ANNEX C - Extract from: Use and Interpretation of Anthropometric Indicators of Nutritional Status. Report of a WHO Working Group (1986a)

Bulletin of the World Health Organization, 64, No. 6, pp929-941.

USE OF THE NCHS POPULATION AS A STANDARD

Discussion has continued in recent years on whether or not it is necessary and appropriate to utilize an international reference (5-7). In analysing this question, it is important to distinguish between a reference and a standard.

A reference is a device for grouping and analysing data. Thus the average weight of a group of children has no meaning unless they happen to be exactly the same age, whereas the average value of the index "weight-for-age" does have meaning. For the construction of such an index a reference population is necessary. In principle, it does not matter what set of reference data is used, provided that it is large enough to contain adequate statistical information and the population is reasonably healthy and well-nourished to avoid major distortions. It is also clearly desirable, for comparative purposes, that there should be a common reference. These principles underlay the recommendation, which was made in 1977 (2) and subsequently endorsed by WHO (8), to adopt the NCHS population as a reference for international use.

A standard embodies the concept of a norm or target - that is, a value judgement. It is this concept that has led to difficulty, since the international reference is widely used also as a standard. The justification for this usage is the evidence collected by Habicht and others (5, 7) that in populations the effect of ethnic differences on the growth of young children is small compared with the effects of the environment. It is accepted that there may be some ethnic differences between groups, just as there are genetic differences between individuals, but for practical purposes they are not considered large enough to invalidate the general use of the NCHS population both as reference and as a standard. This judgement has been endorsed in the report of a recent FAO/WHO/UNU Expert Consultation (9).

There are, however, circumstances in which this usage is felt to be inappropriate and in which local standards are preferred. As a matter of principle, those who are concerned with planning in a particular country may find it unacceptable to base their targets on the characteristics of an alien population. In countries where growth failure in children is widespread and severe, such targets would be unrealistic and unattainable and therefore serve as a hindrance to practical planning.

A realistic target or local "norm" could be set by shifting the international reference downwards. This approach is acceptable if it means simply altering the target, so that, for example, the stated aim would be for the mean height of children to be within 95% rather than 100% of the international reference. It is not acceptable if it means that in the calculation of height-for-age the expected height is taken as 95% of the reference median rather than 100%. When that is done, it is not possible to use the centiles and standard deviations of the reference population, so that the statistical value of the reference is lost.

It is necessary to distinguish between two types of local standards: that derived from an elite, presumably well-nourished group and that which represents the average of the population. A disadvantage of the former is that often an elite group may not be ethnically representative of the population as a whole. Where elite standards have been established in some cases (e.g., Colombia, Mexico, Brazil), they differ little from the NCHS reference. Local standards which represent an average of the population rather than an elite are only useful for identifying groups or individuals who differ from the rest of the population and who may therefore constitute priority targets for intervention. However, many developing countries are experiencing secular trends of increasing weight and height (10), making it necessary to update local population-average references after several years. The development of statistically valid national reference values is costly and often beset with logistic problems, particularly in a very large country such as India. There appear to be no major advantages to offset these drawbacks, and therefore the establishment of local or national reference values is not an urgent priority.

ANALYSIS AND PRESENTATION OF DATA

There are two approaches to the analysis and presentation of data. The first describes the whole distribution; the second provides an estimate of the number or proportion outside the reference distribution. The approaches are complementary and the purpose will determine which is preferred, as discussed in more detail below (pp. 936-937). This type of choice exists in many fields of public health nutrition, and is succinctly described as the choice between shifting the distribution and truncating it.

Whichever approach is to be used, there is then, as discussed in the 1977 report (2), a choice of three ways in which each observed measurement can be related to the reference: by its position within the centile distribution of the reference; as a standard deviation score (Z-score); or as a percentage of the reference median.

Descriptions of the whole distribution

Fig. 1 is an example of how the distribution of the total population may be represented in centiles. The figure is drawn from an actual study and illustrates how a change in the distribution, as the result of an intervention, can be visualized very easily. Statistical methods, such as the chi-square test, can be used for comparing these distributions. However, problems in using centiles for cut-off points are discussed later.

The presentation and statistical treatment of the numbers is the same, whether they represent Z-scores or percentages of the reference median. The simplest descriptor of the whole distribution is the mean Z-score with the SD, or the mean percentage of the reference median with the SD. Standard statistical tests can be applied to these numbers.^a

^a Concern has been expressed about the application of statistical tests when the distribution is skewed. In most populations the distribution of height-for-age is approximately normal (Gaussian), whereas the distributions of weight-for-age and weight-for-height are skewed. In most groups from developing countries the distribution is less skewed than that of the reference population, because the latter contains more overweight children. Therefore, in constructing the NCHS reference tables (3) the population was divided into two halves at the median, and standard deviations calculated separately for each half. Since both observed and reference populations are skewed, relating one to the other will reduce the effect of skewness. Standard statistical tests based on the assumption of a normal distribution can then be applied to the values so derived.

A method of representing the whole distribution, which has been useful in population studies, is to construct a cumulative distribution curve and calculate its slope (Fig. 2). The slopes found for different populations and the position of the curve can then be compared, along with the median Z-scores. However, it is unclear just how much of the cumulative distribution slope can be explained by measurement variability.

Fig. 1. Centile distribution of weight-for-height and height-for-age. (WHO 861466)

Fig. 2. Cumulative distribution curves of Z-scores, the weight-for-height and height-for-age values are for a population that is stunted but not wasted. (WHO 861468)

It appears that the best way of giving a complete picture of the whole distribution which can be compared with that of the reference population is a frequency curve or histogram of Z-scores (Fig. 3). The first step in constructing such a distribution curve would be tabulation of the data in the form shown in Table 1, which can be done for any age group, with any index. The size of the interval used for grouping the data, e.g., 0.5 or 1.0 Z-score unit, will depend on the number of measurements available, the facilities for analysing them, and the extent to which fine grouping is likely to be of practical value. For percentage of the median, the distribution curve is not practical because the data for the reference population are age-dependent when expressed in these terms and are not readily available.

Fig. 3. Distribution curves of weight-for-height and height-for-age in relation to reference Z-scores. (WHO 861467)

Definition of the number at risk and choice of cut-off points

For many purposes the most useful way of describing the nutritional situation is to present an estimate of the number or proportion who might be considered at risk. In principle such an estimate is given by the number outside the reference population. In practice it is conventional to use cut-off points, which are indicators, in the sense defined above; for example, the number below the 3rd centile; the number with Z-scores less than - 2SD; or the number with weight-for-height less than 80% of the median. With centiles and Z-scores it is an advantage that the same cut-off can be used for both weight and height, whereas with percentage of the median the cut-offs are necessarily different.

The disadvantage of using centiles for cut-offs is that the number at extreme degrees of risk cannot be quantified, since centiles below the 3rd or above the 97th cannot be defined from the reference population except by back-calculation from the standard deviations.

It is in the choice of cut-offs that the difference between Z-scores and percentage of the median becomes particularly important. For example, in one survey of weight-for-height of children between 1 and 2 years old, 27% had Z-scores of -2 or below, whereas only 15% were below 80% of the reference median (17). This discrepancy cannot be eliminated simply by adjusting one or the other cut-off, because the coefficient of measurement variation varies with age. By definition, Z-score cut-offs take this into account, percentage of the median cut-offs do not.

Two objections have been made to the use of fixed cut-off points such as those cited above. The first is that at best they represent a purely statistical separation of "malnourished" from "normal". Ideally, cut-off points should be based on biological considerations, such as increased risk of mortality or of functional impairment. The cut-off should distinguish a deficit that matters from one that is of no real significance. This is a valid objection, but the practical problems of establishing a relation to risk are very great. Prospective studies of mortality, such as those of Chen and co-workers in Bangladesh (29), make it possible to determine the predictive value of different indices and to define the cut-off points which produce the optimum combination of sensitivity and specificity (30-32). However, death is not the only outcome which needs to be considered, and even for this particular outcome the results almost certainly cannot be generalized from one region to another. The quantitative relation between mortality risk and anthropometric deficit will vary, among other things, with infectious load. It also varies with age, a given deficit carrying greater risk in younger children (33).

Table 1. Anthropometric data on the distribution of Z-scores in a sample population, used for constructing the distributions in Fig. 1 and 3; the reference distribution in column 4 is a normal distribution, by definition

	Sample population distribution
Z-score range	Weight-for-height of 2-year-olds (%)	Height-for-age of 2-year-olds (%)	Reference distribution (all indices and age groups) (%)
-5.49 to -5.0		0.8
-4.99 to -4.5		1.3
-4.49 to -4.0		4.7
-3.99 to -3.5		5.5
-3.49 to -3.0	0.0	9.4	0.1
-2.99 to -2.5	1.0	11.2	0.5
-2.49 to -2.0	1.3	12.8	1.7
-1.99 to -1.5	5.0	12.8	4.4
-1.49 to -1.0	10.7	12.8	9.2
-0.99 to -0.5	16.4	12.5	15.0
-0.49 to 0	18.6	7.6	19.1
0.01 to 0.5	20.8	5.7	19.1
0.51 to 1.0	13.5	1.0	15.0
1.01 to 1.5	7.6	0.8	9.2
1.51 to 2.0	2.3	0.8	4.4
2.01 to 2.5	1.8	0.3	1.7
2.51 to 3.0	0.3	0.0	0.5
3.01 to 3.5	0.0	0.0	0.1

The second objection is that the conventional cutoff of - 2SD or its equivalent may be unrealistic and of limited use in practice. Thus, in an emergency situation where resources are restricted a lower cut-off point might have to be used to identify the children most in need, i.e., an increase in specificity at the expense of sensitivity (20, 30, 31). Again, if 60% of children in a particular country are described as significantly stunted, because they are below -2SD in height-for-age, this cut-off would defeat one of the aims of concentrating on the tails of the distribution, which is to identify those particularly and exceptionally at risk. In this case, if one wants to determine which children are most severely stunted, a lower cutoff point could be used.

Cut-offs should be chosen at the point most appropriate for the particular purpose in view, the reasons for choice being clearly stated. For most group or population comparisons, where uniformity is important, the standard statistical cut-off points of ±2SD from the mean should be maintained (17). In order to utilize a single method of relating measurements to the reference, it would also be necessary to use Z-scores in the presentation of whole distributions (Fig. 3). This is in accordance with the 1977 report (2), which recommended the use of Z-scores to express both distributions and cut-off points because they have a statistical meaning. Since then, WHO has also recommended to Member countries (8) the use of Z-scores for monitoring nutrition and health progress.

REFERENCES

1. TORUN, B. & VITERI, F. E. Energy requirements of preschool children and effects of varying energy intakes on protein metabolism. Food and nutrition bulletin, Suppl. 5: 229-241 (1981).

2. WATERLOW, J. C. ET AL. The presentation and use of height and weight data for comparing the nutritional status of groups of children under the age of 10 years. Bulletin of the World Health Organization, 55: 4S9-498 (1977).

3. U.S. FOOD AND NUTRITION BOARD. Comparison of body weights and body heights of groups of children. Atlanta, GA, U.S. Department of Health, Education and Welfare, 1974.

4. WORLD HEALTH ORGANIZATION. Measuring change in nutritional status. Guidelines for assessing the nutritional impact of supplementary feeding programmes for vulnerable groups. Geneva, 1983.

5. HABICHT, J.-P. ET AL. Height and weight standards for pre-school children. How relevant are ethnic differences in growth potential? Lancet, 1: 611-615 (1974).

6. GOLDSTEIN, H. & TANNER, J.M. Ecological considerations in the creation and the use of child growth standards. Lancet, 1: 582-585 (1980).

7. GRAITCER, P. L. & GENTRY, M. Measuring children: one reference for all. Lancet, 2: 297-299 (1981).

8. WORLD HEALTH ORGANIZATION. Development of indicators for monitoring progress towards Health for All by the year 2000. Geneva, 1981 ("Health for All" Series No. 4).

9. WHO Technical Report Series No. 724, 1985 (Energy and protein requirements: report of a Joint FAO/WHO/UNU Expert Consultation).

10. EVELETH, P. B. & TANNER, J. M. Worldwide variation in human growth. Cambridge, Cambridge University Press, 1976 (International Biological Programme 8).

11. SEOANE, N. & LATHAM, M. C. Nutritional anthropometry in the identification of malnutrition in childhood. Journal of tropical pediatrics and environmental child health, 17: 98-104 (1971).

12. WATERLOW, J. C. Classification and definition of protein-calorie malnutrition. British medical journal, 3: 566-569 (1972).

13. WATERLOW, J. C. Classification and definition of protein-energy malnutrition. In: Beaton, G. H. & Bengoa, J. M., ed. Nutrition in preventive medicine (WHO Monograph Series, No. 62). Geneva, World Health Organization, 1976, Annex 5, pp. 530-555.

14. ASHWORTH, A. Growth rates in children recovering from protein-calorie malnutrition. British journal of nutrition, 23: 835-845 (1969).

15. PRADER, A. Catch-up growth. Postgraduate medical journal, 54 (Suppl.): 133-243 (1978).

16. KELLER, W. Choice of indicators of nutritional status. In: SchB., ed. Evaluation of nutrition education in Third World communities (Nestloundation Publication Series). Bern, Hans Huber Publishers, 1983.

17. KELLER, W. & FILLMORE, C. M. Prevalence of protein-energy malnutrition. World health statistics quarterly, 36: 129-167 (1983).

18. SOYSA, P. E. & WATERLOW, J. C. Stunting and the prevalence of malnutrition. Lancet, 1: 1430 (1981).

19. ANDERSON, M. A. Comparison of anthropometric measures of preschool children in five developing countries. American journal of clinical nutrition, 32: 2339-2345 (1979).

20. DOWLER, E. A. ET AL. Nutritional status indicators: interpretation and policy-making role. Food policy, 7: 99-112 (1982).

21. NABARRO, D. Social, economic, health and environmental determinants of nutritional status. Food and nutrition bulletin, 6: 18-32 (1983).

22. COLE, T.J. & PARKIN, J. M. Infection and its effect on the growth of young children: comparison of the Gambia and Uganda. Transactions of the Royal Society of Tropical Medicine and Hygiene, 71: 196-198 (1977).

23. WATERLOW, J. C. ET AL. Faltering in infant growth in less developed countries. Lancet, 2: 1176-1177(1980).

24. TANNER, J. M. ET AL. Standards from birth to maturity for height, weight, height velocity and weight velocity. British children, 1963, part II. Archives of disease in childhood, 41: 613-635 (1966).

25. FOMON, S. J. ET AL. Growth and serum chemical values of normal breast-fed infants. Acta paediatrica Scandinavica Supplementa, 273: 1-29 (1978).

26. POMERANCE, H. H. & KRALL, J. M. Linear regression to approximate longitudinal growth curves: revised standards for velocity of weight and length in infants. Pediatric research, 15: 1390-1395 (1981).

27. DUGDALE, A. E. A method for presenting longitudinal growth data. British journal of nutrition, 41: 53-56 (1979).

28. VALVERDE, V. ET AL. Seasonality of nutritional status: a review of findings from developed and developing countries. Archivos Latinoamericanos de NutriciI>, 32: 521-540 (1982).

29. CHEN, L. C. ET AL. Anthropometric assessment of energy-protein malnutrition and subsequent risk of mortality among pre-school children. American journal of clinical nutrition, 33: 1836-1845 (1980).

30. HABICHT, J.-P. Some characteristics of indicators of nutritional status for use in screening and surveillance. American journal of clinical nutrition, 33: 531-535 (1980).

31. HABICHT, J.-P. et al. Overview: indicators for identifying and counting the improperly nourished. American journal of clinical nutrition, 35 (Suppl.): 1241-1254 (1982).

32. BAIRAGI, R. On the best cut-off point for nutritional monitoring. American journal of clinical nutrition, 35: 769-770 (1982).

33. KIELMANN, A. A. & McCORD, C. Weight-for-age as an index of risk of death in children. Lancet, 1: 1247-1250 (1978).

34. BALDWIN, B. T. Weight-height-age standards in metric units for American-born children. American journal of physical anthropology, 8: 1-10 (1925).

35. ROSSO, P. A new chart to monitor weight gain during pregnancy. American journal of clinical nutrition, 41: 644-652 (1985).

36. BRAY, G. A. Obesity in America. Proceedings of the 2nd Fogarty International Center Conference on Obesity. Washington, DC, US Department of Health, Education & Welfare, 1979 (National Institutes of Health, Publication No. 79-359).

37. SOCIETY OF ACTUARIES. The build study. 1979. Chicago, Society of Actuaries, 1979.

38. The incidence of low birth weight: a critical review of available information. World health statistics quarterly, 33: 197-224 (1980).

39. BHARGAVA, V. ET AL. Fetal growth variations. 1. Influence of maternal size and nutrition on identification of fetal growth retardation. Indian pediatrics, 20: 549-559 (1983).

40. MORA, J. ET AL Nutritional supplementation and the outcome of pregnancy. 1. Birthweight. American journal of clinical nutrition, 32: 455-462 (1979).

41. DELGADO, H. ET AL. Relationship of maternal and infant nutrition to infant growth. Early human development, 6: 273-286 (1982).

42. DELGADO, H. ET AL. Nutrition and length of gestation. Nutrition research, 2: 117-126 (1982).

43. MATA, L. J. ET AL. Antenatal events and postnatal survival and growth of children in a rural Guatemalan village. Annals of human biology, 3: 303-316 (1976).

44. NAEYE, R. Weight gain and the outcome of pregnancy. American journal of obstetrics and gynecology, 135: 3-9(1979).