Body mass index (BMI), calculated as body mass in kilograms divided by height in meters squared, is used widely in the classification of adult obesity. According to the Expert Panel on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults (7), a BMI of 25-29.9 kg·m−2 is considered overweight, and a BMI ≥ 30 kg·m−2 is considered obese. These well-established classifications use BMI as a surrogate measure for relative body fatness (% fat). Therefore, adults with a BMI greater than or equal to 25 kg·m−2 are considered to have excess % fat and are considered to be at risk for developing hypertension, high cholesterol, diabetes, and coronary heart disease (7). Large population studies have illustrated that BMI is related to morbidity and mortality. However, no subsequent large studies have confirmed these relationships with % fat. Although BMI is moderately correlated (r = 0.60-0.82) with % fat (26), there is a lack of research regarding the usefulness of BMI as a surrogate for % fat, and their exact biological relationship remains unclear. The classification of obesity on the basis of % fat has not been formally established. In an attempt to address this problem, the American College of Sports Medicine (ACSM) has reported predicted % fat values for BMI in males and females across different age groups (1). ACSM used data by Gallagher and colleagues (10), who developed multiple regression models for predicting % fat on the basis of BMI in 1626 subjects. In developing these equations, the authors used race, age, and sex as predictor variables to help explain the model. Their results state that 20% (in men) and 33% (in women) are acceptable cut points for overfatness corresponding to a BMI of 25 kg·m−2 in young African American and white adults (ages 20-39).
The BMI classification system is derived from cut points obtained from the general population and may not be specific to subgroups such as athletes and young adult nonathletes. Compared with the general adult population, the influence of large muscle mass on BMI in athletes and young adults may misclassify these individuals as overweight and obese. Therefore, the use of % fat may be more effective than BMI in assessing obesity in athletes and young adults. Despite the potential limitations of BMI, it is used to assess obesity in young adults (21) and athletes (14). Therefore, it is critical to understand the accuracy of BMI as a measure of % fat in these populations. However, to our knowledge, no study has assessed the ability of BMI to classify college athletes and nonathletes as overfat. Therefore, the purposes of this study were 1) to describe the relationship between the BMI and % fat of collegiate athletes and college-aged nonathletes, and 2) to determine the accuracy of the BMI category of overfat (≥ 25 kg·m−2) as a measure of excessive body fatness in collegiate athletes and college-aged nonathletes.
A total of 226 (149 male, 77 female) varsity athletes and 213 healthy college nonathletes (78 male, 135 female) from a large midwestern university participated in the study. BMI and % fat data were collected on male athletes participating in football, basketball, hockey, and wrestling and on female athletes participating in basketball, crew, and softball. BMI and % fat were also assessed in undergraduate students majoring in kinesiology who were enrolled in an exercise physiology laboratory class. The study was approved by the university biomedical institutional review board. The requirement for individual subject consent was waived because measurements were taken as part of regular medical screening (athletes) or classroom activities (nonathletes). All data were deidentified after the completion of testing, and as a result, exempt approval status was granted by the university biomedical institutional review board.
All measures were completed in the human energy research laboratory at Michigan State University. Subjects were required to wear spandex clothing and a Lycra cap and were requested not to eat or exercise for 3 h before testing. Standing height was measured to the nearest 0.1 cm using a wall-mounted, calibrated stadiometer. Body mass was measured to the nearest 0.01 kg using a calibrated electronic scale. BMI was calculated as body mass (kg) divided by height (m) squared. Body volume was measured via air displacement plethysmography using the BOD POD version 1.69 (Life Measurement Inc. Concord, CA). Using body mass and volume measures, each subject's body density was calculated and converted to % fat using the modified Siri equation (25). The BOD POD system was calibrated with a cylinder of known volume before testing, and thoracic gas volume was measured in each subject using standard procedures developed by the manufacturer (Life Measurement Inc. Concord, CA). Previous research has shown that the BOD POD provides a valid estimate of body fatness in the general population (9,17,20) and college athletes (2,27). The test-retest reliability of the BOD POD has been shown to be high (r = 0.96-0.99) in athletes and young adult nonathletes (2,4), which is consistent with the repeated measures performed in our lab.
BMI and body fat classifications
On the basis of standard guidelines (7), overweight was defined as a BMI greater than or equal to 25 kg·m−2 and less than 30 kg·m−2, and obesity was defined as a BMI ≥ 30 kg·m−2. Because of the small number of nonathletes and athletes (not including football linemen) with a BMI ≥ 30 kg·m−2, they were included in the overweight classification. We defined overfat as ≥ 20% and ≥ 33% for males and females, respectively. These cut points for overfatness are consistent with the overweight classification using BMI in the young adult population (10).
Because of similarities in the distribution of BMI and % fat, males participating in basketball, hockey, wresting, and football (nonlinemen) were combined into one male athlete group, and females participating in basketball, crew, and softball groups were combined into one female athlete group. As a result of the large discrepancy in size and body mass among football linemen compared with other athletes, the football linemen were analyzed separately. Therefore, for the purpose of analyses, there were three male groups: 1) athletes, 2) linemen, and 3) nonathletes; and two female groups: 1) athletes and 2) nonathletes. One-way ANOVA was used to compare differences in BMI and % fat among the three male groups and between the two female groups. Pearson correlation coefficients (r) were calculated for each group to assess the linear relationship between BMI and % fat.
Using the cut point of BMI ≥ 25 kg·m−2 to define overweight and ≥ 20% (males) or ≥ 33% (females) to define overfat, participants were classified into one of four categories: 1) overweight and overfat (true positive (TP)), 2) overweight and normal fat (false positive (FP)), 3) normal weight and overfat (false negative (FN)), and 4) normal weight and normal fat (true negative (TN)). To determine the accuracy of BMI as a measure of overfatness with % fat as the criterion measure, the sensitivity, specificity, and predictive values of BMI were calculated for each group. Sensitivity was calculated as the proportion of overfat individuals who were identified as overweight by BMI (i.e., TP/TP + FN). Specificity was calculated as the proportion of normal fat individuals who were identified as normal weight by BMI (i.e., TN/TN + FP). Positive predictive value (PPV) was calculated as the probability that a person identified as overweight by BMI was truly overfat (i.e., TP/TP + FP). Negative predictive value (NPV) was calculated as the probability that a person who was identified as normal weight by BMI was normal fat (TN/TN + FN) (18). Therefore, test accuracy increased as the total number of FP and FN classifications decreased. Because no linemen had a BMI less then 25 kg·m−2, and only one had a body fatness value of less than 20%, data were analyzed further using 30 kg·m−2 as the cut point for BMI and 25% fat as the cut point for overfatness. Gallagher and colleagues (10) suggest that 25% fat as an appropriate cut point for obesity (BMI ≥ 30 kg·m−2) in young adult males.
Receiver operator characteristic (ROC) curves, which are plots of the TP rate against the FP rate for the different cut points of a test, were developed and used to define optimal BMI cut points (13). For each group, the optimal BMI cut point represents the point in which the total number of FP and FN are minimized and the combined sensitivity and specificity are maximized (i.e., FP and FN).
Descriptive data (mean ± SD) of height, body mass, BMI, and % fat for all male and female study participants are shown in Tables 1 and 2. The male athlete group was significantly taller, heavier, and had a lower % fat than the nonathlete group. The linemen were significantly taller, heavier, and had greater BMI and % fat values than both the male athlete and nonathlete groups. Within female participants, athletes' height, body mass, and BMI were significantly higher than those of nonathletes. There were significant correlations between % fat and BMI in all male and female groups.
Table 3 shows the prevalence of overfatness, sensitivity, specificity, and predictive values, for the five major subgroups. Figures 1 and 2 are scatterplots of percent body fat versus BMI in each male and female group. In both figures, the quadrants were labeled as FN, TP, TN, or FP. Accuracy of BMI was determined by assessing the proportion of study participants who were FP and FN. In Figure 1, 67% of all male athletes fell within the FP quadrant. As a result, the specificity and PPV of BMI is poor. In addition, 25% of all male nonathletes fell within the FP quadrant (Fig. 1), resulting in a specificity of only 60%. For female athletes (Fig. 2), 31% of total participants were classified as FP, also resulting in poor specificity and PPV for BMI, but not to the extent of the male athletes. In contrast, only 7% of female nonathletes were FP (Fig. 2), resulting in a higher specificity than male athletes, male nonathletes, and female athletes. There were no FN in the male and female athlete groups, resulting in perfect sensitivity and NPV. In addition, only a small proportion of overfat male nonathletes were classified as a FN; hence, the sensitivity was high (83%). However, the percentage of FN results among all overfat female nonathletes was 44%, resulting in low sensitivity.
As shown in Table 3, specificity and NPV could not be calculated for the linemen group because no individual had a BMI less than 25 kg·m−2, and only one individual was less than 20% fat. Because of the large size of the linemen group, data were analyzed using a BMI of 30 kg·m−2 and 25% fat as the cut points for BMI and overfatness. Compared with the original values in Table 3, the prevalence of overfatness within the linemen group was 65%. The sensitivity and specificity (± 95% CI) were 0.94 (0.71, 0.99) and 0.1 (0.01, 0.46), respectively. The PPV (± 95% CI) was 0.65 (0.44, 0.82), and the NPV (± 95% CI) was 0.5 (0.03, 0.97).
The optimal BMI cut points for 20% fat in men and 33% fat in women were calculated via ROC curves, and the results are shown in Table 3. The optimal BMI cut points were 27.9 kg·m−2 for male athletes (sensitivity = 0.92, specificity = 0.77), 27.7 kg·m−2 for female athletes (sensitivity = 1.0, specificity = 0.89), 34.1 kg·m−2 for linemen (sensitivity = 0.78, specificity = 1.0), 26.5 kg·m−2 for male nonathletes (sensitivity = 0.83, specificity = 0.77), and 24.0 kg·m−2 for female nonathletes (sensitivity = 0.71, specificity = 0.79). When using a % fat ≥ 25 as the cut point for overfatness in the linemen group, an optimal BMI cut point of 35.2 kg·m−2 (sensitivity = 0.83, specificity = 0.60) was derived.
The results of our study illustrate that BMI is not an accurate measure of fatness in college athletes and nonathletes. The specificity was low in male college athletes, female college athletes, and male college nonathletes, indicating that BMI misclassifies normal fat individuals a large percentage of the time. In contrast, within female college nonathletes, the specificity was high, but sensitivity was low, suggesting that a large percentage of overfat individuals were classified as normal weight by BMI.
The ability of BMI to accurately reflect adiposity across athletic and nonathletic populations has been assessed previously (19,34). Nevill and colleagues (19) explored the relationship between skinfold measurements, stature, and mass to assess how accurately BMI reflected adiposity in 478 untrained and athletic adults. The authors report that after controlling for differences in body size (approximate BMI), both male and female athletes had lower skinfold measures than their untrained counterparts. These results further illustrate the inability of BMI to effectively represent adiposity in athletic populations. Furthermore, when Witt and Bush (34) examined the relationship between BMI and body fat in college athletes, the authors found that only 20% of women and 4% of men with BMI ≥ 25 kg·m−2 were above the 85th percentile for skinfold measurements. Similar to our results, the studies of both Nevill et al. (19) and Witt and Bush (34) illustrate that elevated BMI does not necessarily represent overfatness across athletic populations. However, we found no previous studies on athletes or young adults that reported sensitivity, specificity, and predictive values for BMI as a measure of body fatness.
Other investigators have examined the diagnostic ability of BMI in relation to % fat in adults (3,5,6,11,15,26,29,31). However, because of the lack of an established % fat criterion for health status and the differences in study design, it is difficult to compare the results of our study with previous research. Previous studies used different methods for measuring % fat, including dual-energy x-ray absorptiometry (DEXA) (3,5,11), skinfolds (6), and hydrodensitometry (15,26,31). In addition, a variety of BMI cut points were used to classify individuals as overweight, including 25 (6,11), 26 (31), 27 (15), and 27.3 (5,26) for women and 25 kg·m−2 (6), 27 kg·m−2 (11), 27.8 kg·m−2 (5,26), and 28 kg·m−2 (15,31) for men. The different % fat cut points used to identify over fatness included 25% (5,26), 30% (6,15), 33% (31), 35% (11), and 38% (3) for females, and either 20% (5,26) or 25% (6,11,31) for males. There was only one multiracial study that included whites, blacks, and Asians (26). The remaining studies examined either Asians (6,11,29) or whites (5,15,31) alone. With the exception of one study that assessed postmenopausal women (3), each study assessed both males and females. The majority of studies included young, middle-aged, and older adults (5,6,11,15,26), whereas an additional study focused primarily on young and middle-aged adults (31). Within the postmenopausal women, BMI seemed to be a good diagnostic test for overfatness (3). However, the remaining research consistently indicated BMI has low sensitivity (0.06-0.60) and high specificity (0.86-1.0) as a measure of % fat in both men and women (5,6,11,15,26,31).
In the present study, specificity and PPV were low for male athletes and moderate for male nonathletes. Therefore, among all normal fat male athletes and nonathletes, 73% and 40% were misclassified as overweight, respectively. Furthermore, BMI ≥ 25 kg·m−2 incorrectly classified male athletes and nonathletes as normal fat 87 and 44% of the time, respectively. Although specificity and PPV were higher in female athletes compared with male athletes, 34% of all normal fat female athletes were classified incorrectly. In addition, BMI ≥ 25 kg·m−2 incorrectly classified female athletes 77% of the time. In contrast to all athletes and male nonathletes, 44% of overfat female nonathletes were classified as normal weight, illustrating low sensitivity in this group.
As stated earlier, direct comparison of our results with those of previous studies is difficult because of the use of different cut points for overfatness. However, two studies that included adults with a wide body size range used cut points for overfatness that were identical to ours (5,31). Curtin and colleagues (5) assessed the validity of BMI as a measure of obesity measured by DEXA in 72 males ages 15-86. Using 20% fat and a BMI of 27.8 kg·m−2 as the cut point for obesity, sensitivity and specificity of BMI were 13% and 100%, respectively. These results are quite different from the high sensitivities found in our male athletes and nonathletes, suggesting that the accuracy of BMI to detect overfatness varied across male populations because of different degrees of body composition. Wellens and colleagues (31) also assessed the accuracy of BMI as a measure of body composition in 511 Caucasian women ages 20-44, using hydrodensitometry. The authors report a sensitivity of 52% and perfect specificity when using a % fat cut point of 33% and a BMI cut point of 26 kg·m−2. These results are nearly identical to those for our female nonathletes, where BMI ≥ 25 kg·m−2 had a low sensitivity and a high specificity. However, the results are in contrast with those of our female athletes. It is possible that the dissimilarities in the diagnostic accuracy of BMI between our college athletes and male nonathletes and in the previous literature on adults are partly explained by the different BMI cut points used in the two comparison studies. However, it is unlikely that these disparate findings are solely attributable to cut-point differences. Other explanations include the wide age range used in the previous studies and the increased number of FP in college athletes and male nonathletes attributable to greater muscle mass at a given body mass. It is possible that the low sensitivity among female nonathletes may be attributable to slightly greater body fat and/or less fat-free mass that results from minimal strength training within this population.
Because of the large size and elevated % fat found in the football linemen, nearly every individual was classified as overweight and overfat using a BMI ≥ 25 kg·m−2 and 20% fat as cut points. Therefore, analysis was extended to evaluate the BMI classification of obesity (BMI ≥ 30 kg·m−2) and a % fat of ≥ 25% as cut points. Gallagher et al. (10) reported that for young adult males, a BMI of 30 kg·m−2 is used to predict a % fat of 25. Results using these cut points showed that BMI correctly classified 94% of overfat linemen but only 10% of normal fat linemen. In a study of 504 nonathlete males between the ages of 20 and 45 yr, Wellens et al. (31) found that a BMI of 28 kg·m−2 classified only 43% of overfat males and nearly 100% of normal fat males correctly using 25% fat as the cut point for overfatness. Compared with the results from our study, the results of Wellens et al. (31) suggest that the accuracy of BMI is considerably different for linemen than for the general adult population. Despite this, the general BMI classification system has been used to assess obesity in professional football players (14). Our results suggest that a BMI ≥ 30 kg·m−2 is incorrect at predicting overfatness 35% of the time in linemen; consequently, a different BMI classification system should be developed to assess overfatness in football linemen.
In our data, ROC suggest the optimal BMI cut points for male athletes (27.9 kg·m−2) and female athletes (27.7 kg·m−2) are higher than the value of 25 kg·m−2 that is currently used. These optimal cut points are consistent with the BMI classifications for overweight established at the 1985 NIH Consensus Development Conference on the Health Implications of Obesity. The consensus panel used a BMI ≥ 27.8 kg·m−2 for men and a BMI ≥ 27.3 kg·m−2 for women to define overweight, which represented 85th percentile of BMI distribution in young adults ages 20-29. Therefore, when using BMI to assess the health status of athletes, the previous classification system seems to be more accurate. Partly because of the lower muscle mass in nonathletes, the optimal BMI cut points for male and female nonathletes were lower than their athletic counterparts. Yet, compared with the current BMI recommendation for overweight, the optimal BMI cut point also was higher for male nonathletes (26.5 kg·m−2). Interestingly, the optimal BMI cut point for female nonathletes (24 kg·m−2) is lower than the current recommendations. In addition, when a cut point of 25% fat was used in the linemen group, ROC analysis indicated that the optimal BMI cut point was 5 kg·m−2 higher than the current classification of 30 kg·m−2. Previous studies on general adult populations have derived optimal BMI cut points of 20.5 kg·m−2 for 20% fat in males between the ages of 15 and 86 (5), 23 kg·m−2 for 33% fat in females between the ages of 20 and 45 (31), and 24.5 kg·m−2 for 25% fat in males between the ages of 19 and 77 (15). Consistent with our results, these differences in optimal BMI cut points further illustrate the limitations in the BMI classification system for assessing overfatness in college athletes and nonathletes.
A limitation to our study is the use of a convenience sample and the resulting lack of generalizability of our college athletes and nonathletes. Previous studies in male athletes participating in basketball (12.4%) (23), hockey (13%) (22), wrestling (8.8%) (24), and football (nonlinemen: 7-16%; offensive linemen: 25%) (16) show % fat values similar to those of our male athletes. The % fat results for female basketball players in our study also are similar to those found in previous studies of female basketball players (30). However, among studies performed more than 20 yr ago, % fat values for crew (14%) (12), and softball (19.2%) (33) were lower than those found in our investigation.
It is possible that our nonathletes are more physically active than the general population and, thus, less likely to be overweight and overfat. A limitation to our study was our inability to compare physical activity levels, athletic history, and general lifestyle of our nonathletes with those of other students who were not kinesiology majors at our university. However, we analyzed self-reported height and body mass data on more than 4800 college students enrolled in the National College Health Risk Behavior Survey. Male and female students enrolled in the National College Health Risk Behavior Survey had average (± SD) BMI values of 25 ± 4.4 and 24.2 ± 5.7 kg·m−2, respectively, which were similar to those of our nonathletes (i.e., 26 and 23.4 kg·m−2). Furthermore, Wetter and Economos (32) reported % fat values of 16.2% in male college students and 25.9% in female college students; these values are similar to the % fat values of the male (17.7%) and female (28.5%) nonathletes in our study.
Another potential limitation of our study is the use of a two-compartment model for the estimation of body composition. Although a two-compartment model cannot distinguish between muscle density and bone density, we assumed that most differences in body density were attributable to differences in muscle mass. Despite the inherent limitation of a two-compartment model compared with a four-compartment model, previous research has illustrated that BOD POD is a valid measure of % fat when compared with DEXA (20). In fact, previous research in female athletes has illustrated a difference of only 0.5% (2). However, comparisons between BOD POD and DEXA have been inconsistent in adults (8) and have slightly underpredicted % fat in athletic populations (4,28). The somewhat inconsistent measures between BOD POD and DEXA may in part be affected by the clothing worn during the BOD POD test (8). To limit this measurement, the methodological procedures for BOD POD measurement were highly stringent in our study.
Results of the current study suggest that a BMI ≥ 25 kg·m−2 is not an accurate predictor of overfatness in college athletes and nonathletes. Because of a larger muscle mass among the male and female athletes, BMI incorrectly classifies normal fat athletes as overweight. According to ROC curves, the current BMI cut point for overweight (BMI = 25 kg·m−2) should be increased to limit these misclassifications within the athletic population. Current BMI classifications also are limited for assessing overfatness in the nonathlete population. In males, a BMI of 25 kg·m−2 may be too low for evaluating overfatness. However, in female nonathletes, a BMI cut point of 25 kg·m−2 may be too high. If future studies can confirm the current findings, then individuals responsible for the assessment of body composition in young adult athletes and nonathletes should consider using BMI cut points derived specifically from these populations.
1. ACSM's Guidelines for Exercise Testing and Prescription
. 7th ed. Baltimore, MD: Lippincott Williams & Williams, pp. 57-66, 2006.
2. Ballard, T. P., L. Fafara, and M. D. Vukovich. Comparison of BodPod and DXA in female collegiate athletes. Med. Sci. Sports Exerc.
3. Blew, R. M., L. B. Sardinha, L. A. Milliken, et al. Assessing the validity of body mass index standards in early postmenopausal women. Obe. Res.
4. Collins, M. A., M. L. Millard-Stafford, and P. B. Sparling. Evaluation of a new air displacement plethysmography for measureing human body composition
. Med. Sci. Sports Exerc.
5. Curtin, F., A. Morabia, C. Pichard, and D. O. Slosman. Body mass index compared to dual-energy X-ray absorptiometry: evidence for a spectrum bias. J. Clin. Epidemiol.
6. Dudeja, V., A. Misra, R. M. Pandey, G. Devina, G. Kumar, and N. K. Vikram. BMI does not accurately predict overweight in Asian Indians in northern India. Br. J. Nutr.
7. Executive summary of the clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults. Arch. Intern. Med.
8. Fields, D., M. Goran, and M. McCrory. Body-composition assessment via air-displacement pplethysmography in adults and children: a review. Am. J. Clin. Nutr.
9. Fields, D., G. Hunter, and M. Goran. Validation of the BOD POD with hydrostatic weighting: influence of body clothing. Int. J. Obes.
10. Gallagher, D., S. B. Heymsfield, M. Heo, S. A. Jebb, P. R. Murgatroyd, and Y. Sakamott. Healthy percentage body fat ranges: an approach for developing guidelines based on body mass index. Am. J. Clin. Nutr.
11. Goh, V. H. H., C. F. Tain, T. Y. Y. Tong, H. P. P. Mook, and M. T. Wong. Are BMI and other anthropometric measures appropriate as indices for obesity? A study in an Asion population. J. Lipid Res.
12. Hagerman, F. C., F. R. Hagerman, and T. C. Mickelson. Physiological profiles of elite rowers. Phys. Sportsmed.
13. Hanley, J., and B. McNeil. The meaning and use of the area under a receiver operating charateristic (ROC) curve. Radiology
14. Harp, J. B., and L. Hecht. Obesity in the National Football League. JAMA
15. Hortobagyi, T., R. G. Israel, and K. F. O'Brien. Sensitivity
of the Quetelet index to assess obesity in men and women. Eur. J. Clin. Nutr.
16. Kraemer, W. J., J. C. Torine, R. Silvestre, et al. Body size and composition of National Football League Players. J. Strength Cond. Res.
17. McCrory, M. A., T. D. Gomez, and E. M. Bernauer. Evaluation of a new air displacement plethysmography for measuring human body composition
. Med. Sci. Sports Exerc.
18. McNeil, B., E. Keeler, and J. Adelstein. Primer on certain elements of clinical decision making. N. Engl. J. Med.
19. Nevill, A., A. Stewart, T. Olds, and R. Holder. Relationship between adoposity and body size reveals limitations of BMI. Am. J. Phys. Anthropol.
20. Nunez, C., A. Kovera, A. Pietrobelli, et al. Body composition
in children and adults by air displacement plethysmography. Eur. J. Clin. Nutr.
21. Ogden, C., M. Carroll, L. Curtin, M. McDowell, C. Tabak, and K. Flegal. Prevalence of overweight and obesity in the United States, 1999-2004. JAMA
22. Rusko, H., M. Havu, and E. Karivien. Aerobic performance capacity in athletes. Eur. J. Appl. Physiol.
23. Siders, W. A., W. W. Bolonchuk, and H. C. Lukaski. Effects of participation in a collegiate sport season on body composition
. J. Sports Med. Phys. Fitness
24. Sinning, W. E. Body composition
assessmetn of college wrestlers. Med. Sci. Sports Exerc.
25. Siri, W. E. The gross composition of the body. Adv. Biol. Med. Phys.
26. Smalley, K. J., A. N. Knerr, Z. V. Kendrick, J. A. Colliver, and O. E. Owen. Reassessment of body mass indices. Am. J. Clin. Nutr.
27. Utter, A. C., F. L. Goss, P. D. Swan, G. S. Harris, R. J. Robertson, and G. A. Trone. Evaluation of air displacement for assessing body composition
of collegiate wrestlers. Med. Sci. Sports Exerc.
28. Vescovi, J., L. Hildebrandt, W. Miller, R. Hammer, and A. Spiller. Evaluation of the BOD POD for estimating percent fat in female college athletes. J. Strength. Cond. Res.
29. Vikram, N. K., A. Misra, R. M. Pandey, et al. Anthropometry and body composition
in northern Asian Indian patients with type 2 diabetes: receiver operator characteristics (ROC) curve analysis of body mass index with percentage body fat as standard. Diabetes Nutr. Metab.
30. Walsh, F. K., V. H. Heyward, and C. G. Schau. Estimation of body composition
of female intercollegiate basketball players. Phys. Sportsmed.
31. Wellens, R. I., A. F. Roche, H. J. Khamis, A. S. Jackson, M. L. Pollock, and R. M. Sievogel. Relationships between the body mass index and body composition
. Obes. Res.
32. Wetter, A. C., and C. D. Economos. Relationship between quantitative ultrasound, anthropometry and sports participation in college aged adults. Osteoporos. Int.
33. Withers, R. T., N. O. Wittingham, K. I. Norton, J. La Forgia, M. W. Ellis, and A. Crockett. Relative body fat and anthropometric prediction of body density of female athletes. European Journal of Applied Physiology
34. Witt, K. A., and E. A. Bush. College athletes with an elevated body mass index often have a high upper arm muscle area, but not elevated triceps and subscapular skinfolds. J. Am. Diet. Assoc.