Ethnic, Anthropometric, and Lifestyle Associations with RegionalVariations in Peak Bone MassJ. W. Davis,1 R. Novotny,2 R. D. Wasnich,1 P. D. Ross31Hawaii Osteoporosis Center, 401 Kamakee Street, Honolulu, Hawaii 96814, USA2Department of Food Science and Human Nutrition, College of Tropical Agriculture and Human Resources, University of Hawaii at Manoa,Honolulu, Hawaii 96822, USA3Merck & Co., Rathway, New Jersey 07065, USA

Received: 6 March 1998 / Accepted: 15 December 1998

Abstract. The study investigated the ability of ethnicityand anthropometric and lifestyle factors to account for dif-ferences within subjects in bone mass at different skeletalsites. The subjects were young, adult, Japanese, Filipino,Hawaiian, and white women ages 2534. In the preliminaryanalyses, they were divided into thirds based on their BMDz-scores. Thirty-five percent exhibited high variability inbone mass: they were in the upper third at one or more bonesites and in the lower third at one or more sites. Otherwomen had more generalized low bone mass: 25% were inthe lowest third for two or more sites, and there were nosites with low bone mass in the upper third. In subsequentanalyses, ethnicity, anthropometry, and lifestyle influenceswere examined as possible predictors of differences in bonemineral content (BMC) between bone sites in bone-sizeadjusted models. White women had greater BMC at theproximal radius and calcaneus than at the distal radius com-pared with other ethnic groups. This may be explained bythe fact that they had exceptionally wide bone widths at thedistal radius. Of the anthropometric variables, fat mass wasassociated with higher bone mass at sites with higher pro-portions of cancellous tissue (calcaneus > spine > radiussites). Muscle mass was associated with greater bone massat the calcaneus and proximal radius than at the spine. Forthe lifestyle variables, women with greater milk consump-tion between the ages of 1024 years had higher spine bonemass than expected from their measurements at the proxi-mal radius. Women 1217 years of age who had been moreactive in sports had higher calcaneous bone mass than ex-pected from their spine measurements. As the study partici-pants were still young women, the results suggest thatregional differences in bone mass may partly derivefrom anthropometric and lifestyle influences during skeletalmaturation.

Key words: Bone density Bone mineral content An-thropometry Lifestyle Ethnicity.

Low bone mass among postmenopausal women may occurwith either a regional or generalized distribution [1].Women with a regional distribution have low bone mass

limited to specific bone sites whereas those with a general-ized distribution have low bone mass throughout much oftheir skeletons. Fracture risk increases with the number oflow bone mass sites, although women with even a singlelow site have an increased fracture risk [1]. Regional dif-ferences in bone mass may result from localized differencesin exposures, as from localized load bearing, regional dif-ferences in responsiveness, or from both. Furthermore, thetiming of the development of regional differences in bonemass is not well understood. Bone mass lower at one sitethan another might arise during growth, accrue throughpostmenopausal bone loss, or develop throughout life.

This study examined the extent that regional and gener-alized low bone mass may already exist among young adultwomen. The participants were a multiethnic population liv-ing on the island of Oahu. Previous articles reported thatbone density varied by ethnicity, and that the exposuresassociated with the amount of bone differed among bonesites [2, 3]. For example, body mass index and muscle masswere associated at the calcaneus and milk use, height, andmuscle mass were associated at the proximal radius. Thisarticle quantifies the associations among ethnicity, anthro-pometric, and lifestyle exposures with the extent that bonemass differs among bone sites within individuals.

Materials and Methods

SubjectsWomen were eligible if they were 2534 years of age and ofHawaiian, Filipino, Japanese, or white ethnicity [3]. One hundredpercent Japanese or white ancestry was required for participationin the study. Filipino women, however, were accepted if they hadat least one full Filipino parent, or two Filipino grandparents; 57%of the Filipino women were of pure Filipino ancestry. All Hawai-ian women were eligible who met the age criteria. Most had amixed ancestry; the median proportion of native Hawaiian ances-try was 33%.

The study was widely advertised on television, radio and busadvertisements, and articles in the major and ethnic newspapers.Brochures were also distributed at community and health centers.A total of 421 women joined the study (144 Japanese, 137 white,74 Filipino, 66 Hawaiian); their characteristics have been previ-ously published [13].

Bone Mass Measurements

Bone measurements were made at four bone sites: the spine andCorrespondence to: J. W. Davis

Calcif Tissue Int (1999) 65:100105

1999 Springer-Verlag New York Inc.

distal radius (common fracture sites among postmenopausalwomen) and the calcaneus and proximal radius, sites of high can-cellous and cortical bone content. The calcaneus and distal andproximal radius were measured using single-energy X-ray absorp-tiometry (Osteon, Wahiawa, Hawaii) [4]. The lumbar spine wasmeasured using dual-energy X-ray absorptiometry (Hologic,Waltham, MA). Both bone mineral contents (BMCs) and boneareas were measured at all four sites. Bone mineral densities(BMDs) were calculated as the ratios of the BMCs to the boneareas.

Anthropometry and Lifestyle ExposuresAnthropometry and lifestyle exposures were previously describedin detail [2]. Anthropometric measurements were made in tripli-cate using standardized techniques. The closest two values wereused in analyses to eliminate any gross measurement errors. Theanthropometric measurements included height, weight, arm andcalf circumferences, and subscapular, biceps, triceps, suprailiac,and calf skinfold thicknesses. Body fat mass and muscle indiceswere calculated from the anthropometric measurements using pub-lished equations [57]. The body muscle index was based on thearm and calf circumference and skinfold measurements. Sportsactivity for ages 1217 was assessed by a questionnaire asking, forall sports activities in which the women had participated, the num-ber of weeks per year, the number of times per week, and thenumber of hours per time. From this information and publishedtables of metabolic equivalents (METs) [8], we calculated totalMETs for sport activities. Subjects were categorized into upper,middle, and lower thirds for sports activity based on the metabolicequivalents for sports activities calculated from their answers.METs were also calculated for physical activity at the time of thestudy; however, current METs were not significantly associatedwith the study outcomes. The subjects were asked separately aboutmilk consumption for ages 1014, 1519, and 2024 years. Foranalysis purposes, their responses were categorized into threegroups depending on their intake of at least 1 glass (237 ml) ofmilk per day during these three age ranges [always (all three ages),sometimes (one or two ages), or never].

Analysis

Both BMC and BMD measurements were converted to z-scores forthe analyses. The means and standard deviations for the entirestudy population were used to calculate the z-scores. Because thebone measurements did not change with age among our young,adult population [3], the z-scores were not age adjusted. For someanalyses the women were divided into thirds based on their BMDz-scores; they were categorized as being in the upper, middle, orlower third at each of the four bone sites. Thirds were selected forcategorization to insure that errors in classification from a lower toupper category were unlikely. Bone densitometry measurementshave a precision of 12%, and measurement errors should rarely

misclassify subjects from the upper to lower third, or vice versa.BMD measurements were chosen for these analyses because theyare commonly used for clinical evaluation, and have strong asso-ciations with fracture risk [9]. Differences in BMC z-scores (onez-score minus another) were calculated for each person as depen-dent variables for multivariable regression analyses. The four bonesites yielded six pairs of differences that were used as outcomes inthe regression analyses. All regression analyses included adjust-ment for bone areas of the two bone sites used in calculating thedifference in z-scores [10]. For variables in the models, the ad-justed analyses measure how much higher (or lower) BMC is atone site compared with a second site among women of similarbone size. A positive regression coefficient indicates that, on av-erage, women with increasing values of the independent variableshad higher z-scores for the first bone site listed in the tables thanfor the second.

Results

To examine the uniformity of BMD across skeletal sites(i.e., whether some women had higher BMD measurementsat some bone sites than at others), the women were cross-classified by the number of their BMD sites that were in theupper and lower thirds for the population. Thirty-five per-cent exhibited high variability, with BMD in the upper thirdon at least one of four bone sites, and in the lower third atone or more other sites (Table 1). Some women were in thelower third at multiple sites: 39% at two or more sites, and20% at three or more sites.

Ethnicity was examined as a possible explanation for thevariability in bone mass and was associated with the differ-ences from five of the pairs of bone sites (Table 2). Com-pared with whites, Japanese had higher z-scores at the distalradius than at the other sites. For the Japanese and whitewomen, both their parents and grandparents had the sameethnicity as the participants. Compared with whites, Hawai-ian women had higher z-scores at the distal radius than atthe spine or proximal radius and higher z-scores at theproximal radius than at the calcaneus. Compared with theJapanese, Hawaiians had higher z-scores at the calcaneusthan at the spine or radius sites. The Hawaiian participants,however, averaged only 33% Hawaiian ancestry, and theresults may not be generalizable to other Hawaiian popula-tions. Compared with whites, Filipinos had greater calca-neus than spine z-scores, and greater distal radius than spineor proximal radius z-scores. Compared with Japanese, Fili-pinos had greater z-scores at the calcaneus than at the spineor radius sites. The Filipino women, like the Hawaiianwomen, were of mixed ancestry, which also limits the gen-eralizability of the results.

Ethnicity may serve as an indicator of differences in

Table 1. Cross-classification of the bone density of 421 young, adult women based on thenumber of bone sites in the upper and lower thirds at four skeletal sites

No. inupper third

Number in lower third

0 1 2 3 4

0 18 (4.3%) 21 (5.0%) 41 (9.7%) 36 (8.6%) 28 (6.7%)1 25 (5.9%) 37 (8.8%) 32 (7.6%) 18 (4.3%)2 42 (10.0%) 35 (8.3%) 8 (1.9%)3 37 (8.8%) 18 (4.3%)4 25 (5.9%)Results give frequencies and percentages for the 421 women based on spine, calcaneus, anddistal and proximal radius BMD measurements

J. W. Davis et al.: Bone Mass Differences Among Bone Sites 101

anthropometry or lifestyle. To investigate this possibility,anthropometry and lifestyle were examined individually aspredictors of z-score differences and, subsequently, in com-bination with ethnicity. Height, fat mass, and muscle masswere chosen for the analyses as distinct, anthropometriccomponents (Table 3). Taller women had higher z-scores atthe proximal radius than at the spine or distal radius.Women with greater fat mass had higher z-scores at thecalcaneus than at the other sites; they also had higher z-scores at the spine than at the radius sites. The magnitudesof the differences were greatest between the calcaneus andradius sites. Women with greater muscle mass had higherz-scores at the calcaneus and radius sites than at the spine.

Sports activity, milk consumption, and menstrual history(age at menarche and menstrual flow) were examined aspotentially important aspects of lifestyle because these vari-ables were associated with BMC in previous analyses [2].Neither age at menarche, menstrual flow, or other menstrualhistory variables examined (including a history of amenor-rhea and oligomenorrhea) were significantly associated (P spine > radius sites), suggesting thatfatness may influence cancellous more than cortical bone.Fat mass may be associated with bone indirectly from as-sociations with hormonal exposures. The onset of menar-che, for instance, is strongly associated with fatness [16].Among the study population, however, menstrual historywas not associated with differences in BMC between sites.Fat mass, alternatively, may affect bone mass by increasingthe supported weight; the more cancellous sites were alsothe more weight-loading sites.

Muscle mass, like fat mass, was associated with differ-ences in BMC. Women with greater muscle mass index hadlower spine BMC than expected from measurements atother sites. The index of muscularity, however, was based

on appendicular measurements, and may not have accu-rately reflected muscularity at the spine. If so, the lowerspine BMC may emphasize the importance of regional mus-cularity in skeletal development.

Compared with the anthropometric measurements,sports activity from ages 12 to 17, and milk intake from ages10 to 24 were less consistently associated with differencesin BMC. Women with limited sports activity at ages 1217,however, did have lower calcaneus BMC than expectedfrom their spine measurements. Many sports activities in-volve running, and may load the heel more than the spine.Women with higher milk intake at ages 1024 had higherproximal radius bone mineral than expected from their spinemeasurements. Studies of calcium supplements duringgrowth have documented benefits before, but not after pu-berty [17, 18]. The peak bone mass of the proximal radius

Table 3. Z-score differences per SD difference in height, fat mass, and muscle mass

Z-score difference

Z-score difference per SD of the listed variable

Height Fat mass Muscle mass

Spineproximal radius 0.135 0.057a 0.180 0.040a 0.112 0.042aSpinedistal radius 0.099 0.063 0.150 0.044a 0.208 0.047aSpinecalcaneus 0.018 0.075 0.127 0.050a 0.159 0.048aProximal radiusdistal radius 0.149 0.050a 0.010 0.041 0.055 0.043Calcaneusproximal radius 0.074 0.082 0.299 0.061a 0.056 0.059Calcaneusdistal radius 0.070 0.083 0.289 0.061a 0.014 0.059Z-score differences represent differences in BMC between the first and second bone siteslisted. All 6 regression models included the z-score difference as the dependent variable, andheight, fat mass, muscle mass, and the bone areas for the two bone sites used to calculate thez-score as independent variablesa P < 0.05

Table 4. Z-score differences among bone sites by sports activity at ages 1217 years and milk consumption at ages 1024 years

Dependent variable(z-score difference)

Mean z-score difference SEM

Sports activity(1217 years)

Milk consumption(1024 years)

Spineproximal radius Lower third 0.171 0.071

may be influenced more than the spine by exposures prior topuberty. Our study did not collect information on activitylevels or calcium intakes before puberty, and cannot directlyaddress prepubertal influences.

In multivariable, prediction models, fat mass remainedassociated with differences in BMC between the more andless cancellous bone sites. The strongest association waswith differences between the calcaneus and the radius sites(0.3 z-score unit differences per SD of fat mass). Musclemass remained associated with differences between thespine and the other sites (0.10.2 z-score unit differencesper SD of muscle mass). The strengths of the associationsof ethnicity, height, milk intake, and sports activity withdifferences in BMC between sites were comparable in mag-nitude to the associations for muscle mass. Taller womenhad higher BMC at the proximal radius than expected fromtheir distal radius measurements. Women with low milkintakes, age 1024, had lower radius BMC than expectedfrom their spine measurements. Ethnicity was associatedwith differences between the distal radius and the calcaneusand proximal radius. The strengths and directions for all ofthe variables in the multivariable models were similar tothose in the simpler models.

The magnitude of the differences in z-scores were mod-estonly fractions of a z-score per SD for continuous vari-ables or between levels of categorical variables. For com-parison, assuming a twofold difference in risk per SD, assupported by some studies of older women [9], a differenceof 0.25 SD would result in an approximately 20% differencein fracture risk. Although modest for the individual, thedifferences could apply to many individuals in the popula-tion. Moreover, these results represent average effects,which may have been underestimated as a result of randommeasurement errors; differences may be much larger forsome individuals.

The results of the various analyses suggest that substan-tial heterogeneity in bone strength already exists within theskeletons of many young, adult women. Other investigatorshave reported associations of factors such as intense or pro-longed sports activity, or amenorrhea, with exceptionallyhigh or low regional bone density [1922]. Our results sug-gest the localized sites of low bone mass are not uncommonamong young, adult women. The results further suggestthat regional differences in bone strength may partly derivefrom anthropometric and lifestyle influences during skeletalmaturation.

Table 5. Associations of z-score differences among bone sites with anthropometric andlifestyle variables in multivariable models

Dependent variables(z-score differences) Independent variables

Regression coefficients SE

Spineproximal radius Height 0.119 0.052Fat mass 0.180 0.040Muscle mass 0.100 0.042Milk intake, ages 1024 years

References

1. Davis JW, Ross PD, Wasnich RD (1994) Evidence for bothgeneralized and regional low bone mass among elderlywomen. J Bone Miner Res 9:305309

2. Davis JW, Novotny R, Ross PD, Wasnich RD (1996) Anthro-pometric, lifestyle and menstrual factors influencing size-adjusted bone mineral content in a multiethnic population ofpremenopausal women. J Nutr 126:29682976

3. Davis JW, Novotny R, Ross PD, Wasnich RD (1994) Thepeak bone mass of Hawaiian, Filipino, Japanese, and whitewomen living in Hawaii. Calcif Tissue Int 55:249252

4. Yano K, Wasnich RD, Vogel JM, Heilbrun LK (1984) Bonemineral measurements among middle-aged and elderly Japa-nese residents in Hawaii. Am J Epidemiol 119:751764

5. Siri WE (1956) Body composition from fluid spaces and den-sity: analysis of methods. Univ of Calif, Berkeley, Donner LabMed Report (UCRL-3349)

6. Ross WD, Crawford SM, Kerr DA, Ward R, Bailey DA, Mir-wald RM (1988) Relationship of the body mass index withskinfolds, girths, and bone breadths in Canadian men andwomen aged 2070 years. Am J Phys Anthropol 77:169173

7. Durnin JVGA, Womersley J (1974) Body fat assessed fromtotal body density and its estimation from skinfold thickness:measurements on 481 men and women aged 16 to 72 years. BrJ Nutr 32:7797

8. Ainsworth BD, Haskell WL, Leon AS, Jacobs DR Jr, MontoyeHJ, Sallis JF, Paffenbarger RS Jr (1993) Compendium ofphysical activities: classification of energy costs of humanphysical activities. Med Sci Sports Exerc 25:7180

9. Marshall D, Johnell O, Wedel H (1996) Meta-analysis of howwell measures of bone mineral density predict occurrence ofosteoporotic fractures. Br Med J 312:12541259

10. Prentice A, Parsons TJ, Cole TJ (1994) Uncritical use of bonemineral density in absorptiometry may lead to size-relatedartifacts in the identification of bone mineral determinants.Am J Clin Nutr 60:837842

11. Honkanen R, Tuppurainen M, Kroger H, Alhava E, Puntila E(1997) Associations of early postmenopausal fractures with

subsequent fractures by sites and mechanisms of fractures.Calcif Tissue Int 60:327331

12. Torgerson DJ, Campbell MK, Thomas RE, Reid DM (1996)Prediction of perimenopausal fractures by bone mineral den-sity and other risk factors. J Bone Miner Res 11:293297

13. Tuppurainen M, Kroger H, Honkanen R, Puntila E, Huopio J,Saarikaski S, Alhava E (1993) Risks of perimenopausal frac-turesa prospective population-based study. Acta Obstet Gy-necol Scand 74:624628

14. Bonjour JP, Theintz G, Buchs B, Slosman D, Rizzoli R (1991)Critical years and stages of puberty for spinal and femoralbone mass accumulation during adolescence. J Clin Endocri-nol Metab 73:555563

15. Eveleth PB (1986) Population differences in growth: environ-mental and genetic factors. In: Falkner F, Tanner JM (eds)Human growth, vol 3. Plenum Press; New York, NY, pp 221239

16. Frisch RE (1985) Fatness, menarche, and female fertility. Per-spect Biol Med 28:611633

17. Johnston CC, Miller JZ, Slemenda CW, Reister TK, Hui S,Christian JC, Peacock M (1992) Calcium supplementation andincreases in bone mineral density in children. N Engl J Med327:8287

18. Bonjour JP, Rizzoli R (1996) Bone acquisition in adolescence.In: Marcus R, Feldman D, Kelsey J (eds) Osteoporosis. Aca-demic Press, pp 465476

19. Huddleston AL, Rockwell D, Dulund DN (1980) Bone massin lifetime tennis athletes. JAMA 244:11071109

20. Slemenda CW, Johnston CC (1993) High intensity activitiesin young women: site-specific bone mass effects among fe-male figure skaters. Bone Miner 20:125132

21. Drinkwater BD, Nelson KL, Chestnut CS, Bremner QJ, Shain-holz S, Southworth MB (1984) Bone mineral content of amen-orrheic and eumenorrheic athletes. N Engl J Med 311:277281

22. Nelson ME, Fisher EC, Catsos PD, Meredith CN, TurksoyRN, Evals WJ (1986) Diet and bone status in amenorrheicrunners. Am J Clin Nutr 43:910916

J. W. Davis et al.: Bone Mass Differences Among Bone Sites 105