time, it is possible that some of the measured deficits in trabecular BMD and cortical dimensions reflect irreversible changes from alloHSCT during adolescence, a critical period for bone accrual.

Normal sex- and maturation-specific skeletal development during childhood requires coordinated actions of growth factors and sex steroids. The growth hormone (GH)/(insulin-like growth factor 1 [IGF-1]) axis is a major determinant of bone mass,44 and IGF-1 plays a critical role in skeletal development through increases in bone length and dimensions.45 AlloHSCT survivors are at risk for abnormalities of the GH/IGF-1 axis as a consequence of TBI exposure,46 with children showing greater susceptibility to GHD development after exposure to low doses of irradiation.47 In this study, the lower height Z-scores in alloHSCT survivors were associated with GHD (p < 0.001) and exposure to TBI (p < 0.01). Furthermore, alloHSCT subjects demonstrated significant pubertal delay as expected due to treatment-related toxic gonadal effects. Gonadal failure related to alloHSCT is well known and multifactorial,48 with deleterious effects of high-dose chemotherapy and TBI.49 Thus, treatment-related GHD and hypogonadism further contribute to compromised skeletal acquisition in pediatric alloHSCT survivors. Importantly, this study demonstrated lasting associations between TBI, GHD, and musculoskeletal deficits despite the fact that all alloHSCT subjects with a diagnosis of endocrinopathy were on appropriate hormone replacement at the time of this study.

During growth, as muscles increase, bones adapt by increasing in dimensions and strength. The capacity of bone to respond to mechanical loading with increased strength is greatest during childhood.50 Given the strong association between muscle mass and cortical bone dimensions, investigators advocate assessment of the functional muscle-bone unit in children with chronic disease.51 Although bone and muscle deficits are highly correlated, this relationship may not be entirely causal, and may be mediated by nutritional, hormonal, or inflammatory factors that directly influence both muscle and bone.52 In the current cohort of alloHSCT survivors, the smaller periosteal circumference and section modulus was no longer significantly reduced, compared with reference participants, when adjusted for the lower muscle CSA Z-scores. It is not known if the bone deficits are a consequence of the significant cachexia, or reflect endocrine deficiencies or inflammation. In this study, muscle CSA Z-scores were not significantly different in subjects with versus without GHD (p = 0.09); however, on average, muscle CSA Z-scores were 0.9 SD lower in subjects with versus without GHD, suggesting potential clinical significance. Muscle CSA Z-scores were significantly lower in alloHSCT subjects after TBI exposure (p < 0.01). Given that TBI exposure is a significant risk factor for GHD, some alloHSCT subjects without a clinical diagnosis of GHD may actually exhibit inadequate muscle mass due to subclinical inadequacies in the GH/IGF-1 axis. In addition, alloHSCT subjects had no greater odds of vitamin D deficiency compared with the reference participants when adjusted for age, race, and winter season. Therefore, it is unlikely that vitamin D deficiency contributed to these lasting musculoskeletal deficits.

The primary limitation of this study is the cross-sectional design. We are unable to determine if bone deficits are progressive in long-term survivors, or if there is potential for recovery. This may account for our failure to demonstrate meaningful alterations in bone biomarkers. Nonetheless, these data demonstrate striking deficits years after transplantation and completion of therapy. An additional limitation is the heterogeneous patient population. However, this is the largest study of its kind, and a sensitivity analysis limited to the AML and ALL patients showed the same results. Finally, the lack of bone histomorphometry data made it impossible to assess bone remodeling rates or mineralization. Additional limitations include the lack of data on dietary or protein intake, physical activity, or laboratory measures of inflammation. Nonetheless, this study establishes the need for further studies of bone deficits in alloHSCT recipients.

In summary, this is the first study to demonstrate substantial deficits in trabecular and cortical bone in children and adolescents after alloHSCT. Future longitudinal studies are necessary to determine if these deficits progress or recover in older adolescence and early adulthood, and to identify associations with fractures. The accurate characterization of skeletal deficits and identification of significant risk factors in childhood alloHSCT survivors are essential to guide future intervention therapies such as antiresorptive therapy (ie, bisphosphonates) for low bone turnover from reduced osteoblast differentiation, or anabolic approaches, such as therapy with growth hormone, intact PTH analogues, or low-magnitude mechanical stimuli53 in this growing population.


All authors state that they have no conflicts of interest.


This work was supported by grants from St. Baldrick's Foundation, NIH (R01 HD040714, R01 DK064966, K24 DK076808), and Clinical Translational Research Center (UL 1-RR-024134). We thank Dan Schiferl for pQCT technical support.

Authors' roles: Detailed substantial intellectual contribution of each listed author according to ICMJE guidelines is given. Each author meets conditions 1, 2, and 3, with specific contribution details outlined: SM-M is the principal investigator and primary author of the manuscript, responsible for conception and study design, patient recruitment, data collection, data analysis, data interpretation, inception, and writing of the entire manuscript; as primary author, she made sure that every listed author read, reviewed, and approved the manuscript's final submitted version. JPG was instrumental in acquisition of data through patient recruitment; she revised the manuscript and specifically contributed to the intellectual content with respect to HSCT late effects after cancer therapy in pediatric survivors; she reviewed and approved the final submitted version of the manuscript. NB contributed to the study conception and interpretation of data with respect to HSCT treatment details, conditioning regimen, and treatment-related complications such as graft versus host disease in study participants, given her expertise in pediatric HSCT as the director of the Bone Marrow Transplantation Program at the Children's Hospital of Philadelphia; importantly, she critically revised the manuscript with important intellectual contribution to HSCT disease and treatment detail presentation as well as interpretation of the results; she read and approved the final submitted version of the manuscript. BZ has expertise in pediatric bone and body composition; she has authored numerous manuscripts on appropriate interpretation pediatric bone abnormalities in chronic disease using DXA and pQCT; importantly, she has participated in several studies with the senior author, MBL; she was instrumental in the study conception, design, and complex data analysis of bone structures specifically representation of cortical geometry and muscle and fat CSA Z-scores that were generated relative to age and subsequently adjusted for tibia length-per-age Z-score using linear regression analyses in the healthy reference population used in this study as well as the HSCT study participants; she provided specific guidance in the study's complex data interpretation and critically revised the manuscript; she read and approved the final submitted version of the manuscript. JS is the study statistician with expertise in analysis and interpretation of pediatric bone and body composition; she has worked extensively with the study primary author for study design and statistical guidance for the complex analysis; she critically revised the manuscript to appropriately reflect the accurate representation of the statistical methods; and she read and approved the final submitted version of the manuscript. MBL, the senior author, was instrumental in direct mentorship and support of the principal investigator, with close supervision of study development, process, data acquisition, and data analysis and interpretation, as well as essential input in the development of the final manuscript; given her expertise in pediatric bone and body composition, she provided detailed critical revision of the manuscript and data interpretation; and she read, reviewed, and approved the final submitted version of the manuscript. Authors SM-M and MLB accept responsibility for the integrity of data analysis.