Validity of skinfold thickness formerly obese adults17

Joanne and ScherfMS,3 Melvyn Rubenfire, Barry MD7 A Franklin, PhD,4 Charles

measures
P Lucas, MD,5

of
Dallas Stevenson, MS,6

ABSTRACT To assess the validity of skinford thickness estimates of body fatness in formerly morbid obese adults, 23 patients (17 women, 6 men) who had completed a protein-sparing modified fast were studied. Mean SD weight loss was 60.7 20.6 kg for men and 42.6 1 1.5 kg for women. Body density and percent body fatness were determined afterweight loss according to four commonly used skinfold equations: Pollock(P); Durnin-Rahaman (D-R); Durnin-Womersley (D-W); and, Jackson-Pollock (J-P). The validity of these measurements was assessed by hydrostatic weighing, which revealed a percent body fatness of2O.4 6.5 for men and 29.8 for women. 8.4 The mean difference and total error (square root of the of squared deviations) between skinfold predicted and hydrostatically-determined percent body fatness for each skinfold equation were: P, 2.0 and 4.9; D-R, 4.2 and 6.6; D-W, 7.1 and 8.4; and, J-P, 0.7 and 4.4. With the exception of the latter equation, all significantly overestimated (p 0.01) hydrostatically-determined < percent body fatness. Conclusion: Select skinfold equations may result in a marked overestimation of body fatness in formerly obese patients. Am J C/in Nuir l986;43: 128-135. KEY WORDS method, body density Protein-sparing modified fast, skinfold prediction equations, hydrostatic criterion

Downloaded from www.ajcn.org by guest on February 14, 2012

Introduction

Individuals undergoing rapid nounced weight loss require accurate sessment of body fatness to determine priate and safe limits for weight reduction. program Empiric experience has shown that after mas- Weight reduction The weight reduction program included behavior sive weight reduction such persons are often and a protein-sparing modified fast, which characterized by excessive unsightly subcuta- modification lasted from 16 to 25 wk. During the rapid weight loss neous tissue, despite attainment of normal or phase, patients consumed 300-500 cal/day of a proteinnear normal body weight and relative fatness. carbohydrate formula diet that contained 45-70 g of proThis situation offers the potential for considerable error when predicting body fatness using I From the Section of Cardiovascular Diseases, Destandard skinfold equations. partment ofMedicine (iS, BAF, MR), Sinai Hospital and The purpose ofthe present study was to as-Wayne State University, Detroit, MI and Obesity & Risk sess the validity of four commonly used skin- Factor Program (CPL, DS), University Health Center, Wayne State University, Detroit, MI. fold thickness equations for determination of 2 Address reprint requests to: Ms J Scherf, Cardiovasrelative body fatness in formerly morbid obese cular Fitness & Rehabilitation Program, Section of Caradults, utilizing hydrostatic weighing as thediovascular Diseases, Department of Medicine, Sinai criterion method. Hospital ofDetroit, 14800 McNichols, Suite 304, Detroit, Methods
Subjects Twenty-three ofage I 28 ( age
=

ically supervised weight reduction program, were studied. Informed consent was obtained in accordance with the Hospitals Medical Research Committee. and pro- policies ofSinai serial as- Mean weight losses were 60.7 kg for men and 42.6 kg for women. Physical characteristics ofthe subjects are shown appro- in Table 1.

patients 42 yr), who

(17 women, had recently

MI 48235. 3 Exercise Specialist, Cardiovascular Fitness and Rehabilitation Program. Program Director, Cardiovascular Fitness and Rehabilitation Program. Medical Director. 6 Behavioral Medicine. Chairman. 6 men) 28 to 59 yr Director, Received February 1, 1985. Accepted for publication July 16, 1985. completed a medJournal ofClinical Nutrition 43: JANUARY 1986, pp 128-1 35. Printed in USA 1986 American Society for Clinical Nutrition

The American

SKINFOLD
TABLE 1

VALIDITY

OF

THE

FORMERLY

OBESE

I 29

by body plethysmography duringthe week ofthe scheduled hydrostatic weighing. This method of measuring residual lung volumes has been found to give reproducible and MIn(n=S) Womn(n=17) accurate values, which in normal subjects, does not differ AtSO VarIabI *$D significantly from values obtained by an open circuit 42.5 6.5 41.6 10.1 Ag (yrs) washout method (3). Although administered separately, 181.6 6.8 164.6 7.9 HIIght(cm) the subjects assumed similar sitting positions for both the 69.0 13.1 WIIght (kg) 93.2 11.6 42.6 I 1.5 Weight Loss (kg) 60.7 20.6 hydrostatic weighing and residual volume measurements. The percentage of body fat for both skinfold and hydrastatic methods was calculated from body density using the Sin (4) equation, where percent body =fat 100 [(4.95/ tein (egg white) and 30 g of sucrose, supplemented with Lean body weight and body fat vitamins and minerals. Strict adherence to the regimen body density) - 4.50]. calculated from total body weight and percent assured a weight loss of 1.4 to 2.3 kgJwk. Upon attainment were body fat. of an arbitrarily determined goal weight, subjects began gradual reintroduction offood followed by a maintenance phase. Statistics
PhysIcal CharacterIstIcs of the Formerly Obese Men and Women

Physical characteristics and body composition measures were expressed as means and standard deviations. The Percent body fatness was determined in all subjects 4 mean difference and total error between skinfold predicted to 8 wk after the weight loss phase was completed and and hydrostatically-determined percent body fatness for duringa period ofstable weight maintained fora minimum each skinfold equation were calculated using the crossof4 wk. Body composition was assessed by four commonly validation procedures of Lohman (5). Total error (E) was used skinfold prediction equations and hydrostatic weighdetermined as E= Y(Y-Yi)2/N, where Y is the skinfold ing. Subcutaneous skinfold thickness was assessed by a predicted body fatness and Yi is the hydrostatically-deLange caliper according to the method of Behnke and termined value (5). t test for paired observations was used Wilmore (1). All measurements were taken by the same to calculate differences between skinfold and hydrostatiinvestigator on the right side ofthe body and recorded to cally-determined percent body fatness for each skinfold the nearest millimeter. The average of three consecutive equation. A p value <0.05 was considered statistically sigliftings and measurements was noted as the skinfold thicknificant. ness at each anatomical site: biceps, triceps, scapula, suprailium, thigh, chest, axilla, and abdomen according to the specific skinfold equation (Table 2). Body density wasdetermined by measuring body weight Results in air and calculating body volume from an underwater weighing according to standard methods (2). Each subject Body composition ofthe 23 formerly obese reported to a swimming pool (average water temperature, subjects, calculated from residual lung volume 28#{176}C) and, while seated in a chair suspended in 1.2 m of body weight, and hydrostaticallywater from a calibrated Chatillion scale, crouched forward measures, body density, is shown Table in in a tuck position until totally submerged. The underwater determined weight was noted after forced expiration during breath 3. Mean SD percent body fatness after rapid holding. A minimum ofeight to ten underwater weighings weight loss was 20.4 6.5 and 29.8 8.4 for were recorded (to the nearest 100 g) and the mean of the men (n = 6) and women (n= 17) respectively. highest three weights was taken as the true underwater values are equivalent to the average of weight. Residual lung volume, employed as a correction These factor in the calculation ofbody density, was determined the adult population. Body composition assessment

Downloaded from www.ajcn.org by guest on February 14, 2012

TABLE

Anatomical Sites for Skinfold Measurements

Pollock Chest Axilla Thigh (W) (M) (6,7) Durnin. Rahaman Triceps Subscapular Supraillium (8) Durnin. Womersley Biceps Triceps Subscapular Supraillium (9) Jackson. Pollock Chest Abdomen Thigh Triceps Supraillium
(M)
=

(10,11)

(M) (M) (W) (W)

Men

(W)

Women

130
TABLE 3

SCHERF

ET

AL

The mean difference and total error between skinfold and hydrostatically-determined percent body fatness for the four skinfold equations are presented in Figures 1-4. Pollocks M.nhi=S) Wom.n(n=17) Vartabl z$D *SD equation (6, 7) yielded a mean difference and RssduaI Volunis, (L) 2.26 .93 1.76 .43 total error of2.0 and 4.9% respectively. Values D.nslty#{149}, (gm/cc) 1.052 0.015 1.032 0.018 of 4.2 and 6.6% were calculated for the DurBody Wei9ht, (kg) 93.26 11.56 69.08 13.10 L#{149}#{149}n Wslght. Body (kg) 73.92 8.52 47.66 5.50 nin-Rahaman formula (8). The DurninFat W.ght, (kg) 19.34 7.48 21 .42 9.51 Womersley prediction (9) showed the highest Body Fate. % 20.4 6.5 29.8 8.4 mean difference (7. 1%) and total error (8.4%), Osr-y cacistsd accodmg ro undsrwatec weEgh.ng pocdur o Go1dm& id overestimating relative fatness in all subjects. Buskt* (2) #{149}P-..-c#{149}r.t caiculatd body lit lrn body d.nssry by the mfhod Sn 01 (4) wh.re % Fat The Jackson-Pollock equation (10, 1 1) exhib= ,fx, 95/body d.nssty)-4 50/ ited only slight variation with the hydrostatic criterion method, with a mean difference and total error of 0.7 and 4.4% respectively. The four skinfold equations yielded a wide Mean percent body fatness calculated from range in estimated body fatness in these sub-the four skinfold equations and hydrostatic jects: Pollock, 14-45%; Durnin-Rahaman, 19-criterion method are shown in Figure 5. Hy43%; Dumin-Womersley, 21-48%; and, Jack- drostatically-determined percent body fatness son-Pollock, 14-40%. With the exception of all subjects for (n 23) averaged 27.3%. Mean the latter equation, all significantly over-skinfold estimates ofrelative body fatness were estimated hydrostatically-determined percent 29, 32, 35, and 28% for the Pollock, Durninbody fatness regardless of subject age or sex. Rahaman, Durnin-Womersley, and JacksonGraphic illustration of overestimation trends Pollock equations respectively. With the cxshowing the individual variation between hyception ofthe latter equation, all significantly drostatic and skinfold-determined body fatoverestimated (p < 0.01) hydrostatically-dcness are shown in Figures 1-4. termined percent body fatness.
Body ComposItion Calculated from Residual Volume Measures, BodyWeight, and Density In Formerly Obese Men and Women
it. ((4

Downloaded from www.ajcn.org by guest on February 14, 2012

0
_1 _1 0 U)
C)

w
2
I-

25
>-

Difference a

2.0%

0 0

Totol
#{163}

Error
(n=6)

4.9%
(n:17)

20 5.

Men

Women .01

0#{149} 0 5 % FIG 1. Comparison of percent body fatness 20

BODY

25
FATNESS

30

35 40
(Hydrostatic)

45

50

determined

by the

Pollock

skinfold

equation

and

hydrostatic

weighing.

SKINFOLD 50

VALIDITY

OF

THE

FORMERLY

OBESE

131

;
4 4 I 4

45.

z z
0 Cl) U)

w
2 4

Difference Total Error (nz6) (n:17)

4.2% 6.6%

IL.

Men

>0 0

#{149} Women p.c.0l

Downloaded from www.ajcn.org by guest on February 14, 2012

l0-

10

20

25

30

35

40 45

50

% FIG 2. Comparison weighing. ofpercent body

BODY

FATNESS

(Hydrostatic)

fatness

determined

by the

Durnin-Rahaman

skinfold

equation

and

hydrostatic

50
>.

w
-J U)

40
0 a 2 2 a

30
U) U)

25a

Difference Total
a
a

7. I 8.4%

I-. 4 IL. 0 0

2O

Error

Men Women

(n:6) (n:l7)

5l00 5
%

p.OOl

20
BODY

25
FATNESS

30

35

40

45

50

(Hydrostatic)

FIG 3. Comparison weighing.

ofpercent

body

fatness

determined

by the Durnin-Womersley

skinfold

equation

and

hydrostatic

132

SCHERF

ET

AL

45 U
0 -J 0

40 35 30 25
20

a.
2 0
U)
U

4
-)

a a

U)
U)

w z
4 IL. 0 0

i Difference
Total
a

0.7 4.4

% %

Error

Men (n:6)
(n:l7)

#{149} Women

5 l0
0

pNS

Downloaded from www.ajcn.org by guest on February 14, 2012

5 %

20
BODY

25
FATNESS

30

3540
(Hydrostatic)

45

50

FIG 4. Comparison weighing.

of percent

body

fatness

determined

by the Jackson-Pollock

skinfold

equation

and

hydrostatic

Discussion Although considered of relative disadvantage 40 35

U)
U,

hydrostatic weighing is generally the criterion method for estimation body fatness, there is considerable to this technique. Hydrostati-

cally-determined percent body fatness is not a popular clinical practice due to the expensive equipment, technical expertise, subject cooperation, additional space, and time required. Furthermore, this criterion method has considerable potential for error associated with the

30

Ui

Key
25 20 A:Hydrostotic

2 ILi >-

B:Pollock

(ps.OI)
( pc.OI) ( p<.00l)

5 0

C:Durnin-Rohomon D: Durnin-Womersley

E=Jockson-Pollock
5 n:23 A
FIG 5. Comparison

(p=NS)

B
of mean percent

C
body

D
fatness

E
determined by hydrostatic weighing and four skinfold equations.

SKINFOLD

VALIDITY

OF

THE

FORMERLY

OBESE

133

the results ofthe Harpenden caliper by assumption that the density of fat-free weight mated approximately 1 to 2 mm/site. While most is 1. 100 g/ml. The use of body density to esoccurs among the new, cheaper vatimate body fat in various populations mayvariance rieties, validity is enhanced when the caliper underestimate or overestimate the actual fat calibrated, and the same that content by 2 to 4% (12). In short, it is impor- is standardized, tant to recognize the inherent limitations of was used to develop the selected regression the hydrostatically-determined body fatness. equation. On the other hand, skinfold thickness mea- Prediction errors of body fatness may also sures provide a convenient and simple method be attributed to the design of the skinfold of body composition assessment. Further- equation used (20, 22). With well over 100 more, prediction of body fatness by the skin- skinfold equations available, all can be clasfold method exhibits a high curvilinear cor-sified as either population-specific or generrelation with the hydrostatic technique (5). alized. Population-specific equations, develSkinfold measurement equations, however, o_ prior to the mid l970s, used samples of tend to be subject to significant error which insufficient size which were homogeneous in requires better standardization and refinement terms of age, sex, and body fatness. A major of techniques for improved validity. For cxlimitation ofthis equation is the small number ample, numerous studies (6-19) have noted of subjects, which reduces validity. Furtherthe reduced accuracy which results whenmore, population-specific equations tend to equations derived from one sample of subjects overestimate body fat of leaner groups and are used to predict the body composition of underestimate a fatter groups. Unless applied to sample other than the one from which the similar a unique population (eg well-defined equations were derived. Prediction errors may athletes), population-specific equations are be due to several factors. Estimates of the likely to be of little value. Generalized equaamount of subcutaneous fat as a percentage tions, on the other hand, are a more recent of total adiposity range from 20 to 70% demethod which is based on large, heterogeneous pending on such biological factors as age, sex,samples varying greatly in age and degree of degree ofadiposity, fat distribution, and mea- body fatness. These newer equations take into surement technique (5). The accuracy of pre- account the potential change in ratio of interdicting body density from skinfold thickness nal to external fat and bone density with age, measures is also subject to large interand inand the nonlinear relationship between skintraexaminer error. Test-retest reliability may fold fat and body density (20). vary from 0.85 to 0.90 for skinfold measures The present study illustrates the potential (20). It appears that good instruction, stan- error in predicting relative body fatness among dardized procedures, and repeated practice can formerly obese adults utilizing several comimprove intrainter-tester reliability. Size of mon skinfold formulas. Skinfold estimates of skinfold grasp, and time delay in reading mea- relative body fatness showed a total error of surements can also cause significant variation. 4.4 to 8.4% when compared to the hydrostatic Interexaminer errors are most problematic criterion method. These data are in agreement with variation due to improper selection of with previous studies which have shown that skinfold sites. In fact, improper skinfold sitewhen skinfold equations are misapplied to selection, technician experience, and subject other populations, the overestimation or unfatness may account for a variation of up to derestimation of percent body fatness may 12 mm on one skinfold site (21). vary from 5 to 10% (23). Furthermore, the validity ofusing skinfolds With the exception of the Jackson-Pollock for estimating body density may be signifi- equation, all skinfold formulas significantly cantly affected by caliper selection and mea- overestimated (p < 0.01) hydrostatically-dcsurement procedure. Caliper design and pres- termined percent body fatness. The accuracy sure exerted will have a significant effect onof this particular equation in the present forboth skinfold measurement and reproducibilmerly obese adults may have been due to simity. According to Lohman (14), there was ilarities in subject age, relative fatness, and variation between Harpenden and Lange cal- level of physical activity, or skinfold sites seipers, wherein Lange consistently overesti- lected. The Jackson-Pollock formula was

Downloaded from www.ajcn.org by guest on February 14, 2012

134

SCHERF

ET

AL

More recently, however, Jackson-Pollock based on a large sample (n 557), 249 women, = mean SD age 31 10.8, and 303 men, mean published equations extending the original of Durnin-Womersley which exhib SD age 33 10.8. By comparison, the mean concept ited accuracy consistent with population-spe SD age ofthe present subjects was 42.5 6.5 for men and 41.6 10. 1 for women. The cific equations. Jackson-Pollock population included a wide Although the Pollock equation was based range of hydrostatically-determined body faton a population that closely resembled the ness for both men and women, 1 to 33% present ( subjects with regard to age, the equa= 17.7%) and 4 to 44% ( = 24. 1%) respection yielded a mean difference and total error tively. The average body fatness of the forthat was slightly higher than that noted for the merly obese adults in the present study, deJackson-Pollock equation, with relative fatness termined by underwater weighing, was 20.4% values significantly greater than those derived for men and 29.8% for women. Many of the from the hydrostatic criterion method. Differsubjects in the Jackson-Pollock sample were ent anatomical sites and the fewer number of engaged in exercise programs, as were the skinfold measurements may have been represent subjects. Finally, it appears that the sponsible for the slightly reduced accuracy of Jackson-Pollock skinfold sites correlate highly this equation. Perhaps the use of circumferwith total adiposity in formerly obese men and ence measurements in addition to skinfold fat women. Middle-aged women are reported to would improve the validity of this prediction have a higher percentage of body fat located equation. The accuracy of estimating body in the chest, trunk, hip, and thigh areas, density in obese populations should be suspect whereas men accumulate the largest depots in when using skinfold measurements, particularly when in excess of 40 to 45 mm per site. their abdomen and trunk (7, 24, 25). The Jackson-Pollock equation evaluated these Furthermore, equations consisting of skinrepresentative areas: chest, abdomen, and folds, circumferences, and diameters had thigh for men, and triceps, suprailium, andhigher correlations with body density than thigh for women. earlier equations using only skinfolds, particGeneralized equations, such as Jackson- ularly in women and middle-aged men (20). Pollock, are valid using large, heterogeneous samples which account for age and fatness difConclusions ferences and the nonlinear relationship between body density and skinfold fat. Another of this study support the conprimary advantage ofthis equation is the con- The results clusions of previous investigators that many sideration of potential change in the ratio of equations are population specific and internal to external fat and bone density withskinfold are valid only when applied to subjects of age (20, 22). age, sex, body composition, and The Durnin-Rahaman equation was based comparable on a British population of 60 men, 18 to 34 demographics. Generalized equations, howare based on large heterogeneous yr ofage ( age = 22 yr), and 45 women, 18 ever, which are valid on populations which vary to 29 yr ( age = 2 1 yr). Although 48 1 British samples, greatly in age and body fatness. In addition, men and women between 16 and 72 yr of age the advantage of the generalized approach is were evaluated by Durnin-Womersley, the can replace several popupopulation was classified as sedentary and only that one equation lation-specific equations without loss of ac36% of the sample were 30 to 49 yr of age. The small sample size ofmature adults in both curacy. The present data further indicate that stanthe Durnin-Rahaman and Durnin-Womersley dard skinfold measurements may overestimate studies, as well as differences in subject debody fatness among formerly obese mographics and caliper selection, may have relative men and women. These findings are clinically accounted for the largest total errors of 6.6 important since such measures may result in and 8.4% respectively. Durnin and Womersley and unsafe recommendations for adwere the first to develop generalized equations spurious weight and fat losses, despite the atwith adjustments for age for predicting body ditional of normal values for these variables. density from skinfold measurements (20).tainment

Downloaded from www.ajcn.org by guest on February 14, 2012

SKINFOLD

VALIDITY

OF

THE

FORMERLY

OBESE

135

predicting body density of men. Br J Nutr 1978;40: Our data also suggest the possibility that rapid 497-504. and pronounced weight reduction may result 1 1. Jackson AS, Pollock ML, Ward J. Generalized equain a disproportionate decrease in internal as tions for predicting body density ofwomen. Med Sci opposed to subcutaneous body fat stores, since Sports l980;l2: 175-82. 12. Lohman TG. Body composition methodology in many of the present subjects demonstrated sports medicine. Phys Sportsmed 1982;l0:47-58. excessive subcutaneous skinfold thickness de13. spite normal relative fatness. #{163}3 Jackson AS, Pollock ML. Predicting accuracy of body 14.

References

1. Behnke AR, Wilmore JH. Evaluation and regulation ofbody build and composition. Englewood Cliffs, NJ: 15. Prentice Hall, 1974. 2. Goldman RF, Buskirk FB. Techniques for measuring 16. body composition. Washington, DC: National AcademyofScience, 1961. 17. 3. DuBois AB, Botelho SY, Bedell GN, Marshall R, Comroe JH. A rapid plethysmographic method for measuring thoracic gas volume: A comparison with a nitrogen washout method for measuring functional 18. residual capacity in normal subjects. J Clin Invest l956;35:322-6. 4. Sin WE. Techniques for measuring body composition. 19. Washington, DC: National Academy ofScience, 1961. 5. Lohman TG. Skinfolds and body density and their relation to body fatness: a review. Hum Biol 198 l;53: 20. 18 1-225. 6. Pollock ML, Hickman T, Kendrick Z, Jackson AS, Linnerud AC, Dawson G. Prediction ofbody density 21. in young and middle-aged men. J Appl Physiol l976;40:300-4. 7. Pollock ML, Laughridge EE, Coleman B, Linnerud 22. AC, Jackson AS. Prediction ofbody density in young and middle-aged women. J AppI Physiol 1975;38: 745-9. 23. 8. Durnin JVGA, Rahaman NM. The assessment of the amount offat in the human body from measurements of skinfold thickness. Br J Nutr l967;21:681-9. 9. Durnin JVGA, Womersley J. Body fat assessed from 24. total body density and its estimation from skinfold thickness: measurements on 481 men and women 25. from 16-72 years. Br J Nutr l974;32:77-97. 10. Jackson AS, Pollock ML. Generalized equations for

density, lean body weight, and total body volume equations. Med Sci Sports 1977;9: 197-201. Lohman TG, Pollock ML, Slaughter MH, Brandon LI, Boileau RA. Methodological factors and the prediction ofbody fat in female athletes. Med Sci Sports Exer 1984; 16:92-6. Sloan AW. Estimation of body fat in young men. J Appl Physiol l967;23:31 1-5. Sloan AW, Burt JJ, Blyth CS. Estimation ofbody fat in young women. J Appl Physiol 1962;l7:967-70. Thorland WG, Johnson GO, Tharp GD, Fagot TG, Hammer RW. Validity of anthropometric equations for the estimation of body density in adolescent athletes. Med Sci Sports Exer 1984;l6:77-8 1. Wilmore JH, Behnke AR. An anthropometiic estimate ofbody density and lean body weight in young men. J Appl Physiol l969;27:25-3 1. Wilmore JH, Behnke AR. An anthropometric estimate ofbody density andlean body weight in young women. Am J Clin Nutr l970;23:267-74. Pollock ML, Jackson AS. Research progress in validation of clinical methods of assessing body composition. Med Sci Sports Exer l984;16:606-13. Pollock ML, Schmidt DH, Jackson AS. Measurement ofcardiorespiratory fitness and body composition in the clinical setting. Compr Ther l980;6:l2-27. Jackson AS. Research design and analysis ofdata procedures for predicting body density. Med Sci Sports Exer l984;l6:6 16-20. Katch F!, Katch VL. Measurement and prediction errors in body composition assessment and the search for the perfect prediction equation. Res Exerc Sport Q l980;5 1:249-61. Edwards DAW. Observations on the distribution of subcutaneous fat. Clin Sci l950;9:259-70. Edwards DAW. Differences in the distribution of subcutaneous fat with sex and maturity. Clin Sci 195 l;10: 305-15.

Downloaded from www.ajcn.org by guest on February 14, 2012