Measured and predicted bone mineral content in

Acta Pædiatr 87: 244–9. 1998

Measured and predicted bone mineral content in healthy boys and girls aged 6–18 years: adjustment for body size and puberty JT Warner 1, FJ Cowan 2, FDJ Dunstan 3, WD Evans 4, DKH Webb 5 and JW Gregory 1 Departments of Child Health 1 and Medical Statistics & Computing 3, University of Wales College of Medicine, Cardiff, Departments of Child Health 2 and Medical Physics 4, University Hospital of Wales Healthcare NHS Trust, Cardiff, and Department of Child Health 5, Llandough Hospital NHS Trust, Cardiff, UK

Warner JT, Cowan FJ, Dunstan FDJ, Evans WD, Webb DKH, Gregory JW. Measured and predicted bone mineral content in healthy boys and girls aged 6–18 years: adjustment for body size and puberty. Acta Pædiatr 1998; 87: 244–9. Stockholm. ISSN 0803–5253 Dual-energy X-ray absorptiometry (DEXA) is a rapid and precise technique for the assessment of bone mineralization in children. Interpretation of the results in growing children is complex as results are influenced by age, body size (height and weight) and puberty. Conventionally, bone mineral data derived from DEXA have been presented as an areal density [BMD; bone mineral content (BMC, g)/projected bone area (BA, cm 2)], yet this fails to account for changes in BMC that result from changes in age, body size or pubertal development. Measurement of BMC and BA of the whole body, lumbar spine and left hip were made in 58 healthy boys and girls using DEXA. The relationship between BMC and BA was curvilinear, with the best fit being that of a power model (BMD ¼ BMC=BAl , where l is the exponent to which BA is raised in order to remove its influence on BMC). The value of l changed when measures of body size and puberty were taken into account (e.g. for lumbar spine from 1.66 to 1.49). Predictive formulae for BMC were produced using regression analysis and based on the variables of age, body size and pubertal development. This provides a method for interpreting the measured BMC which is independent of such variables and a constant reference range for children aged 6–18 y. ` Bone area, bone mineral content, dual-energy X-ray absorptiometry, pubertal stage JT Warner, University of Wales College of Medicine, Heath Park, Cardiff CF4 4XN, UK

There is increasing clinical interest in the measurement of bone mineral density (BMD) in childhood, particularly in the assessment of the adverse effects of disease and also because of the theoretical risk that osteopenia in childhood may predispose to osteopenia and osteoporosis in later adult life. Dual-energy X-ray absorptiometry (DEXA) as a method of assessment has been widely used in adults, but to a lesser extent in children. DEXA is a relatively noninvasive technique for the assessment of bone mineral content (BMC) and BMD and is dependent on the differential attenuation of X-rays of two different energies passing through bone and soft tissue. The precision of the technique (less than 1% for the lumbar spine), the minimal doses of radiation involved (6 mSv, equivalent to 1 d background radiation), and the minimal co-operation required of patients undergoing these measurements makes the method acceptable for children (1). Interpretation of the results in growing children is problematical since bone mineral accretion during childhood is dependent on, and highly correlated with, growth and puberty, the interaction between these variables being complex (2–6). To provide a standardization for differences in bone size, BMD is expressed conventionally as a ratio of the measured BMC and the projected bone area (BA). This is not a true volumetric density but an areal density, as antero–posterior diameter of the bone is not measured. q Scandinavian University Press 1998. ISSN 0803-5253

Furthermore, areal BMD is unable to distinguish between osseous and non-osseous areas within the bone envelope. Mathematical models to derive an apparent volumetric density based on areal measurement have been explored (7, 8). Expression of apparent volumetric density appears to remove the influence of age, but still makes assumptions about bone shape. Mathematically, the expression of biological data as a ratio assumes a linear relationship between the variables in question, with a c-intercept equal to zero (y ¼ mx þ c). In reality, the assumption of a zero intercept is rarely satisfied in biological research and the inaccuracy of the ratio method for the expression of biological data dependent on size has been reported elsewhere (9). The expression of bone mineral data as an areal density also makes the assumption that BMC and BA are proportionally related, i.e. a 1% change in BA is associated with a 1% change in BMC. Recently, Prentice et al. have shown that this is not the case and suggested that the use of areal BMD should be discontinued in epidemiological research (10). If areal BMD adequately adjusts BMC for BA then BMD should be minimally correlated with BA. Furthermore, the power coefficient derived by regressing BMC on BA after transforming the variables to natural logarithms should be close to, and not significantly different from unity (10). However, if the relationship between BMC and BA is not

Bone mineral content in normal children

ACTA PÆDIATR 87 (1998)

directly proportional for children of different ages and body size (height and weight), variation in BMD will occur due to differences in bone size between individuals rather than as a consequence of real variation in bone mineralization. Comparison of bone mineral data between children of different ages, heights, weights and pubertal stages requires adjustment or scaling in order to correct for these differences. A method of interpretation is provided by the relationship between measured BMC and that of a predicted value based on the variables that influence it. This report examines the relationships between BMC and the variables of BA, age, height, weight, pubertal stage and gender in a healthy cohort of children. Predictive formulae are derived for BMC based on these variables in this reference population. Measured BMC can then be expressed as a percentage of a predicted value (%BMC). This allows derivation of a mean and standard deviation %BMC for a reference population which is independent of age, height, weight, pubertal stage and gender. This reference range is constant over the age range used in this study, a provision that cannot be overemphasized in paediatric populations. The model may be used to interpret the bone mineralization of other children, including those suffering from various disease states.

Subjects and methods Measurements were performed on 58 (29M, 29F) healthy children. These comprised 31 (17M, 14F) siblings of longterm survivors of childhood malignancy who were participating in a prospective study into late effects following treatment for childhood cancer and 27 (12M, 15F) relatives of staff within the authors’ department and volunteers from a local school. All children were healthy and well at the time of measurement, and none participated in excessive sporting activity. This study was undertaken in order to establish reference ranges for BMC in normally growing children, so as to allow interpretation of measurement of BMC in children with disease states. Informed consent was obtained from the parents and child and full ethical approval was granted from the ethics committee for South Glamorgan Health Authority. Anthropometry Height (to the nearest 0.1 cm) and weight (to the nearest 0.1 kg) were measured using a wall-mounted Harpenden stadiometer (Holtain, Crymych, Dyfed, UK) and Avery beam balance (Avery, Birmingham, UK), respectively. Standard deviation scores (SDS) for each measurement were calculated from the 1990 British reference standards (11). Puberty staging of pubic hair was assessed according to Tanner’s classification (12).

245

whole skeleton, lumbar spine and the left hip using the Hologic QDR 1000/W bone densitometer (Hologic, Waltham, MA, USA). The analysis is based on the differential attenuation of collimated X-rays, the mean energy of which is rapidly alternated by switching the high voltage of the X-ray tube between 70 and 140 kVp. The attenuation data are converted to BMC by comparison to known bone mineral and soft tissue equivalent standards mounted on a calibration wheel through which the radiation passes (13). Hologic software version V5.67 was used for the whole body analysis and V4.66P for the lumbar spine and left hip. Prior to patient scanning a lumbar spine phantom, supplied by the manufacturer, with known BMC was measured daily. Each child wore light clothing and removed any objects containing metal prior to scanning. Whole body scanning was performed with the child lying supine along the long axis of the DEXA couch. Lumbar vertebrae (L1–L4) were measured with the legs flexed at the hip by resting them on a box supplied by the manufacturer. This reduces the physiological spinal lordosis and aligns the disc spaces with the X-ray beam to improve the separation of individual vertebrae on the image. Positioning of the leg for measurement of the hip is of great importance to obtain a high level of precision. This was achieved by internally rotating the left leg at the hip through an angle of 25–308 and strapping the foot in a footholder supplied by the manufacturer. The Hologic densitometer allows analysis of different sites within the total hip, which can be subdivided into femoral neck, trochanteric and intertrochanteric regions. These regions have been selected to represent critical points in the proximal femur where fractures occur. The location of each region is based on anatomical markers which are defined by the system software to ensure that the same location on the femur is evaluated (13).

Statistical analyses Power coefficients for the relationship between BMC and BA for each anatomical region were obtained by linear regression after transformation of the variables to natural logarithms. Multiple regression analysis was used to create predictive formulae for BMC using the explanatory variables of BA, age, height, weight, pubertal status and gender (after conversion of the continuous variables to natural logarithms). Those variables not reaching significance at the 5% level were discarded in a backward stepwise manner. The measured BMC was expressed as a percentage of the predicted BMC obtained from the predictive formula (%BMC). All statistical analyses were carried out using the SPSS for Windows software program, version 6.0.

Results Dual-energy X-ray absorptiometry DEXA measurements of BMC and BA were made for the

The precision (coefficient of variation) of repeat measurements for the lumbar spine phantom was 0.34% for

246

JT Warner et al.


Table 2. Power coefficients for the relationship between BMC and BA at the different skeletal sites.

Table 1. Anthropometric data.

Age (y) Height (cm) Height SDS Weight (kg) Weight SDS

Mean (SD)

Median (range)

12.9 (3.3) 153.4 (16.6) 0.26 (0.91) 48.9 (19.6) 0.41 (1.04)

12.9 (6.2 to 17.3) 153.8 (113.4 to 182.4) 0.20 (–1.69 to 1.82) 48.0 (19.5 to 121.4) 0.21 (–1.57 to 4.42)

Whole body Lumbar spine Total hip Femoral neck Trochanteric region Intertrochanteric region

Power coefficient

SE

1.39 1.66 1.44 1.63 1.26 1.42

0.02 0.06 0.06 0.11 0.05 0.07

All values were significantly different from 1.0 ( p , 0:01).

measurements taken daily over a period of 2 y during which the study was performed. Patient details are shown in Table 1: 22 children were prepubertal, 6 were at pubertal stage 2, 7 were at stage 3, 4 were at stage 4 and 19 were at stage 5. Expression of data as areal BMD (BMC/BA) failed to remove the influence of BA since BMD remained correlated with BA, r ¼ 0:81 ( p , 0:001) for the lumbar spine, r ¼ 0:91 ( p , 0:001) for the whole body and r ¼ 0:69 ( p , 0:001) for the total hip. BMC was related to BA in a curvilinear fashion which, after conversion of the variables to natural logarithms, became linear (Fig. 1). The power coefficients derived by regressing ln(BMC) on ln(BA) at each anatomical region shown in Table 2 were

all significantly different from 1.0 (p , 0:01). These findings suggested that a more appropriate model for BMD would be BMC/BA l, where l is the power coefficient to which BA is raised in order to remove its influence on BMC. Use of this model removes the influence of BA on BMD, r ¼ 0:03 for the lumbar spine, r ¼ 0:03 for the whole body and r ¼ 0:02 for the total hip. After inclusion of, age, height, pubertal stage and gender into the regression model, the relationship between BMC and BA changed. Table 3 shows the partial power coefficients for the relationship between BMC and BA after adjustment for these variables. For the whole body and lumbar spine the power coefficients remained significantly different from 1.0 ( p , 0:01), but close to, and not significantly different from 1.5. For the total hip and its subdivisions the power coefficients were no longer significantly different from 1.0 (Table 3). The predictive formulae for ln(BMC), including all significant variables, for the various sites are shown in Table 4. Subanalysis revealed no significant differences in the predictive formulae between prepubertal (n ¼ 22) and postpubertal (n ¼ 23; pubertal stage 4 and 5) children. The mean and SD %BMC (measured BMC × 100/predicted BMC) for the whole population at each site is shown in Table 5 and allows calculation of SDS (see worked example). Figure 2 shows the distribution of %BMC at the lumbar spine with age, demonstrating that after allowance for the variables that influence it, the %BMC is no longer dependent on age and therefore allows comparison of %BMC to made from individuals across the age range of children within this study.

Table 3. Partial power coefficients for the relationship between BMC and BA after adjustment for age, body size and pubertal stage.

Fig. 1. Relationship between BMC and BA at the lumbar spine. (A) Curvilinear relationship between BMC and BA before conversion to natural logarithms; (B) linear relationship between BMC and BA after conversion to natural logarithms (B = male, W = female).

Whole body Lumbar spine Total hip Femoral neck Trochanteric region Intertrochanteric region a

Power coefficient

SE

1.49 a 1.49 a 0.93 0.97 0.95 0.92

0.07 0.18 0.13 0.18 0.10 0.09

Significantly different from 1.0 ( p , 0:01).



247

Table 4. Predictive models for ln(BMC) at each anatomical region. Whole body: (R2 ¼ 0:99; p , 0:0001) ln(BMC) = (1.488 × ln(BA)) + (0.145 × ln(age)) – (0.652 × ln(height)) – 0.764 Lumbar spine: (R2 ¼ 0:95; p , 0:0001) ln(BMC) = (1.489 × ln(BA)) + (0.172 × ln(weight)) – (0.977 × ln(height)) + (0.058 × PS) + 1.994 Total hip: (R2 ¼ 0:94; p , 0:0001) ln(BMC) = (0.931 × ln(BA)) + (0.264 × ln(weight)) + (0.033 × PS) – 1.049 Femoral neck: (R2 ¼ 0:86; p , 0:0001) ln(BMC) = (0.970 × ln(BA)) + (0.181 × ln(weight)) + (0.027 × PS) – 0.969 Trochanteric region: (R2 ¼ 0:94; p , 0:0001) ln(BMC) = (0.951 × ln(BA)) + (0.224 × ln(weight)) + (0.028 × PS) – 1.222 Intertrochanteric region: (R2 ¼ 0:94; p , 0:0001) ln(BMC) = (0.916 × ln(BA)) + (0.328 × ln(weight)) + (0.034 × PS) – 1.184 Age in decimal y, height in cm, weight in kg. PS: pubertal status (Tanner classification).

Fig. 2. Variation in %BMC at the lumbar spine with age (B = male, W = female.

Table 5. %BMC at each anatomical site.

Whole body Lumbar spine Total hip Femoral neck Trochanteric region Intertrochanteric region

Mean

SD

Range

100.10 100.10 100.38 100.43 100.38 100.70

4.62 9.35 10.33 9.71 11.77 10.73

90.91–110.65 79.96–130.73 76.18–122.78 73.54–118.10 72.61–126.43 76.44–125.24

Discussion Measurement of a true volumetric bone density requires a three-dimensional scan such as that obtained from computed tomography (14). This, however, is unacceptable in children because of the large radiation doses required. Mathematical models have been derived to produce an apparent bone volume based on an areal measurement but assume that the shape of the bone approximates to that of a cylinder (7, 8). Expression of BMD as an apparent volumetric density appears to remove the influence of age and body size (7, 15). Expression of the data as an areal BMD fails to remove the influence of BA since the variables remain correlated. If the data were not corrected for age, body size and pubertal

status, the exponent to which BA should be raised to remove its influence on BMD was significantly greater than 1.0 for all anatomical sites (Table 2). This means that the expression of BMC per unit projected area is not correct; BMD will be overestimated in large bones and underestimated in small ones. Expression of BMD as BMC/BA l, where l is the variable exponent to which BA is raised, would therefore seem a more appropriate model. However, allowance should be made for gender, age, body size and pubertal status since the value of l changed after including these variables in the model (Table 3), so that for the whole body and lumbar spine it became close to 1.5, whereas at the hip it was close to 1.0. The significance of a power of 1.5 is that it effectively transforms the area into a volume, thereby creating an apparent volumetric density whereas, at the hip, where the exponent is 1.0, the areal BMD would seem an appropriate adjustment provided that the other explanatory variables are taken into account. The reason for the apparent difference in the value for l at different sites is not clear but it only occurs after inclusion of the other explanatory variables. It may be due to differences in bone shape; the lumbar spine and whole body behave like a cylinder (allowing volumetric analysis), whereas the irregular shape of the hip (having different radiological properties) conforms with areal analysis. The use of predicted values for BMC (based on the variables that influence it) to interpret measured values allows comparison of values in subjects with a wide range of body sizes and ages, and does not include assumptions about the shape of the bone. The predictive formulae for BMC are based on measurements of the explanatory variables (age, height, weight, pubertal status and BA), the dependency on these variables being very high, as demonstrated by the R 2 values (Table 4). The explanatory variables entered into the multiple regression model to create the predictive formulae are themselves not independent of one another. The models created, however, were selected in a stepwise manner with the most significant explanatory variables remaining. It is accepted that other models could be created that would fit the data almost as well. For example, in the model for whole body BMC, if ln(age) is replaced with pubertal status the value for R 2 is almost

248

JT Warner et al.

identical (0.98923 against 0.98891), because pubertal status and age are highly correlated (r ¼ 0:90; p, 0:001). In the models presented weight was a significant predictor of BMC at all the anatomical regions, which may simply represent changes due to growth or may suggest that weight bearing has an independent influence on BMC. Height remained a significant predictor of BMC for the whole body and lumbar spine, whereas pubertal stage had significant influences at all regions except for the whole body, where age accounted for the variability. The influence of puberty on BMC has also been documented by others (4, 16). There were no significant gender differences at any site, once adjustment for body size and puberty had been made, in contrast to other studies (17). Regression analysis has been used previously to provide predictive formulae for BMC (18). Hannan et al. derived predictive equations for BMC based on weight, height and age for the whole body and lumbar spine in 216 Scottish girls aged 11.0–17.9 y (18). At the lumbar spine the predicted BMC calculated for girls from the current study (n ¼ 23) using the Scottish formulae were correlated (r ¼ 0:91; p , 0:001) and not significantly different from the predicted BMC calculated from the formulae in Table 4 [44.8 (16.2) vs 43.9 (12.3) g, respectively, p ¼ 0:55]. However, for the whole body, despite a significant correlation (r ¼ 0:95; p , 0:001), there was a small, significant difference [1814.5 (591.5) vs 1901.3 (621.0) g, respectively, p ¼ 0:04]. This small bias in the predicted BMC for the whole body when comparing Welsh with Scottish girls aged 11.0–17.9 y may be accounted for by a disparity in height, weight and body composition, or genetic influences on BMC of similarly aged children from the two regions. More recently, standards for BMC for the whole body have been published based on measurements in 343 Danish boys and girls aged 5–19 y (19). In this study, banded means and SD of BMC for age and BMC for BA allow calculation of SDS for BMC adjusted for gender, age and BA independently. To compare the %BMC derived from the equations in the current study with this much larger cohort, the %BMC for the whole body was converted into individual SDS using the means and SD in Table 5 and compared to the BMC SDS adjusted for age and for BA derived from the Danish study (19). There was no significant difference in the SDS between the two studies for the data adjusted for age [mean (95% CI) difference = 0.14 (–0.25 to 0.52) SDS, p ¼ 0:48], and a small significant difference when adjusted for BA [mean (95% CI) difference = 0.43 (0.23 to 0.63) SDS, p , 0:01]. This comparison highlights the need to adjust BMC simultaneously for all of the different variables that influence it. The advantage of the regression model used in the current study, albeit based on smaller numbers, is that it allows these factors to be taken into account in one equation and compares well to the published data from the two studies described above (18, 19). The mean %BMC for each site is close to 100%, which is to be expected since it is derived from this reference


population (Table 5). The SD at each site may be used to define a normal reference range for the interpretation of %BMC in children undergoing measurement for clinical reasons. For example, at the lumbar spine where the SD is nearly 9.5%, an individual with a %BMC of 81% or less (more than 2SD below the mean) would represent significant osteopenia (less than the 3rd centile compared to this reference population). There is much interest in the influence of various diseases and their treatments on bone mineral accretion (20–22). Interpretation of bone mineral data in children affected by such illnesses is complex, owing to abnormalities of linear growth, body size and pubertal development. The predictive formulae derived in this report, which take into account the influence of these variables, provide a method for interpreting measured values in such children. However, these formulae have been derived from measurements made in a group of healthy, normally growing children across a wide age range and their application to children with various disease states requires further validation. In conclusion, a mathematical model has been used in order to provide a predictive BMC taking into account the influence of the variables of age, body size and puberty. These formulae allow comparison to be made between individuals of differing age, height, weight and stages of puberty. The standard deviation for %BMC at each anatomical site provides a reference, which is constant over the age range used in this study, from which BMC may be interpreted in disease states.

Worked example A 16-y-old prepubertal male with Crohn’s disease was assessed for bone mineralization following a vertebral wedge fracture. DEXA scanning of his lumbar spine showed a BMC of 10.45 g for a projected BA of 34.79 cm 2. His height and weight were 140.1 cm and 33.7 kg, respectively. From Table 4: ln(BMC) = (1.489 × ln(34.79)) + (0.172 × ln(33.7)) – (0.977 × ln(140.1)) + 0.058 + 1.994 Predicted BMC = 22.49 g %BMC = (10.45 × 100)/22.49 = 46.5%. From Table 5, the SDS = (46.5 – 100.1)/9.35 = –5.74. One year later, after treatment with an elemental diet and testosterone to induce puberty, his height was 149.5 cm, weight 42.4 kg, lumbar spine BMC and BA 16.41 g and 35.36 cm 2, respectively. He had progressed to pubertal stage 3. Predicted BMC = 25.27 g %BMC = 64.9% SDS = –3.76. Acknowledgments.—Dr JT Warner is supported by a grant from the local oncology charity, Llandough Aims to Treat Children with Cancer and Leukaemia with Hope (LATCH). We are grateful to Mr JH Pearce, Mr D Coleman and Mrs RJ Pettit for assistance with the DEXA measurements, and also to all the children and their parents who made this study possible.



References 1. Slosman DO, Rizzoli R, Buchs B, Piana F, Donath A, Bonjour J-P. Comparative study of the performance of X-ray and gadolinium 153 bone densitometers at the level of the spine, femoral neck and femoral shaft. Eur J Nucl Med 1990; 17: 3–9 2. Shaw NJ, Bishop NJ. Mineral accretion in growing bones–a framework for the future. Arch Dis Child 1995; 72: 177–9 3. Glastre C, Braillon P, David L, Cochat P, Meunier PJ, Delmas PD. Measurement of bone mineral content of the lumbar spine by dual energy X-ray absorptiometry in normal children: correlations with growth parameters. J Clin Endocrinol Metab 1990; 70: 1330–3 4. Southard RN, Morris JD, Mahan JD, Hayes JR, Torch MA, Sommer A, et al. Bone mass in healthy children: measurement with quantitative DXA. Radiology 1991; 179: 735–8 5. Davie MWJ, Haddaway MJ. Bone mineral content and density in healthy subjects and in osteogenesis imperfecta. Arch Dis Child 1994; 70: 331–4 6. De Schepper J, Derde MP, Van den Broeck M, Piepsz A, Jonckheer MH. Normative data for lumbar spine bone mineral content in children: influence of age, height, weight, and pubertal stage. J Nucl Med 1991; 32: 216–20 7. Lu PW, Cowell CT, Lloyd-Jones SA, Briody JN, Howman-Giles R. Volumetric bone mineral density in normal subjects, aged 5–27 years. J Clin Endocrinol Metab 1996; 81: 1586–90 8. Kro¨ger H, Kotaniemi A, Vaino P, Alhava E. Bone densitometry of the spine and femur in children by dual energy X-ray absorptiometry. Bone Miner 1992; 17: 75–85 9. Tanner JM. Fallacy of per-weight and per-surface area standards, and their relation to spurious correlation. J Appl Physiol 1949; 2: 1–15 10. Prentice A, Parsons TJ, Cole TJ. Uncritical use of bone mineral density in absorptiometry may lead to size-related artifacts in the identification of bone mineral determinants. Am J Clin Nutr 1994; 60: 837–42 11. Freeman JV, Cole TJ, Chinn S, Jones PRM, White EM, Preece MA.

12. 13. 14.

15. 16. 17.

18. 19. 20. 21. 22.

249

Cross sectional stature and weight reference curves for the UK, 1990. Arch Dis Child 1995; 73: 17–24 Tanner JM. Growth at adolescence. 2nd ed. Oxford: Blackwell, 1962 Wahner HW, Fogelman I. The evaluation of osteoporosis: dual energy X-ray absorptiometry in clinical practice. In Fogelman I, editor. Metabolic bone disease. London: Martin Dunitz, 1994 Mora S, Goodman WG, Loro ML, Roe TF, Sayre J, Gilsanz V. Agerelated changes in cortical and cancellous vertebral bone density in girls: assessment with quantitative CT. Am J Roentgenol 1994; 162: 405–9 Scho¨nau¨ E, Wentzlik U, Michalk D, Scheidhauer K, Klein K. Is there an increase of bone density in children? Lancet 1993; 342: 689–90 Gilsanz V, Gibbens DT, Roe TF, Carlson M, Senac MO, Boechat MI, et al. Vertebral bone density in children: effect of puberty. Radiology 1988; 166: 847–50 McCormick A, Ponder SW, Fawcett HD, Palmer JL. Spinal bone mineral density in 335 normal and obese children and adolescents: evidence for ethnic and sex differences. J Bone Miner Res 1991; 6: 507–13 Hannan WJ, Cowen SJ, Wrate RM, Barton J. Improved prediction of bone mineral content and density. Arch Dis Child 1995; 72: 147–9 Molgaard C, Thomsen BL, Prentice A, Cole TJ, Michaelsen KF. Whole body bone mineral content in healthy children and adolescents. Arch Dis Child 1997; 76: 9–15 Cowan FJ, Warner JT, Dunstan FDJ, Evans WD, Gregory JW, Jenkins HR. Inflammatory bowel disease and predisposition to osteopenia. Arch Dis Child 1997; 76: 325–9 Issenman RM, Atkinson SA, Radoja C, Fraher L. Longitudinal assessment of growth, mineral metabolism, and bone mass in pediatric Crohn’s disease. J Pediatr Gastroenterol Nutr 1993; 17: 401–6 Shore RM, Russell MD, Chesney W, Mazess RB, Rose PG, Bargman GJ. Bone mineral status in growth hormone deficiency. J Pediatr 1980; 96: 393–6

Received Apr. 21, 1997. Accepted Nov. 10, 1997