Contents - Previous - Next

This is the old United Nations University website. Visit the new site at

Public health

Observations on the FAO's methodology for estimating the incidence of undernutrition

J. C. Waterlow

The Food and Agriculture Organization's Fifth World Food Survey [1] is, in most respects, a valuable and well-setout report, covering a difficult subject extensively and clearly. The points that I wish to make here are concerned only with the appendix on estimating the number of malnourished people and are directed mainly to some of the biological assumptions and interpretations made therein.

The basis of the methodology is in principle straightforward and involves two stages. First, a cutoff point is determined for energy expenditure, below which a person of any given age and either sex may be regarded as deficient. Second, on the reasonable assumption that over a period of time intake and expenditure must be equal, information is obtained about the distribution of intakes. The third step is to calculate from these two sets of data the number of people whose intake for expenditure is below the cutoff point.

It should be mentioned at the outset that, for practical reasons, the calculations based on individuals' intakes are scanty. To estimate the prevalence of malnutrition it was necessary to assume that the proportion of households with inadequate energy intake represents the proportion of undernourished individuals in the whole population. The FAO was aware of the shortcomings of this procedure in that it ignored the possibility of inequitable distribution within the household.

In this note I am concerned only with certain aspects of the first step. This step itself has two components: determining the basal metabolic rate (BMR) on the basis of body weight, and then arriving at a minimum level of daily energy requirement, expressed as a multiple of the BMR.

Before addressing these two components separately, a general comment is in order about the framework within which the discussion proceeds. It is accepted that a person can be in energy balance, and his requirements met, at different levels of body weight, so that for any individual or group there is a range of requirements. The Food Survey seeks to determine the minimum requirement, or lowest point of the range. The phrase "minimum energy requirement" was avoided by the 1981 Expert Consultation on Energy and Protein Requirements (see its report [2], hereafter referred to as the Rome report). For protein and other nutrients the term "minimum requirement" has generally been taken to mean that an individual should consume this amount or more, although for certain nutrients an upper limit of intake exists at which there is danger of toxicity. In the case of energy, it has generally been held that this concept does not apply, because a person who eats more than the requirement becomes obese.

Indeed, it does seem legitimate to treat energy in the same way as other nutrients and to allow a range of acceptable intakes corresponding to a range of acceptable body weights. The main difference from other nutrients is that the range is rather small.

Body weight as a basis for estimating requirements

The Food Survey points out that if a person's intake is less than the requirement at his existing body weight, the result will be loss of weight and a consequent reduction in energy expenditure until balance is restored at the lower level of intake. This process is referred to as cost-less biological adaptation. The concept involves two value judgements that are not made explicit and are not adequately discussed: the first is a judgement about the lower limit of acceptable weight; the second is the assumption that transition from one body weight to another is costless.

For adults, the Survey takes as its lower limit the minimum desirable weight of the Metropolitan Life Insurance Company tables. Technically, this is not a good choice. Those tables are based on mortality rates obtained many years ago in a selected American population-those who took out life insurance. Today, it is generally recognized among nutritionists that it is preferable to use a purely objective measure of body weight in relation to height, such as the Quetelet index or body-mass index (BMI = weight/height2, with weight expressed in kilograms and height in metres). The problem is to evaluate the acceptable range of BMI.

We have discussed this question in more detail elsewhere [3]. Briefly, for adults at all heights and of both sexes the lower limit of the Metropolitan desirable weight corresponds to a BMI of about 20. This appears to be an acceptable figure for industrialized countries. The Fogarty Conference in the United States and the British Royal College of Physicians [4] chose BMls of 18.6 for men and 20 for women as their lower cut-off points, rounded off to 20 for both sexes. These values, however, based on mortality risk, cannot be extrapolated to the developing nations. It could be argued, however, that for developing countries a cut-off point of BMI at 20 is too high, since in many apparently healthy population groups the average is 18 to 19, so that in half the subjects it will be even lower [5].

Data on the distribution of BMI in several emerging countries collected recently by the FAO suggest that 18.5 would be a reasonable lower limit in both sexes. This is in agreement with our own analysis [3]. The choice of the lower limit clearly would have a large effect on estimation of the number of undernourished people. This is a point to which a great deal of attention must be given when the FAO prepares its next world food survey.

In addition, one may question whether transition to a lower level of body weight is costless even within the acceptable range. Ample evidence suggests that such a transition does not simply represent a loss of fat. Some loss of lean body mass, probably mainly of muscle mass, occurs that may have functional consequences.

It is, of course, a matter of degree. Information on changes in body weight and energy expenditure has been summarized in farmers during the season when food is short and the need for physical work is at its peak [6]. Energy expenditure goes up at the expense of body weight, which may fall by a maximum of 5 kg. This would correspond to about 2 BMI points, so that, if the initial BMI was 20 or 21, in the hungry season it would fall to 18 or 19, or just about the lowest acceptable limit. This degree of change may appear to be costless, but it reduces the body's reserves that may be needed in illness or pregnancy. It has been shown in the Gambia that in the hungry season there is a significant fall in infants' birth weight [7].

The concept of costless adaptation in body weight is unsatisfactory for different reasons when applied to children. It is, indeed, a fact of life that large numbers of children in developing nations are well below the international reference levels of length or height for age [8]. The Food Survey regards this height deficit (stunting) as acceptable at least within limits (see below); it then argues that, as for adults, for any given height there is an acceptable range of body weight, and for assessing the child's requirement it chooses a body weight of one SD (standard deviation) below the reference median weight for height.

These proposals produce a substantial savings in energy requirement. If the requirement per kilogram is taken as 100 kcal per day, a two-year-old child at the reference median weight and length would need 1,260 kcal per day. At one SD below the median reference height for age and reference weight for height, the child's requirement would be 1,085 kcal per day, a saving of 13%.

Are these proposals justified? One has to distinguish between the short term and the long term. Weight for height is concerned with the short term. I do not suggest that one SD below the median in itself represents a weight deficit of any importance. Even in a deprived population, however, the weight for height of many children is above that cut-off point. The Food Survey concept is that such a child, like an adult, can adapt without cost by reducing his body weight until it reaches the cut-off point. This is a dangerous proposition.

In adults, as mentioned above, during the lean season, when energy expenditure exceeds intake, physical activity is maintained at the expense of body weight [6]. Since what matters is activity rather than body weight itself, this adaptation in adults may be regarded as acceptable, if not actually costless. There is evidence that in children the response is different: in one study of pre-school children, when the energy intake was reduced from 87 to 80 kcal per kg, physical activity was decreased while weight gain continued [9]. The profound influence of food supply on the activity level of young children has been documented [10]. Any so-called adaptation that involves a reduction in activity cannot be regarded as costless.

On the question of height, the Survey accepts that young children worldwide have roughly the same potential for growth but argues that there is a growing body of evidence that a growth rate below the genetic potential is not necessarily harmful [11]. This view may be correct within limits, but the evidence quoted to support it is not impressive. It is inappropriate in this context to rely on a study conducted in Bangladesh [12] which was not directed to the point at issue. The Rome report did indeed accept that there is "a wide variation in children's sizes, with no indication that the differences per se are related to health, well-being or physiological function." The question, as always, is one of degree. It is necessary in practice to make value judgements. It would surely be going too far to maintain that deficits of more than two SD below the reference median, such as are found in some 50% of children in many populations [8], are acceptable and costless.

In the short term it may be necessary to accept that stunting is a fact of life in developing countries; it would indeed not make sense to mount a nutritional programme to eliminate stunting, if only because we do not know what its cause is in nutritional terms. However, we are not concerned only with the short term.

The subject of stunting-its natural history, epidemiology, and causal factors-has recently been discussed in depth [13]. Three things are clear: First, stunting is an index of social deprivation, to which many factors contribute. Indeed, in the United Kingdom height has been used as a measure of the health and nutrition of children ever since the pioneering work of Boyd Orr. Second, physical growth retardation is often, although not invariably, accompanied by retardation of mental development [see, e.g., ref. 14]. Third, small body size per se results in reduction of physical working capacity, which is very important in a rural environment [15; 16]. Thus, although to be small may reduce the risk of death when food is short, this adaptation is by no means costless.

To the extent that the efforts of UN agencies are successful in improving the conditions for disadvantaged people, there is likely to be a decrease in the prevalence and extent of stunting and a consequent increase in the food requirements of children. We know that the FAO plans for the future as well as for the present. It is important not to give the impression that the status quo is acceptable and that there is one standard for the rich and another for the poor.


The Food Survey appendix rightly devotes some attention to the effects infection and other undefined environmental factors on requirements, but unfortunately without any clear outcome or recommendations. Although UN committees have defined requirements in terms of the needs of "healthy" people, they have always recognized that this is an abstraction, since very large numbers of people in the world are not healthy. It is naturally impossible to make a general rule that will provide quantitative estimates of the extra requirements, because conditions vary widely. The Rome report took one step forward, giving a concrete example of the extra amounts of energy and protein required to produce catch-up growth in a child exposed to repeated infections.

The basic principle was to determine the degree of growth deficit produced by infection and then to calculate the requirements for catch-up. Quantitative information has to be obtained about the prevalence and duration of infections in any population and the average effect on growth of one day's infection. This information allows calculation of the impact of infection on growth [17]. Two conclusions can be drawn from this kind of exercise: (a) In general, to produce catch-up growth in children, the extra amount of protein needed will be relatively greater than the extra amount of energy. (b) Except in extreme cases, the extra energy needed will not be great compared with the normal energy requirement. This follows from the physiological fact that the energy cost of growth is small compared with the requirement for maintenance.

It is unfortunate that the appendix does nothing more to illuminate the problem. It does not even indicate what kind of information is needed as a basis for calculating, even roughly, an allowance for the effects of infection. It has already been pointed out that, for both adults and children, a so-called costless adaptation that consists in a reduction of body weight will diminish the body's reserves or margin of safety in the face of environmental stress. No such adaptation is truly costless unless it occurs in persons who are initially at or above the upper end of the acceptable range, a situation that is not very common in populations subject to undernutrition.

The maintenance requirement and the cut-off point for undernutrition

Both The Fourth World Food Survey [18] and the fifth have taken the line that the minimum energy expenditure needed for any kind of normal existence (i.e., the maintenance requirement) could be taken as a cut-off point; with intakes below this, people would be undernourished.

The appendix to the fifth Survey discusses fully and frankly the main problems of this approach. First, by definition, maintenance involves minimal physical activity: it allows for washing, dressing, cooking food, and moving about the home, but not for the physical activity demanded in rural communities for the production of food. Thus the maintenance requirement may be a suitable criterion for urban populations who are largely sedentary. In communities that rely on the physical work of their inhabitants to produce food, however, a person who is living just at the maintenance level is maintaining his nutrition at the expense of others, and is in effect increasing their requirements.

It is a difficult and laborious task to quantify physical activity, with all the variations that occur between individuals and communities. It may be simplified when it is possible to apply more widely the doubly labelled-water method, by which total energy expenditure can be measured non-invasively over periods of 10 to 14 days.

The second problem, again well recognized by the fifth Food Survey, has already been alluded to: that the intake data are based on households, and it cannot be assumed that food is evenly distributed within households. Some studies of intra-household distribution have been done, but it would be hazardous to generalize from them.

For the time being, therefore, the use of the maintenance requirement as a cut-off point, in spite of the sources of error, does seem to be the best available approach.

The Rome report proposed that 1.4 x BMR would be a realistic value for the maintenance requirement. Since that report was drafted, several studies in which energy expenditure was measured in a whole-body calorimeter [e.g., ref. 19] have confirmed that this is a reasonable estimate of the expenditure of people leading a very sedentary life, with minimal physical activity. To apply this criterion, the predicted BMR (BMR) for any age and sex group is obtained from the equations of Schofield et al. [20], reproduced in the Rome report. The maintenance requirement is then estimated as 1.4 BMR, with an SD of 0.1 BMR (coefficient of variation about 7%). The calorimeter studies have shown that virtually all the variability in the maintenance requirement can be attributed to variability of the BMR, and that the energy cost of a fixed activity pattern expressed as a multiple of BMR is relatively invariant.

The Food Survey also proposes the use of an alternative cut-off point of 1.2 BMR, based on the hypothesis of Sukhatme and Margen [21]. The essence of the hypothesis is that an individual's BMR is not fixed but may alter from week to week, adjusting to short-term imbalances between energy intake and expenditure. Thus all the variability found in a given set of measurements of BMR would represent within subject variations, because different subjects have been measured at different points within their range of adjustment or adaptation. It is then concluded that all people, if faced with low food intake, could adjust their BMR to the lower end of the range. On this view, the appropriate cut-off point would not be 1.4 BMR but 1.4 BMR- 2 SD = 1.2 BMR. The difference is far from negligible when it comes to determining the number of undernourished people. The Food Survey, very fairly, gives estimates based on both cutoff points. Thus it is calculated that in 1979-1981 the proportion of people undernourished in all developing-market economies taken together would be 15% if the cut-off point is 1.2 BMR, and 23% if it is 1.4 BMR[1, table 3.1]. In terms of energy, the difference between the cut-off points is of the order of 150200 kcal per day.

However, I should point out that, as far as I can see, the difference between the two criteria, cut-offs at 1.2 BMR (A) and 1.4 BMR (B), is not simply a quantitative one, like the difference between 70% and 80% of expected weight for height as a criterion of malnutrition in children: there is also a qualitative difference. By both criteria, anyone with an intake of less than 1.2 BMR would be definitely undernourished. According to A, anyone with an intake greater than 1.2 BMR is not undernourished; according to B, intakes between 1.4 and 1.2 BMR represent a zone of increasing risk of undernutrition, and intakes between 1.4 and 1.6 BMR (mean + 2 SD) represent a zone of decreasing risk. This situation is analogous to that for protein requirements. The contrast is shown in figure 1.

This qualitative difference in the nature of the criteria implies that there should be a difference in the procedures for matching intakes against minimum requirements. According to A, a distribution of intakes for a given age-sex group has to be matched against a fixed cut-off point. According to B, there has to be a matching between two distributions. (Strictly, from information provided by the FAO, the procedure is a "matching within the framework of a bivariate (joint) distribution of intakes and requirements." )

FIG. 1. Alternative approaches to estimating maintenance requirements for dietary energy:

(A) Subjects are assumed to be able to adapt the level of their maintenance requirement to an energy intake of 1.2 x BMR. No subject with an intake above this level would be at risk; all subjects with an intake below it would be deficient. (B) Mean maintenance requirement= 1.4 x mean predicted BMR. The zone from 1.6 down to 1.2 x BMR represents an increasing risk. For a subject with an intake above 1.6 x BMR the risk of being deficient is very small. Subjects whose intake is below 1.2 x BMR are almost certainly deficient.

Recent work does not support the conclusion that within-subject variation accounts for almost all the variability in BMR. Measurements repeated over days, weeks, or months show that the BMR of an individual is remarkably stable. The within-subject coefficient of variation is about 2%, while the coefficient of variation between subjects is 6%-7% [22; 23].

It could be argued that the within-subject variation is small because the BMRs were measured under similar, if not identical, conditions, and that the variation would be greater if they were measured under the stress of major alterations in intake or activity. This is not the place for a detailed discussion of the hypothesis, however. Summaries and critiques appear elsewhere [24; 25].

It should be noted that the Sukhatme-Margen hypothesis is based mainly on results obtained nearly 20 years ago on army recruits over a period of five weeks. It remains to be seen whether it is relevant to the situation of people who throughout their lives are exposed to marginal food intakes. This is the situation with which The Fifth World Food Survey is concerned.

For a long-term adaptation to be acceptable, a sustained decrease in "economic" or "discretionary" activities would not be a viable option. Apart from adjustments in body weight, the only information that we have relates to BMR.

In the world-wide analysis of BMR by Schofield et al. [20], by far the greater part of the data was derived from measurements in industrialized countries. The authors concluded tentatively that in developing countries the BMR might be lower than the overall average by 9%-10%, a conclusion based almost entirely on reports from India. Since then the FAO has itself collected further measurements of BMR in developing countries, and Henry and Rees [26] have made a further analysis of the older literature. The findings show that in developing countries BMR does tend to be lower than the values predicted from the Schofield equations that were used in the Rome report and by the Food Survey, the difference being greater in males than in females. Some examples are given in table 1.

TABLE 1. BMRs predicted by different equations

Age (years) and sex

Body weight (kg)

Predicted BMR (kcal)

A - B as % of A

Aa Bb
male 60 1,61)0 1,490 +6.9
female 50 I ,230 1,200 +2.4
male 60 1,560 1,450 +7.1
female 50 1,250 1,180 +5.6

a. From the equations of Schofield et al. [20]. Average standard error of estimate, ±138 kcal.
b. From the equations of Henry and Rees [26]. Average standard error of estimate, ±132 kcal.


It is not the purpose of this article simply to be critical of an important and valuable document, but rather to indicate topics that require further consideration and areas where more information needs to be collected in preparation for the next world food survey.

As mentioned, a working group of the International Dietary Energy Consultative Group was charged with the task of defining criteria for the diagnosis of chronic energy deficiency, with special emphasis on individuals. This group has reached a tentative conclusion that the most practical criterion for adults is one based not on measurements of energy intake or expenditure but on physical state as measured by the BMI [3]. The difficulty is to define a cut-off point for the lower limit of acceptable BMI. It may well be that, as with the anthropometric criteria of nutritional status in children, it will be advisable to have several cut-off points, corresponding to mild, moderate, and severe deficiency or risk of deficiency.

The current Food Survey approach similarly involves judgements about the acceptable range of body weight, with conclusions about "costless adaptation" that, as we have argued, have not been fully thought out. In relation to children, they are unacceptable. The methodology also requires the collection of data of food intake or proxies for food intake-a difficult and time-consuming task. It is to be hoped that in the future more use will be made of criteria based on physical nutritional status and more consideration given to the functional costs, if any, and the levels of activity that may be related to differences in physical status. This, of course, is more easily said than done. The FAO cannot move faster than the physiologists. This paper will have achieved its aim if it makes clear the needs for further research, often ignored by funding bodies, and the directions in which that research might go.


1. Food and Agriculture Organization. The fifth world food survey. Rome: FAO, 1987.

2. Joint FAO/WHO/UNU expert consultation. Energy and protein requirements. WHO technical report series, no. 724. Geneva: WHO, 1985.

3. James WPT, Ferro-Luzzi A, Waterlow JC. Definition of chronic energy deficiency in adults. Report of a working party of the International Dietary Energy Consultative Group. Europ J Clin Nutr 1988;42:969-81.

4. Royal College of Physicians. Obesity. A report of the Royal College of Physicians. J Roy Coli Phys 1983;17(1): 1-58.

5. Eveleth PB, Tanner JM. Worldwide variations in human growth. Cambridge, UK: Cambridge University Press, 1976.

6. Ferro-Luzzi A, Pastore G, Sette S. Seasonality in energy metabolism. In: Schürch B, Scrimshaw NS, eds. Clinical energy deficiency: consequences and related issues. Lausanne, Switzerland: Nestle Foundation, 1988:3758.

7. Prentice AM. Variations in maternal dietary intake, birth weight and breast-milk output in the Gambia. In: Aebi H, Whitehead R. eds. Maternal nutrition during pregnancy and lactation. Bern, Switzerland: Huber, 1980:167-83.

8. Keller W. The epidemiology of stunting. In: Waterlow JC, ed. Linear growth retardation in less developed countries. New York: Raven Press, 1988: 17-40.

9. Food and Agriculture Organization. Report on the informal gathering of investigators to review the collaborative research programme on protein requirements and energy intake. Doc. ESN/MISC/80/3. Rome: FAO, 1980.

10. Chavez A, Martinez C. Behavioral measurements of activity in children and their relation to food intake in a poor community. In: Pollitt E, Amante P, eds. Energy intake and activity. New York: Alan R. Liss, 1984:30321.

11. Seckler D. "Small but healthy": a basic hypothesis in the theory, measurement and policy of malnutrition. In: Sukhatme PV, ed. Newer concepts in nutrition and their implications for policy. Pune, India: Maharashtra Association for the Cultivation of Science, 1982.

12. Chen LC, Chaudhury AKMA, Huffman SL. Anthropometric assessment of energy-protein undernutrition: subsequent risk of mortality among pre-school-aged children. Am J Clin Nutr 1978;33:1836-45.

13. Waterlow JC, ed. Linear growth retardation in less developed countries. Nestle nutrition workshop series, no. 14. New York: Raven Press, 1988.

14. Richardson SA. The relation of severe malnutrition in infancy to the intelligence of schoolchildren with differing life histories. Pediat Res 1976;10:57-61.

15. Spurr GB. Physical activity, nutritional status and physical work capacity in relation to agricultural productivity. In: Pollitt E, Amante P, eds. Energy intake and activity. New York: Alan R. Liss, 1984:207-61.

16. Satyanarayana K, Naidu AN, Narasinga Rao BS. Nutritional deprivation in childhood and the body size, activity and physical work capacity of young boys. Am J Clin Nutr 1979;32:1769-75.

17. Zumrawi FY, Dimond H, Waterlow JC. Effects of infection on growth in Sudanese children. Hum Nutr Clin Nutr 1987;41C:453-62.

18. Food and Agriculture Organization. The fourth world food survey. FAO statistics series, no. 11. FAO food and nutrition series, no. 10. Rome: FAO, 1977.

19. Garby L, Lammert O, Nielsen E. Energy expenditure over 24 hours on low physical activity programmes in human subjects. Hum Nutr Clin Nutr 1986;40C: 141-50.

20. Schofield WN, Schofield C, James WPT. Basal metabolic rate review and prediction, together with an annotated bibliography of source material. Hum Nutr Clin Nutr 1985;39C(Suppl.1).

21. Sukhatme PV, Margen S. Autoregulatory homeostatic nature of energy balance. Am J Clin Nutr 1982;35:35565.

22. Garby L, Lammert O. Within-subjects between-days-and-weeks variation in energy expenditure at rest. Hum Nutr Clin Nutr 1984;38C:395-97.

23. Soares MJ, Shetty PS. Long-term stability of metabolic rates in young adult males. Hum Nutr Clin Nutr 1987;41C:287-90.

24. Sukhatme PV. Nutritional adaptation and variability. Europ J Clin Nutr 1989 (in press).

25. Waterlow JC, James WPT, Healy MHR. Commentaries on Sukhatme's paper "Nutritional adaptation and variability." Europ J Clin Nutr 1989 (in press).

26. Henry CJK, Rees DG. A preliminary analysis of basal metabolic rate and race. In: Blaxter KL, McDonald 1, eds. Comparative nutrition: an international symposium. London: John Libbey, 1988.

Contents - Previous - Next