Are there ethnic differences in the association between body ... - Nature

International Journal of Obesity (1997) 21, 1085±1092 ß 1997 Stockton Press All rights reserved 0307±0565/97 $12.00

Are there ethnic differences in the association between body weight and resistance, measured by bioelectrical impedance? BL Heitmann1,2, BA Swinburn3, H Carmichael3, K Rowley4, L Plank5, R McDermott6, D Leonard6 and K O'Dea4 1

Danish Epidemiology Science Center, Institute of Preventive Medicine, Copenhagen Hospital Corporation, Municipal Hospital of Copenhagen, Denmark; 2 The Glostrup Population Studies, Medical Dept. C, Glostrup Hospital, Copenhagen, Denmark; 3 Department of Community Health, University of Auckland, Auckland, New Zealand; 4 Deakin Institute of Human Nutrition, Deakin University, 336 Glenferrie Road, Malvern, Australia; 5 Department of Surgery, University of Auckland, Auckland, New Zealand and 6 Tropical Public Health Unit, Queensland Health, Cairns, Queensland, Australia.

OBJECTIVE: To describe ethnic differences in the relationship between body size and body composition. Knowledge about such differences is important when studying obesity-related complications, such as hypertension and noninsulin dependent diabetes, because it may not be possible to generalize results from one study population to other populations. DESIGN: Cross-sectional study. SUBJECTS: Four groups of different ethnic identity (2987 Caucasians (Danes), 243 predominantly Melanesian (Torres Strait Islanders from northern Australia), 206 Australian Aborigines and 146 Polynesians (New Zealand Samoans)), aged 30±70 y, were studied. MEASUREMENTS: We examined associations between body weight and bioelectrical impedance, as a measure of body composition. RESULTS: Except for Australian Aborigines, associations (slopes) between body weight and resistance were generally constant in the different ethnic groups, once height and age differences had been considered, indicating that this relationship may involve a certain universality, that is independent of the population speci®city for impedance measurement. Systematic differences in instrument readings or electrodes did not seem to be responsible for the differences found. CONCLUSION: With the exception of Australian Aborigines, there may be a constant relation between body size and body composition (total body water or fat free mass) of different ethnic groups, that depends on gender and age category only. Keywords: body composition; bioelectrical impedance; ethnic differences

Introduction The relationship between body weight and body fatness is important in both populations and individuals. There is recognition that this relationship may differ between ethnic groups.1±3 However, few studies have examined the effects of ethnicity on the relationship between body size and body composition. Although the prevalence of obesity-related complications, such as hypertension and non-insulin dependent diabetes varies widely among ethnic groups,4 data on body fatdisease vs overweight-disease relationships, is not available. Bioelectrical impedance analysis is an attractive tool in epidemiology for the assessment of body composition, due to its portability, low cost and ease of use for a large number of subjects. However, one of the problems with the method is the apparent population speci®city of the equations applied to Correspondence: Dr BL Heitmann. Received 11 October 1996; revised 13 May 1997; accepted 19 May 1997

estimate body composition.5 In this context, we have examined the associations between bioelectrical impedance, as a measure of body composition, and body weight, in different ethnic groups. While the nature of the relationship between body size and body composition may contain some fundamental characteristics, the aim of this study was to describe the ethnic differences in the relationship.

Subjects and methods Subjects

Four groups, of different ethnic identity, were studied: a Caucasian group (Danes), a predominantly Melanesian group (Torres Strait Islanders from northern Australia), a group of Australian Aborigines and a Polynesian group (New Zealand Samoans). In all instances, representative samples of the populations were studied. In this present study, only subjects aged 30±70 y were included. All analyses were done separately for subjects above or below 50 y of age.

Ethnic differences in bioelectrical impedance BL Heitmann et al

1086

The Danes

The study population included 3608 Danish citizens, all full Caucasian, aged 35, 45, 55 and 65 y. This group was a gender and age strati®ed random sample of the population in the western part of the Copenhagen County.6 The group was selected from the Central Person Register, ®ve years prior to the study, to represent the general population. The study was part of the Danish MONICA project.7±8 The present examination (GEN-MONICA) was carried out from December 1987 to November 1988. All subjects were invited to a general health examination including measurements of height, weight and bioelectrical impedance. A total of 2987 subjects (83%) attended the examination. These subjects were considered the participating Danes. Non-participation has been described in an earlier study.9,10. The Samoans

The Samoans recruited for this study were matched for body mass index (BMI) by gender and 10 y age groups to Samoans surveyed in three communities in different Auckland suburbs. There are no nationally representative data for Paci®c Islands populations living in New Zealand and these Presbyterian church-based communities were considered to be representative of New Zealand Samoans in general. Mean age-adjusted BMI of the study group was 32.6 kg/m2 for males and 34.8 kg/m2 for females compared to the community population of 33.4 kg/ m2 and 36.6 kg/m2, respectively. This difference was signi®cant for the women (P < 0.05) but not the men. Subjects were of at least 75% Samoan origin, did not suffer from any serious illnesses and were not participating in weight lifting or body building. They underwent height, weight and bioelectrical impedance assessment in 1995. The group aged 30±70 y included 146 subjects. The Australian Aborigines11,12

Anthropometric data, including body weight, height and bioelectrical impedance, was collected as part of a population-based health survey in 1995 in a group of remote Australian Aboriginal communities in central Australia. The communities constituted two related population groups de®ned by language and cultural factors. The survey included 90% of the resident adult ( 15 y) population. For the purpose of the present analysis, where exclusion criteria were age below 30 y or above 70 y, a total of 206 subjects were included. The Torres Strait Islanders13

This group of indigenous Australians from The Torres Strait region of far north Queensland is of primarily Melanesian origin and is ethnically distinct from mainland Australian Aboriginal people. As part of a community health screening program (similar to that carried out in the central Australian communities

described above), body weight, height and bioelectrical impedance data was collected in 1993±1995 in ®ve Torres Strait Island communities. The surveys covered from 50±73% of the resident adult population of the ®ve communities surveyed. Ethnic origin was determined by questionnaire and subjects were excluded from present analysis if, in addition to the age criteria de®ned above, they reported their origins to be wholly or partly Australian Aboriginal, or other ethnic group. A total of 243 Torres Strait Islanders were thus included in this analysis. Anthropometric data

Anthropometric measurements were made in accordance with the WHO standards.14 Height (Ht) was measured to the nearest 0.5 cm in Danes and Samoans, and to the nearest 0.1 cm in about half the Australian Aborigines and the Torres Strait Islanders. In the remaining Australian Aborigines, heights were recorded to the nearest 1 cm. All subjects were standing without shoes, heels together and head in the horizontal Frankfurt plane. Body weight (BW) was measured to the nearest 0.1 kg, using either a calibrated SECA scale or an UC300 digital electronic scale. Subjects wore hospital underwear or light clothing during measurements. Measurements of electrical impedance

Different single frequency impedance instruments were used for measuring impedance in the four ethnic groups, but within ethnic group, the same instrument was used. All Danes were measured using a BIA-103 RJL-system-analyser, whereas Samoans were measured with a BIA-101 RJLsystem-analyser (both from RJL-systems, Detroit, USA). The connecting electrodes were of the brand Accu Sensor, (Carbo Cone M45, Lynn Medical Michigan, USA) in the Danes, whereas the Samoans were measured using halved disposable ECG electrodes, supplied by RJL-systems. Australian Aborigines impedance data was collected using an RJL-systems model BIA-101Q (RJL-systems, Detroit, USA) whereas Torres Strait Islanders impedance data came from a SEAC model BIM-3.2 impedance analyser (Inderlec, Sydney, Australia). In both groups, Nikota electrodes (Nikotabs-E, Niko surgical, UK) were used. All measurements were taken following instructions given by the manufacturer (tetra polar electrode placement: electrodes on dorsal surfaces of right hand and right foot, at the distal metacarpals and metatarsals, respectively, and between distal prominences of the radius and the ulna at wrist, and the medial and lateral malleoli at ankle. Subjects were supine with arms and legs not touching the body, or each other). Danes, Aborigines and Torres Strait Islanders had measurements taken after a 12 h overnight fast, whereas measurements in the Samoans were performed without prior fasting.


Systematic differences in impedance readings between instruments used

We assessed differences in impedance readings between three of the instruments used to measure bioelectrical impedance in the ethnic groups (machines used for Danes, New Zealand Samoans and Australian Aborigines). Furthermore, we studied differences in readings from the electrodes used in the different populations. A sample of 20 men and 21 women, of different ethnicity (23 Caucasians, 2 Samoans, 2 Chinese, 5 Tongans, 2 Cook Islanders and 7 Maori) and age (range 21±71 y) from Auckland New Zealand, participated in the study. The project was approved by the Ethical Committees for Copenhagen County, for University of Auckland, for Deakin University, for Alice Springs Hospital and for Cairns Base Hospital, and is in accordance with the Helsinki II Declaration.

Statistical methods

co-variates in the regression models. Resistance is a function of the length (height) of the conductor (the body) and residual age variation may confound the relations, hence, all multiple regression analyses were adjusted for differences in height and residual age. Analyses of variance were used to describe differences between mean values between groups of different gender, age and ethnicity. A standard F-test was applied to examine differences between regression coef®cients and intercepts between ethnic groups. This procedure has been previously described.15 To assure linearity, probability and residual plots were drawn for resistance, log of resistance, weight and log weight, to test for normality and for normal distribution with constant variance of the residuals. Log transformation did not improve linearity and all groups had normally distributed data. All analyses were performed using the SPSS PC or the SAS PC programmes.

Results

Two-way analysis of variance was used to describe group differences between readings of different instruments and electrodes. Agreement between instruments was displayed visually as differences between resistance readings as a function of the mean of the two methods. Multiple regression analyses were carried out for separate stratas of gender, ethnic groups and age (young: 30±50 y; old: 50±70 y). Impedance was included as the dependent and weight as the independent variable. In order to account for interactions on the associations between measures of size (body weight) and measures of body composition (resistance), the regression models included ethnicitydependent intercepts and slopes for all stratas of men and women in the two age-groups (30±50 y and 50±70 y). Height and residual age were included as

The characteristics (mean and s.d.) of subjects by gender, age and ethnicity are shown in Table 1. Within age group, mean ages were similar in all four ethnic groups except that within the 30±50 y age group the Danish men tended to be older, and men from the Torres Strait Island were younger, than other men in this group. In general average values of weight, height and resistance were signi®cantly different between all groups of age, gender and ethnicity, with the highest weights and lowest heights and resistances in the Samoans. Figure 1 displays the distribution of BMI in the four populations by gender and age. The Samoans had the greatest prevalence of high BMI, whereas Danish women and Australian Aboriginal men had the lowest prevalences.

Table 1 Mean and standard deviation (s.d.) of age, weight, height and resistance of the subjects by gender, age and ethnicity Age (y) mean (s.d.) Women (30±50 y) Samoans (n 40) Danes (n 772) Torres St. (n 96) Aborigines (n 76) Women (50±70 y) Samoans (n 36) Danes (n 691) Torres St. (n 36) Aborigines (n 39) Men (30±50 y) Samoans (n 34) Danes (n 786) Torres St. (n 74) Aborigines (n 63) Men (50±70 y) Samoans (n 36) Danes (n 737) Torres St. (n 37) Aborigines (n 28)

Weight (kg) mean (s.d.)

Height (m) mean (s.d.)

Resistance (O) mean (s.d.)

39.4 40.0 39.0 38.8

(6.4) (5.0) (6.0) (5.5)

92.0 64.4 87.6 67.7

(17.0) (10.9) (20.6) (13.5)

1.624 1.654 1.623 1.617

(0.061) (0.059) (0.050) (0.052)

459 (72) 569 (60) 511 (86) 624 (108)

58.5 59.6 59.5 58.4

(5.6) (5.0) (5.4) (4.7)

86.5 66.0 76.8 66.7

(16.2) (12.2) (13.1) (13.0)

1.568 1.612 1.592 1.617

(0.045) (0.056) (0.053) (0.065)

480 567 541 652

(61) (65) (75) (95)

39.9 40.3 38.1 39.2

(5.6) (5.0) (5.1) (6.1)

95.9 80.3 94.1 74.3

(10.4) (11.7) (20.2) (14.6)

1.715 1.778 1.730 1.731

(0.056) (0.065) (0.072) (0.055)

386 464 418 539

(32) (47) (57) (84)

59.3 59.8 58.5 59.0

(5.6) (5.0) (5.7) (5.7)

96.0 79.6 84.1 74.4

(16.7) (12.3) (15.5) (10.4)

1.701 1.735 1.699 1.740

(0.051) (0.066) (0.052) (0.052)

391 468 442 543

(50) (54) (72) (62)

1087


1088

Figure 1 Cumulative distribution curves of body mass index in the four ethnic groups by gender and age.

With the same electrodes (Accu Sensor) for all measurements, virtually identical mean values were obtained with the three instruments (Mean resistance was, using the Danish instrument: 483 O, using the New Zealand instrument: 485 O, and using the Australian instrument: 484 O). Readings were similar for all groups of age, gender and ethnicity. Figure 2 shows the agreement between individual measurements with the three machines. Individual differences were all less than 5% of the mean values and most were within 2%. Different electrodes gave different readings in the 35 subjects where comparison was performed. In general, these differences were quite small (1±2%), however, and did not vary systematically across the weight range (all P for trend > 0.2). The average impedance readings, taken with the type of electrodes used with the Danish instrument (Accu Sensor) were, on average, 7 O lower than the average value of the New Zealand instrument

(492 67 O compared to 498 71 O, P 0.01) and 9 O lower than the average impedance readings with electrodes of the Australian instrument (Nikota) (501 74 O, P 0.05). Table 2 and Figure 3 show relationships for resistance measured by impedance regressed on body weight, after adjustment for height and age. Weight was negatively associated with resistance in all subgroups of gender, age and ethnicity. A standard F-test showed that slopes were different (all P < 0.02) for men and women aged 30±50 y in all ethnic groups, except for Australian Aborigines. Among subjects aged 50±70 y, slopes differed between Danish men and Danish women only (P 0.02, all other P > 0.48). In all men, except the Australian Aborigines, slopes were different between those aged 30±50 y and 50± 70 y (all P < 0.01). Women displayed no signi®cant age differences (all P > 0.17). Within the groups of men and women aged 30±50 y, and in the women aged


aged 30±50 y was: b 72.49; s.e. 0.12, while the slope common to Danes and Samoans aged 50±70 y was: b 72.90; s.e. 0.13 and, Torres Strait Islanders and Australian Aborigines aged 50±70 y: b 74.67; s.e. 0.37). The same F-test showed that intercepts were different for all groups by gender, age and ethnicity, Samoans displaying the lowest intercepts, and Australian Aborigines the highest (all P < 0.05). Figure 3 illustrates the relations between body weight and resistance in men and women of different agegroups, after adjusting for height and residual age differences.

Discussion

Figure 2 Resistance measurements taken with three different instruments used in the present study.

50±70 y there were no differences between the slopes for the Danes, the Samoans and the Torres Strait Islanders (all P > 0.06), whereas slopes for Australian Aborigines were signi®cantly steeper (all P < 0.004). In the men aged 50±70 y, the slopes for the Danes and the Samoans were similar (P 0.31), but both were ¯atter than the slopes of Torres Strait Islanders and the Australian Aborigines (P 0.002) of the same age. No difference was found between the slopes of the Torres Strait Islanders and the Australian Aborigines (P 0.89). The common slopes for women aged 30±50 y and 50±70 y of Danish, Torres Strait Island and Samoan origin were: b 73.18; s.e. 0.14 and b 73.42; s.e. 0.16 respectively, the slope common to the men

The present study was aimed at identifying differences in associations between body weight and resistance measured by bioelectrical impedance in four different ethnic groups, living in Denmark, New Zealand and Australia, after taking into account differences in height and age. The study showed that associations (slopes) between body weight and resistance were generally constant in the different ethnic groups, except for Australian Aborigines, and only differences in the levels (intercepts) of the regression relationships were apparent. The results indicate that this relationship may involve a certain universality, that is independent of the population speci®city for impedance measurement, as proposed in other literature.5 The results of the similar associations between body weight and resistance in Danes, Samoans and Torres Strait Islanders seem to be in general agreement with studies on gene pools suggesting that differences between individuals within one ethnic group are larger than differences between individuals from different ethnic groups.16 The gene pool of Australian Aborigines has been reported to bear the least resemblance to other ethnic groups,16 a fact that may be re¯ected in the present study. Also, differences in relative sitting height, which are probably also an expression of the genes, between the

Table 2 Multiple regression coef®cients (reg. coef.), standard errors (s.e.) and intercepts for the relationship between resistance and weight in four different ethnic groups of men and women aged more than 50 y or less than 50 y Age 30^50 y

Women Samoans Danes Torres St. Aborigines Men Samoans Danes Torres St. Aborigines

Age 50^70 y

Reg. coef.

s.e.

Intercept

Reg. coef.

s.e.

72.94 73.23 73.13 77.28

0.50 0.17 0.28 0.67

635.3 438.9 940.1 203.9

72.30 73.45 74.72 75.47

0.41 0.17 0.73 0.90

218.7 360.7 61.0 548.9

71.23 72.45 73.11 74.85

0.47 0.13 0.33 0.52

1.7 276.9 137.4 365.4

72.77 72.94 74.32 74.83

0.40 0.14 0.52 0.56

437.5 369.1 478.0 7121.2

All multiple regressional analyses performed with statistical adjustment for differences in height and age. All coef®cients (b) signi®cant (P < 0.01).

Intercept

1089


1090

Figure 3 Relationships between body weight and resistance in men and women of different ages and ethnicity. All multiple regression analyses were performed with adjustment for height and residual age differences.

Australian Aborigines and the other groups may explain the present ®ndings. Australian Aborigines have been found to have the lowest mean relative sitting height of any population in which this parameter has been measured (ranges: 46.9±48.3) followed by groups of African ancestry (ranges: 48.0±54.0).17±20 Subjects of European origin, together with Paci®c Islanders, tend to have intermediate values for relative sitting height, whereas Asian subjects have been reported to have the highest values, although recent reports suggest that this is now changing, at least in the Japanese.21 We did not measure relative sitting height in the present study. However, the noted differences in relative sitting height have general implications for the impedance measurements, as resistance measurements vary as a function of resistor length and cross sectional area.5 The relatively longer limbs of Australian Aborigines, relative to their

trunks, would therefore, lead to higher impedance readings in this ethnic group. Hence the different body proportions may, in part, be responsible for the different slopes seen between weight and resistance in this group. In addition, Australian Aborigines tend to have slender limbs compared to other ethnic groups, a circumstance that also leads to higher impedance readings.19 The present data suggest that, once differences in gender, age and height have been taken into account, a given difference in weight seems to correspond to the same difference in electrical conductance, regardless of a person's ethnic identity. It may be hypothesized that, when differences in height have been considered, with the exception of Australian Aborigines, there may be a constant relation between body weight and body composition (total body water or fat free mass) of different ethnic groups, that depends on gender and


age category only. Although a more in-depth body composition analysis is needed to actually establish the validity of this hypothesis, ®ndings from Norgan22 who compared body fat to BMI relationships of several population samples of different ethnic origin, supports this hypothesis. On the other hand, the signi®cantly different intercepts found between the different groups, indicate that resistance, and thereby possibly fat free mass, at a given level of weight, is highly dependent on ethnicity. In this regard, the fact that the Samoans had a lower resistance for given body weight than, for instance the Danes, suggests that Polynesians are leaner (more fat free mass), for given weight than the other groups studied. This has also been reported by others.1 However, the greater leanness in the Samoans to a given weight does not imply that Samoans, as a population, are leaner. The very high prevalence of BMI > 30 kg/m2,23 suggests that there is considerable overweight in the Samoans, and not all of this overweight may be accounted for by high lean/fat ratios. Australian Aborigines, on the other hand, would be assumed to have more body fat, at lower levels of body weight. However, because of the steeper relationship between weight and resistance in the Australian Aborigines, lower levels of body fat would be assumed at higher weights, as is particularly evident in the women aged 30±50 y. Systematic differences in instrument readings or electrodes do not seem to be responsible for the differences found, since differences were small and did not vary consistently across the range of body weight. However, the slightly lower values obtained with the Accu sensor electrodes, may have resulted in slightly lower intercepts in the relationships for the Danes. Also, differences in measurement conditions (Samoans had not fasted), exercise and/or room temperature at measurement may have played a role. However, although a number of authors report that both fasting, exercise and cold room temperatures tend to lower impedance measurements, others have reported that these conditions are of no importance, or ®nd even reversed effects.5,6 The present study indicates that in subjects of similar age and height, a given weight difference would be expected to result in a similar difference in body composition in Caucasians, Melanesians and Polynesians. At this point, it needs to be noted that the present study is cross-sectional in design, and hence, can not determine causal relationships. Only differences in resistance between groups of subjects, and not individual changes in resistance with weight gain, or age, can be predicted. This has to be considered when interpreting the data. However, some trends are evident. For instance, the statistically different slopes in women and men, and in younger and older subjects could imply that changes in weight in general, give rise to large changes in resistance. This is likely to re¯ect a greater proportional increase in fat mass in women than in men, or in older than in younger people. These ®ndings are consistent with the ®ndings

by Norgan.22 However, the data would also imply that a given increase in body weight is accompanied by the same change in body composition in different ethnic groups, and hence, does not lead to a smaller increase in obesity in, for instance, the Samoans than in the other ethnic groups.24,25 Similarities in the relationship between resistance and body weight (except in Australian Aborigines), may also indicate that there is less of a need for population speci®c equations in epidemiological studies, because only the level (intercept) seems speci®c to the given ethnic group after age and gender have been taken into account. This is supported by other research, for instance the ®nding of a common coef®cient for predicting total body water from Ht2/R in individuals ranging from infants to obese adults,26 or by the large number of studies ®nding that equations, developed from a suf®ciently large population sample, seem to predict body composition with similar precision.5

Conclusions We found large differences in the level of body weight corresponding to a given level of resistance, measured by electrical impedance, between the different ethnic groups. These results suggest that large differences exist in body composition. However, in all ethnic groups studied, except for the Australian Aborigines, a similar relationship was found between body weight, as a measure of body size, and resistance, which may be considered a measure of body composition. These ®ndings suggest that the nature of this relationship contains some fundamental characteristics, and imply, that weight gain may be accompanied by similar changes in body composition, independent of ethnic identity. The present ®ndings therefore may have implications for our understanding of variations in body fat-disease relationships among different ethnic groups.

Acknowledgements

Dr Leigh Ward, University of Queensland, Australia is warmly thanked for constructive comments to the manuscript. The present study was kindly supported by The Ciba Foundation, by providing Berit Lilienthal Heitmann with a travel award to New Zealand. The research with the Danes was supported by grants from The Danish Health Insurance Foundation, The Danish Research Councils and Wedell Wedellsborgs Foundation. Research with Australian Aborigines and Torres Strait Islanders was supported by grants from the National Health and Medical Research Council of Australia. Research with Samoans was supported by a grant from the New Zealand Ministry of Health.

1091


1092

References

1 Swinburn BA, Craig PL, Daniel R, Dent DPD, Strauss BJG. Body composition differences between Polynesians and Caucasians assessed by bioelectrical impedance. Int J Obes 1996; 20: 889±894. 2 Wang J, Thornton JC, Russell M, Burastero S, Heyms®eld S, Pierson Jr RN. Asians have lower body mass index (BMI) but higher percent body fat than do whites: comparisons of anthropometric measurements. Am J Clin Nutr 1994; 60: 23± 28. 3 Cohn SH, Abesamis C, Zanzi I, Aloia JF, Yasumura S, Ellis KJ. Body elemental composition between black and white adults. Am J Physiol 1977; 232: E419±E422. 4 Zimmet P. Kelly West Lecture 1991. Challenges in diabetes epidemiology. From West to the rest. Diabetes Care 1992; 15: 232±252. 5 Rising R, Swinburn B, Larson K, Ravussin E. Body composition in Pima Indians: validation of bioelectrical resistance. Am J Clin Nutr 1991; 53: 594±598. 6 Heitmann BL. Impedance: a valid method in assessment of body composition? Eur J Clin Nutr 1994; 48: 228±240. 7 Kirchhoff M, Schroll M, Kirkby H, Hansen B.S., Sanders S, SjoÈl A, Jùrgensen T & Hansen P.F. Screening I. Danmonica. Part of the MONICA Project (Multinational Monitoring of Trends and Determinants in CVD). CVD Epidemiology Newsletter 1983; 34: 32. 8 Jensen KH, Jùrgensen T. Incidence of gallstones in a Danish Population. Gastroenterology 1991; 100: 790±794. 9 Heitmann BL. Body fat in the adult Danish population aged 35±65 years. An epidemiological study. Int J Obes 1991; 15: 535±545. 10 Hollnagel H. The health structure of 40-year-old men and women in the Glostrup area. Denmark. General design, sampling results and referrals for further medical care. Dan Med Bull 1980; 27: 121±130. 11 O'Dea K. Diabetes in Australian Aborigines: Impact of the western diet and lifestyle. J Intern Med 1992; 232: 103±117. 12 O'Dea K, Patel M, Kubisch D, Hopper J, Traianedes K. Obesity, diabetes and hyperlipidemia in a central Australian Aboriginal community with long history of acculturation. Diabetes Care 1993; 16: 1004±1010. 13 Duffy P, Moris H, Neilson G. Diabetes mellitus in the Torres Strait region. Med J Aust Suppl 1981; 1: 8±11.

14 WHO. Measuring obesityÐclassi®cation and description of anthropometric data. Report on a WHO consultation on the epidemiology of obesity, Warsaw, 21±23 October 1987. Nutrition Unit document, EUR/ICP/NUT 125, WHO: Copenhagen, 1989. 15 Heitmann BL. The effects of gender and age on associations between blood lipid levels and obesity in Danish men and women aged 35±65 years. J Clin Epidemiol 1992; 45: 693±702. 16 Cavalli-Sforza L, Menozzi P, Piazza A. The history and geography of human genes. Princeton University Press, Princeton, New Jersey, USA (1994). 17 Martin R, Saller K. Lehrbuch der anthropologie. 2nd edn, Yena, Fischer: 1959; p 84. 18 Eveleth PB, Tanner JM. Worldwide Variation in Human Growth. Cambridge University Press: Cambridge, 1974, pp 284±411. 19 Gallagher MR. Body build: its in¯uence on anthropometric measurements, fat distribution, body composition and carbohydrate metabolism. MSc Thesis, Deakin University, Malvern, Australia, 1996. 20 Abbie AA. Metrical characters of a central Australian tribe. Oceania 1957; 27: 221±243. 21 Micozzi MS. Cross-cultural correlations of childhood growth and adult breast cancer. Am J Phys Anthropol 1987; 73: 525± 537. 22 Norgan NG. Population differences in body composition in relation to the body mass index. Eur J Clin Nutr Suppl 1994; 48/3: S10±S27. 23 Hodge AM, Dowse GK, Toelupe P, Collins VR, Imo T, Zimmet PZ. Dramatic increase in the prevalence of obesity in Western Samoa over the 13 year period 1978±1991. Int J Obes 1994; 18: 419±428. 24 Gallagher D, Visser M, Sepulveda D, Pierson RN, Harris T, Heyms®eld SB. How useful is body mass index for comparison of body fatness across age, sex, and ethnic groups? Am J Epidemiol 1996; 143: 228±239. 25 Kleerekoper M, Nelson DA, Peterson EL, Wilson PS, Jacobsen G, Longcope C. Body composition and gonadal steroids in older white and black women. J Clin Endocrinol Metab 1994; 79: 775±779. 26 Kushner RF, Schoeller DA, Fjeld CR, Danford L. Is the impedance index (Ht2/R) signi®cant in predicting total body water? Am J Clin Nutr 1992; 56: 835±839.