Contents - Previous - Next

Reversibility of stunting: Epidemiological findings in children from developing countries

1. Introduction
2. The timing of stunting
3. Age at menarche
4. Continued residence in the environment that gave rise to stunting
5. Continued residence in the same environment with improvements in nutrition
6. Relocation from the environment that gave rise to stunting
7. Discussion and conclusions

R. Martorell, L. Kettel Khan and D.G. Schroeder

Centre for International Health, Emory University, School of Public Health, 1599 Clifton Road, NE, Atlanta, GA 30329, USA

The growth literature from developing countries is reviewed to assess the extent to which stunting, a phenomenon of early childhood, can be reversed in later childhood and adolescence. The potential for catch-up growth increases as maturation is delayed and the growth period is prolonged. However, maturational delays in developing countries are usually less than two years, only enough to compensate for a small fraction of the growth retardation of early childhood. Follow-up studies find that subjects who remain in the setting in which they became stunted experience little or no catch-up in growth later in life. Improvements in living conditions, as through food supplementation or through adoption, trigger catch-up growth but do so more effectively in the very young. One study cautions that in older adopted subjects, accelerated growth may accelerate maturation, shorten the growth period and lead to short adult stature.

Correspondence to: R. Martorell.

1. Introduction

Stunting, the most easily measured and widespread effect of the malnutrition-infection complex, predicts generalized functional impairment on a wide range of biological, behavioral and social dimensions in children and adults from developing countries (Waterlow, 1992; Pollitt, 1990; Martorell, Rivera & Kaplowitz, 1990; Martorell et al., 1992). Some of these functional correlates, such as poor school performance, are not direct outcomes of growth failure, but instead reflect shared causes, whereas others, such as reduced lean body mass in the adult, undoubtedly are.

Are stunting and the adverse consequences it marks subject to partial or total recuperation in children who survive the turbulent first few years of life? This is a question of great scientific and policy importance. It is also a very broad question, and here only the issue of the reversibility of stunting per se will be addressed. A more comprehensive approach would also assess how reversing stunting affects its functional correlates.

Tanner (1986) has written optimistically about the possibilities for reversing stunting. He states: "Deflect the child from its natural growth trajectory by acute malnutrition or a sudden lack of hormone, and a restoring force develops, so that as soon as the missing food or the absent hormone is supplied again, the child hastens to catch up towards its original growth curve".

It is clear that catch-up growth can occur under specific conditions. What is less clear, however, is the degree to which catch-up growth can compensate for previous losses and whether catch-up growth can occur under a variety of conditions. Is the severity of growth failure an important predictor of the potential for catch-up growth? Does the timing of the growth retardation matter? For example, is it possible to recuperate growth retardation incurred in early childhood during the periods of middle childhood or adolescence? Is the degree to which maturation is delayed, and thus the extent to which the duration of the growth period can be prolonged, a key element in determining the potential for catch-up growth?

The main purpose of this paper is: (1) to assess the degree to which the severe, linear growth failure that commonly occurs in early childhood can be redressed through catch-up growth; and (2) to identify the epidemiological factors or ecological settings which predict the degree of catch-up growth. This is an exemplary review which focuses on epidemiological data from developing countries. Follow-up studies are examined in greater detail, in particular, studies that measured the same subjects in early childhood and later in adolescence or adulthood. Three unique settings are examined: (1) continued residence in the environment that gave rise to stunting; (2) continued residence in the same environment but with improvements in nutrition; and (3) relocation to an improved environment.

This review is not exhaustive on the subject of catch-up growth. Important omissions are studies of the growth response following treatment of such illnesses as celiac disease or hormonal deficiencies (e.g., growth hormone, hypothyroidism). Though this omitted literature is vast, many reports deal only with selected cases, and statistical descriptions of defined populations are rare. This is likely to lead to a bias towards the publication of cases with spectacular or more complete catch-up growth than commonly is the case. In commenting on this literature, Tanner (1981) observed that the potential for catch-up diminishes as a function of the severity, duration and timing (i.e., earlier in life limits potential) of the growth retarding influence. Michael Golden, in the following paper, deals with the clinical and endocrinological literature.

The studies reviewed used a variety of approaches to assess the degree to which catch-up growth had taken place. Some used an internal comparison; for example, contrasting the stature of adults grouped according to their degree of stunting at 5 years of age. In other cases, height during childhood, adolescence and young adulthood was directly compared to values from an external reference population, usually from the United States. This latter approach raises some concern. Whereas growth potential in early childhood is surprisingly similar across ethnic groups (Martorell, 1985), little is known about the importance of ethnic or genetic differences in growth during adolescence. Thus, inferences about catch-up growth from analyses which use an external reference population to assess growth during adolescence need to be made with extreme caution 1.

1 Where appropriate, we specify which reference population is used. The "WHO/NCHS" reference population is used worldwide to assess the growth of pre-pubescent children (Hamill et al., 1979). Another, identified simply as "US reference values", is needed because curves for adolescents and adults are not included in the WHO/NCHS reference. It should be noted that these two reference populations are distinct despite the fact that both use NCHS data. The WHO/NCHS is a combination of Fels Research Institute data with various NCHS surveys, including NHANES I but not NHANES II. The US reference curves published by Frisancho (1990) include only NHANES I and H. Percentile values for pre-pubescent children are very similar in both references.

2. The timing of stunting

Stunting is a phenomenon of early childhood and a direct result of poor diets and infection (Martorell & Habicht, 1986). Though the prevalence of low birthweight is increased in poor societies, newborn lengths are often similar to well-to-do samples. The intense period of growth retardation is generally between 3 and 12 or 18 months. In some countries growth retardation continues into the third year or longer but to a lesser extent. At the end of this process, marked departure from normality will have often occurred. Median regional prevalence of stunting, defined as children with lengths two or more standard deviations below the WHO/NCHS median, in children 12-23 months of age is 41 % for Africa, 3 i% for Latin America, 47% for Asia and 39% for the Eastern Mediterranean (Victora, 1992).

There are multiple reasons why stunting occurs in early childhood and not later. In childhood nutritional needs are greater, in relation to weight, than at any time later. One of the reasons that nutritional requirements are high is that growth velocities are the highest they will ever be, including peak height velocity in adolescence. Thus, the opportunity for growth retardation is great in early childhood, partly because more growth is taking place. Because of high nutritional requirements and limited gastric capacity, energy and nutrient dense foods are required to complement breast milk during weaning, yet caretakers in developing countries often serve foods that are too bulky for young children. In addition, infections, particularly gastrointestinal ones, limit growth in young children because episodes are more frequent and more severe in the very young, especially the malnourished. Finally, young children are totally dependent on others for their care and are hence most vulnerable to poor caretaking. Older children can make their needs known more easily and are better able to take the initiative to procure more food.

3. Age at menarche

Maturation can be assessed through a variety of approaches including skeletal age, appearance of secondary sexual characteristics and menarche. However, the majority of studies of adolescents, such as the ones reviewed here, only report menarche.

Tanner (1981) has summarized the available data for average age at menarche in 19th century Europe for several groupings of countries. Average ages at menarche in the United Kingdom, Scandinavia, Germany and Russia were between 15.0 and 16.8 years in working women and between 14.3 and 15.0 years in middle class women. For France, Spain and Italy, corresponding ranges were 14.5 to 15.4 years for working class samples and 13.5 to 14.3 years for middle class samples. The most complete overview of age at menarche for modern populations is given in Eveleth and Tanner (1990). The range given for European samples (all social classes) collected since 1970 is 12.1 to 13.5 years, indicating a decline of several years with respect to 19th century data. In Oslo, where serial data are most complete, the mean menarchal age declined from 15.6 years in 1860 to 13.2 in 1960. Recent values for the US are 12.8 years for whites and 12.5 for Blacks.

Eveleth & Tanner (1990) also provide menarche data for various countries and regions of the world. For African samples since 1970, values range from 13.1 to 14.5 years. Values for India are 12.5 to 14.6 years, while for Latin American countries they are 12.0 to 13.4 years, except in a sample of Oaxacan Indians from Mexico where mean age at menarche was 14.3. The data available are incomplete but they do suggest that the environment has a greater effect on maturation than ancestry.

For purposes of this review, samples with mean ages at menarche of 15 years or higher are assumed to have had marked maturational delays, whereas those between 13.5 and 14.9 are assumed to have had mild to moderate delays.

4. Continued residence in the environment that gave rise to stunting

The first situation to be considered is one in which subjects continue to live in the environment in which they became stunted. In this instance, two types of populations may be considered: those in more extreme settings, in which puberty and/or menarche are markedly delayed (i.e., age at menarche of 15 years or more) and those in populations more typical of agricultural societies in developing countries, where delays in menarche and/or maturation are less marked.

4.1. Populations from developing countries where delays in menarche/maturation are presumed to be marked

Substantial catch-up growth is possible in unusual circumstances where menarche and maturation are very delayed. Studies of nineteenth century American slaves by Steckel (1987), and of Turkana pastoralists by Little, Galvin & Mugambi (1983) demonstrate this phenomenon. A possible exception is an account of the Bundi of Papua New Guinea by Malcolm (1970). In each of these populations, there is marked stunting in early childhood compared to US reference values as well as a pronounced delay in maturation.

The American Black slaves described by Steckel (1987) were approximately 15 cm below the NCHS reference mean before puberty (Fig. 1). Differences began to increase at about 9 years in females and 11 years in males, reflecting delayed puberty, but were diminished considerably by the later teens. The initial differences of about 15 cm were reduced to about 5 cm at the end of the growth period. Nutritional status and prevalence of illness data are not available for this sample population, but other reports note that poor prenatal care, early weaning, inadequate and contaminated food supplements, and a heavy disease burden hampered growth in early childhood (Swados, 1942; Postell, 1951; Kiple & King, 1977). However, the diets were deliberately improved when the slaves entered the labor force at 8 to 12 years of age. Still, age at menarche appears to have been markedly delayed; Trussel & Steckel (1978) report values of 15 years for some US slave populations. Another indication of delayed maturation is the observation that growth seems to have continued until about age 21 years in males and 19 in females.

Fig. 1. Differences in mean height between 19th century American slaves and US reference values (Steckel, 1987).

In the American Black slaves, a prolonged growth period appeared to provide an opportunity for marked catch-up in growth. Differences in growth potential with respect to the reference population are not a likely explanation. American Blacks today attain adult heights that are similar to those of US Whites.

Little, Galvin & Mugambi (1983) examined Turkana pastoralists in Kenya who are tall, lean people. The basic diet of milk, meat and blood is derived from their cattle; however, hungry periods are common. The degree of pre-pubescent growth retardation is initially about 10 cm but increases to about 15 cm by age 13 years in both sexes (Fig. 2) 2.

2 Readers are cautioned that the data in Fig. 2 are based on small sample sizes and uncertain ages.

The pattern in older boys is erratic, probably due to small sample sizes, but it does suggest that differences continue to increase till about 17 years of age. Final achieved stature is nearly the same as in the US reference population, suggesting nearly complete catch-up. An explanation for this is that age at menarche was 15 years, delayed compared to well-off populations, permitting a longer growth period. An alternative explanation for the apparent catch-up growth is that growth potential in the Turkana and other Nilotic peoples differs from that of the US population and what might be interpreted as catch-up growth during adolescence may, in fact, be an expression of genetic potential. If this were the case, a Turkana adult population which did not experience early childhood growth retardation would be significantly taller than the US population.

Fig. 2. Differences in mean height between Turkana pastoralists from Kenya and US reference values (Little et al., 1983).

The Bundi subsist on sweet potatoes and the diet has a very low fat content; both energy and protein are probably limiting in young children and perhaps at older ages as well (Malcolm, 1970). Growth is extremely retarded, with the differences with respect to the NCHS reference mean increasing progressively with age, until reaching a high of 37.7 cm in girls at 12 years of age and of 40.8 cm in boys at 14 years of age (Fig. 3). Differences with respect to US reference means diminish later in adolescence and in adults are 20.8 cm in males and 15.4 cm in females. Maturation is markedly delayed, based on assessment of skeletal age and of sexual maturation. Menarche, for example, occurs at 18 to 19 years of age. Peak height velocity is delayed by about 2 years in comparison to British children, and growth continues into the early twenties. The prolongation of growth provides an opportunity for catch-up growth but not as much seems to be achieved through this mechanism as in the two populations discussed above (i.e., the growth failure of early childhood is not made up by later growth). However, growth potential during adolescence may be less for the Bundi than for US populations (Malcolm, 1990); this would underestimate the extent of catch-up growth.

Fig. 3. Differences in mean height between Bundi from Papua New Guinea and US reference values (Malcolm, 1970).

4.2. Populations from developing countries where delays in menarche/maturation are presumed to be moderate/minor

The circumstances described above are unusual and are not characteristic of what is generally observed in developing countries. Marked to moderate stunting in early childhood with limited delays in menarche and other maturation indicators is a more common pattern.

4.2.1. Cross-sectional studies illustrating common patterns of growth retardation The diet of rural Embu Kenyans is based on cereals and tubers, and energy and many micronutrients are probably deficient (Neumann, Bwibo & Sigman, 1992). Retardation in growth is pronounced at an early age. At 18 months of age, the average height-for-age Z score (with reference to the WHO/NCHS reference data) was -2.0, decreasing to about -1.8 at 84 months (values similar for boys and girls). As adults, only a minor degree of catch-up growth seemed to have taken place; average Z scores (with reference to the 1959 Metropolitan Life Insurance Data) were -1.5 in men and -1.4 in women. In other words, the retardation of early childhood growth appears to persist into adulthood with only modest attenuation.

Fig. 4. Differences in mean height between Malawians and US reference values (Pelletier et al., 1991).

In a rural population from Malawi, linear growth retardation increased throughout the first 60 months of age (Fig. 4) (Pelletier, Low & Msukwa, 1991). This is unusual; in most studies, differences are constant from about 24 months. In adulthood, there is a slight increase in the average differences in height with respect to the US age-specific mean compared to those seen at 60 months of age in males, but a slight decrease in females. Menarche data were not reported in this or the Kenyan study. According to Eveleth & Tanner (1990), age at menarche in African samples ranges from 13.1 to 14.5 years.

Cross-sectional data from rural Guatemala, where menarche occurs approximately one year later (13.5 years) than in well-off populations (Khan et al., in press), show that prior growth retardation is not recuperated (Fig. 5). The difference in height at 11 years of age with respect to the US reference population is nearly constant from 11 years to adulthood; the increasing differences in males at ages 12-15 years are probably caused by delayed timing of the adolescent growth spurt in the Guatemalan sample. It remains unclear whether growth potential during adolescence of Mexicans and Guatemalans is the same as that of the US population (Martorell, Mendoza & Castillo 1989). If data from the Hispanic Health and Nutrition Examination Survey (HHANES) are used for comparison instead, the difference at the end of adolescence in Guatemalans becomes 8-9 cm instead of the 13-14 cm shown in Fig. 5. Since there are no differences between HHANES and the US reference population prior to 11 years, this would suggest modest catch-up in growth.

Fig. 5. Differences in height between Guatemalans and NCHS reference data.

The use of cross-sectional studies to infer catch-up growth, as done above, assumes that all age groups suffered the same degree of stunting in early childhood. This may not be the case in populations undergoing significant changes in the quality of life. For example, Mexican-Americans have increased in height dramatically in the last 20 years, probably as a result of better growth in early childhood. The smaller size of Mexican-American adolescents with respect to the reference population may reflect, not reduced growth during adolescence per se, but poorer growth in early childhood than experienced by younger cohorts. Follow-up studies of individuals initially measured in childhood would therefore appear to be better in design.

4.2.2. Follow-up studies Satyanarayana and colleagues conducted a longitudinal study in children from 26 villages near Hyderabaad, India. Heights of 197 girls at 5 years and at 18 years of age were analyzed (Satyanarayana et al., 1981). Subjects were divided into four groups according to the degree of stunting at 5 years of age as follows: I (above -2.0 SD), II (-2.0 to -3.0 SD), III (-3.0 to -4.0 SD), and IV (below -4.0 SD) using the Harvard reference data (Reed & Stuart, 1959). The height differential of 14 cm seen at age five years between the most retarded group (90.1 cm) and the least retarded group (104.3 cm) was reduced to 7.7 cm by 18 years, 148.2 and 155.9 cm, respectively. In all groups, the differences with respect to the Harvard reference mean were less at 18 years than at 5 years, particularly in groups III and IV (Fig. 6). The authors suggest that the most plausible explanation for the apparent catch-up growth is delayed menarche (Satyanarayana et al., 1981). The authors state that the pubertal growth spurt was delayed and that the growth period was prolonged by about 1.5 years, though no information on age at menarche or incremental growth was given.

In the same population, Satyanarayana, Nadumuni Naidu & Narasinga Rao (1980) examined the growth pattern of 667 Indian boys. The boys were classified in the same fashion as the girls. The height differential of 16.5 cm seen at age 5 years between the most retarded group (88.5 cm) and the least retarded group (105.0 cm) was only slightly improved (15.5 cm) by 18 years (149.0 and 164.5 cm, respectively). The boys who were shortest (Group IV) at age 5 continued to be so at age 17 years. The most retarded group (below -4 SD) had a lower peak height velocity (6.9 cm/year) than British boys (7.3 cm/year) and its timing was delayed (14 vs 16 years). In spite of the maturation delay, the differences with respect to the Harvard reference data increased slightly form 5 to 18 years (Fig. 7). In summary, the growth retardation of early childhood was slightly increased by adulthood in males but decreased in females.

Fig. 6. Differences between groups* of Indian girls at 5 and 18 years and US reference values** (Satyanarayana et al., 1981).

* Groups: I (> -2.0 SD), II (-2.0 to -3.0 SD), III (-3.0 to -4.0 SD), IV (< -4.0 SD).
**Harvard reference data (Reed & Stuart, 1959).

Fig. 7. Differences between groups* of Indian boys at 5 and 18 years and US reference values** (Satyanarayana, 1980).

* Groups I (> -2.0 SD), II (-2.0 to -3.0 SD), III (-3.0 to -4.0 SD), IV (< -4.0 SD).
**Harvard reference data (Reed & Stuart, 1959).

In a later analysis of boys from the same population, Satyanarayana et al. (1989) described the dynamics of growth during puberty through 19 to 24 years using a Preece and Baines model. Subjects who were previously measured biannually throughout their preschool years during 1965-1969 were measured again annually from 1976 to 1984 (n = 323). The boys were classified according to the degree of stunting in childhood. Three groups (not four as in earlier studies) were formed: I (above -2.0 SD), II (-2.0 to -4.0 SD), and III (below -4.0 SD). Adult height was defined as height at 18 years or older (the range was 18 to 24 years). The gain in height from 5 to 18 years exceeded that of British children only in the case of the most stunted group (Table 1). Growth after 5 years was related to the degree of stunting, such that the differences between groups I and III was reduced from 14.4 cm at 5 years to 9.7 cm at 18 years or older, suggesting a moderate degree of catch-up growth.

Table 1. Height (cm) at 5 years and at adulthood in an Indian sample*

Group at 5 yr**

Height at 5 yr

Adult Height

Gain 5 yr to Adult

















* Adapted from Satyanarayana et al., 1989.

** Definition: I (>-2.0 SD); II (-2.0 to -4.0 SD); and III (< -4.0 SD) of Harvard reference standards (Reed and Stuart, 1959) for height.

The parameters of the Preece-Baines model are given in Table 2. Age at take-off is only slightly later in the most stunted subjects compared to the rest of the Indian sample and to British children (11.3 years vs 10.6/10.7 years).

Table 2. Comparison of adolescent growth in Indian and British boys using the Preece-Baines model*

Group at 5 yr**

Age at take off (yr)

Height at take off (cm)

Peak height velocity (cm)

Puberty duration (yr)

Growth during puberty (cm)

End of growth (yr)

Adult height (cm)

































* Adapted from Satyanarayana et al., 1989.

** Definition: I (> -2.0 SD); 11 (-2.0 to -4.0 SD); and III (< -4.0 SD) of Harvard reference standards (Reed and Stuart, 1959) for height.

Fig. 8. Differences between US reference values and Guatemalan groups* at 5 and 18+ years.

* Groups: I (> -2.0 SD), II (-2.0 to-3.0 SD), III (< -3.0 SD).

The primary factor leading to the short adult stature in this population is the growth retardation already present at take-off. During adolescence, all Indian groups grew more during adolescence compared to British children, with the most stunted groups growing slightly more than the least stunted Indian group. The additional growth is apparently achieved, not by greater peak height velocity, but by a lengthening of the puberty growth period which is 1.3 years longer in the most stunted Indian children than in British children.

Data from Guatemala indicate that there is no catch-up growth during later childhood and adolescence (Martorell, Rivera & Kaplowitz, 1990). As in the Indian studies, subjects were grouped by the degree of stunting at 5 years of age: I (above -2.0 SD), II (-2 to -3 SD) and III (below -3 SD) using US reference values. Growth from 5 years to adulthood was similar in all three groups, so that the differences between groups were similar in adulthood to those at 5 years of age (Fig. 8). It can be observed that the relative ranking of groups is maintained (5 years vs adulthood) and that the differences with respect to the reference data increased by about the same amount in all three groups.

Contents - Previous - Next