This Article Abstract Full Text (PDF) Services Download to citation manager PubMed PubMed Citation

Body Composition Prediction Equations for Elderly Men

^a University of Houston-Clear Lake, Houston, Texas.
^b Texas A&M University, College Station, Texas (retired)

Terry L. Dupler, Fitness and Human Performance Laboratory, University of Houston-Clear Lake, 2700 Bay Area Blvd., Houston, TX 77058 E-mail: dupler{at}cl.uh.edu.

Decision Editor: William B. Ershler, MD

	Abstract

Top Abstract Methods Results Discussion Conclusion References

 
Background. The purpose of this study was to develop and validate anthropometric body composition prediction equations for elderly (i.e., 65 years old) men. This was necessary because of a lack of accurate and reliable predictive equations specifically developed for this population.

Methods. Seventy-five elderly men were randomly assigned to either an equation development sample () or an equation validation sample (). Subject anthropometric measures were analyzed in a regression procedure with hydrodensitometry-determined body density, percentage of fat, fat-free mass, and fat weight to develop prediction equations for each body composition variable. The equation estimates were then validated against the hydrostatically determined measures.

Results. Four equations were developed and validated for the estimation of elderly male body composition variables [one each for body density (, where SEE is the standard error of estimate), percentage of fat (), fat-free mass (, and fat weight (R)]. The equations provided estimates of body density, percentage of fat, fat-free mass, and fat weight, which were not statistically different from the hydrostatically determined criterion variables.

Conclusions. The results of this study indicate that accurate and reliable anthropometric predictive equations can be developed for an active and healthy elderly male population. These equations may be used for accurate epidemiological testing of this group's body composition variables.

KNOWLEDGE of the actual body composition changes associated with aging and the relationship of these changes with nutritional and health factors are required for valid medical diagnoses and prognoses or nutritional and pharmacological treatment ⁽¹⁾. The assessment of body composition is important in determining elderly people's nutritional status, protein mass, skeletal mineral status, energy expenditure, and relative level of hydration ^(2)(3)(4). Changes within these factors may signal declines in organ function, metabolism, nutrient intake and utilization, and resistance to traumatic and disease stresses ⁽¹⁾. Being overfat (i.e., obese) is a well-known contributor to medical conditions such as diabetes, cardiovascular disease, and hypertension. Researchers have demonstrated that negative changes in body composition (i.e., increased percentage of fat) precede the onset of many debilitating diseases such as diabetes, coronary artery disease, and hypertension ⁽¹⁾. Therefore the ability to accurately and reliably monitor the body composition changes associated with aging is imperative in the medical or clinical fields ⁽¹⁾⁽⁴⁾ and would be beneficial in the exercise and fitness fields ⁽⁵⁾. Indeed, in the fitness area, exercise intervention has been shown to slow the aging process per se ⁽⁶⁾; however, accurate predictive equations do not exist that allow the testing of these results (i.e., fat loss and muscle gain) in an epidemiological fashion. Nutritional and exercise interventions may be significant, but in order to be more effective, the age-associated changes in body composition must be understood ⁽⁷⁾.

Without adequate knowledge of the patient's current body composition status, medical personnel and exercise scientists may be hindered in their attempts to prescribe accurately medical, nutritional, and exercise intervention treatments. In summary, there is a need in the medical, clinical, and fitness fields for accurate predictive methods of determining elderly people's percentage of body fat (%F), fat-free mass (FFM), fat weight (FW), and body density (Bd) ⁽¹⁾.

	Methods

Top Abstract Methods Results Discussion Conclusion References

 
Subjects
Subjects for this study consisted of 75 healthy Caucasian elderly men between the ages of 65 and 81 years. The subjects were recruited from Houston (Texas) area senior adult groups. These groups, in general, offer various programs to senior adults such as exercise and nutrition classes, and the membership consists of a general cross-section of the local elderly adults. Every effort was made to ensure that a random cross-sectional sample was chosen, but one cannot overlook the fact that the more active and outgoing senior adults are more likely to participate in such programs and volunteer for such a study. All subjects were required to read and sign a University-approved informed consent form before any questioning or testing. Any questions pertaining to procedures or protocols were answered, and the subjects were informed they were free to discontinue the study at any time without penalty. Each subject chosen to participate in the study was free from any metabolic or musculoskeletal diseases, which may affect FFM or FW. Examples of such medical disorders include diabetes mellitus, muscular dystrophies, and Parkinson's disease. The health status of each subject was determined by the completion of a detailed medical history survey. Subjects were considered generally healthy if they were free from metabolic/musculoskeletal diseases, were not currently or had not been hospitalized within the previous 6 months, did not show obvious signs of edema, and had not undergone significant weight loss or gain (i.e., 10 lb) within the previous 8 weeks. Subjects with any diagnosed hypertension, hyperthyroidism, or hypothyroidism were not excluded from the data set. Of the 75 subjects, 3 reported being diagnosed with hypothyroidism, none with hyperthyroidism, and 22 with hypertension (of which 12 were taking medication for hypertension control).

The subjects were divided into equation development and validation samples. The sample for equation development consisted of 50 randomly chosen subjects from the 75 possible. The sample used for equation validation comprised the remaining 25 subjects.

For each subject, body weight (WT) and height (HT) were measured to the nearest 0.01 kg and 0.05 m, respectively, with a Detecto scale/stadiometer (Model 01151HTK). Body mass index (BMI) was calculated from the resulting HT and WT measures with the formula (where WT is in kilograms and HT is in meters).

Hydrodensitometry
Hydrodensitometry was used to obtain the criterion variables by which the equations were developed and validated. A modified hydrostatic weighing procedure, as described by Kohrt and colleagues ⁽⁵⁾, was used to obtain each subject's Bd. The procedure of Kohrt and colleagues (developed with an elderly sample group) allowed the subject to retain a partial lung volume (LV) during the immersion. Specifically, the modified procedure involved measuring the subject's vital capacity (VC) while the subject was seated in the hydrostatic tank with the water level just below the chin. When performing the immersion, the subject would perform a maximal inhalation, a measured exhalation, then underwater immersion. The volume of air exhaled was subtracted from the VC to determine the volume of air remaining in the lungs during immersion. The calculated remaining LV was added to the residual volume (RV) and gastrointestinal volume (GI) and the sum used in the calculation of the Bd. Kohrt and colleagues ⁽⁵⁾ reported a correlation coefficient of 0.99 between the conventional hydrodensitometric method (i.e., maximal exhalation) and their modified procedure. Kohrt and colleagues ⁽⁵⁾ also reported an intraindividual difference of <1%F between the two densitometric procedures. The modified procedure was utilized in this study to reduce the discomfort and apprehension associated with performing a maximal exhalation during hydrostatic weighing. A Collins Modular Lung Analyzer (Model 05015) was used to determine subject VC and to provide measures of the expired air before submersion.

The RV was determined with the closed-circuit oxygen dilution procedure described by Wilmore ⁽⁸⁾. The RV procedure was repeated until two tests resulted in measured volumes within 25 ml. GI was estimated at 100 ml ⁽⁸⁾.

Each subject's hydrostatic weights were determined with the use of a computerized hydrostatic weighing system. The system consisted of a Lebow 25-lb load cell (Model 3167-25) connected to a National Instruments analog–digital data collection system. National Instruments LabVIEW software for Windows (Model 776671-01) was used to collect, display, and analyze the hydrostatic weight (UWW) of each subject. The UWW was reported in kilograms and was the net weight after the tare weight was accounted for. A total of 10 trials was performed on each subject, with the five most consistent weights being averaged as the subject's mean UWW (e.g., typically the last five trials). Bd was determined by the formula ()}). The Brozek ⁽¹⁰⁾ formula was used to calculate %F. Multiplying the %F and the WT calculated the subject's FW. The FFM was calculated as WT - FW.

Skinfold and Circumference Measures
Skinfold and circumference measures were taken according to the procedures described by Heyward ⁽⁹⁾. Each subject, while standing, was measured on the right side of the body for determination of circumference measures and skinfold thickness. Skinfold and circumference sites were marked to improve the accuracy of each remeasurement. Skinfold thickness was determined with the use of Lange calipers at the following sites: subscapular (SS), at the inferior angle of the scapula with a diagonal fold; suprailiac (SI), at the iliac crest posteriorly of the midaxillary (MA) line with an anterior diagonal fold; MA, at the xiphisternal junction in a line vertically inferior of the axilla with a horizontal fold; and abdomen (AB), 3 cm laterally of the umbilicus with a vertical fold. Skinfold thickness measurements were taken in triplicate at each site and the mean used as the measure of skinfold thickness. Skinfold and circumference measures were obtained by one technician, who used the same calipers and tape throughout the duration of the study.

Circumferences of waist (WC), hip (HC), and thigh (TC) were determined with the use of a plastic-coated fiberglass tape (Grafco Model 17-1340-2) at the umbilicus, point of maximum posterior protrusion of the buttocks, and just inferior of the gluteal fold, respectively. The tape was spring loaded and marked to ensure that each circumference measure received identical tension. Circumference measures were taken in triplicate and averaged with the mean used as the measure of circumference. The waist-to-hip ratio (WHR) was calculated from the WC and HC means.

Data collection standardization procedures were included to minimize subject body hydration variation. Specifically, the procedures were (a) to require each subject to consume 1 cup of water immediately after signing the informed consent form; (b) to make sure the subjects did not perform any exercise or semistrenuous activity within 2 hours of testing; (c) to test each subject at least two hours after he had consumed a meal; and (d) to test only during the middle of the day (e.g., 10:00 AM–2:00 PM). Because total body water was not measured or estimated, an assumption was made that the aforementioned standardization procedures minimized subject hydration variation.

Statistical Analysis
A hold-out sample validation method was used to develop and test the validity of the new equations. According to this technique, the sample in the current study () was split randomly into two subsamples consisting of an equation development sample () and a validation sample (). The contained in the development sample was sufficient for prediction models with approximately 15 independent variables ⁽¹¹⁾. Multivariate analyses were utilized to test the differences between the equation development and equation validation samples. Specifically, Wilk's lambda was used to compare the independent and the dependent variables of both the equation development and validation samples.

The new prediction equations were developed with the equation development sample and validated by a test of their accuracy in predicting the dependent variables in the validation sample. Hydrodensitometrically determined %F, FW, FFM, and Bd were used as the criterion variables in the development and validation of the anthropometric prediction equations. Anthropometric independent variables (age, HT, WT, BMI, SS, SI, AB, MA, WC, HC, TC, WHR) were analyzed in a regression procedure with each dependent variable (%F, FFM, Bd, and FW) to develop the prediction equations. The model of variables deemed to provide the highest R² with the appropriate Mallow's Cp (i.e., Cp p) statistic was chosen as the anthropometric prediction equation for that specific dependent variable. Before an equation's inclusion as a possible prediction equation, a plot of the residuals tested each equation's lack of fit. The occurrence of any trends in the plot would indicate the need for a higher-order model or data transformation ⁽¹¹⁾. The equations chosen exhibited residual plots without trends and error rates within one standard deviation (SD).

The validity of the prediction equations was determined by a test of the predicted dependent variables [i.e, predicted fat percent (P%F), predicted fat-free mass (PFFM), predicted fat weight (PFW), and predicted body density (PBd)] to the criterion variables of the validation sample by a paired-comparisons t test of the differences.

The coefficients of reliability for the measures acquired during this study were calculated with a test–retest procedure for each of the independent variables. The r values indicate that there was little variation between measurement trials. The r values for the sample groups (equation development and equation validation combined) were .997 for SS, .996 for MA, .994 for SI, .992 for AB, .999 for WC, .999 for HC, .998 for TC, and .982 for underwater weight.

All statistical tests and analyses were performed with the software package Statistical Analysis System ⁽¹²⁾. The specific test and procedures were the procedure Proc-RSquare for the development of the prediction equations; Proc ANOVA (analysis of variance) with MANOVA (multivariate analysis of variance) options for the Wilks' lambda test of sample differences; and Proc Means with the T and PRT options for the paired-comparisons t test of the differences between the criterion measures and the new predicted measures ⁽¹²⁾.

	Results

Top Abstract Methods Results Discussion Conclusion References

 
The Wilks' lambda global test comparing the male sample equation development and validation sample variables was not significant (p .904), so univariate F tests were not required for each separate variable. However, summaries of the physical and anthropometric characteristics as well as the comparison tests between the sample groups are displayed in Table 1 and Table 2 , respectively. All values were similar between the two groups, with no variable being statistically different between the equation development and validation subsamples (Table 1 and Table 2 ).

The FFM of the subjects was most accurately predicted by the independent variables of WT, SI, WC, and HC (Table 3 ). The mean ± SD of the prediction for the FFM equation was 54.54 ± 7.36 kg (Table 4 ). The SEE for the equation was ±3.94 kg (Table 3 ). The difference between the FFM estimate and the hydrostatically determined FFM was shown to be not significantly different (p .7801) (Table 4 ).

The group's FW was most accurately estimated by the independent variables of HT, MA, and HC (Table 3 ). The mean ± SD of the prediction for the equation was 24.16 ± 6.43 kg (Table 4 ). The SEE for the FW equation was ±4.11 kg (Table 3 ). The FW estimate was shown to be not significantly different (p .9214) from the hydrostatically determined FW (Table 4 ).

	Discussion

Top Abstract Methods Results Discussion Conclusion References

 
Equation Validity
The results of this study indicate that accurate and reliable body composition predictive equations may be developed for active and healthy elderly men. This is important because of the increased awareness of elderly men with regard to their health and fitness. If some reliable measure of body composition status can be provided to elderly men they will then be better prepared to make dietary and exercise decisions in order to promote a healthier lifestyle.

Lohman ⁽¹³⁾ reported two criteria necessary for body composition estimate procedures. First, the new estimate procedures must produce estimates not significantly different from the criterion measures. And second, the SEE of the estimates must be comparable with expected values. The equations presented here meet these established criteria. The estimates produced by the four new equations were not significantly different from the hydrostatically determined criteria values, and the SEEs for the estimates were comparable (if not smaller) with the SEEs reported by other body composition prediction studies.

Previous researchers have published equations that were purported to be of use for elderly adults. Specifically, Deurenberg and colleagues ^(14)(15)(16), and Durnin and Womersley ⁽¹⁷⁾ have published equations that may be utilized to estimate elderly male body composition. However, these various equations have inherent problems that bring about increased prediction error and thus large inaccuracies. A primary reason for the lack of validity and reliability for these previously published equations is that they were developed with subjects other than elderly adults. Deurenberg and colleagues, for example, utilized a subject sample with wide age ranges (e.g., 7–83 years) for their equation development. Their total number of elderly subjects (i.e., >65 years, men and women) was only 50. Like Deurenberg and colleagues, Durnin and Womersley used a sample with wide age ranges (e.g., 17–72 years). Durnin and Womersley did not report the number of subjects over 65 years of age, but the mean age is well below what is generally considered elderly. Durnin and Womersley and Deurenberg and colleagues utilized these age ranges with the expectation of building general equations that may be used on subjects of virtually any age. However, in body composition analysis, especially methods that utilize anthropometric measures, the predictive equations must be population specific if measure validity and reliability are of interest ⁽²⁾⁽³⁾.

A second difference between the equations provided in this study and the previously published elderly body composition equations was the choice of anatomical site(s) chosen for measurement. Specifically, equations that utilize certain body sites may not provide accurate estimates of body composition for the elderly. For example, as aging progresses, elderly people are expected to see a change in subcutaneous fat storage ⁽¹⁸⁾. This rearrangement of fat storage from peripheral subcutaneous locations to intraabdominal locations will lessen the predictive power of typical skinfold sites ⁽¹⁸⁾. Body sites such as the biceps, triceps, subscapula, and pectoral can be expected to provide less insight to elderly adults' total body fatness ⁽¹⁸⁾. This same rearrangement of fat storage, however, may increase the predictive power of other anthropometric measures. Body measures such as WC, HC, SI, MA, WT, and HT may be expected to provide greater equation accuracy. The equations presented in this article represent these expected changes in prediction power. The Bd and %F equations utilize the WT, MA, WC, and WHR independent variables. Likewise, the FFM and FW equations use similar anthropometric sites (i.e., WT, SI, WC, and HC for FFM; HT, HC, and MA for FW). Unlike the previously published equations that use skinfold sites and anthropometric measures that may be adversely affected by the aging process, the current set of equations utilize anthropometric measures and skinfolds that should be more accurate predictors of elderly male body composition.

Equation Development and Equation Validation Sample Homogeneity
The pool of subjects chosen for this study represented a very homogeneous set of volunteers. Table 1 and Table 2 provide the mean ± SD for all measures. There were no statistical differences between the equation development and equation validation sample groups. In fact, using samples with such homogeneity will provide for less group variation, and this small amount of variation leads to small R². Historically, the R² value has been a criterion for choosing the predictive method to use. Many of the previously published body composition procedures possess R² greater than .90. These previously published procedures have such large R²s because of the variability of the subject pools used to develop the specific procedure. As may be seen in Table 3 , the equations presented here do not possess overly large R². This is due mainly to the difficulty in accounting for the extremely small amount of variation within the subject pool utilized to develop and validate the equations.

A possible problem with sample groups with such homogeneity is that it would be easier to cross-validate any given equation or predictive measure developed with the other group. With that in mind, one should be aware that the equations presented in this article were developed and validated with an active and healthy senior adult sample.

Criterion Measures
The use of hydrodensitometry as a means of deriving criterion measures on elderly adults has been criticized by some ⁽²⁾. Baumgartner reports that the changes in bone mineral, body hydration, and loss of lean mass (i.e., muscle) will cause overestimates of elderly body density when hydrodensitometry is used ⁽²⁾. Fuller and colleagues ⁽¹⁹⁾ and Heymsfield and colleagues ⁽²⁰⁾ report finding no differences in elderly Bd when determined by hydrodensitometry and a four-compartment model. Other groups ^(3)(21)(22) believe that the changes associated with aging do not necessarily invalidate the hydrodensitometric process. Cohn and colleagues ⁽²¹⁾, Going and colleagues (23), and Lohman ⁽²²⁾ report that elderly Bd does not necessarily change due to the way each specific factor changes. Specifically, the contribution of body hydration, mineral status, and protein changes in such a way as to keep elderly FFM density relatively stable ^(3)(21)(22).

Several groups of researchers have utilized hydrodensitometry to determine criterion measures ^(5)(15)(16). These groups did report concerns with the hydrodensitometric process and the associated changes in elderly Bd. However, with the lack of inconclusive data regarding hydrodensitometry and elderly Bd, these groups decided that hydrodensitometry was a valid and reliable procedure for criterion measures ^(5)(15)(16).

	Conclusion

Top Abstract Methods Results Discussion Conclusion References

 
In conclusion, this study has presented four equations by which the body composition of active and healthy elderly men may be determined in an epidemiological manner. Specifically, the equations were developed with the intent of being used to promote the health and well-being of elderly men.

Received January 8, 1999

Accepted September 2, 1999

	References

Top Abstract Methods Results Discussion Conclusion References

 

1. Kuczmarski RJ, 1989. Need for body composition information in elderly subjects. Am J Clin Nutr. 24:1150-1157.
2. Baumgartner RN, 1993. Body composition in elderly persons: a critical review of needs and methods. Prog Food Nutr Sci. 17:223-260. [Medline]
3. Going S, Williams D, Lohman T, 1995;23. Aging and body composition: biological changes and methodological issues. Holloszy JO, , ed.Exercise and Sports Science Reviews 411-458. Williams and Wilkins, Baltimore.
4. Heymsfield SB, Wang J, Lichtman S, Kamen Y, Kehayias J, Pierson RN, Jr 1989. Body composition in elderly subjects: a critical appraisal of clinical methodology. Am J Clin Nutr 50:1167-1175.
5. Kohrt WM, Malley MT, Dalsky GP, Holloszy JO, 1992. Body composition of healthy sedentary and trained, young and older men and women. Med Sci Sports Exercise 24:832-837. [Medline]
6. Rogers MA, Hagberg JM, Martin WH, III Ehsani AA, Holloszy JO, 1990. Decline in VO₂max with aging in master athletes and sedentary men. J Appl Physiol 68:2195-2199. [Abstract/Free Full Text]
7. Morley JE, Moorandian AD, Silver AJ, Heber D, Alfin-Slater RB, 1988. Nutrition in the elderly. Ann Intern Med. 109:890-904.
8. Wilmore J, 1969. A simplified method for determination of residual lung volumes. J Appl Physiol. 27:96-100. [Free Full Text]
9. Heyward VH. Advanced Fitness Assessment and Exercise Prescription. Champaign, IL: Human Kinetics Books; 1991:147–150.
10. Brozek J, Grande F, Anderson JT, Keys A, 1963. Densitometric analysis of body composition: revision of some quantitative assumptions. Ann New York Acad Sci 110:113-140.
11. Ott L. An Introduction to Statistical Methods and Data Analysis. 3rd ed. Boston: PWS-Kent; 1988:540–619.
12. SAS Institute; 1985. SAS Users guide: Statistics. Version 5 edition SAS Institute Inc, Cary, NC.
13. Lohman T, 1981. Skinfolds and body density and their relation to body fatness: a review. Hum Biol. 53:181-225. [Medline]
14. Deurenberg P, van der Kooy K, Evers P, 1990. Assessment of body composition by bioelectrical impedance in a population aged > 60 y. Am J Clin Nutr. 51:3-6. [Abstract/Free Full Text]
15. Deurenberg P, van der Kooy K, Leenen R, Westrate JA, Seidell JC, 1991. Sex and age specific prediction formulas for estimating body composition from bioelectrical impedance: a cross-validation study. Int J Obesity 15:15-25.
16. Deurenberg P, Westrate JA, Seidell JC, 1991. Body mass index as a measure of body fatness: age and sex specific prediction formulas. Br J Nutr 65:105-114. [Medline]
17. Durnin JVGA, Womersley J, 1972. Body fat assessed from total body density and its estimation from skinfold thickness: measurements on 481 men and women aged from 16–72 yr. Br J Nutr. 32:77-97.
18. Baumgartner RN, Rhyne RL, Troup C, Wayne S, Garry PJ, 1992. Appendicular skeletal muscle areas assessed by magnetic resonance imaging in older persons. J Gerontol Med Sci. 47:M67-M72.
19. Fuller NJ, Jebb SA, Laskey MA, Coward WA, Elia M, 1992. Four-compartment model for the assessment of body composition in humans: comparison with alternative methods, and evaluation of the density and hydration of fat-free mass. Clin Sci. 82:687-693. [Medline]
20. Heymsfield SB, Wang J, Lichtman L, Kamen Y, Kehayias J, Pierson R, Jr 1989. Body composition in elderly subjects: a critical appraisal of clinical methodology. Am J Clin Nutr. 50:1167-1175.
21. Cohn SH, Vartsky D, Yasumura S, Sawitsky A, Zanzi I, Vaswani A, Ellis KJ, 1980. Compartmental body composition based on total-body nitrogen, potassium, and calcium. Am J Physiol. 239:E524-E530. [Abstract/Free Full Text]
22. Lohman T. Advances in body composition. Current Issues in Exercise Science. Monograph 3. Champaign, IL: Human Kinetics, 1992.

This Article Abstract Full Text (PDF) Services Download to citation manager PubMed PubMed Citation