Multiple underlying health problems and associated treatments can impair bone mineral accrual during childhood and adolescence. We sought to identify the risk factors associated with low bone density in subjects referred for DXA measurements of BMD at a tertiary care pediatric hospital to help inform a bone health patient safety initiative. Using a multivariable logistic regression model, we found that low BMI Z-score, history of BMT, and 25OHD insufficiency were the medical conditions most predictive of a BMD Z-score ≤ -2, a threshold deemed by expert consensus to represent a significant low bone density for age .
We identified that history of BMT significantly increases the risk of low BMD. Multiple mechanisms have been postulated to mediate the low BMD seen after BMT, including glucocorticoid use, chemotherapy, irradiation, hypogonadism, decreased activity levels, and chronic graft versus host disease . Prior studies support our findings, demonstrating that BMT during childhood increases the risk of low BMD and fracture [29, 30], and that bone formation is decreased and bone resorption is increased in the months immediately following BMT .
Our multivariate analysis demonstrated that 25OHD insufficiency (< 30 ng/mL) is associated with a nearly 4-fold increased risk of low BMD. This finding is consistent with the observation of Cheng et al. demonstrating that low 25OHD levels correlate with lower cortical BMD in pre-pubertal and pubertal girls . Unlike many of the underlying medical conditions included in our analysis, 25OHD insufficiency may be improved with 25OHD supplementation and thus represents an area amenable to intervention for improved bone health in children and adolescents. Our selection of values less than 30 ng/mL for defining 25OHD deficiency was based upon the lower limit of the normal range of our hospital laboratory. After we designed and initiated our study, the Institute of Medicine reported on dietary reference intakes for calcium and 25OHD, and concluded that an adequate 25OHD level is ≥ 20 ng/mL . However, controversy regarding adequate serum 25OHD levels remains (e.g., above 20 or 30 ng/mL or higher)  and additional research is needed to identify the optimal therapeutic target.
We found that low BMI is a significant risk factor for low BMD, yet despite the robust relationship between underweight and low BMD, we did not find a relationship between malnutrition and low bone density. Prior studies have demonstrated a strong relationship between malnutrition in young women with anorexia nervosa, a model of malnutrition-induced bone loss, and low BMD [35–37], which we did not observe in this sample. Bone density in these patients can vary depending on weight loss or gain, and severity of disease. The patients with anorexia nervosa included in this study could have represented a skewed subgroup, with less severe disease and bone loss. A prospective study of larger sample size would likely be powered to reveal the expected relationship between malnutrition and low BMD.
The multivariate analysis demonstrated that low height Z-score is a risk factor for low BMD. Any child with an underlying health problem may have short stature, confounding bone density measurements obtained by DXA . It is important to recognize that DXA measurements may be confounded by bone size, an important issue to consider when studying small underweight adolescents with anorexia nervosa or other pediatric chronic diseases . Zemel et al. recently proposed a method for height adjustment in calculating DXA measurements of bone mass in growing children  which will be useful to account for the effect of stature on BMD measurement. Despite these limitations in smaller children, our findings suggest that children who are underweight may be at risk for decreased BMD.
Consistent with prior reports of gender differences in fracture risk [40, 41], we found that fractures occurred in a higher percentage of boys than girls in a population referred for DXA measurement. This gender difference has been attributed to higher rates of sports-related injuries in boys [40, 41]. Surprisingly, in our univariate analysis of factors associated with BMD Z-score ≤ - 2, we found that male gender was a risk factor for low BMD. This finding is questionable, as there is a known association between male gender and higher BMD , and as gender is no longer predictive in the multivariate model, there may be other confounders influencing this finding. As our subjects were selected from a retrospective sample of patients referred for DXA measurements, this observation may also represent selection bias and merits further study.
Prior studies have demonstrated that IBD is both associated with bone loss [43–45], and increased fracture risk in adults . Thus, the finding that a history of malabsorption or IBD was associated with decreased risk of low BMD in the univariate and multivariate analyses was unexpected. The retrospective design of our study may have introduced bias, or referral bias may have influenced the selection of subjects with IBD referred for DXA. Additionally, the timing of the BMD measurements in relation to the IBD diagnosis and associated malabsorption and therapy may have contributed to this finding.
Limitations of this study should be acknowledged. The study design was a retrospective cohort study; therefore, the findings represent associations and causality cannot be inferred. Our results are limited by information obtained solely from subjects referred for DXA measurements at a tertiary care medical center. However, DXA is the assessment tool used most commonly in clinical practice around the world for bone density measurements in children and adults. Thus, we are hopeful that the information gleaned will be useful for clinicians, and in particular, experts in bone health. Measurements of volumetric BMD of the peripheral skeleton were not available and would be valuable to obtain, especially in a cohort of young patients, although this tool is only infrequently used at the present time for clinical purposes. By design, our study was not able to include or identify subjects who were not able to have DXA measurements performed. The retrospective nature of our study required us to rely upon diagnoses and referrals made by providers and the samples sizes in many diagnostic categories were small, limiting our ability to detect differences in BMD. Additionally, due to the reliance on the medical record, we were not able to include assessments of physical activity. We acknowledge that physical activity data would have been useful, as exercise may positively impact BMD  and chronic disease often limits activity for a child or adolescent. Obtaining an accurate family history of osteoporosis was also limited by the retrospective nature of this study, as some charts did not include documentation of this information. However, despite these limitations, our results identified the relative risks of a variety of medical conditions known to be associated with low BMD in a sample of patients referred for DXA at a tertiary care center. The information gained from our study highlights pediatric populations who may be at highest risk for a low BMD and may be candidates for bone density screening.