Body composition of anorexia nervosa patients assessed by underwater weighing and skinfold-thickness measurements before and after weight gain

¹ From the University Center St-Jozef, Kortenberg, Belgium, and the Departments of Rehabilitation Sciences and Kinesiology, Faculty of Physical Education and Physiotherapy, Catholic University of Leuven, Belgium.

² Address reprint requests to M Probst, University Center St-Jozef, B-3070 Kortenberg, Belgium. E-mail: michel.probst{at}flok.kuleuven.ac.be.

ABSTRACT  
Background: Weight restoration is a crucial element in the treatment of patients with anorexia nervosa. Therefore, the validity of different methods for measuring body composition is important.

Objective: We tested the concurrent validity of hydrodensitometry (underwater weighing) and anthropometry (12 skinfold thicknesses) and assessed body composition and subcutaneous fat before and after a refeeding program and a multifaceted program of therapy in a specialized inpatient unit for eating disorders.

Design: The body composition of a large sample of anorexia nervosa patients (97 restricting type, 33 binging-purging type) was studied by using 2 methods both before and after weight gain. We applied a behavioral contract for weight restoration with a minimum weekly gain of 700 g and a maximum of 3 kg. Bland-Altman analysis of agreement, Pearson correlation analysis, t tests, and analysis of covariance were used.

Results: There was good agreement between the results obtained by underwater weighing and by skinfold-thickness measurement (r = 0.76, P < 0.001); the results produced by the 2 methods did not differ significantly. On average, a significant weight gain (11.9 kg) was observed, composed of 6.6 kg fat and 5.3 kg fat-free mass.

Conclusions: Body fat estimation by skinfold-thickness equation appeared to be as accurate as underwater weighing. The refeeding program led to a significant increase in body weight, of which 55.5% was body fat. The mean ratio of fat-free mass to fat mass at the end of the treatment was 3.4:1.

Key Words: Anorexia nervosa • body composition • body fat measurement • hydrodensitometry • skinfold-thickness measurement • underwater weighing • weight gain

INTRODUCTION  
Weight restoration is one of the most important goals in the treatment of patients with anorexia nervosa (AN). Through food restriction, excessive physical activity, or purging (self-induced vomiting, laxative abuse, or both), AN patients reach their desired goal of extreme slimness, resulting in a body weight 15% below the expected weight-for-height (1, 2). The patients' perception of body weight and shape is distorted (3). From both a physiologic and psychological viewpoint, weight and body composition are crucial elements in treatment because the patients are dissatisfied with their body appearance and have unrealistic ideas about the influence of food intake and dieting on body weight and shape. The study of body fat distribution can have important implications for the treatment of AN because of the patients' intense fear of weight gain and becoming fat, which is enhanced by the belief that this would be accompanied by fat distribution in the abdominal region. Hence, the amount of body fat at discharge may influence the further evolution of the disease (4). An earlier study showed that age and duration of illness had no influence on body composition in AN patients before weight restoration (5). However, studies of body composition in AN patients are sparse and mostly based on small samples of 6–32 patients (6–12). Moreover, the effect of weight restoration (pretreatment compared with posttreatment) has been neglected. Orphanidou et al (13) studied short-term weight gain (n = 26) and Zamboni et al (14) assessed the effects of weight gain on abdominal fat distribution in AN patients (n = 14).

A wide variety of methods are available to measure body composition (for a review see reference 15). In this study, the more traditional methods of body-composition measurement, such as the densitometric technique of underwater weighing (UWW) and the anthropometric method of measuring skinfold thicknesses (SFTs), were applied. The aims of this study were 1) to compare these 2 methods (and test their concurrent validity) for evaluating body composition before and after weight gain, and 2) to examine the effects of weight changes on body composition and subcutaneous fat after an inpatient treatment that included a refeeding protocol. We hypothesized that the SFT equations would accurately estimate body fat in AN patients and that refeeding would lead to a significant increase in body weight and subcutaneous fat and that body fat would increase most significantly.

SUBJECTS AND METHODS  
Subjects
One hundred thirty female patients consecutively admitted to the inpatient Eating Disorder Unit of the University Center Kortenberg, Belgium, and who finished the whole treatment program were included in the study. Data were collected between 1994 and 1999 by following a standard procedure. All patients met the Diagnostic and Statistical Manual of Mental Disorders, 4th edition, diagnostic criteria for AN (1). Patients were classified into 2 subtypes: the "restricting" type, who lost weight only through fasting and eventually rigorous physical exercise (n = 97), and the "mixed" or "binging-purging" type, who intermittently showed episodes of binge eating, self-induced vomiting, or laxative use (n = 33). Patients with concurrent organic illnesses were excluded. At the time of admission, all patients were amenorrheic. Study procedures were integrated into the treatment program and were approved by the Ethical Committee of the University Center Kortenberg.

Procedures
UWW is considered to be the method of reference for body-composition assessment and is considered to have excellent reliability and validity. The measurement of SFTs is a classic technique that measures only subcutaneous adipose tissue, with generally weaker reliability (but usually r > 0.90) and lower validity than UWW. The correlation between UWW and SFT measures ranges from 0.65 to 0.93 (15).

A standard protocol for all anthropometric and densitometric assessments was executed within 3 d of admission and once again in the last week before discharge from the hospital by the same highly experienced female investigator (MG). Height was measured to the nearest 0.5 cm by using a stadiometer. Body weight was measured to the nearest 100 g, with the subject wearing only a swimsuit, on a beam balance. The body mass index (BMI; kg/m²) was calculated.

Twelve SFTs were measured (5, 15–17): biceps, triceps, subscapular, suprailiac, chin, side, waist, abdomen, thigh anterior and posterior, and calf lateral and medial on the left side of the body by using a Harpenden skinfold caliper with electronic readout (18; HERO; Human Biology Laboratory of the Department of Human Sciences, University Loughborough, United Kingdom, and British Indicators, Ltd, St Albans, United Kingdom). The test-retest precision of our SFT measurements fluctuated from r = 0.93 (for thigh posterior) to r = 0.97 (for suprailiac). The percentage body fat was estimated from the logarithm of the sum of the SFT measurements taken at 4 sites (biceps, triceps, subscapular, and suprailiac) and by using the age-specific prediction equations (density = 1.1599 - 0.0717 x log 4 SFTs) of Durnin and Womersley (19). For girls <16 y of age (n = 18), we used the same formula (density = 1.1549 - 0.0678 x log 4 SFTs) as for girls <20 y of age.

UWW calculates body volume as the difference between body weight measured in air and that measured in water (based on Archimedes' principle that body volume is equal to the loss of weight in water). Modern hydrodensitometry systems consist of a scale within a large heated (37°C) tank of water. The subject exhales maximally, while totally immersed, and body weight is then recorded. Body weight (accurate to the nearest 10 g) in water is measured 6 times. The highest value of the measurements is used. Body density (D) is calculated with the following formula:

RESULTS  
Descriptive statistics
The patients' physical characteristics and body-composition variables, before and after weight gain, are listed in Table 1. The mean (±SD) BMI at admission was 14.9 ± 1.5, indicating severe underweight. The initial percentage body fat was 12.8 ± 3.6%, which is far below normal values for females (23–25%; 33, 34). The period of refeeding was 164 ± 19.8 d (range: 113–215 d).

View this table:
TABLE 1. Physical characteristics and body composition of 130 female anorexia nervosa patients before and after weight gain¹  
Body composition measured by anthropometry and densitometry before and after weight gain
The correlations between SFT and UWW percentage body fat values were 0.76 before weight gain and 0.71 after weight gain (Table 2). The difference between the mean percentage body fat by SFT measurement and by UWW before refeeding was -0.040%, with a 95% CI of -0.43%, 0.51% (t test statistic 0.17, with 129 df and an associated P = 0.9). After refeeding, the difference between the SFT and UWW measurements was -0.050%, with a 95% CI of -0.35%, 0.46% (t test statistic 0.25, with 129 df and an associated P = 0.8). The results of a t test showed that neither before nor after refeeding were the SFT percentage fat values significantly different from those by UWW.

View this table:
TABLE 2. Comparison of mean percentage body fat (%fat) estimated by underwater weighing (UWW) and by skinfold-thickness (SFT) equation in anorexia nervosa patients before and after weight gain¹  
E values are statistical manipulations that assess the total error derived when one procedure is validated against another. They include variation due to the lack of association between the 2 methods and variation due to the degree of mean difference between predicted and measured body fat (35–37). Therefore, E values will generally be similar in magnitude to SEEs for effective prediction equations. The E values before weight gain were larger than those after weight gain and were within acceptable limits (3.8% fat; 34, 38). The relation between percentage body fat before and after treatment, measured by rank correlation, was r = 0.44 (P < 0.0001). These results indicate that the ranking order before therapy was at least partially maintained after therapy (Figure 1).

View larger version (20K):
FIGURE 1. . Agreement assessment (Bland-Altman method) between percentage body fat estimated by underwater weighing and skinfold-thickness prediction in patients with anorexia nervosa before and after weight gain (n = 130). Means are plotted against the difference between the 2 procedures. The center line represents the mean difference between the 2 methods and the other 2 lines represent 2 SDs from the mean (95% CI).

 
Changes in body composition before and after refeeding
The refeeding protocol induced significant weight gain in all subjects as predicted. The mean weight gain was 11.9 kg (P < 0.0001) after 23 wk, which included significant increases in body fat (6.6 kg), percentage fat (9.8%), and fat-free mass (5.3 kg). We found a greater increase in fat mass (5.3 kg compared with 11.9 kg) than in fat-free mass (35.4 kg compared with 40.7 kg). Of the weight regained, 55.5% was fat. The ratio of fat-free mass to fat mass was 6.7:1 at admission and 3.4:1 at discharge, when 22.5% of body weight was fat. After the refeeding program, the mean body weight was only 6% below the ideal body weight (compared with 27% below ideal body weight at admission), according to the tables of the Metropolitan Life Insurance Company (39). After refeeding, only 10 patients had a BMI < 18. One patient reached a BMI >21.

A one-way ANOVA with repeated measures among the 3 subgroups based on the percentage body fat before weight gain for BMI and percentage body fat separately (Table 3) showed a significant effect for BMI (P < 0.0001) and percentage body fat (P < 0.0001). A post hoc Scheffe test showed significant differences before and after weight gain for all the variables in all groups. The 3 subgroups differed significantly from each other for the variables percentage body fat and BMI before weight gain. After weight gain, there was only a significant difference between the subgroup with >15% body fat and the other 2 subgroups for the variable percentage body fat.

View this table:
TABLE 3. Percentage body fat (%fat) and BMI before and after weight gain in 3 groups on the basis of %fat before weight gain¹  
Before weight gain, the AN patients of the restricting type (n = 97) had lower values and differed significantly from the binging-purging type (n = 33) for the variables weight, percentage body fat, fat mass, and fat-free mass measured by UWW (Table 4). After treatment, no significant differences were found between the subtypes of AN patients. The results for the estimation by SFT measurement were comparable. ANOVA did not show an effect of duration of illness or duration of treatment on body-composition variables measured by UWW after weight gain. There was no significant relation between duration of illness and the increase in weight, fat mass, or fat-free mass. The increase in percentage body fat was not related to length of treatment (r = 0.13, NS).

View this table:
TABLE 4. Body composition in anorexia nervosa (AN) patients of restricting type and binging-purging type before and after weight gain¹  
Changes in subcutaneous fat
A repeated-measures ANCOVA (Wilks = 0.15, P < 0.001) indicated that all the SFTs were significantly greater after weight gain. The lowest relative increases in subcutaneous fat thickness (<100%) were at the biceps, subscapular, chin, calf lateral, and calf medial sites. The greatest relative increases (>130%) were measured at the side, waist, thigh anterior, and thigh superior sites (Table 5).

View this table:
TABLE 5. Skinfold thickness in anorexia nervosa patients before and after weight gain  
Changes measured by SFT were highly correlated with weight gain. The highest correlation was found for the chin (r = 0.66, P < 0.001) and the sum of the 12 SFTs (r = 0.69, P < 0.001). The lowest correlation was found for the calf medial site (r = 0.35, P < 0.001). SFT showed a small correlation with fat-free mass (r = 0.20, P < 0.05 and r = -0.29, P < 0.01, respectively, before and after weight gain) but was moderately to highly correlated with percentage body fat (before weight gain: r = 0.51, P < 0.001 for abdomen to r = 0.74, P < 0.001 for triceps; after weight gain: r = 0.31, P < 0.001 for calf medial to r = 0.54, P < 0.001 for the side). These correlations did not differ markedly among the common sites.

DISCUSSION  
The estimation of body composition over time is difficult (34). Therefore, body-composition studies, including assessments of the changes after weight gain in large groups of AN patients, are rare. Nevertheless, this information is important from a medical and psychological viewpoint.

This study simultaneously compared 2 methods for assessing body-composition changes in a large group of AN patients. Bland and Altman (31) stated that the statistical procedure for assessing agreement between 2 methods is more sensitive to differences than to correlations. Two methods may be highly related yet have poor agreement. The agreement assessments in this study clearly show that the SFT method for estimating body fat agreed with the UWW method. For effective estimates of percentage body fat from SFT equations, the SEE should be similar to that obtained when estimating percentage body fat with a recognized standard such as UWW. The SEE values for percentage body fat were within the recommend limits (2.5–3.8%). Some reports (34, 35, 40, 41) suggest that effective SFT equations will generally estimate body fat to within 3.5% of fat values estimated by UWW. In the present study, SFT percentage body fat values within 3.5% of UWW values were measured in 81.5% and 85.4% of the patients, respectively, at the start and the end of treatment. The range of percentage body fat values within the 95% CIs (±2 SDs from the mean of the difference) at the end of treatment (-0.35%, 0.46%) was smaller than the range of 95% CIs for the differences in fat values (-0.43%, 0.51%) before treatment.

Because there was sufficient agreement between and among the results obtained by UWW and SFT measurement, our findings suggest that for an undernourished population, such as AN patients with relatively little body fat, simple SFT measurements are as useful as more sophisticated measures for the estimation of body composition. This is an important finding because SFT measurements can be used more easily in screening than can many other methods. Moreover, SFT measurement costs less and provides the opportunity to look at fat distribution. This agrees with the findings in the literature for normal subjects (42) and for AN patients (5). To this point, there has been no SFT equation developed specifically for female AN patients. Therefore, we used already published, generalized equations for white female subjects. Heyward and Stolarczyk (33) stated that the use of the Siri (23) equation to estimate percentage body fat from density can be questionable in an anorectic population. The Siri equation assumes that the density of fat-free mass equals 1.10 kg/L, that fat has a density of 0.90 kg/L, and that fat-free mass consists of 72% water (D = 0.9937 kg/L), 21% protein (D = 1.34 kg/L), and 7% mineral (D = 3.00 kg/L). Heyward and Stolarczyk (32) estimated that in AN patients, the components of fat-free mass were 76% water (D = 0.9937 kg/L), 17.7% protein (D = 1.34 kg/L), and 6.3% mineral (D = 2.73 kg/L). They proposed the following formula for calculating percentage body fat from density: percentage body fat = (5.26/D - 4.83) x 100. Compared with the values derived with this formula, the average percentage body fat of anorectic women will be systematically overestimated by 3–4% when the Siri equation is used to estimate percentage body fat from density. However, 2 problems arose when the Heyward and Stolarczyk formula was used: 1) for some test subjects (n = 4; AN restricting type; mean age: 25 y) a negative percentage body fat was obtained when density was calculated on the basis of SFT, and 2) when we applied this formula at admission, we wondered whether it would still be valid at discharge, when the test subjects would more closely resemble the general population.

The main conclusion of our study is that an inpatient refeeding program led to a significant increase in body weight and that body fat increased most significantly. The refeeding program was well accepted by the patients and no medical problems or adverse effects were observed. The mean weight after refeeding was 4 kg below the average ideal body weight according to the tables of the Metropolitan Life Insurance Company (39). More than half (55%) of the weight increase was attributed to an increase in fat mass. These findings agree with those of previous studies (10, 11, 13, 43) but not with others (6, 7). The low-to-very low percentage body fat findings before refeeding were similar to those reported in other studies that used the same methods. According to the criterion of Lohman (34), ie, percentage body fat <8%, 9.2% of our patients were at serious health risk. The percentage body fat increased on average by 9.8% to 22.6% after 6 mo. Although there is no general consensus in the literature about the normal percentage body fat for the general female population, the results of the refeeding program are in the range of normality (BMI > 18; percentage body fat > 22.5%). The ratio of fat-free mass to fat mass (3.4:1) after the refeeding program is in accordance with that of a normal-weight female population in the United Kingdom. Furthermore, we showed that patients with a low percentage body fat at admission usually had a low percentage body fat at discharge as well, and vice versa.

An analysis of the changes in percentage body fat and BMI between the groups with differing percentages of body fat before weight gain showed that the differences at admission disappeared at discharge. The same is true for the differences in the AN subtypes. Our findings are in accordance with those of Orphanidou et al (13). Additional analysis showed no effect of length of treatment or duration of illness on body composition after treatment. More rapid weight gain does not change the amount of fat gained. The greatest increase is not in the abdominal region, as many patients believe, but instead on the thighs, the waist, and the sides. Orphanidou et al (13) found a larger increase in fat in the abdominal region and less in the arms and calves. Ross et al (44) mentioned that relative increases in SFTs during weight gain may not be the same at different sites. In our study, changes in SFTs correlated highly with weight gain. The weight gain in our patients was substantial but it occurred gradually.

The amount of body fat may also affect the clinical outcome. Menses return with refeeding when patients reach higher percentages of body fat (18%); however, the changes in body fat and the return of menses may also enhance the psychological stress in patients, increasing the likelihood of relapse. The changes in body fat are also in contradiction with what patients want themselves, at least as long as they are anorectic. The mean percentage body fat after refeeding was still lower than in normal-weight females (23–25%), although most patients believe they are too fat and are afraid of becoming overweight. Therefore, it is important to include some sort of body-oriented therapy in the latter part of the treatment in addition to the refeeding program (27). The aim of this therapy is to educate the patients about their bad physical condition and to help them to accept the physical and psychological changes that result from increasing weight. It is possible that a fitness training program (20 sessions, 2 times/wk) during refeeding, supervised by a therapist who is familiar with the physical consequences of undernutrition, could increase their fat-free mass and redirect the patients' hyperactivity in a healthy way, allay their fears of weight gain, and improve their sense of self control (27, 45). The possible influence of an adapted fitness training program on body composition and body experience of AN patients has not been assessed in detail and will be an interesting topic for future research.

The therapeutic implications of this study are that patients have to be informed about the physiologic meaning of body fat. The therapist has to emphasize the importance of body fat for the return of menses and the normalization of reproductive functions. The therapist has to address the effects of low body fat on health, but in young patients in particular, also the effects of low weight on maturation, growth, and osteoporosis. Therapists should be aware that a normalization of body fat may also increase the patient's fat phobia, which led to her pursuit of thinness.

REFERENCES  

1. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th ed. Washington, DC: American Psychiatric Association, 1994.
2. Becker AE, Grinspoon SK, Klibanski A, Herzog DB. Eating disorders. N Engl J Med 1999;340:1092–8.
3. Probst M, Vandereycken W, Vanderlinden J, Van Coppenolle H. The significance of body size estimation in eating disorders: its relationship with clinical and psychological variables. Int J Eat Disord 1998;24:167–74.
4. Krahn DD, Rock C, Dechert RE, Nairn KK, Hasse SA. Changes in resting energy expenditure and body composition in anorexia nervosa patients during refeeding. J Am Diet Assoc 1993;93:434–8.
5. Probst M, Goris M, Vandereycken W, Van Coppenolle H. Body compositon in female anorexia nervosa patients. Br J Nutr 1996;76: 639–47.
6. Forbes GB, Kreipe RE, Lipinski BA, Hodgman CH. Body composition changes during recovery from anorexia nervosa: comparison of two dietary regimes. Am J Clin Nutr 1984;40:1137–45.
7. Russell D, Prendergast PJ, Darby PL, Garfinkel PE, Whitewell JR, Jeejeebhoy KN. A comparison between muscle function and body composition in anorexia nervosa: the effect of refeeding. Am J Clin Nutr 1983;38:229–37.
8. Russell JD, Mira M, Allen BJ, et al. Protein repletion and treatment in anorexia nervosa. Am J Clin Nutr 1994;59:98–102.
9. Fohlin L. Exercise performance and body dimensions in anorexia nervosa before and after rehabilitation. Acta Med Scand 1978;204:61–5.
10. Vaisman N, Corey M, Rossi MF, Goldberg E, Pencharz P. Changes in body composition during refeeding of patients with anorexia nervosa. J Pediatr 1988;113:925–9.
11. Polito A, Cuzzolaro M, Raguzzini A, Censi L, Ferro-Luzzi A. Body composition changes in anorexia nervosa. Eur J Clin Nutr 1998;52: 655–62.
12. Shinder SE, Shephard RJ. Relationship of premorbid mass and energy intake to increase of body mass during the treatment of anorexia nervosa. Int J Eat Disord 1993;14:65–73.
13. Orphanidou CI, McCargar LJ, Birmingham CL, Belzberg AS. Changes in body composition and fat distribution after short-term weight gain in patients with anorexia nervosa. Am J Clin Nutr 1997; 65:1034–41.
14. Zamboni M, Armellini F, Turcato E, et al. Body fat distribution before and after weight gain in anorexia nervosa. Int J Obes Relat Metab Disord 1997;21:33–6.
15. Heymsfield SB, Allison DB, Heshka S, Pierson RN. Assessment of human body composition. In: Allison DB, ed. Handbook of assessment methods for eating behaviors and weight related problems. Measures, theory and research. London: Sage Publications, 1995: 515–60.
16. Weiner JS, Lourie JA. Human biology. A guide to field methods. Edinburgh: Blackwell Scientific Publications, 1969.
17. Brown WJ, Jones PR. The distribution of body fat in relation to habitual activity. Ann Hum Biol 1977;4:537–50.
18. Jones PRM, Marshall WA, Branson SJ. Harpenden electric read-out (HERO) skinfold calipers. Ann Hum Biol 1979;6:159–62.
19. Durnin JV, Womersley J. Body fat assessed from total body density and its estimation from skinfold thickness: measurements on 481 men and women aged from 16 to 72 years. Br J Nutr 1974;32:77–96.
20. Leusink JA. A comparison of the body composition estimated by densitometry and total body potassium measurements in trained and untrained subjects. Pflugers Arch 1974;348:357–62.
21. Lasser RB, Levitt MD, Bond JH. The role of intestinal gas in functional abdominal pain. N Engl J Med 1975;293:524–6.
22. Lasser RB, Bond JH, Levii MD. The role of intestinal gas after ingestion of a standard meal. Gastroenterology 1976;70:48/906 (abstr).
23. Siri WE. The gross composition of the body. In: Lawrence JH, Tobias CA, eds. Advances in biological and medical physics. New York: Academic Press, 1956:239–80.
24. Vandereycken W, Meermann R. Anorexia nervosa: a clinician's guide to treatment. New York: Walter de Gruyter, 1984.
25. Vandereycken W. The management of patients with anorexia nervosa and bulimia: basic principles and general guidelines. In: Beumont PJV, Burrows GD, Casper RC, eds. Handbook of eating disorders. Part 1: anorexia and bulimia nervosa. Amsterdam: Elsevier, 1987:235–53.
26. Vandereycken W, Depreitere L, Probst M. Body-oriented therapy for anorexia nervosa patients. Am J Psychother 1987;61:252–9.
27. Probst M, Van Coppenolle H, Vandereycken W. Body experience in anorexia nervosa patients: an overview of therapeutic approaches. Eating Disord J Treatment Prev 1995;3:186–98.
28. Van Vreckem E, Vandereycken W. Sexual education programme for women with eating disorders. In: Dolan B, Gitzinger I, eds. Why women? Gender issues and eating disorders. London: Athlone, 1994:110–8.
29. Vandereycken W, Kog E, Vanderlinden J. The family approach to eating disorders: assessment and treatment of anorexia nervosa and bulimia. New York: PMA Publications, 1989.
30. Van den Broucke S, Vandereycken W, Norré J. Eating disorders and marital relationships. London: Routledge, 1997.
31. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307–9.
32. Statsoft Inc. Statistica for Windows (computer program manual). Tulsa, OK: Statsoft, Inc, 1998.
33. Heyward VH, Stolarczyk LM. Applied body composition assessment. Leeds, United Kingdom: Human Kinetics, 1996.
34. Lohman TG. Advances in body composition assessment. Champaign, IL: Human Kinetics, 1992.
35. Thomas KT, Keller CS, Holbert KE. Ethnic and age trends for body composition in women residing in the U.S. South West: II. Total fat. Med Sci Sports Exerc 1997;29:90–8.
36. Guo SS, Chumlea C. Statistical methods for the development and testing of predictive equations. In: Roche AF, Heymsfield SB, Lohman TG, eds. Human body composition. Champaign, IL: Human Kinetics, 1996:1191–202.
37. Mayhew JL Clark BA, McKeown BC, Montaldi DH. Accuracy of anthropometric equations for estimating body composition in female athletes. J Sports Med 1985;25:120–6.
38. Lohman TG. Applicability of body composition techniques and constants for children and youths. Exerc Sports Sci Rev 1986;14:325–57.
39. Metropolitan Life Insurance Company. Height and weight tables. Stat Bull Metropol Life Insur Co 1983;64:3–9.
40. Katch FI, Katch VI. Measurement and prediction errors in body composition assessment and the search for the perfect prediction. Res Q Exerc Sport 1980;25:120–6.
41. Katch FI, Katch VL, Behnke AR. The underweight female. Physicians Sports Med 1980;8:55–60.
42. Womersley J, Durnin JV, Boddy K, Mahaffy M. Influence of muscular development, obesity and age on the fat-free mass of adults. J Appl Physiol 1976;41:223–39.
43. Mitchell PB, Truswell AS. Body composition in anorexia nervosa and starvation. In: Beumont PJV, Burrows GD, Casper RC, eds. Handbook of eating disorders. Part 1: Anorexia and bulimia nervosa. Amsterdam: Elsevier-North Holland, 1987:45–7.
44. Ross R, Leger L, Marliss EB, Morris DV, Gaugeon R. Adipose tissue distribution changes during rapid weight loss in obese adults. Int J Obes 1991;15:733–9.
45. Beumont PJV, Arthur B, Russell JD, Touyz SW. Excessive physical activity: proposals for a supervised exercise program. Int J Eat Disord 1994;15:21–36.