点击显示 收起
1 From the Childhood Nutrition Research Centre (JCKW) and the Centre for Paediatric Epidemiology & Biostatistics (TJC), Institute of Child Health, London, United Kingdom, and the Department of Computer Science, University College London, London, United Kingdom (PT)
2 Supported by a consortium of UK industry and the Department of Trade and Industry of the United Kingdom. 3 Reprints not available. Address correspondence to JCK Wells, Childhood Nutrition Research Centre, Institute of Child Health, 30 Guilford Street, London WC1N 1EH, United Kingdom. E-mail: J.Wells{at}ich.ucl.ac.uk.
ABSTRACT
Background: Human body shape is a rich source of information about health and the risk of disease. Measuring anthropometry manually is time-consuming, however, and only a few indexes of shape (eg, body girths and their ratios) are used regularly in clinical practice or epidemiology, both of which still rely primarily on body mass index (BMI). Three-dimensional (3-D) body scanning provides high-quality digital information about shape.
Objectives: The objectives of the study were to investigate the relation of shape and BMI and to examine associations between age, sex, and shape.
Design: In a cross-sectional study of 9617 adults (45% male) aged 16–91 y who were participating in the UK National Sizing Survey, body girths and their ratios were obtained with the use of a 3-D body scan. Data on weight and height were also obtained.
Results: BMI was significantly associated with chest and waist in men and with hips and bust in women. In early adulthood, the sexes differed significantly in shape; however, these differences declined with age. Whereas male shape remained highly stable through adulthood, upper body girths, particularly waist, increased in women, but thigh decreased. After adjustment for other girths, waist was significantly and inversely associated with height, particularly in men. Waist varied widely in both sexes for a given BMI value.
Conclusions: Relations between BMI and shape differed significantly between the sexes, particularly in association with age. The inverse association between height and waist in men suggests either a genetic contribution or a link between early growth pattern and predisposition to obesity. The 3-D scans offer a novel approach for epidemiologic research into associations between body shape and health risks and outcomes.
Key Words: Body mass index • body shape • waist circumference • obesity • 3-dimensional body scanning
INTRODUCTION
Measurements of the human body have been used in medical practice and research for centuries. The most widely used measurements are weight and height, which are often combined as body mass index (BMI; in kg/m2) to provide a proxy for nutritional status. BMI is used to categorize underweight (1), normal weight, overweight, and obesity (2), and much research has illustrated associations between BMI and the risk of cardiovascular disease (CVD) and other diseases (3-6). However, abdominal fat shows the strongest association with disease risk, and several studies have reported associations with later CVD and mortality (7-10) and type 2 diabetes (11).
Body shape contrasts with size in providing information on weight distribution. BMI represents a very crude index of shape, whereas waist circumference (WC) gives a clearer indication of relative abdominal shape. More sophisticated information can be obtained from ratios of different body girths, such as the waist-hip (WHR) or waist-chest (WCR) ratio, to act as a proxy for central adiposity. Such measurements are easy to perform and are often highly informative. Studies of both adults and children have shown that the combination of WC or WHR and BMI is a better predictor of CVD risk and mortality than is BMI alone (12-15). Furthermore, clinical focus on shape may achieve greater "connection" with the patient than does BMI, which is difficult for the layperson to calculate and interpret.
Despite awareness of the utility of shape as a marker of health status, little detailed information exists on how age-associated changes in shape influence disease risk. Multiple manual measurements quickly become time-consuming and hence negate their value of being simple and quick. Some manual measurements, such as bust girth in women, may also be considered invasive. Medical research and practice has therefore directed less attention to shape than to internal imaging performed with techniques such as computed tomography scanning or magnetic resonance imaging. Moreover, existing methods do not obtain shape data in digital 3-dimensional (3-D) format, which constrains their utility and integration with other methods.
Highly accurate 3-D photonic body scanners have been developed for the capture of surface topography (16, 17). Initially developed by the clothing industry, they are now ready for exploitation in medical contexts (18). A light-stripe scanner, together with its output, is illustrated in Figure 1. Regional surface scanning of specific parts of the body with the use of either photogrammetric or laser scanning has been emerging in many medical disciplines over recent decades, and fast and accurate whole-body scanners represent a further major advance. We have also reviewed the potential health and medical applications of this technology (JCK Wells, A Ruto, PC Treleaven, unpublished observations, 2006). Here, we have used 3-D data from the recent cross-sectional UK National Sizing Survey of adults (SizeUK), to investigate the associations between age, sex, BMI, and shape. Our aims were, first, to examine the relation between shape and BMI and, second, to identify the interactive effects of age and sex on shape.
FIGURE 1.. A: White-stripe scanner output, showing the projection of concentric strips of light on the body from which shape is reconstructed. Image courtesy of [TC]2, Cary, NC. B: The process of digital reconstruction of shape from raw photonic point-cloud data (left), which leads to surface reconstruction with automatic landmark location (middle) and application of an electronic tape measure (right).
SUBJECTS AND METHODS
Subjects
The SizeUK survey, conducted during 2001 and 2002 by a consortium of retailers in collaboration with University College London, took place in 8 cities distributed throughout the United Kingdom (ie, Birmingham, Cardiff, Edinburgh, Leeds, London, Manchester, Nottingham, and Southampton). The survey was designed to recruit a minimum number of subjects of each sex in the following age groups: aged 16–25, 26–35, 36–45, 46–55, 56–65, 66–75, and 76 y; these subjects were stratified in equal numbers by the 5 conventional UK government categories of socioeconomic status according to education and income criteria.
A 2-stage recruitment strategy was used. Volunteers filled in a short questionnaire to provide contact details, age, sex, and ethnic origin. Recruitment from these subjects was first conducted randomly, and then quota sampling was used to complete cell sizes defined by criteria of age, ethnic origin, socioeconomic status, and geographic area. Seventeen thousand persons registered through the website, 10 000 returned questionnaires distributed via retail outlets and mass mailings to homes, and 20 000 used the customized help line.
Participants gave written informed consent and agreed to allow the use of their anonymized data in statistical analyses. Ethical approval of the study protocol was granted by the Ethics Committee of Great Ormond Street Hospital National Health Service Trust and the Institute of Child Health, London.
Methods
Cell sizes were calculated to estimate mean height in each cell with a 95% CI of 1 cm, equivalent to an SE of 0.5 cm. With a minimum cell size of 188 and adjustment for 3 geographical areas and 7 age categories, the total minimum sample size for each sex was calculated to be 3948.
Participation involved undergoing selected manual measurements and a 3-D body scan and responding to questions on lifestyle and personal profile. For the body scan, subjects wore tight-fitting underwear and adopted a standardized position, using stabilizing handholds to maintain the position. The scanner was a TC2 model ([TC]2, Cary, NC), which projects strips of safe "white light" (in the visible part of the electromagnetic spectrum) onto the body form and records distortions by using 6 nonmoving cameras over a period of 6 s (Figure 1). The scanner can accommodate objects up to 2.1 m in height, 1.2 m in width, and 1 m in depth. The software used was BODY MEASUREMENT SYSTEM software (version 5.3; [TC]2). Three scanners were used for the survey, and they were moved among the 8 cities as required.
The [TC]2 scanner obtains a cloud of raw photonic data points and then reconstructs the skin topography by using computer algorithms. More detailed descriptions of this process have been published elsewhere (19, 20). The technical precision of all measurements was within 0.5 cm. The software automatically extracts >130 measurements from each scan. Measurements used in this analysis comprised manual measurements of weight, height, and head circumference and 3-D measurements of the circumferences of the chest, bust (women only), midupper arm, waist, hips, midthigh, and knee. For limbs, data from the right side of the body were used.
Agreement between manual and scanner measurements of waist and hip was assessed in a subset of 1800 subjects. The mean (±SD) difference between methods was 1.3 ± 4.0 cm and 5.7 ± 2.6 cm for waist and hip, respectively. These mean differences are attributable to 2 factors: first, inconsistency in the exact location of the measurement and, second, the fact that manual measurements involve a degree of compression of the skin. Pearson correlation coefficients between methods were 0.96 for waist and 0.97 for hip, which indicated high consistency in ranking between techniques, despite the mean bias.
For examination of the effects of age on shape, the data for each sex were grouped in the following age categories: 16–20, 21–30, 31–40, 41–50, 51–60, 61–70, and 71 y old. The group aged 16–20 y was separated from older groups because adolescents may have greater variability in shape as a result of their continuing growth. Subjects with missing data for weight and height were excluded from the analyses.
Data on BMI were used to define overweight (25 BMI < 30) and obesity (BMI 30) (2). Linear regression analysis was used to evaluate the association between the prevalence of overweight and obesity and age. To assess variability in WC for a given BMI value, waist z scores were calculated for each sex. The range of waist z scores was then determined in each sex for persons with BMIs between 24.00 and 24.99. Correlation analyses between shape variables were conducted. Linear trends of shape outcomes in relation to age were then examined. Mean (±SD) values were provided for each component of shape for each age group, and the P value and regression coefficient were given for the effect of age as a continuous variable in the whole sample. Finally, models were constructed to investigate the relation between age, height, and components of shape and WC. Multiple regression analysis was used to investigate the contributions of height, age, and shape to weight in each sex, and all data were natural log transformed to look for proportional associations. Graphic analysis was used to illustrate differences between the sexes in the effects of age on shape by using WHR, WCR, and ratios of waist to thigh and bust as appropriate and the ratio of thigh to arm. For comparison with other populations, we used average data on women from the 1951 National Sizing Survey based on manual measurements of 4995 women aged 18–64 y (21) and the 2003 US Sizing Survey based on 3-D scanning and with a study design identical to that of SizeUK.
RESULTS
Complete data were available for 9617 persons (45% were male), and those persons were used in subsequent analyses. The average BMI for males was 25–26 in all age categories, which is equivalent to the cutoff for overweight (2). The average woman was in the normal BMI range for the age categories <21 y and 21–30 y, but was at or over the threshold for overweight for all subsequent age categories. The proportion of those who were overweight or obese rose significantly (P < 0.0001) with age in each sex, but it rose more steeply in women (age x sex interaction: P < 0.0001) (Table 1). Because there were markedly fewer overweight women than men in younger age groups, this steeper increase resulted in approximate parity between the sexes from the 51–60 age category upward. The proportion of women with BMI <20 declined with age, from 23.7% (16–20 y) to 3.3% (71 y), whereas the proportion of men with BMI <20 was relatively stable [8.5% (16–20 y), 6.2% (71 y)]. Digital girth measurements stratified by sex and age, along with manual measurements of weight and height, are given in Table 2. In men, height, head circumference, and thigh decreased with age, whereas knee and waist increased. In women, all dimensions increased with age except height, thigh, and head, and the trends were stronger than in men.
View this table:
TABLE 1. Prevalence of overweight and obesity stratified by age and sex1
View this table:
TABLE 2. Raw values for anthropometry and shape stratified by sex and age1
A correlation matrix for weight, height, and all girths except bust is shown in Table 3. In general, girths were highly correlated with each other and with weight in both sexes; however, head girth showed the weakest correlations with other variables. Correlations between height and girths ranged from 0.18 to 0.35 in men (P < 0.001 in all cases) except that between height and waist, which was 0.04 (P < 0.05). Likewise in women, girths corresponding to fat depots had relatively weak correlations with height, ranging from 0.04 for waist to 0.21 for thigh (all: P < 0.05).
View this table:
TABLE 3. Correlation matrix for weight, height, and body girths in men and women1
Results for the log-log regression of weight on components of size and shape, after adjustment for age, are shown in Table 4. The combination of height and circumferences explained 91.7% and 94.8% of the variance in weight in men and women, respectively. Age (men) or the combination of age and age squared (women) was significant in these models, but it did not change the r2 value by even 0.1% and had negligible effect on the other regression coefficients. In both sexes, height was the strongest predictor of weight; however, whereas the 2 next strongest predictors in men were waist and chest, those in women were hip and bust. Thus, men and women differ in the principal aspects of their shape that account for variability in weight.
View this table:
TABLE 4. Multiple log-log regression analysis of weight on components of size and shape, ranked in terms of significance1
Results for the log-log regression of waist on components of size and shape are shown in Table 5. The models presented had r2 values of 82.7% and 83.4% in men and women, respectively. In both sexes, hip, arm, and knee circumferences were positively related to waist, whereas height and thigh circumferences were negatively related. Age was significant in both models. However, the strengths of the associations differed between the sexes: the effects of height and thigh were much stronger in males, and the influence of age was greater in females.
View this table:
TABLE 5. Multiple log-log regression analysis of waist on components of shape, ranked in terms of significance1
The log-log regression of height on shape variables is shown in Table 6. In men, all aspects of shape were positive predictors of height except waist, which was inversely associated and was the strongest predictor. Three girths that relate to body fatness (thigh, arm, and waist) were also negatively associated with height in women but much more weakly so than in men.
View this table:
TABLE 6. Multiple log-log regression analysis of height on components of shape, ranked in terms of significance1
For all subjects with BMI between 24 and 25, waist ranged from 75.5 to 109.8 cm in men and from 72.7 to 113.6 in women. These ranges were equivalent to ranges of 3.5 z scores in waist in both sexes, which showed the wide range in WC for a narrow range of BMI.
Key measurements for women from this survey and from equivalent surveys conducted by using manual measurements in the United Kingdom in 1951 and with 3-D scanning in 2003 in the United States (using a study design identical to that of SizeUK) are shown in Table 7. In each survey, the average woman was in her mid-30s. The average UK woman from the current survey was 3 kg heavier than her counterpart in 1951 and had a 16-cm larger waist, and the average US woman weighed 6 kg more than the average UK woman, despite being 3 cm shorter.
View this table:
TABLE 7. Comparison of anthropometric values for an average woman from the 1951 and 2001 United Kingdom (UK) surveys and the 2003 US survey
Sex differences in the effect of age on shape are shown by using ratios between different girths (Figures 2 and 3). Figure 2 shows that the WHR and WCR increased only moderately in men after 30 y and started to decrease again from 70 y. In women, in contrast, the WHR, WCR, and waist-bust ratios all increased more steeply and continue to rise even from age 60 y. Figure 3 shows that the waist-thigh and thigh-arm ratios were fairly stable across the life-course in men, whereas in women the waist and arm increased systematically relative to the thigh, which implies a tendency to deposit fat both upwards and centrally. This tendency also appears to be in part a redistribution, because thigh circumference declined in women from around age 50 y. Overall, sexual dimorphism in shape was greatest in young adults, and lessened with age. Figure 3 shows that sexual dimorphism in some aspects of shape had vanished by the 8th decade.
FIGURE 2.. Mean (±SE) differences in the cross-sectional abdominal body shape ratios in 4316 men (A) and 5228 women (B), showing relative stability across age categories in men, but increases in waist relative to bust, chest, and hips in women. n = 289 men and 208 women per age group.
FIGURE 3.. Mean (±SE) differences in the cross-sectional distribution of weight between abdomen and limbs with age, as shown by waist-thigh and thigh-arm ratios; relative stability is seen in men (n = 4316) but waist and arm increased relative to thigh in women (n = 5228). n = 289 men and 208 women per age group.
DISCUSSION
The shape comparison of average women obtained from 3 different surveys gives an indication of the profound changes that have occurred in anthropometry over the last half-century. The average UK woman has increased substantially in weight and body girths since 1951 (21), gaining 16 cm in WC despite being only 4 cm taller. The average contemporary US woman has even greater waist and weight than her UK counterpart, despite being 3 cm shorter. As is well recognized, the US population began the trend toward obesity earlier than did European populations, and, without progress in obesity prevention, the UK population is likely to continue to expand in weight and girths.
Our investigation highlights the limitations of BMI for assessing the health effect of such secular trends in weight. Whereas BMI is widely used as an indicator of CVD risk, our analyses suggest that BMI conveys different information about men and women. The 2 main factors associated with weight in men after adjustment for height are chest and waist, whereas in women they are hip and bust. In each sex, both physique and fatness appear to be important, but the sexes differ in the body region of most influence. We suggest that chest in men but hips in women reflect physique, whereas waist in men and bust in women reflects fatness. In both sexes, for those within a narrow BMI band on the borderline of overweight, subjects varied widely in WC, which is indicative of wide variations in CVD risk.
Moreover, BMI is insensitive to age-associated changes in the distribution of body weight, which also differ markedly between the sexes. In both sexes, BMI increased steadily with age group, but, whereas older women had greater upper-body girths than did younger women, such a trend was not observed in men. These findings provide further evidence that BMI is insufficient as an index of nutritional status for use in monitoring the risk of CVD and related diseases. This perspective is supported by reports showing that secular trends in children's WC are greater than those in BMI (22-24).
Our findings also illustrate important differences between the sexes in the relation between WC, a measurement previously shown to be highly correlated with visceral fat (25), and other components of size and shape. First, women showed a negative association between thigh and waist, but the equivalent association was not significant in men. This suggests a tradeoff between the 2 fat depots in women, corresponding to the "apple" or "pear" type of fat distribution. Second, after adjustment for other girths, waist was strongly negatively associated with height in men, whereas the association was much weaker in women. When height was predicted from multiple shape measurements in men, all components of shape showed a positive association except waist, which was strongly negatively associated. Such an association may derive from genetic factors or from a developmental link between growth patterns and obesity risk. An association between stunting and central body fatness has been observed in other populations, eg, Brazil, where stunted children have been observed to have impaired fat oxidation that predisposes them to excess central weight gain (26). Why this association should be stronger in men is not clear, but the question merits further investigation.
Associations of shape with age were significantly stronger in women than in men. Whereas men were significantly heavier with increasing age category, the only dimensions of shape to show accompanying linear trends were WC and thigh circumference. The directions of these effects were opposing, such that waist increased whereas thigh decreased. However, the differences were small: the difference in waist between the highest and lowest age groups was only 3.3 cm. In contrast, women showed much stronger associations between age and shape. The difference between the highest and lowest age groups was 13.6 cm, which is 4 times the difference in men. As shown in Figures 2 and 3, ratios between different body regions—girths of the arms, chest, bust, waist and hips—were greater in older women, whereas thigh circumference was reduced in those aged >50 y. This association of shape and age implies a shift in fat toward the upper body with age and, within the upper body, to a preferential distribution around the waist rather than the arm and bust. In premenopausal women, overweight and obesity are associated with a lower risk of breast cancer (27, 28); however, in postmenopausal women, overweight increases the risk (29). The redistribution of weight to the upper body with increasing age, which increases the bust, may contribute to this changing association between BMI and the risk of cancer.
Our analyses have some limitations. First, although the sample derived from 8 different urban regions throughout the United Kingdom, the subjects are not necessarily representative of UK adults. Second, the data are cross-sectional and do not show how individual subjects change over time. For example, the negative trend of height with age is probably due not only to the effects of aging on posture but also to the effect of secular trends in height during the 20th century. Nevertheless, the dataset is large and capable of identifying important biological associations between sex, shape, and age.
The association of different variables of shape with different diseases illustrates the potential that monitoring changes in shape has for informing clinicians about the risk of multiple outcomes. Several studies have linked body shape with the risk of components of the metabolic syndrome (30-32), as well as other health outcomes (33, 34). A single 3-D body scan, repeated at intervals during adult life, may provide sensitive information about progression in risk and allow improved targeting of those most at risk.
Body scanning represents a sophisticated technology in this context, because the digital nature of the data facilitates rapid archiving and retrieval of successive scans, and software also allows the superimposition of repeat scans to highlight shape changes. Further software development will allow the evaluation of more sophisticated 3-D shape indexes, which will supercede girths that are one-dimensional and thus provide relatively crude shape outcomes. 3-D scanning is particularly valuable because of its suitability for use in children as well as adults, which facilitates both life-course research and long-term clinical monitoring. Thus, we suggest that this technology is capable of making a major contribution to health monitoring in the general population.
In conclusion, the current study is the first to use sophisticated measurements of body shape in a large survey of adults. We found that relations between shape and BMI differed between the sexes, particularly in association with age. BMI does not reflect variability in WC, an index of shape closely linked with cardiovascular health, and it is insensitive to age-associated changes in weight distribution. 3-D body scanning represents a cheap and noninvasive technology that is appropriate for use in the monitoring of CVD risk.
ACKNOWLEDGMENTS
PT planned and directed SizeUK; the analyses were conceived by all authors and conducted by JCKW and TJC; JCKW wrote the draft of the manuscript, and all authors contributed to revisions. PT is a Director of Bodymetrics Digital Fashion Technology, a company that provides body-related software and services to the clothing industry, such as the sale of the SizeUK data, analysis of body data for clothing, custom-made fit mannequins, made-to-measure clothing and virtual try on services. JCKW and TJC had no personal or financial conflicts of interest.
REFERENCES