Subjects and methods

SUBJECTS

Schoolchildren from five schools in Copenhagen participated in the study. A letter of information was given to their parents, who completed an informed consent form before examination. Adolescents aged 18 years or older completed the consent form themselves. The study was approved by the ethics committee for Copenhagen and Frederiksberg. Among those who were approached, 219 (27%) of 798 girls and 155 (22%) of 701 boys agreed to participate. For this analysis, 31 were excluded because of non-white origin (n=17), chronic disease (n=8), medication that might affect bone mineralisation (n=2), incomplete DXA scanning (n=2) or a large metal ornament, that could not be removed, interfering with the DXA results (n=2). The remaining 201 females (age 4.9–19.3 years) and 142 males (age 5.6–19.1 years) participated in the study. The sex and age specific medians and ranges of height and weight are shown in table 1.

ANTHROPOMETRY

Height and weight were measured before DXA scanning. Height was determined to the nearest 1 mm using a stadiometer. Weight was measured to the nearest 0.1 kg using a digital electronic instrument. Subjects wore only pants and a cotton T-shirt when weighed.

BONE MINERAL ASSESSMENT

Whole body BMC measured in g hydroxyapatite and bone size expressed as anterior-posterior projected bone area measured in cm² was determined by DXA scanning using a Hologic 1000/W (Hologic Inc, Waltham, MA). For analysis, software version 5.61 was used. Subjects wore only pants and a cotton T-shirt during the scan. For quality control, spine phantoms were scanned daily. The coefficient of variation for these BMC and bone area measurements on a spine phantom over a period of two years (n=358) was 0.37% and 0.28%, respectively. The entrance radiation dose level was 15 μSv, with an effective dose not more than 10 μSv, equal to about one day’s background radiation in Denmark.

STATISTICAL METHODS

The weight for height and the height for age distribution of the individuals was compared graphically with a Danish sex specific standard.21 The fractions of the data in each of the six groups delimited by the 10th, 25th, 50th, 75th, and 90th centile in the weight for height standard were compared with the theoretical probabilities by a χ² test with five degrees of freedom. The height for age SD score (Z score) was symmetrically distributed when plotted against age in both sexes (data not shown) and the sex dependent overall mean and SD of the Z scores were calculated.

Smooth cross sectional centile curves for bone area and BMC versus age, bone area versus height, and BMC versus bone area were derived using the LMS method.22 This summarises the distribution of the dependent variable given the covariate by a Box-Cox power (L) depending on the skewness of the distribution, the median (M), and coefficient of variation (S) giving smooth curves for the dependence of L, M, and S on the covariate. The present data seemed to be fairly symmetrically distributed given the value of the covariate. Therefore, the value of L was fixed to 1 corresponding to the normal distribution, while the curves of M and S were estimated. The Z score of an observed value of for example, BMC may be calculated in the usual way as Z = (BMC-mean)/SD using the mean (mean = M) and SD (SD = M*S) based on the estimated M and S curves. The ranges of the covariates were reduced in order to eliminate edge effects due to single extreme values. Thus we considered the age interval 6–19 years (both sexes), height intervals 120–181 cm (girls) and 120–190 cm (boys), and bone area intervals 700–2400 cm² (girls) and 700–2500 cm² (boys).

To test whether BMD calculated as BMC/bone area provides an adequate correction of BMC for bone area, the regression coefficients of BMC on bone area (both log transformed), were compared to 1 by Wald’s test. Confidence intervals for parameter estimates were based on Wald’s test.

Results

Both girls and boys were heavier for height and heigher for age than the present Danish standards based on data from 1974.21 The numbers of individuals in each of the six groups delimited by the 10th, 25th, 50th, 75th, 90th centile in the weight for height standard were 18, 21, 37, 46, 48, 31 and 15, 22, 32, 33, 20, 20, for girls and boys, respectively. The difference from the expected distribution was significant for the girls (χ²(5) = 23.3, p<0.001) but not the boys (χ²(5) = 3.0, p>0.2). The mean (SD) of the height for age Z scores was 0.35 (0.99) for girls and 0.45 (1.04) for boys, both significantly different from zero (both p<0.001).

The sex specific age dependent centile curves for bone area (fig 1) and BMC (fig 2) showed a large variation around the median. The curves were very similar for the two sexes from 6 to 11 years of age for both bone area (fig 1C) and BMC (fig 2C). From 11 to 15 years of age the BMC for age curves were slightly higher for the girls than for the boys (fig2C). However, the main sex difference was an earlier flattening of the girls’ curves around the age of 15 years for both bone area (fig 1C) and BMC (fig 2C). The bone area and BMC curves were of similar shape except that bone area was flatter at the top end. Thus there was a distinct flattening of the bone area curve for boys after the age of 16 years, whereas for BMC the deflection was less evident. At 18.5 years, the median bone area was 12% higher in boys than in girls (2335 cm² v 2083 cm²), and the median BMC was 21% higher in boys than in girls (2698 g v 2238 g). The estimated mean and SD values by sex and age appear in table 2for bone area and in table 3 for BMC.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Centile curves for whole body bone area according to age in girls (A) and boys (B). The solid curves are the median and the 3rd and 97th centiles of the normal distribution with mean and SD as shown in table 2. The dotted lines show the interquartile ranges. The median and outer centiles for girls (dashed lines) and boys (solid lines) are shown together in (C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Centile curves for whole body BMC according to age in girls (A) and boys (B).The solid curves are the median and the 3rd and 97th centiles of the normal distribution with mean and SD as shown in table 3. The dotted lines show the interquartile ranges. The median and outer centiles for girls (dashed lines) and boys (solid lines) are shown together in (C).

The sex specific height dependent centile curves for bone area (fig 3) showed that bone area was much closer associated with height than with age for both girls (figs 1A and 3A) and boys (figs 1B and 3B). Bone area for height was higher in girls than boys (fig 3C). Table 4 shows the estimated mean and SD values for bone area for height by sex.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Centile curves for whole body bone area according to height in girls (A) and boys (B).The solid curves are the median and the 3rd and 97th centiles of the normal distribution with mean and SD as shown in table 4. The dotted lines show the interquartile ranges. The median and outer centiles for girls (dashed lines) and boys (solid lines) are shown together in (C).

The sex specific bone area dependent centile curves for BMC (fig 4) showed a very close association between BMC and bone area for both girls and boys, especially for the smaller values of bone area. For bone area above 1500 cm², BMC for bone area was higher for girls than boys (fig 4C). Table 5 shows the estimated mean and SD values of BMC for bone area by sex.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Centile curves for whole body BMC according to bone area in girls (A) and boys (B). The solid curves are the median and the 3rd and 97th centiles of the normal distribution with mean and SD as shown in table 5. The dotted lines show the interquartile ranges. The median and outer centiles for girls (dashed lines) and boys (solid lines) are shown together in (C).

The regression coefficient of BMC on bone area was 1.43 (95% confidence interval (CI) 1.41 to 1.46) for the girls and 1.37 (95% CI 1.34 to 1.40) for the boys.

Discussion

We have presented centile curves of whole body bone area and whole body BMC for healthy school age children. We have chosen not to use BMD calculated as BMC/bone area because BMD measured with anterior-posterior osteodensitometry represents a mixture of true density and skeletal size.20 23 24 Note that the regression coefficient for log transformed BMC on log transformed bone area may be interpreted as a power for bone area. In the present data, the regression coefficient was highly significantly different from 1, indicating that the use of BMD would only partially adjust for bone area.

We suggest that our data are used as a reference for assessing bones of children for clinical or research purposes. To our knowledge it is the most comprehensive data on whole body measurements available at present, but there are some limitations to the use of this data set as a reference. Due to the limited number of children in each age and sex group, we cannot evaluate the tail probabilities of the distributions, indicated by the outer centiles in figs 1-4. Only 25% of the children approached agreed to participate. We have no data to compare the social background of participating and non-participating children, but it is our impression that there were no major differences. In a study that also measured weight and body composition, we had expected that overweight children would be less inclined to participate, that is that overweight children would be under-represented. The weight for height distribution indicated that this was probably not the case. The children were heigher for age and heavier for height compared with the reference data of Andersen et al,21 in accordance with the general pattern that both the height of children and the number of overweight children has increased in Scandinavia.25-27 The sex specific height for age and weight for height standards of Andersen et al are 20 years old, but they are the only Danish standards available.

Three studies have previously reported data on children’s whole body BMC measured with the Hologic 1000/W. Lloyd et al found a median BMC of 1218 g (range 799–2083 g) and an average BMC of 1276 g (SEM 26 g) in American girls whose median age was 11.9 years (range 10.7–13.3 years).28 Considering the girls of the same age range in our data (median age 12.2 years), the median BMC was 1300 g (range 779–2080 g), and the average BMC was 1352 g (SEM 46 g). The difference in average BMC was not significant, and the direction was in accordance with the slightly higher median age in our data. Faulkneret al published data for whole body BMC in Canadian girls and boys aged 8–16 years.16 Our age specific averages of whole body BMC were within 267 g below and 173 g above the average values of Faulkner et al, with no systematic age dependent difference, and the individual differences were not significant (Wald’s test with Bonferroni correction). Hannan et alreported regression models for the dependence of BMC on age alone and on age, height, and weight simultaneously in 11–18 year old Scottish girls.17 We have re-estimated the coefficients in their regression models based on those of our girls aged 11–18 years who were within the ranges of height, weight, and body mass index of their data (n=115). For the model consisting only of a second degree polynomial in age, our coefficients were not significantly different from those of Hannan et al, whereas for the model including weight and height, the two sets of coefficients were significantly different. Thus the relation between BMC and age may not be very different between countries, partly because of the large variation in BMC at each age.

The lower trend in whole body BMC with age in girls during the years after puberty in our study corresponded with other cross sectional studies of whole body BMC6 8 17 and lumbar BMC.7 It has previously been shown that lumbar BMC (L2–L4) in boys, in contrast to girls, continues to rise after puberty.7 29 In a cross sectional study (age 4–26 years), Ogle et al found that whole body BMC increased until the late teens with a flattening after 17.4 years in males while peak BMC was attained at 15.7 years in females.8 This is in agreement with our results for girls, while the flattening for boys was less clear, possibly because of our narrower age range. In a cross sectional study, Mosekilde found sex differences in vertebral bone growth in adults, as men but not women have a continued periostal bone growth.29 The modest increase in BMC after puberty in girls indicates that prepuberty and puberty, especially in girls, are important periods for achieving optimal peak bone mass.

Bonjour et al have previously published data on lumbar bone area by age in boys and girls aged 9–18 years.30 To our knowledge, data on whole body bone area have not been published previously. The pattern of bone area by age was similar to that of BMC in both girls and boys, except for a more pronounced levelling off of the bone area curves after puberty. Apparently, bone area peaks earlier than BMC in both girls and boys. The narrowing of the centiles for higher ages for girls may be an artifact due to edge effects.

The BMC for bone area curves were fairly similar for the two sexes, with little variation in BMC for given bone area. Thus it seems that the differences in BMC in these healthy children to a large extent is due to differences in bone size.

In clinical paediatric practice, it is insufficient to express BMC according to age alone, both because of the considerable variation in bone and body size for a given age, and because chronic disease affecting bone mineralisation often affects body and bone size as well. Several studies have shown that it is unsatisfactory and might lead to misinterpretation if BMC is expressed according to age only. Atkinsonet al found that preterm infants had lower BMC growth than term infants at the same gestational age, but when corrected for weight, there was no difference.31 Issenman et al found that bone density (radius) in children with Crohn’s disease was less abnormal when compared with height matched references than when compared with age matched references.32 Younget al found a lower bone density in the arm in both ballet dancers and girls with anorexia compared with normal girls of the same age, but the difference disappeared when they corrected for weight.33 Russell-Aulet et al found that the difference in BMC between white and Asian women could be explained by body size.34

The centile curves presented here can be used to validate whether a reduced whole body BMC is primarily due to a reduction in bone size or primarily due to a reduction in size adjusted BMC. A reduced bone mass may be due to short bones, narrow bones, or light bones, possibly with different health implications. We, therefore, recommend that evaluation of whole body BMC for children be separated into three parts as outlined in the appendix: (1) Is the child’s height appropriate for age? (‘short bones’); (2) Is the bone size (bone area) appropriate for height? (‘narrow bones’); (3) Is the BMC appropriate for bone area? (‘light bones’). For daily clinical use, an individual may be plotted on centile charts like figs 1-4. Single copies of large centile charts may be obtained from the authors. If a more accurate centile or Z score is required, the Z score may be calculated in the usual way using the estimated mean and SD values presented in tables2-5. An example of calculation of Z scores and centiles is presented in the , evaluating the BMC of a young man with galactosaemia and a girl with myositis treated with corticosteroid. Note, the latter example shows that a normal BMC for age does not preclude an abnormally low BMC for bone area, if the bone size of the individual is above average.

In conclusion, BMC depends on bone length, bone width, and bone density. The health implication of a reduced value is likely to depend on which of the three factors is responsible for the reduction. Therefore, it may be valuable to distinguish between these three factors, both when bone mass is assessed, and when the effect on BMC of some independent variables like nutrition, physical activity, etc is examined.