want. le “ede,
For the remaining equations, correlations with laboratory measures for the appropriate age-sex-race groups were satisfactory but in only one instance were they as
high as those obtained in the present study.
The linear regression equations developed for five race-sex—age categories esti-
to
R. C, STEINKAMP, ef al,
tg Men a ye
Lau
mate total body fat from four variables with multiple correlations of 0.918 to 0.980.
cad
The standard errors in kilograms body fat range from 2.044 to 3.665. The highest
standard error was obtained for the Negro men, perhaps because the age range was
greater for this group. Even so, on the basis of a 70 kg man, the error of fat prediction in this group would be +5.2 per cent of total body weight. These regression
on determinations of body fat from laboratory measurements of total body water
(tritium) and body density by helium dilution on 167 subjects randomly selected
from a total of 2053 subjects who were measured anthropometrically. Multiple
correlation coefficients on kilograms body fat for all equations were 0.897 or greater.
of
For 68 subjects, comparison has been made of the lean body weight obtained
from K'° measurements with lean body mass calculated from total body weight and
body fat determinations.
Calculations of either body fat, total body weight, fat-free body weight, percentage
body weight as fat or potassium equivalents according to anthropometric formulas
ee ee
inclusive: the lowest was 0.918, obtained for white men of the same age.
ee
With four variables, the highest r was 0.980 obtained for women aged 35-44 years,
have been correlated with the appropriate measured variable for the data obtained
in this study.
The methodology of predicting body composition from anthropometric measure-
ments provides a useful tool in the further epidemiological and clinical study of
chronic diseases associated with overweight.
Acknowledgements—We are indebted to the subjects whose participation made this study
bot
2
BFse St
Appreciation is expressed to Miss Mary Sprotr and Mr. Don CHAFFEE for computer programming and to Mrs. OLGA PoLivKa and Mrs. ZELNER HANDLEY for their careful clerical work.
icalaA
possible. Our grateful acknowledgement is made to management for allowing employee time
to be used.
1at
Bisbee
~
il
Taito aedonaes Se Cee ton cet PhtSic Tiaa at Pe ecm ngs AB aden THe
peas
Cin
na
Study of body composition is important in epidemiology, medicine, nutrition,
surgery, physiology and other related fields. While laboratory methods are sufficiently refined to determine body fat and lean body mass with accuracy, they remain
specialized research tools. In this and the preceding paper, regression equations
using anthropometric measurements are developed by which total body fat and lean
body mass can be predicted easily with an accuracy suitable for clinical and
epidemiological studies.
Regression equations for estimating body fat from two, three and four anthropometric variables are presented for five race-sex—age categories of healthy adults and
for combinations of these subject categories. Validation of the equations was based
a eee
SUMMARY
Le
[10, 15, 16, 26]. The predictability of body fat provided by the regression equations
is completely adequate for epidemiological and mostclinical purposes.
tn
equations for estimating body fat provide greater accuracy and have been derived
from laboratory measurements of greater validity than others previously developed