(introductory text...)

Reynaldo Martorell and Meera Shekar

Abstract

Criteria used in growth-monitoring programmes in developing countries were applied to three-month weight-gain data for children 12-24 months old in three populations: in Berkeley, California; Guatemala; and Tamil Nadu. A significant proportion of the Berkeley children showed growth faltering: 18% had at least one period in which they failed to gain any weight, and 37% had at least one in which they gained less than 300 g in three months. The frequency of faltering, however, was appreciably greater in Guatemala (45% and 82% respectively) and Tamil Nadu (42% and 74% respectively). These data raise concerns that growth-monitoring criteria, as used in most settings, identify too many children for special attention, perhaps more than can be handled by most programmes. Adjusting the criteria to select fewer children necessarily means departure from the simplest guidelines about focusing on the growth trajectory (i.e., up, flat, or down) to those that specify rates of weight gain by age. This may make growth monitoring impractical in many settings.

Introduction

Considerable debate continues to take place about the merits of growth monitoring as a component of programmes aimed at improving child health and nutrition in developing countries [1-3]. Some see growth monitoring as the cornerstone to which education, nutrition, and health interventions can be anchored, whereas others see it as a time-consuming, ineffectual ritual.

An often-cited advantage of growth monitoring is the simplicity of measuring weight. A scale plus pencil and paper are all the equipment required. However, analysing this information is not simple, and therefore great effort has been directed at developing easy-to-follow methods for the appropriate plotting of the serial data and interpretation of the results [4, 5]. A much greater problem is achieving adequate coverage rates at timely intervals among children who are likely to grow poorly. How best to use this information to initiate remedial and preventive actions to improve child health is yet another important area of focus and concern [2, 6].

Qualitative as well as quantitative criteria for judging growth have been proposed. The simplest method for interpreting the adequacy of weight gain stresses the direction of the growth trajectory [3]. If it is going up, it is good. A flat line or, even more so, a downward trend suggests concern. Quantitative criteria for identifying children who falter in growth have been proposed and used as well. These generally specify a certain amount of gain over a given period of time; criteria are usually specified by age (e.g., infants as opposed to older children). Some feel that these quantitative criteria are impractical for most programmes [7].

Although much concern has been expressed about the large number of children identified by growth-monitoring programmes [6, 7], studies on the percentages of children in developing countries who are found to be growing poorly according to commonly used criteria are lacking. Also, there is no information on the proportion of healthy children from developed societies who would be identified as faltering in growth by these same criteria. Knowledge about rates of faltering in the populations of interest, as well as in healthy populations, may be useful, together with other information (e.g., about resources available for focusing actions on the at-risk group identified), in defining the criteria to use. A further issue is whether the criteria, once defined, can be applied effectively and at a tolerable cost.

Methods

The objective of this study was to compare rates of growth faltering in children from developed and developing countries. Criteria for the selection of samples included adequate sample sizes, the availability of individual-level data, and serial weight data at three-month intervals from 12 to 24 months of age.

The age range 12-24 months was selected for study because growth rates over this time are more or less constant, permitting one to apply uniform criteria for growth faltering. Furthermore, the second year of life coincides with weaning and is associated with significant health and nutrition problems in developing countries. Also, together with infants, children 12-24 months old are often the priority target group of growth-monitoring programmes.

Many definitions of growth faltering have been proposed and used for children in the second year of life. For this investigation, we used two definitions of growth faltering over a three-month period: failure to gain weight or actual loss of weight, and weight gain less than 300 g. These criteria are identical to those used in the Tamil Nadu Integrated Nutrition Project (TINP) to select children for supplementary feeding.

 

Populations

Three populations were selected for study: in Berkeley, California, USA; Guatemala; and Tamil Nadu state, India.

Berkeley

The data from Berkeley are from a guidance study and refer to a core sample of subjects born in 1928 and 1929 for whom complete longitudinal data were kept from birth to adolescence [8]. The 66 boys and 70 girls in the sample were white and apparently from middle- and upper-class backgrounds. At 18 years of age, the average height was 179.0 cm for the boys and 166.6 cm for the girls. National data from NHANES I and 11 indicate that the average stature of United States white adults 18.0-24.9 years old was 176.9 cm for men and 163.3 cm for women [9], values lower than those of the Berkeley sample. The Berkeley sample is among the tallest in the world, taller than the English and as tall as present-day Dutch and Scandinavians [10].

All of the anthropometric data for the core sample are available in the public domain [8]. From these records, length and weight data were abstracted for subjects at ages 12, 15, 18, 21, and 24 months. Length was measured to the nearest millimetre and weight to the nearest tenth of a kilogram.

Guatemala

The Guatemalan data are from the INCAP longitudinal study [11]. The children came from four Ladino (i.e., Spanish-speaking, of mixed Spanish-Indian ancestry) villages from eastern Guatemala. Growth retardation was marked, particularly in the first three years of life [12]. The causes of these effects were infections, particularly diarrhoeal diseases, and dietary factors [13].

Examinations took place within ±15 days of the target date. Length was measured to the nearest millimetre and weight to the nearest hundredth of a kilogram. All available data from this study were used. A total of 1,021 individuals contributed data for one or more child-periods. For some analyses, only the 637 subjects with complete serial data were selected.

Tamil Nadu

The Tamil Nadu data were generated by the TINP through its activity of monthly weighing of children under 36 months of age [14]. The data were from the subset abstracted by Shekar [15] from 42 villages in the Kottampatti Block of Madurai district [15]. The information was extracted by trained personnel from village-based records for the 12 months from April 1986 to March 1987. Weight data at ages 12, 15, 18, 21, and 24 months were selected for analyses. Weight was measured to the nearest 50 g. Length was not measured. A total of 1,348 individuals contributed data for one or more child-periods.

As noted below, some analyses focus on the subset of children with complete data for all four periods. Because the information abstracted from the Tamil Nadu records was only for 12 months [15], the maximum number of child-periods for which an individual may have information is three; for this research, the ages 12-15, 1518, and 18-21 months were selected for analysis. A total of 209 children had data for all these intervals. To obtain data that would provide four child-periods per individual, one would have to extend data collection to 13 months.

 

Analyses

The WHO/NCHS reference curves [16] were used to contrast growth patterns in the three populations. The data for children under two years of age come from the Fels Research Institute and were collected from 1928 to 1978 [17]. There were no secular trends in increments over this period [18].

The distribution of weight increments in Berkeley, Guatemala, and Tamil Nadu were compared after classifying the values into any of four categories: negative or 0, 1-299, 300-499, and 500 g and above (the exit criterion from supplementation in the TINP is weight gain greater than 500 g over a three-month period).

Some analyses used all the available period data, and some used only those for children with data for all possible periods. The latter approach was followed to estimate the number of episodes of growth faltering, defined alternatively as a weight gain either of 0 g or less or of less than 300 g in a three-month period per child. For Berkeley and Guatemala, the possible number of episodes of faltering ranged from zero to four, but for Tamil Nadu, for the reasons already given, the range was zero to three. The raw data in Tamil Nadu were adjusted to make the results comparable to those of Berkeley and Guatemala. The probability of faltering in n periods (F_n) can be expressed as

F_n = 1 - pⁿ,

where p is the probability of not faltering in any one period. The TINP data available for analyses provided values for F₃ (i.e., for three periods), and this allows one to solve for p. The probability of faltering in four periods (F₄) was then estimated as

F_n = 1 - p⁴

The above method assumes that the probability of faltering in different periods is independent of other periods and that p is constant across periods. Although neither of these assumptions may be true, the degree to which they are violated is minor (e.g., successive increments are only weakly correlated after adjustment for common error terms) and should not appreciably affect the estimates provided.

Also, estimates of the probability of faltering in four periods were obtained through a different procedure. The average probability of not faltering (p) was obtained using pooled data for all child-periods, and F₄ was estimated as above. These values were only slightly higher than those obtained from longitudinal data.

Results

The median weights for the Berkeley, Guatemala, and Tamil Nadu children are plotted relative to the WHO/NCHS reference curves (ninety-fifth, fiftieth, and fifth percentiles) in figure 1 (see

Figure. 1. Median weights for boys in three populations compared with WHO/NCHS reference curves) and figure 2 (see

Figure. 2. Median weights for girls in three populations compared with WHO/NCHS reference curves). The medians for Berkeley were consistently above the reference median, while those for both the developing country samples were below the fifth percentile. Also, the medians for Tamil Nadu were consistently lower than those for Guatemala.

The differences in median weight for boys at 24 and 12 months were 2.8 kg in Berkeley and 1.7 in both Guatemala and Tamil Nadu, compared with 2.4 kg in the WHO/NCHS reference population. In girls, the weight gains were 2.6, 1.9, and 1.5 kg respectively in Berkeley, Guatemala, and Tamil Nadu, compared with 2.4 kg in the reference sample. Compared with incremental growth charts that use Fels Research Institute data [18], the weight growth velocities in boys and girls exceeded the fiftieth percentile in the Berkeley sample but were around the tenth percentile for the other two samples.

Length values were not available for Tamil Nadu. The median lengths for Berkeley exceeded the reference median, particularly after 12 months of age, but the medians for Guatemala were much lower than the fifth percentile.

The distribution of three-month weight increments is given in table 1. Here, the unit of analysis is child-period, as each subject may contribute as many as four values to the total. Analyses restricted to only individuals with complete data for all periods give results similar to those shown in the table. As expected, the weight increments were larger in Berkeley than in either Tamil Nadu or Guatemala. The degree of faltering was similar in boys and girls in all three populations.

TABLE 1. Distribution of three-month weight increments measured between 12 and 24 months of age

	<=0 g		1-299 g		300-499 g		>=500 g		Total
	No.	%	No.	%	No.	%	No.	%
Berkeley
all	38	7.0	37	6.8	65	12.0	403	74.2	543
boys	24	9.1	21	8.0	27	10.2	192	72.7	264
girls	14	5.0	16	5.7	38	13.6	211	75.6	279
Guatemala
all	453	13.7	601	18.2	622	18.9	1,623	49.2	3,299
boys	244	13.8	307	17.3	356	20.1	864	48.8	1,771
girls	209	13.7	294	19.2	266	17.4	759	49.7	1,528
Tamil Nadu
all	348	14.8	345	14.7	757	32.3	897	38.2	2,347
boys	174	14.2	184	15.0	404	33.0	464	37.9	1,226
girls	174	15.5	161	14.4	353	31.5	433	38.6	1,121

Periods are 12-15, 15-1A, 18-21, and 21-24 months.

Chi-square tests with three degrees of freedom were applied to the following: Berkeley boys vs. girls (5.66, NS), Guatemala boys vs. girls (4.93, NS), Tamil Nadu boys vs. girls (1.35, NS); Berkeley vs. Guatemala (120.76, p<.001), Berkeley vs. Tamil Nadu (231.81, p< .0001), and Guatemala vs. Tamil Nadu (149.13, p<.001)

The definition of faltering in the second year of life as a weight gain of less than 300 g over a three-month period is widely used. The percentages of all three-month increments that indicate faltering according to this definition are shown in figure 3 (see

Figure. 3. Percentages of all three-month weight increments of less than 300 g (based on data from table 1)). Some 17.1% of the increments in boys and 10.7% in girls in Berkeley were classified as faltering. The rates were two to three times as great in Guatemala and Tamil Nadu.

In table 2 the children are classified according to the number of episodes of growth faltering they experienced within the age ranges studied (12-24 months in Berkeley and Guatemala, with a possible range of 0-4 episodes; 1221 months in Tamil Nadu, with a possible range of 0-3 episodes), using both definitions of faltering: a weight gain of 0 g or less (i.e., no change or loss of weight) over three months, and one of less than 300 g. The analysis is restricted to individuals with complete information for all possible child-periods. The sexes are combined because of the general similarity in the distribution, particularly in Guatemala and Tamil Nadu. Although faltering was considerably more frequent in Guatemala and Tamil Nadu, it did occur in many Berkeley children, a few of whom had repeated episodes of faltering.

TABLE 2. Children classified by number of episodes of growth faltering, 12-24 months of age (Berkeley, Guatemala) or 12-21 months (Tamil Nadu) according to two criteria—growth <=0 g and growth < 300 g over three months

Episodes	Berkeley		Guatemala		Tamil Nadu
	No.	%	No.	%	No.	%
<=0 g
0	112	82.4	348	54.6	139	66.5
1	13	9.6	232	36.4	63	30.1
2	8	5.9	53	8.3	6	2.9
3	3	2.2	4	0.6	1	0.5
4	0	0.0	0	0.0	—
Total	136	100.1	637	99.9	209	100.0
< 300 g
0	86	63.2	112	17.6	77	36.8
1	31	22.8	290	45.5	96	45.9
2	14	10.3	184	28.9	34	16.3
3	4	2.9	46	7.2	2	1.0
4	1	07	5	0.8	—
Total	136	99.9	637	100.0	209	100.0

For Berkeley and Guatemala, based on weight measurements taken at ages 12,15,18, 21, and 24 months. For Tamil Nadu, the data were abstracted from the records in such a way that three-month increments can be determined only for ages 12-15, 15-18, and 18-21 months The sexes are combined in all the groups.

The percentage of children who faltered (<= 0 g and < 300 g) in at least one period from 12 to 24 months is shown in figure 4 (see

Figure. 4. Percentages of children faltering in growth in at least one three- month period between the ages of 12 and 24 months (based on data from table 2; for Tamil Nadu, the values based on three periods are projected to four).). The values for Tamil Nadu are adjusted for the fact that faltering was observed over only three periods; using the methods described earlier, the values shown project the rate of faltering to four periods of observation.

With a weight gain of 0 g or below as the criterion for faltering, 17.6% of the children in Berkeley faltered at least once between 12 and 24 months of Figure. 4. Percentages of children faltering in growth in at least one three-month period between the ages of 12 and 24 months (based on data from table 2; for Tamil Nadu, the values based on three periods are projected to four) age, compared with 45.4% and 42.0% in Guatemala and Tamil Nadu respectively. In other words, faltering in at least one period was about 2.5 times as likely to occur in Guatemala and Tamil Nadu as in Berkeley. When a gain of less than 300 g is used as the criterion, the percentage of children in Berkeley who faltered at least once rises to 36.8%, compared with 82.4% in Guatemala and 73.6% in Tamil Nadu. Using the second definition, the rate of faltering in at least one period was twice as great in Guatemala and Tamil Nadu as in Berkeley.

Discussion

Our objective was to compare rates of growth faltering in developed and developing countries. To this end, data from the Berkeley guidance study were compared with data for Guatemala and Tamil Nadu. An objection might be raised because the subjects from Berkeley were born in 1928 and 1929. Any infections that may have occurred in these children would have been treated symptomatically, because no effective antibacterial drugs existed at the time. Vaccines (e.g., for measles) were also not available. Whereas this suggests that growth faltering might be less common in Berkeley today than it was in 1928, there are other considerations to the contrary. These children were taller than the WHO/NCHS median, and as adults they were taller than today's average for the United States and as tall as northern Europeans. This is not surprising since the subjects were from an affluent university community. Clearly, the Berkeley population cannot be considered to have been constrained in physical growth in an important way. They grew faster than the Fels Research Institute sample (1928-1978), which is the basis for the WHO/NCHS reference curves for young children. Another important observation is that increments in the Fels data set did not change over the 50 years of data collection [18]. Perhaps growth is nearly the same in well-to-do children in Berkeley today as it was in 1928. Another potential objection is that both the Guatemalan and Tamil Nadu populations were benefiting from nutrition and health interventions. Undoubtedly, other populations lacking these services might show even greater rates of growth faltering. These potential objections do not undermine the present study. Choosing well-to-do samples with somewhat less growth faltering and malnourished samples with more growth faltering would not alter the general nature of our observations and conclusions.

One interesting aspect of the results is that, when the growth-faltering criteria that are used in Tamil Nadu to select children for supplementary feeding are applied to the subjects from Berkeley, we find that a significant percentage of them would have been selected. From 12 to 24 months of age, 17% of the children in Berkeley had at least one episode of faltering, using the criterion of a gain of 0 g or less over three months. All of these children would have been admitted to feeding in Tamil Nadu. Some 36.8% of them would have been placed at risk because they had at least one period when they gained less than 300 g, and would have been admitted to feeding if the next monthly weighing also indicated a similarly limited gain. However, the data are not available to check how many ultimately would have been chosen for supplementary feeding.

These findings do not invalidate the use of growth monitoring to select children for intervention in developing countries. The rates of faltering were considerably higher in Guatemala and Tamil Nadu. For example, the percentage of children with at least one period with failure to gain weight (<= 0 g) was 45.4% in Guatemala and 42.0% in Tamil Nadu, compared with 17.6% in Berkeley. When the criterion of a gain of less than 300 g is used, the majority of the subjects in Guatemala (82.4%) and Tamil Nadu (73.3%) were found to falter at least once, compared with 36.8% in Berkeley.

Some of the children in well-to-do populations who falter according to the criteria given are probably constrained in their growth by infections and dietary problems, among other factors. Paediatricians in industrialized societies should continue to monitor growth for the same reasons as in developing countries—to identify such children, to assess the possible reasons for their faltering growth, and to recommend appropriate actions. On the other hand, some children may have been classified as faltering because of measurement factors: actual measurement error or day-to-day variation (e.g., differences in weight caused by hydration levels, micturition, and defecation). Still another possibility is that growth may not be linear; healthy children may grow in spurts rather than continuously [19]. In other words, some sporadic growth faltering may have no functional implications, suggesting that, like any diagnostic tool, growth monitoring, as applied in Tamil Nadu, probably picks up a number of false positives.

Ultimately, programme planners must take the availability of resources into account in deciding what percentage of the population can be targeted for special attention or intervention. The application of either of the criteria discussed—weight gain of less than 300 g or of 0 g or less over three months during the second year of life—to Tamil Nadu and Guatemala identifies very large percentages of the population; for many settings, tending to the needs of so many would overwhelm resources. On the other hand, the TINP, an exceptionally well-planned and well-staffed programme, has applied these criteria successfully for many years [14].

Even if one is interested in growth monitoring only as an educational tool for promoting better child care, one should be concerned about a criterion that identifies too many in the population. If the instrument is too all-inclusive, there is little justification for its use, particularly if it requires considerable time and resources to apply.

Age should be considered as a targeting criterion. There is a consensus that growth retardation in developing countries is limited to the first two years of life in most populations and up to three years in others [20]. Monitoring the weight of children older than three years would appear to be of less priority in societies where this pattern of growth retardation applies. Therefore, focusing on children under the age of three years may be one way of decreasing the burden of growth monitoring. In addition, a means of identifying the most needy young children may be required, growth monitoring being one possibility.

In summary, the results indicate that commonly used criteria to define growth faltering in developing countries also identify a significant portion of children in developed societies; however, the proportion is two to three times greater in developing countries. Programmes in developing countries should consider selecting criteria that identify only that portion of the population that can be effectively handled. This necessarily implies departing from the simplest guidelines about the trajectory of the growth curve (i.e., whether it is going up, flat, or going down) and using instead guidelines that specify rates of weight gain at different ages, and this may make growth monitoring impractical in many settings.

Acknowledgements

This study was funded by grant HD22440 from the US National Institutes of Health and grant 9202716-000 from the Pew Charitable Trusts.

Useful comments and suggestions were given by E. Frongillo, M. Latham, D. Pelletier, and M. Ruel.

References

1. Growth monitoring: intermediate technology or expensive luxury? [Editorial] Lancet 1985;2:1337-8.
2. Nabarro D, Chinnock P. Growth monitoring: inappropriate promotion of an appropriate technology. Soc Sci Med 1988;26(9):941-8.
3. Rohde JE. Beyond survival: promoting healthy growth. Indian J Pediatr 1988;55(1,suppl):3-8.
4. Latham MC. Growth monitoring and promotion. In: Himes JH, ed. Anthropometric assessment of nutritional status. New York: Wiley-Liss, 1991:287-99.
5. Hendratta L, Rohde JE. Ten pitfalls of growth monitoring and promotion. Indian J Pediatr 1988;55(1, suppl):9-15.
6. Gerein N. Ross D. Is growth monitoring worthwhile? an evaluation of its use in three child health programmes in Zaire. Soc Sci Med 1991;32(6):667-75.
7. Gopalan C. Growth monitoring: some basic issues. Bull Nutr Found India 1988;8(2).
8. Tuddenham RD, Snyder MM. Physical growth of California boys and girls from birth to eighteen years. Berkeley and Los Angeles, Calif, USA: University of California Press, 1954.
9. Frisancho AR. Anthropometric standards for the assessment of growth and nutritional status. Ann Arbor, Mich, USA: University of Michigan Press, 1990.
10. Buckler J. A longitudinal study of adolescent growth. London: Springer-Verlag, 1990.
11. Martorell R. Habicht J-P, Klein RE. Anthropometric indicators of changes in nutritional status in malnourished populations. In: Underwood BA, ed. Methodologies for human population studies in nutrition related to health. NIH publication 82-2462. Washington, DC: US Government Printing Office, 1982:99-110.
12. Martorell R. Klein RE. Food supplementation and growth rates in preschool children. Nutr Rep Int 1980;21 :447-54.
13. Martorell R. Rivera J. Lutter CK. Interaction of diet and disease in child growth. In: Atkinson SA, Hanson LA, Chandra RK, eds. Breastfeeding, nutrition, infection and infant growth in developed and emerging countries. St. John's, Newfoundland, Canada: ARTS Biomedical Publishers, 1990:307-21.
14. Berg A. Malnutrition: what can be done? Iessons from World Bank experience. Baltimore, Md, USA: Johns Hopkins University Press, 1987.
15. Shekar M. Positive and negative deviance in child growth: a study in the context of the Tamil Nadu lntegrated Nutrition Project in rural south India. Doctoral dissertation, Cornell University, Ithaca, NY, USA, 1990.
16. World Health Organization. Measuring change in nutritional status. Geneva: WHO, 1983.
17. Hamill PVV, Drizd TA, Johnson CL et al. Physical growth: National Center for Health Statistics percentiles. Am J Clin Nutr 1979;32:607-29.
18. Roche AF, Himes JH. Incremental growth charts. Am J Clin Nutr 1980;33:2041-52
19. Lampl M, Veldbuis JD, Johnson ML. Saltation and statis: a model of human growth. Science 1992;258(5083): 801-3.
20. Beaton G. Kelly A, Kevany J. Martorell R. Mason J. Appropriate uses of anthropometric indices in children. ACC/SCN State-of-the-Art Series, Nutrition Policy Discussion Paper no. 7. Geneva: ACC/SCN, 1990.