13.0 Introduction

Standardized methods of evaluating anthro­pometric indices are essential to assess nutritional status and to identify mal­nut­rition in individuals or populations. Two methods are used based on Z‑scores or per­cent­iles, both calculated in relation to the distri­bution of the anthro­pometric indices for a reference population. The World Health Organization (WHO) recom­mends the use of the WHO Child Growth Standards for young children 0–5y inter­nationally. They were compiled using a prescriptive approach depicting physiological human growth under optimal conditions. Hence, they represent how young children should grow, rather than as a “reference” describing how children do grow. The WHO Child Growth Standards are recom­mended for inter­national use because the effects of ethnic and genetic differences on the growth of infants and young children are small compared with the environmental, nutritional, and socio-economic effects on growth, some of which may persist across generations. The reference population recom­mended for assessing the growth of older children is the WHO Growth Reference data for school-age children and adolescence.

A series of prescriptive standards for monitoring fetal, newborn growth, and gestational weight gain have also been developed by the INTERGROWTH-21^st project for inter­national use. This project adhered to the WHO recom­mendations for assessing human size and growth (WHO Working Group on Infant Growth, 1995), and followed healthy pregnant women longitudinally from 9wks of fetal life to 2y (Papageorghiou et al., 2018). Populations from urban areas of eight countries in which maternal health care and nutritional needs were met (Brazil, China, India, Italy, Kenya, Oman, the UK and the USA) were involved to ensure universal multi-ethnic growth standards were generated that represent how fetuses should grow (i.e., the standards are prescriptive). Box 13.1 lists the prescriptive growth standards generated from the INTERGROWTH-21^st project with the associated references.

In clinical setting, serial anthro­pometric measure­ments can be used to track the pattern of growth and body composition over time. They can also be used to identify and classify an individual as “mal­nourished” or of “normal” nutritional status by comparison with either pre­determined reference limits, often termed cutoff points. Individuals may also be classified into risk categories indicative of the severity or type of mal­nut­rition.

However, serial measure­ments are often not possible in public health. Instead, screening systems, which include at least one anthro­pometric measurement or index in combination with the recom­mended reference limit or cutoff point are used to identify individuals in the population “at risk” to mal­nut­rition. Examples of measurements or indices used in this way and discussed below include weight-for-height, body mass index (BMI), MUAC (to identify severe acute mal­nut­rition), and a composite index of anthro­pometric failure (CIAF). Data derived from screening systems can be used to compare and/or monitor the preva­lence of mal­nut­rition across populations, and to assess the need for, or impact of, inter­ventions both within and across countries. In some cases, reference limits can be com­bined with “preva­lence thresholds” to describe and compare populations according to levels of severity of mal­nut­rition, and identify priority populations for action. Prevalence thresholds can also be used by governments to trigger action, and to monitor and evaluate the impact of interventions.

Anthropometry can also be used to characterize the nutritional status of a population or to highlight changes in their nutritional status with successive measure­ments over time (i.e., surveillance). To achieve these objectives, a comparison of the frequency distri­bution of anthro­pometric indices of the entire population in relation to the recom­mended reference data using Z‑scores (in developing countries) or per­cent­iles is recom­mended.

Details of the methods recom­mended by WHO for evaluating anthro­pometric indices, followed by a review of the systems used to identify and classify risk of mal­nut­rition at the individual and population levels, are discussed below. Examples of the applications of anthropometry in public health are also included together with details of the designs and statistical procedures that can be used to evaluate inter­vention programs.

13.1 Anthro­pometric indices and their modes of expression

Details of the standardized procedures for the measure­ments of body size and body composition are outlined in Chapters 10 and 11. Raw measure­ments, however, have little meaning unless they are related to, for example, the age or sex of an individual (WHO, 1995). Hence, the correct interpretation of anthro­pometric measure­ments requires the use of anthro­pometric indices. They are usually compiled from two or more raw measure­ments. Details of indices of body size and body composition are also available in Chapter 10 and 11, respectively. In children, the four most commonly used anthro­pometric indices are weight-for-height, height-for-age, weight-for-age, and body mass index — defined as weight / (height)² (kg/m²).

Weight-for-height measures body weight relative to height. Low weight-for-height is described as “thinness” and reflects a pathological process referred to as “wasting”. It arises from a failure to gain sufficient weight relative to height or from losing weight. High weight-for-height in children is referred to as “over­weight” and arises from gaining excess weight relative to height or from gaining insufficient height relative to weight.

Height-for-age is a measure of achieved linear growth and 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 normal variation or a pathological process involving a failure to reach a linear growth potential. The outcome of the latter process is referred to as “stunting” or the gaining of insufficient height relative to age (WHO, 1995). The etiology of stunting is multifactorial, often resulting from a combination of extended periods of inadequate food intake, poor dietary quality, and increased morbidity.

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. High weight-for-age can be described as “heaviness” but this term is seldom used: high weight-for-height is more useful as a proxy for “overweight” (WHO, 1995).

A major limitation of weight-for-age is that it reflects both weight-for-height and height-for-age. It fails to distinguish tall, thin children from those who are short with adequate weight. Thus, children with a low weight-for-age may be genetically short or their low weight-for-age may result from “stunting” or nutritional growth failure. Consequently, weight-for-height and height-for-age are the preferred anthro­pometric indices of body size for children because, in combination, they can distinguish wasting and stunting.

Table 13.1 The relative usefulness of different anthro­pometric indices on a scale from 1 (excellent) to 4 (poor). ^a Depends partially on the preva­lence of stunting and wasting in the population. Modified from Gorstein et al. (1994).

	Wt.- for- ht.	Ht.- for- age	Wt.- for- age	BMI- for- age
Usefulness in populations where age is unknown or uncertain	1	4	4	4
Usefulness in identifying wasted childrena	1	4	3	1
Sensitivity to weight change over a short period of time	1	4	2	2
Usefulness in identifying stunted children	4	1	2	4

Body mass index (BMI) is the inter­nationally recom­mended index of over­weight and obesity in children and adults and calculated as: \[\small\mbox{weight(kg)/ height(m)}^{2} \] An increased BMI-for-age in childhood and adolescence is associated with a higher percentage of body fat, which is a known risk factor for cardiovascular disease (de Onis and Lobstein, 2010). However, BMI alone does not distinguish between weight associated with muscle, and weight associated with body fat. Hence, in some circum­stances, an elevated BMI may result from adiposity, muscularity, or edema. Furthermore, BMI is insensitive to the actual distri­bution of body fat (Thomas et al., 2012). For more details, see Chapter 10. Table 13.1 summarizes the information that can be obtained from each of the four growth indices discussed above.

Note that to interpret any single measure­ment of either weight or stature in relation to the reference data, the exact age of the child at the date of the measure­ment must be known. Software such as WHO AnthroPlus can calculate exact ages in decimal fractions of a year, from birth and visit dates. 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 event calendar. Information on the develop­ment of a local event calendar is available from FAO (2008).

For studies of both individuals and populations, the anthro­pometric indices can be compared to the reference population using per­cent­iles or Z‑scores derived from the reference data. The reference populations recom­mended are the WHO Child Growth Standards (WHO, 2006) for infants and children 0–5y, and the WHO Growth Reference data for school‑age children and adolescents 5–19y (de Onis et al., 2007).

The WHO Child Growth Standards were based on a multicenter growth reference study for infants and children aged 0–5y from six diverse geographic sites (Brazil, Ghana, India, Norway, Oman, and the U.S.), as noted earlier. The study com­bined a longitudinal design from 0–24mos with a cross-sectional study of children aged 18–71mo from the same sites (de Onis et al., 2004). A prescriptive approach was taken for the develop­ment of these growth standards with the goal of providing a single inter­national reference that represents the best description of physiological growth for all children ≤ 5y and to establish the breastfed infant as the normative model for growth and develop­ment.

The WHO Growth Reference for school-age children and adolescents 5–19y was developed from the 1977 National Center for Health Statistics data set (Silveira et al., 2011), supple­mented with data 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. More details of these child growth standards are available from WHO.

In most industrialized countries, per­cent­iles are used, whereas in low-income countries, the use of Z‑scores is recom­mended. The proportion of individuals with indices below predefined reference limits based on per­cent­iles or Z‑scores can also be assessed and are widely used as indicators of community status.

In some circum­stances, per­cent­iles and Z‑scores cannot be calculated and, instead, the anthro­pometric indices are expressed in terms of percent-of-median. Percent-of-median, however, provides only limited information on the relative position of the value within the population and as a consequence, is rarely used today.

13.1.1: Percentiles

A per­cent­ile refers to the position of the measure­ment value in relation to all the measure­ments for the reference population, ranked in order of magnitude. Figure 13.1

is a cumulative frequency distri­bution of the heights of boys aged 13y, illustrating the use of per­cent­iles. In this figure, a height of 152cm represents the 50^th per­cent­ile. This means that 50% of the boys in the reference population have a height at or below this value and 50% at or above this value. Similarly, 10% of the boys have a height at or below the 10^th per­cent­ile that, in Figure 13.1, falls at 141.8cm. For data with a Gaussian distri­bution (e.g. height for age), the 50^th per­cent­ile corresponds to both the mean and the median; for skewed data (e.g. weight-for-age), the 50^th per­cent­ile corresponds to the median.

The per­cent­ile for an individual of known age and sex can be calculated exactly, if the numerical per­cent­ile values are available for the reference data. Computer programs (e.g. WHO AnthroPlus for personal computers) are available to calculate exact per­cent­iles (1^st to 99^th per­cent­iles) based on the WHO Child Growth Standards (0–5y) (WHO, 2006) and the WHO Reference (2007) for school-aged children and adolescents (5–19y) (de Onis et al., 2007). Alternatively, the updated US Center for Disease Control and Prevention (CDC) 2000 Growth Reference in the United States (Kuczmarski et al., 2000) can be used from which numerical per­cent­ile values can be calculated based on the method described by Vidmar et al. (2004). Alternatively, the per­cent­ile range within which the measure­ment of an individual falls can be read from sex-specific graphs or tables of the appropriate reference data.

For children (0–5y), per­cent­ile charts are for: weight-for-length from 45–110cm; weight-for-height from 65–120cm; weight-for-age (0–60 completed months); length / height-for-age (0–60mos); arm-circumference-for age (0–60mos); BMI-for-age (0–60mos); head-circumference-for-age (0–60mos); triceps skinfold-for-age (0–60mos); subscapular-skinfold-for-age (0–60mos). Growth charts are available from WHO. For field use, simplified field tables are also available from WHO. For the WHO Growth Reference 2007 for school-aged children and adolescents (5–19y), percentile charts for weight-for-age (5–10y) and height-for-age and BMI-for-age (5–19y) are discussed in de Onis et al. (2007).

Sometimes, when calculating per­cent­iles, adjustments are made for parental stature, to distinguish between genetic and pathological effects in a child who is unusually short. The magnitude of the adjustment tends to be greater for older children. An example of adjustments based on mid-parent height is given in Himes et al. (1985).

Adjustments are also important when evaluating the growth per­cent­iles of both low birth weight (i.e., < 2500g) and preterm infants (i.e., < 38wk gestation). Specialized growth charts are available for determining the growth per­cent­iles of preterm infants (Villar et al., 2015).

For population studies, the number and percentage of individuals falling within specified per­cent­iles of the reference data can be tabulated or presented graphically to provide an estimate of “relative status”, as shown in Figure 13.2.

This approach highlights critical features of the distri­bution of the study population compared to that of the reference population (WHO, 1986).

The per­cent­ile reference limits commonly used for designating individuals as “at risk” to mal­nu­trition are either below the 3^rd or 5^th per­cent­iles or above the 97^th or 95^th per­cent­iles. The limits chosen depend on the reference data used.

Use of per­cent­iles is recom­mended for the evaluation of anthro­pometric measure­ments from relatively well-nourished populations from industrialized countries, as no errors are introduced if the data have a skewed distri­bution. Weight-for-age, weight-for-height, and many circumferential and skinfold indices have skewed distri­butions.

Percentiles are not recom­mended for evaluating anthro­pometric indices of individuals or populations from low-income countries if reference data from industrialized countries, such as the US CDC 2000 (Kuczmarski et al., 2000), are used. In such circum­stances, an individual, or many participants with large deficits in weight and height, may have indices below the extreme per­cent­ile of the reference distri­bution (i.e. below 3^rd or 5^th per­cent­ile), making it difficult to characterize the magnitude of the deficit for the individual or accurately classify large numbers of individuals (Waterlow et al., 1977). Moreover, because data from a population, expressed in terms of per­cent­iles, are often not normally distributed, such populations cannot be correctly described in terms of means and standard deviations of the per­cent­iles.

13.1.2 Z‑scores

The World Health Organization recom­mends the use of Z‑scores for evaluating anthro­pometric data from low-income countries (Gorstein et al., 1994), because Z‑scores can be calculated accurately beyond the limits of the original reference data. This is an advantage in low-income countries because individuals with indices below the extreme per­cent­iles of the reference data can then be classified accurately.

The method measures the deviation of the anthro­pometric measure­ment from the reference median in terms of standard deviations (SD) or Z‑scores. The score is a measure of an individual's value with respect to the distri­bution of the reference population.

Z‑scores are calculated differently for measure­ments that are distributed normally (e.g., height- for-age) and non-normally in the reference population (e.g., weight-for-age). The Z‑score for a measure­ment that is normally distributed corresponds to:
\[\small \mbox{Z‑score =} \frac{\mbox{(observed value)−(median reference value)}}{\mbox{ standard deviation of the reference population}}\]

As an example, a boy called Sam is 96.1cm tall, and aged 2y and 4mo

• the observed value is Sam's height (96.1cm).
• the median reference value for height of all boys who are Sam's age (2y and 4mo) is 90.4cm.
• the standard deviation of boy's height at 2y and 4mos is 3.3. The standard deviation of boy's height can be thought of as the average of differences from the median for each member of the reference population.
• inserting the above numbers in the formula, Sam's Z‑score is calculated as: \[\small \frac{(96.1− 90.4)}{ 3.3 } = 1.73\]

Note: Sam's height-for age could also be plotted manually on the WHO Child Growth Chart for boys aged 2–5y shown in Figure 13.3.

However, the above formula cannot be used for measurements that are not normally distributed such as weight-for-age, weight-for-height, and BMI-for-age, because the distances between adjacent Z‑scores are not constant. As a result, an LMS formula is used to calculate the Z‑scores for measure­ments that are skewed. For more details, see Cole (1990) and Vidmar et al. (2004). Note that for measure­ments that fall beyond −3 and +3 Z‑scores a modified version of the LMS formula must be applied (incor­porated within WHO Anthro software).

These calculations can be carried out using the WHO software program (WHO AnthroPlus 2009) which uses the WHO Child Growth Standard for those 0–60mos (WHO, 2006) and the WHO Growth Reference for children and adolescents aged 5–19y (de Onis et al., 2007) as the reference populations. As noted earlier for per­cent­iles, for clinical use, the Z‑score range within which the measure­ment of an individual falls can also be read from graphs or tables of the appropriate reference data for the same indicators for infants and children age 0–60mos and children and adolescents age 5–19y. Simplified field tables are also available.

WHO has also developed a tool for the application of the WHO Child Growth Standards which includes instructions of how to take the measure­ments, interpret growth indicators, investigate causes of growth problems, and how to counsel caregivers. A child age calculator is available as part of the course materials for the trainers and available in WHO Regional Offices. An anthro­pometry training (video) is also available. Motor milestone windows of achievement have also been developed and are available from (WHO.)

Guidelines have also been developed on the use of the fixed exclusion method to identify implausible Z‑score value for each anthro­pometric growth index based on the WHO Child Growth Standard. Z‑score values outside the following intervals are considered implausible:

• HAZ (below −6, or above +6);
• WHZ (below −5, or above +5);
• WAZ (below −6, or above +5).

When using WHO AnthroPlus these “implausible” values for each anthro­pometric Z‑score are flagged independently and the percentage of implausible values for each index calculated. Note: the date of birth (month and year) is required to calculate Z‑scores and errors in this date information are a major source of implausible values. A percentage of implausible values that exceeds 1% is said to indicate poor data quality. However, inaccurate values may still exist within the rather generous flag range set by WHO, so even a low percentage does not necessarily imply adequate data quality.

WHO also recom­mends calculating the standard deviation (SD) of Z‑scores for each anthro­pometric growth index as an additional assessment of data quality. This is important because an inflated SD arising from poor quality data results in preva­lence estimates that are likely to be overestimated. Currently, WHO has not set cut-offs at which SDs for each anthro­pometric index are related to poor data quality. Instead, WHO suggests that strata specific SDs that are greater than the SD for the national estimate should be examined and highlighted in any survey report. More details on standardizing the analysis, interpretation, and reporting of anthro­pometric data for infants are available from (WHO).

Z‑scores for anthro­pometric indices using alternative reference populations such as U.S. CDC 2000 Growth Reference in the United States, (Kuczmarski et al., 2000) can also be calculated based on the method described by Vidmar et al. (2004).

In population studies, the number and the proportion of individuals within a specified range of Z‑scores for each age and sex group can be tabulated or presented graphically to provide an estimate of “relative” status, as shown in Figure 13.4

and as noted previously for per­cent­iles (Figure 13.2). In addition, the proportion of individuals with anthro­pometric indices below or above some predetermined reference limits defined by Z‑scores can also be determined. Often scores of below −2 or above +2 are used to designate individuals with either unusually low or unusually high anthro­pometric indices. This approach is used because statistically 95% of the inter­national reference population fall within the central range. Theoretically the proportion of children with a Z‑score less than −2 or greater than +2 in a study population should be ≈ 2.3%. Clearly, if the proportion in the study population with such low or high Z‑scores is significantly greater than this, then the study population is seriously affected.

The WHO Global Database on Child Growth and Malnutrition uses a Z‑score less than −2 to classify low weight-for-age as “under­weight”, low length/height-for-age as “stunted”, and low weight-for-length as “wasted” (WHO, 2006). In children aged 0–5y, a Z‑score above +2 is used to classify weight-for-height as “over­weight” and as “obese” with a weight-for-height Z‑score above +3.

Figure 13.5

presents the preva­lence of wasting, stunting, and over­weight based on these WHO indicators in children at age 1 year in 1982, 1993, 2004, and 2015 in population-based cohorts in Pelostas, Brazil, 1982–2015. The preva­lence of stunting declined by 53% (from 8.3% to 3.9%) from 1982 to 2015. Wasting preva­lence remained stable at low levels (1.8% in 1982 and 1.7% in 2015), whereas over­weight increased by 88% (6.5% to 12.2%). Note that the current preva­lence of stunting in Pelostas is over twice, and that of under­weight over five times higher than the 2.3% that would be expected in a well-nourished population. Details on the interpretation of these preva­lence estimates to classify levels of mal­nut­rition in global surveillance are given in Section 13.4.5.

The WHO indicators stunting, over­weight, and wasting are three of the six global nutrition target indicators shown in Box 13.2.

Indicators for over­weight and obesity in both young and older children and adolescents are also based on  Z‑scores for BMI-for-age of the WHO Child Growth Standard (WHO, 2006) and the WHO growth reference 2007. For children (0–5y) a Z‑score for BMI-for-age above +1 is described as being “at risk of over­weight”, above +2 as “over­weight” and above +3 as “obese”. For children greater than ≥ 5y, de Onis and Lobstein (2010) suggest using the adult classification scheme; children with BMI Z‑scores above +1 are described as being “over­weight” and above +2 “as obese”.

An important advantage of using Z‑scores for population-based applications is that it is valid to calculate the mean and standard deviation for a group of Z‑scores. This allows the nutritional status of the entire population to be described. In addition, because the observed Z‑scores from a population are often normally distributed, statistical analytical procedures that assume normality such as t-tests and regression methods can be used (Gorstein et al., 1994).

In some circum­stances, however, a mean Z‑score will not be useful. For example, in some studies of refugees, the death rate among those most severely malnourished can be high and the lower end of the distri­bution can be truncated: the mean is hardly affected (Yip and Sharp, 1993). In such cases, a comparison of the distri­bution of indices of the entire population in relation to the reference population is more useful (Figure 13.4).

If the distri­bution of the reference values is normal, per­cent­iles and Z‑scores are directly related, as shown in Table 13.2:

Table 13.2. Equivalents of per­cent­ile and Z‑scores in a normal distri­bution.

Below mean		Above mean
Percentile	Z‑score	Percentile	Z‑score
5.0	–1.645	55.0	0.126
10.0	–1.282	60.0	0.253
15.0	–1.036	65.0	0.385
20.0	–0.842	70.0	0.524
25.0	–0.675	75.0	0.675
30.0	–0.524	80.0	0.842
35.0	–0.385	85.0	1.036
40.0	–0.253	90.0	1.282
45.0	–0.126	95.0	1.645
50.0	–0.000

a Z‑score of −1.645 corresponds to an obser­vation on the 5^th per­cent­ile. Conversely, a value that is at the 85^th per­cent­ile equates to Z‑score of +1.036. The commonly used −3, −2, and −1 Z‑scores are the 0.13^th, 2.28^th, and 15.8^th per­cent­iles. An easy way to assess whether the distri­bution is skewed is to compare the values of the mean and the median. For normal distri­butions the mean and median will be very similar. As the distri­bution becomes more skewed, the difference between mean and median increases. Several tests are available for testing normality. Examples include the Kolmogorov-Smirnoff and the Cox tests, of which the latter is generally preferred.

The characteristics of the three anthro­pometric reporting systems — Z‑scores, per­cent­iles, and percent-of-median — are compared in Table 13.3. The inter­pretation of reference limits based on per­cent­iles and Z‑scores, is believed to be consistent across all ages and indices. This means that a reference limit of 3^rd per­cent­ile or −2 Z‑scores represents the same degree of mal­nut­rition, irrespective of the anthro­pometric index used (e.g. weight-for-age or weight-for-height) or the age of the child (Waterlow et al., 1977). Nevertheless, the physiological meaning or consequence of extreme values may differ with age and height.

Table 13.3. The characteristics of three anthro­pometric data-reporting systems. From: WHO (1995).

Characteristic	Z‑score	Per- centile	% of median
Adherence to ref- erence distri­bution	Yes	Yes	No
Linear scale permitting summary statistics	Yes	No	Yes
Uniform criteria across all ages and indices	Yes	Yes	No
Detects changes at ext- remes of distri­butions	Yes	No	Yes

Software can be used to compute per­cent­iles and Z‑scores for each of the indices. Of the software available, both the manual and the program for WHO AnthroPlus 2009 are in the public domain and may be freely copied, translated, and distributed.

13.1.3 Height-for-age difference

There is some debate about using absolute (centimeters) vs. relative (Z‑scores) scales to describe changes in growth in populations of children over time. Leroy and colleagues (2014) argue that although HAZ is useful to assess childen's attained height at a given age, HAZ is inappropriate to evaluate changes in height as children age at both the individual and population level. As shown earlier in the example of the Z‑score calculation of an individual, the construction of a Z‑score for height-for-age uses as a denominator the Z‑score of the reference population. The Z‑score is based on cross-sectional data, and reflects the scatter of height values at a specific age. Such Z‑scores of the reference population are not constant, however, but actually increase linearly during childhood, thus limiting the usefulness of HAZ to assess changes in height over time.

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 is defined as a child's height minus the median height of the growth standard at that age, expressed in centimeters. Leroy and colleagues compared changes in height as populations age based on absolute height-for-age differences (HADs) and HAZs using data from 51 nation­wide surveys from low- and middle-income countries (Leroy et al., 2014). A comparison of these changes based on the two methods for children 1–59mo is shown in Figure 13.6.

In this figure the mean HAZ started below the WHO growth standard at approx­imately −0.4 Z‑scores and fell steadily up to 24mo, after which it stabilized and increased only slightly.

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 and 18mo, 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 18mo, albeit at a lower rate than before 18mo. Indeed, the slopes of the HAD curve provide no indication that the process of growth faltering leveled off even at 5y. The bumps noted in the curves just after 24, 36, and 48mo 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 Carribean, Africa South of the Sahara, South Asia), although the magnitude of the deficits varied between regions. See Leroy et al. (2014) for more details.

Globally, 70% of the absolute deficit accumulated in height (HAD) at 60mo could be attrib­uted to falter­ing during the first “1000” days (conception to 24mos), but 30% was due to continued increases in deficit from age 2–5y. However, such continued increases were masked by the use of HAZ with the age-related changes in  Z‑scores. These findings emphasize the importance of using the mean HAD to provide an accurate assessment of population-level catch-up in linear growth at all ages. For more discussion see Leroy et al. (2014; 2015) and Lundeen et al. (2014).

13.1.4 Percent-of-median

Growth indices can be expressed as percent-of-median value of the reference data when the distri­bution around the median value is unknown. Percent-of-median is the ratio of a measured anthro­pometric value (e.g. weight) in the individual, to the median value of the reference data for the same age or height, expressed as a percentage. It is especially useful when the distri­bution of the reference data has not been normalized, such as the earlier Harvard reference data, and explains why so many of the first classification schemes based on the Harvard reference data (e.g., Gomez and Wellcome classifications) used percent-of-median (Gorstein et al., 1994).

Use of percent-of-median does not provide the same information as the Z‑score, and the relation­ship between these two criteria differs with age and height. Such discrepancies may be especially important if inter­vention activities are prioritized across populations according to the preva­lence of low anthro­pometric indices. A comparison of selected Z‑scores and percent-of-median curves for weight-for-height, height-for-age, and weight-for-age is given in Gorstein et al. (1994).

A further limitation of the percent-of-median is that, unlike per­cent­iles or Z‑scores, the inter­pretation of specific percent-of-median varies across age groups and growth indices Table 13.3. This arises because the calculation of the percent-of-median does not take into account the distri­bution of the data within the reference set, and particularly the differing widths of the distri­butions of the weight-for-age, weight-for-height, and height-for-age indices (Waterlow et al., 1977). Indeed, it is this variability that inhibits the universal use of a constant percentage of the reference median (e.g. 70%) across all ages and for all growth indices. For example, 60% of median weight-for-age represents a more severe state of mal­nut­rition for younger than for older children. Moreover, when the index weight-for-height is used, 60% of the median is inappropriate; such a deficit at any age is incompatible with life (Dibley et al., 1987).

13.2 Use of anthro­pometric indices in clinical settings

Anthro­pometry in clinical settings has the following applications for individuals: screening for early detection of abnormal changes in growth, and assessing the response of an individual to therapy (WHO, 1995). Increasingly, the three conventional growth indices (weight-for-age, length/height-for-age, and weight-for-length/height) are used together with BMI for children age 0–19y, all expressed as either per­cent­iles or Z‑scores. Recumbent length is the recom­mended measure­ment for children younger than 24mo, whereas in those aged ≥ 24mo, standing height should be used.

Health care providers who measure and assess the growth of children or supervise these activities can access the WHO Training Course on Child Growth Assessment which provides instructions on both measuring a child's growth and inter­preting growth indicators manually using the WHO inter­national growth reference charts. The course is available from (WHO).

Alernatively, software can be used to compute per­cent­iles or Z‑scores for the three growth indices (weight-for-age; length/height-for-age; and weight-for-length/height) and BMI-for age for a child aged from 0–19y using two modules — Anthro­pometric calculator and Individual assessment — both available in WHO AnthroPlus 2009. Calculations are based on the WHO Child Growth Standard for children 0–5y (WHO, 2006) and the WHO Growth Reference for children aged 5–19y (de Onis et al., 2007).

13.2.1 Screening to identify abnormal growth

Accurate measure­ments of weight and length or height on each child are needed. WHO has developed an Anthro­pometry Training Video which can be downloaded from WHO. Box 13.3 summarizes some of the points emphasized by WHO to ensure the collection of accurate growth data.

Table 13.4. summarizes examples of growth indicators based on Z‑scores relative to the WHO Growth References which have been shown to reflect growth problems. More insight on the nature of the growth problems of a child can be gained by plotting and examining all the growth measure­ments together on the WHO growth charts.

Table 13.4. Interpretation of growth indicators based on Z‑scores indicative of growth problems. From: (WHO Training Course on Child Growth Assessment). Module C Interpreting Growth Indicators. Notes:
1. A child in this range is very tall. Refer a child in this range for further assessment if you suspect an endocrine disorder (e.g., if parents of the child are of normal height).
2. A child whose weight-for-age falls in this range may have a growth problem, but this is better assessed from weight-for-length/height or BMI-for-age.
3. A child above 1 shows possible risk. A trend towards a Z‑score of + 2 shows definite risk.
4. It is possible for a stunted or severely stunted child to become over­weight.

Z‑score	Length/height- for-age	Weight-for age	Weight-for- length/height	BMI-for-age
Above 3	See note 1	See note 2	Obese	Obese
Above 2			Over­weight	Over­weight
Above 1			Possible risk of over­weight (See note 3)	Possible risk of over­weight (See note 3)
0 median
Below −1
Below −2	Stunted (See note 4)	Under­weight	Wasted	Wasted
Below −3	Severely stunted (See note 4)	Severely under­weight	Severely wasted	Severely wasted

13.2.2 Screening to identify abnormal changes in growth

In general, to identify abnormal changes in growth, measure­ments in a child should be repeated at monthly intervals during the first 6mo, then every 2–3mo during later infancy and early childhood, and then every 6–12mo (Guo et al., 1991). Special attention must be given when measuring length in children below 2 years of age; guidelines are available in FANTA Anthro­pometric Guide (2018).

Such serial measure­ments provide information on the pattern of growth over time. This enables a child with failure to thrive to be identified. Abnormal growth is apparent from the direction of the growth curve: an abnormal height-for-age growth curve is horizontal or, in the case of weight-for-height, moves downwards. The former indicates that the child is not growing, and the latter shows that the child is losing weight. Such a trend may indicate that the child has been ill. Alternatively, the trend may be desirable if the child is over­weight. Consult the WHO Training Course on Child Growth Assessment for more examples.

As an example, Figure 13.7

depicts diagram­matically the growth history of a 7y girl who was following a macrobiotic diet, and whose growth, based on height-for-age, was measured and plotted at least annually on a distance growth chart. Curves for −2 Z‑score, the median, and +2 Z‑score for the growth reference for height-for-age are shown. Growth indices that fall below −2 Z‑score or above +2 Z‑score are defined as abnormal. In Figure 13.7 the child's height-for-age growth curve fell below the −2 Z‑score reference at 2y, at which point the curve became nearly horizontal, indicating that the child was growing very, very slowly.

When a child exhibits deceleration or slow growth of length or height, as shown in Figure 13.7, then the child must be evaluated to distinguish between a shifting of linear growth, sometimes termed “rechanneling”, constitutional growth delay, and under­lying pathology. Any shift in linear growth is determined by genetic influences, and tends to occur within the first 18mos of life, when the growth of an apparently healthy child may cross one or more Z‑score or per­cent­ile lines. In contrast, the exact cause of constitutional growth delay is unknown. It occurs later than rechanneling, and is characterized by a temporary delay in skeletal growth and height of a child, and is not accompanied by any other physical abnormalities causing the delay. Alternatively, slow growth can arise from many pathological conditions, such as mal­nut­rition (Figure 13.7), chronic disease (e.g., tuberculosis), non-organic failure to thrive (caused by a severe psychosocial disturbance in the family), or organic disorders. The latter may encompass congenital and acquired disorders, or abnormal function of the endocrine system or metabolism, and gastrointestinal dysfunction (e.g., malabsorption); details are given in WHO (1995). Guidelines on how to investigate causes of under­nutrition in a child are given in the WHO Child Growth Training Course described earlier.

Although changes in a child's growth are apparent on distance growth charts, abnormal changes in the rate of growth of a child can be detected much earlier when growth velocity charts, rather than distance growth charts, are used: growth velocity measure­ments are much more sensitive to recent changes in growth but require careful measure­ment procedures to be of value. WHO has developed a set of growth velocity charts which are recom­mended for inter­national use (de Onis et al., 2011).

The preva­lence of over­weight and obesity is increasing globally in children younger than 5y, and is an important contributor to diabetes and other non­communical diseases in later life. The WHO Child Growth Training Course provides guidelines on how to investigate causes of over­weight and obesity in infants and children 0–6mo; 6mo–2y; and 2–5y.

Trends in pediatric over­weight and obesity in the United States are an ongoing serious health concern as about 17% of U.S. children and adolescents are obese. In the U.S., the use of BMI in association with the CDC (2000) BMI per­cent­iles are recom­mended by the U.S. Endocrine Society to diagnose over­weight or obesity in children and adolescents > 2y. Over­weight is diagnosed in the U.S. if a child or adolescent > 2y has a BMI > 85^th per­cent­ile but < 95^th per­cent­ile for age, as obese if the BMI is > 95^th per­cent­ile, and as extremely obese if the BMI is > 120% of the 95^th per­cent­ile or > 35kg/m² (Styne et al., 2017). They recom­mend calculating, plotting, and reviewing a child's or adolescent's BMI per­cent­ile at least annually during well-child and/or sick visits, and if a child or adolescent has a BMI > 85^th per­cent­ile, then they should be evaluated for potential comorbidities. The Endocrine Society also provides advice to prevent obesity; for more details see Styne et al. (2017).

13.2.3 Assessing response to therapy

Serial measure­ments are also useful when assessing the response to therapy, as shown in Figure 13.7; again at least two anthro­pometric measure­ments are required. Selection of the indicator of response to the therapy must be made carefully. It must take into account the possible time lag between the start of the inter­vention and the time when a response is apparent.

In the example shown in Figure 13.7, the child was enrolled in a 2y inter­vention program because at age 2y the child's length-for-age growth curve fell below the −2 Z‑score reference, at which point the curve tended to flatten out, indicating that the child was growing slowly. After 2 years of treatment, catch-up growth in response to the therapy is clear, although the child did not approach the −2 Z‑score reference until she was 4y.

An infant with severe acute mal­nut­rition (SAM) is known to respond to a nutrition inter­vention by first putting on weight, only later “catching” up in linear growth. Consequently, the recom­mended discharge criteria from a therapeutic feeding program for a child with SAM defined by a weight-for-height < −3 Z‑score (based on the WHO Child Growth Standard) is when the child has reached a weight-for-height above −2 Z‑score and no edema (WHO/UNICEF, 2009). These discharge criteria are based on the lower risks of mortality compared to those children below −3 Z‑score, as shown in Figure 13.8.

If individuals do not respond to the therapy, then some other medical treatment may be necessary to improve the poor growth, and the earlier therapy should be dis­con­tinued.

13.3 Screening systems to identify individuals at risk in a population

Anthro­pometry has a range of applications that can be used at the population level. Increasingly, screening systems are being used to map countries according to levels of severity of mal­nut­rition (UNICEF/WHO/World Bank, 2021), and to identify priority countries for action. Screening systems are also being used to monitor progress across countries in an effort to mobilize countries to achieve the Sustainable Development Goals.

In public health, it is often not possible to obtain serial measure­ments on the same individuals to identify abnormal changes in growth, as described in the clinical setting in Section 13.2.2. Instead, screening systems are used to identify individuals in a population “at risk” to mal­nut­rition and who may require inter­vention. Some screening systems are used only in emergency settings, when low-cost portable equip­ment and the use of trained unskilled personnel are primary considerations. In such circum­stances, only one anthropmetric measure­ment (e.g., MUAC) may be taken. All the systems use at least one anthro­pometric meaurement and one or more reference limits drawn from appropriate reference data. In some cases, cutoff points based on functional impairment or clinical signs of deficiency or mortality risk may be used.

An optimal screening system is one in which no misclassification occurs; all the individuals designated as “at risk” are actually malnourished (i.e. there are no false positives; sensitivity is 100%). Similarly, those individuals classified as “not at risk” are truly unaffected (i.e. there are no false negatives; specificity is 100%).

In practice, classification schemes are not perfect; some misclassification will always occur so that some individuals identified as “at risk” to mal­nut­rition will not be truly malnourished (false positives), and others classified as “not at risk” to mal­nut­rition will in fact be malnourished (false negatives). Unfortunately, specificity and sensitivity data for the anthro­pometric indices selected are usually not known for the target population. Instead, when selecting a screening system, values for sensitivity and specificity obtained elsewhere are often used and assumed to be appropriate for the target population (Habicht et al., 1979; 1982).

Unfortunately, literature values for sensitivity and specificity are unlikely to apply to the target population when the indices assess an outcome (e.g. low birth weight) which is influenced by several biological factors. For example, low birth weight can be due to both prematurity and intrauterine growth retardation, and its sensitivity as an indicator of neonatal mortality is greater for the former than the latter (WHO, 1995). Factors such as the severity of the deficiency, age, and sex may also affect the magnitude of the expected response of the chosen indicator. A detailed discussion of the selection criteria for indicators can be found in Habicht et al. (1982) and Habicht and Pelletier (1990).

A more general way of selecting an optimal screening system is shown in Figure 13.9,

that depicts the relation­ship between the sensitivity and specificity for both height-for-age and weight-for-height in predicting 2y survival. The position of the Receiver-Operating Characteristic (ROC) curves indicates that, in this case, height-for-age is the better index. For more details on the use of ROC curves, see Chapter 15. Brownie et al. (1986) also provide details of statistical methods that can be used to select the best indicator.

In many of the early classification systems, cutoff points were defined as a designated percentage of the median of the reference population, for the reason outlined earlier. Now the approach preferred by WHO is to use reference limits based on Z‑scores (WHO, 1995). Examples of screening systems that use this approach and are based on a single anthro­pometric index or a combination, and that can be used to classify individuals in a population “at risk” to mal­nut­rition are outlined below.

13.3.1 Use of weight-for-length/height for young children

As noted earlier, low weight-for-length/height is described as “thinness” and reflects a pathological process referred to as “wasting”. It arises from a failure to gain sufficient weight relative to length or height, or from losing weight. The preva­lence of wasting in a population is defined as the proportion of children with a Z‑score for weight-for-length or height less than −2 (i.e., < 2SDs below the age- and sex-specific WHO reference mean). Wasting is said to represent acute mal­nut­rition which should be treated with an acute inter­vention to reverse the wasting and prevent death. The onset of wasting is often highly seasonal with an onset that is fairly acute and often associated with both changes in the food supply and the preva­lence of infectious diseases. The highest preva­lence of low weight-for-length often occurs during the postweaning period (12–23mos) (WHO, 1986).

WHO further classify low weight-for-length/height as:“moderate acute mal­nut­rition” when the weight-for-length/height is between −2 and −3 Z‑score, and “severe acute mal­nut­rition” (SAM) if the Z‑score is below −3 of the WHO Child Growth Standards (WHO/UNICEF, 2009). Treatment of children with moderate acute mal­nut­rition involves improving their existing diets by nutritional counseling to ensure consumption of nutrient-dense foods that meet their additional needs for weight and length gain and functional recovery. Only in settings where there is a high preva­lence of wasting or food insecurity at the community or household level should food supple­ments, usually fortified blended flours, be provided (WHO/UNICEF/WFP/UNHCR, 2009). For more details on the nutrient needs of children with moderate acute mal­nut­rition, see Golden (1990).

Currently, children with SAM (weight-for-length/height Z‑score below −3) are treated most frequently with Ready-to-Use Therapeutic Foods or F75 and F100 milk-based diets; for more treatment details see WHO (2013). WHO recom­mends that all children with SAM who also have medical complications, severe edema, poor appetite, or present with one or more Integrated Management of Childhood Illness danger signs, should be admitted to hospital for inpatient care. Transfer of children with SAM from inpatient to outpatient care should be based on the clinical condition of the children and not on specific anthro­pometric outcomes. Justification for the use of weight-for-length/height Z‑score below −3 for identifying SAM is shown in Box 13.4.

WHO provide weight-for-length reference cards for children less than 87cm, and 87cm and above, by sex, that can be used to assess the Z‑score of a child in the field once their weight (kg) and height (cm) have been measured. These cards are available from WHO.

As noted earlier, the WHO recom­mended criterion used for discharge when weight-for-height has been used as an admission criterion is when weight-for-height reaches > −2 Z‑score with no pitting edema (WHO, 2013). This discharge criterion is based on the lower risks of mortality compared to those children below −3 Z‑score as shown in Figure 13.8. Percentage weight gain should no longer be used as a discharge criterion (WHO, 2013).

Note that use of the new WHO Child Growth Standard in low income countries has resulted in a 2–4 times increase in the number of infants and children falling below −3 Z‑score compared to using the former NCHS / WHO reference population published by WHO in 1983.

Over­weight in young children is becoming increasingly common, as noted earlier. Although WHO defines any child as over­weight with a weight-for-length/height Z‑score above +2, and as obese with a weight-for-length/height Z‑score above +3, increasingly BMI-for-age Z‑scores are being used to define risk of over­weight, over­weight, and obesity in children.

13.3.2 Use of MUAC for children

Mid-upper-arm circumference (MUAC) is used for screening individual children for targeting interventions in the community when weight and stature measure­ments are impossible and the precise age of the child is unknown. In such cases, a fixed cutoff point for MUAC is often used. For example, to diagnose children with SAM aged 6–60mo, WHO (2009) recom­mends a fixed MUAC cut-off point of 115mm as one of the independent criteria that can be used by trained community health workers and community members. This MUAC cutoff was chosen because children with a MUAC less than 115mm were observed to have a highly elevated risk of death compared to those with a MUAC above 115mm (Myatt et al., 2006). In addition, the preva­lence of SAM based on a MUAC cutoff of 115mm and that defined by a weight-for-height below −3 Z‑score based on the new WHO Child Growth Standard is believed to be similar, based on an analysis of measure­ments on more than 450,000 children aged 6–60mo in 31 countries.

Table 13.5.

Table 13.5. Diagnostic criteria for severe acute mal­nut­rition in children aged 6–60mos. From: WHO/UNICEF, 2009.
¹ Based on Child Growth Standards (WHO).
^2,3 Independent indicators of Severe Acute Malnutrition that require urgent action. From: WHO/UNICEF, 2009.

Indicator	Measure	Cut-off
Severe wasting2	Weight-for-height1	< −3 Z‑score
Severe wasting2	MUAC	< 115mm
Bilateral edema3	Clinical sign

summarizes the WHO diagnostic criteria for SAM in children aged 6–60mos. To identify children in the community with SAM for treatment, WHO recom­mends the measure­ment of MUAC together, where possible, with examination for bilateral pitting edema as the preferred dignostic criteria. Provision to mothers of MUAC tapes with a cut-off value of 115mm have been used in an effort to detect mal­nut­rition early before the onset of complications, and thus reduce the need for inpatient treatment (Blackwell et al., 2015).

The WHO recom­mended discharge criteria from treatment for children diagnosed with SAM using the MUAC cutoff is based on a MUAC > 125mm and no edema for at least 2wk. If children have been admitted based only on bilateral pitting edema, then the criteria for discharge should be the one used in programs. To avoid any relapse, children with SAM who have been discharged from treatment should be monitored periodically (WHO, 2013).

Concerns have been raised over the use of the MUAC cutoff alone as a prognostic indicator of mortality. As an example, Grellety and Golden (2018) claim that up to 45% of SAM children at high risk of death will not be identified using this approach. This finding was based on an examination of the relative mortality rates of children 6–60mos who had SAM (n=76,887) classified by the three diagnostic criteria depicted in Table 13.5. Instead, these investigators urge that the use of weight-for-height Z‑scores should be retained as an independent criterion for diagnosis of SAM, and innovative methods developed to identify those children with a low weight-for-height Z‑score but not a low MUAC in community screening programs. In the interim, they suggest that use of both a weight-for-height < −3 Z‑score and MUAC < 115mm should be retained and used routinely as diagnostic criteria for SAM in an effort to identify those children who should receive treatment.

Use of a fixed cutoff assumes that MUAC is relatively independent of age for children 1–5y (Burgess and Burgess, 1969). This assumption has been questionned by several investigators. Data from both affluent and nonaffluent populations show that MUAC is age dependent, resulting in over-diagnosis of wasting among younger children and under-diagnosis among older ones, when a fixed cutoff point is used (de Onis et al., 1997). In view of these concerns, WHO has developed MUAC-for-age reference data for children 6–60mo (Figure 13.10).

The curves show important age-specific differences, as well as significant sex-specific differences for boys and girls < 24mo. In the field, the use of MUAC-for-age reference data is no more difficult than using weight-for-height, the conventional index of wasting.

13.3.3 Use of MUAC and Calf Circumference for adults

The importance of adult under­nut­rition is becoming increasingly recognized. Both the Food and Agricultural Organization (FAO) and WHO have adopted BMI to assess adult under­nut­rition and have defined cutoffs grading the severity of low BMI in adults. However, in emergency situations, measure­ments of weight and height, required to calculate BMI, are difficult to obtain from severely emaciated individuals who often cannot stand unaided. Therefore, MUAC and calf circumference,  measure­ments that only require a tape measure, have been investigated as a substitute for BMI.

Significant correlations between measure­ments of MUAC and BMI have been demonstrated in women of reproductive age (Khadivadeh, 2002), and severely under-nourished adult populations from low-income countries (Collins, 1996). Reports of positive associations between calf circumference and BMI have been primarily reported in clinical settings and mostly in the elderly (Aparecida et al., 2012). However, there is little known about the relation­ships between both MUAC and calf circumference (CC) and health outcomes in adults.

WHO has proposed universal cutoff values for MUAC of 22cm for females and 23cm for males, and below 31cm for calf circumference (for both sexes) to screen for under­nutrition (WHO, 1995). However, in view of the reported regional and ethnic variations in body composition, these cutoff values may not be appropriate for Asian populations who have a higher proportion of body fat than other races (e.g., Caucasians and Africans) (Deurenberg et al., 1998) (see Chapter 11).

As an example, in a study in northern Vietnam of women of reproductive age (n=4,981), a higher MUAC cutoff (23.5cm) was used to screen for under­weight in comparison to the BMI. This 23.5cm MUAC cutoff had a sensitivity of 89.1%, a specificity of 71%, and a positive predictive value of 84% when analyzed using ROC curves. If the lower WHO cutoff (22cm) had been used, many women at risk for under­nutrition would have been misclassified as healthy. By contrast, the calf circumference cutoff proposed by WHO (31cm) was found to be appropriate, although slightly less accurate than the MUAC cutoff for predicting under­weight (Nguyen et al., 2014). More studies are needed to examine the utility of MUAC and calf circumference in predicting functional outcomes.

MUAC during pregnancy has also been investigated as a screening tool for risk of low birth weight and late fetal and infant mortality. Maternal arm circumference is relatively stable during pregnancy, so its measure­ment is independent of gestational age (Mohanty et al., 2006). Several investigators have reported that MUAC, when used as a proxy of nutritional status of women during pregnancy, is a strong predictor of maternal mortality (Christian et al., 2008) and fetal outcomes such as low birth weight (Mohanty et al., 2006; Assefa et al., 2012). In contrast, calf circumference is not recom­mended as a proxy measure­ment of status during pregnancy. Edema is increasingly common as pregnancy advances, so calf circumference may be increased particularly in late pregnancy (WHO, 1995).

13.3.4 MUAC-for-height: QUAC stick

The Quaker arm circumference measuring stick (QUAC stick) was developed as a rapid, cheap, and simple screening tool for the nutritional assessment of children. It avoids the use of cumbersome scales and minimizes errors in measuring height (Arnhold, 1969). The measuring stick is used to compare a child's MUAC with the two reference limits of the MUAC reference data corresponding to the child's height.

The modified QUAC stick developed by WHO uses MUAC-for-height data from the MUAC-for-age reference for children < 5y (de Onis et al., 1997; Mei et al., 1997). Three sets of MUAC-for-height curves were developed: one for boys, one for girls, and one for both boys and girls com­bined. The curves are for children of height 65–145cm. Recumbent length measure­ments were adjusted to standing height when the MUAC-for-height reference data were constructed to eliminate any discrepancy caused by differences in the two stature measure­ments.

Mei et al. (1997) also provide instructions on the construction and use of the modified QUAC stick shown in Figure 13.11

that uses the WHO MUAC-for-height reference data. Construction of two sticks, one for the boys and one for girls, is recom­mended.

The WHO-modified QUAC stick consists of a vertical stick, about 150cm long and 3 × 3cm in cross-section. One face is marked from 0 at the bottom up to 145cm near the top in 0.5cm increments.   On one adjacent face, the corresponding reference values for the median −2 Z‑score of MUAC-for-height are marked; the other face, adjacent to the height measure­ments, shows the values for median −3 Z‑score of the MUAC-for-height.

To use the QUAC stick, it is first placed firmly upright on a platform against a vertical wall. The child is then asked to stand straight with his or her back against the height measure. Next, the MUAC of the child is measured using the technique described in Chapter 11. Then a note is made about whether the value is below that given on the “median −3 Z‑score” face at the point on the stick marked by the child's height. If the measure­ments is below this value, the child is noted as having a severe nutritional deficit; if above and yet still below the appropriate value on the “median −2 Z‑score” face, the deficit is recorded as moderate. In this way, all children are identified as having severe, moderate, or no deficit. Note that the method eliminates the necessity of recording both the height and MUAC of the child for subsequent comparison with reference data.

The performance of MUAC-for-height for screening was evaluated using weight-for-age < −2 Z‑score as the “gold standard” for identifying malnourished children (Mei et al., 1997). Results indicated that the sensitivity and specificity for MUAC-for-height was better than MUAC based on a fixed cutoff for screening malnourished children 6–59mo from Sri Lanka, Nepal, Togo, and Malawi.

13.3.5 Weight-for-height wall chart

A wall chart developed by Nabarro and McNab (1980) is based on weight-for-height and three percentage ranges of the corresponding median weight-for-height of the earlier NCHS reference data. These are 70%–80%, 80%–90%, and 90%–110%. Wasted children are identified as those with a percentage weight-for-height less than 80% of the reference median. The chart has the advantage of visually identifying the extent of wasting in a very simple manner. In addition, during nutritional rehabilitation, a target weight can be readily identified.

The chart is fixed to a smooth wall, and the child is positioned on a level floor against the chart at a point corresponding to the child's weight (Figure 13.12).

The top of the child's head will fall into one of three color-coded categories on the chart, equivalent to 70%–80% (red), 80%–90% (yellow), and 90%–110% (green) of the expected weight-for-height and representing various degrees of mal­nut­rition. The color should be read off the chart at the point at which the examiner's fingers touch the column marked with the child's weight.

For small children, two persons may be required to position the child correctly for the measure­ment. To avoid misclass­ifying children with edema or ascites (fluid in the abdominal cavity), a brief clinical examination should also be under­taken. The charts can be modified to include different colors, numbers of colors, and alter­native growth reference data and reference limits, depending on the resources and needs of the population.

13.3.6 Use of Body Mass Index for children

In view of the increasing worldwide preva­lence in under-five children of over­weight and obesity, and their subsequent role in diabetes and other chronic diseases in adulthood, WHO has endorsed “reduce and maintain childhood overweight to less than 3%” as one of the six global nutrition targets (Box 13.1; WHO, 2014). The worldwide preva­lence of over­weight and obesity among preschool children has increased from 4.2% (95% CI:3.2%, 5.2%) in 1990, to 6.7% (95% CI: 5.6%. 7.7%) in 2010, and is expected to have reached 9.1% (95% CI:7.3%, 10.9%), or about 60 million in 2020 (de Onis & Lobstein, 2010).

Body mass index (BMI) is considered the most practical, universally applicable, and non-invasive index of over­weight and obesity. An increased BMI-for-age in childhood and adolescence is associated with higher percentages of body fat and known risk factors for cardiovascular disease.

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 > +3, based on the WHO Child Growth Standard (WHO, 2006; de Onis & Lobstein, 2010). For children aged 5–19y, BMI-for-age Z scores above +1 and above +2 based on the WHO 2007 growth reference data (de Onis et al., 2007) are applied to define over­weight and obesity, respectively.

In the United States, use of BMI in association with the CDC (2000) BMI per­cent­iles are recom­mended by the U.S. Endocrine Society to diagnose over­weight or obesity in children and adolescents > 2y (Kuczmarski et al., 2000), as noted earlier (Section 13.2.2); (Figure 13.13). The BMI-for-age per­cent­iles (5^th, 10^th, 25^th, 50^th, 75^th, 90^th, and 95^th) are available for children aged 2–20y. The weight data for children > 6y who participated in the U.S. NHANES III survey were excluded from these BMI growth charts because the inclusion of these data shifted the upper per­cent­ile curves. Consequently, the CDC (2000) growth reference is not a purely descriptive growth reference because it does not contain representative data for the BMI variables.

13.3.7 Use of Body Mass Index for adults

Over the past few decades both low- and middle-income countries have faced a double burden of malnutrition as a result of the rapid increase in the preva­lence of overweight together with a persistent preva­lence of under­weight. For example, in 2016, the global estimate among adults for overweight or obesity was more than 33% while under­weight was 10% (NCD Risk Factor Collaboration, 2017). Both conditions are associated with increased morbidity and mortality and increase the risk of developing non-communicable diseases (Zhang et al., 2021; NCD Risk Factor Collaboration, 2017). This increasing trend for a double burden of malnutrition has been attrib­uted to unequal changes in food production, dietary patterns, and physical activity as communities have encountered the nutrition transition (Hoque et al., 2015).

Body mass index (BMI) is the most widely used index of over­weight and obesity in adults as well as children. Body mass index is relatively unbiased by height and generally correlates with laboratory-based measures of adiposity in most younger and older adults, as shown in (Figure 13.14).

WHO has recom­mended the use of a graded classification of over­weight and obesity to:

• Compare weight status within and between populations
• Identify individuals and groups who are at increased risk of morbidity and mortality
• Identify priorities for inter­vention at both the individual and community levels
• Allow interventions to be evaluated appropriately

Table 13.6 shows the classification for under­weight as well as the graded classification of over­weight and obesity in adults based on BMI and recom­mended for inter­national use by WHO (2000).

Table 13.6. WHO classification of obesity in adults according to body mass index (BMI). From: WHO (2000).

Classification	BMI (kg/m2)	Risk of comorbidities
Under­weight	< 18.50	Low (but risk of clinical problems is increased)
Normal range	18.50–24.99	Average
Over­weight	≥ 25.00
Pre-obese	25.00–29.99	Increased
Obese class I	30.00–34.99	Moderate
Obese class II	35.00–39.99	Severe
Obese class III	≥ 40.00	Very severe

WHO (2000) also recom­mend that the optimum population mean BMI should be between 21.0 and 23.0kg/m². The classfication in Table 13.6 is based primarily on the association between BMI and mortality, after taking into account the confounding effects of cigarette smoking and co-existing chronic disease.

The general term “over­weight” is used by WHO (2000) to describe all individuals with BMI ≥ 25kg/m². Individuals with BMI from 25.0 to 29.9 are further termed “pre-obese”. Three classes of obesity are defined based on BMI cutoffs of 30, 35, and 40kg/m². This classification system is not intended for use with pregnant and lactating women, or persons < 18y.

Although WHO recom­mends the universal use of these cutoffs inter­nationally (WHO Expert Consultation, 2004), nevertheless WHO recognized that a substantial proportion of Asians were at high risk of type 2 diabetes and cardiovascular disease with BMIs lower than the existing WHO cutoff for over­weight (i.e., ≥ 25kg/m²). (Figure 13.15).

Indeed, some experts have set lower cutoffs for over­weight (i.e., 23–27.4kg/m²) and obesity (> 27.5kg/m²) for Asians based on their cardiovascular and type 2 diabetes risks (Al Kibria, 2019). However, WHO has not set a lower cutoff for Asians in view of the evidence that the risk appears to vary in different Asian populations. Instead, WHO recom­mends that a broad range of BMI categories should be used for reporting the preva­lence of over­weight in countries.

Several investigators have studied the deter­minants of both under­weight and over­weight in low- and middle-income countries in an effort to develop country-specific strategies to reduce or halt the double burden of mal­nut­rition. (Al Kibria, 2019), for example, using data from the national Demographic and Health Survey (DHS) in Nepal, examined the preva­lence and factors affecting under­weight, over­weight, and obesity among adults in 2016. They followed the WHO recom­mendation and estimated the preva­lence by sex over a broad range of BMI categories which included three for under­weight and seven for over­weight, as shown in Table 13.7.

Table 13.7. Prevalence (with 95% confidence interval) of BMI categories among adult men and women, Nepal Demographic and Health Survey 2016. BMI: body mass index in kg/m², IQR: Inter-quartile range.
^a Asian cutoffs for over­weight and obesity start from these ranges, respectively.
^b World Health Organization cutoffs for over­weight and obesity start from these ranges, respectively. Both cutoffs define under­weight as < 18.5kg/m². From: (Al Kibria, 2019).

BMI cut-offs	Men	Women
≤ 16.00	2.1 (1.7–2.5)	3.5 (3.1–4.0)
16.00–16.99	2.9 (2.5–3.4)	3.9 (3.5–4.4)
17.00–18.49	10.1 (9.3–10.9)	10.5 (9.8–11.2)
18.50–22.99	49.8 (48.5–51.2)	43.2 (42.0–44.3)
23.00–24.99 a	15.9 (14.9–16.9)	14.1 (13.3–14.9)
25.00–27.49 b	11.5 (10.7–12.4)	11.6 (10.9–12.3)
27.50–29.99 a	5.2 (4.6–5.8)	7.7 (7.1–8.3)
30.00–32.49 b	1.9 (1.6–2.3)	3.4 (3.0–3.8)
32.50–34.99	0.1 (0.0–0.7)	1.5 (1.3–1.8)
35.00–37.49	0.1 (0.0–0.7)	0.5 (0.3–0.7)
37.50–39.99	0.1 (0.0–0.2)	0.1 (0.1–0.2)
≥ 40.00	0.1 (0.0–0.1)	0.1 (0.0–0.2)

In Nepalese adults, the overall preva­lence of under­weight (I.e., BMI < 18.5kg/m²) was 16.7%, higher than the global preva­lence, but comparable to the preva­lence in South Asia (NCD Risk Factor Collaboration, 2017), whereas that of over­weight (18.2%) and obesity (4.3%) was lower than the global average; 45.9% had a BMI within the normal range. Age, sex, and education level, household wealth status, place, ecological zone, and provinces of residence were all deter­minants of the extreme body weight categories (i.e., under­weight, over­weight, and obesity). Overall, higher education level and wealth status were positively associated with over­weight / obesity, and inversely associated with under­weight. Women had a higher preva­lence of being both under­weight and over­weight / obese compared to males, consistent with earlier studies (Flegal, 2006).

Successive national DHS surveys have been used to track the rates of change in the preva­lence of under­weight and over­weight in some countries. As an example, based on a meta-analysis of data on women of reproductive age from five successive national DHS surveys in Bangladesh from 1996 to 2011, the overall total preva­lence of under­weight decreased from 47% to 24% over this time period, whereas that of over­weight increased from 4% to 16%. Moreover, this shift in BMI from under­weight to over­weight from 1996 to 2011 was most strongly associated with urban residence, age, higher socioeconomic status, and higher education attainment (Hoque et al., 2015). (Figure 13.16).

The increasing double burden of under­weight and over­weight in women during the preconception period is especially concerning as both are known to contribute to maternal and fetal complications during pregnancy (Salihu et al., 2009; Han et al., 2011). In a systematic review and meta-analysis of 34 studies, pre-pregnancy under­weight (defined by a BMI < 18.5) contributed to a 32% higher risk of preterm birth, while for over­weight women, the risk of preeclampsia and gestation diabetes mellitus approx­imately doubled, an effect that was even greater among women with pre-pregnancy obesity. These adverse effects of both under­weight and over­weight in the preconception period on pregnancy outcomes are likely to be amplified in adoles­cents or women with closely-spaced pregnancies and thus depleted nutrient reserves. These findings emphasize the importance of appropriate interventions to optimize maternal BMI in the pre-conception period (Dean et al., 2014).

13.3.8 Composite Index of Anthro­pometric Failure (CIAF)

The three conventional growth indices — weight-for-age, weight-for-height, and height-for-age — are described in Section 13.1; additional details are available in Chapter 10. Each index provides important information on different aspects of under­nutrition (e.g., chronic versus acute) and can be used to select the most appropriate inter­vention.

In many low income countries, weight-for-age is the most commonly used growth index for children, with the proportion under­weight (i.e., weight-for-age Z‑score < −2) being used extensively globally as a proxy indicator to monitor progress in meeting the Millennium Development Goal to eradicate hunger. However, as discussed earlier and in detail in Chapter 10, under­weight can be caused by stunting or wasting or a combination thereof and does not distinguish between them. Therefore, under­weight when used on its own does not identify the sum of those children who are stunted and/or wasted, and thus provides an under­estimate of the overall magnitude of under­nutrition among children in the population. This is unfortunate because for policy and planning in a country, it is important to under­stand the overall scale of the problem of under­nutrition to ensure sufficient resources can be allocated.

Consequently, an alter­native classification system — termed the Composite Index of Anthro­pometric Failure (CIAF), has been developed by Nandy and Svedberg (2012), and is presented in Table 13.8.

Table 13.8. Categories of the composite index of anthropometric failure (CIAF). Modified from Savanur and Ghugre (2015).

Group	Description of the group	Definition
A	No anthro- pometric failure	Normal WAZ, HAZ, WHZ
B	Wasting only	WHZ < −2 SD but normal WAZ and HAZ
C	Wasting and under­weight	WHZ and WAZ < −2 SD but normal HAZ
D	Wasting, under- weight & stunted	WHZ, WAZ and HAZ < −2 SD
E	Stunting and underweight	HAZ and WAZ < −2 SD normal WHZ
F	Stunting only	HAZ < −2 SD normal WAZ and WHZ
G	Stunted and overweight	HAZ < −2 SD and WHZ > +2 SD
H	Overweight only	WHZ > +2 SD and WAZ > +2 SD normal HAZ
Y	Underweight only	WAZ < −2 SD normal HAZ and WHZ

The original CIAF provided an aggregate measure to estimate the overall burden of under­nutrition based on children whose weight and height were below the WHO Child Growth Standard.

Comparison of the percentage of children classified as under­nourished, based on the three conventional growth indices, and the CIAF using the new WHO Child Growth Standards for children under 5y have consistently shown more under­nourished children using the CIAF than all the conventional indices (Nandy and Svedberg, 2012; Savanur and Ghugre, 2015). Children exhibiting multiple anthro­pometric failures are especially vulnerable.

With the emergence of the dual burden of mal­nut­rition, the original CIAF model of classification for under-nutrition is no longer sufficient. This has led to the extension of the CIAF model include two new groups: “stunted and over­weight”; and “over­weight only”, as shown in Table 13.8. In some studies, BMI‑Z score has been used to define over­weight instead of WHZ > +2 SD, with the BMI‑Z score cutoff varying according to the age group of the children (Bejarano et al., 2019). The CIAF extended model provides policy-relevant information on the pattern of child malnutrition in any country.

As an example, the use of the extended CIAF (ECIAF) to classify mal­nut­rition due to both under­nutrition and over-weight in children (n=10,679) age 3–13.99y was explored in a study involving six Argentine provinces. The percentage of preschool children 3–4.99y with mal­nut­rition based on ECIAF was lower than in the school children 5–13.99y (15.1% vs. 28.4%). In the whole sample, about 25% of the mal­nut­rition was caused by under­nutrition and 75% by over-weight (Bejarano et al., 2019). Kuwornu et al. (2020) also used the ECIAF to measure the overall burden of early childhood malnutrition in Ghana from 2008–2011 for children 6–59mos based on three national surveys.

13.4 Applications of anthropometry in public health

The applications of anthropometry commonly used in public health are itemized in Box 13.5. Details of these applications are discussed briefly below. Recommendations for data collection, analysis and reporting of anthro­pometric indicators in children under‑5y are available from WHO.

The reader is strongly recom­mended to follow these WHO guidelines to ensure harmonized methodology and reporting and thus facilitate comparisons of growth inter­nationally. The WHO recom­mended age groups, (Table 13.9), Z‑score cutoffs for abnormal growth, and the use of flags to identify implausible values should always be used. Details on how to identify and report implausible Z‑score values are given in Section 13.1.2.

Table 13.9. Suggested age groups for use in the presentation of anthro­pometric data. From: WHO/UNICEF (2019).

Group	Age Classes
Newborn children, infants and pre-school children	0 to < 6mos 6 to < 12mos 12 to < 24mos 24 to < 36mos 36 to< 48mos 48 to < 69mos

Three methods have been recom­mended by WHO to evaluate cross-sectional anthro­pometric data for use in public health. These three methods are itemized in Box 13.6. To calculate the preva­lence estimates for stunting and over­weight based on national population data, the reader is advised to use the WHO tracking tool to set national targets and monitor progress. Note wasting is excluded from the tracking tool because of its high short-term variability. See WHO Global Target Tracking tool for more details. However, these methods can also be augmented with more complex statistical techniques to identify the deter­minants and consequences of mal­nut­rition, and to set targets and/or assess the response to interventions, depending on the study objectives.

13.4.1 inter­national comparisons of growth

(UNICEF/WHO/World Bank Group, 2017) highlight the levels and trends in child mal­nut­rition across countries based on the thresholds of severity shown in (Table 13.10).

Table 13.10. Prevalence thresholds and corresponding labels for wasting, over­weight and stunting in children under 5 years using the “novel approach”. From: de Onis et al. (2019).

Wasting		over­weight		Stunting
Prevalence thresholds (%)	Labels	Prevalence thresholds (%)	Labels	Prevalence thresholds (%)	Labels
< 2·5	Very low	< 2·5	Very low	< 2·5	Very low
2·5–< 5	Low	2·5–< 5	Low	2·5–< 10	Low
5–< 10	Medium	5–< 10	Medium	10–< 20	Medium
10–< 15	High	10–< 15	High	20–< 30	High
≥ 15	Very high	≥ 15	Very high	≥ 30	Very high

As an example, Figure 13.17 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 13.10. Currently, no estimates exist on children suffering simul­taneously from two forms of mal­nut­rition such as stunting + over­weight or stunting + wasting.

As an alter­native to only depicting a subset, the mean Z‑score and standard deviation of the entire distri­bution of HAZ‑scores can be calculated. If the mean Z‑score is significantly lower than zero — the expected value for the reference distri­bution — then usually the entire distri­bution has shifted downward suggesting that most, if not all of the individuals, have been affected. This can be confirmed by plotting the entire distri­bution of the Z‑scores against the reference distri­bution as shown in Figure 13.18. Here all the children, with an overall mean Z‑score of −1.83, were affected by some degree of linear growth retardation and an inter­vention is required for the entire community. This is termed a “population approach to targeting”. In contrast, the stunted children, as defined by HAZ < −2 were only a subset of those with linear growth retardation.

With the recognition that anomalies in the distri­bution of abdominal fat are as great a risk factor for disease as is excess body fat per se, waist circumference is increasingly used as a surrogate estimate of intra-abdominal fat content in adults, often in combination with BMI as a surrogate indicator of obesity. However, to date, the use of this combination in large scale population studies is limited.

Adult cutoffs for waist circumference that vary by sex and ethnicity have been developed (M: ≥ 102cm, F: ≥ 88cm, WHO, 2000) with lower cutoffs proposed for Asians (M: ≥ 90cm, F: ≥ 80cm, Lear et al., 2010) than Europeans, as reported for BMI. Nevertheless, more research is needed to establish whether specific cut-offs are needed for African-American, Hispanic, and Middle Eastern populations (Lear et al., 2010); see Chapter 11 for more details.

13.4.2 Identifying the deter­minants and consequences of mal­nut­rition

The main direct deter­minants of mal­nut­rition are frequently dietary intake and infection, both of which are affected by under­lying deter­minants such as household socio­economic conditions, food security, education, health services, or water, sanitation and hygiene (i.e., WASH). Anthro­pometric indices can be used in conjunction with household and WASH data, to examine factors associated with mal­nut­rition as well as its consequences. Anthro­pometric indices are especially useful in extreme acute situations such as famine, as the deter­minants of very high levels of wasting are severe energy deficits. The latter can be confirmed by a rapid reduction in the preva­lence of low weight-for-height following re-feeding.

In non-disaster situations, however, the causes of wasting, stunting, or over­weight are much more difficult to identify because multiple factors may be involved, making it important to collect information or measure all potential risk factors. Aguayo et al. (2016) investigated the deter­minants of stunting and poor linear growth in children in the state of Maharashtra, India where 22.7% of children 0–23mos were stunted, of whom 7.4% were severely stunted. They used data from the 2012 comprehensive nutrition survey which included questionnaires focussed on the households, mothers, and infant and young child feeding (IYCF) practices, using the standardized IYCF indicators developed by WHO (2008).

In the state of Maharashtra, the mean HAZ was significantly lower (−1.74 vs. −0.25) and the preva­lence of stunting significantly higher (40.5% vs. 9.2%) among children age 18–23mos than among children age 0–5mos. Multivariate regression analyses suggested that the most consistent deter­minants of stunting and poor linear growth in these Indian children < 23mos were birth­weight and child feeding, women's nutrition and status, and household sanitation and poverty; for more specific details see Aguayo et al. (2016). Based on these results, policies, programs, and investments to improve linear growth and reduce the preva­lence of stunting in the State of Maharashtra should be prioritized in four key areas, as shown in Box 13.7.

In a more recent study, factors associated not only with stunting but also with wasting and under­weight, were each examined in children aged 12–59mos using Demographic and Health Survey data (2007–2018) from 35 low- and middle-income countries (Li et al., 2020). The comparative importance of both direct and under­lying factors associated with anthro­pometric deficits were investigated. In most countries, irrespective of the anthro­pometric indicator applied, household socio-economic status (assessed by household wealth and maternal education) and parental nutritional status (i.e., height and BMI) were the strongest factors associated with anthro­pometric under­nutrition, consistent with findings elsewhere (Li et al., 2017). Nevertheless, several other factors associated with anthropometric under­nutrition were also identified and were heterogeneous among countries. These findings highlight the importance of under­standing the under­lying mechanisms of child under­nutrition within a specific country prior to formulating inter­vention policies and programs to reduce child under­nutrition.

Over­weight and obesity in childhood is also of increasing public health concern in both developed and developing countries, as noted earlier. Hence, identifying modifiable risk factors for child over­weight and obesity that can be targeted through public health interventions is also important. In a report of Kenyan preschool children between 3–5y (n=1495) based on data from a nation-wide Demographic Health Survey (DHS), Gewa (2010) examined the associations between childhood over­weight and obesity and selected maternal and child-related factors. Maternal over­weight and obesity, higher levels of maternal education, being a large or very large child at birth, and being stunted were each associated with higher odds of over­weight and obesity among these Kenyan preschoolers. Some of these factors are modifiable and could be targeted through appropriate public health interventions. For more discussion, see Gewa (2010).

Nevertheless, the interpretation of such cross-sectional data on the deter­minants of stunting and over­weight is difficult. Anthropometry and the data from the households, mothers, and young children are collected at the same time during a cross-sectional survey, whereas both stunting and over­weight reflect long-term cumulative influences which may have existed for a long time before the households were surveyed.

Child stunting, wasting, under-weight or over­weight each have short- and long-term health consequences which are well documented (Black et al., 2008; Nahar et al., 2020). The capacity of anthro­pometric indicators to predict the risk of these adverse health consequences has been the subject of much research; it depends on several factors. These include the indicator selected, the specific risk under question, the age of the children, the baseline levels of the anthro­pometric deficit, and the health status in the study population at baseline. The findings vary according to the study setting. Nonetheless, information from these investigations is important because they can promote the survival of these “at risk” children through appropriate interventions.

Many settings in developing countries are characterized by poverty, infections, and inadequate dietary intakes. As a result, children may exhibit multiple anthro­pometric deficits simultaneously, which may amplify their risk for adverse health outcomes such as morbidity or mortality. To address this concern, the association between simultaneous anthro­pometric deficits and the risk of adverse health consequences have been studied by some investigators.

Table 13.11. Subgroups of anthro­pometric failure among Indian children 0–3y (n=24,396), with scores of < 14: low standard of living; scores of 15–24: medium standard; and scores 25–67; high standard. Data from Nandy et al. (2005).

Group	% of Children	Mean SLI (95% CI)
A (no failure)	40.2	22.4 (22.2, 22.6)
F (stunting only)	10.1	20.1 (19.7, 20.5)
Y (underweight only)	5.9	19.4 (18.9, 19.9)
B (wasting only)	2.6	19.3 (18.6, 20.1)
C (wasting and underweight)	6.1	18.3 (17.8, 18.8)
E (stunting and underweight)	27.9	16.9 (16.7, 17.1)
D (wasting, stunting and underweight)	7.2	15.3 (14.9, 15.7)

As an example, Nandy and co-workers (2005) employed the composite index of anthro­pometric failure (CIAF, see Section 13.3.8) and a standard of living index (SLI) for each child to investigate the deter­minants and consequences of under­nutrition among children < 3y in the 1998–1999 Indian National Family Health Survey. The SLI was based on a range of 30 household assets and possessions and used as a proxy for household wealth and economic status. Table 13.11 shows that the under­nourished children in each of the six subgroups were found to have a lower mean SLI score compared to children in the referent group (Group A). Note the subgroup of children with wasting, stunting and under­weight (Group D) had the lowest mean SLI score, and those children with multiple anthro­pometric failures (i.e., those in subgroups C, E, and D) had lower mean SLI scores than those with single failures (Groups F, Y and B), or no failure (Group A).

Nandy et al. (2005) also collected information on child morbidity from maternal reports on the two weeks preceeding the survey, as recom­mended by WHO (1994; 1999). Three indicators of morbidity were constructed: diarrhea (i.e., reported in last two weeks); severe diarrhea (i.e., blood in stools); and acute respiratory infection (i.e., cough followed by breathing difficulties). Age-adjusted binary logistic regressions for anthro­pometric failure, diarrhea, severe diarrhea and acute respiratory infection (Table 13.12)

Table 13.12. Age-adjusted binary logistic regression analysis for anthro­pometric failure, diarrhea, severe diarrhea, and acute respiratory infection among Indian children 0–3y. Information on morbidity was provided by the child's mother or caregiver.
CI = confidence interval.      ^b p < 0.001     ^c p < 0.05.
Data from Nandy et al. (2005).

	Diarrhea	Severe diarrhea	Acute res- piratory infection
Group	Odds ratio 95% CI	Odds ratio 95% CI	Odds ratio 95% CI
A (no failure)	1.00	1.00	1.00
B (wasting only)	1.06 (0.86–1.30)	0.45 (0.19–1.10)	0.92 (0.75–1.14)
C (wasting and under­weight)	1.45b (1.27–1.65)	1.19 (0.82–1.73)	1.28b (1.12–1.46)
D (wasting, stunting and under­weight)	1.72b (1.52–1.95)	1.95b (1.45–2.61)	1.39b (1.23–1.58)
E (stunting and under­weight)	1.54b (1.42–1.67)	2.03b (1.66–2.49)	1.34b (1.24–1.46)
F (stunting only)	1.04 (0.92–1.17)	1.08 (0.78–1.48)	1.02 (0.91–1.14)
Y (under­weight only)	1.19c (1.03–1.37)	1.64 (1.17–2.29)	1.16c (1.01–1.33)

showed children who were simultaneously wasted, stunted, and under­weight (i.e., Group D) had the highest risk of both diarrhea and acute respiratory infection. These findings suggest that children with multiple anthro­pometric failures are at a greater risk of morbidity and more likely to come from poorer households, as reported elsewhere (Baig-Ansari et al., 2006). However, because this was a cross-sectional study with the morbidity cases recorded before the anthro­pometric measure­ments were taken, the possibility that the associations observed arose from reverse causality (i.e., restricted growth could be due to previous infection), cannot be ignored.

The associations of child stunting, wasting, and under­weight with increased risk of mortality are well recognized (Fawzi et al., 1997). In many cases, these studies have been cross-sectional, as noted earlier, with mortality estimates based on the effects of individual anthro­pometric indicators. However, as highlighted by Nandy et al. (2005), multiple anthro­pometric deficits may occur simultaneously, which may amplify the risk of mortality.

Khan and Das (2020) used the CIAF model (Section 13.3.8) to classify under­nutrition and investigate risk of child mortality based on data on children < 5y from the National Family Health Survey in India in 2015–2016. Although no causal inferences could be drawn, their results indicated that districts with a higher burden of multiple anthro­pometric failures based on the CIAF index were associated with an elevated risk of infant and child mortality; see Khan and Das (2020) for more discussion of the findings.

Prospective studies, unlike cross-sectional studies, provide the best information about causation of disease and the most direct measure­ment of the risk of developing disease. McDonald et al. (2013) were the first to estimate the mortality risk associated with multiple anthro­pometric deficits in prospective cohorts based on a meta-analysis of data from 10 prospective studies in Africa, Asia, and Latin America. Unfortunately, cause-specific mortality was not investigated due to insufficient statistical power. In this study, the recent anthro­pometric status of the children was classified into seven subgroups shown in Table 13.13.

Table 13.13. Baseline distri­bution of anthropometric status based on data from 10 prospective studies in Africa, Asia, and Latin America. From: McDonald et al. (2013).

Classification of anthropometric status	n (%)
Not stunted, wasted, or under­weight	27,794 (51.7)
Stunted only	7475 (13.9)
Wasted only	1025 (1.9)
Underweight only	1358 (2.5)
Stunted and under­weight but not wasted	11,501 (21.4)
Wasted and under­weight but not stunted	1655 (3.1)
Wasted, stunted, and under­weight	2953 (5.5)

Mean age of the children at baseline ranged from 5wk to 2.8y and the mean follow-up time ranged from ≈ 35wk to 85wk. A significantly elevated risk of all-cause mortality was reported 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; 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.

13.4.3 Targeting interventions

Anthropometry can be used as a screening tool to identify the areas of greatest need (e.g., areas with a high preva­lence of stunting or wasting or multiple anthro­pometric deficits) and, hence, likely to gain the most benefit from any inter­vention. WHO and UNICEF have classified the severity of mal­nut­rition in young children < 60mos into five categories based on the percentage preva­lence threshold of wasting, over­weight, and stunting (de Onis et al., 2018) (Table 13.14).

Table 13.14. Prevalence thresholds, corresponding labels, and the number of countries (n) in different preva­lence threshold categories for wasting, over­weight and stunting in children under 5 years using the “novel approach”. From: de Onis et al. (2018).

Wasting			over­weight			Stunting
Prevalence thresholds (%)	Labels	(n)	Prevalence thresholds (%)	Labels	(n)	Prevalence thresholds (%)	Labels	(n)
< 2·5	Very low	36	< 2·5	Very low	18	< 2·5	Very low	4
2·5 – < 5	Low	33	2·5 – < 5	Low	33	2·5 – < 10	Low	26
5 – < 10	Medium	39	5 – < 10	Medium	50	10 – < 20	Medium	30
10 – < 15	High	14	10 – < 15	High	18	20 – < 30	High	30
≥ 15	Very high	10	≥ 15	Very high	9	≥ 30	Very high	44

The fifth threshold labeled “very low” and of no public health concern was included across all three indicators to reflect the expected preva­lence of 2.3% (rounded to 2.5%) below/above 2 SDs from the median of the WHO Child Growth Standard. The higher four preva­lence thresholds are defined by multipliers of this “very low” level rounded to 2.5%. Note that for wasting and over­weight, the thresholds are the same: “very low” (< 2.5%), “low” (≈ 1–2 times 2.5%); and “very high” (> ≈ 6 times 2.5%). For stunting, the new thresholds are “very low” (< 2.5%); “low ”(≈ 1–4 times 2.5%); “medium” (≈ 4-8 times 2.5%); “high” (≈ 8–12 times 2.5%) and “very high”(≈ 12 times 2.5%).

The number of countries in the different threshold categories for wasting, over­weight, and stunting, shown in Table 13.14, is based on data accessed in 2018 from the WHO Global Database on Child Growth and Malnutrition and drawn from nationally representative surveys in 134 countries of children < 5y.

Comparison of the preva­lence estimates for each anthro­pometric indicator can trigger countries to identify the most appropriate inter­vention program to achieve “low” or “very low” preva­lence threshold levels. For example, in Table 13.14 , in the countries in which the preva­lence of stunting was “very high” (i.e., n=44), inter­vention activities should focus on resolving factors associated with stunting. These often include increasing the availability of food, and improving diet quality, hygiene, and potable water supplies, as well as programs to prevent and treat infectious diseases (Gorstein et al., 1994).

13.4.4 Evaluating response to an inter­vention

When using anthropometry to evaluate the performance of an inter­vention, at least two measure­ments, taken before and after the inter­vention period, should be recorded, where feasible. In child growth monitoring and promotion programs, for example, serial measure­ments of weight, and sometimes length/height, are taken on children attending child welfare clinics or community health centres. These measure­ments are then plotted manually on distance growth charts, preferably those developed by WHO (Figure 13.3), to provide information on the pattern of growth over time and the response to the inter­vention, where appropriate; see Section 13.2.2 and Section 13.2.3 for more details. When choosing an anthro­pometric indicator to monitor the response, the time lag between the start of the inter­vention and the time when a response can be anticipated must always be considered. As noted in Section 13.2.3, children who are wasted, respond to a nutrition inter­vention by gaining weight first, only later catching up in linear growth. Hence, plotting weight-for-length/height (if available) or weight-for-age on the WHO Child Growth Standard distance chart would be appropriate.

For large scale public health programs, the purpose of an evaluation is to influence decisions about the effectiveness of the programs. These decisions may include to continue, expand, modify, strengthen, or discontinue a program. Habicht et al. (1999) have reviewed the various designs recom­mended for evaluating public health programs. The designs vary depending on the purpose of the evaluation and the level of precision required and are summarized in Box 13.8.

Adequacy evaluation is under­taken when it is not feasible to include a comparison or control group. Instead, the inter­vention is evaluated on the basis of a cross-sectional survey in which the preva­lence of abnormal growth at a certain time-point is compared with the preva­lence of abnormal growth defined by a pre-established criterium of adequacy. This could include, for example, the “very low” preva­lence threshold for abnormal growth shown for children under 5y in Table 13.10. Adequacy of the inter­vention would be achieved if the evaluation showed that the preva­lence of stunting, for example, at the time of the evaluation, was less than the pre-deter­mined “very low” preva­lence threshold of 2.5%.

In an adequacy evaluation, program activities cannot be causally linked to the observed changes. Nevertheless, the adequacy evaluation may provide the necessary reassurance that the expected goals of the program are being met, thus ensuring the continued support for the program. Examples of public health programs in which adequacy evaluation has been applied include a 10‑year Micronutrient and Health (MICAH) program implemented from 1996–2005 in selected areas of Ethiopia, Ghana, Malawi, and Tanzania by World Vision (Berti et al., 2010), and the large-scale program to improve infant and young child feeding (IYCF) and nutrition in Ethiopia by the Alive and Thrive initiative (Menon et al., 2013).

Plausibility evaluation is conducted when decision makers require a greater degree of confidence that the observed changes were indeed due to the program. For an evaluation to be plausible, it must be able to show that the inter­vention appears to have an effect above and beyond other confounding factors which might have caused the observed effects. Hence, although this design does not require randomization of inter­vention and control activities to the comparison groups, nevertheless, attempts must be made to control for the influence of possible confounding factors. This can be achieved at the inter­vention design stage by the careful selection and matching of the comparison or control groups before the evaluation is begun. Alternatively, the influence of confounders or modifiers can be assessed afterwards at the analysis stage, when stratification or multivariate analyses can be used, provided information on potential confounders or modifiers has been collected or measured (Habicht et al., 1999). Of the plausibility designs summarized in Box 13.8, those of choice are the quasi-experimental or case-control designs. For details on selecting appropriate control or comparison groups for plausibility evaluation, see Rogers (2014) and White & Sabarwal (2014). For guidance on the statistical strategies to control for confounders and modifiers, the reader is advised to consult a biostatistician.

Probability evaluation designs provide the highest level of confidence that the inter­vention caused the outcome, and are considered the gold standard method. They require the use of a randomized, controlled, double-blind design which aims to ensure there is only a small known probability (usually < 0.05) that the observed difference in the preva­lence between the inter­vention and control communities (or individuals) is due to confounding, bias, or to chance. With the double-blind randomization of the subjects to inter­vention and control groups, baseline differences between the inter­vention and control groups are minimal.

Figure 13.19.

provides an example of the method used in a probability evaluation design of Alive and Thrive. This design is especially appropriate when evaluating long-running programs, such as 6y for the Alive and Thrive initiative. Note in particular that the inter­vention and control groups in this design are similar at baseline as a result of the randomization. Over such a long time frame, non-inter­vention-related changes in under­lying deter­minants of nutrition, such as income, food security, education, health services, or water and sanitation may occur. These may generate changes in the control group during the duration of the inter­vention as shown in Figure 13.19, high­lighting the importance of a control group, without which the true impact of the inter­vention would have been over­estimated.

In order to attribute the changes in the outcomes to the inter­ventions implemented, Alive and Thrive used the difference-in-difference (DID) or double difference method. This method compares the difference in changes in outcomes over time between the inter­vention and comparison groups to estimate impact, and is said to yield a reliable attribution of impact to an inter­vention. For more details see White and Sabarwal (2014).

Challenges may arise however, when interpreting the data of the impact of the inter­vention when noninter­vention-related changes occur, and additional statistical techniques may be required to under­stand and inform policy makers about which under­lying factors are most important in explaining the improvements in linear growth; see Nguyen et al. (2017) for more details. Gertler et al. (2011) also provide details on statistical methods to evaluate large-scale programs.

Table 13.15

Table 13.15. Growth at 18mos by inter­vention group in Peruvian children. Differences adjusted for socio­economic status, hygiene score, and birth­weight.  ^*Odds ratio (95%CI)    ^**Mean of the regression gradient for length-for-age.    From: Penny et al. (2005).

	inter­vention	Control	Adjusted differences*	P
Mean weight at 18mos (kg)	10.77 (1.16)	10.48 (1.02)	0.199 (-0.033 to 0.431)	0.093
Mean length at 18mos (cm)	79.36 (2.74)	78.29 (2.66)	0.714 (0.146 to 1.282)	0.014
Mean WAZ at 18mos	-0.33 (0.90)	-0.62 (0.83)	0.194 (0.008 to 0.380)	0.041
Mean LAZ at 18mos	-0.81 (0.80)	-1.19 (0.83)	0.272 (0.099 to 0.445)	0.002
Mean WLZ at 18mos	0.15 (0.87)	0.05 (0.79)	0.048 (-0.139 to 0.237)	0.609
Linear growth velocity **	-0.031 (0.043)	-0.057 (0.050)	0.021 (0.012 to 0.031)	<0.0001
Number stunted at 18mos	8 of 171 (4.7%)	26 of 165 (15.8%)	3.035 * (1.207 to 7.636)	0.018

shows an example of a probability evaluation based on a cluster-randomized controlled trial that examined the effectiveness of a nutrition education inter­vention program in Peru. The cohort of young children, recruited at 6mos, were from a poor periurban area. The inter­vention was delivered through the health services with the aim to enhance the quality and coverage of nutrition education programs. The primary outcome was growth, and the impact of the inter­vention program on anthro­pometric indicators was assessed using the “difference-in-difference” (DiD) approach and is shown in Table 13.15. Note that the growth differences were adjusted for socio­economic status, hygiene score, and birth­weight, after accounting for cluster randomisation. The adjusted mean changes in weight gain, length gain, and Z‑scores at 18mos were all significantly greater for the children in the inter­vention area than for those in the control area. In addition, children in the inter­vention group were less likely to be stunted at 18mos that those in the control group; see Penny et al. (2005) for more details.

Abnormal changes in the rate of growth can be identified much earlier when growth velocity instead of distance growth is assessed, as noted previously (Section 13.2.2). However, growth velocity can only be assessed in inter­ventions when the same children have been measured on more than one occasion, as was performed in the study of Penny et al. (2005). In this study in Peru, length velocity was very significantly greater in the inter­vention group than in the controls at 18mos.

For short-term inter­ventions, assessment of weight gain or weight velocity rather than length or height velocity is more appropriate and easier to measure. As an example, in a randomized controlled trial, Sie et al. (2018) evaluated the effect of a 5-day course of three different antibiotics (amoxicillin, azithromycin, and cotrimoxazole) on growth among preschool children in Burkina Faso. Short-term growth was assessed from weight and height measure­ments taken at baseline and 30 days following the last medication dose from which weight velocity (g/kg/d), change in weight (g) and height (in cm), together with WHZ, WAZ, and HAZ- scores, were calculated. Results confirmed that children randomized to amoxicillin gained significantly more weight, had higher weight velocity and WHZ and WAZ‑scores compared to their placebo siblings, but there were no differences in weight gain in those children randomised to azithromycin or cotrimoxazole. Moreover, as expected, no difference in height gain or HAZ were detected over the 30day study period across any of the study arms. These findings highlighted that amoxicillin may have short-term growth-promoting effects in children in settings with a high preva­lence of under­nutrition and child mortality.

13.4.5 Nutritional surveillance

The overall purpose of a nutritional surveillance program is to gather, interpret, and disseminate information about nutrition. WHO (1995) defines nutritional surveillance as:

The continuous monitoring of the physical status of a population, based on repeated surveys or on data from child health or growth-monitoring programs.

The nutrition information collected can be used to achieve the objectives shown in Box 13.9.

The U.S. has conducted a comprehensive national surveillance system since 1959 on the health status of US residents as part of the National Health and Nutrition Examination Surveys (NHANES). Data on anthropometry, demographic and socioeconomic status, dietary and health-related measures are collected.

The U.K. also has a surveillance system designed to assess the food consumption, nutrient intakes and nutritional status of the general population 18mos and older living in private households. The system is termed the National Diet and Nutrition Survey Rolling Program and began in 2008.

Successive national DHS surveys have also been used to track the rates of change in the preva­lence of stunting, under­weight, and over­weight in some countries. As noted earlier (Section 13.3.7), in a meta-analysis of data on women of reproductive age from five successive national DHS surveys in Bangladesh from 1996 to 2011, the preva­lence of under­weight was reported to decrease from 47% to 24% over this time period, whereas that of over­weight increased from 4% to 16% (Figure 13.16). This shift in BMI from under­weight to over­weight from 1996 to 2011 was most strongly and positively associated with urban residence, age, higher socioeconomic status, and higher education attainment (Hoque et al., 2015).

Recently, WHO has provided countries with tools to develop or strengthen their surveillance systems so they have the national capacity to monitor changes in the global nutrition target indicators shown in Box 13.2. These indicators focus specifically on childhood stunting, anemia in women of reproductive age, low birth­weight, over­weight, and exclusive breastfeeding.

The tools developed by WHO/UNICEF include the provision of six policy briefs, each covering one of the global targets with details of the actions that should be prioritized to meet the global target. In addition, a web-based tracking tool has also been developed to assist countries to set national targets and chart their progress of achieving the six global goals. Details of the indicators recom­mended to monitor each target are available from WHO (2013). More details of these tools are available from WHO (2019).

The rising global burden of chronic, non-communicable diseases (NCDs) has also led WHO to adopt a Global Monitoring Framework that includes 25 key indicators to track progress in the prevention and control of NCDs. This initiative prompted WHO in 2002 to develop a stepwise approach to NCD surveillance (STEPS) (WHO STEPS Surveillance Manual, 2005). The approach involves a standardized but flexible framework for countries to monitor the main NCD risk factors based on national household surveys of adults aged 18 to 69y, preferably repeated every 3 to 5 years.

In step 1 of this approach, self-reported information on demographics and behavioral risk factors are collected through in-home interviewer-administered questionnaires. This is followed by step 2 in which physical measure­ments of height and weight (to calculate BMI), waist circumference, and blood pressure are under­taken. Step 3 is performed at a local clinic or health center for biochemical assessment for blood glucose, blood lipids, and urinary sodium (Riley et al., 2016).

Since 2009, WHO has introduced eSTEPS for the data collection which has led to a more streamlined and standardized approach to quality control. WHO also provides technical support through training workshops for the implementation of STEPS, as well as STEPS “Data to Action” workshops to strengthen the in-country develop­ment of national NCD policies and programs.

Data generated through STEPS has been used in a number of projects.

Ng et al (2014) provided a systematic analysis of the global, regional, and national preva­lence of over­weight and obesity in children and adults from 1980–2013. Globally, the proportion of children, adoles­cents, and adults with a BMI indicative of over­weight and obesity has increased over this time period in both developed and developing countries, although for adult obesity the increase has stabilized since 2006. Moreover, age and sex-related differences in patterns of adult over­weight and obesity exist, with men having higher rates than women in developed countries in 2013, peaking at ≈ 55y, whereas in developing countries, rates are higher in women than men, peaking at ≈ 45y (Figure 13.20). Nevertheless, the levels of over­weight and obesity for both men and women in developing countries continues to be lower than that of developed countries.

There has been a concerted effort over the past decade to strengthen nutritional surveillance globally. This effort is not only good public health practice, but it has also increased public and government awareness of the extent of the problems of mal­nut­rition in both developing and developed countries. Future develop­ments in surveillance may include methods that quantify intra-abdominal and subcutaneous adipose tissue, and measure physical activity patterns using wearable technology, and assay hemoglobin A1c as an indicator of diabetes. For more discussion of the challenges, limitations, and future directions of STEPS, see Riley et al. (2016).

As noted in Section 13.1.3, questions have been raised about the validity of interpreting changes in HAZ scores across time as representative of growth falter­ing or catch-up growth. Several studies have reported that HAZ scores increased despite increasing height deficits, highlighting that that the use HAZ scores rather than absolute height differences from the median (HADs) may result in opposite conclusions about the occurrence of catch-up growth (Lundeen et al., 2014). Hence caution is required when defining catch-up growth. The authors recom­mend that absolute height differences should always be included along with HAZ‑scores when evaluating catch-up growth in height in populations with a high preva­lence of growth failure. More research is needed to deter­mine which of the two metrics is better at predicting different outcomes.

Acknowledgements

RSG would like to thank past collaborators, particularly my former graduate students, and is grateful to Michael Jory for the HTML design and his tireless work in directing the trans­lation to this HTML version. Support from “Nutrition International” is also gratefully acknowledged.