Gibson RS1  and Meredith-Jones K2
Principles of Nutritional           
Assess­ment:   Body Size

3rd Edition
January 2023


Body size can be assessed from measurements of stature (height or length), weight, and head circumference using stan­dard­ized procedures described here. When height cannot be measured, published equations based on knee height, lower leg length, arm span or demi-span are used. To interpret the measurements, indices such as weight-for-age, body-mass-index-for-age, stature-for age, weight-for stature, and body mass index (BMI, weight / height2) are constructed.

To evaluate the anthropometric indices, they can be compared with appropriate growth reference data using Z‑scores or per­cent­iles. For international use, the WHO prescriptive Child Growth Standard for children (0–5y) and the WHO growth reference for school-age children and adolescents (5–19y) should be used. Statistically determined reference limits or cutoffs can also be applied to generate “anthropometric indicators” such as stunting (stature-for-age < −2 Z‑score), wasting (weight-for-stature < −2 Z‑score), over­weight (weight-for-stature > +2 Z‑score), and thinness, over­weight, and obesity based on BMI-for-age < −2, > +2, and > +3 Z‑scores, respectively. Both stature-for-age at < − Z‑score and weight-for-stature at < −2 Z‑score are recommended by WHO because together they can distinguish between stunting and wasting. To assess changes in stature over time, however, use of stature-for-age difference rather than stature-for-age Z‑score is now preferred.

To assess and compare the severity of malnutrition across countries, five prevalence thresholds for wasting, over­weight, and stunting can be used. Recognition of an elevated mortality risk for children suffering from multiple anthropometric deficits simultaneously has led to the development of the Composite Index of Anthropometric Growth Failure (CIAF). This is used to characterize the overall burden of under‑ and ‑over-nutrition in children age < 5y so they can be identified and prioritized for intervention.

Changes in body weight over time are used in both clinical and public health settings. In clinical settings lack of weight gain and weight loss are used identify pediatric undernutrition, whereas weekly weight gain is used to monitor response of children with severe acute malnutrition (SAM) to feeding programs. In public health, changes in body weight over time are often monitored because of their increased risk of mortality. New U.S guidelines for gestational weight gain are now available which have been adopted by WHO for low-income populations until specific country cutoffs are available.

BMI is currently the indicator of choice for defining under­weight, over­weight, and obesity in adults and children. However, the limitations of BMI alone has led to the use of both BMI and waist circumference because the distribution of abdominal visceral fat has an even greater cardio­metabolic risk than excess body fat per se. The BMI cutoffs used to classify adults as under­weight (BMI < 18.5kg/m2), over­weight (BMI > 25kg/m2), and obese (BMI > 30kg/m2) are based on adverse health risks and sometimes risk of mortality. In adults, WHO recommends the use of universal BMI cutoffs for over­weight (BMI > 25kg/m2) and obesity (BMI > 30kg/m2); three classes of obesity are also defined. This classification has been adopted by several countries, including Canada and the United States, with the highest class indicative of severe obesity (i.e., BMI > 40kg/m2). Cutoffs to define grades of low BMI values indicative of under­weight in adults are also recommended together with a classification to identify populations with a public health problem. Increasingly, low- and middle-income countries are impacted by both under- and over-nutrition, a condition termed the “double burden of malnutrition”.

In children, no simple cutoff to define thinness, over­weight, or obesity can be used because BMI has a characteristic curvilinear shape with age. Instead, countries have compiled their own ways of defining thinness, over­weight, and obesity in children. Three classification schemes are described with cutoffs defined by BMI-for-age Z‑scores or per­cent­iles based on international or national growth reference data. They include those set by the International Obesity Task Force (IOTF), the new WHO classification based on their international growth standard and growth reference, and the U.S classification based on the 2000 CDC growth charts. Clearly, there is an urgent need for consensus on BMI cutoffs for thinness, over­weight, and obesity in childhood that are defined by adverse health risks. Only in this way can valid international comparisons across countries on the prevalence of thinness, over­weight, and obesity in childhood be made.

CITE AS: Gibson RS and Meredith-Jones K. Principles of Nutritional Assess­ment: Body Size. https://nutritionalassess­
Email: Rosalind.Gibson@Otago.AC.NZ
Licensed under CC-BY-SA-4.0

10.0 Anthro­pometric assess­ment of body size

The most widely used anthro­pometric measure­ments of body size are those of stature (height or length) and body weight. These measure­ments can be made quickly and easily and, with care and training, accurately. Head circum­ference measure­ments are often taken in associ­ation with stature. Details of stan­dard­ized proce­dures for these measure­ments of body size are summarized below and are given in detail in (Lohman et al., 1988).

Indices such as head-circum­ference-for-age, weight-for-age, weight-for-stature, and stature-for-age and the ratio of weight to stature are derived from these measure­ments. Of these, stature-for-age and weight-for-stature have been recom­mended by the World Health Organization (WHO, 1995a) for use in low-income countries. In combi­nation, they can distin­guish between stunting and wasting. There is also recognition that indi­vidual chil­dren may be simul­taneously wasted and stunted, which has prompted interest in ident­ifying the combi­nation of anthro­pometric deficits with the greatest risk of mortality (McDonald et al., 2013).

In epidemiological studies body mass index (BMI) (wt/ht2, as kg/m2) is used to define thinness, over­weight, and obesity. Increas­ingly, BMI is accompanied by waist circum­ference to provide an assess­ment of abdom­inal fat in chil­dren, adoles­cents, and adults. In hospital settings, anthro­pometric indices of body size are used primarily to identify under- or over-nutrition and obesity, and to monitor changes after a nutrition intervention.

10.1 Measure­ments of body size

Of the various measure­ments of body size, head circum­ference is impor­tant because it is closely related to brain size. It is often used with other measure­ments to detect the pathological conditions asso­ciated with either an unusually large (macro­cephalic) or small (micro­cephalic) head.

Recum­bent length is measured in infants and chil­dren < 2y. Height is mea­sured in older chil­dren and adults. The inter­pretation of length (and weight) at birth and during later infancy requires a valid and precise estimate of gesta­tional age: making such an estimate is often difficult in many low-income countries. To assess growth over short time periods, lower-leg length in infants and children can be mea­sured. In adults, knee-height measure­ments are used to estimate height in those persons with severe spinal curvature and in those who are unable to stand. Alterna­tively, arm span or demi-arm span can be mea­sured when actual height cannot be performed.

Weight in infants and young chil­dren can be measured using a suspended scale and a weighing sling, or for greater precision (within 10g), a pediatric scale. For older chil­dren and adults, portable elec­tronic digital scales with a taring capacity are recom­mended. Elbow breadth is used as a measure of frame size, which is relatively inde­pen­dent of adiposity and age (Frisancho, 1990); it should be measured with flat-bladed sliding calipers.

For details on the measure­ment tech­niques and standard­ization protocols used in the World Health Organization (WHO) Multi­center Growth Refer­ence Study (MGRS), the reader is referred to: the WHO Anthropometry training course available from WHO (Training Course).

10.1.1 Head circum­ference

For the measure­ment of head circum­ference, a narrow, flexible, non-stretch tape made of fiberglass or steel (range 0–200cm, cali­brated to 1mm, and about 0.7cm wide) should be used. Any head-bands or hair-pins should be removed and braids undone for the measure­ment.

An infant or child below the age of two years should be held on the mother's lap, whereas older chil­dren can stand with the left side facing the measurer, with arms relaxed and legs apart. The child must look straight ahead so the line of vision is per­pen­dic­ular to the body and the Frankfurt plane — that is, an imaginary plane which passes through the external auditory meatus (the small flap of skin on the forward edge of the ear) and over the top of the lower bone of the eye socket immedi­ately under the eye — is horizontal. The tape is placed just above the supraorbital ridges (i.e., slightly above the eye­brows) covering the most prominent part of the frontal bulge and over the part of the occiput that gives the maximum circum­ference (i.e., fullest protuberance of the skull at the back of the head) Figure 10.1. Care must be taken to ensure that the tape is at the same level on each side of the head and pulled tightly to compress the hair and skin. Measure­ments are recorded to the last completed 1mm (de Onis et al., 2004).
Figure 10.1
Figure 10.1. Measurement of head circum­ference.

Where possible, serial measure­ments of head circum­ference should be incor­porated into routine well-child care, to establish if head circum­ference is within normal limits, too large (mega­cephaly), or too small (micro­cephaly) (Harris, 2015). For inter­national use, the WHO Child Growth Standards that include head circum­ference measure­ments are recom­mended (Growth Standards).

10.1.2 Gestational age

The assess­ment of gesta­tional age is necessary for the inter­pretation of any size-for-age measure­ment of infants and for following the neurodevelop­mental progress of preterm infants. It is also essen­tial for the management of pregnancy and treat­ment of the new­born.

Several strategies are available for estimating gesta­tional age. Prenatal measures of gesta­tional age include calculating the number of completed weeks since the beginning of the last men­strual period, prenatal ultra­sonography, and clinical methods. Of these, the definition of gesta­tional age on the basis of the last men­strual period (LMP) is most frequently used in low-income countries, but it is asso­ciated with several problems: errors may occur because of irregular menses, bleeding early in preg­nancy, and incor­rect recall by mothers. Macaulay et al. (2019) concluded that use of LMP for gesta­tional age dating during early preg­nancy was not sensitive in ident­ifying late- and post-term preg­nan­cies based on a compar­ison of gesta­tional age estimates from LMP and ultra­sonography.

Prenatal ultra­sonography during the first or second trimester is considered by many to be the gold standard for assess­ment of gesta­tional age. Estimates are based on different ultra­sonic measures of fetal size, with crown-rump length considered the best single parameter for the first trimester, whereas for a second- or third-tri­mester scan, a combi­nation of multiple bio­metric parameters such as the bipari­etal diameter, head circum­ference, femoral length, and abdom­inal circum­ference are recom­mended (Butt & Lim, 2019). Measure­ments are most accurate when made early in gesta­tion.

Fetal growth charts based on ultra­sound biometric measure­ments from multi-national pro­spec­tive observational longitu­dinal studies have been devel­oped by WHO (Kiserud et al., 2017) and by the INTER­GROWTH-21st project (Papageorghiou et al., 2014). These fetal growth charts are based on serial ultra­sound exam­in­ations during preg­nancy in which women with obstetric condit­ions that may influence growth were excluded. The standards of the INTER­GROWTH-21st project are based on popu­lations from urban centers in eight countries in which maternal health care and the nutritional needs of women were met (Brazil, China, India, Italy, Kenya, Oman, the UK and the USA). The objective was to generate universal multi-ethnic growth standards that represent how fetuses should grow (i.e., the standards are prescriptive) when nutritional, environmental, and health constraints on growth are minimal. Fetal growth charts for estimated fetal weight and common ultra­sound biometric measure­ments at 14–42wks of gestation are presented (Papageorghiou et al., 2014).

Unfortunately, ultra­sonography is not universally available, especially in low-income countries, and the quality of both the equipment used and the tech­nical training varies. Instead, clinical methods of prenatal assess­ment such as measure­ment of symphysis-fundal height (in cm) are often used, which should correlate with the week of gestation after 20wks. Inter­national standards for symphysis-fundal height based on serial measure­ments from the fetal growth longitu­dinal study of the INTER­GROWTH-21st Project are also available (Papageorghiou et al., 2016). Alternative clinical methods some­times used include the auscul­tation of fetal heart tones (audible at 11–12wks), and the recording of fetal move­ments (Alexander & Allen, 1996; (Butt & Lim, 2019).

Several scoring systems, based on external and neurological criteria, have been devel­oped to estimate maturity — and thus gesta­tional age — post­natally. The scoring systems initially devised by Dubowitz et al. (1970) and later modified by Ballard et al. (1979) have been widely adopted. Both methods appear to have limited accuracy at the extremes of gestation. Some (Tergestina et al., 2021) but not all (Stevenson et al., 2021) investigators advocate foot length, measured post­natally (< 48h) with calipers, as useful in resource-limited settings. To date, the measure­ment of gesta­tional age post­natally where ultra­sound is not possible continues to be problem­atic (Stevenson et al., 2021).

10.1.3 Recum­bent length

For infants and chil­dren < 2y (i.e., < 85cm), recum­bent length is measured, preferably using an infanto­meter with a range of 30–110cm, and preferably equipped with an elec­tronic digital reader. Recum­bent length should be recorded to the nearest millimeter, or even more precisely (i.e., 0.1mm) when possible. Wooden or perspex length measuring boards (Figure 10.2)
Figure 10.2
Figure 10.2. Device for the mea­sure­ment of recumbent length. Available from: UNICEF.ORG.
can also be used, although they are rarely fitted with digital counters so are less reliable. Note that recum­bent length for a child of about 2y is approximately 5mm greater on average than standing height for the same child (Haschke & van't Hof, 2000).

Two examiners are required to correctly position the child and ensure accurate and reliable measure­ments of length. Prior to the measure­ment, braids must be undone and any hair ornaments removed. Diapers must also be removed so that both of the child's legs can be outstretched in the correct position. The subject is placed face upward, with the head toward the fixed end of the board and the body parallel to the board's axis. The shoulders should rest against the surface of the board. One examiner applies gentle traction to bring the crown of the child's head into contact with the fixed headboard and positions the head so that the Frankfurt plane is vertical (Figure 10.3).
Figure 10.3
Figure 10.3. Measurement of recumbent length. Redrawn from Robbins GE, Trowbridge FL, in: Nutrition Assess­ment: A Comprehensive Guide for Planning Intervention by M.D. Simko, C. Cowell, and J.A. Gilbride (eds.), p.75, Aspen Publishers, Inc., 1984.
The second examiner holds the subject's feet, without shoes, toes pointing directly upward, and keeping the subject's knees straight, brings the movable footboard to rest firmly against the heels. To ensure the soles of the feet are flat on the footboard, the examiner should run the tip of their finger down the inside of the child's foot. The reading is taken to the nearest millimeter. If the subject is restless, only the left leg should be positioned for the measure­ment. To encour­age and comfort the child during the measure­ment, the parent or guardian should stand between the examiner and recorder so that they can make eye contact and talk to the child during the measure­ment procedure.

10.1.4 Standing height

Chil­dren aged ≥ 2y and adults should be measured in the standing position, if possible (WHO, 1995a), using a free-standing stadio­meter (range 65–206cm) with a fixed vertical back­board and an adjustable head piece, preferably equipped with a digital reader capable of measuring stature to 0.1mm.

Clothing should be minimal when measuring height so that posture can be clearly seen. Shoes and socks should not be worn, hair ornaments should be removed, and braids undone. The subject is asked to stand straight with the head in the Frankfurt plane (Figure 10.4), knees straight, with heels together and toes apart pointing slightly out­ward at a 60° angle. The back of the head, shoulder blades, but­tocks, and heels must be in contact with the vertical back­board of the sta­diome­ter. Arms should be hanging loosely at the sides with palms facing the thighs. For younger subjects, it may be necessary to hold the heels to ensure they do not leave the ground.
Figure 10.4
Figure 10.4. Positioning of subject for height mea­sure­ment. The horizontal line is the Frankfurt plane, which should be in a horizontal position when height is measured. Reproduced from Robbins GE, Trowbridge FL, in: Nutrition Assess­ment: A Comprehensive Guide for Planning Intervention by M.D. Simko, C. Cowell, and J.A. Gilbride (eds.), p.77, Aspen Publishers, Inc., 1984.

Some investigators recom­mend applying gentle upward pressure to the mastoid processes to stretch the spine and minimize effects produced by diurnal variation (Tanner et al., 1966). Subjects are asked to take a deep breath and stand tall to aid the straight­ening of the spine. Shoulders should be relaxed. The movable headboard is then gently lowered until it touches the crown of the head. The height measure­ment is taken at maximum inspiration, with the examiner's eyes level with the headboard to avoid parallax errors. Height is recorded to the nearest millimeter, or even more precisely with more modern digital equipment.

The time at which the measure­ment is made should be recorded; diurnal variations in height occur due to compression of the spine as the day progresses. Conse­quently, for cross-sectional and longitu­dinal studies, heights should be measured at the same time of day, preferably in the afternoon. In cases where large amounts of adipose tissue prevent the heels, buttocks, and shoulders from simul­taneously touching the wall, subjects should simply be asked to stand erect with the head in the Frankfurt plane.

In the field, vertical surfaces are not always available. In such circum­stances, modified tape measures such as the Microtoise, which measure up to 2m, can be used. To use the Microtoise, it is first placed on the floor, after which the tape is pulled out to its fullest extent and released, and the end is fixed with a nail to a door or doorway. The subject is then instructed to stand erect directly below the point of attachment. An anthro­pometrist should position the subject's head correctly in the Frankfurt plane, as described in Section 10.1.1, before the tape is lowered by a second person until the head-bar touches the crown of the head and compresses the hair. A direct reading of height to the nearest millimeter may then be obtained.

Recum­bent length for a child of about 2y is approximately 5mm greater on average than standing height for the same child (Haschke & van't Hof, 2000). Hence, if standing height rather than recum­bent length is measured, 5mm must be added to the standing height value when recum­bent length refer­ence data are used. In general self-reported heights tend to produce slightly higher estimates of measured height, the magnitude of the discrepancy varying depending on age, race/ethnicity (Ekström et al., 2015; Hodge et al., 2020).

When measuring recum­bent length or standing height, attempts should be made to minimize measure­ment errors. In longitu­dinal studies involving sequential measure­ments on the same group of indi­viduals, it is preferable when­ever possible, to have one person carrying out the measure­ments throughout the study to eliminate between-examiner errors. This is especially critical when growth velocity is calculated; growth increments are generally small and are asso­ciated with two error terms, one on each measure­ment occasion. Recommen­dations of the minimal intervals necessary to provide reliable data on growth increments during infancy and early childhood (de Onis et al., 2004) and adoles­cence are available. In the WHO Multi­centre growth refer­ence study (MGRS) for example, length was measured every two weeks from aged 2–6 weeks, monthly for ages 2–12 months, and every two months for toddlers aged 14–24 months. Increments measured over 6 months are the minimum interval that can be used to provide reliable data during adoles­cence. For shorter intervals, the combined errors may be too large in relation to the expected mean increment. Length or height velocity is usually expressed as cm/y.

In large regional surveys, it is often necessary to use several well-trained anthro­pometrists. In such circum­stances, the anthro­pometrists should be rotated among the subjects to reduce the effect of measure­ment bias of the indi­vidual examiners. Further, regular standard­ization sessions with the assess­ment of both within- and between-examiner reliability throughout the data collection should be conducted to maintain the quality of the measure­ments, and identify and correct systematic errors in the measure­ments; details of the procedures used in the WHO MGRS are given in de Onis et al. (2004). The WHO MGRS recom­mends that the maximum allowable difference in length for acceptable precision between measure­ments by two examiners is 7.0mm. This figure is based on the technical error of the mea­sure­ment (TEM) obtained in the initial standard­ization session conducted at the Brazil site (de Onis et al. 2004). Details of the measure­ment tech­niques and standard­ization protocols for both recum­bent length and stature are also available in an anthro­pometric training video prepared for the WHO MGRS and available on request from de Onis et al. (2004). Statistical methods exist for removing anthro­pometric measure­ment errors from cross-sectional anthro­pometric data; details are given in Ulijaszek and Lourie (1994).

10.1.5 Knee height in chil­dren aged over 3y

The measure­ment of knee height in chil­dren is termed “knemometry”. The measure­ment is taken from the distance between the heel and knee of the right leg (i.e. the lower leg length) when the child is sitting. At least three and preferably six measure­ments are needed for an accurate deter­mination of the lower leg length in a child (Hermanussen et al., 1988). A training period of several weeks is required prior to routine measuring of lower leg length in chil­dren > 3y (Ahmed et al., 1995).

The main application of knemometry is for physically disabled chil­dren and in certain pediatric units specializing in growth disorders. Knemometry can also be used to assess the effect of thera­peutic interventions (e.g., steroid therapy etc) on short-term growth (Gradman & Wolthers, 2010; Battin et al., 2012). By measuring knee height, growth increments in chil­dren are said to be detected more readily and over a shorter time frame than by conven­tional height measure­ments. Moreover, knee height measure­ments can be made with greater precision.

Several factors other than growth influence the measure­ment of lower leg length, and hence must be controlled. For example, as diurnal variations influence the measure­ment, it is preferable for all measure­ments to be performed during the afternoon, and by one trained operator. Before the measure­ments, chil­dren should avoid vigorous physical activity for at least 2h, and instead, stand or walk slowly for 5–10min.

To measure lower leg length in chil­dren > 3y, the chil­dren must be able to sit quietly and co-operate. The first knemometer for measuring chil­dren > 3y was devel­oped by Valk in 1971, and modified in 1983 to improve its accuracy. Portable devices, termed a knee height measuring devices (KHMDs), are now also available for measuring changes in lower leg length in chil­dren > 3y (Cronk et al., 1989). Extremely low TEMs have been reported for the original Valk knemometer (0.09 and 0.16) (Hermanussen et al., 1988), whereas the newer, less costly portable knemometers, have a slightly poorer performance (Cronk et al., 1989). For chil­dren < 3y, a mini-knemometer can be used to measure lower leg length (Section 10.1.6).

Measurement using a knemometer

Figure 10.5
Figure 10.5. Diagram of the knemometer. Movable chair with adjustable height (A) and back (B); foot plate (C); Measuring board (D); Counterweight (E); Visual elec­tronic display (F). The coordinates of the child's sitting position are defined by: the sitting height (H); the distance between the chair and the measuring board (M); and the distance between the lateral condyle (marked “X”) and the back of the chair (P). Redrawn from Hermanussen et al., Annals of Human Biology 15: 1–16, 1988.

(Figure 10.5) depicts the tech­nique of lower leg length measure­ment using a modified Valk knemometer. To take the measure­ment, the child is asked to sit on the chair. The sitting height (A) and the chair back (B) and chair position can be adjusted. The foot of the child is placed on the foot rest (C) and the angle and the length of the foot are recorded. The sitting position of the child is stan­dard­ized by recording the following:

Next, the measuring board (D) is lowered onto the child's knee, and then the chair on which the child is sitting is gently moved forward and back­ward. The knee is moved passively under the measuring board (D) and the foot plate (C), while the length is displayed continuously on an elec­tronic display. A counter­weight (E) ensures a constant pressure of the measuring board of about 200g.

The actual lower leg length is defined as the maximum distance that can be reached during the movements of the child's knee. The minimum distance that can be discriminated by this instrument is 0.1mm. Chil­dren should be lightly dressed when these measure­ments are taken.

Portable knee height measure­ment device (KHMD)

The portable instrument is also designed to measure short-term growth of the lower leg. This device is less costly and easier to use than the knemometer. Again, the measure­ments are taken while the child is sitting. The chair used with this device should have a seat height of 33cm and a seat length of 26cm.

The KHMD consists of a standard industrial elec­tronic height gauge (precision of 0.01mm) to measure the distance between the knee and foot plate. A plexiglass knee plate is mounted on the height gauge. A dial indicator is attached to the height gauge, so accurate and repeatable force can be applied to the knee during each measure­ment. The reading on the dial indicator corresponds to the torque flexion of the knee plate during the measure­ment.

The height gauge with the knee plate attached is mounted on a base comprised of a foot plate, wheels, and a handgrip to move the device. Both heel and toe plates are mounted on the foot plate to ensure that the position of the foot is repro­ducible for each measure­ment. The knee is also fixed in position during the measure­ment by an adjustable knee brace. The knee brace position is recorded by the operator from a numerical scale attached to the upright support. This allows the operator to position the subject repro­ducibly at each measure­ment session.

The position of the chair relative to the KHMD must also be fixed for each subject at each measure­ment session. This is achieved by placing the chair on a plexiglass grid with numbered and lettered squares. This marked grid, together with the braces for the knee, foot, and toe, ensure that the position of the subject at each measure­ment session can be replicated (Cronk et al., 1989).

For the measure­ment, subjects should sit with buttocks and the small of the back firmly in contact with the chair back, with the foot on the foot plate flush with the heel brace. The top of the knee brace is then adjusted until it has contact with the nonboney space between the patella and the tibial tuberosity. The chair is then pushed forward so that the knee is in full contact with the knee brace, with the leg centered in relation to the knee plate. The opposite foot is positioned so it is flush with the instrument and parallel with the foot plate. Next, the knee plate is lowered onto the top of the knee until it is deflected 0.03mm (equivalent to about 2kg of pressure on the knee which yields the smallest degree of error without causing discomfort). The knee height measure­ment is elec­tronically recorded. Subjects should remain seated between measure­ments but remove their foot from the foot plate and then return it again. The position of the child in the chair and the foot on the foot plate should be readjusted, and three additional indepen­dent measure­ments should be taken.

The performance of the KHMD has been compared with that of the Valk knemometer on 103 chil­dren 6–10y. Growth was measured at 28d intervals with both devices. The within-examiner error for the measure­ments was 0.295mm for the KHMD and 0.206mm for the knemometer. The correlation between devices for measure­ments of growth rate on each child was 0.73 (p< 0.002) (Cronk et al., 1989).

Knee-height appears to be a valid index of linear growth in well-nourished chil­dren. For example, in a study of healthy Indonesian chil­dren aged 7–12y, two equations to predict height based on knee height for boys and girls have been devel­oped (Rumapea et al., 2021). However, whether knee height is a valid index of linear growth in chil­dren recovering from severe malnu­trition is uncertain (Doherty et al., 2001).

10.1.6 Lower leg length in infants and chil­dren aged < 3y

Lower leg length (i.e., knee-heel length) is also used to measure short term changes in linear growth over 1–8wks during infancy and early childhood. For preterm infants, normal infants, and toddlers < 3y, a hand-held elec­tronic knemometer (Michaelsen et al., 1991), a mini-knemometer (Hermanussen & Seele, 1997), or an inexpensive vernier or elec­tronic caliper (Skinner et al., 1997; Engström et al., 2003) can be used. Reports for the TEM for these three instruments vary across studies, in part depending on whether the readings are recorded blinded or not, and the number of consecutive readings taken (Hermanussen & Seele, 1997; Skinner et al., 1997; Engström et al., 2003). In general, the inexpensive electronic caliper has a lower TEM than the handheld mini-knemometer (Figure 10.6) and could be used when assessing lower leg length in preterm infants over a short time period (Engström et al., 2003).

The mini-knemometer contains a commercially available elec­tronic slide that discriminates intervals of 10µm. The slide is connected to two measuring arms (A, B) with metallic holders, as shown in (Figure 10.6). The infant's knee and heel are placed between the holders. Knee and heel holders of different sizes can be fitted, depending on the age of the infant.
Figure 10.6
Figure 10.6. Diagram of a mini-knemometer: measuring arms (A and B); metallic holders for knee and heel (C and D). The arms are spring-loaded so that a constant pressure is applied during mea­sure­ment. Redrawn from Hermanussen and Seele, Annals of Human Biology 24: 307–313, 1997.
When serial measure­ments are being taken using mini-knemometry, care must be observed to ensure that the infant is always measured in the same position. A spring mounted within the instrument between the arms ensures a constant soft tissue pressure between 2.0 and 3.0N. The measure­ment is painless for the infant and is best made during breast-feeding. Measure­ments should be performed without refer­ence to previous recordings. Four–to–five indepen­dent measure­ments of lower leg length should be taken on each child within 1–3min, and both the mean and standard deviation calculated. More readings are needed when the child is restless.

10.1.7 Knee height in adults

Knee height is highly correlated with stature and may be used to estimate height in persons with severe spinal curvature or who are unable to stand. Knee height is measured with a caliper consisting of an adjustable measuring stick with a blade attached to each end at a 90° angle.

Recum­bent knee height is measured on the left leg, which is bent at the knee at a 90° angle, while the subject is in a supine position (Figure 10.7). One of the blades is positioned under the heel of the left foot and the other is placed over the anterior surface of the left thigh above the condyles of the femur and just proximal to the patella. The shaft of the caliper is held parallel to the shaft of the tibia, and gentle pressure is applied to the blades of the caliper.
Figure 10.7
Figure 10.7. Measurement of knee height in adults.
Some of the knee-height calipers are equipped with a locking mech­anism to retain the measure­ment after removing the caliper from the leg. At least two successive measure­ments should be made, and they should agree within 5mm; the mean is then calculated. Details for measuring knee height in elderly persons seated in wheel chairs are given in (WHO, 1995a).

Formulae are used to estimate adult stature from knee height. Separate stature prediction equations using knee height and age were devel­oped for non-Hispanic white, non-Hispanic black, and Mexican-American elderly persons based on the NHANES III data (Chumlea et al., 1998). However, their appropriateness for estimating stature among other ethnic groups has been questioned. Silva de Lima et al. (2018) compared the validity of 16 equations to estimate height based on knee-height in elderly nursing home residents (n=168) in Brazil. None of the equations examined were applicable for the estimate of height of indi­viduals > 70y, emphasizing that popu­lation-specific equations may be necessary. Conse­quently, an assess­ment of stature based on knee height should be used only for indi­viduals for whom a direct measure­ment of stature is not possible, or is likely to be inaccurate, because of vertebral flexions or other skeletal deformities.

10.1.8 Arm span and demi-span

Arm span, like knee height, is also highly correlated with stature and, hence, can be used as an alter­native measure­ment when actual height cannot be used such as in elderly persons when degen­erative and osteoporotic changes give rise to spine curvature (Goswami et al., 2018). Arm span is also often used as an alternative to height for the calculation of body mass index (BMI) in older adults (Arlappa et al., 2016), and as a reliable surrogate of both recum­bent length and height in healthy chil­dren, when these measure­ments are unobtainable or unreliable (Forman et al., 2014).

The measure­ment of arm span is easier if carried out against a flat wall (Figure 10.8), to which is attached a fixed marker board at the zero end of a horizontal scale. Sliding on the scale is a vertical movable arm. The horizontal scale should be positioned so that it is just above the shoulders of the subject. Two examiners are needed to measure arm span: one is at the fixed end of the scale; the other positions the movable arm and takes the readings. In elderly persons, an assistant may be needed to support and maintain the arm being assessed in a 90° position (Silva de Lima et al., 2018).
Figure 10.8
Figure 10.8. Measurement of arm span.

For the measure­ment, the indi­vidual should stand with feet together, back against the wall, with the arms extended laterally in contact with the wall, and with the palms facing forward. The arms should be kept at shoulder height and outstretched maximally. The measure­ment is taken when the tip of the middle finger (excluding the fingernail) of the right hand is kept in contact with the fixed marker board, while the movable arm is set at the tip of the middle finger (excluding the fingernail) of the left hand. Two readings are taken for each measure­ment, which is recorded to the nearest 1.0mm (Lohman et al., 1988).

Arm span is difficult to measure in non-ambulatory elderly persons and in indi­viduals with significant chest and spinal deformities and stiffness. (Silva de Lima et al., 2018). Zhang et al. (1998) concluded that in a group of elderly Chinese, for example, knee height provided a more valid estimate of maximum stature during early adulthood than arm span.

Instead of arm span, demi-span is some­times used to assess body mass index in older adults when an actual height measure­ment is not possible. Demi-span is the distance between the mid-point of the sternal notch (or jugular notch) — a large visible dip in the neck in humans between the clavicles — and the finger tips, with the right arm outstretched laterally. The measure­ment is performed using a retractable metal tape and taken to the nearest mm. Again, in some elderly persons, an assistant may be needed to support and maintain the arm being assessed in a 90° position (Silva de Lima et al., 2018).

Demi-span measure­ments were included in the Health Survey for England (HSE) because they can be measured easily without causing discomfort or distress. New demi-span sex- and age-specific regression equations for estimating adult height have been devel­oped by Hirani and Aresu (2012) and are shown below (standard errors are given in parentheses): \[ \small \mbox{Men: DEH}_{\mbox{age}}\mbox{(cm) = 73.0 + 1.30 (0.04) × demi-span − 0.10 × age (0.02)}\]

\[ \small \mbox{Women: DEH}_{\mbox{age}}\mbox{(cm) = 85.7 + 1.12 (0.05) × demi-span − 0.15 × age (0.05)}\] These equations are based on data for adults ≥ 65y who participated in the HSE. However, as 98% of the sample were white, whether the proposed equations are valid for other ethnic groups is unknown.

10.1.9 Weight in infants and chil­dren

In field surveys, a suspended scale and a weighing sling may be used for weighing infants and chil­dren < 2y (Figure 10.9). They should be weighed naked or with the minimum of clothing. After slipping the subject into the sling, the weight is recorded as soon as the indicator on the scale has stabilized.
Figure 10.9
Figure 10.9. Measurement of weight in infants and chil­dren using a spring balance. The bowl for the baby is made from a metal or bamboo ring and netting. See comparable devices available from: UNICEF.ORG.

Alterna­tively, for greater precision, a pediatric scale (within 10g) with a pan may be used (Figure 10.10A). Care must be taken to ensure that the infant (preferably nude) is placed on the pan scale so the weight is dis­trib­uted equally on each side of the center of the pan. Once the infant is lying quietly, weight is recorded to the nearest 10g. In cold weather, the infant can be wrapped in a blanket of known weight, which is subtracted after­wards to obtain the weight of the child nude. If the scale has a taring capacity, then the weight of the blanket can be tared before placing the infant on the pan scale.

If there is no alternative, the mother can be weighed alone, and then again holding with the child, using battery-operated precision digital elec­tronic weight scales. The child's weight can then be calculated by subtraction. Alterna­tively, if the scale has a taring function, then the weight of the mother can be set so the scale readout is zero, enabling only the child's weight to be recorded. If the child cannot be undressed, then standard light clothing of known weight should be worn. This clothing weight should be subtracted after­wards from the child's weight.

10.1.10 Weight in older chil­dren and adults

The measure­ment of weight in older chil­dren and adults should be done preferably after the bladder has been emptied and before a meal. A portable digital elec­tronic weight scale, preferably one that has a taring capacity, and has been cali­brated to 0.1kg, should be used (Figure 10.10B).

Figure 10.10B
Figure 10.10. Measurement of weight: (A) Digital pediatric scale for infants, (B) Digital balance for a child or adult. Available from: UNICEF.ORG.
The weight scale should be placed on a hard, flat surface (not carpet) and checked and adjusted for zero-balance before each measure­ment. The subject should stand in the center of the platform and look straight ahead, standing unassisted, relaxed but still, and preferably nude. If nudity is not possible, the subject can wear light underclothing or a paper examination gown, and the weight of these garments should be recorded for later subtraction; standard corrections for clothing should not be used. The presence of visible edema should also be recorded. Body weight should be recorded to the nearest 0.1kg. Again, the time at which the measure­ment is made should be recorded because diurnal variations in weight occur.

Elec­tronic weight scales should be cali­brated with a set of standard weights over the full weight range, both regularly throughout the year and when­ever they are moved to another location. Special equipment, such as a movable wheelchair balance beam scale, bed scales, or specialized beds with integrated weighing scales (Mishra et al., 2021) is needed for weighing non-ambulatory persons (Chumlea et al., 1984).

Estimates of weight for the U.S. elderly popu­lation can be derived from calf circum­ference (calf circ), knee height (knee ht), mid-upper-arm circum­ference (MUAC), and subscapular skinfold (subscap), using equations devel­oped by Chumlea et al. (1989); examples are given below: \[ \small \mbox{Weight (M) = (0.98×calf circ) + (1.16×knee ht) + (1.73×MUAC) + (0.37×subscap) - 81.69 }\]

\[ \small \mbox{Weight (F) = (1.27×calf circ) + (0.87×knee ht) + (0.98×MUAC) + (0.40×subscap) - 62.35 }\] The above equations were devel­oped from a selected popu­lation living in the United States and hence are inappropriate for estimating weights of other popu­lations. Instead, popu­lation-specific equations may be required. Quiroz-Olguin et al. (2013) have devel­oped and validated formulae for predicting body weight using circum­ference-based equations for Mexican adults.

Estimates of weight among multi-racial/ethnic infants and chil­dren 0–5.9y in the United States based on ulna length and forearm width and circum­ference using simple and portable tools have been devel­oped. More investigation of their validity for physically impaired or non-ambulatory chil­dren is needed (Zhu et al., 2019).

Use of self-reported weights in adoles­cents and adults may lead to bias and should be avoided (Ekström et al., 2015; Pérez et al., 2015). In the U.S. Women's Health Initiative, on average women under-reported their weight by about 0.91kg although the discrepancies varied by age, race / ethnicity, education, and body mass index (Luo et al., 2019).

10.1.11 Elbow breadth

Elbow breadth is a good measure of skeletal dimensions and, hence, frame size. The measure is less affected by adiposity than many other anthro­pometric dimensions and is highly asso­ciated with lean body mass and muscle size (Frisancho, 1990).

Figure 10.11
Figure 10.11. Measurement of elbow breadth.
Elbow breadth is measured as the distance between the epicondyles of the humerus. For the measure­ment, the right arm is raised to the horizontal and the elbow is flexed to 90°, with the back of the hand facing the measurer (Figure 10.11). The measurer then stands in front of the subject and locates the lateral and medial epicondyles of the humerus. The two blades of a flat-bladed sliding caliper are applied to the epicondyles, with the blades pointing upward to bisect the right angle formed at the elbow. Care must be taken to ensure that the caliper is held at a slight angle to the epicondyles and that firm pressure is exerted to minimize the influence of soft tissue on the measure­ment. The latter is taken to the nearest millimeter (Lohman et al., 1988). Elbow breadth is used to calculate the Frame Index: \[ \small \mbox{Frame Index = elbow breadth (mm) × 100 / height (cm)}\]

This index is used as a measure­ment of external skeletal robustness in current and past popu­lations (Frisancho, 1990). Studies have confirmed a decline in the Frame Index in recent decades among young chil­dren in Germany (Scheffler & Hermanussen, 2014), Russia (Rietsch et al., 2013a), and Argentina (Navazo et al., 2020), a trend that appears to parallel an increase in over­weight and obesity. This decline in Frame Index has been asso­ciated with a reduction in physical activity, also known to decrease external skeletal robustness (Rietsch et al., 2013b). In 2018, refer­ence per­cent­iles for Frame Index were published for European chil­dren and adoles­cents (Mumm et al., 2018).

10.2 Growth indices, indicators, and recom­mended growth refer­ence data

The correct inter­pretation and grouping of anthro­pometric measure­ments require the use of anthro­pometric indices (WHO, 1995a). They are usually calculated from two or more raw anthro­pometric measure­ments. In the simplest case the indices are numerical ratios such as wt/ht2 (kg/m2). Combi­nations such as weight-for-age, length / height-for-age (i.e., stature-for-age), and weight-for-stature are more complex. These latter growth indices are not ratios and, to avoid confusion with numerical ratios, should not be written as “wt/age”, “ht/age”, and “wt/height”.

Anthro­pometric indices are often evaluated by comparison with the distri­bution of appropriate refer­ence data using standard deviation scores (Z‑scores) or per­cent­iles. From this, the number and proportion of indi­viduals (as %) with anthro­pometric indices below or above a pre­deter­mined refer­ence limit or cutoff are often calculated. A commonly used statis­tically defined refer­ence limit for the three main growth indices is a Z‑score of −2 (i.e., 2SD below the WHO growth refer­ence median). When used in this way, the index and its asso­ciated refer­ence limit become an “indicator”. Growth indicators are often used for public health or socio-medical decision making at the popu­lation level. They are also used in clinical settings to identify indi­viduals at risk of malnu­trition. Examples of frequently used anthro­pometric growth indicators and their corresponding application are shown in Table 10.1.
Table 10.1 Anthro­pometric indicators and their corres­ponding applications.
Anthro­pometric indicator Application
Proportion of chil­dren (of defined
     age and sex) with WHZ < −3
Preva­lence of severe wasting
Proportion of chil­dren (of defined
     age and sex) with WHZ < −2
Preva­lence of wasting
Proportion of chil­dren (of defined
     age and sex) with WHZ > +2
Preva­lence of over­weight
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 of under­weight
Proportion of children (of defined
     age and sex) with BMIZ +1 to +2
Prevalence of “at risk of over-
     weight”(for those 0–5y)
Proportion of children (of defined
     age and sex) with BMIZ > +2
Prevalence of over­weight
     (for those aged 0–5y)
Proportion of children (of defined
     age and sex) with BMIZ > +3
Prevalence of obesity
     (for those aged 0–5y)

The WHO recom­mends the use of the WHO Child Growth Standards for young chil­dren from birth to 5y for inter­national use (Growth Standards), given the small effect of ethnic and genetic differ­ences on the growth of infants and young chil­dren compared with the environ­mental, nutritional, and socio-economic effects, some of which may persist across gener­ations. These new growth standards were devel­oped because of the tech­nical and bio­logical limit­ations iden­tified with the earlier NCHS/WHO growth refer­ence; for more details, see Garza & de Onis, (1999).

A prescriptive approach depicting normal early childhood growth under optimal environ­mental conditions was used for the new WHO Child Growth Standards so they represent how young chil­dren should grow, rather than as a “refer­ence” describing how chil­dren do grow. For details of the methods used in the compilation of the new WHO Child Growth Standards, see de Onis et al. (2004).

WHO has devel­oped a tool for the application of the WHO Child Growth Standards which includes instructions on how to take the measure­ments, interpret growth indicators, investigate causes of growth problems, and how to counsel caregivers. For more details see: WHO Child Growth Training Module (Training Module).

For boys and girls charts (0–6mos; 0–2y; 6mos–2y; 0–5y) for head circum­ference-for-age, length/ height-for-age, weight-for-age, weight-for-length (45–110cm), weight-for-height (65–120cm), and body mass index (BMI)-for-age are available, expressed as per­cent­iles or Z‑scores. Data tables for Z‑scores and per­cent­iles can also be downloaded from WHO (Growth Standards).

The WHO growth refer­ence for school-age chil­dren and adoles­cents 5–19y (or 61–228mos) is recom­mended for inter­national use (de Onis et al., 2007). This is based on a reconstruction of the original U.S. National Centre for Health Statistics (NCHS) data set supple­mented with data for under-fives from the WHO Child Growth Standard. The statistical methodology used to construct this refer­ence was the same as that used for the WHO Child Growth Standard. For boys and girls charts for height-for-age (5–19y), BMI-for-age (5–19y) and weight-for-age (5–10y) are available, expressed as Z‑scores or per­cent­iles. Data tables for Z‑scores and per­cent­iles are also available (Growth Reference Data).

Concern over the impact of using the growth charts based on the new WHO Child Growth Standard compared to the earlier NCHS/WHO inter­national growth charts on the estimates of malnu­trition for chil­dren from birth to age 5y prompted investigations by several researchers (de Onis et al., 2006; Martorell & Young, 2012). Such information is impor­tant because the resulting pattern of child growth, following an intervention designed to improve child health, depends on the growth charts used. Based on the growth indices outlined above, de Onis and co-workers (2006) compared the rates of under­weight, stunting, wasting, severe wasting, and over­weight generated from the two sets of growth charts using two cross-sectional datasets for chil­dren 0–5y representative of different nutritional status profiles from Bangladesh and the Dominican Republic (for over­weight only). Impor­tant differ­ences were revealed between the WHO standard and the NCHS refer­ence that varied by age group, growth indicator, and the nutritional status of the popu­lations examined. These same differ­ences in the age-related patterns of stunting and wasting have also been reported between the two growth refer­ence data when compared using nationally representative data from India and Guatemala (Martorell & Young, 2012).

In general, for popu­lations with nutritional status profiles similar to those of Bangladesh, rates of under­weight (Figure 10.12) during the first half of infancy, and stunting (Figure 10.13) at all ages, will be higher when based on the WHO standard compared with those of the NCHS refer­ence, whereas after the first six months, rates for under­weight will be lower.
Figure 10.1213 Figure 10.1415
Figs 10.12 – 10.15. Prevalence of under­weight, stunting, wasting, and over­weight, showing the differences between the WHO and NCHS standards. From: de Onis et al. (2006).
For wasting (i.e., weight-for-stature Z‑core < −2), the rates will be greater at all ages, with the difference being especially striking during early infancy (0–6mos), as shown in Figure 10.14. A similar pattern exists for severe wasting, (i.e., weight-for-length/ height Z‑score < −3) and applied as the criterion for enrolling chil­dren in thera­peutic feeding programs (WHO/UNICEF, 2009).

For over­weight (i.e., weight-for-stature > +2 Z‑scores), the expected difference in over­weight for chil­dren 0–60mos assessed by the WHO standard will be an increase in the rates of over­weight as shown in Figure 10.15, although the pattern varies by age group, and will also vary with the nutritional status of the popu­lation under study. Differ­ences in over­weight defined by BMI-for-age Z‑scores could not be evaluated because BMI-for-age data are not available for preschool chil­dren in the older NCHS growth refer­ence.

The reported differ­ences in the age patterns of growth between the two growth charts, partic­ularly during infancy, are likely because the WHO standard represents the growth of only term breast-fed infants who have greater weights for length than the predominantly formula-fed Caucasian infants of the NCHS refer­ence, although they do become thinner during the second year and beyond (de Onis et al., 2006). An additional factor is the shorter intervals between measure­ments during the periods of rapid growth for the WHO standards compared to the NCHS refer­ence (de Onis et al., 2006).

The U.S. have also released updated CDC 2000 childhood (growth charts), based primarily on physical measure­ments taken during five nationally representative surveys conducted between 1963 and 1994, although some supple­mental data were also used. During the creation of these revised CDC 2000 Growth Charts, two data sets were excluded: growth data for very low birth­weight (VLBW) infants (< 1500g) whose growth differs from that of normal birth­weight infants, and weight data for chil­dren > 6y who participated in the NHANES III survey. The latter data were excluded from both the revised weight and BMI growth charts because the inclusion of these data shifted the upper per­centile curves. With the exclusion of these selected data, a modified growth refer­ence that is not a purely descriptive growth refer­ence was created because it does not contain representative national data for all vari­ables (Kuczmarski et al., 2002). Users can compute exact per­cent­iles, Z‑scores, and BMI values (from weight and data on length or height) using WHO Anthro software and the CDC refer­ence. For more details, see Kuczmarski et al. (2002). A comparison of these CDC 2000 Growth Charts with the WHO Child Growth Standards is available in de Onis et al. (2006).

A series of prescriptive standards are now available for monitoring fetal, new­born growth, and gesta­tional weight gain for inter­national use devel­oped by the INTER­GROWTH-21st Project. As noted earlier, this project adhered to the WHO recommen­dations for assessing human size and growth and followed healthy pregnant women longitu­dinally from 9wks of fetal life to 2y (Papageorghiou et al., 2018). Box 10.1. lists the prescriptive standards generated from the INTER­GROWTH-21st project with the asso­ciated refer­ences.
Box 10.1. INTER­GROWTH-21st inter­national standards for monitoring growth and develop­ment from early pregnancy to 2y.

Of the indicators listed in Table 10.1, three are also included as components of the six Global Nutrition Targets for 2030 set by WHO/UNICEF and shown in Box 10.2.
Box 10.2 Global nutrition targets for 2030 From WHO/UNICEF (2021),

Policy briefs provide details of the extensions of each of the 2025 Global Nutrition Targets. WHO has also devel­oped a web-based tracking tool to assist countries to set national targets and chart progress for achieving the six global targets (See Section 10.2.6).

Additional factors that must be considered when selecting an index or combi­nation of indices to eval­uate growth include the avail­ability of accurate measuring equipment, the training of examiners to collect accurate inform­ation and to interpret the results correctly, and the time required to take the measure­ments. Finally, often over­looked are the costs of not ident­ifying under­nourished chil­dren or incor­rectly ident­ifying adequately nourished chil­dren as under­nourished (Gorstein et al., 1994).

Details of the growth indices, the indicators derived from them and their asso­ciated applications, together with their advantages and limitations, are discussed below. Information on the available interpretive criteria for each growth index is also included. .

10.2.1 Head circum­ference-for-age

Intrauterine growth retardation, or chronic malnu­trition during the first few months of life, may hinder brain develop­ment and result in an abnormally low head circum­ference. Hence, head circum­ference is a widely used proxy of neural growth and brain size. When brain size is outside of normal values, it is an impor­tant risk factor for cognitive and motor delay.

Head circum­ference-for-age can be used as an index of chronic malnu­trition for chil­dren < 2y but is not sensitive to less extreme malnu­trition (Yarbrough et al., 1974). Beyond age 2y, growth in head circum­ference is slow and its measure­ment is no longer useful, so an indicator based on head circum­ference-for-age is not included in Table 10.1 (Harris, 2015). Certain non-nutritional factors, including some diseases and pathological conditions (e.g., micro­cephaly), genetic variation, and cultural practices such as binding of the head during infancy, as well as a difficult or forceps-assisted delivery at birth, may also influence head circum­ference.

Interpretive criteria

Microcephaly in an infant is defined as a measure­ment of head circum­ference that is more than 2SD below the mean of an age- and sex-appropriate growth chart, whereas in severe micro­cephaly, head circum­ference is more than 3SD below the mean (Harris, 2015). For inter­national use, the WHO Child Growth Standards that include head circum­ference by age and sex from birth to 13wks, birth to 2y, and birth to 5y are recom­mended (Growth Standards). These standards, unlike the INTER­GROWTH-21st standards described below, only include term new­born infants.

Papageorghiou et al. (2014) have produced international standards for fetal growth in which head circumference was measured using ultra­sound from 14wks to 42wks gestation. The 3rd, 5th, 10th, 50th, 90th, 95th, and 97th smoothed per­centile curves for fetal head circumference are available.

Inter­national standards for new­born head circum­ference by gesta­tional age (33–42wks) and sex are also available from the INTER­GROWTH-21st Project. Pregnancies of all women who met strict eligibility criteria for a popu­lation at low risk of impaired fetal growth were selected and followed pro­spec­tively. Gestational age was estimated using ultra­sound. Hence, these multi-ethnic growth standards represent how fetuses should grow (i.e., the standards are prescriptive) and can be used to diagnose fetal growth restrictions world-wide and allow comparisons of new­born size across multi-ethnic popu­lations (Villar et al., 2014). Free software is available through the INTER­GROWTH-21st Project website to calculate Z‑scores and centiles.

10.2.2 Weight-for-age

Weight-for-age reflects body mass relative to chronological age. Low weight-for-age is described as “lightness” and reflects a pathological process referred to as “under­weight”, arising from gaining insufficient weight relative to age, or losing weight (WHO, 1995a). Because of its simplicity and the availability of scales in most health centers in low- and middle-income countries, weight-for-age is widely used in chil­dren from 6mo to 7y to assess under­weight.

Under­weight is defined by the indicator weight-for-age < −2 Z‑score (relative to the WHO Child Growth Standard) and was selected as one of the indicators to track the progress in addressing poverty and hunger for the UN Millennium Develop­ment Goals (MDGs). These are now replaced by the UN Sustainable Develop­ment Goals (SDGs) (2015–2030), which include the indicators to monitor progress shown in Box 10.2.

A major limitation of weight-for-age is that it is influenced by both the height and weight of a child, making inter­pretation difficult. For example, weight-for-age fails to distin­guish tall thin chil­dren (i.e., those with low weight-for-height) from those who are short (i.e., low height-for-age) but with adequate weight. As a result, the use of weight-for-age alone to estimate the preva­lence of under­nutrition leads to a gross underestimate of the problem in popu­lations where the preva­lence of low height-for-age (i.e., linear growth retardation) is high but that of low weight-for-height (i.e., wasting) is low (e.g., Guatemala) (WHO, 1995a). Conversely, in countries undergoing the nutrition transition and experiencing progressive increases in childhood over­weight and obesity, use of weight-for-age alone will result in overstating their progress in reducing under­weight and mask stunting (Uauy et al., 2008).

To interpret any single measure­ment of weight (or height) in relation to the refer­ence data, the exact age of the child at the date of the measure­ment must be calculated from the date of birth. Software, such as WHO AnthroPlus, can calculate exact ages in decimal fractions of a year, from birth and visit dates (AnthroPlus 2009). Details of WHO Anthro Growth Standards are given here (Growth Standards).

Even when information on the date of birth is available, ages are some­times reported following rounding off the most recently attained whole month. This practice should not be followed, because it results in systematic errors (Gorstein et al., 1989). In the event that documentary evidence of the date of birth is not available, it may be necessary to obtain at least the month and year of birth using a local events calendar. Information on the develop­ment of a local events calendar is available (FAO, 2008). Details of other methods that can be used to assess the age of chil­dren are given in Chapter 9.

Birth­weight is also measured in health centers in low-income countries as well as maternity hospitals because it is an impor­tant indicator of fetal and neonatal health. Low birth­weight (LBW) is defined as < 2500g at birth, and very low birth­weight as < 1500g. South Asia is a region where 25% of births are LBW, which is of concern given that LBW infants are at risk of poor health and develop­ment outcomes. Conse­quently, as noted earlier, one of the Global Nutrition Targets is to achieve a 30% reduction in low birth­weight by 2030 in recognition of the importance of LBW for survival, develop­ment and health in the lifespan (WHO, 2014).

Low birth­weight may be a consequence of premature birth (i.e., before 37wks gestation), being intra­uterine growth restricted (IUGR), or both. To assess fetal growth restriction, the weight of the new­born is compared to the achievement of the expected weight for a given gesta­tional age. Small for gesta­tional age (SGA), defined as being born below the 10th per­centile of a sex-specific birth­weight distri­bution at a specified gesta­tional age, is often used as a proxy to iden­tify IUGR neonates. Methods for estimating gesta­tional age have been described in Section 10.1.2.

Interpretive criteria

For inter­national use, data for weight-for-age, expressed as Z‑scores and per­cent­iles, are available for term infants and chil­dren (0–6mos; 0–2y; 6mo–2y; 2–5y; 0–5y) based on the WHO Child (Growth Standards), as noted earlier. These data only include term new­born infants (i.e., not gesta­tional age specific at birth), unlike the birth­weight data of the INTERGROWTH 21st Project (Villar et al. 2014).

Data for weight-for-age for chil­dren beyond age 10y are not available (de Onis et al., 2007) because beyond age 10y, chil­dren are experiencing the pubertal growth spurt and may appear to have excess weight based on weight-for-age when in fact they are just tall. The U.S. CDC 2000 growth refer­ence data, however, provide weight-for-age growth charts for boys and girls from birth to 36mos and 2–20y (Kuczmarski et al., 2002). As emphasized earlier (Section 10.2), weight data for children > 6y who participated in NHANES III survey were excluded from these revised CDC 2000 weight growth charts because the inclusion of these data shifted the upper per­centile curves.

Differ­ences in the prevalence of under­weight (i.e., weight-for-age Z‑score < −2) according to the WHO standard and the older NCHS refer­ence have been reported based on the same dataset, as noted in Section 10.2. Such differ­ences have affected the ranking of countries with respect to under­weight. For example, using the 1996–1997 DHS survey from Bangladesh, the preva­lence of under­weight was much higher (i.e., 2.5 times greater) during the first six months but lower thereafter when based on the WHO standard compared to the NCHS refer­ence (de Onis et al., 2006). As noted earlier (Section 10.2), this difference has arisen because the WHO standard is based on breast-fed infants whose growth pattern in infancy differs substantially from that of the predominantly formula-fed infants of the NCHS refer­ence (de Onis et al., 2006).

Inter­national standards for optimal fetal growth and new­born weight (and length and head circum­ference) by sex and gesta­tional-age (33–42wks) are available from the INTER­GROWTH-21st Project (Villar et al., 2014; Papageorghiou et al., 2014). Reliable estimates of gesta­tional age were obtained by ultra­sound measure­ments and new­born anthro­pometric measure­ments obtained within 12h of birth. These international fetal growth standards should be used worldwide to diagnose fetal growth restriction uniformly and monitor growth from early pregnancy through to the neonatal period (Papageorghiou et al., 2018), rather than locally-produced reference data. Use of these international fecal growth standards derived from a healthy population reduces the risk of under-diagnosing fetal growth restriction, which may occur when locally-produced reference data that includes high-risk mothers are used. The inter­national standard for weight by gesta­tional age and sex for new­born infants allows the accurate assess­ment of the preva­lence of SGA infants worldwide. The 3rd, 10th, 50th, 90th, and 97th smoothed per­centile curves and the numerical values for birthweight according to gestational age (33 to 43wks) are available (Villar et al., 2014). For more details of the INTER­GROWTH-21st Project, see Box 10.1.

10.2.3 Weight-for-stature

Weight-for-stature, often referred to as “Weight-for-length/height”, measures body weight relative to length or height. Low weight-for-stature in chil­dren is described as “thinness” and reflects a patho­logical process referred to as “wasting”. It arises from a failure to gain sufficient weight relative to stature or from losing weight. High weight-for-stature in chil­dren is termed “over­weight” and arises from gaining excess weight relative to length or height or from gaining insufficient length or height relative to weight (WHO, 1995a).

The preva­lence of wasting is defined as the proportion of chil­dren with a Z‑score for weight-for-stature < −2 (i.e., below −2SD from the WHO median for weight-for-length/height). Wasting often develops very rapidly, and is asso­ciated with both changes in the food supply and the preva­lence of infectious diseases. Wasting can be reversed quickly with an appropriate intervention. As a result, weight-for-stature is the preferred anthro­pometric index for ident­ifying young chil­dren who are most likely to benefit from a feeding program, or for evaluating the benefits of intervention programs. It is more sensitive to changes in nutritional status than stature-for-age.

A Z‑score between −2 and −3 is often defined as “moderate acute malnu­trition”. Chil­dren with moderate acute malnu­trition have specific nutrient needs, details of which are available in Golden (2009). When the Z‑score for weight-for-stature falls below −3, severe acute malnu­trition (SAM) is present (WHO / UNICEF, 2009). Chil­dren are some­times discharged following treat­ment for SAM when weight-for-stature reaches a Z‑score of > −2 relative to the WHO Child (Growth Standards) with no pitting edema (WHO, 2013). This discharge criterion is based on the lower risk of mortality reported compared to those chil­dren with a Z‑score more negative than −2.0.

In the past, the peak preva­lence of wasting was said to occur in chil­dren at aged 12–23mos, triggered by inappropriate complementary feeding and the depleting effects of infectious diseases, partic­ularly diarrhea (WHO, 1986). However, these early studies were based on the use of the NCHS refer­ence in which the majority of the infants were formula-fed, with a different growth pattern compared to their breast-fed counterparts, as noted earlier. As a result, use of the new WHO Child Growth Standard based on breast-fed infants has revealed an alarming level of wasting during infancy, partic­ularly in South Asian countries, a trend that was masked when the earlier NCHS refer­ence was applied, as shown in Figure 10.16.
Figure 10.16
Figure 10.16. Percentage of chil­dren (N=41,306) with wasting (weight-for-length / height Z‑score < −2) by age and refer­ence (NCHS and WHO) in India. Data are from the Indian National Family Health Survey 3 (2005–2006). Redrawn from Martorell and Young (2012)

As discussed earlier in Section 10.2, the striking differ­ences in the preva­lence of wasting are likely due to the inclusion of only breast-fed infants in the WHO sample, whereas the NCHS refer­ence was based predominately on formula-fed chil­dren, as noted earlier (see Figure 10.14). Differ­ences in the measure­ment intervals used may also be a factor.

Numerous preconceptual / prenatal and post­natal factors have been asso­ciated with wasting. Of the preconceptual / prenatal factors, short maternal stature, low BMI, and intra­uterine growth restriction are well recognized (Black et al., 2008), whereas maternal depression has only recently been explored (Ashaba et al., 2015). Post­natal attributable factors comprise poor infant and young child feeding, recurrent infections, and small body size in early life. In some circum­stances, household wealth and sub-optimal child care practices have also been identified (Martorell & Young, 2012).

Seasonality is another factor influencing the preva­lence of wasting. Rates are often higher in the rainy season when staple foods stored from the previous year's harvest may be depleted, morbidity is increased, and there is greater participation by women in the labor market (Hillbruner & Egan, 2008; Baye & Hirvonen, 2020).

Because the fifth Global Nutrition Target for 2030 (Box 10.2) includes “reduce and maintain childhood wasting to less than 3%” (WHO/UNICEF, 2021), WHO and UNICEF have devel­oped a screening tool that includes five preva­lence thresholds for wasting for global monitoring and for ident­ifying priority countries for action; see Section 10.2.6 for more discussion of this tool.

Studies using in vivo laboratory methods of body composition have shown that wasting is asso­ciated with major deficits in both the fat-free mass and fat mass. The methods employed include deuterium dilution (D2O), dual-energy X-ray absorptiometry (DXA), or bioelec­trical impedance analysis (BIA) (Chapter 14). In severe wasting, the extent of the decline in fat-free mass appears to be in proportion to the severity of wasting, whereas the decline in fat mass is more modest. Moreover, following treat­ment, although levels of fat may recover, those of fat-free mass remain low in the longer-term even following treat­ment; this pattern is shown in Figure 10.17
Figure 10.17
Figure 10.17. Association between body composition outcomes expressed as absolute fat-free mass (FFM) and fat mass (FM) and the severity of wasting, categorized by weight-for-height scores. Effects for the three categories of low weight-for-height z-score (WHZ) are expressed relative to a refer­ence group with WHZ > 0. Redrawn from Wells (2019)
with data from a randomized trial in Cambodia in which infants received four ready-to-use thera­peutic foods (RUTFS) from age 6–15mos. Body composition in this study was assessed using deuterium dilution (D2O) at 6mos and 15mos of age (Skau et al., 2019). For a detailed review of body composition in under­nourished chil­dren, see Wells (2019)

Over­weight and obesity in young chil­dren is becoming increasingly common worldwide (de Onis & Lobstein, 2010). Recognition of the role of childhood over­weight and obesity, and their subsequent causal link with diabetes and other chronic diseases in adulthood, has led WHO to include preva­lence thresholds for over­weight in chil­dren < 5y (defined as weight-for-height > 2 Z‑score) as well as wasting. Moreover, WHO/ UNICEF has endorsed “reduce and maintain childhood over­weight to less than 3% in 2030” as one of the six Global Nutrition Targets shown in Box 10.2 (WHO/UNICEF, 2021).

Differ­ences also arise in the preva­lence of over­weight (defined as > +2 Z‑scores weight-for-stature for children), with a higher preva­lence for all age groups when estimated by the WHO standard compared to the older NCHS reference, as noted in Section 10.2; for more discussion see (de Onis et al., 2006).

Interpretive standards

WHO Child Growth Standards provide data for weight-for-length for boys and girls age 0–2y; 2–5y; 0–5y expressed as Z‑scores or per­cent­iles and available as charts or tables (Growth Standards). However, during adoles­cence, the weight-for-height relation­ship changes dramatically with age and also with maturational status. For this reason, the WHO growth refer­ence for school-age chil­dren and adoles­cents 5–19y does not include weight-for-height refer­ence data. Instead, they include BMI-for-age (5–19y) Z‑scores and per­cent­iles (de Onis et al., 2007). The full set of charts and tables displayed by sex and age (years and months), per­centile and Z‑score values and related information are available (Growth Reference Data).

The U.S. CDC 2000 growth charts are available for weight-for-length data by sex from birth to 36mos, as well as weight-for-height refer­ence data for prepubescent boys and girls that extend from a height of 77–121cm (growth charts),

Note that increasingly, body mass index-for-age (BMI), expressed as a Z‑score or as a per­centile, is being used to assess over­weight and obesity in both childhood (0–5y) and in school chil­dren and adoles­cents (5–19y). The BMI indicator is replacing the index weight-for-stature used in the past to define over­weight and obesity in childhood (0–5y). The WHO defines a child 0–5y as “at risk of over­weight” if BMI Z‑score is > +1 and < +2; “over­weight” if BMI Z‑score > +2 and “obese” if BMI Z‑score is > +3, based on the WHO Child Growth Standard (de Onis & Lobstein, 2010). For children aged 5–19y, however, BMI-for-age Z‑scores above +1 and +2 based on the WHO 2007 growth reference data (de Onis et al., 2007) are applied to define over­weight and obesity, respectively. In contrast, based on the U.S. CDC 2000 growth charts, over­weight in a child is defined as a BMI > 85th per­centile but < 95th per­centile for age and gender, whereas a BMI > 95th per­centile is indicative of obesity;

The CDC2000 BMI-for-age growth charts did not extend beyond the 97th per­centile. With the increasing prevalence of severe obesity in the United States, CDC has developed new per­centiles to monitor very high BMI values by including data on children and adolescents from 1988–2016. These extended BMI per­centiles include the 95th, 98th, 99th, and 99.9th percentiles by age (2–20y) and sex (Extended BMI-for-Age Charts).

10.2.4 Length or height-for-age

Length or height-for-age (i.e., stature-for-age) is a measure of achieved linear growth that can be used as an index of past nutritional or health status. Recum­bent length is measured in infants and chil­dren less than 2y, and height in older chil­dren. Low height-for-age is defined as “shortness” and reflects either a normal variation or a pathological process involving failure to reach linear growth potential. The outcome of the latter process is the gaining of insufficient height relative to age and is referred to as linear growth retar­dation (or linear growth faltering) (Leroy & Frongillo, 2019).

The number of chil­dren suffering from linear growth retardation is much higher than the number of chil­dren who are stunted (Roth et al., 2017). Stunting is defined as having a length or height-for-age Z‑score < −2SD based on the WHO Child Growth Standard for chil­dren age 0–5y (Growth Standards). When the preva­lence of stunting is greater than 40%, it is considered a severe public health problem (WHO, 1995a). In popu­lations with a high preva­lence of stunting, usually the entire height distri­bution has shifted down­ward, suggesting that most, if not all indi­viduals, have been affected and are not growing to their full potential. However, when the preva­lence is much lower and approximates the expected level (i.e., ~ 2.5% of a healthy popu­lation), then those with low height-for-age are likely to be genetically short. WHO/UNICEF, 2021 has set a 50% reduction in stunting for chil­dren <5y by 2030 as one of the global nutrition targets (Box 10.2).

The preva­lence of linear growth retardation generally peaks during the second or third year of life (Roth et al., 2017; Leroy et al., 2015). This age pattern is much less affected by the growth refer­ence data used than wasting (de Onis et al., 2006), although the preva­lence estimates for stunting do appear to be higher based on the WHO Child Growth Standard compared to the older NCHS refer­ence data. The difference is attrib­uted to the tighter variability of the WHO standards which affects the place­ment of the usual refer­ence limit for stunting (i.e., −2SD)( de Onis et al., 2006). These discrep­ancies in the trends for stunting are evident in Figure 10.18, based on data from India. Here, the greatest difference between the WHO and NCHS refer­ence is apparent in the 0–5mos age group (i.e., 20% vs. 10%).
Figure 10.18
Figure 10.18. Percentage of chil­dren (N=41,306) with stunting (length-for-age < −2Z‑score by age and refer­ence (NCHS and WHO) in India (National Family Health Survey, 2005–2006). Redrawn from Martorell and Young (2012).

Unfortunately, in many low-income countries, length at birth or in the neonatal period, is seldom reported because of the difficulty of recording at the time of birth and the lack of the appropriate equipment and measure­ment training (Solomons et al., 2015). As a result, assess­ment of rates of linear growth retardation in many low-income countries are often based on those for chil­dren aged from 6–59mos, so the age of onset of growth failure is uncertain.

In a recent longitu­dinal birth cohort study, length (n=1197) was measured shortly after birth (mean 7.7d post-delivery) in 7 resource-poor settings in Bangladesh, Brazil, India, Nepal, Peru, South Africa, and Tanzania; gesta­tional age was not assessed (MAL‑ED Network Investigators, 2017). Here, shortly after birth the preva­lence of a length-for-age Z‑score below −1 of the WHO Child Growth Standard was 43% (range 37%–47% across sites), and that of stunting (i.e., length-for-age Z‑score below −2) 13% (range 10%–16% across sites).

It is of interest that in the MAL‑ED longitu­dinal birth cohort, there was an almost uniform decrease in length-for-age with age (except for Brazil), with the greatest increase in stunting in all sites (except Brazil) occurring after 6mos. The preva­lence of stunting reached a plateau at about two years. Hence, despite diverse cultures and geography, the patterns of stunting with age were the same across study sites (except for Brazil), although the magnitude differed, as shown in Figure 10.19.
Figure 10.19
Figure 10.19. Length-for-age stratified by age and site. Site-specific length-for-age trajectories with age were smoothed using a smoothing spline. Sites include BGD, Dhaka, Bangladesh; BRF, Fortaleza, Brazil; INV, Vellore, India; NEB, Bhaktapur, Nepal; PEL, Loreto, Peru; SAV, Venda, South Africa; TZH, Haydom, Tanzania. Redrawn from MAL‑ED study (2017)

The stunting rates at birth reported in the MAL-ED cohort study are in striking contrast to the stunting preva­lence within 6wks of birth reported for young infants of Mayan indigenous origin in Guatemala. Here stunting rates were 36-47% indi­cating that impaired fetal growth was the major predictor of early infant linear growth failure among these infants of Mayan indigenous origin (Bernagard et al., 2013; Solomons et al., 2015).

The etiology of linear growth faltering is multi­factorial, and is asso­ciated with many of the same prenatal and post­natal factors identified for wasting (Section 10.2.3). For example, in the MAL‑ED longitu­dinal birth cohort, five factors were deter­minants of linear growth falter­ing during early child­hood: lower enrol­ment weight, shorter maternal height, higher prev­al­­ence of entero­pathogen detection, lower socio­economic status, and consum­ption of a lower percent of energy from protein in non-breast-milk foods (MAL‑ED Network Investigators, 2017). For a comprehensive review of the factors recognized as causal for stunting, see Bhutta et al. (2008).

Interestingly, some of the predictors of stunting may not be the same as those linked to the recovery from this condition. In the MAL‑ED longitu­dinal cohort, the timing of stunting was significantly asso­ciated with recovery from stunting. Their findings suggested chil­dren who are stunted at an older age have a higher chance of recov­ering from stunted growth than those children who are stunted earlier, i.e, at age 6mos. (Das et al., 2021).

Identifying and preventing linear growth retardation or stunting during childhood is impor­tant because impaired linear growth results in a reduction in adult size, which, in turn, is causally linked with difficulties in child birth and poor birth outcomes, notably increased risk of low-birthweight infants (Sinha et al., 2018). Delays in cognitive and motor develop­ment in childhood, reduced earnings in adulthood, and chronic diseases have also been presented as con­se­quences of linear growth retardation and stunting during early childhood. Based on the current evidence, however, these outcomes are unlikely to be causally related, but instead correlates of linear growth retardation and stunting (Leroy & Frongillo, 2019).

Several early cross-sectional studies have linked stunting with an elevated risk of over­weight in chil­dren when clas­sified by BMI or weight-for-height (Popkin et al., 1996; Sawaya et al., 1995), although longitudinal studies have reported contrasting findings (Walker et al., 2007; Kagura et al., 2012). Caution is needed when using BMI (and weight-for-height) to examine associations between stunting and body composition, given that height is incorporated in the mea­sure­ment of both stunting (HAZ) and BMI. Wells (2019) emphasizes that this will generate an autocorrelation between short stature and high BMI. Instead, it is preferable to measure adiposity directly using such methods as bioelec­trical impedance analysis (BIA) or dual-energy X-ray absorptiometry (DXA). To date, whether stunting is casually associated with later adiposity is unclear in view of the many method­ologicall challenges that remain. For a discussion of possible pathways underlying the association between early life stunting and subsequent body composition and nutritional status, see Wells (2019).

Interpretive criteria

The distri­bution of height measure­ments at a given age within most popu­lations is often narrow, so that accurate measuring tech­niques are essential. More­over, a deficit in length takes some time to develop, so assess­ment of nutritional status based on length-for-age alone may under­estimate malnu­trition in infants in some settings.

Data for length-for-age for boys and girls age 0–6mos, 0–2y, 6mos–2y and height-for age 2–5y and length/height for age 0–5y expressed as Z‑scores or per­cent­iles based on the WHO Child Growth Standards are recom­mended for inter­national use, as noted earlier. Both charts and tables are available (Growth Standards). The U.S. CDC 2000 growth charts also provide length-for-age from 0–36mos and stature-for-age 2–20y (24–240mo) (Kuczmarski et al., 2002).

Note the use of the new inter­national standard for length at birth for gesta­tional age devel­oped in the INTER­GROWTH‑21st Project by Villar et al. (2014) will provide a method for the early diagnosis of linear growth failure, provided gesta­tional age can be assessed correctly. The WHO Child Growth Standard can then be used to monitor linear growth failure during infancy and childhood.

10.2.5 Height-for-age difference

Height-for-age Z‑scores (HAZ) are widely used to assess chil­dren's attained height at a given age (see Chapters 9 and 13 for more details on Z‑scores) However, some investigators have used positive changes in attained mean height-for-age Z‑scores to identify popu­lation-level catch-up growth in chil­dren (Crookston et al., 2010). Leroy et al. (2014) have raised concern over the appropriateness of using height-for-age Z‑scores to evaluate such changes in linear growth with age over time because the cross-sectional standard deviations used in the denominator of height-for-age Z‑scores and shown in Box 10.3 are constructed from cross sectional data and are not constant over time, but increase linearly from birth to 5 years of age. As a result, a child with a constant absolute height deficit will appear to improve with age based on the HAZ. Instead, Leroy et al. (2014) recom­mend using height-for-age difference (HAD) to describe and compare height changes as popu­lations of chil­dren age. Height-for-age difference (in cm) is defined as: child's height compared to standard, expressed in centimeters. It is calculated by subtracting the sex- and age-specific median height (from the WHO Child Growth Standard) from the child's actual height as shown in Box 10.3.
Box 10.3 Height-for-age Z‑scores (HAZ) and height-for-age difference (HAD)

Figure 10.20 compares changes in growth in popu­lations of chil­dren between 0–60mos based on HAD and HAZ using data from 51 nationwide surveys from low- and middle-income countries (Leroy et al., 2014).
Figure 10.20
Figure 10.20. Mean HAZ and HAD relative to the WHO standard (1–59mo) by completed month and kernel-weighted local polynomial smoothed values. Data from n = 309,215 chil­dren from 51 demographic and health surveys and multiple indicator cluster surveys. HAD, height-for-age difference; HAZ, height-for-age Z‑score. Redrawn from Leroy et al. (2014).
Note in the Figure 10.20, the mean HAZ started below the WHO Child Growth Standard (at approximately −0.4 Z‑scores) and fell markedly up to 24mos, after which it stabilized and increased only slightly. In contrast, based on the mean HAD curve, the chil­dren started with an average height deficit of 0.8cm, with the most pronounced faltering (i.e. steepest slope) evident between 6–18mos, as shown for the HAZ curve. Nevertheless, in contrast with the trend in the HAZ curve, the deficits in linear growth depicted by the HAD curves continued to increase after 18mos to 60mos, although at a lower rate than before 18mo. Indeed, the slopes of the HAD curves provide no indication that the process of growth faltering leveled off even at 5y. The bumps in the curves just after 24, 36, and 48mos are reportedly due to the tendency to report age in completed years rather than exact months (Victora et al., 2010). Such growth deficits were also evident in five regions (E. Europe and Central Asia, North Africa and Middle East, Latin America and Caribbean, Africa South of the Sahara, South Asia), although the magnitude of the deficits varied between regions; see Leroy et al. (2014) for more details. For more examples comparing changes in growth in popu­lations of chil­dren age 2–5y using HAD vs. HAZ based on both cross-sectional and longitu­dinal data, see Leroy et al. (2015). For a discussion of whether linear growth retardation (or “catch-up growth”) and the asso­ciated negative effects are reversible, consult Leroy et al. (2020).

10.2.6 Classification of the severity of malnu­trition based on the preva­lence of wasting, over­weight, and stunting

UNICEF/WHO/World Bank Group (2021) update the joint global and regional estimates of malnutrition among chil­dren under 5 years of age each year. These estimates of prevalence and numbers affected for child stunting, over­weight, wasting and severe wasting are derived for the global population as well as by regional groupings of United Nations. A new country-level model was used to generate the country, regional and global estimates for stunting and over­weight, for the 2021 edition and annual estimates from 2000 to 2020 are newly available. These estimates do not account for the impact of COVID-19, but the pandemic is expected to exacerbate all forms of malnutrition. This is likely due to worsening household income especially in vulnerable populations, constraints in the availability and afford­ability of nutritious food, disruptions in essential nutrition services, and reduced physical activity.

The severity of malnu­trition in young chil­dren age < 5y was originally classified into five classes of severity based on the preva­lence (as %) of wasting, stunting, and over­weight. This practice was adopted to highlight the levels and trends across countries, and identify the areas of greatest need and hence likely to gain the most benefit from an intervention (de Onis et al., 2019). The five preva­lence levels are shown in (Table 10.2).
Table 10.2 Preva­lence thresholds, corresponding labels, and the number of countries (n) in different preva­lence threshold categories for wasting, over­weight and stunting in chil­dren under 5y using the “novel approach”. Wasting: weight-for-length/height < −2 Z‑score; Over­weight: weight-for-length/height > +2 Z‑score; Stunting: length/height-for-age < −2 Z‑score; From: de Onis et al. (2019).
Wasting over­weight Stunting
Labels(n) Preva­lence
Labels (n)Preva­lence
< 2·5Very low 36< 2·5 Very low 18 < 2·5 Very low4
2·5 – < 5 Low 33 2·5 – < 5 Low 33 2·5 – < 10 Low26
5 – < 10 Medium39 5 – < 10 Medium50 10 – < 20 Medium30
10 – < 15 High14 10 – < 15 High18 20 – < 30 High 30
≥ 15 Very high10 ≥ 15 Very high9 ≥ 30Very high 44
The fifth threshold labeled “very low” of no‑public health concern was included across all three indicators to reflect the expected preva­lence in a healthy popu­lation. Theoretically, the proportion of chil­dren with a Z‑score < −2  or > +2 relative to the median of the WHO Child Growth Standards in a healthy popu­lation is 2.3% (rounded to 2.5%). The higher four preva­lence thresholds are defined by multipliers of this “very low” level.
Figure 10.21
Figure 10.21. Modeled per­cent­age of chil­dren < 5y affected by stunting in 2020. From: UNICEF, WHO, WB (2021).

As an example, Figure 10.21 depicts a modeled per­cent­age of chil­dren under 5y affected by stunting in countries worldwide in 2020, categorized by the five preva­lence thresholds for stunting presented in Table 10.2.

10.2.7 Composite Index of Anthro­pometric Failure (CIAF)

Currently, the UNICEF/WHO/World-Bank classification scheme provides no estimates on chil­dren suffering simul­taneously from multiple anthro­pometric deficits such as stunting plus wasting. However, there is increasing recognition that indi­vidual chil­dren may be at risk of both conditions simul­taneously, might be born with both, pass from one state to the other over time, and accumulate risks to their health and life through their combined effects.

In response to these concerns, Nandy and Svedberg (2012) proposed the Composite Index of Anthro­pometric Failure (CIAF). This is an aggregate measure aimed to estimate the overall burden of under­nutrition in chil­dren age < 5y that incorporates chil­dren who are wasted and/or stunted and/or under­weight based on the WHO Child Growth Standard (Growth Standards). The original CIAF model iden­tifies six groups of chil­dren defined by the categories B to G, shown in Table 10.3. The overall CIAF excludes those chil­dren not in anthro­pometric failure (i.e., group A) and counts all chil­dren who have wasting, stunting, or are under­weight (i.e., groups B to F), thus providing a single measure to estimate the overall preva­lence of under­nutrition (Porwal et al., 2021) (Table 10.3).
Table 10.3 Classification of anthro­pometric failure as per Composite Index of Anthropometric Failure (CIAF) categories, together with the distri­bution of participants per CIAF category. Note: the combination of wasted and stunted is not included as it is physically impossible because a child cannot have a low weight-for height and a low height-for-age and not be underweight. Data from from Porwal et al. (2021).
Categories of Under­nutrition Wasting Stunting Under-
N =
A No failure No No No 18,434 51.8
B Wasting (Low weight-
for-height WTHT) only
Yes No No 1392 4.6
C Wasting and Under­weight
[Low weight-for-height (WTHT)
and low weight-forage (WTA)]
Yes No Yes 1729 6.5
D Wasting, Stunting and Under­weight
(All three anthro­pometric failures)
Yes Yes Yes 1222 6.0
E Stunting and Under­weight
[Low height-for-age (HTA)
and Low weight-for-age (WTA)]
No Yes Yes 3552 16.6
F Stunting only
[Low height-for-age (HTA)]
No Yes No 3467 11.5
Y Under­weight only
[Low weight-for-age (WTA)]
No No Yes 704 3.0
B + C + D+ E + F + Y 12,066 48.2

Anthro­pometric data comparing the per­cent­age of Indian chil­dren classified as under-nourished when based on one of the three conven­tional growth indices and using the CIAF as an aggregate indicator of under­nutrition are also shown in Table 10.3. The data are based on 38,060 Indian chil­dren aged < 5y collected in the 2016–2018 Compre­hensive National Nutrition Survey (CNNS) (Porwal et al., 2021). As shown in Table 10.3, the preva­lence rate (as %) of under­nutrition is consistently less when only one of the three conven­tional indices is employed, compared to the overall CIAF preva­lence of 48.2% among chil­dren < 5y. Of these Indian chil­dren, 19.1% had only one form of anthro­pometric failure (groups B, F and Y), whereas 22.1% had two forms of anthro­pometric failure simul­taneously (groups C and E), and 6% had all three forms of anthro­pometric failures (Group D).

Nandy and Svedberg also argued that chil­dren who were simul­taneously stunted, wasted, and under­weight were likely to be at the greatest risk of mortality. However, most of the studies which have investigated risk of child mortality using the CIAF model to classify under-nutrition, have been cross-sectional (Khan & Das, 2020) so no causal inferences could be drawn. McDonald et al. (2013) were the first investigators to estimate mortality risk asso­ciated with multiple anthro­pometric deficits based on pro­spec­tive cohort data and classified according to the CIAF model (Table 10.3). The data were from ten pro­spec­tive cohorts in Africa, Asia, and Latin America. Pro­spec­tive cohort designs, unlike cross-sectional studies, provide the best information about causation of disease and the most direct measure­ments of the risk of developing disease.

McDonald et al. (2013) from their meta-analysis, reported a significantly elevated risk of all-cause mortality among chil­dren with 1, 2 and 3 anthro­pometric deficits compared with chil­dren with no deficits, with a dose response relation­ship in relation to the number of deficits. Moreover, for chil­dren wasted and under­weight but not stunted, risk of mortality was greater than for those who were stunted and under­weight but not wasted. Cause-specific mortality, however, could not be investigated due to insufficient statistical power; see McDonald et al. (2013) for further details.

More studies are needed to establish whether there are unique determinants of multiple anthro­pometric deficits, as well as to identify the combi­nation of anthro­pometric deficits asso­ciated with the greatest risk of mortality. With this knowledge, interventions could be targeted to the subgroup of chil­dren who may benefit the most.
Table 10.4 Preva­lence of malnu­trition based on the CIAF among chil­dren aged 6–59mos, in the Ghana Socioeconomic Panel Survey (GSPS). N = 1608, calculated using the composite index of anthro­pometric failure (CIAF) and the extended composite index of anthro­pometric failure (eCIAF). Data from Kuwornu et al. (2020).
GroupAnthro­pometric StatusCIAF - Preva­lence (95% CI)
A No failure 54.5 (50.9–58.1)
B Wasting only 4.2 (3.0–5.5)
C Wasting and under­weight 9.0 (7.2–10.8)
D Stunting, wasting, and under­weight 5.6 (4.3–7.0)
E Stunting and under­weight 7.6 (5.9–9.2)
F Stunting only 18.4 (15.8–21.0)
Y Under­weight only 0.8 (0.3–1.2)
B to Y Overall under­nutrition (CIAF) 45.5 (42.0–49.1)
eCIAF - Preva­lence (95% CI)
A No failure 42.7 (39.4–46.1)
B Wasting only 4.2 (3.0–5.5)
C Wasting and under­weight 9.0 (7.2–10.8)
D Stunting, wasting, and under­weight 5.6 (4.3–7.0)
E Stunting and under­weight 7.6 (5.9–9.2)
F Stunting only 12.3 (10.2–14.4)
Y Under­weight only 0.8 (0.3–1.2)
G Stunting and over­weight 6.1 (4.7–7.4)
H Over­weight only 11.7 (9.5–14.0)
B to H Overall malnu­trition (eCIAF) 57.3 (53.9–60.6)

With the emergence of the dual burden of malnu­trition, the original CIA model of classification for under-nutrition is no longer sufficient. This has led to the exten­sion of the CIAF model to include two new groups: “stunting and over-weight” and “over­weight only”, as shown in Table 10.4. Over­weight is defined as a BMI‑for‑age Z‑score > +2 for chil­dren < 5y based on the WHO Child Growth Standard (de Onis & Lobstein, 2010). The preva­lence of malnu­trition based on the eCIAF among chil­dren aged 6–59mos who participated in the 2011 Ghana Socio­econ­omic Panel Survey (GSPS) (n=1608) is also presented in Table 10.4 (Kuwornu et al., 2020). From the estimate of eCIAF in Table 10.4, more than one half of all Ghanian chil­dren aged 6–59mos had at least one form of malnu­trition over the study period.

10.2.8 Weight changes

Body weight is the sum of the protein, fat, water, and bone mass in the body. Changes in body weight do not provide any inform­ation on the relative changes among these components. Increasingly changes in the components of body compos­ition are assessed using laboratory-based methods, such as dual-energy X-ray absorp­tiometry (DXA), deuterium dilution, or bio­elec­trical impedance analysis. For details of these techniques, see Chapter 14.

Body weight is the sum of the protein, fat, water, and bone mass in the body. Changes in body weight do not provide any information on the relative changes among these components. In normal adults, there is a tendency for increased fat deposition with age, concomitant with a reduction in muscle protein. Such changes are not evident in body weight measurements but can be seen by determining either body fat or the fat-free mass. In healthy persons, the daily variations in body weight are generally small (i.e., less than ±0.5kg). In conditions of acute or chronic illness, however, negative energy-nitrogen balance may occur as the body can use endogenous sources of energy (including protein) as fuel for metabolic reactions. Consequently, body weight declines. In conditions of total starvation, the maximal weight loss is approximately 30% of initial body weight, at which point death occurs. In chronic semistarvation, body weight may decrease to approximately 50%–60% of ideal weight. In contrast, when persistent positive energy balance occurs, there is an accumulation of adipose tissue, and body weight increases.

Body weight can only be used to assess the severity of under­nutrition in subjects with relatively uncom­plicated, non­edematous forms of semi­starvation (Heymsfield et al., 1984). In disease conditions in which edema, ascites (fluid in the abdominal cavity), dehydration, diuresis, massive tumor growth, and organomegaly occur, or in obese patients undergoing rapid weight loss, body weight is a poor measure of body energy-nitrogen reserves. In such conditions, a relative increase in total body water, for example, may mask actual weight loss that results from losses of fat or skeletal muscle. Massive tumor growth may also mask losses of fat and muscle tissue, which may occur during severe undernutrition. Hence, additional anthropometric measurements (e.g., mid-upper-arm circumference and triceps skinfold thickness) should be taken to obtain more information on the origin of any change in body weight (Heymsfield et al., 1984).

To assess weight changes, the actual and usual weight of an indi­vidual must be known. From these two measure­ments, the per­cent­age of usual weight, per­cent­age of weight loss (or weight gain), and rate of change can be calculated using the equations shown in Box 10.4. The indi­vidual's actual weight can also be compared with appropriate age- and sex-specific refer­ence data.
Box 10.4 \[ \small \mbox{% Usual wt. =}\frac {\mbox{actual wt.}}{\mbox{usual wt.}} \mbox{× 100%}\]

\[ \small \mbox{% Wt. loss =}\frac {\mbox{usual wt. − actual wt.}}{\mbox{usual wt.}} \mbox{× 100%}\]

\[ \small \mbox{Rate of change =}\frac {\mbox{body weight present − body weight initial }}{\mbox{day present − day initial}} \mbox{ (kg/d)}\]

Weight changes in children

In low income countries, where there is a high prevalence of under-nutrition, weekly weight gain has often been used to monitor short-term response of children to a feeding program. For example in the past, in children with severe acute malnu­trition (SAM), percen­tage weight gain was applied as the discharge criterion. However, use of a weight-based assess­ment can be misleading in children with SAM. Such children may have diarrheal disease accom­panied by dehydration or edema, both of which have an affect on body weight (Modi et al., 2015).

Both lack of weight gain and weight loss in a child have been shown to be independ­ently and more closely related to mortality than other indicators of under­nutrition, such as BMI-for-age based on survival data from 2402 rural children aged 0–24mos fom the Democratic Republic of Congo (O'Neill et al., 2012). These two indicators are also included in a set of diagnostic indicators developed to identify pediatric under nutrition by the the U.S. Academy of Nutrition and Dietetics and the American Society for Parenteral and Enteral Nutrition (ASPN) (Becker et al., 2015). Pediatric under­nutrition in developed countries generally occurs in a setting of acute or chronic illness. Never­theless more studies are needed to evaluate the feasibility, suitability, validity, and reliability of these indicators to identify pediatric undernutrition (Mogensen et al., 2019).

With the growing concern about the dual burden of nutrition, changes in body weight are being monitored in apparently healthy children. For example, changes in body weight over a 15y period (both weight loss and weight gain) have been monitored in apparently healthy children and adolescents (n=> 150,000) in Germany (Vogel et al., 2022). A change outside ±0.2 BMI‑SDs per year was classified as a high positive or high negative change. Between 2005–2020, there was a small but stable positive trend in the proportion of children with a high positive weight change, especially within those children with obesity. This trend was accompanied by a decrease in the proportion with a high negative weight change. During COVID‑19, the weight gain was substantial across all weight and age groups (i.e., 1–6y; 6–12y; 12–18y), as noted by others (Jia, 2021; Brooks et al., 2021): COVID-19 has aggravated the childhood obesity pandemic.

Weight changes in adults

The U.S Academy of Nutrition and Dietetics and the American Society for Parenteral and Enteral Nutrition (ASPN) have also devised a series of diagnostic characteristics to identify and document malnu­trition (under­nutrition) in adults in acute, chronic, and transitional care settings in devel­oped countries. Six clinical charateristics were identified to support a a diagnosis of malnutrition in adults, one of which was percentage weight loss from baseline (White et al., 2012). For more details of these diagnostic indicators, the reader is referred to Chapter 27, Nutritional assess­ment of hospital patients.

Marked changes in body weight in healthy adults with age have been documented in several national surveys in industrialized countries. Such changes may have long-term consequences on health. In general during the period from young to middle adulthood, adults gain weight more rapidly and accrue excess adi­pos­ity, whereas from middle to late adulthood, weight begins to stabilize or even decrease, espe­cially around age 70y, when the decrease may continue for the remainder of life.

Several investigators have examined the association between weight changes across adulthood and mortality. For example, in a pro­spec­tive cohort study of US adults participating in NHANES from 1988 to 2014, Chen et al. (2019) reported stable obesity across adulthood, weight gain from young (age 25y) to middle adulthood (mean age 47y), and weight loss from middle to late adulthood, were associated with increased risk of mortality. These findings suggest that main­tain­ing a normal healthy weight across adult­hood, and especially pre­vent­ing weight gain in early adult­hood, is impor­tant for preven­ting premature deaths in later life.

The potential effect of weight loss and weight gain among older adults has also been investigated. In a systematic review and meta-analysis of community-dwelling older adults aged > 65y, the effect of weight loss on increased risk of all-cause mortality was stronger than weight gain (Alharbi et al., 2021), as noted by others (Karahalios et al., 2017). However, the point at which weight loss may indicate increased mortality risk has not been clearly defined. Further research is needed to determine whether these associations vary with gender, initial weight, and whether or not the weight loss/gain was intentional.

Nevertheless, caution when interpreting the results of some of these epidemiological studies is warrented. Methodological issues of concern include confounding by smoking, reverse causation due to existing chronic diseases, especially among the elderly, and non-specific loss of lean mass and function in the frail elderly (Fontana & Hu, 2014).

Gestational weight changes

The U.S. National Research Council (2009) have compiled new guidelines for gesta­tional weight gain in response to the increase in the proportion of U.S. women of reproductive age who are over­weight and obese. Gestational weight gain includes gains in maternal and fetal fat-mass and fat-free mass, as well as the placenta and amniotic fluid. How these different components of gesta­tional weight gain influence both maternal and offspring health remains uncertain: see Widen and Gallagher (2014) for more details.

For the first time, the new U.S guidelines consider those health outcomes of both mother and child during and after delivery which are plausibly related to gesta­tional weight gain, and the trade-offs between them. For the mother, the most important health outcomes considered were postpartum weight retention and unscheduled cesarean delivery, whereas for the infants the outcomes were preterm birth, extremes of birthweight (expressed as small- or large-for gesta­tional age), and childhood obesity (Rasmussen et al., 2010).

Table 10.5
Table 10.5. New recommen­dations for total and rate of weight gain during pregnancy, by prepregnancy BMI. * Calculations assume a 0.5–2kg (1.1–4.4lbs) weight gain in the first trimester (Rasmussen et al., 2010 and based on Siega-Riz et al., 1994; Abrams et al., 1995; Carmichael et al., 1997).
Total Weight
Rates of Weight Gain*
2nd and 3rd Trimester
Prepregnancy BMI Range
in kg
in lbs
Mean (range)
in kg/week
Mean (range)
in lbs/week
Under­weight (< 18.5 kg/m2) 12.5–18 28–40 0.51 (0.44–0.58) 1 (1–1.3)
Normal weight (18.5–24.9 kg/m2) 11.5–16 25–35 0.42 (0.35–0.50) 1 (0.8–1)
Over­weight (25.0–29.9 kg/m2) 7–11.5 15–25 0.28 (0.23–0.33) 0.6 (0.5–0.7)
Obese (≥ 30.0 kg/m2) 5–9 11–20 0.22 (0.17–0.27) 0.5 (0.4–0.6)
presents the new recommen­dations for total weight gain ranges for U.S women according to pre-pregnancy BMI (kg/m2) based on the BMI categories of WHO. The rates of weight gain (mean and range) for the 2nd and 3rd trimester are also stated. These recommen­dations have been adopted by WHO for low-income popu­lations, until specific country cutoff values are available. Note the recommen­dations that were included in the earlier IOM (1990) guidelines for short, young, or black women are no longer included in the new U.S. IOM/NRC (2009) recommen­dations.

The new U.S. recommen­dations include a relatively narrow range of recom­mended weight gain for obese women. However, because of the lack of data, no recommen­dations were made to reduce the guide­lines for obese women to below 5.0–9.0kg, or specific guidelines according to obesity classes II and III. Women with a pre-pregnancy BMI indicative of obesity are recom­mended to lose weight before and not during pregnancy, and clinicians are encouraged to monitor weight gain at each prenatal visit. Interventions may be needed to assist women, especially those who are over­weight or obese at the time of conception, to meet these guidelines.

10.3 Body mass index in adults

Table 10.6. Indices for weight relative to height. The power p in Benn's index is calculated to minimize the direct relation­ship with height (Benn, 1971).
Weight/height ratio wt/ht
Body Mass Index wt/(ht)2
Ponderal index ht / ∛wt
Benn's index wt/(ht)p
Weight to height ratios indicate body weight in relation to height and are partic­ularly useful for providing a measure of over­weight and obesity in adult populations. Hence these ratios are sometimes referred to as obesity indices.

At the present time, the ratio most used in this way is the body mass index (BMI) (also termed Quetelet's index), calculated as weight (kg) / height2 (m). Body mass index is used in preference to other weight / height indices, including the weight / height ratio, the Ponderal index, and Benn's index (Table 10.6). BMI is now widely used to classify over­weight and obesity in adults.

Body mass index is relatively unbiased by height and correlates reasonably well with other anthro­pometric measures of adiposity in adults such as waist circum­ference (WC) and abdominal visceral fat (AVF), as shown in Figure 10.22. For details on the importance of waist circum­ference and abdominal visceral fat, see Chapter 11.

Figure 10.22
Figure 10.22. Correlations among BMI, FM, WAIST and CT-assessed AVF. The arrows are bidirectional to indicate that no causal relation­ships are implied in the model. Each coefficient represents the weighted mean of six correlations computed in samples of black men, black women, white men and white women from the HERITAGE Family Study, and of Caucasian men and women from the Quebec Family Study. AVF, Abdominal visceral fat; FM, fat mass; waist, waist circum­ference. From Bouchard (2007).

Body mass index is employed in large-scale nutrition surveys and epidemio­logical studies as a measure of over­weight and obesity; mea­sure­ments of weight and height are easy, quick, relatively noninvasive, and more precise than skinfold thickness mea­sure­ments. Nevertheless, BMI does not quantify total body adiposity or distinguish between weight associated with muscle and weight associated with body fat. Hence, in some circumstances, an elevated BMI may result from excessive adiposity, muscularity, or edema, and is thus an inaccurate mea­sure­ment for adiposity at the individual level.

Certain population groups may also be misclas­sified as over­weight or obese based on BMI, due to differences in body proportions and the relation­ship between BMI and body fat content. Māori and some Polynesian groups, for example, appear to have less body fat than Caucasians of the same age, sex, and BMI (Craig et al., 2003; Rush et al., 2009), but in some Asian populations, the reverse is true (Deurenberg et al., 1998; WHO Expert Consultation, 2004) (Section 10.3.1). Other examples include athletic populations and individuals who are very short (< 150cm). Abnormal relation­ships between leg and trunk length can also lead to some misclas­sification (Deurenberg et al., 1999), although BMI can now be corrected to consider unusual leg length based on the ratio of sitting height to standing height (Norgan & Jones, 1995). Such adjustments are necessary for example, among Australian Aboriginals.

In addition, BMI conveys no information about the actual distribution of body fat. Anomalies in the distribution of abdominal visceral fat, and in particular ectopic fat (i.e., fat depots in liver, heart, muscle, pancreas and kidney) are now recognized to be an even greater cardio­metabolic risk than excess body fat per se (Neeland et al., 2019). Conse­quently, BMI alone can misrepresent the burden of abdominal visceral fat in a subgroup of individuals who are at highest cardiometabolic risk. Therefore, a BMI within the normal range does not necessarily preclude an increased cardio­metabolic risk. In a community-dwelling adult men in Korea, for example, assess­ment of cardiometabolic risk based on high blood pressure, hyperglycemia and dyslipidemia, was highest among those with a BMI < 25kg/m2 but who had a high waist circum­ference (i.e., > 90cm). Moreover, in this group the amounts of both total fat mass and per­cent­age body fat (assessed via bioelec­trical impedance analysis) were similar to the group with a BMI > 25kg/m2 and a waist circum­ference < 90cm (Kim et al., 2020).

Therefore, increasingly a combination of both waist circum­ference (as a surrogate estimate of abdominal visceral fat) and BMI are being used to predict risk of cardiovascular disease (Nazare et al., 2015). In the future, simple clinically applicable tools to monitor changes over time in abdominal visceral fat and ectopic fat need to be developed for use with BMI and waist circum­ference. Currently, the laboratory methods available are used mainly in research settings; they are described in Chapter 14.

10.3.1 BMI and measures of body fat and disease risk

The validity of BMI as an index of the percentage body fat and fat mass (in kg) in adults has been assessed by comparing BMI with body fatness estimated initially using a 2‑component model, although more recently multi-component models of body composition are being used (Gallagher et al., 1996; Wells et al., 2010; Silva et al., 2013).

In the simpler 2‑component model, total body mass is partitioned into total body fat and fat-free mass, applying the principle that if one of the components is measured, the other can be estimated. However, when using this approach, the calculation of body fat depends on certain theoretical assumptions. The hydration of fat-free mass, for example, is assumed to be constant within and between individuals even though inter-individual variability is known to exist, especially during growth and maturation in children, in pregnancy, and among adults with varying adiposity (Wells et al., 2010; Most et al., 2018; Gutiérrez-Marin et al., 2021; Gallagher et al., 1996).

Figure 10.23
Figure 10.23. The relation­ship between body mass index and percentage of body fat in both younger and older women from New York City. Body fat determinations based on a four-compartment bodycomposition model. Data from Gallagher et al., (1996).
In contrast, use of multi-component models for measuring body composition minimizes the use of theoretical assumptions for key body properties. In Figure 10.23, for example, the relation between BMI (kg/m2) and body fat (as a percentage) was assessed using a 4‑component model in young (< 35y) and elderly (> 65y) women. A comparable relation between BMI and body fat mass (kg) was also observed. In this study, measurements were collected on body weight, body volume, total body water, and bone mineral mass, allowing the inter-individual variability in the composition of the fat-free mass to be considered rather than assuming fixed constants for the water content, bone mineral content, or density of fat-free mass (Wells & Fewtrell, 2006). Consequently, the four-component model is the established criterion reference method for generating accurate data on body fat and fat-free mass. For more details of these methods used to assess body composition, see Chapter 14

Over the last few decades, validation studies have analyzed the performance of BMI to detect body adiposity by comparison with a variety of techniques considered to measure body composition accurately. They have included a combination of statistical approaches including correlation and regression analyses, paired t‑tests, and more recently, Bland-Altman analyses. In some cases, the sensitivity and specificity of the methods have also been compared (Okorodudu et al., 2010). For details of the statistical techniques now applied to evaluate agreement between body composition methods used in validation studies, see Earthman (2015).

A major assumption of BMI has been that it is an independent index of body fat. This means that after adjusting for body weight-for-stature, all subjects with the same BMI have the same relative fatness, irrespective of their age, sex, or ethnicity. However, it is now widely known that the relation­ship between BMI and per­cent­age body fat is influenced by age, sex, and ethnicity, all of which have implications for the use of BMI as an index of body fatness (Norgan, 1994; Gallagher et al., 1996; Jackson et al., 2002).

It is recognized that the relation­ship between BMI and body fat is both age‑ and sex-dependent (Deurenberg et al., 1991; Gallagher et al., 1996). For example, in the cross-sectional study shown in Figure 10.23, older women tended to have a relatively greater per­cent­age of body fat than their younger counterparts with a comparable BMI (Gallagher et al., 1996). A similar age-related trend was noted for men; these age-related trends persist up to 60–65y of age in both sexes. According to a recent critical review based on data from the U.S NHANES survey, such age-related changes in body composition by both men and women possibly reflect senescence-mediated hormonal changes, lowering of physical activity levels, reductions in trunk length owing to osteoporosis, and a variety of other related mechanisms (Heymsfield et al., 2016). Moreover, women have significantly greater amounts of total body fat than do men for an equivalent BMI suggesting that BMI cannot be used as a comparable measure of fatness in males and females. These sex differences are substantial and are maintained throughout the entire adult life span (Gallagher et al., 1996). Further, they can be seen in children as young as 3–8y (Taylor et al., 1997).

The relation­ship between BMI and the per­cent­age of body fat appears to differ among certain race/ethnic groups, as noted earlier. Figure 10.24 shows results of a meta-analysis among different race/ethnic groups conducted by Deurenberg et al. (1998).
Figure 10.24
Figure 10.24. Ethnic differences in calculated values for body mass index (BMI) (mean and 95% confidence interval) which reflect equal levels of body fat, adjusted for age and sex. The means are relative to the results for Caucasians (set to 0.0). Data from Deurenberg et al. (1998).
Of the 15 studies included in the meta-analyses, 11 used those considered by the investigators to be a “reference criterion method” for the mea­sure­ment of body fat, and included hydrostatic weighing, deuterium dilution, and DXA. Results of the meta-analysis indicate that in some populations (e.g., some Chinese, urban Thais, Indonesians), levels of obesity in terms of per­cent­age of body fat will be greater at the 30kg/m2 obesity cutoff suggested by WHO (2000) than for Europeans.

The existence of such race/ethnic-related differences has been confirmed by several other investigators (Wagner & Heyward, 2000; Deurenberg-Yap et al., 2000; Hsu et al., 2012). For example, Asian Indians are also reported to have more body fat than Europeans at any given BMI (WHO Expert Consultation 2004; Low et al., 2009), whereas for Chinese from Beijing and rural Thais, values are like those of Europeans (He et al., 2015). In contrast, in some Pacific populations, the per­cent­age of body fat at a given BMI is lower (see Polynesians in Figure 10.13) (Swinburn et al., 1999). Possible reasons for these race / ethnic-related discrepancies may include differences in body shape and body composition, some of which may be inherited and others due to environ­mental and lifestyle factors. There is an urgent need for more in-depth studies of race / ethnic differences in body shape and composition and how these differences relate to clinically meaningful risks; see Heymsfield et al. (2016) for more discussion.

Body mass index has been used in numerous international population studies to assess disease risk among adults. Increasing BMI is clearly associated with a higher risk of high blood pressure, type 2 diabetes mellitus, cardiovascular disease (coronary heart disease and stroke) and cancer (WHO Expert Consultation 2004; Abdullah et al., 2010; Ni Mhurchu et al., 2004 ; Whitlock et al., 2009; Wiseman, 2008). Indeed, the relative risk for cardiovascular disease risk factors and cardiovascular disease incidence increases in a graded fashion with increasing BMI in all population groups. Nevertheless, the risk of type 2 diabetes appears to vary according to race/ethnicity and presence or absence of other traditional risk factors, with Chinese Americans, Hispanics, African Americans over a 10y period having a similar risk at lower BMI values compared with those for white participants (Rodriguez et al., 2021). These findings are based on data from a large prospective cohort analysis from the Multi-Ethnic Study of Atherosclerosis of US adults aged > 45y who were free of diabetes at baseline and in whom incident type 2 diabetes was defined as fasting glucose > 7.0mmol/L, or use of any diabetes medications.

Over­weight and obesity as defined by BMI has also been associated with increased mortality based on both large scale prospective studies in single countries (Jee et al., 2006; Patel et al., 2014) and systematic reviews and meta-analysis in several regions or continents For example, Aune et al. (2016) conducted a systematic review and meta-analysis of 230 cohort studies of BMI and the risk of all-cause mortality in five regions (North America, South America, Europe, Australia, and the Pacific). Over­weight and obesity were associated with increased risk of all-cause mortality, consistent with the findings of Di‑Angelantonia et al. (2016), from a study across four continents (Europe, North America, east Asia and Australia and New Zealand). In the systematic review and meta-analysis of Aune et al. (2016), the lowest mortality was observed in those participants with a BMI of about 25.0

In addition, associations between BMI and musculo­skeletal disorders, impairments in respiratory and physical functioning, quality of life, and mental illness such as clinical depression, anxiety, and other mental disorders have been reported by the (CDC)

10.3.2 WHO classification of over­weight and obesity in adults

The WHO International Obesity Task Force defined obesity as a condition in which abnormal or excessive fat accumulation may impair health. (WHO, 2000). They have recommended the use of a graded classification of both over­weight and obesity to:
Table 10.7. Classification of over­weight in adults according to body mass index (BMI). From (WHO, 2015) and (CDC, 2022)
Classification BMI
Risk of
Underweight < 18.50 Low (but risk of
other clinical prob-
lems is increased)
Normal weight > 18.5 to < 25.0 Average
Over­weight 25 to < 30.0 Increased
Class 1 obesity 30 to < 35.0 Moderate
Class 2 obesity 35 to < 40.0 Severe
Class 3 obesity > 40.0 Very severe
Table 10.7 shows the classification of adults according to BMI recommended by WHO (2015). The classification is based primarily on the association between BMI and mortality. WHO considers BMI as the most useful population-level measure of over­weight and obesity as it is the same for both male and female adults and for all ages of adults. This classification system is not intended for use with pregnant and lactating women, or persons < 18y.

WHO (1995a) have also suggested that a universal cutoff point of BMI ≥ 30 could be used to estimate and compare the prevalence of obesity within and between populations. Figure 10.25 shows the trends in BMI for adults in some selected regions in 1980, 2008, and 2016 by sex. The figure emphasizes the wide variation in prevalence across regions: the low prevalence in South Asia is particularly marked. The increase in obesity from 1980 to 2016 is clearly a world-wide phenomena and independent of sex and region (NCD-RiskC, 2017).

Figure 10.25
Figure 10.25. Global trends in the prevalence of obesity defined as BMI ≥ 30kg/m2 in men and women in 1980, 2008 and 2016 from selected regions of the world. Data from the Noncommunicable Diseases Risk Factor Collaboration (NCD-RisC). (1998).
WHO (2000, 2004) recognized that these BMI cutoff points for over­weight and obesity may not correspond to the same d.egree of fatness and risk of obesity-related co-morbidities across different race/ethnic populations, as noted earlier. In particular, in their 2004 report, WHO emphasized the mounting evidence that certain Asian populations have a partic­ularly high risk of type 2 diabetes, cardiovascular disease, and mortality from other causes at relatively lower BMIs. However, WHO did not consider there was enough evidence for a single BMI cutoff point for all Asians for over­weight or obesity at that time. Nevertheless, a report cosponsored by the WHO Western Pacific Region (WPRO) and led by the International Association for the Study of Obesity and the International Obesity Task Force (WHO/IASO/IOTF, 2000) did propose lower cutoff points for all Asian adults for over­weight and obesity, based primarily on the relation­ship between BMI and morbidity; these are presented in Table 10.8. The use of these lower cutoffs for both over­weight and obesity for Asians was supported by the findings of a prospective cohort study in Taiwan. (Wen et al., 2004).
Table 10.8. Proposed classification of weight by BMI for adult Asians. (WHO/IASO/IOTF, 2000 ).
Classification BMI
Risk of
Underweight < 18.50 Low (but risk of
other clinical prob-
lems is increased)
Normal weight > 18.5 to 22.9 Average
Over­weight ≥ 23
At risk 23 to 24.9 Increased
Class 1 obesity 25 to 29.9 Moderate
Class 2 obesity ≥ 30.0 Severe

More recent research findings in South Asian populations suggest that a BMI ≥ 23kg/m2 is not sensitive enough to identify the obesity-associated disease risk in these populations, and indicate that an even lower cutoff (i.e., BMI ≥ 21kg/m2) may be more appropriate in South Asian countries (Jauawardena et al., 2013).

10.3.3 Canadian classification of over­weight and obesity in adults

The BMI classification system for adults used by Health Canada has been adapted from that of WHO (2000) The classification applies to adult women (nonpregnant and nonlactating) and men between the ages of 18 and 65y (Douketis et al., 2005).

Table 10.9. Health Canada (2003) classification of adults according to BMI. For persons ≥ 65y, the “normal” range may begin slightly above BMI 18.5 and extend into the “over­weight” range.
Health Canada
Risk of developing
health problems
Underweight < 18.50 Increased
Normal weight 18.50 to 24.99 Least
Over­weight 25.00 to 29.99 Increased
Obese Class I 30.00 to 34.99 High
Obese Class II 35.00 to 39.99 Very high
Class 2 obesity ≥ 40.00 Extreme high

Health Canada (2003) cautions that the classification system may misclas­sify certain population groups such as adults with a very lean body build or who are highly muscular, young adults who have not reached full growth, those over 65y, and certain ethnic and racial groups. Several health problems associated with over­weight and obesity are also specified. They include coronary heart disease, type 2 diabetes, hypertension, dyslipidemia, gallbladder disease, obstructive sleep apnea, and certain cancers. A combination of BMI with waist circum­ference as a measure of abdominal visceral fat are now used to screen for over­weight and obesity and to assess a person's risk of health problems (i.e., Type 2 diabetes, hypertension, dyslipidemia and the metabolic syndrome, which combines three of these conditions) (Douketis et al., 2005).

10.3.4 U.S. classification of over­weight and obesity in adults

The cutoff points that define normal weight, over­weight, and the various categories of obesity in the United States are now the same as those used by WHO (2015) and are presented in Table 10.7.

Obesity is often subdivided into three categories in the United States, with class III (i.e., BMI > 40) designated as “severe obesity”. These BMI cutoff points for adults define over­weight and class I to III obese individuals and identify those who may be at increased risk for cardiovascular disease and other obesity-related conditions. However, within these categories, personal additional risk assess­ment is also needed because degree of risk can vary. For more details on the assess­ment and treat­ment of cardiovascular risk factors and obesity-comorbidities for individuals with over­weight or class I to III obesity, the reader is referred to the AHA/ACC/TOS guideline (2013). Note that these guidelines also recommend the mea­sure­ment of waist circum­ference in patients with BMI 25–34.9kg/m2. The waist circum­ference cutoff points indicative of increased cardiometabolic risk used in the United States are > 88cm for women and > 102cm for men. These same waist circum­ference cutoffs are also used for adult Canadians. See Chapter 11 for more details of the mea­sure­ment of waist circum­ference.

10.3.5 BMI and chronic energy deficiency in adults

In low-income countries, low values for BMI in adults have been consistently associated with a decline in work output, productivity, and income generating ability, as well as a compromised ability to respond to stressful conditions (Ferro‑Luzzi et al., 1992). Underweight has also been associated with increased mortality. For example, in an early study in India, a progressive increase in mortality rate below a BMI of 18.5 was reported, with an almost threefold higher rate after ten years for those with BMI below 16 (Naidu & Rao, 1994). Whether the individuals were ill prior to the mea­sure­ment is unknown, however, making it difficult to assign the trend to any causal relation­ship. Nevertheless, some evidence for such a relation­ship has been reported based on more recent meta-analysis of prospective studies across continents (Di Angelantonio et al., 2016; Wild & Byrne, 2016), as well as from an analysis of the Global Burden of Disease Study (2017).

In developed countries, some of the early studies of low BMI and mortality failed to control for two principal sources of bias: cigarette smoking, and the elimination of thin adults with early mortality, many of whom may had been already ill when measured (Manson et al., 1987). More recently, Flegal et al. (2005) investigated excess deaths associated with under­weight (as well as over­weight and obesity) in the U.S NHANES data. In their study, they noted an increase in mortality with both under­weight (BMI < 18.5) and with obesity (BMI ≥ 30), but not with over­weight (i.e., BMI ≥ >25 but < 30). In most developed countries today, underweight is common in adults in hospitals and nursing homes, almost invariably secondary to disease, although in some more affluent countries under­weight has been associated with osteoporosis, infertility, and impaired immuno­competence (Health Canada, 2003).

A single cutoff point of < 18.5kg/m2 is frequently used to define low BMI values, indicative of under­weight in both men and women in both low-income and more affluent countries (Naidu & Rao, 1994; Health Canada, 2003; NCD Risk Factor Collaboration, 2017). FAO has categorized three degrees of under­weight based on BMI and adopted the term “chronic energy deficiency” for under­weight (Shetty & James, 1994). The FAO classification of adult chronic energy deficiency based on the three degrees of under­weight are shown in Table 10.10.

Table 10.10. Simple classification of adult chronic energy deficiency using BMI (kg/m2). From Shetty & James (1994).
Chronic energy
deficiency grade
Normal > 18.50
Grade I 17.0 to 18.4
Grade II 16.0 to 16.9
Grade III < 16.0
WHO (1995a; 2004) adopted these same cutoffs to define three grades of low BMI, referred to as “thinness” in WHO (1995a) and as “under­weight” in WHO (2004), rather than the FAO definition of “chronic energy deficiency.” In WHO (2004), Grade I (17.0–18.4kg/m2) is termed “mild under­weight,” Grade II (16.0–16.9kg/m2) “moderate under­weight, and Grade III (< 16kg/m2) “severe under­weight”. WHO (1995a) also devised a classification system to identify populations with a public health problem on the basis of low BMI; details are shown in Table 10.11. However, challenges still exist in relation to the use of low BMI cutoffs for defining thinness or chronic energy deficiency in adult populations. It is likely that the cutoff for under­weight, like over­weight, might also differ with
Table 10.11. WHO classification of populations on the basis of low BMI. Compiled from WHO (1995a).
Situation Percent of population with BMI
(kg/m2) < 18.5
Low prevalence
(warning sign,
monitoring required)
5 to 9
Medium prevalence
(poor situation)
10 to 19
High prevalence
(critical situation)
20 to 39
Very high prevalence
(critical situation)
≥ 40
ethnicity and environ­ment, as well as sex and duration of undernutrition, and more research is required.

In any healthy adult population, 3%–5% can be expected to have a BMI < 18.5kg/m2, but in countries with food insecurity, an excessive proportion of the population may have BMIs < 18.5, with very few obese subjects. However, in general as socio­economic conditions improve, there is a tendency for a population to shift from thinness to over­weight (de Onis & Habicht, 1996). In many low‑ and middle-income countries, this trend has now occurred with a rising prevalence of over­weight and obesity (Figure 10.25), which is surpassing the fall in the global prevalence of under­weight (NCD Risk Factor Collaboration, 2017). For example, in 2014, more men were obese than under­weight in 136 (68%) of countries based on a pooled analysis of trends on adult BMI from 1975–2014 from 200 countries and terri­tories. For women, obesity surpassed under­weight in 165 (83%) of countries and severe obesity surpassed under­weight in 135 countries.

This emerging situation in which low and middle-income countries are impacted by both under‑ and over­nutrition is termed the “double burden of malnutrition” (DBM) (Shrimpton & Rokx, 2012), and is receiving increasing attention (Wells et al., 2020; Hawkes et al., 2020; Nugent et al., 2020). The WHO refers to the double burden of malnutrition as “characterized by the coexistence of under­nutrition along with over­weight, obesity or diet-related noncommunicable diseases (NCDs), within individuals, households and populations, and across the life-course” (WHO, 2017). This broad definition describes the current state of global health where rising rates of over­weight and obesity overlap with stagnant or slowly decreasing rates of under­weight and chronic energy deficiency.

10.4 BMI in children and adolescents

Body mass index is also now the inter­nation­ally recommended screening indicator of over­weight and obesity in both children and adolescents (WHO, 1995a). The recommendation arises from the following observations:
Figure 10.26
Figure 10.26 Relation­ship between BMI and total body fat (via DXA) in 90 boys and 98 girls aged 5–19y. Regression lines for boys (Model A:solid line) and girls (Model C;dashed line). Modified from Pietrobelli et al. (1998).
Notwithstanding the relation in some children of BMI and body fatness measured by DXA, BMI is not a perfect indicator of body fatness. Dietz and Bellizzi (1999) have cautioned that BMI will falsely classify some children of normal fatness as over­weight and some over­weight children as not over­weight. Although errors in the mea­sure­ment of per­cent­age of body fat may account for some of the variability in the relation­ship between BMI and body fatness in children, several other factors may be implicated. These include enhanced muscular development, large head size, and a high torso-to-leg ratio. All these factors may falsely elevate BMI into an over­weight range in some “non-over­weight” children (Roberts & Dallal, 2001).

Defining over­weight and obesity in children is more difficult than in adults. In children BMI has a characteristic curvilinear shape with age so that no simple cutoff can be used to define over­weight or obesity, so the age-independent adult BMI classifications are not appropriate. Indeed, the 50th per­centile for BMI for younger children and adolescents increases markedly from birth to early adulthood. These challenges have led many countries to use their national reference to compile their own ways of defining over­weight and obesity in children making it impossible to make comparisons across countries (Rolland-Cachera, 2011).

There are three main classification schemes that are used inter­nation­ally to define over­weight and obesity in children. They are all based on BMI per­cent­iles or SD scores derived from growth charts to account for gains in weight relative to height and each is described below.

10.4.1 International Obesity Task Force (IOTF) mass index cutoffs for over­weight and obesity in children

In 1998, WHO convened an International Task Force on Obesity to define over­weight and obesity in children and adolescents. This expert group agreed to recommend the adoption of an earlier approach developed by the European Childhood Obesity group (Poskitt, 1995) but based on an international reference population.

To accomplish this recommendation, Cole et al. (2000) compiled data for BMI for children 2–18y from six large nationally representative cross-sectional growth studies (Brazil, Great Britain, Hong Kong, the Netherlands, Singapore, and the United States). The data were used to construct two country-specific per­centile curves passing through the adult definitions of both over­weight (BMI > 25kg/m2) and obesity (> 30kg/m2) at age 18y for each sex. The curves were then averaged across countries by age to yield sex-specific curves for each cutoff.

A major disadvantage of these international (IOTF) cutoffs, however, was that they could not be expressed as per­cent­iles (e.g., 85th per­centile) or standard deviation scores (+2 Zscores). As a result they could not be compared with other reference data in which the BMI cutoffs were defined by age-sex- specific BMI per­cent­iles (e.g., 85th or 95th per­centile) or standard deviation scores (e.g., +2SDs). Conse­quently, Cole and Lobstein (2012) reformulated their original international cutoffs using the LMS curves which permit BMI to be expressed as a per­centile or SD score. Table 10.12 presents the sex-specific BMI cut- offs at age 18y for over­weight (BMI 25), obesity (BMI 30), and morbid obesity (BMI 35), expressed as a SD score equivalent or per­centile equivalent. Only small differences (< 0.2% on average) were observed when these new international IOFT cutoffs were compared with the original cutoffs based on prevalence rates for over­weight and obesity from US and Chinese data.
Table 10.12. SD score cutoffs corresponding to the international BMI cutoffs. Note the new cutoff for morbid obesity (BMI=35). From Cole and Lobstein (2012).
BMI cutoff at
18y (kg/m2)
SD score
25 1.31090.5
30 2.28898.9
35 2.93099.83
25 1.24489.3
30 2.19296.6
35 2.82299.76

A worked example on how to derive these new cutoffs for obesity in children is given in the appendix of Cole and Lobstein (2012), allowing researchers to construct their own cutoffs for any required BMI (e.g., BMI 23 for Asians) at 18y. The adoption of this reformulated approach has several important advantages. In summary:

10.4.2 WHO classification of over-weight and obesity in children

With the development of international growth references for both young children 0–5y and school-aged children and adolescents for clinical and epidemiological use, WHO has adopted cutoffs points for over­weight and obesity for the two age groups based on standard deviation scores (i.e., Z‑scores) derived from the corresponding BMI-for-age curves of the corresponding reference data.

For infants and children 0–5y, the WHO Child Growth Standards based on the multi­center study of children aged 0–5y from 6 diverse geographic sites (Brazil, Ghana, India, Norway, Oman, and the United States) should be used (WHO, 2006). whereas, for school-age children and adolescents the WHO international growth reference 2007 is recommended. The latter is a reconstruction of the original 1977 National Center for Health Statistics (NCHS) data set supplemented with data from the WHO Child Growth Standard to ensure a smooth transient at age 5y (de Onis et al., 2007). The statistical methodology used to construct this reference was the same as that used for the WHO Child Growth Standard (2004). The full set of tables and charts for the BMI-for-age curves by sex for both preschool and school-aged children, including application tools, is available (WHO international growth reference 2007).

For a child 0–5y, WHO defines “at risk of over­weight” as a BMI Z‑score of > +1 but < +2; “over­weight” as a BMI Z‑score > +2 and < +3, and “obese” as a BMI Z‑score > +3, based on the WHO Child Growth Standard (WHO, 2006; de Onis & Lobstein, 2010). For a school-aged child aged 5–19y, WHO defines a BMI Z‑score above +1 but < +2 as over­weight, and +2 as obesity, based on the WHO 2007 growth reference data (de Onis et al., 2007; de Onis & Lobstein, 2010).

Details for the justification of these WHO cutoffs were not provided. The more stringent WHO cutoffs for children aged < 5y compared to those ≥ 5y were proposed to avoid over­diagnosis of obesity in younger children; the increase in average BMI observed in all age groups of children has been greater in the older children since the 1990s. Hence, the current trend in BMI in children no longer reflects what is termed “age agnosticism”, in which the risk of obesity is assumed to be constant throughout infancy and childhood (de Onis & Lobstein, 2010; Wright et al., 2022).

Cole and Lobstein (2012) compared the prevalence rates for over­weight and obesity defined by the WHO BMI Z‑scores and their extended international IOFT cutoffs using US and Chinese data. There were systematic differences in the prevalence of over­weight and obesity; the WHO prevalence rates were lower for children age 2–5y, but higher for those 5–18y, probably due to the method of construction of the WHO per­centile curves; see Cole and Lobstein (2012) for more details.

10.4.3 U.S. classification of over­weight and obesity in children

The approach of Cole et al. (2000; 2012) has not been adopted in the United States. Instead, the U.S 2000 CDC growth charts are used, with over­weight among children aged 2–19y defined as BMI > 85th but < 95th per­centile, and obesity as BMI > 95th per­centile of the sex-specific CDC BMI-for-age growth charts. Two sets of clinical charts for children are available. Set 1 has the outer limits of the curves at the 5th and 95th smoothed per­cent­iles and includes the 85th per­centile for BMI-for-age as shown in Figure 10.27.
Figure 10.27
Figure 10.27.U.S 2000 CDC Body mass index-for-age per­cent­iles: boys 2–20y.
Set 2 has the outer limits of the curves at the 3rd and 97th per­cent­iles and are for specialist use. For details on the compilation of the U.S 2000 CDC growth charts, see Section 10.7. For details of the BMI-for-age charts see: 2000 CDC growth charts.

The smoothed per­cent­iles of the original U.S CDC reference growth charts extended only up to the 97th per­centile. The marked rise in the prevalence of extreme obesity among children in the U.S (i.e, BMI > 99th per­centile of the CDC growth charts) has led to alternative definitions of severe obesity in U.S children aged > 2–19y. They include: BMI > 120% of the 95th per­centile (1.2 × 95th per­centile) for age, or an absolute BMI > 35kg/m2 for adolescents, whichever is lower based on age and sex. The inclusion of an absolute BMI > 35kg/m2 threshold was chosen by the American Heart Association because it aligns the pediatric definition of severe obesity with the U.S class II obesity in adults (Table 10.7), in whom there is a high risk of early mortality (Kelly et al., 2013). However, these alternative definitions of severe obesity in children aged > 2 to 19y have some limitations which have been discussed in detail in a recent CDC report by Hales et al. (2022). CDC has now extended their BMI distribution on their BMI-for-age growth charts by using additional data from NHANES (1999–2016), rather than relying on extrapolation. These BMI-for-age charts extend up to a BMI of 60kg/m2 with four additional per­centile curves above the 95th (i.e., 98th, 99th, 99.9th, and 99.99th) by gender.

The CDC (2022) recommend using these extended BMI-for-age growth charts (Figure 10.28) for monitoring obese children and adolescents in place of the severe obesity growth charts that are used in clinical care (Gulati et al., 2012; Kelly et al., 2013). For monitoring growth of U.S children without obesity, however, the 2000 CDC BMI-for-age growth charts (e.g., Figure 10.27) should be used.

Figure 10.28
Figure 10.28. Extended CDC BMI-for-age growth chart for boys. From CDC (2022)

Unlike the two different cutoffs for children aged < 5y versus ≥ 5y set by WHO (de Onis et al., 2007), the CDC per­centile cutoffs assume that the risk of obesity is constant throughout childhood, thus reflecting “age agnosticism”, an approach suggested to result in overdiagnosis of obesity in younger children (Wright et al., 2022). However, the CDC does caution that “in‑depth assessments” are required to determine whether children and adolescents with BMI-for-age > 95th per­centile are truly overfat, and at increased risk for health complications.

Skinfold thickness and waist circumference have been proposed as alternatives for quantifying adiposity in over­weight and obese children and adolescents but are challenging in children with severe obesity and not recommended. For a review of the existing evidence for an association of excess adiposity with cardiovascular and metabolic risk in children, see Chung et al. (2018).

10.4.4 Comparisons of over­weight and obesity in children among countries

BMI is accepted as a valid indirect measure of adiposity in children. However, unlike adults, the 50th per­centile BMI varies with age as well as weight. Consequently, BMI values in childhood must be compared with age and sex-specific reference data sets. Unfortunately, differences exist in the methods that have been used to construct the curves for the reference data sets, some of which are based on representative national data (e.g., 2000 CDC reference data), whereas others provide international reference data (Cole et al., 2000; WHO, 2006). Marked variations across countries also occur in the terminology used to define the levels of BMI indicative of over­weight and obesity. For example, based on the French BMI reference data, “over­weight” corresponds to values above the 97th per­centile and no cutoff for “obesity” exists (Rolland-Cachera et al., 2002), whereas the US CDC now define over­weight as a BMI ≥ 85th per­centile but < 95th per­centile, and obesity as ≥ 95th per­centile. In contrast, WHO apply Z‑scores to indicate “risk of over­weight”, “over­weight”, and “obesity”, and these differ for children less than and greater than aged 5y (Section 10.4.2).

As a result of these differences, comparison of the prevalence of over­weight and obesity across countries is challenging, and the prevalence estimates differ even when compared in the same population of children. For example, estimates often appear highest when using the WHO classification systems, intermediate with the 2000 CDC definition, and lowest based on the IOTF definition (Shields & Tremblay, 2010).

An additional challenge complicating the interpretation of the data is the absence of international validated gold-standard upper BMI cutoffs for over­weight and obesity in childhood. The existing prospective studies in which excess adiposity in children has been associated with risk of cardiovascular disease are limited by small sample sizes (Chung et al., 2018). Moreover, to date, there are no large-scale longitudinal data that relate over­weight in childhood with adverse health outcomes in adulthood because of the long-time span required before they appear (i.e., middle-age or beyond) (Wright et al., 2022). This has led to the conventional statistical definitions for the BMI cutoffs for over­weight and obesity, with the assumption that the prevalence of over­weight or obesity is the same for each age‑ and sex-specific group, even though it appears that older children now experience a greater increase in BMI than younger children.

Clearly, an international consensus is necessary before meaningful comparisons between studies of over­weight and obesity during childhood can be made (Rolland-Cachera, 2011; Flegel & Ogden, 2011). The current recommendation by the European Childhood Obesity Group (ECOG) is to use the IOTF and WHO definitions for international comparisons of the prevalence of childhood over­weight and obesity, despite their limitations. For comparison within countries, however, national age‑ and sex-specific BMI reference data and definitions of over­weight and obesity are appropriate (Rolland-Cachera, 2011). Further discussion on the use of international versus national BMI reference data are available in Chinn & Rona (2002) and Reilly et al. (2010).

Unfortunately, the evidence to identify children who are truly at risk for future adverse health outcomes arising from over­weight or obesity remains insufficient. It appears that the conventional statistical approach to define over­weight and obesity is no longer valid and results in overdiagnosis of obesity in younger children when a true measure of total fat mass is determined based on dual energy x‑ray absorptiometry (Vanderwall et al., 2017) or deuterium dilution (Wright et al., 2022). Therefore, more research is urgently needed to improve the diagnosis of childhood obesity and define its adverse health consequences.

10.4.5 Using BMI to define thinness in children and adolescents

The term “thinness” was adopted by Cole et al. (2007) to describe a low BMI to avoid confusion with the terms “wasting” (defined by a weight-for-stature< −2 Z‑score) and “under­weight” (defined by a weight-for- age < −2 Z‑score).

Cole et al. (2007) used the same methods to define grades of thinness in children and adolescents as they used previously to define over­weight and obesity in this age group (Section 10.4.1). For each dataset, per­centile curves were drawn to pass through the BMI cutoff values of 18.5, 17, and 16kg/m2 at age 18y, consistent with the grades 1, 2, and 3 of thinness used by WHO for adults (Table 10.12). The per­centile curve passing through a BMI value of 17kg/m2 at age 18y gave a mean BMI close to a Z‑score of −2, matching the existing WHO criteria for wasting in children less than 5y of age (i.e., weight-for-height below −2 Z‑score). Hence, the authors proposed that this per­centile curve should be a basis for an international definition of thinness in children and adolescents.

However, a major disadvantage of these childhood BMI cutoffs for thinness is that they are not expressible as BMI per­cent­iles. Conse­quently, Cole and Lobstein reformulated their international cutoffs for thinness for children aged 2–18y so that BMI could be expressed as per­cent­iles or Z‑scores for comparison with the WHO BMI cutoffs (Cole & Lobstein, 2022). Table 10.13 shows the sex-specific Z‑score cutoffs and their corresponding per­centile equivalents to the original international BMI cutoffs (16, 17, and 18.5kg/m2 at age 18y) which are consistent with the grades 3, 2, and 1 of thinness used by WHO for adults (Section 10.3.5). For a comparison of these new cutoffs for thinness with the originals, see Cole & Lobstein (2012).
Table 10.13. SD score cutoffs corresponding to the international BMI cutoffs for thinness. From Cole and Lobstein (2012).
BMI cutoff at
18y (kg/m2)
SD score
16 -2.5650.52
17 -18773.0
18.5 -1.01415.5
16 -2.4360.74
17 -1.7893.7
18.5 -0.97516.5

The BMI-for-age reference data as sex-specific Z‑scores or as per­centiles for both children aged 0–5y and school-aged children and adolescents (5–19y) are available as charts and tables from WHO (Child growth standards; WHO international growth reference 2007). Thinness and severe thinness are defined as: BMI −2 Z‑score and BMI −3 Z‑scores, respectively. Raw mea­sure­ments can be calculated as BMI-for-age per­cent­iles and Z‑scores using the WHO AnthroPlus application (AnthroPlus 2009).

In an effort to avoid overestimating the prevalence of thinness among South Asian children, de Wilde et al. (2013) have suggested that ethnic specific BMI-cutoffs for thinness should be developed, as proposed for the BMI cutoffs for over­weight and obesity.


RSG and KM-J would like to thank past collaborators, and is grateful to Michael Jory for the HTML design and his tireless work in directing the translation to this HTML version.