Introduction
Osteoporosis is a multifactorial skeletal disorder resulting in bone fragility and is associated with fractures, morbidity, and mortality [
22]. Pediatric osteoporosis is generally categorized in primary and secondary forms. Secondary osteoporosis and thus also secondary low bone quality in children is caused by systemic disease, their treatment, or indirect effects of systemic disease, such as immobilization, reduced time spent outdoor, and poor nutrition, or a combination of these factors [
28].
Osteoporosis in children is defined differently than in adults, because bone mass (bone mineral density, BMD) varies greatly with age. Therefore, bone densitometry uses pediatric
Z-scores that refer to an age-appropriate cohort of healthy children and adolescents, instead of adult
T-scores [
1,
26].
Z-scores ≤ − 2.0 are defined as low bone density or mass for age [
14]. According to the 2013 International Society of Clinical Densitometry, pediatric osteoporosis is defined as low bone density for age in combination with a clinically significant fracture history, i.e., ≥ 2 long bone fractures before age of 10 or ≥ 3 long bone fractures before age of 19, or the presence of one or more vertebral compression fractures occurring without major trauma or local disease [
2].
Peak bone mass is a strong predictor of fracture risk and osteoporosis in adulthood. Because 90% of peak bone mass is acquired by the age of 18, any (in)direct effects of pediatric disease influence bone status in both childhood and adulthood [
22,
28]. Incidence records of secondary low bone mineral density (BMD: an equivalent of bone mass) vary; however, the incidence of low BMD in non-ambulatory children with cerebral palsy is reported up to 97% [
7,
11]. Children with limited ambulation typically have low BMD and many will sustain fractures [
8].
Dual-energy X-ray absorptiometry (DXA) is the golden standard for bone quality measurement in children as well as adults, due to precision, reproducibility, and availability of normative data [
14,
19]. DXA measurements give information about BMD of the site studied. Different skeletal sites are described for BMD measurement in children. DXA of the lumbar spine L1–L4 (DXA
LS) is a recommended site and is superior to DXA of the femur or fore arm [
14]. Nevertheless, DXA has its limitations. Disrupting factors as movement during measurement, metallic implants, contractures, and sometimes even scoliosis can cause results to be non-interpretable. In addition, the
Z-scores are based on calendar age and do not take bone age into account, hence, may provide inaccurate findings [
8,
12,
13]. Finally, DXA provides measurement of areal BMD (g/cm
2), rather than volumetric density (g/cm
3), which may result in underestimation of BMD in children with small or narrow bones and overestimation of BMD in children with tall stature [
1,
8]. These limitations feel the need for alternative methods.
The clinically available digital X-ray radiogrammetry (DXR) of the hand seems a feasible alternative. Using a web-based software like BoneXpert, it can assess both bone age and bone quality, expressed as bone health index (BHI), a measure of cortical thickness and mineralization, which may result in an accurate representation of bone quality. The BHI reference values are gender and bone age specific. DXR is less stressful compared to DXA, easy to obtain, and often does not involve additional exposure to ionizing radiation, since hand radiographs for the assessment of bone age are regularly obtained in disabled children that are prone to low bone quality [
13,
19,
27].
Several studies in pediatric, adolescent, and adult populations showed that bone quality measured by DXR may correlate well with DXA measurements [
3,
17‐
19,
24]. Other studies mentioned sensitivity and specificity of DXR compared to DXA of the lumbar spine and/or total body bone mineral density as the “golden standard,” varying from 40–90 to 79–93%, respectively [
15,
16]. However, these studies were performed in specific (i.e., children with juvenile idiopathic arthritis or intestinal failure) and small (
n = 24–35) pediatric populations [
15,
16]. This raises the question whether DXR is a viable alternative for DXA for bone health assessment in children with high probability of secondary low bone quality, which has not yet been proven.
This study compares the measurements of DXR and DXA performed in children with high probability of secondary low bone quality, thus determining the diagnostic accuracy of DXR as a method for determining bone health in these children.
Discussion
Our results show that DXR may be a promising alternative for measuring bone health in children with high probability of secondary low bone quality. BHI has a significant and positive correlation with all DXA measurements. Agreement between BHI and as well BMDLS as bone age-adjusted BMDLS is high, especially for Z-scores ≤ − 2.0. BHI Z-scores show best diagnostic performance when compared with BMDLSZ-scores without correction for bone age.
To our knowledge, this is the first prospectively planned cross-sectional study comparing DXR and DXA in a diverse group of children with high probability of secondary low bone quality and the first study to compare DXA scores corrected for bone age. We used DXR with automated calculation of bone age and bone quality using the BoneXpert software. The advantages of DXR over DXA are its low burden on patients, no influence of soft tissue thickness on bone quality calculations, and since many children with suspicion of low bone quality have an indication for a hand radiograph to analyze bone age, the extra software postprocessing does not involve additional exposure to ionizing radiation [
10,
13,
27]. Additionally, conventional radiographs of the hand can be performed in any hospital in the Netherlands, in contrary to DXA [
13].
Previous studies compared DXR and DXA measurements in different groups of children; however, only a few compared DXA
Z-scores with DXR. The correlation coefficients between BMD
LS and BHI
Z-scores in our study were higher than those found in children with juvenile idiopathic arthritis or children with suspected secondary low bone mineral density [
16,
19]. A study in children with inflammatory bowel disease showed comparable correlation coefficients between BMD
LS and BHI
Z-scores [
4]. Neelis et al. showed slightly higher correlation coefficients in children with intestinal failure [
15]. That study also attempted to show agreement between BMD
LS and BHI. Limits of agreement were comparable with our study; however, their variability was non-consistent, most likely due to their small number of participants. Sensitivity and specificity of DXR compared to DXA
LS were similar [
15].
Unfortunately, DXR assessment was not possible in 10 of 118 (8%) of eligible patients, due to overprojection of metacarpals, anatomical bone deformities, and missing bone age. This percentage is higher than in previous studies that showed 1.4 to 7.6% [
15,
19,
23]. However, in our study DXR only failed in severely disabled patients. If ambulatory status is taken into account, DXR had a feasibility of 100% in full ambulatory patients and was successful in 27/37 (73%) severely disabled patients. This is better than the 63.2% reported in a feasibility study in severely disabled children [
13].
We investigated correlation, agreement, and distinguishing features of DXR compared to DXA
LS, since DXA
LS is a recommended skeletal site for BMD measurement in children and is superior to DXA of the femur or fore arm [
14]. BHI and BMD
LSZ-scores correlated well and showed moderate to good agreement. Percentage similarity showed good agreement for
Z-scores ≤ − 2.0. Nonetheless, in 27 (27%) patients BHI and BMD
LSZ-scores differed greatly (> 2
Z-scores). This difference was caused by a large discrepancy between bone and calendar age in 9/27 (33%) individuals. In 16 of the 18 (89%) remaining patients, BMD
LS was higher than the BHI
Z-score, probably because DXR is described to be more sensitive to irregularities than DXA [
15]. Another possible explanation is that DXR is sensitive to a decrease in the amount of bone tissue, but unsuitable for bone mineralization defects, especially when affecting the cortical bone [
19,
24]. Certain syndromic disorders and genetic disorder that are known for altered bone mineralization may cause differences between DXA and DXR. Furthermore, DXR only measures cortical bone [
24], and therefore, disease and/or medication that alter trabecular bone may create differences in BMD measurements by DXR and DXA. However, for clinical practice, it is most important to determine low bone quality (
Z-scores ≤ − 2.0). Our results showed good agreement between BHI and BMD
LS and bone age-adjusted BMD
LS for
Z-scores ≤ − 2.0. Additionally, in this series with high prevalence of children with low bone quality, DXR showed good negative predictive value of 82% and 89% when compared to BMD
LS and bone age-adjusted BMD
LSZ-scores. These values were even higher for ambulatory patients (BMD
LSZ-score, 86%; bone age-adjusted BMD
LSZ-scores, 94%), most likely because immobility induces altered bone geometry, i.e., thinner cortices and reduced cortical diameters [
5], resulting in lower BHI values for non-ambulatory patients.
In the present study, bone density was significantly impaired. As shown in Table
2, it is presumable that ambulatory status had a large contribution to the impaired mineral bone density in all patients. Ambulatory status did not influence bone age. Bone age was significantly impaired compared to a healthy cohort; hence, we adjusted the DXA
Z-scores for bone age and compared these with the BHI
Z-scores, as was suggested earlier [
15].
Bone age is associated with pubertal maturation and could therefore be an advantage [
29]. Unexpectedly, DXA scores adjusted for bone age were slightly worse correlated with BHI
Z-scores than DXA Z-scores. In addition, comparison of bone age-adjusted DXA
LSZ-scores with BHI
Z-scores showed a significant mean bias of 0.48. Although this mean bias differed significantly from the mean difference between DXA
LSZ-scores with BHI
Z-scores, differences from the Bland-Altman analyses showed a very strong positive correlation. Similarity percentage between BHI and bone age-adjusted DXA
LSZ-scores were comparable for
Z-scores ≤ − 2. We hypothesize that these results are most likely due to lack of reference range data for bone age-adjusted DXA
LS scores [
5]. In the present study, we used the same reference database for as well the DXA
LS as the bone age-adjusted DXA
LSZ-scores. It is probable that normal distribution between bone age and BMD diverge; therefore, bone age-adjusted DXA
LSZ-scores can differ from measured DXA
LSZ-scores. In addition, the bone age was significantly impaired in our cohort, resulting in higher bone age-adjusted DXA
LSZ-scores than measured DXA
LSZ-scores. Moreover, it is likely that differences between bone age-adjusted DXA
LS and measured DXA
LSZ-scores tend to be more pronounced when there is a larger difference between bone age and calendar age, as for our population.
Strengths of our study include its prospectively planned cross-sectional design. The amount of patients included is substantially larger than in previous studies [
10,
15‐
17]. In addition, our patient group is heterogeneous in both ambulatory status and medical conditions that could lead to secondary low BMD. Therefore, our study population is representative for a large amount of patients that might need bone quality assessment.
Further, the median time interval between DXA and hand radiographs was 0 day. This makes our comparison more reliable, especially for the bone age-adjusted DXA
Z-scores, compared to studies where the time interval was up to 8 months [
19].
Some limitations of our study should be addressed. Subgroup analyses were difficult to interpret, because the non-ambulatory group was smaller than the full ambulatory patient group. Also, since we only included patients who underwent both DXA and DXR, there may be a selection bias. Some parents did not want to make an additional DXR. In addition, clinicians’ interpretation of a high probability of low bone mineral density could vary, and therefore, there could be a selection bias.
In conclusion, DXR and DXA measurements correlate well. BHI and as well BMDLS as bone age-adjusted BMDLSZ-scores show good agreement for Z-scores ≤ 2.0, especially the comparison of BHI and BMDLSZ-scores. Our results suggest DXR to be a promising alternative for DXA for determining low bone quality in children with suspected secondary low bone quality or osteoporosis. Future research should include gathering of reference data for bone age-adjusted DXA Z-scores, the value of DXR in predicting future fracture risk, and the value of DXR in measuring the therapeutic effects of different interventions. For these last two reasons, prospective, longitudinal studies are required.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.