You are on page 1of 3

Replicability of measurements of density

of the human body as determined


by underwater weighing

J. V. G. A. DURNIN AND A. TAYLOR1


Institute of Physiology, University of Glasgow, Scotland; and the Ministry
of SupPly, Chemical Defence Experimental Establishment, Porton, England

DURNIN, J. V. G. A. AND A. TAYLOR. Replicability of measure- while the subjects were in apparent calorie balance.
ments of density of the human body as determined by underwater weighing. This is a brief description by Keys and his colleagues
J. Appl. Physiol. r5( I) : 142-144. rg6o.-Ten healthy young (4, section on methods), where some of their subjects
men were maintained for a period of 2 weeks on a diet, the
had duplicate determinations of densitv / made about
calorific value of which exactly equaled their energy expen-
diture. During this period measurements of body density by
4 days apart.
underwater weighing were made on several occasions to
Since it is common to repeat measurement of densitv
determine the reproducibility. The standard error of a single after periods of weeks or months in order to find whether
observation was 0.0023 units of density. The replicability of the proportion of fat in the body is altered, it would also
the method seems satisfactory, since the formulas used to be useful and interesting to know the reproducibilitv , of
estimate total body fat do not justify a technique where the these measurements of density over a period of I or 2
accuracy would be better than that obtained here. weeks. This involves certain added difficulties. Regard-
less of other reasons for variability, if there are am-.
changes in the net body weight (excluding water) then
this would be reflected in a true change in the density.
S INCE THE DESCRIPTION BY Behnke, Feen and Welham By preventing any calorie imbalance, that is, by ensuring
that the calorie intake equals the calorie expenditure,
(I) of the possible use of the measurement of density for
calculating the amount of fat in the human body, there then the sources of variability in the measurement of
have been many descriptions of the technique of densi- density would still remain those of technique. It is true
tometry, and its accuracy has been discussed in some that other small chemical changes might occur, but it is
detail by several authors, e.g. Keys and Brozek (2) and highly improbable that in any normal, healthy, voung
Von Dijbeln (3). There have also been attempts to man these changes would be of measurable significance.
determine the replicability of these measurements. In the present experiment a group of IO men was
However, ‘replicability’ has usually meant the repetition maintained for a period of 2 weeks on a diet, the calorific
of the measurement of the underwater weight at intervals value of which exactly equaled the energy expenditure,
of either a few minutes or, occasionally, I or 2 hours.
so that throughout the period of study they remained
This of course is a useful assessment, although obviously in fairly precise calorie balance. The body density of
the results will vary, in different laboratories, with the these men was measured by weighing the subjects in
fundamental accuracy of the technique used and with air and under water on several occasions by the usual
the care which is paid to details of the technique. This technique.
may be important to remember since the very elaborate
technique of some laboratories can never be generally METHODS

reproduced and the replicability described by some of The subjects were healthy young men of the Roval ,
the best-known exponents of these measurements may Air Force Regiment. Their ages, heights and weights are
not be universally applicable. shown in table I. Measurements of body density were
There appears to be only one description of an attempt made before IO:OO A.M., the subjects having fasted
to find the reproducibility of measurements of the overnight.
density of the human body over a period of a few days For the underwater weighing, each man sat on a
Received for publication May I I, I 959.
weighing platform completely immersed in a tank of
l Present address: Dept. of Physiology, London Hospital Medi- water, 2.5 m x 1.5 m and 1.5 m deep. He wore a nose-
cal College, Turner Street, London, E. I, England. clip and mouthpiece connected via a remotely operated

Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (130.063.180.147) on December 7, 2018.


Copyright © 1960 the American Physiological Society. All rights reserved.
REPLICABILITY OF HUMAN DENSITY MEASUREMENTS ‘43

valve and two pairs of flexible hoses to the atmosphere TABLE 2. Analysis of Variance of Results
on the one hand and to a reservoir of pure oxygen and a Detailed in Table I
collecting bag for expired air on the other. The subjects Variation Due to DF Sum of Mean Square
were completely accustomed to the procedure before Squares
any experimental measurements were done. The Subjects 9 3302.7 366 * 967
temperature of the water was taken carefully and never Error 33 180.5 5.470
varied more than between 33O and 35OC. The weight Total 42 3483 * 2
while under water was taken from a calibrated spring Coefficient of variation = 0.227&
balance at the end of a full expiration and, at this
moment, the breathing tube connections were changed rection was made for the percentage of nitrogen re-
so that the next inspiration was of pure oxygen from the maining in the alveolar air at the end of the washout
reservoir. The residual air was assessed by the nitrogen period. A small correction was also used for the nitrogen
washout method, the percentage of nitrogen being liberated from solution in the blood. This residual air
analyzed by means of a Van Slyke apparatus. Cor- estimation is essentially similar to that referred to by
Brozek, Henschel and Keys (5). Weighings were made
TABLE I. Ages, Heights, Weights in Air and Bodv on the ten subjects on alternate days; in seven men
Densities of Subjects these consisted of five separate measurements, in one
Subj. Age, Y’* Ht., cm Wt. in Air Density man of four and, in the case of two subjects, only two
kg kg/m3 readings were obtained (these last two subjects refused
WI 22 ‘74 64.13 1066 to continue with the dietary restrictions and were
64.62 Id9 returned to their unit. The subject on whom four
D 22 170 64.12 I 064
1071 measurements were made on alternate days, developed
64.05
H 21
169 64.25 I 078 influenza before the experimental period could be
63.70 ‘073 completed.). On a separate occasion, four measurements
63.96 I 078
of density were made during the space of 155 hours on
63.68 ‘074 one of the men.
63 -9 I 076
SI 22 54.60 I 076 An analysis of variance was carried out on the results
‘75
54.49 obtained.
54.94 1070
54.86 1068 RESULTS
54.49 1068
D ‘9 I85 64.60 I 086 Table I shows the ages and heights of the subjects
63.81 1081 and also their weights in air and their density on the
63.99 I 084
occasion of each underwater measurement. There was a
63 -58 I 083
I 084
small variation in the weights of the subjects in air
63 -41
s2 23 189 75.42 1086 which, since the men were in calorie balance and since
75.46 1081 they were always weighed under identical conditions
75.9 1087 with regard to time of day, with the bladder empty and
74.84 1086
I 084
in the fasting condition, was presumably due to changes
74.73
18 in the fluid balance.
s3 ‘77 62.94 ‘077
62 .gr ‘075 The four measurements made within I 35 hours on one
62.60 1078 subject would appear to show that the replicability of
62 .OI 1076 our measurements was comparable in accuracy to those
62 -59 ‘075 described by other authors using a rather more elaborate
I 056
w2 36 174 77.20
76.79 1051 technique.
76.33 ‘055 Table 2 summarizes the analysis of variance of the
I 056
76.05 results on the density measurements. The standard
76.02 ‘053 error of a single observation was 0.0023 units of density
L 19 ‘73 64.78 ‘075
so that the go % limits of confidence of one observation
63.94 ‘077
640 ‘4 IO77 is likely to be =t twice the standard error, i.e. 0.0046
64.11 ‘074 units. Since the standard error is inversely proportional
63.82 ‘075 to the square root of the number of observations, if
A 21
I73 70.08 ‘075 four observations were made, then the standard error
69.72 ‘079
69.60 1076 0.0023
would be -/- = 0.001 I 5.
- . 69.44 ‘073 1/4
s3 4 times within 135 hr. 1079
1078
I 080
DISCUSSION
I 078
There is now widespread use of the concept of using
Weighings were performed on alternate days. fat-free body mass as a reference unit instead of gross

Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (130.063.180.147) on December 7, 2018.


Copyright © 1960 the American Physiological Society. All rights reserved.
‘44 J. V. G. A. DURNIN AND A. TAYLOR

body weight. There is a certain amount of confusion of changes of body weight would appear to make nonsense
thought in the use of reference standa .rds of this natu re, of attempting to define techniques to the extent where
particularly when applied to the measurement of the error is smaller than 0.004 per unit of density. The
energy expenditure. Some of these have been discussed results of Keys and his colleagues (4) showed that their
by Durnin (6). It is by no means sure that the fat-free duplicate values for density differed in go % of the cases
body mass is necessarily a more useful unit than gross by less than &0.003 units. Our own results, using a less
body weight in many circumstances and it has almost elaborate technique, showed that the error of a single
certainly been frequently misused. However, knowledge measurement in go % of the cases was likely to be
of the fat-free body mass is often of value, and it is ~to.004 units. The replicability of the method seems,
therefore important to be aware of the accuracy both of therefore, to be of a satisfactory degree of accuracy.
the theoretical validitv of its derivation and of the
technique. The former has been discussed on numerous
occasions, notably by the Committee on Foods (7) and This work was carried out at the Ministry of Supply Chemical
Defence Experimental Establishment at Porton, England, while
by Von Diibeln (3).
one of the authors (J. D.) was there as a Vacation Consultant
The difference of 0.001 of a unit of density will make Physiologist. We are much indebted to Dr. W. S. S. Ladell,
about 0.4% difference in the calculation of the per- Superintendent, Medical Division, for the facilities and laboratory
centage of fat in the human body. Therefore, 0.004 space provided. We are also very grateful to Dr. R. A. Robb of
units of density will affect the end results of the amount the Department of Statistics, University of Glasgow, for advice
and assistance with the statistical analysis of the results. During
of body fat by about 155 %. The formulas used in the period of these experiments one of us (J. D.) was in receipt
estimating the total body fat and for estimating the of a research grant from the Scottish Hospital Endowments
composition of the tissues gained or lost during the Research Trust.

REFERENCES

I. BEHNKE, A. R., B. G. FEEN AND W. C. WELHAM. J.A.M.A. 5. BROZEK, J., A. HENSCHEL AND A. KEYS. 3. App/. Ph_ysiol. 2:
I 18: 495, 1942. 240, 1949.
2. KEYS, A. AND J. BROZEK. Physiol. Rev. 33: 245, 1953.
6. DURNIN, J. V. G. A. Brit. 3. Nutrition I 3 : 68, 1959.
3. VON D~BELN, W. Acta physiol. scandinav. 37 : Suppl. I 26, I, 1956.
7. SPECTOR, H., M. S. PETERSON AND T. E. FRIEDEMANN (editors).
4. KEYS, A., J. BROZEK, A. HENSCHEL, 0. MICKELSEA AND H. L.
TAYLOR. The Biolog of Human Starvation. Minneapolis: Univ. Methods for Evaluation of Nutritional Adequacy and Status. Nat.
of Minnesota Press, I 950, p. 1080. Acad. Sci.-Nat. Res. Council, Washington, D. C., 1954.

Downloaded from www.physiology.org/journal/jappl by ${individualUser.givenNames} ${individualUser.surname} (130.063.180.147) on December 7, 2018.


Copyright © 1960 the American Physiological Society. All rights reserved.

You might also like