点击显示 收起
1 From the Occupational Physiology Division, US Army Research Institute of Environmental Medicine, Natick, MA, and the Division of Human Biology, Department of Community Health, Wright State University School of Medicine, Dayton, OH.
2 The opinions and assertions in this article are those of the authors and do not necessarily represent the views and policies of the Department of the Army. 3 Address reprint requests to KE Friedl, PO Box 1779, Frederick, MD 21702. E-mail: karl.friedl{at}det.amedd.army.mil.
ABSTRACT
Background: Healthy young women who engage in an exercise program may lose fat that is not reflected in body weight changes because of concurrent gains in fat-free mass (FFM).
Objective: This study addressed the question of how well anthropometry-based predictive equations can resolve these changes.
Design: Several widely used skinfold-thickness- or circumference-based equations were compared by using dual-energy X-ray absorptiometry to study 150 healthy young women before and after 8 wk of Army basic combat training (average energy expenditure: 11.7 MJ/d).
Results: Women lost 1.2 ± 2.6 kg fat (
Key Words: Anthropometry • weight reduction • body composition • dual-energy X-ray absorptiometry • body circumferences • skinfold thicknesses • exercise • generalized equations • women • military personnel • physical training • fitness program
INTRODUCTION
Young men and women in a physical training program can show beneficial changes in their body composition, such as reductions in body fat and increases in muscle mass (1). Body weight is not a suitable measure for assessing these changes because an increase in weight due to an increase in fat-free mass (FFM) can be misinterpreted as an increase in body fatness. To correctly identify changes and to provide positive and correct feedback, it is necessary to be able to assess progress without the use of expensive equipment or technically complex procedures. For men, fat loss is well represented by a simple decrease in the abdominal girth measurement, even for the leanest men (2). For women, there is no equivalent single site of fat deposition that adequately represents fat loss; changes occur in the thighs, hips, abdomen, and upper arms (3). Skinfold thicknesses may also be inadequate reflections of total body fat change because visceral fat cannot be measured and limb circumference measurements are complicated by concomitant changes in the underlying muscle mass. A simple method for assessing fat loss would be useful to the many individuals who engage in fitness programs. This has particular importance to women in occupations in which weight or fat standards are imposed to motivate fitness and good nutrition habits for the prevention of obesity. For example, in the US Armed Forces, all members are weighed at least annually to ensure compliance with weight control regulations (4). Individuals who exceed an established weight-for-height limit are then assessed for percentage body fat by using circumference-based prediction equations. Overfat service members are encouraged to achieve standards through a reasonable program of exercise and dietary modification, and they risk separation from the military if they do not make satisfactory progress toward meeting the standards (5, 6).
Anthropometrically based prediction equations have not been rigorously tested in men or women for their ability to detect body-composition changes. Several reports commented on the inadequacy of various skinfold-thickness equations to predict changes in body composition, whether modest (7, 8) or large (9, 10). Abdominal circumference is responsive to reduction in fat energy stores in men, and neck circumference provides an adjustment factor for body size or lean mass changes (2, 11). However, there is no such simple set of predictive measurements for women, whose principal sites of fat deposition and responsiveness to fat mobilization vary according to genetics, hormonal influences, and the reproductive life cycle (12, 13). Because typical fat depots in women (ie, hips, thigh, abdomen, and upper arm) may not be equally responsive to exercise, the measurement sites that best predict adiposity in an individual at one point in time may not even be the same sites that best detect change in fat weight.
The goal of the present study was 2-fold. First, changes in body composition during 8 wk of intense exercise training as part of the basic combat training (BCT) program were determined in healthy women. Factors such as initial body fat, previous pregnancy, and ethnicity were investigated in relation to the changes. Second, the ability of existing anthropometry-based equations for women to detect changes in percentage body fat in these women was evaluated.
SUBJECTS AND METHODS
Experimental design
The research protocol for this study was approved by the Human Use Review Committee of the US Army Research Institute of Environmental Medicine, Natick, MA, and by the Human Use Review Office of the US Army Medical Research and Materiel Command, Fort Detrick, Frederick, MD. The study subjects were volunteers from an all-woman basic training unit at Fort Jackson, SC, in the spring of 1993. Prospective study subjects were informed of the purposes, risks, and benefits of the study, and those who volunteered to participate gave their written consent before the study.
Data were collected at the start and the end of the training, with 56–59 d between measurements. BCT is 8 wk in duration and includes orientation of new recruits to soldier skills with an intensive period of physical training. Average energy requirements during training totaled 11.7 MJ/d, on the basis of careful assessments of intakes of a subsample of subjects (n = 40) and body energy balance calculations. Additional data from the parent study, of which this study was one part, is available in a technical report (14, 15).
Study subjects
Complete data were available for 150 women before and after BCT. Descriptive characteristics of the subjects are summarized in Table 1. The mean (±SD) age of the participants was 21.4 ± 3.6 (17–33) y; 57% identified themselves as non-Hispanic white (n = 87), 25% as black (n = 41), and 13% as Hispanic (n = 19). On arrival at BCT, 19.1% of the women were smokers; all were required to quit abruptly because smoking is not permitted during the 8 wk of BCT. At the start of BCT, 28% of the women exceeded Army female body-composition standards. These standards involve screening tables based on body mass index (BMI; in kg/m2) and then an upper limit of percentage body fat predicted by the Army circumference-based equation. For the youngest women (18–20 y of age), the weight screen threshold that prompts a percentage body fat assessment is a BMI value of 23.5; the upper limit for percentage body fat is 30%. For the next age group (21–27 y), these values are 24.3 and 32% body fat, respectively. Many participants in this study were unaware of their status with respect to Army body-composition standards, and 18% of recruits still exceeded the Army standards by the end of BCT. Soldiers who exceed the standard are required to lose weight to achieve the body fat standards within 6 mo of their arrival at their first unit of assignment.
View this table:
TABLE 1. Body composition and anthropometry measures for 150 young women before and after 8 wk of basic combat training1
Menstrual history indicated that only 15% of the soldiers had ever missed a period unrelated to pregnancy; in general, women who are not eumenorrheic are rejected from service on the basis of medical standards of fitness (16). Thirty-five percent of the women had been pregnant previously (n = 54), and 23% were using contraceptive steroids, primarily norethindrone, levonorgestrel, and norgestrel combined with low-dose ethinyl estradiol, as well as levonorgestrel implants (n = 4). BCT did not suppress or lengthen menstrual cycles, on the basis of daily self-reported menses and the absence of a change in pre-, mid-, and posttraining mean serum estradiol and progesterone concentrations (17).
Anthropometry and body composition
Anthropometric measurements were performed before and after BCT. Height was measured with a calibrated anthropometer (GPM no. 101; Seritex Inc, Carlstadt, NJ). The women were dressed in physical training uniforms (T-shirts and shorts) during the measurement sessions. Body weight was measured by using electronic balances calibrated onsite and accurate to 0.04 kg. Measurements included 15 circumferences (neck, shoulders, chest, abdomen at the natural waist, abdomen at the navel, abdomen at the iliac crest, hips, thigh, biceps flexed, biceps, forearm, wrist, knee, calf, and ankle) and 9 skinfold thicknesses (chest, subscapular, triceps, biceps, midaxillary, suprailiac, abdominal, thigh, and calf). Eight circumference and 6 skinfold-thickness measurements are reported here; more detail is available in the technical report (14). Circumference measurements were taken in duplicate by using a 0.25-in (0.64 cm) wide fiberglass tape measure; skinfold thicknesses were made in triplicate by using block-calibrated Harpenden calipers (British Indicators, Ltd, London). All measurements were made on the right side of the body by using techniques described in the Anthropometric Standardization Reference Manual (18). Total body composition was determined by dual-energy X-ray absorptiometry (DXA; DPX-Plus, software version 3.6; LUNAR Corporation, Madison, WI). Total body fat mass (TBF) and FFM were calculated from DXA-assessed percentage body fat and scale-assessed body weight.
Statistical analysis
Descriptive statistics including means, SDs, minimums, and maximums were computed for each body-composition variable from DXA and for anthropometric measurements for all ethnic groups combined before and after BCT. Descriptive statistics for the changes in these variables before and after BCT were also computed. Significant tests for these changes were performed by using paired t tests. Analysis of variance was performed to evaluate ethnic group differences and the effects of previous pregnancy on the baseline measures and the changes. Changes in percentage body fat, TBF, and FFM were regressed against baseline percentage body fat to evaluate the effects of baseline body fat on subsequent changes in body composition. Five commonly used prediction equations for women were used to estimate percentage body fat (19–23). We also formulated an equation for predicting percentage body fat from selected anthropometric measurements in the present study. Detailed procedures for deriving this equation were described elsewhere (24); briefly, all possible subsets of regression with selection criteria: minimum root mean square errors, minimum Cp (an index of the appropriate number of independent variables in an equation), and significant regression estimates at P < 0.05 were used.
The sensitivity and specificity of these equations were calculated. The sensitivity is the conditional probability of detecting changes in percentage body fat with the equation, assuming there are significant changes in the observed data; specificity is the conditional probability of no changes in percentage body fat as detected by the equation given there are no significant changes in the observed data.
RESULTS
Physical characteristics at the start of BCT
The mean body weight, height, and other anthropometric data for this sample of women at the start of BCT are shown in Table 1. There were 54 women (35%) who were pregnant previously. There were no significant differences in body composition, circumferences, or skinfold thicknesses with previous pregnancy except that women who had been pregnant (age: 22.7 ± 1.5 y) tended to be older than women who had never been pregnant (age: 20.6 ± 2.9 y) and they had a smaller mean thigh skinfold-thickness measurement (P < 0.05). Ethnic group differences in anthropometric and body-composition variables for white (n = 87), black (n = 41), and Hispanic women (n = 19) were compared by analysis of variance. No major differences were found except for height and FFM (blacks > non-Hispanic whites > Hispanics). These 2 measures remained significantly different at the end of training.
Changes as a result of BCT
In general, the fat variables decreased during BCT whereas the FFM variables increased (Table 2). Weight increased significantly (P < 0.05) during BCT by 0.8 ± 2.9 kg. Values for BMI increased significantly by 0.2 ± 1.1. The increase in weight during BCT was due mainly to an increase in FFM (2.5 ± 1.5 kg), with a smaller mean decrease in TBF (1.2 ± 2.6 kg). Evaluation of the physical training effects based only on body weight could be misinterpreted because the conventional reason for an increase in weight is an increase in fat.
View this table:
TABLE 2. Changes in body composition and anthropometric measures of 150 young women after 8 wk of basic combat training
BCT also resulted in a decrease in waist-to-hip ratio of 0.01 ± 0.03 (P < 0.05). There was a significant increase in some circumferences after BCT and a decrease in skinfold thicknesses at various sites. When data only from those women who lost fat weight were examined, the abdominal circumference measurements stood out with significant decreases compared with significant increases in most of the other circumference measurements (data not shown).
Regressions of changes in weight, TBF, and FFM on baseline percentage body fat were performed separately (Figure 1). The recruits with the highest percentage body fat at baseline lost the largest amount of fat weight (r = 0.47), but there was no relation between increase in FFM during BCT and initial percentage body fat. The recruits with the highest percentage body fat at baseline also lost the largest amount of weight (r = 0.39). Effects of a previous pregnancy on the changes in body composition after BCT were examined and no significant differences were noted in women who had been pregnant.
FIGURE 1. . Regression of changes in weight, total body fat (TBF), and fat-free mass (FFM) during basic combat training versus initial percentage body fat (%BF). Women who started the training with between 25% and 35% body fat tended to lose body fat but gained weight because of the universal gain in FFM. RSME, root mean square error.
Application of existing equations for percentage body fat
Selected existing prediction equations were applied to the data and the SDs of the residual were computed. The equations used were those developed for the Marine Corps (19), the Navy (20), and the Army (21) on the basis of samples from each of these respective services and the frequently cited equations of Durnin and Womersley (22) and Jackson et al (23) (Table 3). The resulting mean residuals and residual SDs for each equation, by ethnic group, are presented in Table 4. There were no differences by previous pregnancy status (data not shown). Equations that used body density as the dependent variable were converted to percentage body fat by using the Siri equation (25). For all ethnic groups combined, the residual SDs ranged from 1.9% to 2.8% body fat. The Marine equation yielded the largest residual SD. All equations had a tendency to underestimate percentage body fat except for the equation of Durnin and Womersley (Figure 2). The equation of Jackson et al had the second largest residual SD. The equations of Durnin and Womersley and of the Navy performed significantly better than did the other equations (Table 5). When the residual SDs were separated by whether the women had previously been pregnant, the Navy equation performed significantly better than did the other circumference-based equations (data not shown).
View this table:
TABLE 3. Anthropometrically based prediction equations tested1
View this table:
TABLE 4. Cross-validation of existing equations with measurements from 150 women at the start of basic combat training1
FIGURE 2. . Percentage body fat (%BF) predicted by the Marine (19), Navy (20), Army (21), Durnin and Womersley (22), and Jackson et al (23) equations versus that measured by dual-energy X-ray absorptiometry and percentage body fat predicted by the respective equations versus the residuals.
View this table:
TABLE 5. Comparison of the performance of the existing equations for prediction of percentage body fat within ethnic groups and for the combined set of young women at the start of basic combat training1
Testing the sensitivity of the equations in detecting changes
Ideally, a prediction equation should be sufficiently precise to detect changes in the predicted variables over time. The sensitivity to detect changes in percentage body fat was evaluated and is summarized in Table 6. All the equations had relatively low sensitivity (Figure 3). Among the observed significant changes, the proportions of changes detected by the existing equations were between 12% and 55% for body fat. The specificity was in general greater than the sensitivity. The values for specificity ranged from 48% to 86% for body fat. There were no effects of a previous pregnancy on the sensitivity and specificity of these prediction equations. Even the equation developed from the data collected at the start of BCT did not have better sensitivity or specificity than some of the other existing equations for detecting change during BCT.
View this table:
TABLE 6. The sensitivity and specificity of the anthropometric equations in predicting changes in percentage body fat of 150 young women at the end of 8 wk of basic combat training
FIGURE 3. . Predicted versus measured (by dual-energy X-ray absorptiometry) change (post - pre) in percentage body fat (%BF).
DISCUSSION
Body weight is not an effective or accurate way to assess the results of a physical training program. Several women either gained weight or had little weight change after 8 wk of BCT, even though they lost body fat. This effect was due partly to the results of a desirable change: an increase in FFM that averaged 2.5 kg. The fattest women lost the greatest amount of fat and weight. Women who had between 25% and 35% body fat were the ones most likely to register little or no change in weight when they may have actually lost several kilograms of fat. Regardless of their initial percentage body fat, all of the women tended to gain FFM. There were no ethnic differences in the predictability of body composition or changes in body composition.
Direct measures of percentage body fat and FFM are currently impractical for widespread use in screening for general health and fitness standards. Indirect or clinical methods usually rely on estimation of body composition from easily measured variables such as circumferences or skinfold thicknesses and use of prediction equations. Anthropometric measures provide convenient estimates of body fat in individuals; however, these predictions have not been developed or validated for prediction of change in body composition. Furthermore, body fat prediction from one or more regional measurements of fat deposition is more difficult in women than in men (in whom the abdominal site provides a single reliable focus). In a previous sample of 237 women, abdominal circumference alone characterized the fattest women (>34% body fat), but average proportions for all other women remained similar between abdomen, hips, thigh, and biceps (3). This suggests that abdominal circumference in women distinguishes only those above some threshold of body fat or has greater individual variation than does the direct relation between adiposity and abdominal circumference in men. Note that in the current study, the best circumference equation for body fat estimation and prediction of change in body fat assesses both waist and hip circumferences (20). It is convenient that these are the 2 sites that are most susceptible to environmental (ie, diet and exercise) changes, compared with the upper arm and thigh fat depots, which are more influenced by pregnancy and postpartum hormones (3).
Cross-validation of the existing equations with the data in this study indicated relatively large errors in the prediction of percentage body fat ranging from between 3.5% and 8.1%. This implies that exercise training would have had to produce a minimum change in percentage body fat of 3.5–8.1% (depending on the equations applied) to register a change with use of these equations. Our results also indicated a lack of sensitivity and specificity of these existing prediction equations in detecting changes in body composition. Change in the criterion method for body-composition assessment may account for some of the error in body fat estimation but should not play an important role in prediction of relative changes (ie, prediction of body fat change with use of these equations is not likely to be better when compared with underwater weighing instead of DXA). In fact, the use of DXA as the criterion method would be expected to reduce some of the prediction variability of these equations because both DXA and the circumference measurements are free of the variability produced by differences in bone density. This is not the case for the 2-compartment-model interpretations of hydrodensitometry, against which each of the equations was originally calibrated, and hydrodensitometry measurements also produce body fat estimates with the lowest precision (26).
Ethnic group differences can potentially affect the performance of the existing equations. Nonetheless, in the present study, we did not find ethnic group differences in anthropometry and body composition in our sample. Statistical analysis of the existing equations was also performed separately for each ethnic group and for all ethnic groups combined. The results of cross-validation for the combined set were similar to those for the ethnic groups separately. In the present study, the number of blacks and Hispanics was small and this may have contributed to our findings of no major ethnic group differences.
Body-composition assessment in the US military is a 2-tiered process. Personnel who exceed the weight-for-height limits are required to receive a second evaluation. The second evaluation involves an individual percentage body fat estimation made with use of prediction equations for percentage body fat derived from underwater weighing. Body fat standards have been enforced in the military services since the early 1980s (5), when the Marine Corps (19), the Navy (20), and the Army (21) each developed their own percentage body fat equations. For men, each of these equations measured the abdominal circumference at the umbilicus and used the neck circumference as an adjustment for body size or FFM (11). For women, the equations were considerably more complex and each of the military services produced their own equations that relied on different circumference measurements to capture fat stores, emphasizing thigh and abdomen (19), hips and abdomen (20), or hips and body weight (21). Each of these equations also had factors that adjust for body size or FFM. The skinfold-thickness equations of Durnin and Womersley (22) were used by the Army as an interim method while a circumference-based equation was being developed. This experience showed the difficulty of standardizing measurements by using a method with any complexity in technique such as skinfold-thickness measurement, although standardization of seemingly simple circumference measurements has also proven to be a challenge, both in tape handling and proper landmark identification. Currently proposed revisions of the fitness and body composition directive for all military services require adoption of an equation that uses the measurement sites used in the equation developed by the Navy; the US Marine Corps has already adopted an equation using the recommended measurement sites.
The present study highlights the challenge faced by the military in evaluating body composition in servicewomen. Even if BMI proved to be a satisfactory estimate of adiposity in women (because of smaller variability in lean mass compared with hypertrophic responses to exercise in men), modest success in weight loss programs would not be adequately reflected in body weight changes. Methods for assessing fat change, such as the current predictive equations, increase the likelihood of detecting actual fat weight loss. Some of these circumference-based equations may categorize individuals with excess fat reasonably well but may not adequately predict compliance with standards after a period of fat loss. This requires, at a minimum, that changes in either weight or fat prediction be accepted as measures of progress in fat loss, and marked changes should not be expected within short intervals (ie, assessments repeated in <6–8 wk are unlikely to reflect actual progress).
The present findings suggest that, for healthy young nonobese women, 1) a change in body weight alone is not a good measure to assess the effectiveness of physical training programs because a lean mass gain can mask fat weight loss (but reduction in body weight would strongly suggest fat weight loss), 2) the circumference-based equation of Hodgdon and Beckett (Navy) and the skinfold-thickness-based equation of Durnin and Womersley are the best of the equations tested for predicting body fat in this sample of women, and 3) none of the equations tested were very accurate in detecting changes in percentage body fat in women after physical training.
ACKNOWLEDGMENTS
We are grateful to the young soldiers whose willing participation made this study possible. We thank Bob Mello and Sabrina Carson for their assistance in data collection.
REFERENCES