Archives of Disease in Childhood 1995; 73: 305-3 10

Determination of body composition from skinfoldthickness: a validation study

J J Reilly, J Wilson, J V G A Durnin

AbstractMeasurement of body composition isproving increasingly important in clinicalnutrition and research. Skinfold thicknessis a simple means ofestimating body com-position which is widely used in children,but there is little information on itsvalidity. There has been a proliferation ofequations for estimation of body com-position from skinfolds, but some doubt asto their general applicability. The aim ofthe present study was to validate five cur-rently used equations for this purpose in asample of 98 healthy prepubertal children(64 boys, 34 girls), mean (SD) age 9-1 (1-7)years by comparison of estimates fromeach equation with measurements offatness derived from hydrodensitometry.Differences between methods were deter-mined by calculation of biases andlimits of agreement. Limits of agreementbetween predicted and measured fatnesswere wide, particularly in the girls, andsome distinct biases were apparent.Choice of prediction equation thereforehas a substantial influence on the estimateoffatness obtained when using skinfolds inchildren. The existing published equa-tions are associated with large randomerrors or significant systematic errors.For the time being skinfolds might bestbe regarded as indices (rather thanmeasures) of body fatness in individuals,or means of estimating body fatness ofgroups. Estimating the total body fatnessof individual prepubertal children usingskinfolds, on the basis of this evidence, isnot advisable at present.(Arch Dis Child 1995; 73: 305-310)

Keywords: body fatness, fat free mass, validation,nutritional status.

University of GlasgowDepartment ofHumanNutrition, YorkhillHospitals, GlasgowG3 8SJJ J ReillyJ WilsonJV G A Durnin

Correspondence to:Dr Reilly.Accepted 2 May 1995

A recent editorial in this journal noted the'inability of classical measures of growth anddevelopment to meet the requirements ofmodern clinical medicine and research', andsuggested that measurement of body composi-tion could meet these requirements.1 Therehas been a resurgence of interest in bodycomposition in childhood as a result ofincreased awareness of its importance' 2 and theproliferation of new measurement techniques.

Measuring the body composition of childrenis problematic for practical and theoretical rea-

sons. Many standard methods of assessment(for example, hydrodensitometry) can beimpractical, particularly in young or disabledchildren. Theoretical problems arise largely

from the 'chemical immaturity' of children.Assumptions of constant composition of fatfree mass (FFM), made routinely in adults, areinvalid in children2 3 and approaches that takeaccount of the changing composition of FFMwith advancing age2 4 are therefore necessary.

There is a need for simple 'bedside' methodsof body composition measurement which havebeen validated and are generally applicable inchildren.5 In adults skinfold thickness has beenshown to be as valid as any other method forthe measurement of absolute fat mass andFFM,6 and changes in fat mass and FFM,7even in populations where the composition ofFFM is particularly variable.8Measurement of skinfold thickness is cur-

rently widely used in children for clinical,research and epidemiological purposes, butthere are doubts about its validity in infancy,8 9no empirical demonstrations of its validity inprepubertal children, and concerns over agerelated variability in the composition of FFM,2and variability in skinfold compressibility,10both of which have a bearing on the validity ofthe technique. Furthermore, many equationsexist for the estimation of body fatness fromskinfold thickness in children. There is evi-dence that some skinfold equations are highlypopulation specific11 12 but also reports thatsome equations are considerably less so13 andso are generally applicable. In summary, theuser of skinfold thickness in childhood is there-fore faced with a choice of equations for esti-mating fatness, but these may not be applicableto the population under study. More funda-mentally, the technique itself is of unknownvalidity in the prepubertal child. The aims ofthe present study were to cross validate themore commonly used equations for predictionof body fatness from skinfold thickness inhealthy prepubertal children, and to describebiases and errors in estimation by comparisonwith fatness measured using hydrodensito-metry, the reference method. The precision ofthe skinfold technique was also assessed.

Subjects and methodsSUBJECTSThe study sample consisted of 98 healthy, selfselected children; there were 64 boys (mean(SD) age 9-3 (1-7) years) and 34 girls (9-0(1-7) years). Subjects were recruited from localschools and informed consent was obtainedfrom parents. Ethical approval was obtainedfrom the Yorkhill Hospitals ethics committee.Physical characteristics of subjects aredescribed in table 1. Body fatness (% of bodyweight, BF%) ranged from 4 0 to 39 1% inboys and from 4-1 to 32.8% in girls with the

305

on May 10, 2021 by guest. P

rotected by copyright.http://adc.bm

j.com/

Arch D

is Child: first published as 10.1136/adc.73.4.305 on 1 O

ctober 1995. Dow

nloaded from

Reilly, Wilson, Durnin

Table 1 Characteristics ofsubjects

Boys Girls

No Mean (SD) Range No Mean (SD) Range

Age (years) 64 9-3 (1-7) 5-8-11-7 34 9-0 (1-7) 6-3-11-3Weight (kg) 64 32-0 (8-6) 203-59-2 34 31-0 (7-6) 21-9-50-1Height (m) 64 1-36 (0-10) 1-18-1-57 34 1-33 (0-11) 1-15-1-59Density (kg/l)* 57 1-055 (0-016) 1-003-1-077 24 1-037 (0-016) 1-012-1-068BFO/%* 57 13-2 (8-0) 4-0-39-1 24 19-8 (7-9) 4-1-32-8

*Density, and density derived estimates of fatness, available for 57/64 boys and 24/34 girls.

reference method, indicating heterogeneity ofthe sample with respect to fatness.

ANTHROPOMETRYBody weight was measured to 0 1 kg using a

standard beam balance. Height was measuredto 0 5 cm using a Holtain wall mountedstadiometer. In all subjects four skinfold thick-nesses (biceps, triceps, subscapular, suprailiac)were measured (as previously described'4) intriplicate, by the same trained observer.Measurements were made on the right handside of the body using a Holtain caliper. In 17subjects all anthropometric procedures were

repeated within seven days of the initial visit.The test-retest data were then used to calculatethe precision of skinfold thickness measure-

ment.

MEASUREMENT OF BODY DENSITYIn 57 boys and 24 girls body density was

measured in triplicate, using hydrodensito-metry. Residual volume determination was

made by nitrogen wash out, as previouslydescribed.'4 Seven of the 64 boys and 10 of the34 girls were unable or unwilling to undertakethe underwater weighing procedure.

CONVERSION OF MEASURED OR PREDICTEDDENSITY TO BODY FATNESSIn all children measured or predicted densitywas converted to an estimate of BF% using a

modification ofthe SiriI5 equation proposed byWestrate and Deurenberg:

BF%=[562-4-2 (age-2)]/d-[525-4-7 (age-2)]

where age is in years, d=density (kg/lI). Use ofthe Siri'5 equation in children, with itsinherent assumption of FFM density of1-1 kg/I, leads to systematic overestimation ofbody fatness from density. Two alternativemodels have been constructed,2 4 based onempirical data, which are unbiased and more

physiological, assuming increasing FFMdensity during childhood. These models pre-dict similar fatness at the same density and theequation of Westrate and Deurenberg4 was

chosen for convenience.

PREDICATION OF BODY DENSITY OR FATNESSFROM PUBLISHED EQUATIONS: VALIDATIONSTUDYFive empirically derived equations for the pre-diction of density or BF% from skinfolds were

tested.

(1) Equations ofDurnin and Rahaman'6Predicted density (kg/i):

Boys= 1 1533-0 0643X (log sum of 4 skinfolds)Girls= 1-1369-0-0598x (log sum of 4 skinfolds)

These equations were derived from empiricalrelationships between skinfold thickness andbody density in adolescents (48 boys age range12-7-15-7 years; 38 girls age range 13-2-16-4years). Their validity in younger children wasunknown, and as the methodology used wasidentical to that in the present study, a test ofcross validation was justified.

(2) Equations ofSlaughter et al'7BF% for children with triceps and subscapularskinfolds <35 mm:

Boys= 1-21 (sum of 2 skinfolds)-0-008 (sum of 2 skinfolds2)-1-7Girls= 1-33 (sum of 2 skinfolds)-0-013 (sum of 2 skinfolds2)-2.5

BF% for children with triceps and subscapularskinfolds >35 mm:

Boys=0-783 (sum of 2 skinfolds)- 1-7Girls=0 546 (sum of 2 skinfolds)+9 7

These equations are based on an empiricallyderived multicomponent method utilisingmeasurement of body density, total bodywater, and bone mineral content of radiusand ulna. The sample used to derive theseparticular equations consisted of 50 boys(mean age 9-8 years) and 16 girls (mean age10 0 years) from the USA. The cross validityof the Slaughter et al equations has beenreported to be high,'3 and these are effec-tively the 'standard' equations used in NorthAmerica.

(3) Equations ofJ7ohnston et al'8Predicted density (kg/l):

Boys= 1-1660-0-0070X(log sum of 4 skinfolds)Girls= 1-144-0 060X(log sum of 4 skinfolds)

These equations are based on empiricallyderived relations between skinfolds and densityof Canadian children aged 8-14 years (140boys, 168 girls). The cross validity of theseequations has not been reported.

(4) Equations ofBrook19Predicted density (kg/l):

Boys= 1-1690-00788x (log sum of 4 skinfolds)Girls= 1-2063-0-0999X(1og sum of 4 skinfolds)

These equations were derived from empiricalrelations between total body water and bodydensity (predicted from equations for adoles-cents'6), in a sample\of 13 obese children and10 with short stature, age range 1-11 years.The equations are widely used in the UK.

(5) Equations ofDeurenberg et a120Predicted density (kg/l) (prepubertal):

Boys= 11133-0-0561 X(log sum of4 skinfolds)+ 17 (ageX10-3)Girls= 1-1187-0 063x (log sum of 4 skinfolds)+ 1 9 (ageX 10- )

These equations are based on empiricallyderived. relationships between skinfolds anddensity in Dutch children (114 boys mean age1 1 0 years; 98 girls mean age 10 5 years). Theequations are now widely used, but their crossvalidity has not been reported.

306

on May 10, 2021 by guest. P

rotected by copyright.http://adc.bm

j.com/

Arch D

is Child: first published as 10.1136/adc.73.4.305 on 1 O

ctober 1995. Dow

nloaded from

Determination of body composition from skinfold thickness: a validation study

Table 2 Group mean (SD) estimates ofBF% byfiveequations and densitometry in girls and boys

Boys (n=57) Girls (n=24)

Densitometry 13-2 (8 0) 19-8 (7-9)Equations

(1) Dumin and Rahaman 10-4 (6-1) 16-6 (4-5)(2) Slaughter et al 15-5 (5-5) 19-9 (4 9)(3) Johnston et al 8-2 (6 5) 13-1 (4 4)(4) Brook 13-1 (7-1) 13-3 (7-3)(5) Deurenberg et al 16-2 (4 9) 19-3 (4 9)

STATISTICAL ANALYSISAgreement in estimates of BF% between thevarious methods, and agreement between thevarious methods and hydrodensitometry(selected as the reference method), was deter-mined by calculation of biases, and limits ofagreement (bias ± 2SD) following themethod of Bland and Altman.21 This focuseson the individual differences betweenmethods and summarises agreement on thebasis of how wide or narrow are the limits ofthe differences between methods.2' Groupmean estimates by the various methods arealso presented. Sample size for comparisonbetween the five equations and densitometrywas 57 boys and 24 girls; for comparisonbetween the five equations sample size was 64boys and 34 girls.

ResultsMEASUREMENT RELIABILITYFor the 17 children included in the assessmentof reliability, the mean (SD) of the sum of twoskinfolds (triceps plus subscapular) was 1 3-8(3 2) mm, and mean of the sum of four skin-folds was 25-0 (6.2) mm.The mean (SD) difference between

measurements 1 and 2 for two skinfolds was05 (08) mm (range -0-8 to 1-8 mm), with

95% confidence interval (CI) for the difference-0 2 to 0-8 mm. For four skinfolds mean (SD)difference between measurements was 0-8(1-4) mm (range -2-6 to 2-7 mm), with 95%CI for the difference -0 1 to 1-2 mm. The dis-tribution of differences was normal.The value mean ± (2SD) for differences

were used to propagate errors in estimatedBF% for two and four skinfolds. For two skin-folds this represented maximum differences of1 8 BF%. For four skinfolds the maximum dif-ference in BF% estimates was 1-5 BF%. Thesedifferences were similar to those reported foradults.7

DIFFERENCES BETWEEN EQUATIONS TESTEDAND DENSITOMETRY: BOYSSome distinct biases were apparent, andlimits of agreement (mean ± 2SD of differ-ences) between all five equations and the ref-erence method were wide, indicating pooragreement between predictions.2' Body fat-ness was overestimated by the equations ofSlaughter et al and Deurenberg et al. Theequation of Brook had the smallest bias andnarrowest limits of agreements relative todensitometry (table 2). The equations ofDurnin and Rahaman and Johnston et alunderestimated fatness relative to densitome-try, and the latter equation predicted anumber of very low fatness estimates (somenegative estimates). Limits of agreementbetween the various equations and each otherwere generally narrower (that is agreementbetter) than those between the variousmethods and densitometry. As an example ofthe range of estimates which were obtainedfrom the five prediction equations in the samechild, in a typical boy predicted fatnessranged from 5 3% (Johnston equation) to

(1)20X15 +2SD10 * X a)

0.5- ', . *Bias ag CO- .. ~~* . .'

-5 . .** -2SDLi-10-15 r= 0-43 (p<0-01) X-20 1 2 25

10 15 20 25 30 35

BF%

(2)20-15-10-5-0-

-5-

-10--15-

5

(3)?n",

* +2SD

. ...- Bias

-2SD

r= 0.57 (p<0.01)10 15 20 25 30 35

BF%

+ 155)

+- 10o

5

C.C 0-

0

ui.. -10C -15-)Bo -20-

+2SD

.. _ __ ...........---Bias.. .. + .

-2SD

r= 0.32 (p<0.05)10 15 20 25 30 35

BF%

__+2SD

:*...Bias.- .* . .

.

*,- . * *+* _-2SD

r= 0.23 (NS)10 15 20 25 30 35

BF%

4)m

(U

Q)0)

4- L

CU5,

0.0

a)'-5)w.

L-o 0

20151050

-5-10-15-20

.+2SD

(5)

)-. - . *. ....... - Biasi .;*r -. _ '. _ _

)- -tS5.t

-2SDr= 0.62 (p<0.01)

5 10 15 20 25 30 35

BF%Figure 1 Biases (reference method minus prediction) in BF% estimates in relation to body fatness in 57 boys; (1)-(5) refer to equations in subjects andmethods.

'

C

E

o

CU

a)

V-~--

(4)20]15-

xsO104~- 0,o5

._

o -5. c

iZ .n -10-15-20

s9v-zv I- II. .

307

-1

on May 10, 2021 by guest. P

rotected by copyright.http://adc.bm

j.com/

Arch D

is Child: first published as 10.1136/adc.73.4.305 on 1 O

ctober 1995. Dow

nloaded from

Reilly, Wilson, Durnin

~~~-~~+2S--~-~~~--~-~ +2SD

* B-~.~ "~~ ~~~~ Bias

r = 0.61 (p<0.01)10 15 20 25 30 35

BF%

Co

Co)O =

a, )o

cn

nv

(2)20115---- --- - +2SD10 -

5.0o Bias

-5 . .

-101-~~~~~ ~~~~~~-S-15

-20Ir= 0.55 (p<0.01)

5 10 15 20 25 30 35

CZ

~4-o

C .c4-0

C=

a)-

BF%

(3)2015 A 6

2S

10-5--- -----. ---- .-------- Bias

0o- ..

10-lb r= 0.64 (p<0-01)-20 -

5 10 15 20 25 30 35

BF%

(4)20 -r-- - -------15-10-

5-0-

+2SD <

C .

Bias nCo°

'- ED_-0

-10 - --- -2SD

-15- 0.21 (NS)

-20 1o 15 20 25 30 35

-L]

._

enCo>Vt

BF%

(5)20 l15 _ _ ~ ~+2SD10 + .# +

5-

0-

-5

-10

-1 5 -r = 0-60 (p<0.01 )-20 10 15 20 25 30 35

BF%Figure 2 Biases (reference method minus prediction) in BF% estimates in relation to body fatness in 24 girls; (1)-(5) refer to equations in subjects andmethods.

13-8% (Deurenberg equation) with fatnessfrom densitometry 11 9%.The relationships with age and body fatness

were tested using correlation of the biasobserved with fatness derived from the mean oftwo methods.2' In general, there was littleevidence of influence of age on the biasesobserved, but fatness had a significant influ-ence (fig 1) for all the equations tested otherthan that of Brook, where there was a trendwhich was not statistically significant.

DIFFERENCES BETWEEN EQUATIONS TESTEDAND DENSITOMETRY: GIRLSSome distinct biases were apparent and limitsof agreement between the five equations anddensitometry were wider than those in theboys. Estimates ofBF% between the equationswere in closer agreement than those betweenthe equations and densitometry. Highest esti-mates ofBF% were produced by the equationsof Slaughter et al, followed by (in descendingorder) those of Deurenberg et al, Dumin andRahaman, then Brook and Johnston et al (table2). As in the boys, estimates from each equa-tion in individuals produced widely differentvalues for BF%. In one typical girl, forexample, BF% estimates ranged from 10%(Brook equation) to 19-8% (Deurenberg equa-tion) with true body fatness (densitometry)17 4%.

In the girls there was no evidence that biasesin estimation of BF% relative to densitometrywere age-related except for that of Deurenberget al (bias and age positively correlated,r=0-42, p<0 05). There was evidence of aninfluence of fatness on the magnitude anddirection of bias (fig 2) for all equations testedwith the exception of that ofBrook where, as in

the boys, there was a trend which was notstatistically significant.

DiscussionThe present study confirms that, for theindividual child, there is poor agreementbetween reference estimates of body fatnessand predictions based on measurements ofskinfolds. While some of the differencebetween skinfold methods and densitometryis attributable to variability in the composi-tion of the FFM, and hence to 'error' in thereference method,2 15 the study neverthelesssuggests that considerable caution should beexercised in predicting body fatness fromskinfolds in children, and that the validity ofpublished predictive equations cannot betaken for granted. Poor cross validity has beendescribed for other anthropometric equationsin children,"I 12 but was not previously estab-lished for the equations tested in the presentstudy. Furthermore, more casual users ofanthropometric methods are largely unawareof this problem and the methods are widelyused with no reference to the validity of themethod in particular population groups. Thestudy demonstrates that lack of validity inhealthy children is a serious problem andthis is likely to be compounded when themethods are applied to patient populations.Alterations in fat distribution22 or systematicdifferences in the composition ofFFM associ-ated with disease,23 are likely to exacerbatelack of cross validity and lead to poorer pre-diction.'8 Users of skinfolds in childhoodmust consider whether the equations availableare valid in their particular group. For manyclinical settings equations will be invalid, andthey also appear to be invalid in healthy

1)- (C 20-Eco

15-

x.- 5-co._o 0 -

"C

o E- -5-iLJ a -io0

a, -20-V

oL~o_O .

C I

, coi-CLZi.1)

308

* +

-51-----, -2SD

I

., , .1 B ias

on May 10, 2021 by guest. P

rotected by copyright.http://adc.bm

j.com/

Arch D

is Child: first published as 10.1136/adc.73.4.305 on 1 O

ctober 1995. Dow

nloaded from

Determination of body composition from skinfold thickness: a validation study 309

prepubertal children on the basis of thepresent evidence.The reasons for the poor validity (and lack of

general applicability) of the equations testedmust lie in methodological or biologicaldifferences between studies. The most obviousbiological difference between the samplerecruited and those recruited in the otherstudies is age: the equations of Durnin andRahaman'6 were based on adolescents and sothe evidence of systematic errors in prediction(table 2) are perhaps not surprising. Thesamples used to derive the Deurenberg et al,20Slaughter et al,17 and Johnston et al 18 predic-tive equations included some children asyoung as 7 years, but the age distribution of thepresent sample was generally younger than inthat of these published studies. Since age ormaturational state has been identified as animportant predictive variable17 19 24 this mayaccount for some of the differences observed.

Differences in body fatness between thepresent sample and those used to derive thevarious equations may also have contributed tothe lack of validity, particularly in view of theevidence that bias was related to the bodyfatness of the child (figs 1 and 2). This findinghas relevance for those wishing to choose equa-tions for particular groups: users of skinfoldsshould be aware that magnitude and directionor error for some prediction equations willdepend on the body fatness of the childrenbeing studied. It may be significant that theprepubertal boys recruited by Deurenberget a120 were fatter than those in the presentstudy, and the boys (mean 23-5% fat) and girls(mean 27-5% fat) recruited by Slaughter et al 17were considerably fatter than those recruited tothe present study.

There were methodological differencesbetween the present study and those used toderive the predictive equations tested. In thepresent study, and most others, skinfolds weremeasured on the right hand side, but this wasnot the case for the equations of Deurenberget a120 (left hand side). In the present studyresidual lung volume could not be measuredsimultaneously with density measurements insitu because most children could not complywith the rebreathing required on surfacingafter immersion. This procedure was commonto some17 18 but not all'6 20 other studies. Thedifferent approaches may have introducedsystematic differences between methods.

While considerations of body compositionof individual children are often of primeimportance this is not always the case.Attention has recently been drawn to the lackofvalidated methods for quantification ofbodyfatness in children in epidemiology, in order toquantify prevalence of obesity for example.25In such circumstances, absence of systematicdifferences (biases) may be the prime concernand the evidence of poor agreement forindividuals may be of secondary importance.The present study suggests that, for groups, theequations of Deurenberg et al and Slaughteret al predicted fatness with negligible bias inthe girls and the equation of Brook predictedfatness with negligible bias in the boys (table

2). Measurement of single or multiple skin-folds may also be of value as relative indices offatness, and may be more appropriately used asindices rather than measures of body fatness ofindividual children.

In conclusion, the results indicate thatprediction of body fatness by measurement ofskinfold thickness in prepubertal children isassociated with large errors at the individuallevel. The prediction equation chosen can havea profound effect on the estimate obtained, andcare is required in deciding which equation touse. The simplicity of skinfold thickness isattractive and, as the method can be reasonablyprecise, improvements could be made in themethodology. These are likely to requirederivation of predictive relationships on largeheterogenous samples of children with externalcross validation. Until this has been done,choice ofprediction equation will continue to bethe major influence on skinfold estimates offatness in childhood, and the method wouldseem unsuitable as a means of accurately esti-mating body fatness of individual prepubertalchildren.We thank the volunteers and their families for their cooperation.The Institute of Physiology University of Glasgow providedaccess to laboratory facilities, Lawrence Weaver commented onan earlier draft of the manuscript. The Department of HumanNutrition is supported by the Rank Prize Funds and RankFoundation. The study was funded by the Scottish OfficeHome and Health Department.

1 Davies PSW. Body composition assessment. Arch Dis Child1994; 69: 337-8.

2 Lohman TG. Advances in body composition assessment.Monograph No 3. Champaign, Illinois: Human KineticsPublishers, 1993.

3 Hewitt MJ, Going SB, Williams DP, Lohman TG.Hydration of the fat-free body in children and adults:implications for body composition assessment. Am JPhysiol 1993; 265: E88-95.

4 Westrate JA, Deurenberg P. Body composition in children.Am J Clin Nutr 1989; 50: 1104-15.

5 Elia M. Body composition analysis: an evaluation of twocomponent models, multicomponent models, and bed-side techniques. Clinical Nutrition 1992; 11: 114-28.

6 Fuller NJ, Jebb SA, Laskey MA, Coward WA, Elia M. Afour component model for the assessment of body com-position in humans. Clin Sci 1992; 82: 687-93.

7 Jebb SA, Murgatroyd PR, Coward WA, Goldberg GR,Prentice AM. In vivo measurement of changes in bodycomposition. Am J Clin Nutr 1993; 58: 455-62.

8 Reilly JJ, Murray LA, Wilson J, Durnin JVGA. Measuringthe body composition of elderly subjects: a comparison ofmethods. BrJfNutr 1994: 72: 33-44.

9 Davies PSW, Lucas A. Prediction of body fatness in earlyinfancy. Early Hum Dev 1989; 21: 193-8.

10 Davies PSW, Preece MA. Body composition in children:methods of assessment. In: Tanner JM, Preece MA, eds.The physiology of growth. Cambridge: CambridgeUniversity Press, 1989: 95-107.

11 Frerichs RR, Horsha DW, Berenson GS. Equations forestimating percentage body fat in children 10-14 yearsold. Pediatr Res 1979; 13: 170-4.

12 Boileau RA, Wilmore JH, Loman TG, Slaughter MH,Riner WF. Estimation of body density from skinfoldthicknesses, circumferences, and skeletal widths in boysaged 8-11 years: comparison of two samples. Hum Biol1981; 53: 575-92.

13 Janz KF, Nielsen DH, Cassady SL, Cook JS, Wu YT,Hansen JR. Cross-validation of the Slaughter skinfoldequations for children and adolescents. Med Sci SportsExerc 1993; 25: 1070-6.

14 Durnin JVGA, Womersley J. Body fat assessed from totalbody density and its estimation from skinfold thicknessmeasurements on 481 men and women aged 16-72 years.BrJ7Nutr 1974; 32: 77-97.

15 Siri I. Body composition from fluid spaces and density:analysis of methods. In: Brozek J, Henschel A, eds.Techniques for measuring body composition. Washington,DC: National Academy of Sciences, 1961: 223-34.

16 Durnin JVGA, Rahaman MM. The assessment of theamount of fat in the human body from measurements ofskinfold thickness. Br3'Nutr 1967; 21: 681-8.

17 Slaughter MH, Lohman TG, Boileau RA, et al. Skinfoldequations for estimation of body fatness in children andyouth. Hum Biol 1988; 60: 709-23.

18 Johnston JL, Leong MS, Checkland EG, Zuberbahler PC,Conger PR, Quinney HA. Body fat assessed from bodydensity and estimated from skinfold thickness in normal

on May 10, 2021 by guest. P

rotected by copyright.http://adc.bm

j.com/

Arch D

is Child: first published as 10.1136/adc.73.4.305 on 1 O

ctober 1995. Dow

nloaded from

310 Reilly, Wilson, Durnin

children and children with cystic fibrosis. Am Jf Clin Nutr1988; 48: 1362-6.

19 Brook CGD. Determination of body composition ofchildren from skinfold measurements. Arch Dis Child1971; 46: 182-4.

20 Deurenberg P, Pieters JJL, Hautvast JGAJ. The assessmentof the body fat percentage by skinfold thickness measure-ments in childhood and young adolescence. Br J Nutr1990; 63: 293-303.

21 Bland JM, Altman DG. Statistical methods for assessingagreement between two methods of clinical measurement.Lancet 1986; i: 307-10.

22 Spender QW, Cronk CE, Stallings VA. Fat distribution inchildren with cerebral palsy. Ann Hum Biol 1988; 15:191-6.

23 Reilly JJ, Workman P. Is body composition an importantvariable in the pharmacokinetics of anticancer drugs?Cancer Chemother Pharmacol 1994; 34: 3-13.

24 Parzikova J. Total body fat and skinfold thickness in child-ren. Metabolism 1961; 10: 794-807.

25 Hammond J, Rona RJ, Chinn S. Estimation in communitysurveys of total body fat of children using bioelectricalimpedance or skinfold thickness measurements. EurJr ClinNutr 1994; 48: 164-71.

The malaria tragedy: a pen torch in the darkness

Eight jumbo jets crashed today. There were no survivors. Ninety percent of the victims were children. The same happened yesterday, andthe day before, and the day before that: in fact it has happened andwill happen every day this year. That is the news. But it isn'tjumbo jets crashing, if it were it would be the single fact foremostin the thoughts of almost every person in the developed world.The killer is malaria: 3000 deaths a day, nine out of every 10 achild. Malaria eradication has not happened, control is the aim,and clinical management the practical reality. A report fromKenya (Kevin Marsh and colleagues, New England Journal ofMedicine 1995; 332: 1399-404) shows how children withfalciparum malaria can be selected for more concentratedcare.

Over a period of 2-5 years 7538 children were admitted toKilifi District Hospital, 1866 with malaria. Eighteen of thechildren with malaria were moribund on arrival and died quicklyand another four died later of causes other than malaria. Theclinical features and outcome of the remaining 1844 childrenwere analysed. Sixty four of the 1844 children (3 5%) died. Theimportant new finding is that respiratory distress is a mainindicator of severe malaria and can be used together withimpaired consciousness to choose children for more intense care.World Health Organisation (WHO) criteria for severe andcomplicated malaria consist of 10 major clinical and laboratorycriteria, including pulmonary oedema, and five minor,supporting criteria. The work in Kenya showed that the WHOcriteria identified 80% of the children who died but a simpleclinical assessment based on two criteria, prostration (inability tosit unaided) and respiratory distress (alar flaring, chest recession,use of accessory muscles of respiration, or deep breathing)identified over 90%. Six hundred and ninety eight childrensatisfied the WHO criteria and the mortality in them was 7-3%compared with 8-9% for the 650 with prostration and/orrespiratory distress.The major cause of respiratory distress in these patients is

thought to be metabolic acidosis and the main treatments bloodtransfusion and rehydration although severe anaemia was not anecessary concomitant of severe respiratory distress. Thepathophysiology of respiratory distress in these children remainsto be fully elucidated.

ARCHIVIST

on May 10, 2021 by guest. P

rotected by copyright.http://adc.bm

j.com/

Arch D

is Child: first published as 10.1136/adc.73.4.305 on 1 O

ctober 1995. Dow

nloaded from