Spinal Bone Texture Assessed by Trabecular Bone Score in Adolescent Girls With Anorexia Nervosa

Abstract

Context:

Trabecular bone score (TBS) is a bone assessment tool that offers information beyond that afforded by dual-energy x-ray absorptiometry (DXA) bone mineral density (BMD) measurements. Adolescents with anorexia nervosa (AN) are known to exhibit compromised bone density and skeletal strength.

Objectives:

This study aimed to determine TBS among adolescents with AN and evaluate the correlation with anthropometric, clinical and densitometric variables.

Design:

Areal BMD spinal measures were analyzed for TBS. Findings were compared with clinical (height, weight, body mass index [BMI], age, pubertal development, 25-hydroxyvitamin D) and self-reported data (illness duration, amenorrhea, exercise, fracture, family history of osteoporosis, and antidepressant use), and BMD measures by DXA and peripheral quantitative computed tomography (pQCT).

Setting and Participants:

This was an urban adolescent program consisting of 57 females with AN, age 11–18 y.

Interventions:

Interventions included DXA (absolute BMD and Z-score), pQCT (volumetric BMD [vBMD] and stress-strain index [SSI]), laboratory evaluation, and questionnaire administration.

Main Outcome Measures:

Main outcome measures included TBS, areal and vBMD, SSI, fracture history, disease duration.

Results:

The TBS of six participants (11%) showed degraded and 19 (33%) partially degraded microarchitecture. Spinal TBS was correlated (P < .05) with age, height, weight, BMI, pubertal stage, BMD, and body composition by DXA, and BMD and SSI by pQCT. TBS was not correlated with disease duration, fracture, vitamin D status, race, or ethnicity, and self-reported health data.

Conclusions:

TBS showed evidence of degraded microarchitecture in over 40% of this study sample, and strongly correlated with anthropometric data and measures of BMD and skeletal strength. TBS is a novel tool that captures another dimension of bone health in adolescents with AN.

The most commonly used tool for evaluating bone health in both adolescents and adults is dual-energy x-ray absorptiometry (DXA), a safe, widely available densitometric technique (1, 2). Two-dimensional lumbar spine DXA images can be used to extract a trabecular bone score (TBS), a gray-level textural metric that mathematically captures elements of three-dimensional bone architecture that are not obtained in a standard DXA bone mineral density (BMD) measure (3, 4). This new method of assessing bone health offers information about a bone's solid volume fraction and mean solid thickness (3), in addition to providing information in older adults about fracture risk (4). A higher TBS indicates healthier skeletal microarchitecture, whereas a lower value suggests a weaker bone structure.

Previous studies using TBS have been conducted in adults and have shown declining TBS in postmenopausal women (5–8), and in men with a history of fragility fractures (9, 10). TBS is also low in some adults with fragility fractures whose standard BMD measurement by DXA did not predict osteoporosis (11, 12). Of high clinical relevance, TBS has been shown to be associated with fracture risk in adults with a variety of conditions associated with low BMD or abnormal bone quality (13–16).

Adolescents with anorexia nervosa (AN) represent a patient group at high risk for inadequate bone accrual, bone loss, and fractures (17–20). AN is one of the most common chronic illnesses of adolescence, and affected individuals exhibit multiple hormonal abnormalities that may alter bone turnover, ultimately leading to both suboptimal bone accretion and skeletal losses (21–23). To our knowledge, no previous studies have used TBS to evaluate bone health among a pediatric population. The objective of the current study was to evaluate the correlation between lumbar spine TBS and areal measures of spinal and whole-body BMD, body composition, tibial BMD by peripheral quantitative computed tomography (pQCT), anthropometric measures, duration of disease, and reported fracture among a sample of adolescent girls with AN. As secondary outcomes, we examined the relation of spinal TBS to duration of disease, age, pubertal development, vitamin D status, family history of osteoporosis, race/ethnicity, exercise history, and antidepressant use.

Materials and Methods

Study population

Data were obtained from the baseline visit of a clinical trial of adolescent girls with AN. Participants were recruited from an urban eating disorders program and were deemed to be hemodynamically stable for study participation by their health provider. Exclusion criteria included presence of other chronic disease that could affect bone health, and use of glucocorticoids or other medications known to alter bone metabolism over the previous 6 months. A urine pregnancy test was also obtained and confirmed negative prior to participation. Measures for this analysis were all obtained prior to randomization and treatment for the clinical trial. Fifty-seven adolescent girls age 11–18 years were recruited who had undergone baseline areal BMD measures of the spine (L1–L4, in g/cm²), hip (total hip region), and whole body (total body and total body less head) by DXA (Table 1). The two-dimensional gray-scale images of the spine were retrospectively analyzed for TBS. The study was approved by the hospital institutional review board and all participants gave informed assent or consent, and in those less than age 18 years, a parent or guardian also gave informed consent.

Table 1.

Sample Anthropometrics, Clinical Parameters, and Bone Measures, Correlated With Trabecular Bone Score

Characteristic	N	Distributiona	Correlation with TBS (r)	P
TBS	57	1.35 ± 0.10
Age, yr	57	15.5 ± 1.9	0.28	.03
Height, cm	57	159.4 ± 7.1	0.45	.0005
Weight, kg	57	48.1 ± 6.5	0.64	<.0001
BMI, kg/m2	57	18.9 ± 1.8	0.57	<.0001
Duration of anorexia, mo	57	4 (1–36)	0.03	.81
Duration of amenorrhea, mo	49	3 (1–17)	0.03	.85
Exercise, h/wk	49	4 (0–25)	−0.12	.41
25(OH) vitamin D, ng/mL	56	31 ± 9	−0.13	.32
Bone mineral density, g/cm2
Lumbar spine	57	0.903 ± 0.126	0.81	<.0001
Hip	57	0.895 ± 0.130	0.74	<.0001
Femoral neck	57	0.783 ± 0.114	0.67	<.0001
Whole body	57	1.059 ± 0.105	0.77	<. 0001
Whole body, less head	57	0.906 ± 0.102	0.77	<.0001
Bone mineral density z score
Lumbar spine	57	−0.45 ± 1.31	0.58	<.0001
Hip	57	−0.30 ± 1.15	0.69	<.0001
Femoral neck	57	−0.51 ± 1.05	0.62	<.0001
Whole body	57	0.13 ± 1.25	0.67	<.0001
Lean mass, kg	57	34.2 ± 5.0	0.57	<.0001
Fat mass, kg	57	11.8 ± 3.4	0.37	.0045
Fat mass, %	57	24.4 ± 5.2	0.11	.40
Total vBMD at 3% site, mg/cm3	57	307 ± 48	0.62	<.0001
SSI (38% site), mm3	57	1310 ± 285	0.58	<.0001
SSI (66% site), mm3	57	2008 ± 477	0.57	<.0001

^aMean ± sdwith Pearson correlation; or Median (min–max) with Spearman correlation (skewed measure). P tests hypothesis of zero correlation.

Data sources

Data were obtained via a confidential participant interview between the principal investigator (C.M.G.) or research assistant (J.M.O.) and the study participant. Questions were posed regarding past medical history (including gynecologic history and duration of eating disorder), medication use, lifetime fractures, and family history of osteoporosis. Information regarding weekly exercise and dietary intake was obtained via the Youth/Adolescent Questionnaire, a comprehensive, semiquantitative food frequency questionnaire validated for use in this age group (24, 25). A serum 25-hydroxyvitamin D (25[OH]D) measurement was also obtained at the time of the baseline visit using a Diasorin chemiluminescent assay (LIAISON; DiaSorin, Inc.; interassay coefficient of variation, 8.6–10.0%).

Measurement of BMD and body composition

All DXA scans were performed in the Bone Densitometry Unit at Women and Infants Hospital in Providence, RI, using a Hologic scanner (Hologic Inc.; Discovery W, software version 23.1). DXA measures of absolute areal BMD (g/cm²) were obtained of the lumbar spine (L1–L4), hip (total hip region), and whole body (total body and total body less head), and BMD Z-scores were calculated using pediatric software (26). Body composition data obtained by DXA included body fat mass, lean mass, and body fat percentage. The scanner used exhibited stable short- and long-term in vitro precision (coefficient of variation [CV], < 0.5%). In vivo precision (CV) was 0.6% for the lumbar spine, 1.8% for the total hip, and 1.0% for the whole body.

Measurement of tibial BMD and SSI by pQCT

All pQCT scans were performed in the Clinical Research Center at Rhode Island Hospital, using a Stratec XCT 3000 device (Orthometrix). All scans were analyzed with Stratec software version 6.20. Scans were obtained of the 3, 38, and 66% (percentage distance of tibial length from the distal growth plate) regions of the left tibia. Volumetric BMD (vBMD) at the 3% metaphyseal site, and strength-strain index at the 38 and 66% diaphyseal sites were generated. Quality control was monitored daily using the hydroxyapatite phantom provided by the manufacturer (CV < 0.5%). In vivo precision (CV) of total vBMD measurements ranged from 2.1% (3% tibial site) to 0.9% (66% site).

Measurement of lumbar spine TBS

TBS measurements were obtained from DXA spinal images using TBS iNsight Software (version 2.1.1.0, Medimaps).

Statistical analysis

All data including questionnaires and imaging results were deidentified prior to analysis. All statistical analyses were performed using SAS (version 9.4, SAS Institute). P =.05 was set as the threshold for statistical significance. The association of TBS with clinical characteristics and DXA measures was determined using Pearson and Spearman correlation analyses, as appropriate for variables with a nonskewed or skewed distribution, respectively. TBS was compared between participant subgroups using Student t tests.

Results

A total of 57 participants were included in the analysis, with data obtained from the baseline visit of a clinical trial. Participant mean age was 15.5 years old (SD, 1.9 y) and the median grade level was ninth grade (range, grade 6–13). Seven participants were nonwhite by self-report (12%) and four were Hispanic (7%). Participants were late in development, with 68% at Tanner stage IV and 25%, Tanner stage V for breasts; 36% at Tanner stage IV and 63% Tanner stage V for pubic hair. Twenty of the 57 participants (35%) had a BMD Z-score at the lumbar spine that was ≤−1; of those, 5 (9%) were ≤−2. The whole-body Z-score was ≤−1 in nine participants (16%) and ≤−2 in five of those (9%). The participants had a low mean body mass index (BMI) of 18.9 kg/m² (SD, 1.8 kg/m²), and the duration of amenorrhea was variable (Table 1).

Mean TBS, lumbar spine, and whole-body areal BMD and BMD Z-score, and baseline characteristics were determined (Table 1). To assess TBS status, established norms for adults (ie, age ≥ 18 y) were applied. TBS at the lumbar spine showed degraded microarchitecture (TBS ≤ 1.20) in six (11%) of participants, partially degraded microarchitecture (TBS > 1.20 and < 1.35) in an additional 19 (33%), and normal values (TBS ≥ 1.35) in 32 (56%) subjects.

Lumbar spine TBS was shown to correlate significantly with age, height, weight, and BMI (Table 1). TBS was also shown to correlate significantly with BMD at the lumbar spine, hip, and whole-body BMD (total body and total body less head), as well as the accompanying BMD Z-score at all sites (Figure 1). Spinal TBS was also correlated significantly with vBMD by pQCT at the 3% tibia site, SSI at the 38 and 66% tibial sites, as well as with fat mass and lean mass. Three subjects had an extremely low BMI (< 15 kg/m²). Removing these participants from the analysis attenuated the correlation of TBS with age and fat mass, but had a negligible effect on the strong correlation of TBS with BMD and BMD Z-scores.

Figure 1.

Correlation between trabecular bone score and BMD Z-score. Correlation between TBS and BMD Z-score–lumbar spine (X) and whole body (filled circle) in 57 adolescent girls with anorexia nervosa. Dashed horizontal lines provide indication of the adult guidelines for the TBS: degraded microarchitecture (TBS ≤ 1.20), partially degraded microarchitecture (TBS > 1.20 and < 1.35), and normal microarchitecture (TBS ≥ 1.35).

Thirty percent (n = 17) of the study participants reported a history of fracture and 24% (n = 12) reported a family history of osteoporosis. Fifty percent (n = 28) exhibited vitamin D insufficiency (defined as 25[OH]D < 30 ng/mL), and 9% (n = 5) were vitamin D deficient (25[OH]D < 20 ng/mL). These clinical variables did not correlate significantly with TBS, nor did duration of AN or amenorrhea, weekly exercise, or 25(OH)D concentration (Table 1). TBS showed a weak but significant correlation with pubertal stage (Spearman r = 0.38, P = .004 for breasts; Spearman r = 0.30, P = .03 for pubic hair), but not with nonwhite race (P = .43) or Hispanic ethnicity (P = .13).

Discussion

This is the first study to test the feasibility of deriving TBS from DXA scans and its potential utility as a skeletal assessment tool in adolescents with AN, as well as to examine its use as a measurement technology in a pediatric population. Our TBS results showed evidence of degraded microarchitecture in greater than 40% of the adolescents studied, using adult reference standards, suggesting abnormal trabecular architecture to be present in this patient group. TBS was also strongly correlated with areal measures of absolute BMD and the accompanying BMD Z-score by DXA, as well as vBMD and SSI by pQCT. However, TBS was not correlated with fracture history, marking a departure from previous studies of adults.

In the current study, using normative data obtained in adults, an abnormal TBS was common among the subjects studied whereas the mean TBS for the sample was exactly at the threshold of normal. An explanation for this finding may be the weight status of the current participants, as the mean BMI was 18.9 kg/m². Although low for a healthy adolescent population, this BMI is relatively high for a sample with AN. The higher BMI could reflect a study sample with either less severe AN or shorter duration of disease than anorexic populations reported previously. Similarly, the spinal BMD Z-scores were <−1 in approximately one third of the sample, and a definitive low bone mass for age (ie, BMD Z-score <−2) was present in only a small percentage. These data are again consistent with a sample of less severely affected patients with AN. This Z-score threshold has been associated with higher fracture risk among children and adolescents, and expert consensus has endorsed it for reporting of densitometry results (2). The mean BMD Z-scores at all skeletal sites among the current participants were within the normal range. The entry criteria for this study were guided by standard diagnostic criteria for AN including adolescents with distorted body image cognition, weight restrictive behaviors, 3 or more months of amenorrhea, and a BMI less than 20 kg/m². We were interested in the effects of bone health across a broader range of BMI than has been conducted in previous studies. However, our recruitment strategy, although uncovering some skeletal abnormalities, could be interpreted as providing more reassurance than is warranted for these patients. Future studies should be carried out among healthy youth to provide reference data for TBS, and among additional patients with AN, include more subjects with a very low BMI and longer duration of disease. There seem to be inherent challenges with TBS in patients of severe underweight. Noteworthy is the fact that three of the current subjects had a BMI less than 15 kg/m², a threshold below which assumptions of this methodology may be violated and the accuracy of the information is less clear. Thus, the current exploratory study will hopefully contribute new knowledge on how TBS operates among pediatric patients, thereby informing the direction of future research in this area.

In contrast with the current findings among adolescents, previous studies in adults have not consistently yielded a correlation between TBS and measures of areal BMD (11, 12). The current study found these measures to be strongly and significantly correlated. TBS and measures of the peripheral skeleton by pQCT were both strongly correlated with volumetric bone density and SSI, the latter a variable that estimates skeletal strength. The strength of these correlations, even among a relative small sample size, provides some validation of TBS among this pediatric group. The strong correlation of TBS with anthropometric measures and body composition further supports the validity of this method among this patient population.

Prior studies using TBS, many of a large sample size, have found this methodology to predict fracture risk in adults who had a variety of conditions associated with low BMD or poor bone quality (13–16). Other studies in adult populations have explored the utility of evaluating TBS and BMD together to enhance fracture risk prediction (11, 27, 28). The current findings did not support this body of adult literature; TBS was not correlated with history of fracture among adolescent girls with AN, an illness that is clearly associated with poor bone health (17–19). Possible explanations for these findings may be both the relatively small sample size and difference in fracture risk between postmenopausal and adolescent female populations, as children and adolescents generally sustain fractures less frequently than older adults. Although fracture risk is known to be increased in adolescents with eating disorders (17–19), in general, pathologic fractures are less common among youth vs older adults with known osteoporosis. The fracture threshold for patients in the pediatric age range is also less clear than for adults, and in particular, the elderly (1, 2). As discussed previously, severity of illness could have contributed to the finding. For the current sample, the average duration of AN reflected a relatively short duration of illness. Thus, skeletal compromise may not have been advanced enough to provide valid data on fracture risk at the time of data collection. Longer-term, longitudinal assessments of fracture among children and adolescents, clearly defining the normative range for TBS among youth and including those with a low to low-normal TBS would better inform the extent to which this methodology predicts fracture risk in a pediatric population. In addition, this study was not powered to predict fracture; a much larger sample size would be needed to examine definitively the association between TBS in affected patients and reported fractures.

This study is the first to explore the use of TBS to inform bone health evaluations in subjects with AN. TBS was shown to correlate with multiple anthropometric variables in this sample, including height, weight, and BMI. These findings are of high interest as AN is known to be a classic model of malnutrition-induced bone loss and low bone accrual (17–20), with weight and BMI known to be inversely correlated with areal measures of BMD. Zemel et al (29) have pioneered the use of height adjustments for DXA measures, showing height-for-age Z-score to be the most informative tool for the adjustment of BMD Z-scores as generated by DXA among healthy youth. Similarly, it is of interest that height was found to be strongly correlated with TBS among adolescents with AN, a group known to possess many risk factors for both growth delay and poor bone health. Future studies will need to explore the relation between TBS and anthropometric and other clinical variables among both healthy children and adolescents, as well as those with chronic disease.

This study examined the relation of TBS with multiple clinical variables including vitamin D status, weekly exercise, and family history of osteoporosis; none of these secondary variables were found to be correlated with TBS. Race and ethnicity were also examined, and similarly showed no relation to TBS. However, the current sample and this study population are racially and ethnically homogeneous, with most participants being non-Hispanic white. These secondary outcome variables warrant further examination within the context of a larger sample size. Of interest, in one prior study, vitamin D deficiency was found to be less common among patients with AN compared with healthy youth (30), suggesting that the effect of low vitamin D status on bone health may not always be generalizable to all patients with AN. This prior study concluded that the findings were in part due to exquisite vitamin and supplement compliance among participants (30). Thus, the role of vitamin D deficiency in skeletal deficits among adolescents with AN is equivocal, and the current study adds TBS data to the clinical outcomes that have previously been examined.

Although the current study showed no association between antidepressant use (typically provided as a selective serotonin reuptake inhibitor [SSRI]) and TBS, these medications are commonly prescribed for adolescents with AN. Studies to date on SSRI use and bone health have been carried out primarily in adult patients (31, 32). One prior study in adolescents with AN (33) showed reported SSRI use greater than 6 months to be associated with low BMD Z-scores at the spine, femoral neck, and whole body. The current study did not document duration of SSRI use, and brief use, reflecting short duration of illness or recent onset of disease, could have contributed to the lack of association. However, given that an SSRI is often used to manage comorbid anxiety and depression among adolescents with AN, continued monitoring for adverse skeletal effects is warranted.

Limitations of the current study include the relatively small sample size, which limits the generalizability of the findings and may have contributed to nonsignificant relationships observed. In addition, the cross-sectional design prevents the ability to establish causality. Many of the data were obtained via self report, which may have introduced bias, noteworthy in a patient group whose disease is characterized by denial. The exploratory nature of this study, inherent in the first use of TBS among pediatric subjects, must be acknowledged, including the inclusion of some subjects with an extremely low BMI (ie, < 15 kg/m²). All participants in this study were female, which eliminates the opportunity to inform regarding the utility of TBS in bone health evaluation among adolescent males with AN. Finally, as mentioned previously, a key limitation that merits discussion is the current lack of normative data for the pediatric population. This limitation does not change results from the current study, including results of correlation analyses examining the relation between the absolute TBS and various clinical variables. In addition, it is important to note that the current subjects were either in late puberty or completely developed. Ninety-three percent were at Tanner stage IV or V for both breasts and pubic hair, thus near or past epiphyseal fusion and more similar to an adult from a skeletal standpoint, than younger children. However, additional research is needed to determine whether the normative range established for adults is appropriate for use in children and adolescents.

In this first study of TBS in a pediatric population, the current findings suggest that TBS is strongly correlated with both DXA measures of the axial skeleton and pQCT measures of the peripheral skeleton, as well as anthropometric variables and body composition. The most informative data that a skeletal assessment tool can provide is information on fracture risk. Although TBS did not predict fracture among adolescents with AN in this initial study, further research is warranted that explores this question in pediatric and eating disordered populations. Future studies should strive for a larger sample size, and explore similar measures of bone health among male as well as female pediatric patients. It will also be important to generate additional data to validate this methodology against standards measures of areal BMD, vBMD by computed tomography (standard and high-resolution), and include fracture assessments in a large sample of both healthy and diseased youth. These future studies will afford a more definitive understanding of the application of TBS to children and adolescents as a novel measurement technology.