Body Composition Comparisons in Female Athletes

1

Body Composition Comparisons in Female Athletes

Body Composition

EVALUATION OF THREE SKINFOLD EQUATIONS BY USING THE BOD POD AS THE CRITERION IN CAUCASIAN FEMALE ATHLETES

JENNY FRUTH1, AMY MORGAN1, LYNN DARBY1,DAVID TOBAR1

1School of Human Movement, Sport, and Leisure Studies/

Bowling GreenStateUniversity, Bowling Green, OH, USA

ABSTRACT

Fruth J, Morgan A, Darby L, Tobar D. Evaluation of Three Skinfold Equations using the BOD POD as the Criterion in Caucasian Female Athletes. JEPonline 2008;11(1):28-37. The use of valid skinfold (SF)equations for female collegiate athletes allows exercise physiologists to accurately and easily determine players’ percentages of body fat (%BF) in a field setting. The BOD POD® has been shown to be a valid measurement of body composition for female athletes when compared to underwater weighing (UWW) and dual-energy x-ray absorptiometry (DXA), but has not been used as a criterion measure to validate skinfold regression equations. The purpose of this investigation was to compare %BF estimated from three skinfold regression equations to %BF measured using the BOD POD®for female athletes from basketball, soccer, softball, swimming, and volleyball. Seventy-five collegiate athletes (mean ± SD) [age = 19.3 ± 1.2 yrs, height = 166.7 ± 6.8 cm, body mass = 68.5 ± 9.4 kg] were measured. Each participant’s %BF was assessed via the BOD POD® and three skinfold equations based on skinfold measurements at the triceps, suprailiac, abdomen, and thigh. Significant differences were noted between all body composition techniques except one (BOD POD® vs.skinfold equation for the general population). Significant effect sizes were most meaningful between the BOD POD® and the skinfold equation for female athletes based on UWW, suggesting that this skinfold equation underestimates %BF. Therefore, skinfold equations for the general population or developed from DXA analysis are best for estimating %BF in female collegiate athletes when the criterion measure is %BF from the BOD POD®.

Key Words: Percent Body Fat, Body Composition, Dual-Energy X-Ray Absorptiometry (DXA), Hydrostatic Weighing, Densitometry

INTRODUCTION

Body composition testing is important for athletes as an indicator of their fitness and health (1). Using body composition measurements one can monitor preseason and postseason changes, and also track weight loss or gain as indicators of poor health or eating disorders (1,2). The monitoring of body fat is especially important in female athletes because of their increased risk for developing symptoms of the female athlete triad (2,3). As the result of disordered eating behaviors, amenorrhea and osteoporosis may be developed at an early age in female athletes. Early identification of these potentially serious disorders is crucial for an athlete to prevent harmful diseases or injury (2,3).

There are many methods available for measuring body composition. The most common methods used to predict percentage body fat (%BF) are two-compartment models that divide the body into fat mass (FM) and fat free mass (FFM) (1). The optimal method to measure body composition depends on the end sought and the practical possibilities for measurement (4). In athletics, tests for predicting %BF must be valid, reliable and readily accessible for coaches, athletic trainers, or exercise scientists to use with multiple athletes in a short amount of time (5).

Hydrostatic (underwater) weighing (UWW) has long been considered the “gold standard” in body composition testing (6). UWW has been used in multiple investigations either being compared to another form of body composition measurement or being used as a criterion measure (7). Despite its widespread use, UWW has many limitations such as its expensive equipment, time-consuming measurement (e.g., 30-60 minutes depending on residual volume measurement), and the necessity of submersing the head under the water (6).

Dual-Energy X-ray Absorptiometry (DXA) is another form of body composition analysis. DXA has been deemed a potential replacement of UWW as the gold standard because of its capability of testing multiple body compartments compared to UWW’s two-compartments (7,8). There are also limitations to using this method that include equipment and testing being expensive andavailable mainly in clinical settings. In addition, there is a lack of agreement between available versions ofDXAhardware andsoftware for determining %BF (5,9,10).

Air-displacement plethysmography is a newer method available for testing body composition using the commercially available product, the BOD POD® (Life Measurement, Inc., Concord, CA). This method is a two-compartment model like UWW, but it removes many of the limitations associated with UWW testing such as submersing the head under water, time to complete the procedure (e.g., 5 min vs. 30-60 min), and it requires less training time for the technician. One of the major disadvantages to using this method is the expensive cost of the equipment (~ $30-40K), but overall it is deemed advantageous for use when compared to UWW and DXA (10).

UWW, DXA and air-displacement plethysmography are accurate methods of testing body composition but are not routinely available for testing in a field setting due to the large size of the equipment and lack of portability. Measuring skinfolds is a practical and accessible method for testing body composition in a field setting. There are also disadvantages to using this method that include different caliper types and intra- and inter-tester reliability (11). Skinfold measurements are a good estimate of body composition when used in an equation appropriate for the population being assessed (12,13). Skinfold models have previously been validated using UWW and DXA as the criterion measure (1), but to date no research is available using the BOD POD® as the criterion measure.

Valid skinfold equations for female collegiate athletes allow individuals such as exercise physiologists,coaches, and athletic trainersto accurately and easily predict players’ %BF in field settings. Accurate determination of body fat for female athletes will enable pre- and post-season assessments to monitor increases or decreases in body fat. To date the BOD POD® has not been used as a criterion measure to validate skinfold regression equations. Therefore, the primary purpose of this investigation was to compare %BF as estimated from three previously developed skinfold regression equations for female athletes to %BF measured using the BOD POD®.

METHODS

Subjects

The subjects were 75 female collegiate athletes recruited from a National Collegiate Athletic Association (NCAA) DivisionIIIUniversity. Subjectinclusioncriteria included participation in a female collegiate sport, and age ranging between 18 and 24 years. These athletes were recruited through contact with various intercollegiate coaches. The participants (n = 75) were athletes involved in basketball (n = 17), soccer (n = 18), softball (n = 15), swimming (n = 17) and volleyball (n = 8). Subject characteristics are shown in Table 1; all subjects were Caucasian. Each subject was provided with a list of pretest procedures(i.e., no alcohol, caffeine, food or exercise 2 – 3 hours prior to testing) before the testing date. The procedures as well as the potential benefits and risks were explained before obtaining written informed consent from each participant. Institutional Review Board approval was obtained for the use of human subjects.

PROCEDURES

BOD POD®

A two-point volume calibration was performed on the BOD POD® at baseline using a 50-liter calibrated cylinder in an empty chamber for 20 seconds. In addition, the electronic scale was calibrated by the BOD POD® computer system. As described by Dempster and Aitkens (14), each subject was instructed to wear a swimsuit and a swim cap to compress air in the hair. Height (cm), measured on a physician’s scale, and weight (kg), measured on the calibrated scale, were taken for each subject after voiding. The subject then entered the chamber and sat down breathing normally for 20 seconds. This was the initial test for body volume. At the end of the first test a second test was performed to verify consistency. The door was opened and shut between trials. The data were accepted if the two measurements were within 150 ml. If the two measurements were different by more than 150 ml, a third test was performed. Db was converted to %BF utilizing the Siri (15) equation.

Included in the BOD POD® system is the predicted measurement of average thoracic gas volume. McCrory,Gomez, Bernauer, & Mole (16) found that, on average, predicted thoracic gas volume did not differ significantly from the measured volume in men and women18-56 years of age. Because the average predicted thoracic gas volume is not different from measured thoracic gas volume, the predicted value was used in this investigation. However, it is important to note that some have cautioned that the failure to account for thoracic gas volume when using BOD POD® software will result in the overestimation of body density (Db) (16). All %BF procedures were performed as recommended by Life Measurement, Inc., Concord, CA, manufacturer of the BOD POD®.

Skinfold Measurements

The four skinfold sites (triceps, abdomen, suprailiac, thigh)were measured in rotating order three times each on the right side of the body to the nearest 0.5 mm; the median value was used for analysis (17). Each skinfold was grasped firmly with the thumb and index finger holding the caliper perpendicular to the fold approximately one centimeter away from the thumb and finger (18). The identificationof measurement sites were as explained by Jackson and Pollock (18). All skinfold measures were taken by one tester.

Data from the skinfold measurements were utilized in three different skinfold equations. The first and third skinfold equations were converted from body density (Db) to %BF via the equation of Siri (15). The three different skinfold equations used for this study are given below:

1) The first equation utilized (SF-UWW) was developed using UWW as the criterion method and has been recommended for female athletes (19, 20):

Db = 1.0960950 – 0.0006952 (X) + 0.0000011 (X)2 – 0.0000714 (age)

where X is the sum of triceps, abdomen, suprailiac and thigh skinfolds in mm.

2) The second equation utilized (SF-DXA)was developed from female athletes using DXA as the criterion measure (1):

FFM = 8.51 + (0.809 * wt) – (0.178 * abdominal skinfold) – (0.225 * thigh skinfold)

3) The third equation utilized (SF-Gen) was developed using UWW and is recommended for women between the ages of 18 and 55 years (19):

Db = 1.0994921 -0.0009929 (X) + 0.0000023 (X) 2 – 0.0001392 (age)

where X is the sum of the triceps, thigh and suprailiac skinfolds in mm.

To verify accuracy in SF measurements, criterion-related concurrent validity and intraclass reliability were calculated. Criterion-related concurrent validity was measured by comparing skinfold measurements taken by the tester with those taken by an individual with more than 15 years of experience with skinfold measurements. Fifteen women were tested at the same time of day by both measurers, with the newer tester always measuring prior to the experienced tester.The validity coefficients between testers for each skinfold measurement were above r = 0.80, which is recommended in order to substitute the tester for the experienced technician (21): triceps, r = 0.96; suprailiac, r = 0.94; abdomen, r = 0.95; thigh, r = 0.96.

Intraclass reliability was calculated for the three trials on all four skinfold sites by calculating a mean square representing the sum of changes in the mean and error (22). This was done to measure the repeatability or consistency of the tester’s ability to take skinfold measurements. All of the trials were highly correlated: triceps, R = 0.99; suprailiac, R = 0.99; abdomen, R = 0.99; thigh, R = 0.99. The coefficient of variation was computed for each subject’s skinfold results by dividing the standard deviation by the mean and multiplying by 100 (23). The mean coefficient of variation for each site reflected a small variability relative to the mean for each subject: triceps = 4.33%; suprailiac = 5.74%; abdomen = 3.98%; thigh = 3.41%.

STATISTICAL ANALYSES

Criterion-related concurrent validity and intraclass reliability were assessed usingStatView for Macintosh (Abacus Concepts, Inc., Berkeley, CA) before data collection began to ensure that the investigator could obtain valid and reliable data when taking skinfold measurements.

All other data were analyzed using SPSS 14.0 for Windows (SPSS Inc., Chicago, IL). One –way analysis of variance (ANOVA) was performed with repeated measures on the four body composition techniques. The dependent variable was estimated %BF. If the F-test was significant, Tukey’s HSD post hoc tests were used for pairwise comparisons of the means for each technique. Cohen’s d was used to calculate effect size for mean differences (24). Multiple linear regression (MLR) was also performed to predict the BOD POD® estimate of %BF, and the predictor variables included the three skinfold estimates of %BF. All statistical analyses were evaluated at an alpha of p < 0.05.

G-Power was used to determine a target sample size. Using the following parameters of a large effect size (Cohen’s d) of 0.80, an alpha level of 0.05 and four groups, an estimated sample size of 76 was computed for a power of 0.80 in this study (25).

RESULTS

The one-way ANOVA with repeated measures revealed that the assumption of sphericity was violated based on Mauchly’s Test of Sphericity. The Greenhouse-Geisser method was used to correct the degrees of freedom in order to determine statistical significance of the within subjects effect. Based on the statistical analysis, a significant effect was observed for the body composition techniques (F(3,222) =93.71, p < 0.001, η2 = 0.56). Body composition results (Mean ± SD) are presented in Table 2. Eta2 (η2) is the ratio of the variance due to the independent variable and the total variance (22). This means that 56% of the total variance in %BF can be attributed to the body composition techniques, and 44% of the total variance in %BF is attributed to other factors (e.g., genetics, diet, physical activity).

Table 1. Subject Characteristics (Mean ± SD) Table 2. Percentage of Body Fat (%BF) Results (Mean ± SD)

Total (n = 75)
Age (yrs) / 19.3 ± 1.2
Height (cm) / 166.7 ± 6.8
Body Mass (kg) / 68.5 ± 9.4
BMI (kg/m2) / 24.7 ± 3.5
Method * / Total (n=75)
SF-UWW / 19.33 ± 4.0
SF-DXA / 23.03 ± 3.50
SF-Gen / 25.09 ± 5.03
BOD POD® / 24.39 ± 6.42

BMI = Body Mass Index; SD = StandardDeviation All pairwise differences are significant different

(p < 0.05 [F(3, 222) =93.71, η2 = 0.56])except SF-Gen vs. BOD POD (N.S.) SF = Skinfold; UWW = Underwater weighing;

DXA = Dual-energy X-Ray Absorptiometry;

Gen = Generalized

A Tukey HSD post hoc analysis showed significant differences among all pairwise comparisons except between the BOD POD® and SF-Gen. The magnitude of these differences (Cohen’s d) for BOD POD® to skinfold techniques were small to large (see Table 3). Cohen (24) defines a small effect as 0.2, a moderate effect as 0.5, and a large effect as 0.8. This post hoc analysis also showed that significant differences were found between all three skinfold equation estimates of %BF. The magnitude of these differences were moderate to large and are also shown in Table 3.

Since the equation coefficients were the only difference between the SF-UWW and SF-Gen equations for estimating %BF, the two estimates would yield different but highly correlated values (see Table 4). To address the problem of multicollinearity, two multiple linear regressions (MLR) were conducted to determine which of the three skinfold estimates of %BF were predictors of the BOD POD® estimate of %BF. The first MLR included SF-UWW and SF-DXA as the predictor variables, and the results indicated a model of one significant predictor (SF-UWW), R = .819, R2 = .671, R2adj = .662, F(2,72) = 73.40, p < 0.001.

Table 3. Effect Sizes for Skinfold Techniques [Cohen’s d (23)]

SF-DXA / SF-Gen / BOD POD®
SF-UWW / 0.98* / 1.27* / 0.95*
SF-DXA / ------/ 0.48* / 0.26*
SF-Gen / ------/ ------/ 0.12

*p < 0.05; SF = Skinfold; UWW= Underwater weighing (19);

DXA = Dual-Energy X-ray Absorptiometry; Gen = Generalized

This model accounted for 67.1% of the variance in the BOD POD® estimate of %BF. The second MLR included SF-Gen and SF-DXA as the predictor variables, and the results indicate a model of one significant predictor (SF-Gen), R = .814, R2 = .662, R2adj = .653, F(2,72) = 70.63, p < 0.001. This model accounted for 66.2% of the variance in the BOD POD® estimate of %BF.

Table 4. Pearson Correlation Coefficients for Skinfold Techniques

SF-DXA / SF-Gen / BOD POD®
SF-UWW / 0.849* / 0.999* / 0.817*
SF-DXA / ------/ 0.855* / 0.727*
SF-Gen / ------/ ------/ 0.811*

*p < 0.05; SF = Skinfold; UWW= Underwater weighing (19); DXA = Dual-Energy X-ray Absorptiometry

(30); Gen = Generalized (19)

DISCUSSION

% Body Fat Determined from the BOD POD® vs. Skinfold Equations

The key finding of this investigation is that there were significant differences between the BOD POD® and other estimates of %BF, specifically regression equations validated for female athletes utilizing both DXA and UWW as the criterion. Although the differences between the BOD POD® and the SF-DXA equation were statistically significant, it is unlikely that these differences were large enough to suggest a meaningful difference, as indicated by the small effect size (see Tables 2 and 3). The difference between the SF-UWW equation and all other measures is important, however, as the results indicate that this equation underestimates %BF in female athletes.

There are several reasons that can be suggested as potential explanations for the difference between SF-UWW and the other measures of %BF. First, the sample in the current investigation was composed entirely of Division III athletes, while the SF-UWW equation and the SF-DXA equation were validated on Division I athletes. Due to the level of competition, training practices, and differences in scholarship opportunities, it is suggested that Division I athletes would be more muscular, with less body fat than their Division III counterparts, potentially explaining this difference. Secondly, the SF-UWW equation was cross validated in 1984, while the SF-DXA equation was developed in 2004. One could argue that there has been a substantial transformation in the body composition of the female athlete during this 20 year time period, perhaps rendering the SF-UWW equation less accurate for today’s female athletes. It is also possible that the SF-UWW equation is the accurate measure, and our other measures overestimate %BF. However, since previous research has shown that the BOD POD® is comparable to both DXA (5, 9) and UWW (9, 26, 27), this is unlikely.

It is also important to note that there were no differences between the %BF measures between the BOD POD and the SF-Gen equation, indicating that both are good measures of %BF in Caucasian female athletes. Further, as stated above, while a significant difference was calculated between the BOD POD and SF-DXA measures, the effect size indicates that this is not a meaningful difference, and all SF measures were strongly correlated with the BOD POD®. Therefore, it is suggested that the BOD POD, SF-DXA, and SF-Gen are all appropriate for determining %BF in Caucasian female athletes. As all of these are appropriate, the selection of method will depend upon factors such as time for test administration, cost, equipment maintenance, ability to accommodate people with limitations, ability to monitor changes over time, and ease of use (9).