3rd Edition
July 2024
Abstract
Anthropometric measurements are used to assess body sizeand body composition. The measurements are simple, safe,and non-invasive and provide information on past exposure,but cannot detect short-term disturbances or deficiencyof a specific nutrient. There are three major sourcesof error in anthropometry: (i) measurement errors,(ii) alterations in the composition and physicalproperties of certain tissues, and, (iii) use of invalidassumptions in the derivation of body compositionfrom the measurements.
Anthropometric indices arederived from two or more raw measurements and areessential to interpret and group the anthropometric data.Selection of indices must take into account their sensitivity,specificity, predictive value, and any potentialmodifying factors.Examples of indices include weight-for-height,body mass index (weight kg) / (height m)2,and waist-hip circumference ratio.Anthropometric indices are often evaluatedby comparison withpredetermined reference limits or cutoff points.Calculation of the number and proportion of individuals (as %) withanthropometric indices below or above adesignated reference limit or cutoff, generates “anthropometricindicators” that can be used in clinical and public health settingsto classify individuals at risk of malnutrition.Examples of indicators used in this wayinclude mid‑upper‑armcircumference (MUAC) with a cutoff <115mm toidentify severe acute malnutrition (SAM) in children 6–60mos,and WHZ <−2,BMIZ >+2, and HAZ <−2,used by WHO and UNICEF to definewasting, overweight, and stunting respectively inchildren <5y and to defineprevalence thresholds and identify priority countries.Cutoffs, unlike statistically derived reference limits,are based on functional impairment or clinicalsigns of malnutrition, and occasionally mortality.
The reference growth data recommended by WHO for international use are theprescriptive WHO Child Growth Standards for 0–5y, and the WHOgrowth reference data for older 5–19y.Updated childhood growth charts are also availablefor U.S.infants age 0–36mos and children 2–20y.Local reference data are preferred for body composition,although few are available. Instead, whor*commends using reference data for MUAC,triceps, and subscapular skinfolds collectedfor the WHO Child Growth Standards.
CITE AS: Gibson, R.S., Principles of Nutritional Assessment.Introduction to Anthropometry https://nutritionalassessment.org/intant/
Email: [email protected]
Licensed under CC-BY-4.0
( PDF ).
The term “nutritional anthropometry” firstappeared in “Body Measurements and Human Nutrition”(Brožek, 1956)and was later defined by Jelliffe(1966)as:
“measurements of the variationsof the physical dimensions and the gross composition of the human bodyat different age levels and degrees of nutrition”
Subsequently, a number of publications made recommendations on specific bodymeasurements for characterizing nutritional status, standardizedmeasurement techniques, and suitable reference data(Jelliffe, 1966;WHO, 1968;Weiner and Lourie, 1969).Today, anthropometric measurementsare widely used for the assessment of nutritional status and health, at both theindividual and population levels. One of their main advantages is thatanthropometric measurements may be related to past exposures, to presentprocesses, or to future events(WHO, 1995).
For individuals in low-income countries,anthropometry is particularly useful when there isa chronic imbalance between intakes of energy,protein, and certain micronutrients.Such disturbances modify the patterns of physical growth and the relativeproportions of body tissues such as fat, muscle, and total body water.For individuals in clinicalsettings, anthropometry can be used to diagnose failureto thrive in infants and young children,and monitor overweight and obesity in children and adults.
At the population level, anthropometry has an important role in targetinginterventions through screening, in assessing the response tointerventions, in identifying the determinants and consequences ofmalnutrition, and in conducting nutritional surveillance.Increasingly, anthropometry is also being used to characterizeand compare the health and nutritionalstatus of populations across countries(WHO/UNICEF, 2019).
9.1 Measurements, indices, and indicators
Anthropometric measurements areof two types. One group of measurementsassesses body size, the other group appraises bodycomposition. The most widely used measurements of body sizeare stature (length or height), weight, and headcircumference; see Chapter10 for more details. The anthropometricmeasurements of body composition are based onthe classical “two component model” in whichthe body is divided into two major compartments,fat mass and the fat free mass.Skinfold thickness measurements are used toestimate of the size of the subcutaneous fat depot, which,in turn provides an estimate of total body fat: overone third of total body fat is estimated to besubcutaneous fat. The distribution of body fat isalso important, with the measurement of waistcircumference used increasingly as aproxy for the amount of intra-abdominal visceral fat. Waist circumferenceis recommended for use in population studies(WHO, 2011),as well as in clinical practice for the evaluation andmanagement of patients with overweight or obesity(Ross etal., 2020).
The fat-free mass consists of the skeletal muscle,non-skeletal muscle, soft lean tissue, and the skeleton.A major component of the fat-free mass is body muscle.As this is composed of protein, assessment of muscle masscan provide an indirect assessment of the protein reservesof the body. Measurements of thigh circumference andmid-upper-arm circumference (MUAC) can be used toassess skeletal muscle mass(Müller etal., 2016).Measurement of MUAC is especially useful for youngchildren <5y in emergency settings such as faminesand refugee crises. In such settings, children often have a smallamount of subcutaneous fat, so changes in MUACtend to parallel changes in muscle mass; see Chapter11 for more details.
Anthropometric indices are usually calculated from two ormore raw measurements, and are essential forthe interpretation and grouping of measurementscollected in nutritional assessment. For example,the measurement of a child's body weight ismeaningless unless it is related to the age orheight of a child. In young children the threemost commonly used growth indices are weight-for-age,height-for-age, and weight-for-height. The first twoindices reflect body weight or heightrelative to chronological age, whereas weight-for-heightassesses body weight relative to height.
Body mass index (BMI) is also widely used in children and adultsto assess underweight, overweight, and obesity, and is calculated as(weight kg) / (height m)2. When height cannot be measured,as may occur in bed-bound or frail individuals,published equations based on a range of body measurementssuch as knee height, lower leg length,arm span, and ulna length can be used to providean approximate estimate of height. Examples of equationsfor estimating height from these body measurements in adultsare given in Madden etal.(2016).However, their usefulnessfor hospitalised patients may be questionable, if the equationshave been derived from young and healthy populations(Reidlinger etal., 2014).
Examples of body composition indices include acombination of triceps skinfold and mid-upper-armcircumference, which together can be used to estimatemid-upper-arm fat area and mid-upper-arm musclecircumference or area, surrogates for total body fatcontent, and muscle mass, respectively.Other measurement combinations include the waist-hip ratio(i.e., the waist circumference divided by the hip circumference),an additional index of the distribution of body fat whichcan be measured more precisely than skinfolds.Moreover, measurementsof waist-hip ratio as a surrogate for abdominal obesity,appear to be a stronger independent risk factor forrisk of myocardial infarction, stroke and prematuredeath than BMI, especially among men(Larsson etal., 1984; Lapidus etal., 1984).
In an effort to obtain more reliable estimates of percentagebody fat and fat-fat-free mass based on anthropometricmeasurements in healthy adults,the sum of skinfold thickness measurementsfrom multiple anatomical sites is also used inconjunction with population-specific or generalizedregression equations to predict body density, andin turn, the percentage of body fat using one ofthree empirical equations. Once the percentage of bodyfat is calculated, total body fat content and thefat-free mass can be derived (see Chapter11for more details). Again, many of theprediction equations were developed on young, healthy, lean Caucasianpopulation groups and, hence, are less appropriate for malnourished,obese, or elderly subjects or for other racial groups.
Anthropometric indices are often evaluated bycomparison with the distribution of appropriate anthropometricreference data using standarddeviation scores (Z‑scores) or percentiles.(seeSection9.4.3).From this, the number andproportion of individuals (as%) with anthropometricindices below or above a predeterminedreference limit or cutoff are often calculated.A commonly used reference limit for the three main growthindices is a Z‑score of −2 (i.e., below the whor*ference median) (Section9.4.2).When used in this way, the index and its associatedreference limit or cutoff become an “indicator”; theseare discussed below.
Anthropometric indicators are constructed fromanthropometric indices, with the term “indicator” relatingto their use in nutritional assessment, often for public health,or socio-medical decision-making at the population level. Indicators are alsoused in clinical settings to identify individuals at risk of malnutrition.To be valid, a substantial proportion of the variabilityof an anthropometric indicator should be associatedwith differences in nutritional status. WHO(1995)provide a detailed classification of recommendedanthropometric indicators based on their uses forboth targeting and assessing response to interventions,identifying determinants of malnutrition,or predicting malnutritionin populations of infants and children.
Anthropometric indicators should be chosen carefully in relation toboth their proposed use and their attributes.Indicators vary in their validity, sensitivity, specificity, and predictivevalue; these characteristics are discussed briefly inSection9.4.3.For example, although the indicatorweight-for-age <−2Z‑score isstill widely used in health centers in many low-income countriesfor screening young children at risk of malnutrition, it is inappropriate.Children who are stunted but of normal weight,or alternatively, tall and thin may beincorrectly diagnosed as “healthy”.Instead, in these countries, the indicatorlength/height-for-age <−2Z‑scoreshould be used(Ruel etal., 1995).
Anthropometric indicator | Application |
---|---|
Proportion of children (of defined age and sex)with WHZ <−2 | Prevalence of wasting |
Proportion of children (of defined age and sex) with HAZ <−2 | Prevalence of stunting |
Proportion of children (of defined age and sex) with WAZ <−2 | Prevalence ofunderweight |
Proportionof children 0–5y (of defined age and sex) with BMIZ >+2 orBMIZ >+3 | Prevalence of overweight or obesity |
Proportion of adult women or men with waist- hip ratios >0.85 (F) and >0.90(M) | Prevalence of abdominal obesity and thus risk of metabolic syndrome |
Proportion of children6–60mos with MUAC <115mm | Prevalence ofsevere acute malnutrition (SAM) |
Proportion of children with SAM who have MUAC >125mm and no edema for at least 2wk after receiving treatment for SAM | Prevalence of childrenready for dis- charge following treatment for SAM |
Further, several factors will affect the magnitude of theexpected response of an anthropometric indicator. These may includethe degree of deficiency, age, sex, and physiologicalstate of the target group. Someexamples of frequently used anthropometric indicatorsand their corresponding application are shown inTable9.1.
9.2 Advantages and limitations of anthropometry
Anthropometric measurements are of increasing importance in nutritional assessment asthey have many advantages (Box9.1). However, anthropometric measuresare relatively insensitive and cannot detect disturbances in nutritionalstatus over short periods of time. Furthermore, nutritionalanthropometry cannot identify any specific nutrient deficiency and,therefore, is unable to distinguish disturbances in growth and bodycomposition induced by nutrient deficiencies (e.g., zinc) from thosecaused by imbalances in protein and energy intake.
Certain non-nutritional factors (such as disease, genetic influences,diurnal variation, and reduced energy expenditure) can lower thespecificity and sensitivity of anthropometric measurements(Section1.4),although such effects generally can be excluded or taken intoaccount by appropriate sampling and experimental design.
Nevertheless, nutritional anthropometry can be used to monitor changes in both growthand body composition in individuals (e.g., hospital patients) and inpopulation groups, provided sources of measurement error and the effectsof confounding factors are minimized(Ulijaszek & Kerr, 1999).
Box9.1. The advantages of anthropometry measurements in nutritional assessment
- Simple, safe, noninvasive techniques are involved,which can be used at the bedside of a single patient, but are alsoapplicable to large sample sizes.
- Inexpensive equipment is required.It is portable, and durable and can be made or purchased locally.
- Relatively unskilled personnel can perform the measurementprocedures if adequately trained
- Methods can be precise and accurate, if standardized techniques andtrained personnel are used.
- Retrospective information is generated onpast long-term nutritional history, which cannot be obtained with equalconfidence using other techniques.
- Mild to moderate undernutrition, aswell as severe states of under- or overnutrition, can be identified.
- Changes in nutritional status over time and from one generation to thenext, a phenomenon known as the secular trend, can be evaluated.
- Screening tests that identify individuals at high risk to under- orovernutrition can be devised.
9.3 Errors in anthropometry
Errors can occur in nutritional anthropometry which mayaffect the precision, accuracy, and validity of the measurements, andthus indices and indicators. Three major sources of error are significant: measurementerrors, alterations in the composition and physical properties ofcertain tissues, and the use of invalid assumptions in the derivation ofbody composition from anthropometric measurements(Heymsfield and Casper, 1987).
Measurement errors arise from examiner error resulting frominadequate training, instrument error, and difficulties in making themeasurement (e.g., skinfold thicknesses). The major sources ofmeasurement error in anthropometry are shown in Boxes 9.2 and 9.3.Both random and systematic measurement errors may occurwhich reduce the validity of the index and any indicatorconstructed from the index; they have been extensively reviewed byUlijaszek and Kerr(1999).
Box9.2: Common errors and possible solutions when measuringlength, height, and weight.
All measurements
- Inadequate instrument: Select method appropriate to resources.
- Restless child: Postpone measurement or involve parent in procedure oruse culturally appropriate procedures.
- Errors in reading equipment: Training and refresher exercises, stressingaccuracy, with intermittent oversight by supervisor.
- Errors in recording results: Record results immediately after measurement and haverecord checked by a second person.
Length
- Incorrect method for age: Use length only when child is <2y.
- Footwear/headwear not removed: Remove as local culture permits(or make allowances).
- Head not in correct Frankfurt plane: Correct position of child before measuring.
- Child not straight along board and/or feet not parallel with movable board:Have assistant and child's parent present; don't take the measurement while the child isstruggling; settle child.
- Board not firmly against heels: Correct pressure should be practiced.
Height
- Incorrect method for age: Use only when child is ≥2y.
- Footwear/headware not removed; Remove aslocal culture permits (or make allowances).
- Head not in correct Frankfurt plane, subject not straight,knees bent, or feet not flat on floor:Correct technique with practice and retraining; provide adequate assistance; calm noncooperative children.
- Board not firmly against head: Lower head board to compress hair.
Weight
- Room cold, no privacy: Use appropriate clinic facilities.
- Scale not calibrated to zero: Re-calibrate after every subject.
- Subject wearing heavy clothing: Remove or make allowances for clothing.
- Subject moving or anxious as a result of prior incident: Wait until subject is calm or remove the cause of anxiety.
From:(Zerfas AJ, 1979)
9.3.1 Random measurement errors and precision
Random measurement errors limit precision or the extentto which repeated measurements of the same variable give the same value.Random measurement errors can be minimized by trainingpersonnel to use standardized techniques and precise,correctly calibrated instruments(Lohman etal., 1988).Furthermore, the precision (and accuracy)of each measurement techniqueshould be firmly established prior to use. To improve precision,two or three measurements on each individual should be conducted.
A description of the measurement techniques used in theWHO Multicenter Growth Reference Study (MGRS)are available in deOnis etal.(2004),as well as in an anthropometric training video from WHO. In the WHO MGRSthe equipment was calibrated regularly, using standard weightsover the full weight range for the portable electronicweighing scales, metal rods of known length for both theinfantometer and stadiometer, and calibration blocks ofvarying widths for the skinfold calipers.
Poor precision often reflects within-examiner error, but between-examinererror may also be significant in surveys with multiple examiners. Theprecision of a measurement technique can be assessed by calculating:
- Technical error of the measurement (TEM)
- percentage technical error (%TEM)
- Coefficient of reliability (R)
These parameters can becalculated for each anthropometric measurement technique from repeatedmeasurements on each subject made within a few minutes to avoidphysiological fluctuations. A minimum of 10 subjects is recommended.
Box9.3: Common errors and possible solutions when measuringmid-upper-arm circumference,head circumference, and triceps skinfold.
It is particularly important with these measurements to use the correct techniques.This requires training, supervision,and regular refresher courses. Always take intoaccount any cultural problems, such as thewearing of arm bands etc.
Arm circumference
- Subject not standing in correct position: Position subject correctly.
- Tape too thick, stretched, or creased: Use correct instrument.
- Wrong arm: Use left arm.
- Mid-arm point incorrectly marked: measure and remark midpoint carefully.
- Arm not hanging loosely by side during measurement: ask subject to allow arm to hang loosely.
- Examiner not comfortable or level with subject: position subject correctly relative to examiner.
- Tape around arm not at midpoint: reposition tape.
- Tape too tight (causing skin contour indentation): loosen tape.
- Tape too loose: carefully tighten tape.
Head circumference
- Occipital protuberance / supraorbital landmarkspoorly defined: position tape correctly.
- Hair crushed inadequately: carefully tighten tape.
- Ears under tape, or tensionposition poorly maintained at time of reading: repeat after positioning tape correctly.
- Headwear not removed: remove as local culture permits.
Triceps fatfold
- Wrong arm: use left arm.
- Mid-arm point or posterior plane incorrectly measuredor marked: measure and remark midpoint carefully.
- Arm not loose by side during measurement: ask subject to allow arm to hang loosely.
- Finger-thumb pinch or caliper placement too deep (muscle) or too superficial (skin): correct technique with training and supervision.
- Caliper jaws not at marked site; reading done too early,pinch not maintained, caliper handle not released: correct technique with training and supervision.
- Examiner not comfortable or level with subject: ensure examiner is correctly positioned.
From:(Zerfas AJ, 1979)
The technical error of the measurement (TEM) is the square root of the measurementerror variance. TEM is expressed in the same units as that of the anthropometricmeasurement under study and is often agedependent. The value is also related to the anthropometriccharacteristics of the study group. The calculation varies according tothe number of replicate measurements made. For one examiner making twomeasurements,TEM =where D = the difference between two measurements and N =number of subjects. For more than two measurements, the equation ismore complex, and TEM =where N = number ofsubjects, K is the number of determinations of the variable taken oneach subject, and Mn is the nth replicate of the measurement, where n varies from 1 to K.
Subject | Stature (m) as determinedon repeat | (1) | (2) | Diff. | |||
---|---|---|---|---|---|---|---|
no. | 1 | 2 | 3 | 4 | ΣM2 | (ΣM)2/K | (1) − (2) |
1 | 0.865 | 0.863 | 0.863 | 0.864 | 2.984259 | 2.984256 | 0.000003 |
2 | 1.023 | 1.023 | 1.027 | 1.025 | 4.198412 | 4.198401 | 0.000011 |
3 | 0.982 | 0.980 | 0.989 | 0.985 | 3.873070 | 3.873024 | 0.000046 |
4 | 0.817 | 0.816 | 0.812 | 0.817 | 2.660178 | 2.660161 | 0.000017 |
5 | 0.901 | 0.894 | 0.900 | 0.903 | 3.236446 | 3.236401 | 0.000045 |
6 | 0.880 | 0.876 | 0.881 | 0.881 | 3.094098 | 3.094081 | 0.000017 |
7 | 0.948 | 0.947 | 0.947 | 0.946 | 3.587238 | 3.587236 | 0.000002 |
8 | 0.906 | 0.905 | 0.907 | 0.908 | 3.286974 | 3.286969 | 0.000005 |
9 | 0.924 | 0.924 | 0.926 | 0.924 | 3.418804 | 3.418801 | 0.000003 |
10 | 0.969 | 0.987 | 1.002 | 0.993 | 3.942343 | 3.942210 | 0.000133 |
Σ = 0.000282 | |||||||
TEM = = = 0.00307 |
shows the calculation of TEM frommeasurements of stature performed four times on 10 subjects by a singleanthropometrist.
Note that the size of the measurement also influencesthe size of the associated TEM, so that comparisonsof precision of different anthropometric measurementsusing TEM cannot be made easily. This is highlighted inTable9.3in which the TEM for five anthropometricmeasurements taken during theinitial standardization session conducted at the Braziliansite of the WHO Multicentre Growth Reference Study (MGRS)are presented(deOnis etal., 2004).Table9.3also depicts the maximum allowable differences betweenthe measurements of two observers that were used inthe WHO MGRS, and set based on TEMs achievedduring the standardization session.
measurement | Brazil TEM from pilot study | Maximum allowable difference |
---|---|---|
Weight | Not available | 100g |
Length | 2.5mm | 7.0mm |
Head circumference | 1.4mm | 5.0mm |
Arm circumference | 1.8mm | 5.0mm |
Triceps skinfold | 0.44mm | 2.0mm |
Subscapular skinfold | 0.43mm | 2.0mm |
Percentage TEM has been recommended to overcome the difficulty of the TEM being dependent onthe size of the original measurement(Norton and Olds, 1996).The percentage technical error ofthe measurement is analogous to the coefficient of variation and is calculated as:\[\small\mbox{%TEM = (TEM/mean) × 100% } \]Note that %TEM has no units and can be used to makedirect comparisons of all types of anthropometric measurements. Itcannot be used, however, when more than one examiner is involved, asthen both within- and between-examiner errors are involved. Ulijaszekand Kerr(1999)describe ways to deal with this more complex case.
The coefficient of reliability (R)is an alternative approach that is widely usedfor comparing measurement errorsamong anthropometric measurements.It ranges from 0to1 and can becalculated using the following equation:\[\small\mbox{ R = 1 −((TEM)}^{2}/ \mbox{s} ^{2}) \]where s2 is the between-subject variance. The coefficientindicates the proportion of between-subject variance in a measuredpopulation which is free frommeasurement error. Hence, a measurementwith R=0.95 indicates that 95% of the variance is due to factors otherthan measurement error.
Whenever possible, a coefficient of reliability>0.95 should be sought. Coefficients of reliability can be used tocompare the relative reliability of different anthropometricmeasurements, and the samemeasurements in different age groups, as wellas for calculating sample sizes in anthropometric surveys.
More details of standardization procedures and calculation of precision using TEM,percentage TEM, and coefficient of reliability are given in Lohman etal.(1988).In general, the precision of weight and heightmeasurements are high. However, for waist and hip circumferences,between-examiner error tends to be large and it is preferable for onlyone examiner to take these measurements. Because skinfolds arenotoriously imprecise, both within- and between-examiner errors can belarge. Therefore, rigorous training using standardized techniques andcalibrated equipment are critical when skinfold measurements are taken.
9.3.2. Systematic measurement errors and accuracy
Systematic measurementerrors affect the accuracy of anthropometric measurements or how closethe measurements are to the true value. The most common form ofsystematic error in anthropometry results from equipment bias. Forexample, apparent discrepancies in skinfold measurements performed onthe same person but with different calipers may be due to compressiondifferences arising from variations in spring pressure and surface areaof the calipers(Schmidt & Carter, 1990); Harpenden and Holtainskinfold calipers consistently yield smaller values than Lange calipers(Gruber etal., 1990).Errors arising from bias reduce the accuracy of the measurementby altering the mean or median value, but have no effect on the variance. Hence, sucherrors do not alter the precision of the measurement.
The timing of some anthropometric measurementsof body size and composition is also known to be critical, particularlyfor short-term growth studies: progressive decreases in the height ofan individual during the day as a consequence of compressionof the spinal column, for example, may seriously compromise the accuracy ofheight velocity measurements.
The determination of accuracy inanthropometry is difficult because the correct value of anyanthropometric measurement is never known with absolute certainty. Inthe absence of absolute reference standards, the accuracy ofanthropometric measurements is estimated by comparing them with thosemade by a criterion anthropometrist(Ulijaszek & Kerr, 1999), a personwho has been highly trained in the standardized measurement techniquesand whose measurements compare well to those from another criterionanthropometrist.
In preparation for the compilation of the new WHO Child Growth Standard,four anthropometrists were trained and standardized against acriterion anthropometrist, designated as the “lead”anthropometrist; see deOnis etal.(2004)for more details. All anthropometric measurements were taken andrecorded independently by two designated anthropometrists, andtheir measurement values compared for maximum allowabledifferences(Table9.3).Targets for sports anthropometrists are also available(Gore etal., 1996).
Attempts should always be made tominimize measurement errors. In longitudinal studies involvingsequential anthropometric measurements on the same group of individuals(e.g., surveillance), it is preferable, whenever possible, to have oneperson carrying out the same measurements throughout the study toeliminate between-examiner errors. This is particularly critical whenincrements in growth and body composition are calculated; suchincrements are generally small and are associated with two error terms,one on each measurement occasion. Recommendations for the minimumintervals necessary to provide reliable data on growth increments duringinfancy and early childhood(Guo etal., 1991)and adolescence(WHO, 1995)are available.
In large regional surveillance studies, it isoften necessary to use several well-trained anthropometrists. In suchcirc*mstances, the between-examiner differences among anthropometristsmust be monitored throughout the study to maintain the quality of themeasurements and thereby toidentify and correct systematic errors inthe measurements. This practice was followed during theWHOMGRS(deOnis etal., 2004).
In studies involving two longitudinal measurements,the TEM can be calculated to estimate the proportion of the differencethat can be attributed to measurement error. For example, with a TEM of0.3 for a given anthropometric measurement,the TEM for the differencebetween two measurements is:because both TEM values contribute to thevariance in the difference. Only if the differenceexceeds 2 × 0.42 = 0.84is there a 95% probability that the difference exceeds themeasurement error alone.
Once assured that such differences are not afunction of measurement error, then any changes in growth and bodycomposition can be correlated with factors such as age, the onset ofdisease, response to nutrition intervention therapy, and so on.
The collection of longitudinal anthropometric data is more time consuming,expensive, and laborious than from cross-sectional surveys, and, as aresult, the sample size is generally smaller. Hence, the probability ofsystematic sampling bias(Section1.4.2)is generally greater than in more extensive cross-sectional surveys.
For cross-sectional studies,the examiners should be rotated among the subjects to reduce the effectof measurement bias of the individual examiners. Statistical methodsexist for removing anthropometric measurement error from cross-sectionalanthropometric data; details are given in Ulijaszek and Lourie(1994).
Cross-sectional surveys are useful for comparing population groups,provided that probability sampling techniques have been used to ensurethat the samples are representative of the populations from which theyare drawn (Chapter1). Recently WHO has provided countries with toolsto develop or strengthen their surveillance systems so theyhave the capacity to monitor changes in the Global NutritionTargets for 2030. They include the followinganthropometric indicators: stunting,wasting, low birthweight, and childhood overweight.For more details see:WHO Nutrition Tracking Tool.
9.3.3 Errors from changes in tissue composition and properties
Variation in the composition and physicalproperties of certain tissues may occur in both healthy and diseasedsubjects, resulting in inaccuracies in certain anthropometricmeasurements. Even among healthy individuals, body weight may beaffected by variations in tissue hydration with the menstrual cycle(Heymsfield and Casper, 1987;Madden and Smith, 2016).
Skinfold thickness measurements may beinfluenced by variations in compressibility and skin thickness with age,gender, and the level of tissue hydration(Martin etal., 1992;Ward and Anderson, 1993).For example, repeated measurements of skinfolds, overa short period (i.e., 5min), may actually decrease accuracy ofskinfolds because later measurements are more compressed due to theexpulsion of water from the adipose tissue at the site of the earliermeasurement(Ulijaszek and Kerr, 1999).
The accuracy of waist circumference is affected by both the phaseof respiration at the point of measurement and by the tensionof the abdominal wall. The phase of respiration is importantbecause it determines the extent of fullness of the lungs andthe position of the diaphragm at the time of the measurement.Increasing the tension of the abdominal wall (by sucking in)is frequently an unconscious reaction which is also importantbecause it reduces the waist measurement. To minimize theseerrors, WHO(2011)recommends advising the subject torelax and take a few deep, natural breaths before the actualmeasurement and at the end of normal expiration.
In addition, during aging, demineralization of thebone and changes in body water may result in a decrease inthe density of the fat-free mass(Visser etal., 1994;JafariNasabian etal., 2017),which are not always takeninto account when calculating total body fat and hencefat-free mass from skinfolds via bodydensity (see Chapter11 for more details).
9.3.4 Invalid models and errors in body composition
Invalid assumptions may lead to erroneous estimates of body compositionwhen these are derived from anthropometricmeasurements, especially inobese or elderly patients and those with protein-energy malnutrition orcertain disease states. For instance, use of skinfold thicknessmeasurements to estimate total body fat assumes that (a) the thicknessof the subcutaneous adipose tissue reflects a constant proportion of thetotal body fat and (b) the sites selected represent the averagethickness of the subcutaneous adipose tissue. In fact, the relationshipbetween subcutaneous and internal fat is nonlinear and varies with bodyweight, age, and disease state. Very lean subjects have a smallerproportion of body fat deposited subcutaneously than do obese subjects,and in malnourished persons there is probably a shift of fat storagefrom subcutaneous to deep visceral sites. Variations in thedistribution of subcutaneous fat also occurs with age, sex, and ethnicityor race(Wagner and Heyward, 2000;He etal., 2002).
Estimates of mid-upper-arm muscle area areused as an index of total body muscle andthe fat-free mass (Chapter11), regardless of age and health statusof the subjects. Such estimates are made, despite the known changes inthe relationship between arm muscle and fat-free mass with age andcertain disease states(Heymsfield and McManus, 1985),and the questionable accuracy of the algorithms used(Martine etal., 1997).Moreover, even the corrected algorithms developed for adultsoverestimate arm muscle area in obese persons when compared with thedetermination by computerized tomography(Forbes etal., 1998).
Increasingly, body composition is assessed by laboratory methods; theseare described in Chapter14. Even laboratory methods are based oncertain assumptions that have been challenged in recent years. Forexample, until recently, densitometry, frequently using underwaterweighing, has been the gold standard reference method for thedetermination of the percentage of body fat. The assumptions used indensitometry are that the densities of the fat mass and fat-free massare constant at 0.90 and 1.10kg/L, respectively (Chapter11).Several researchers have questioned the validity of using a constantdensity of the fat-free mass for groups who vary in age, gender, levelsof body fatness, and race or ethnicity(Visser etal., 1997).During aging, the density of the fat-free mass may decrease due todemineralization of the bone and changes in bodywater, as noted above(Visser etal., 1994;JafariNasabian etal., 2017),which are not always taken into when calculating total body fatfrom skinfolds via body density, leading to a1%–2%overestimate of the body fat content in such subjects(Deurenberg etal., 1989);see Chapter11 for more details.
In contrast, persons of African descent have a larger fat-free massbecause they have a greater bone mineral density and body proteincontent compared to Caucasians(Wagner and Heyward, 2000).Such differences lead to an underestimate of body fat,when generalized equations developed for Caucasians are used.
Percentage of body fat can also be determinedusing an isotope dilution technique and dual‑energy X‑rayabsorptiometry (DXA) (Chapter14). Both of these methods assume aconstant hydration of the fat-free mass (i.e., 73.2% watercontent), despite knowledge that itvaries with age(Wang etal., 1999),obesity, and pregnancy(Hopkinson etal., 1997),and throughout the course of a clinical condition (e.g., inflammation)(Müller etal., 2016).When the actual hydration offat-free mass is higher than the assumed value, then the percentage ofbody fat is underestimated by isotope dilution techniques (Chapter14)(Deurenberg-Yap etal., 2001).For example, even whenpregnancy-specific values for hydration have been appliedto account for the increased accretion of water that occursduring pregnancy, individual estimates of fat mass usingisotope dilution differed by >3kg from valuesbased on the four-compartment model(Hopkinson etal., 1997).In contrast, hydration effects on estimates of fat byDXA are not significant(Pietrobelli etal., 1998).
Fortunately, the advent of multicomponentmodels (i.e., the 4‑compartment-model) with minimal assumptionsfor assessing body composition circumvent the use of older methods, which useassumptions that are not always valid for certain ethnic groups or the elderly(Müller etal., 2016).Nevertheless, the use of multicomponent models is expensive,requiring more time and facilities.
9.4 Interpretation andevaluation of anthropometric data
Anthropometric indices are derived from two or moreraw measurements, as noted earlier. Normally, it is these indices thatare interpreted and evaluated — not the raw measurements.Anthropometric indices can be used atboth the individual and the population levels to assessnutritional status and to screen and assess a response duringinterventions. In addition, in populations, anthropometry can be usedto identify the determinants andconsequences of malnutrition and fornutritional surveillance. To achieve these objectives, knowledge offactors that may modify or “condition”the interpretation of abnormalanthropometric indices is generally required. These conditioningfactors are briefly discussed below,together with anthropometric reference data andmethods to evaluate anthropometricindices, including classification systems that identifyindividuals and populations as “at risk” for malnutrition.
9.4.1 Conditioning factors
A variety of factors are known to modify or condition the interpretationof anthropometric data and must be taken into account. Some importantexamples include age, birthweight, birth length, gestational age, sex,parental stature, and feeding mode during infancy. Maturation duringadolescence, prepregnancy weight, maternalheight, parity, smoking, pregnancy, and ethnicityare major conditioning factors for adults(WHO, 1995).
Information on some of these conditioning factors can be obtained by physicalexaminations, questionnaires, or self-reports. An accurate assessmentof age is especially critical for the derivation of many anthropometricindices used to identify abnormal anthropometry, notably height-for-ageand weight-for-age. See:WHO Child Growth Standards.
Age is also important for categorizing the data intothe age groups recommended by WHO for analysis andinterpretation(WHO/UNICEF, 2019);see Chapter13 for details.In more affluent countries, the assessment of age using birth certificatesis generally easy, but in some low-income countries, local calendars ofspecial events are often constructed to assist in identifying the birthdate of a child.
Alternatively, for young children, age is sometimesassessed by counting deciduous teeth. This method is most appropriateat the population level because of the wide variation among individualsin the timing of deciduous eruption(Delgado etal., 1975).For individuals, bone age can be estimated from the left-hand-wristradiograph using the Tanner WhitehouseII method(Tanner etal., 1983).Gorstein(1989)has highlighted the marked discrepancies that may occurin the prevalence estimates of undernutrition during infancy whendifferent methods are used to determine age.
With infants, an accurateassessment of birth weight, and, if possible, birth length andgestational age, is also important(Hediger etal., 1999). Assessmentof gestational age is especially critical for the interpretation ofboth size-for-age measurements during infancy and the neurodevelopmentalprogress of preterm infants. It is also essential for the management ofpregnancy and the treatment of new-born infants.
Several strategies areavailable for estimating gestational age. Prenatal measures ofgestational age include calculating the number of completed weeks sincethe beginning of the last menstrual period, prenatal ultrasonography,and clinical methods; they are all described inChapter10. In public health settings,the definition of gestational age on the basis of the lastmenstrual period is most frequently used, although it isassociated with several problems: errors may occurbecause of irregular menses, bleeding early in pregnancy,and incorrect recall by mothers. Prenatal ultrasonographyduring the first or second trimester, although consideredthe gold standard method for assessment of gestationalage, is not universally available, especially in low-income countries.Furthermore, the quality of both the equipment used and the technical training varies.
For studies of the adolescent age group, defined by WHO(1995)as 10–19y,information on maturation should also be collected in view of the markedvariation in the timing of the maturational changes during adolescence.The best measure of maturity is bone age — often termed skeletalmaturation — because it can be obtained for both sexes over a wide agerange. However, special equipment and expertise are required for theassessment of bone age. Hence, instead, surrogate measures of somaticmaturation are generally used in nutrition surveys. WHO(1995)recommends the use of two maturational events for each sex to assist ininterpreting anthropometric data during adolescence: one markersignaling the beginning of the adolescent growth spurt in each sex, andone indicating that the peak velocity for height and associated changeshave passed. In girls, the indicator that can be used to signal that theadolescent growth spurt has begun is the start of breast development,which precedes peak height velocity by about 1y.The marker indicating that most of the adolescent growthspurt has been completed is the attainment of menarche,which begins a little more than 1y after peak heightvelocity (Figure9.1). In boys, the corresponding indicatorssignaling the beginning and completion of the adolescent growthspurt are adolescent changes in the penis,characterizing G3, followed by the attainment of adult voice, respectively(Figure9.1).
When an assessment of somatic maturationcannot be obtained by physical examination and questioning, then a self-administeredquestionnaire containing drawings illustrating Tanner's stages ofdevelopment of breasts and pubic hair for females, or pubic hair andmale genitalia, may be used. Adolescents are requested to select thedrawing closest to their stage of development, as described in Morrisand Udry(1980).
9.4.2 Appropriate anthropometric reference data
In public health settings, appropriate anthropometric reference datafacilitate international comparisons of anthropometricindices across populations and enable the proportionof individuals with abnormal indices to be determinedrelative to the reference population. Such comparisonsenable the extent and severity of malnutrition in the studygroup to be estimated. In surveillance studies, referencedata allow the evaluation of trends over time, as well asthe effectiveness of intervention programs to be assessed.Reference data can also be used in clinical settingsto monitor growth of individuals, detect abnormalchanges in growth, and assess response to treatment(WHO, 1995).
The WHO recommends the use of the WHO Child Growth Standards foryoung children from birth to 5y for international use(WHO, 2006)in view of the small effect of ethnic and genetic differences on the growth ofinfants and young children compared with theenvironmental, nutritional, and socio-economic effects,some of which may persist across generations.The WHO Child Growth Standard was developedas a result of the technical and biological limitationsidentified with the earlier NCHS/WHO growth reference(Garza and deOnis, 1999).A prescriptive approachdepicting physiological human growth under optimal conditionswas used for the new Child Growth Standards so they representhow young children should grow, rather than asa “reference” describing how children do grow.To achieve this goal, a set of individual eligibility criteriawere developed: term singleton infants with non-smoking mothers,a health status that did not constrain growth, and mothers whowere willing to follow current WHO feeding recommendations.The design combined a longitudinal study from birth to 24moswith a cross-sectional study of children aged18–71mosbased on pooled data from 6 participating countries(Brazil, Ghana, India, Norway, Oman, and the UnitedStates)(deOnis etal., 2004).WHO has developed a tool forthe application of the WHO Child Growth Standardswhich includes instructions on how to take themeasurements, interpret growth indicators, investigatecauses of growth problems, and how to counsel caregivers.An anthropometry training video is also available. For more details, see:WHO Child Growth Training Module.
For older children, the WHO growth reference data for school-agechildren and adolescents 5–19y should be used(deOnis etal., 2007).This is a reconstruction of the original 1977 National Centre forHealth Statistics (NCHS) data set supplemented withdata from the WHO Child Growth Standard. The statisticalmethodology used to construct this reference was thesame as that used for the WHO Child Growth Standard.
A series of prescriptive standards for monitoring fetal,newborn growth, and gestational weight gain have alsobeen developed for international use by theINTERGROWTH‑21st project.This project adhered to the WHO recommendations forassessing human size and growth, and followed healthypregnant women longitudinally from 9wks of fetal lifeto 2y(Papageorghiou etal., 2018; Ismail etal., 2016).Populations from urban areas of 8 countries in whichmaternal health care and nutritional needs were met(Brazil, China, India, Italy, Kenya, Oman, the UK and the USA)were involved to ensure universal multi-ethnic growthstandards were generated that represent how fetusesshould grow. Postnatal growth standards for preterminfants were also developed by this group(Villar etal., 2015).
Updated childhood growth charts have been preparedby the Center for Disease Control (CDC) for U.S children for two age groups:0–36mos and2–20y.These CDC2000 growth charts are based primarily onphysical measurements taken duringfive nationally representative surveys conducted between 1963 and 1994,although some supplemental data were also used.When creating these revised growth charts, two data sets were excluded:growth data for very low birthweight infants (<1500g)whose growth differs from that of normal birth-weight infants,and weight data for children >6y who participatedin the NHANESIII survey. The latter data wereexcluded from both the revised weight and BMI growthcharts because their inclusion shifted the upper percentile curves.Hence, the exclusion of these selected data resultedin a modified growth reference that is not a purelydescriptive growth reference because it does notcontain representative national data for all variables(Kuczmarski etal., 2000).A comparison of theseCDC 2000 Growth Charts with the WHO Child GrowthStandards is available in deOnis etal.(2007).
For body composition indices, use of localreference data are preferred becauseracial differences exist in both body proportions, and theamount and distribution of subcutaneous andintra-abdominal fat(Wagner and Heyward, 2000;He etal., 2002;Lim etal., 2019).In practice, however, only a few countrieshave local body composition referencedata. In the absence of such local data,WHO recommend the use of the reference datafor mid-upper-arm circumference (MUAC),triceps and subscapular skinfolds based on thedata collected during the WHO MGRS on children age0–5y.Electronic copies of the WHO tables and charts ofpercentiles and Z‑scores for MUAC‑for-age,triceps-for-age, and subscapular-for age by sexare available for children from age 3mos to 71mosand are included inWHO Child Growth Standards.
In clinical settings, abnormal changes in the rate ofgrowth of a child can be detected much earlier whengrowth velocity charts, rather than distance growth charts,are used; see Chapter13 for more details.Growth velocity charts are based on longitudinalstudies during which the same child is measured serially,and the growth rate calculated for each interval.WHO has developed a set of growth velocity charts basedon the WHO MGRS described earlier forinternational use; see deOnis etal.(2011)for more details.
Several different distance growth standards have beencompiled, depending on the specific determinants ofgrowth. For example, the new international postnatalgrowth standards should be used for preterminfants(Villar etal., 2015)in view of the differencein weight, length, and head circumference between pretermand full-term infants. Moreover, the time period over which thesedifferences extend varies with the growth measurement.Differences are significant until 18 mos for head circumference,until 24mos for weight-for-age, and up to 3.5y for length/height-for-age.
Alternatively, tempo-conditional growth charts can be used for monitoring thegrowth of individual children during adolescence(Figure9.2).
These growth charts are based on mixed cross-sectionaland longitudinal data, and take into account differencesin the timing of the adolescent growth spurt — termedthe “phase difference effect”(Tanner & Buckler, 1997).
Parent-allowed-for growth reference data are available forchildren from2–9y when nonfamilial short stature isof concern. Cole(2000)developed a novel parent-allowed-for-heightchart that adjusts for mid-parent, single parent, or siblingheight based on the UK90-height reference. Specialgrowth charts have also been compiled for childrenwith certain genetic disorders such as Down's syndromeor other developmental disorders in which growth patternsdiffer from the reference growth curves.
9.4.3 Methods of evaluating anthropometric indices
For studies of both individuals and populations, the anthropometricindices can be compared to the reference population usingpercentiles or Z‑scores derived from the distributionof the anthropometric reference data. A percentile refers to the positionof the measurement value in relation to all the measurementsfor the reference population, ranked in order of magnitude.A Z‑score (or standard deviation score) measuresthe deviation of the value for an individual from the medianvalue of the reference population, divided by the standarddeviation of the reference, as shown below:\[\small \mbox{Z‑score or SD score = } \frac {\mbox {(observed value) − (median reference value)}}{\mbox {(standard deviation of reference population)}}\]In most industrialized countries, percentiles are used becauseno errors are introduced if the data have a skewed distribution.Weight-for-age, weight-for-height, and many circumferentialand skinfold indices have skewed distributions.
percentiles are not appropriate for use inlow- and middle-income countries (LMICs) wheremany children may fall below the lowest percentile.
In these settings Z‑scores should be usedbecause they can be calculated accurately beyond thelimits of the original reference data. Hence, individualswith indices below the extreme percentiles of thereference data can then be classified accurately.For more discussion of percentiles and Z‑scores, see Chapter13.Both percentiles and Z‑scores based on the WHOChild Growth Standards (0-5y) and the WHOGrowth Reference for school-aged children andadolescents (5-19y) can be readily calculatedin population studies using the WHO softwareprogram: WHO AnthroPlus (2009). Alternatively, forindividuals in clinical settings, the percentile orZ‑score range within which the measurement of anindividual falls can be read from sex-specificcharts or tables of the appropriate reference data, as shown inFigure9.3.
Three methods have been recommended by WHO to evaluatecross-sectional anthropometric data for use inpublic health; these are summarized in Box9.4.
Box9.4. Methods recommended by WHOto assess anthropometry
- Comparison of the frequencydistribution of anthropometricindices with appropriate WHOgrowth reference data using Z‑scores
- Summary statistics of the Z‑scores:mean; median; SD; standard error(SE) with the 95% confidence interval(CI) for each growth indicator
- Calculation of number and proportionof individuals (as %)with anthropometricindices below or above a designated cutoff (95% CIs)for each age and sex group.
is an example of the first method summarizedin Box9.4. Here the frequency distribution of theZ‑scores for height-for-age children from the IndianNational Family Health Survey (2005-2006) are comparedwith the corresponding reference distribution of height-for-ageZ‑scores for the WHO Child Growth Standards.The figure highlights that nearly all the children surveyedwere affected by some degree of linear growth retardationand would benefit from an intervention; this approach is termed a “population approach to targeting”.
Summary statistics can also be used whenanthropometric indices are expressed as Z‑scoresin population studies. In the example given in Figure9.4,the calculated mean height-for-age Z‑score (-1.83)for the children was markedly lower than zero —the expected value for the reference distribution.This statistic alone indicatesthat the entire distribution has shifted downward,suggesting that most, if not all of the individuals,are affected by linear growth retardation, as was clearfrom the frequency distribution of the height-for-ageZ‑scores comparedwith the corresponding reference distribution depicted, in Figure9.4.
Note, summary statisticscannot be calculated in this way for data from a populationexpressed in terms of percentiles which are often not normally distributed.
Alternatively, calculation of the mean Z‑score and SDcan be used to compare directly different populationsor the status of the same population at different times(Goldstein and Tanner, (1980).However, even if the populations have thesame mean Z‑score, their SDs may differ, with the populationwith the larger SD having a greater proportionbelow the reference limit or cutoff point, as shown in(Figure9.5).Here, the growth reference is not being usedfor comparative purposes as shown in Figure9.4.
The third method itemized in Box9.4 involvescalculating the percentage of individuals withanthropometric indices below or abovepredetermined reference limits or cutoff points.When used in this way, the anthropometric indexand its associated reference limit or cutoff pointare termed an “indicator” as described earlier.This approach is used to classify individuals as “at risk”to malnutrition and isused by governments and International Agencies (e.g., WHO and UNICEF)to generate prevalence estimates of malnutritionfor information and comparisons across countries,as well as for advocacy. The approach is describedin more detail below.
9.4.4 Classification systems
Classification systems are used in both clinicalsettings and in public health. All use at least oneanthropometric measurement and one or morereference limit derived from appropriate referencedata (i.e., indicator) to classify at risk individuals.Alternatively, cut-off points are used. In practice,classification schemes are not perfect, somemisclassification will always occur so that someindividuals identified as “at risk” tomalnutrition will not be truly malnourished (false positives),and others classified as “not at risk” tomalnutrition will in fact be malnourished (false negatives).Misclassification arises because there is alwaysbiological variation among individuals (and hence inthe normal levels defined by themeasurement)(Fraser, 2004); see Chapter13 for more details.
Reference limits for anthropometric indices arederived from a reference distribution and canbe expressed in terms of Z‑scores or percentiles.In low income countries reference limits defined byZ‑scores are frequently applied, with scoresbelow −2or above +2Z‑scoresof the WHO Child Growth Standard or WHO Growth Referenceused to designate individuals with eitherunusually low or unusually high anthropometricindices(WHO, 1995).This approach is usedbecause statistically 95% of the international referencepopulation fall within the central range assumed tobe “healthy”.Therore, theoretically,the proportion of children with a Z‑score lessthan −2or greater than+2 in astudy population should be ≈2.3%.Clearly, if the proportion in the study populationwith such low or high Z‑scores is significantlygreater than this, then the study population isseriously affected. The WHO uses thebelow −2Z‑scores of the whor*ference median for weight-for-age, length/height-for-age,or weight-for-height to classify children as underweight,stunted, or wasted, respectively.
In industrialized countries,the percentiles commonly used for designatingindividuals as “at risk” to malnutritionare either below the 3rd or 5th andabove the 97th or 95th percentiles.The limits chosen depend on the reference data used: see Chapter13 for more details.
There is nothing immutable about a referencelimit at −2Z‑score, despite theiruse by countries and agencies to generate prevalenceestimates, most notably for stunting and wasting.As a consequence, attempts have been made toestablish “cutoffs” for someanthropometric indicators to improve the abilityto discriminate between children who aremalnourished and those who are “healthy”.These cutoffs have beenestablished by a review of the anthropometriccharacteristics of individuals with either clinicallymoderate or severe malnutrition or who subsequently die.However, many other characteristics of individuals suchas age, sex, life-stage, race/ethnicity, genetics, andmorbidity or nutritional status may affect the relationship under study(Hondru etal., 2019; Yaghootkar etal., 2020; Wright etal., 2021).Hence, in practice, defining cutoffs is difficult becausethe relationship between the indices and the biologicalfactors cannot be generalized from one region to another(WHO Expert Consultation, 2004)..Consequently, in some studies universal cutoff points are used,whereas in others the methods used to identifythe cutoff points applied are not always well documented.
As an example in a study of low BMI and morbidityin Pakistan, a reported a cutoff of <18.5 wasassociated with higher morbidity, whereas in Calcuttait was <16.0(Campbell and Ulijaszek, 1994;Kennedy and Garcia, 1994).
Cut-offs for BMI, waistcircumference, and waist-hip ratio associatedwith risk of cardiovascular disease and type2 diabeteshave also been extensively investigated among differentethnic groups(Ding etal., 2020; WHO Expert Consultation, 2004; Lear etal., 2010).Some expert groups have defined lower waist circumference cutoffs for adults ofAsian descent compared to Europeans(IDF, 2006).
Receiver operating characteristic (ROC) curves are oftenused to determine cutoff points. This is a graphical methodof comparing indices and portraying the trade-offs thatoccur in the sensitivity and specificity of a measurementor index when the cutoffs are altered.To use this approach, a spectrum of cutoffs over theobserved range of the indicator results is used, and thesensitivity and specificity for each cutoff calculated.Next, the sensitivity (or true-positive) rate is plotted on thevertical axis against the true negative rate (1−specificity)on the horizontal axis for each cutoff point, as shown in(Figure9.6).
The closer the curve follows the left-hand of the ROC space,the more accurate is the cutoff under study indistinguishing the health or nutritional status conditionunder investigation from optimal status. The optimalROC curve is the line connecting the points highest and farthestto the left of the upper corner. The closer the curve comesto the 45° diagonal of the ROC space,the less accurate the indicator cutoffs(Søreide, 2009).Most statistical programs (e.g., SPSS) provide ROC curve analysis.Details of alternative statistical methods for selecting thebest cutoff point are given in Brownie etal.(1986).
The choice of the cutoff may vary depending on the circ*mstances.When resources are scarce, a low cutoff point may be selected.As a consequence, the sensitivity decreases, which mean thatmore truly malnourished children are missed. However, at thesame time, the specificity increases, which means thatfewer well-nourished children are misdiagnosed as malnourished.Conversely, when resources are generous, the cutoffcan be high, because it does not matter if some children receivetreatment when they do not need it. The inverse relationshipbetween sensitivity and specificity and relative risk of mortalityassociated with various values for MUAC in children6–36mosin rural Bangladesh is shown inTable9.4.
Arm circum- ference (mm) | Sensitivity (%) | Specificity (%) | Relative Risk of death |
---|---|---|---|
≤100 | 42 | 99 | 48 |
100–110 | 56 | 94 | 20 |
110–120 | 77 | 77 | 11 |
120–130 | 90 | 40 | 6 |
More details of the inter-relationships between these variables is given in Chapter1.
Unfortunately, because sensitivity and specificity datafor the anthropometric indices selected are usuallynot known for the population under study, the datarequired to plot ROC curves areoften obtained elsewhere, even though the valuesmay not be appropriate for the population understudy because of the many factors known toinfluence cutoff values, as noted earlier; see Chapter13for more details. Ultimately the choice of anindex and an associated cutoff point (i.e., an indicator)depends on the resources available and the purpose for which it is being used.The latter can range from screening for disease or monitoringto detect changes in prevalence of malnutrition etc.For more discussion on the selection of anthropometric indicators, see Brownie etal.(1986)and WHO(1995).
Screening tools to identify those who are malnourishedin clinical or public health settings can be basedon single or multiple measurements and associatedreference limits or cutoff points.An example of a screening tool used in clinical settings is theMalnutrition Universal Screening Tool (MUST),widely used to identify adults who are at riskof undernutrition or obesity (See Chapter27).MUST is based on height and weight measurementsfrom which both a BMI score (from 3 cutoffs)assessed via a chart and weight loss in previous3–6mos,can be derived, followed by establishing acute diseaseeffect and score. From the sum of the three scores,the overall risk of malnutrition is calculated,with management guidelines provided according to the levelof risk (low; medium; high). More details are available:Malnutrition Universal Screening Tool .
In public health emergencies, a screening toolbased on a single measurement (i.e., MUAC)and associated cutoff (i.e., <115mm)is often used to identify severe acute malnutrition (SAM)in children 6–60mos. This MUAC cutoff was chosenbecause children with a MUAC <115mm wereobserved to have a highly elevated risk of death compared to those with aMUAC >115mm(Myatt etal., 2006).
For defining overweightand obesity in children and adolescents,WHO recommends the use of BMI-for-ageZ‑scores and reference limits based on Z‑scores.For children (0–5y), a Z‑score for BMI-for-ageabove+1 is described as being “at riskof overweight”, above+2 as “overweight”,and above+3 as “obese”based on the WHO Child Growth Standard. Forchildren 5–19y,BMI-for-age Z‑scores above+1and above+2 based on the WHO 2007growth reference data are recommended(deOnis & Lobstein, 2010).To classify overweight and obesity in adults,however, WHO recommends a graded classification scheme, as shown in(Table9.5).
Classification | BMI (kg/m2) | Risk of comorbidities |
---|---|---|
Underweight | <18.50 | Low (but risk of clinical problems is increased) |
Normal range | 18.50–24.99 | Average |
Overweight | ≥25.00 | |
Pre-obese | 25.00–29.99 | Increased |
Obese class I | 30.00–34.99 | Moderate |
Obese class II | 35.00–39.99 | Severe |
Obese class III | ≥40.00 | Very severe |
Increasingly, it is recognized that in low-income countries,multiple anthropometric deficits may occursimultaneously in children and amplify theirrisk of morbidity and mortality. Consequently, a composite index of anthropometricfailure (CIAF) has been developed, andis described in Chapter13.
In public health, screening tools are also used to mapcountries according to levels of severity ofmalnutrition(UNICEF/WHO/World Bank, 2021)in orderto identify priority countries.Five prevalence thresholds (as %) basedwasting (i.e., WHZ <−2),overweight (BMIZ >+2), andstunting (i.e., HAZ <−2) have beendeveloped by WHO and UNICEF; these aredepicted in(Table9.6).The fifth thresholdlabelled “very low” and of no publichealth concern was included across all threeindicators to reflect the expected prevalenceof 2.3% (rounded to 2.5%) below/above 2SDsfrom the median of the WHO Child Growth Standard.
Wasting | Overweight | Stunting | ||||||
---|---|---|---|---|---|---|---|---|
Prevalence thresholds (%) | Labels | (n) | Prevalence thresholds (%) | Labels | (n) | Prevalence thresholds (%) | Labels | (n) |
<2·5 | Very low | 36 | <2·5 | Very low | 18 | <2·5 | Very low | 4 |
2·5 – <5 | Low | 33 | 2·5 – <5 | Low | 33 | 2·5 – <10 | Low | 26 |
5 – <10 | Medium | 39 | 5 – <10 | Medium | 50 | 10 – <20 | Medium | 30 |
10 – <15 | High | 14 | 10 – <15 | High | 18 | 20 – <30 | High | 30 |
≥15 | Very high | 10 | ≥15 | Very high | 9 | ≥30 | Very high | 44 |
The number of countries in different thresholdcategories for wasting, overweight, and stunting,also shown in(Table9.6),is based on data from 134countries.Comparison of the prevalence estimatesfor each anthropometric indicator can triggercountries to identify the most appropriateintervention program toachieve “low” or “very low” prevalence thresholds.
Details of the techniques used to measure body size andbody composition, together with the indices derived from thesemeasurements, are discussed in Chapters10 and11,whereas Chapter13 discussesmethods used for evaluation of anthropometric indicesand their application at the individual and population level.