The emphasis on thinness and appearance has been reported to cause female gymnasts to engage in pathogenic weight-control behaviors such as self-induced vomiting and the use of laxatives, diuretics, and/or diet pills to lose weight and reduce their percent body fat (%fat)(27,28,33). The use of these techniques can result in excessive weight loss and have serious implications for health, as well as performance. A pre-season assessment of body composition and the assignment of a minimal weight based on the gymnast's%fat may reduce excessive weight loss and decrease the use of potentially dangerous weight-control behaviors. These procedures have been successfully implemented for high school wrestlers who also reduce weight for competition(5,26). The success of assigning a minimal weight, however, is, in part, dependent on an accurate and practical assessment of body composition.
Underwater weighing (UWW) is a valid method for estimating body composition in young subjects (18,25). Although a laboratory technique such as UWW is preferred for determining body composition, it is often necessary to use a more practical field-based method such as skinfolds. Because there are many skinfold equations available, the decision regarding which one to use should be based on its accuracy for the population being examined (14). Lohman (19) recommends that, because of the confounding influences of such factors as age, gender, and ethnicity on the relationship between skinfolds and body fatness, skinfold equations should be cross-validated on various populations and in different laboratories from the ones in which they were developed. Therefore, the purpose of the present investigation was to cross-validate existing skinfold equations on a sample of high school female gymnasts.
The subjects for this study were 73 Caucasian high school female gymnasts([horizontal bar over]X age ± SD = 15.7 ± 1.2 yr; range, 13-17.8 yr). The subjects included varsity and junior varsity competitors in the 9th through 12th grades from three Class A high schools (largest schools in the state). The laboratory testing was performed during pre-season training and 1-2 wk prior to the competitive season. The study was approved by the Institutional Review Board for Human Subjects, and written informed consent was obtained from the subjects and their parents prior to testing.
Body density (BD) was assessed from UWW with correction for residual lung volume (RV) using the oxygen dilution method of Wilmore(39). Residual volume was determined on land with the subject seated in a position similar to that assumed during UWW. The average of similar scores (within 100 ml) from two to three trials was used as the representative RV. Underwater weight was measured in a hydrostatic weighing tank in which a metal swing seat was suspended from a Chatillon 9-kg scale. The average of the three highest values from six to ten trials was used as the representative underwater weight. Percent body fat was calculated from BD using the conversion constants of Brozek et al. (3). Previous test-retest reliability data for UWW from the authors' laboratory indicated that, for young adult male subjects (N = 16) measured 24-72 h apart, the intraclass correlation (R) was 0.98, with a standard error of measurement of 0.9%fat. These values are comparable with those reported by Jackson et al. (10) and Thomas and Cook(34). Furthermore, the UWW procedures used in the present investigation were standardized as a part of an interuniversity study(38) and found to be highly consistent with those from three other laboratories.
Predicted%fat was calculated from the skinfold equations (EQ) listed inTable 1. EQ 1 and 2 are generalized equations developed on adult samples and were selected for the present study because of their widespread popularity. EQ 3, 8, and 11 were selected because of their accuracy with young adult females. EQ 4 was derived for use with college-age female athletes, and EQ 9 and 10 were selected because of their accuracy with adolescent female athletes. EQ 5, 6, and 7 have been recommended for use with nonathletic children or adolescents. Skinfold measurements were taken on the right side of the body using Lange calipers at the triceps, subscapular, axilla, anterior suprailium, abdominal, and thigh sites and using the landmarks described by Jackson and Pollock (8) by an investigator who had previously showed test-retest reliability of r > 0.90. The medial calf and suprailiac skinfolds were measured in accordance with the procedures of Slaughter et al. (31).
All body composition determinations (UWW and skinfolds) were performed on the same day following a 12-h fast (ad libitum water intake was allowed). The subjects were also instructed to avoid exercise for a minimum of 12 h prior to testing.
The cross-validation analyses of the 11 equations in this study(Table 1) were based on an evaluation of the predicted%fat vs the criterion%fat from UWW via calculation of the constant error (CE = mean difference for criterion%fat - predicted%fat), r, standard error of estimate (SEE = SD √1 - r2), total error (TE =√Σ(criterion%fat -%fat predicted)2/n), and the similarity between the standard deviations of the predicted%fat and criterion%fat. The CE for each equation was analyzed using a dependent t-test with the Bonferroni correction (16) to adjust the per comparison alpha (0.05/11 = P < 0.0045).
Descriptive characteristics of the subjects are presented inTable 2. The mean%fat determined by UWW was 18.6 ± 4.5%fat. Table 3 includes the skinfolds used in the prediction equations.
Table 4 presents the results of the cross-validation analyses. The CE values ranged from -4.3 (EQ 11) to 1.7 (EQ 2)%fat. Equations 1, 2, 5-7, and 11 resulted in significant (P < 0.0045) CE values. The validity coefficients ranged from 0.58 (EQ 4) to 0.77 (EQ 9). The SEE values ranged from 2.9 (EQ 9) to 3.7 (EQ 4 and 6)%fat. However, TE, which combines the errors associated with both the CE and SEE(17), ranged from 3.3 (EQ 3 and 8) to 5.5 (EQ 11)%fat. Seven equations (EQ 1-3 and 7-10) resulted in TE values of ≤3.9%fat. In addition, the SD values for predicted%fat for five of the equations (EQ 1-3, 8, and 11) were condensed by 0.6 (EQ 2) to 2.5 (EQ 11)%fat compared with the distribution from UWW (SD = 4.5%fat; Table 2).
The gymnasts in the present study were approximately the same height and body weight (BW) but had less%fat than reference adolescent females of the same age (7). Compared with gymnasts of approximately the same age (4,23,35), the present subjects were taller, heavier, and had greater%fat. These differences in BW and%fat between samples of gymnasts may have been attributable to differences in training status. The subjects of Claessens et al. (4), Moffatt et al. (23), and Thorland et al.(35) were highly trained, while the present sample was measured during the preseason and prior to competition.
The following criteria were used in the present study to evaluate the results of the cross-validation analyses(6,9,14,17,29): (a) there should be a nonsignificant mean difference (CE) between the criterion and predicted%fat values; (b) the criterion and predicted%fat values should be significantly correlated; (c) there should be close agreement between the SD values of criterion and predicted%fat; (d) the SEE and TE values should be≤4%fat, and there should be a close similarity between the SEE and TE because it reflects the relationship between the regression line of criterion vs predicted%fat and the line of identity; and (e) if all other criteria are equal, equations that use the sum of skinfolds are preferred over those that use individual skinfold measures due to less intertester error. While all of these criteria should be considered, Sinning et al. (29) have stated that the TE is the best single criterion for determining the true difference between criterion and predicted%fat values.
There were significant CE values (Table 4; P< 0.0045) for EQ 1, 2, 5-7, and 11, which ranged from -4.3 to 1.7%fat. Equation 4 had the lowest CE value (0.0%fat), and nonsignificant CE values were also found for EQ 3 and 8-10 (-0.9 to 0.6%fat). The quadratic sum-of-three skinfold equations (EQ 1 and 2) of Jackson et al.(11), which were derived from adult females 18-55 yr of age, underestimated%fat for the high school gymnasts by ≥1.6%fat, while the linear equation (EQ 11) of Wilmore and Behnke (40), which was also derived on adult females 17.8-47.8 yr of age, overestimated%fat by 4.3%fat. Thus, for these adult equations (EQ 1,2,and 11), there was no consistent pattern of over- or underestimation of%fat for the present sample of high school gymnasts. Equations 5-7, however, which were developed for use with children and nonathletic adolescents(24,30), overestimated%fat by ≥2.2%fat. These findings likely reflected maturational changes in the BD vs skinfold relationship as described by Slaughter et al. (31). In general, as children and adolescents mature, a given skinfold value corresponds to higher BD values (31).
All of the validity coefficients in the present study(Table 4) were significant at P < 0.001 and ranged from r = 0.58 (EQ 4) to 0.77 (EQ 9). With the exception of EQ 11 (r = 0.70), the validity coefficients were less than the values from the original derivation analyses (Table 1). Furthermore, five equations (EQ 1-3, 8, and 11) resulted in SD values (Table 4) that were less than (by 0.6 to 2.5%fat) the value from UWW (SD = 4.5%fat). A large difference between the SD value for predicted and criterion%fat from UWW would likely result in errors at the extremes of the%fat distribution (17).
Lohman (17) estimates that because of biological and technical sources of error, the expected SEE associated with the prediction of BD from skinfolds within a specific population is approximately 0.0098 g·cm-3 (4%fat). In the present investigation, all of the equations (Table 4) had SEE values that were ≤3.7%fat, and seven equations (EQ 1-3 and 7-10) had TE values that ranged from 3.3 to 3.9%fat.
Previous studies have cross-validated EQ 1-4, 6, and 8-11 on adolescent athletes (36,37), college athletes(20), and nonathletic children and adolescents(12). Thorland et al.(36,37) report TE values for EQ 1, 2, and 8-11 that ranged from approximately 2.2 (EQ 9) to 7.3 (EQ 11)%fat in samples of highly trained adolescent athletes (age range, 11.17-19.58 yr). As in the present study, EQ 1, 2, 9, and 10 resulted in low TE values of approximately 2.5, 2.6, 2.2, and 2.4%fat, respectively, while the equation (EQ 8) of Sloan et al. (32) has a TE of 4%fat(36,37). Mayhew et al. (20) cross-validated EQ 1-3, 8, and 11 on a sample of 111 college athletes([horizontal bar over]X age ± SD = 19.66 ± 1.24 yr) and found TE values of approximately 4, 4.3, 4.2, 3.9, and 5.1%fat, respectively. Except for EQ 11, these values were higher than those from the present investigation(Table 4). Janz et al. (12) report TE values of approximately 3.8 and 3.5%fat for EQ 6 and 7, respectively, in a sample of 55 children and adolescents ([horizontal bar over]X age ± SD= 11.5 ± 1.8 yr). These values were 0.4 to 1.4%fat less than those from the present investigation (Table 4).
In the present study, the quadratic sum-of-three skinfold equation (EQ 10) of Thorland et al. (37) satisfied the most cross-validation criteria and, therefore, is recommended for estimating body composition in high school female gymnasts. The other six equations with TE values ≤3.9%fat (EQ 1-3 and 7-9) do not meet as many of the criteria but should be considered acceptable alternatives to EQ 10. EQ 10 resulted in a nonsignificant CE (0.6%fat; P > 0.0045), as well as low SEE(3.1%fat) and TE values (3.4%fat). Furthermore, the SD of predicted%fat(4.8%fat) was similar to the SD of%fat from UWW (4.5%fat), and EQ 10 uses the sum of skinfolds, which is preferred over individual skinfold measures such as those used in EQ 3 and 8 due to less intertester error(9). Compared with EQ 9, EQ 10 uses fewer skinfold measures (3 vs 7 skinfolds) and more closely approximated the%fat distribution from UWW (SD from UWW = 4.5%fat; EQ 9 = 5.4%fat; EQ 10 = 4.8%fat).
A preseason estimation of body composition using EQ 10 may be valuable to coaches and high school female gymnasts for determining a safe minimal competitive body weight and reducing the use of the potentially pathogenic behaviors associated with excessive weight loss. It has been recommended that high school wrestlers compete at not less than 5-7%fat(1,38). Females, however, have essential fat (bone marrow, spinal cord, liver, heart, spleen, kidneys, etc.) plus sex-specific fat (breasts, genitals, lower body subcutaneous, intramuscular, etc.) levels that total approximately 9-14%fat (2,13). Furthermore, it has been suggested that 12-14%fat represents a realistic lower limit of fatness for maintaining good health in females(22). This range is consistent with the mean%fat(13.1-14.8%fat) of adolescent female gymnasts at their optimum training level(23,35). Therefore, it is recommended that: (a) EQ 10 be used for a preseason assessment of body composition for high school female gymnasts if a laboratory procedure such as UWW is not available, (b) a minimal body weight at not less than 12%fat be determined from the preseason body composition assessment, and (c) high school female gymnasts receive medical clearance before being allowed to compete at less than 12%fat. Using the mean data from the present study, the minimal body weight from UWW at the recommended 12%fat would be 50.5 kg (fat-free weight/0.88). According to EQ 10, the predicted minimal body weight would be 50.9 kg (±1 TE of 3.4%fat = 48.8-53 kg).
1. American College of Sports Medicine. Position statement on weight loss in wrestlers. Sports Med. Bull.
2. Behnke, A. R. New concepts of height-weight relationships. In: Obesity
, N. L. Wilson (Ed.). Philadelphia: F. A. Davis, 1969, pp. 25-53.
3. Brozek, J., F. Grande, J. T. Anderson, and A. Keys. Densitometric analysis of body composition: revision of some quantitative assumptions. Ann. N. Y. Acad. Sci.
4. Claessens, A. L., R. M. Malina, J. Lefevre, et al. Growth and menarcheal status of elite female gymnasts. Med. Sci. Sports Exerc.
5. Clark, R. R., J. M. Kuta, and R. A. Oppliger. The Wisconsin wrestling minimal weight project: cross-validation of prediction equations. Pediatr. Exerc. Sci.
6. Graves, J. E., M. L. Pollock, A. B. Calvin, M. Van Loan, and T. G. Lohman. Comparison of different bioelectrical impedance analyzers in the prediction of body composition. Am. J. Hum. Biol.
7. Haschke, F. Body composition during adolescence. In:Body Composition Measurements in Infants and Children
, W. J. Klish and N. Kretchmer (Eds.). Columbus, OH: Ross Laboratories, 1989, pp. 76-82.
8. Jackson, A. S. and M. L. Pollock. Practical assessment of body composition. Physician Sportsmed.
9. Jackson, A. S., M. L. Pollock, and L. R. Gettman. Intertester reliability of selected skinfold and circumference measurements and percent fat estimates. Res. Q
. 49:546-551, 1978.
10. Jackson, A. S., M. L. Pollock, J. E. Graves, and M. T. Mahar. Reliability and validity of bioelectrical impedance in determining body composition. J. Appl. Physiol.
11. Jackson, A. S., M. L. Pollock, and A. Ward. Generalized equations for predicting body density of women. Med. Sci. Sports Exerc.
12. Janz, K. F., D. H. Nielsen, S. L. Cassady, J. S. Cook, Y. T. Wu, and J. R. Hansen. Cross-validation of the Slaughter skinfold equations for children and adolescents. Med. Sci. Sports Exerc.
13. Katch, V. L., B. Campaigne, P. Freedson, S. Sady, F. I. Katch, and A. R. Behnke. Contribution of breast volume and weight to body fat distribution in females. Am. J. Phys. Anthropol.
14. Katch, F. I. and V. L. Katch. Measurement and prediction errors in body composition assessment and the search for the perfect prediction equation. Res. Q. Exerc. Sport
15. Katch, F. I. and W. D. McArdle. Prediction of body density from simple anthropometric measurements in college-age men and women.Hum. Biol.
16. Keppel, G. Design and Analysis: A Researcher's Handbook
, 3rd Ed. Englewood Cliffs, NJ: Prentice-Hall, 1991, p. 167.
17. Lohman, T. G. Skinfolds and body density and their relationship to body fatness: a review. Hum. Biol.
18. Lohman, T. G. Applicability of body composition techniques and constants for children and youths. In: Exercise and Sport Sciences Reviews
. New York: Macmillian, 1986, pp. 325-357.
19. Lohman, T. G. Advances in Body Composition Assessment
. Champaign, IL: Human Kinetics, 1992, pp. 43-45.
20. Mayhew, J. L., B. A. Clark, B. C. Mckeown, and D. H. Montaldi. Accuracy of anthropometric equations for estimating body composition in female athletes. J. Sports Med.
21. Mayhew, J. L., F. C. Piper, J. A. Koss, and D. H. Montaldi. Prediction of body composition in female athletes. J. Sports Med.
22. McArdle, W. D., F. I. Katch, and V. L. Katch.Exercise Physiology: Energy, Nutrition, and Human Performance
. Philadelphia: Lea and Febiger, 1991, p. 603.
23. Moffatt, R. J., B. Surina, B. Golden, and N. Ayres. Body composition and physiological characteristics of female high school gymnasts. Res. Q. Exerc. Sport
24. Mukherjee, D. and A. F. Roche. The estimation of percent body fat by maximum R2
regression equations. Hum. Biol.
25. Nielsen, D. H., S. L. Cassady, K. F. Janz, J. S. Cook, J. R. Hansen, and Y. T. Wu. Criterion methods of body composition analysis for children and adolescents. Am. J. Hum. Biol.
26. Oppliger, R. A., R. D. Harms, D. E. Herrmann, C. M. Streich, and R. R. Clark. Grappling with weight cutting: the Wisconsin wrestling minimum weight project. Physician Sportsmed.
27. Petrie, T. A. and S. Stoever. The incidence of bulimia nervosa and pathogenic weight control behaviors in female collegiate gymnasts.Res. Q. Exerc. Sport
28. Rosen, L. W. and D. O. Hough. Pathogenic weight-control behaviors of female college gymnasts. Physician Sportsmed.
29. Sinning, W. E., D. G. Dolny, K. D. Little, et al. Validity of “generalized” equations for body composition analysis in male athletes. Med. Sci. Sports Exerc.
30. Slaughter, M. H., T. G. Lohman, R. A. Boileau, et al. Skinfold equations for estimation of body fatness in children and youth.Hum. Biol.
31. Slaughter, M. H., T. G. Lohman, R. A. Boileau, et al. Influence of maturation on relationship of skinfolds to body density: a cross-sectional study. Hum. Biol.
32. Sloan, A. W., J. J. Burt, and C. S. Blyth. Estimation of body fat in young women. J. Appl. Physiol.
33. Teitz, C. C. Sports medicine concerns in dance and gymnastics. Pediatr. Clin. North Am.
34. Thomas, T. R. and P. L. Cook. A simple inexpensive method for estimating underwater weight. Br. J. Sports Med.
35. Thorland, W. G., G. O. Johnson, T. G. Fagot, G. D. Tharp, and R. W. Hammer. Body composition and somatotype characteristics of Junior Olympic athletes. Med. Sci. Sports Exerc.
36. Thorland, W. G., G. O. Johnson, G. D. Tharp, T. G. Fagot, and R. W. Hammer. Validity of anthropometric equations for the estimation of body density in adolescent athletes. Med. Sci. Sports Exerc.
37. Thorland, W. G., G. O. Johnson, G. D. Tharp, T. J. Housh, and C. J. Cisar. Estimation of body density in adolescent athletes.Hum. Biol.
38. Thorland, W. G., C. M. Tipton, T. G. Lohman, et al. Midwest wrestling study: prediction of minimal weight for high school wrestlers. Med. Sci. Sports Exerc.
39. Wilmore, J. H. A simplified method for determination of residual lung volume. J. Appl. Physiol.
40. Wilmore, J. H. and A. R. Behnke. An anthropometric estimation of body density and lean body weight in young women. Am. J. Clin. Nutr.
Keywords:©1996The American College of Sports Medicine
BODY COMPOSITION; FEMALE ATHLETES; CROSS-VALIDATION; GYMNASTICS