Evaluation of the possibility to assess bone age on the basis of DXA derived hand scans—preliminary results
- First Online:
- Cite this article as:
- Płudowski, P., Lebiedowski, M. & Lorenc, R.S. Osteoporos Int (2004) 15: 317. doi:10.1007/s00198-003-1545-6
- 153 Views
The classical method of skeletal age assessment is based on the recognition of changes in the radiographic appearance of the maturity indicators in hand-wrist radiographs by comparison with a reference atlas. The purpose of this study was the evaluation of the possibility to assess bone age using a less invasive method such as dual-energy X-ray absorptiometry (DXA). Bone ages of 50 children free of any chronic diseases (5–18 years old) and ten with multihormonal pituitary deficiency (MPD) (8–20 years old) were assessed using an Expert-XL densitometer. Hand scans and classical hand-wrist radiographs were evaluated by two independent observers for bone age by visual comparison with reference standards of skeletal development published in the atlas. The precision errors of duplicate bone age ratings were low both for radiographs (<1%) and DXA hand scans (<0.9%). A high degree of agreement between bone age ratings done by two observers was assessed by intraclass correlation coefficients. The same bone age based on radiographs and DXA hand scans was assessed in 44 of 60 cases (73.3%); in 16 cases the differences between bone age were no higher than 0.5 year. No significant difference between mean bone age based on radiographs and DXA hand scans was observed (P>0.05). Moreover, there was a very strong correlation between bone age results (r=0.998; r2=0.996; P<0.0001), indicating agreement of bone age assessments based on DXA and radiographic images. Remarkable differences (up to 3 years) between bone age and chronological age were observed in healthy subjects, probably reflecting the effect of the secular trend towards earlier maturation or alterations in pubertal development. The study indicates that evaluation of skeletal maturity using DXA images is less invasive (up to 8 µSv) than radiography, giving results comparable to the classical method.
KeywordsBone ageBMDDXAHand densitometry
It is well known that children with the same chronological age often have different rates of skeletal maturation as a consequence of various genetic and social factors [1, 2, 3, 4]. Bone age is a useful index of biological maturation, used by pediatricians, orthodontists, anthropologists and others interested in the process of children’s growth. Bone age can be assessed at different skeletal sites.
The classical methods for the assessment of bone age are based on the recognition of changes in the radiographic appearance of the maturity indicators in hand-wrist radiographs by comparison with a reference atlas (Greulich and Pyle method), estimation of skeletal maturity scores (Tanner-Whitehouse method) or bone age computed on the grades of maturity indicators (FELS method) [5, 6, 7]. The main difficulty for the assessment of bone age using these methods is the lack of a standardized technique, subjectivity of evaluation and variability of bone age results assessed by different methods . The most important is still the fact that there are ethical considerations against exposure to irradiation involved in hand-wrist or even knee radiography procedures (40–100 µSv) . In developed countries, for example, radiologists have a legal obligation to keep doses “as low as readily achievable” , and guidance has been issued on the potential for reducing doses . Dose reduction is therefore particularly important, especially in childhood, because in a number of chronic diseases, repeated X-rays of the hand are necessary for the assessment of bone age and monitoring of skeletal development [11, 12].
The purpose of this study was to investigate the possibility of assessing bone age using less invasive method such as dual energy X-ray absorptiometry (DXA), typically utilized for diagnosis and monitoring of bone disorders .
The methods were used in healthy children and evaluated in subjects with altered maturation, resulting from multihormonal pituitary deficiency (MPD).
Materials and methods
The subjects investigated in our study were 60 Caucasian children aged 5–20 years (mean 12.2±3.7 years), including 30 females (11.8±3.9 years) and 30 males (12.5±3.4 years). Fifty participants were recruited in medical centers located in the Warsaw area, where they underwent routine medical care examination by pediatricians due to minor skeletal pain incidents, scoliosis or fracture (forearm) and afterwards sent to our hospital for further diagnosis. In all cases, chronic diseases were excluded, both body height and weight were within the 10th and 90th percentiles for a normal age-related population, and general health was sufficiently good to treat this group as healthy.
Patients with MPD (n=10) were recruited from Children’s Memorial Health Institute. The patients had maximal stimulated GH levels <10 µg/l and/or reduced spontaneous GH secretion for 24 h (<3 µg/l) at the onset of the disease. The treatment regimens for MPD subjects varied concerning the number of GH injections per week (6–7) and GH dosage (0.5–0.7 IU/kg per week), but were not considered to be relevant for our study and in consequence were disregarded. Ethical approval was obtained for the study from Ethical Committee of Children’s Memorial Health Institute. All subjects or their parents gave their informed consent to participate.
Assessment of bone age—reference method
The estimation of skeletal maturity was done on the basis of 60 hand-wrist radiographs of the left hand of recruited children (30 females, 30 males).
The radiographs were performed in different medical centers in the Warsaw area, with various exposure conditions (FFD 100–105 cm, tube voltage 50 kV or 60 kV, 400 mA and 0.8 s or 300 mA and 0.5 s) and various entrance surface doses (40–100 µSv). All radiographs, blinded to chronological age and subject’s name, were sent to two independent observers (O1, O2) for the evaluation of bone age. The results obtained by the other observer were also blinded. Before the final estimation of our subject’s bone age, a number of discussions were held to minimize inter- and intra-observer errors. The method of bone age evaluation was equivalent to the Greulich and Pyle method in terms of agreement with recent recommendations for interpretation of X-ray data according to Mora  and Groell . Each hand-wrist radiograph was compared to an original series of standard plates obtained from standard hand-wrist radiographs of Caucasian Polish girls and boys, published as a reference atlas of skeletal development .
The reference radiographs published in the atlas represent the average skeletal development for chronological age and gender. The bone age of all subjects was determined by identification of appearances, sizes and shapes of skeletal maturity indicators. Maturity indicators were visually analyzed by comparison with standards, and the age given to the standard that matched most closely with the evaluated radiograph was assigned as the bone age of the subject. When the appearances lay between two standards, a bone age between these was given . Skeletal age of 60 children was assessed with a precision of 0.5 years (according to the reference atlas). Bone age assessments done on the basis of hand-wrist radiographs were treated as reference data for comparison analysis to determine the potential of DXA derived images in skeletal age evaluations.
Assessment of bone age—evaluated method
Sixty left hand-wrist scans were performed using Expert-XL (GE Lunar, USA) densitometer. The resolution of hand-wrist image was 0.0025 cm2 as an effect of fan-beam technology.
Bone age of DXA hand-wrist scans was assessed by the same procedure as used in the reference method, and compared with results obtained from radiographs.
Evaluation of reproducibility of bone age assessments for each of two observers (O1, O2) was done on the basis of duplicate readings of 28 hand-wrist radiographs and 50 DXA hand scans.
The minimum interval between first and second skeletal age rating was 14 days. Precision of bone age assessments for each observer (O1, O2) was calculated as the root mean square coefficient of variation (CV) for the repeated measurements. The final bone age results assessed by O1 and O2 were compared and the significance of the differences was evaluated by paired t-tests. The intraclass correlation coefficients (ICC) were calculated to evaluate the inter-observer concordance between the readings of radiographs as well as DXA-derived hand scans, according to recommendations previously described [2, 15].
For the calculation of differences between bone age results estimated by reference radiography and evaluated DXA, as well as between bone age and chronological age, repeated measures ANOVA and Newman-Keuls multiple comparison tests were performed. The Pearson correlation analysis was performed to assess the concordance of results obtained by two methods.
All P-values <0.05 were considered significant. Results are presented as mean±SD (95% confidence interval).
Reproducibility of bone age readings based on radiographs and DXA derived hand scans
For 24 radiographs assessed by the first observer (O1) (85.7%) and 26 radiographs assessed by the second observer (O2) (92.9%), the repeated assessments gave the same bone age. In four cases assessed by O1 and 2 cases assessed by O2, differences equal to 0.5 years were observed.
The percentages of identical bone age ratings on duplicate assessments of 50 DXA images were 78% (O1) and 84% (O2). The bone age results of 11 assessments (22%) performed for the second time by the first observer (O1) differed in comparison with the first examination by 1 year (in two cases) and by 0.5 years (in nine cases). The difference between the first and the second reading done by O2 was not higher than 0.5 years (eight cases).
Each of those six radiographs and 19 hand-wrist scans were re-evaluated for the estimation of bone age, and the result of that examination was treated as the final bone age estimation.
The precision errors expressed as CV%, calculated on the basis of duplicate bone age readings based on 28 radiographs and 50 DXA derived hand scans, were 0.92% and 0.88% for O1 and 0.84% and 0.78% for the second observer, respectively.
Inter-observer variability of bone age readings based on radiographs and DXA derived hand scans
Bone age values assessed by the two observers showed a high degree of agreement. The means of differences between duplicate bone age readings based on radiographs and DXA hand scans were 0.02 years (−0.04–0.08, 95%CI, P=0.573) and 0.05 years (−0.03–0.13, 95%CI, P=0.184), 0.07 years (−0.01–0.15, 95%CI, P=0.103) and −0.03 years (−0.14–0.08, 95%CI, P=0.536), 0.04 years (−0.01–0.09, 95%CI, P=0.096) and 0.07 years (−0.06–0.07, 95%CI, P=0.799), for girls, boys and whole group of children, respectively. The intraclass correlation coefficients (ICC) computed for bone age readings of radiographs and DXA scans separately for girls and boys were r=0.999 and r=0.998, r=0.998 and r=0.997, respectively. Because the bone age values assessed by the two observers were highly concordant, the means between the bone age assessed by O1 and O2 were used for further analyses.
Assessment of skeletal maturity on the basis of DXA derived hand scans
Chronological age (CA) and bone age (BA) assessed on the basis of radiographs and DXA hand scans, and anthropometrical characteristics of studied girls. MPD indicates children with multihormonal pituitary deficiency
Healthy girls (n=24)
MPD girls (n=6)
BA based on hand-wrist radiographs (years)
BA based on DXA hand scans (years)
Chronological age (CA) and bone age (BA) assessed on the basis of radiographs and DXA hand scans, and anthropometrical characteristics of studied boys. MPD indicates children with multihormonal pituitary deficiency; NS not significant
Healthy boys (n=26)
MPD boys (n=4)
BA based on hand-wrist radiographs (years)
9.7 (5.9) NS
BA based on DXA hand scans (years)
9.6 (5.9) NS
Individual results of multihormonal pituitary deficiency subjects. MPD indicates children with multihormonal pituitary deficiency
MPD girls (n=6)
MPD boys (n=4)
Bone age based on hand-wrist radiographs (years)
Bone age based on DXA hand scans (years)
The maximum difference between duplicate bone age ratings for individual subjects was not higher than 0.5 years. In 44 cases (73.3%), bone age values estimated on the basis of DXA hand-wrist scan were the same as assessed by the reference method. Lower bone age values based on DXA scans were identified in nine subjects (15.0%) and higher in seven subjects (11.6%) in comparison to bone age assessed on the basis of radiographs. The results of paired t-tests indicated no significant difference between bone age ratings based on radiographs and DXA derived hand-wrist scans (P>0.05).
Bone age values estimated by both methods were highly correlated (r=0.998; r2=0.996; P<0.0001; 95% confidence interval from 0.996 to 0.998), indicating the similar potential of DXA hand-wrist scans and radiographs for bone age assessments.
In several cases, substantial differences (at least 1 year) between bone age and chronological age were noticed. Delayed bone age was observed in eight female subjects, including six MPD and two healthy cases. In the group of boys, delayed bone age was observed in three of six MPD subjects and five healthy ones. The increased rate of skeletal maturation was noticed in 24 healthy subjects (13 females). The means and significance of differences between bone age and chronological age are presented in Tables 1 and 2 for girls and boys, respectively. A higher difference between means of bone age and chronological age was observed in the group of healthy girls (0.9 years, P<0.001), than in boys (0.5 years, P<0.05).
Skeletal maturation, the process of transformation of the skeleton from cartilage into bone tissue, is regulated by endocrine, genetic and nutritional factors and can be slowed or accelerated by systemic disease. Moreover, in healthy children of the same chronological age, the grade of skeletal maturation, as well as pubertal development, may be different.
Skeletal age can be assessed even in newborn children, while Tanner stages are useful only during the adolescent years. Different techniques for estimation of skeletal maturity have been available since last century. Bone age may be assessed at various skeletal sites using more or less subjective methods. In spite of this, all of them are based on traditional radiographs.
Therefore, the aim of this study was to evaluate the possibility of bone age determinations using less invasive DXA technology with standardized imaging procedures.
The method of bone age determination used in this study was similar to the Greulich and Pyle  method, with modifications related to the original standards . Several standards have been developed for Japanese, German and Scandinavian populations [16, 17, 18]. In our study, bone age was assessed according to standards obtained between 1960 and 1963, and published by Kopczynska-Sikorska in the Atlas of Hand-Wrist Bone Development of White Polish Children .
In this study, both hand-wrist radiographs and DXA hand-wrist scans were compared visually to original standards. Because the method was based on the visual comparison of ossification centers in carpal bones and epiphyses of tubular bones including proximal, middle, and distal phalanges, as well as radius and ulna (Fig. 1), the selection of the proper standard was inevitably subjective. Despite the fact that the assessment was additionally affected by the quality of published standards, the reproducibility of duplicate assessments was high and evaluation of interobserver variability showed a high degree of consistency between the two independent observers.
It should be noted that hand-wrist radiographs were carried out under various conditions, i.e. exposition values, distance between the radiation source and the object and localization of the central beam. Nevertheless, the reproducibility of bone age rating reached in our study is in agreement with data published by Cameron and others [19, 20, 21].
The problems with unification of bone age assessment procedure have been discussed in the literature [3, 18, 19, 20, 21], with reproducibility errors comparable for both classical methods and evaluated in this study. Despite the wide experience of laboratories specialized in assessment of the skeletal age, the results include both measurement and interpretation errors .
In our study, bone age results established on the basis of left hand-wrist radiographs (treated as the reference) and left hand-wrist DXA scans performed in parallel in 60 children showed a high correlation.
The lack of significant differences between bone age assessments done by the same rating method but based on radiographs or DXA derived images suggests that the quality of DXA hand-wrist scans is sufficient for visual determination of skeletal age, at least in our study conditions. Moreover, the evaluated DXA method shows advantages over classical radiography because of sufficient resolution of the fan-beam images, standardized measurement conditions eliminating repeats due to incorrect exposures or low image quality, with much lower radiation dose (up to 8 µSv).
Surprisingly, in a number of studied healthy subjects, increased bone age was observed. The phenomenon is probably due to secular trends of earlier maturation, predominantly linked to rapid improvement of socioeconomic conditions in Poland. In addition, differences of this magnitude may be the consequence of some limitations of this study. First, the study was performed on a relatively small group of children. Second, the study population was enrolled only from Warsaw city area, where socioeconomic status is thought to be higher than in the country as a whole. Third, we did not verify whether the pubertal development of studied healthy children was normal, according to Tanner stage standards. The data published in the study of applicability of the Greulich and Pyle standards showed no significant differences between chronological and bone ages of healthy American children. However, the range of differences between bone age and chronological age in healthy children up to 2.62 years when grouped by stage of sexual development indicated marked individual variations in prepubertal, pubertal and postpubertal girls and boys. Concerning the apparent differences between the generations of 1960s and 2000s, there is a strong need to establish actual standards of skeletal maturation [4, 22].
The skeletal age estimations using DXA hand-wrist scans are very promising at a time when increasing numbers of fan-beam DXA devices are being installed. The feasibility of bone age estimation based on DXA derived hand scans with the use of the Hologic 4500A machine has already been confirmed . Future pediatric densitometry methods should include computer-assisted bone age rating along with the BMD measurement allowing increase in objectivity of diagnosis. This idea should be considered by DXA machine manufactures.
Evaluation of skeletal maturity using DXA images appears to be simple, easy, and non-invasive, giving results comparable to the classical method. Nevertheless, radiographic images appear to be still the method of choice in cases when bone age determination needs to be more precise than 0.5 years.
The authors would like address special thanks to Piotr Chądzyński, Joanna Marowska, Kenneth G. Faulkner and Thomas J. Beck for their valuable suggestions.