Total Body Bone Mineral Density in Young Children:Influence of Head Bone Mineral Density*

ARLENE TAYLOR,1 PATRICIA T. KONRAD,2 MICHAEL E. NORMAN,3 and H. THEODORE HARCKE2

ABSTRACT

Dual-energy X-ray absorptiometry (DXA) with its short scan time, low radiation dose, and high precision andaccuracy have made this technique particularly suitable for measuring total body bone mineral density (TBMD)in children. Other published reports have related TBMD to age in children 2–18 years of age. However, in youngnormal children aged 2–9 years (51 girls, 43 boys), we found that regression equations for TBMD with age as thepredictor did not explain enough of the variance to warrant their use for predicting TBMD (adjusted R2 0.47,females; 0.41, males). Subtotal BMD (TBMD2 head BMD) is predicted better by age because of a possibly invalidadult algorithm for head BMD (adjusted R2 0.73, females; 0.71, males). (J Bone Miner Res 1997;12:652–655)

INTRODUCTION

DUAL-ENERGY X-RAY ABSORPTIOMETRY (DXA) with itsshort scan time, low radiation dose, and high precision

and accuracy have made this technique particularly suitablefor measuring total body bone mineral density (TBMD) inchildren. Several published papers relate TBMD to age, asmeasured by DXA, in normal children 2–18 years ofage.(1–3) However, in young, normal children scanned at theduPont Institute to establish local laboratory norms, wefound that age accounted for less than 50% of the variabil-ity in TBMD. We report our investigation of this finding.

MATERIALS AND METHODS

Subjects were 51 girls and 43 boys (age range 2–9 years)recruited through local elementary schools or Institute em-ployees for the purpose of developing local laboratorynorms. There were 10 African-Americans whose TBMDdid not differ by age from our Caucasian population, so the

data from both races were pooled together. The study wasconducted between 1993 and 1995, and subjects were stud-ied in either the spring or fall months to avoid any winter/summer variation.(4,5) A short medical and dietary ques-tionnaire was completed for each subject to rule outillnesses and unusual dietary patterns that would affectbone density or body composition. Height, measured on awall-mounted stadiometer, and weight were plotted onstandard growth curves, and any child falling below the 5thor above the 95th percentile for either height or weight wasexcluded from the study. Two subjects were excluded be-cause of increased height (one girl, one African-Americanboy), and two were excluded because of increased weight(one girl, one boy). The age groups were constructed aspreviously reported.(1)

This study was approved by the Institute’s Clinical Re-search Review Committee. Children 5 years of age and overand parents of all subjects signed the informed consentdocument.TBMD was measured by DXA using a Hologic QDR-

2000 (Hologic Inc., Waltham, MA, U.S.A.) in the array-beam scanning mode (Enhanced Array Whole Body, ver-sion 5.60A). Quality control spine phantom scans wereperformed daily (CV 5 0.47%). All children wore hospitalgowns, and any metal objects were removed. Subjects were

*Presented in part at the Annual Meeting of the AmericanSociety for Bone and Mineral Research, Kansas City, MO, U.S.A.,1994.

1Department of Research, Alfred I. duPont Institute, Wilmington, Delaware, U.S.A.2Department of Medical Imaging, Alfred I. duPont Institute, Wilmington, Delaware, U.S.A.3Department of Pediatrics, Carolinas Medical Center, Charlotte, North Carolina, U.S.A.

JOURNAL OF BONE AND MINERAL RESEARCHVolume 12, Number 4, 1997Blackwell Science, Inc.q 1997 American Society for Bone and Mineral Research

652

placed on the table in the supine position with their toestaped to aid in placing the hips and legs in true anteropos-terior position. No children received sedation prior to orduring the scans. In a few cases there was movement on thefirst scan and a second scan was performed without move-ment. Children were kept still by telling them stories and/orby verbal reinforcement to hold still (e.g., “pretend you’re astatue and statues don’t move”).Statistical analysis was performed using Pstat (Princeton,

NJ, U.S.A.). Analysis of variance (ANOVA) was done usinga gender 3 age factorial design. Statistical significance wasconsidered to be p , 0.05. There was no significant inter-action between gender and age. Linear regression analyseswere also performed. Since the ordinary R2 will alwayseither increase or stay the same when a new term is addedto the regression model,(6) the adjusted R2 is presented.

The adjusted R2 accounts for using unnecessary terms in themodel.

RESULTS

Age, height, and weight characteristics by gender and agegroup are shown in Table 1. There were no statisticallysignificant differences between females and males withinany age group.Table 2 gives the TBMD, subtotal BMD, and head BMD

values. There was a significant ( p , 0.0001) age effect forTBMD for both sexes. There was an increase in TBMDwith increasing age, except for age 7 years, where there wasa slight decline from the previous year for both females andmales. Males also showed a slight drop at 9 years.

TABLE 1. AGE, HEIGHT, AND WEIGHT BY GENDER AND AGE GROUP

Agegroup n

Age(years)

Height(cm)

Weight(kg)

Females 2–3 2 3.3 (0.0) 95.4 (0.2) 14.9 (0.5)4 5 4.0 (0.4) 101.4 (2.3) 17.2 (1.0)5 10 5.0 (0.3) 109.4 (3.8) 19.4 (2.6)6 6 6.0 (0.4) 115.0 (2.2) 22.4 (1.2)7 8 6.9 (0.2) 119.6 (5.6) 22.1 (2.3)8 11 8.0 (0.2) 129.6 (3.9) 28.4 (3.5)9 7 8.8 (0.2) 131.8 (5.6) 29.5 (4.0)

Males 2–3 2 2.8 (0.8) 94.4 (4.1) 14.5 (0.4)4 7 3.9 (0.3) 102.4 (2.8) 17.1 (1.7)5 8 5.0 (0.3) 110.1 (3.4) 18.5 (1.3)6 6 5.9 (0.3) 114.9 (4.5) 21.7 (3.0)7 6 6.8 (0.3) 117.0 (6.6) 21.5 (3.4)8 10 7.9 (0.3) 125.4 (3.8) 24.8 (3.3)9 2 8.8 (0.5) 132.5 (0.4) 25.1 (2.3)

Values are means with standard deviations in parentheses.

TABLE 2. BONE MINERAL DENSITY (GM/CM2) BY GENDER AND AGE GROUP

Agegroup n Head Subtotal body Total body

Females 2–3 2 0.998 6 0.031 0.415 6 0.032 0.661 6 0.0374 5 1.112 6 0.149 0.486 6 0.029 0.692 6 0.0495 10 1.170 6 0.069 0.516 6 0.034 0.714 6 0.0266 6 1.268 6 0.061 0.570 6 0.022 0.752 6 0.0307 8 1.189 6 0.075 0.552 6 0.039 0.724 6 0.0328 11 1.309 6 0.126 0.642 6 0.032 0.785 6 0.0349 7 1.321 6 0.087 0.651 6 0.049 0.790 6 0.048

Males 2–3 2 1.044 6 0.054 0.427 6 0.026 0.679 6 0.0004 7 1.105 6 0.111 0.480 6 0.033 0.701 6 0.0475 8 1.168 6 0.118 0.505 6 0.035 0.717 6 0.0476 6 1.337 6 0.124 0.569 6 0.039 0.775 6 0.0607 6 1.261 6 0.093 0.569 6 0.045 0.760 6 0.0408 10 1.327 6 0.055 0.601 6 0.031 0.786 6 0.0239 2 1.310 6 0.043 0.600 6 0.006 0.764 6 0.014

Values are mean 6 SD.

TOTAL BODY BMD IN CHILDREN 653

In deriving a regression equation based on age, age alonewas as good as age 1 age2 based on adjusted R2. Theregression equations and adjusted R2s using age as thepredictor, and Pearson’s correlations (r) were:

Females: TBMD 5 0.021 age 1 0.609, adjusted R2 0.47,

r 5 0.69, p , 0.001

Males: TBMD 5 0.020 age 1 0.626, adjusted R2 0.41,

r 5 0.65, p , 0.001

If height and/or weight were added to the equation, therewas no improvement in adjusted R2 for males. For females,height, and weight slightly increased this value (height, 0.52;weight, 0.47; height 1 weight, 0.52).Further regression analyses of the data were then per-

formed in an attempt to explain why age accounted for lessthan 50% of the variance. The analyses revealed that headBMD accounted for the most variance in TBMD (adjustedR2 . 0.85) and age accounted for less than 50% of thevariance in head BMD. However, age accounted for at least60% of the variance in BMD for the upper limbs, lowerlimbs, and pelvis. Therefore, regression analyses were runusing subtotal BMD (TBMD 2 head BMD) (Table 2),which is included as part of the Hologic patient analysisprintout. The regression equations and Pearson’s correla-tions using age were improved over TBMD and were:

Females: Subtotal BMD 5 0.037 age 1 0.327, adj. R2 0.73,

r 5 0.86, p , 0.001

Males: Subtotal BMD 5 0.031 age 1 0.360, adj. R2 0.71,

r 5 0.85, p , 0.001

Age 1 age2 did not improve the adjusted R2. However,contrary to TBMD, height, and/or weight improved thisvalue for both females (height, 0.85; weight, 0.83; height 1weight 0.86) and males (height, 0.78; weight, 0.81; height 1weight 0.81). Height and weight alone actually accountedfor more variance in subtotal BMD than did age for females(0.85 and 0.80, respectively) and for males (0.78 and 0.74,respectively).

DISCUSSION

In children (as with adults) it has been customary torelate TBMD to age when developing standards.(1–3) Theimportant finding in our data on normal children 2–9 yearsof age was the effect of head BMD on the predictability ofTBMD by age. Head BMD accounted for most of thevariance in TBMD, yet head BMD variance was not ac-counted for by age, height, or weight, suggesting that inclu-sion of head BMD in the calculation of TBMD may con-tribute to the low predictive value of age, height, and weightfor this measurement. We showed that age, height, andweight were much better predictors when subtotal BMDwas used rather than TBMD. Subtotal BMD removes thecontribution of the skull, and in children this may be ofimportance. The algorithm for determining skull bone min-

eral content (BMC) was developed and optimized foradults. The algorithm for determining head BMC uses thebody composition step phantom to determine the soft tissuebaseline of the skull. This is necessary because there isinsufficient soft tissue surrounding the calvarium to enablethe same methodology employed for BMC determination inother parts of the skeleton to be used in the head region(Thomas L. Kelly, personal communication). Skull size inproportion to the body is not changing in the adult. Inchildren the skull does not develop in proportion to bodyweight and other organ weights. Eighty percent of adultskull volume is achieved by 3 years of age, whereas otherbody structures grow linearly with age.(7) Our adjusted R2

for head BMC versus age was only 0.52 for both males andfemales. Because of these differences, it may be more ap-propriate to use the subtotal BMD result, which excludesthe head region, as the reference measurement with whichto compare children with bone disorders. Whether or notthe subtotal BMD result will improve the diagnostic sensi-tivity of the whole body measurement in children needs tobe investigated.Another issue concerns the use of BMC instead of BMD.

This is not within the scope of this paper. However, welooked at the regression of TBMC and subtotal BMC withage and found the adjusted R2 values to be essentially thesame as those obtained with subtotal BMD and age.At 7 years of age, there was a slight drop in TBMD in

both females and males. We could not explain this dropbased on differences in either height or weight. Two previ-ous reports have shown drops in BMD around this age.Ponder et al.(8) showed by dual-photon absorptiometry aslight median drop in spinal BMD for females at age 8 yearsand for males at age 7 years, and Glastre et al.(9) showed aslight decrease in spinal BMD at age 7 in females but not inmales by DXA. This variation provides another rationalefor needing an age-specific pediatric reference base againstwhich to compare patients of similar ages.In summary, in young children, regression models for

TBMD with age as the predictor are presented, but theamount of variance explained by age is too low to warranttheir use for predicting TBMD. Subtotal BMD, however, ispredicted much better by age, and because of a possiblyinvalid adult algorithm for head BMD, subtotal BMD maybe a more reliable measurement in younger children.

REFERENCES

1. Faulkner RA, Bailey DA, Drinkwater DT, Wilkinson AA,Houston CS, McKay HA 1993 Regional and total body bonemineral content, bone mineral density, and total body tissuecomposition in children 8–16 years of age. Calcif Tissue Int53:7–12.

2. Lu PW, Briody JN, Ogle GD, Morley K, Humphries IRJ, AllenJ, Howman-Giles R, Sillence D, Cowell CT 1994 Bone mineraldensity of total body, spine, and femoral neck in children andyoung adults: A cross-sectional and longitudinal study. J BoneMiner Res 9:1451–1458.

3. Zanchetta JR, Plotkin H, Alvarez Filgueira ML 1995 Bone massin children: Normative values for the 2–20 year-old population.Bone 16:393S–399S.

654 TAYLOR ET AL.

4. Bergstralh EJ, Sinaki M, Offord KP, Wahner HW, Melton LJIII 1990 Effect of season on physical activity score, back exten-sor muscle strength, and lumbar bone mineral density. J BoneMiner Res 5:371–377.

5. Rico H, Revilla M, Cardenas JL, Villa LF, Fraile E, Martin FJ,Arribas I 1994 Influence of weight and seasonal changes onradiogrammetry and bone densitometry. Calcif Tissue Int54:385–388.

6. Montgomery DC, Peck EA (eds.) 1992 Introduction to LinearRegression Analysis. John Wiley and Sons, Inc., New York, NY,U.S.A., pp. 160.

7. Webster EW, Alpert NM, Brownell GL 1974 Radiation doses inpediatric nuclear medicine and diagnostic X-ray procedures. In:James AE, Wagner HN, Cooke RE (eds.) Pediatric NuclearMedicine. W.B. Saunders Company, Philadelphia, PA, U.S.A.,pp. 36.

8. Ponder SW, McCormick DP, Fawcett HD, Palmer JL, McKer-nan MG, Brouhard BH 1990 Spinal bone mineral density in

children aged 5.00 through 11.99 years. Am J Dis Child144:1346–1348.

9. Glastre C, Braillon P, David L, Cochat P, Meunier PJ, DelmasPD 1990 Measurement of bone mineral content of the lumbarspine by dual energy x-ray absorptiometry in normal children:Correlations with growth parameters. J Clin Endocrinol Metab70:1330–1333.

Address reprint requests to:Arlene Taylor

Alfred I. duPont InstituteDepartment of Medical Imaging

P.O. Box 269Wilmington, DE 19899 U.S.A.

Received in original form April 19, 1996; in revised form October29, 1996; accepted November 21, 1996.

TOTAL BODY BMD IN CHILDREN 655