Contents - Previous - Next
2. Duration of the insult
3. What is the height potential?
4. Bone age vs height age
5. Follow-up of malnourished children
6. Change of environment
7. Secondary malnutrition
8. Slave studies
9. Why do most subjects with post-natal stunting fail to catch up?
Department of Medicine and
Therapeutics, University of Aberdeen, Aberdeen AB9 2ZD, Scotland, UK
It has been consistently demonstrated that children in the developing world are, on average, shorter than those in affluent societies. Indeed, many view this poor height growth as a global measure of socioeconomic deprivation in these countries, just as it was in Europe and North America in the past and, to a lesser extent, the present.
Stunting starts in early childhood, or before, and usually persists to give rise to a small adult. The question to be addressed is whether the persistence of stunting is inevitable and permanent, a sort of 'growth scar', or whether catch-up can occur if conditions are conducive to it. Widdowson's studies, in which rat pups and piglets that were malnourished for a period shortly after birth never caught up, suggest that stunting in humans may indeed be permanent. However, rats are born at a much more immature stage than humans and grow throughout the* lives; they may be an inappropriate model for human stunting.
In children, Martorell et al. (1979) concluded that "catch-up growth in chronically malnourished children is limited and related to maturity delays". This conclusion was based on the fact that the delay in height appeared to be much greater than the delay in skeletal maturation. His later conclusion (Martorell, Rivera & Kaplowitz, 1990) that stunting is "a condition resulting from events in early childhood and which, once present, remains for life... There is no catch-up growth in later childhood and adolescence as some might have expected" is much stronger. He made this assertion because the growth increment of short, and not-so-short, Guatemalan children from 5 to 18 years was greater than the increment of US citizens of Mexican ancestry and only 3 cm less than NCHS standards, but totally independent of the subject's attained height at age 5 years. This assertion was further reinforced by the finding of a very high correlation between the height of children at 3 years and their final adult height and performance (Martorell et al., 1992). Such a high correlation between height in childhood and final adult heights is well known in other populations in both the developed and developing world (Mills et al., 1986; Satyanarayana, Prasanna Krishna & Narasinga Rao, 1986; Binkin, Fleshood & Trowbridge, 1988).
All these data could be interpreted
to show that a period of malnutrition in the first 2-3 years of childhood irrevocably
changes the child so that he is 'locked into' a lower growth trajectory with a lower
potential for future growth. The alternative hypothesis is that full catch-up growth is
possible, but is not observed in practice because the correct conditions for catch-up are
not satisfied; in most populations the environment and diet, associated with poor growth
performance initially, do not change.
Clearly, if an insult sufficient to retard growth is imposed after the major components of growth have occurred, then there will be no discernable effect. Equally, if the insult operates for only a short time, then there are unlikely to be any sequelae, unless it permanently changes genetic expression. The insult has to be imposed for a sufficiently long time for a discernable difference from the normal child (the standard) to be established. Just how chronic the insult has to be is often not appreciated.
Suppose that a completely normal child, of a certain age, reduces his growth rate so that he either ceases growth altogether (0%) or continues to grow at either 30%,50% or 70% of the rate at which he should have grown. The curves in Fig. 1a show the length of time, at each starting age, before the child falls 2 SD below the reference. As he gets older, the reduction of growth has to last longer and longer for him to be recognised as stunted. Thus, if a child stops growth completely at 12 months he reaches -2 SD after nearly 6 months (age 17.7 months); if he stops at 36 months, it takes 13 months (age 49 months) to reach the same point. However, an insult usually retards but does not stop growth. If growth continues at a reduced rate, then it will take longer for a child to fall below -2 SD. This is illustrated in the figure for various degrees of reduced, but continued, growth. Thus, if the 12 months old child reduces his growth rate to 30% of normal (reduced by 70%) he will take 10 months to become stunted (age 22 months), if the growth rate is only reduced to 70% of normal (reduced by 30%) he cannot be diagnosed as stunted until he is 52 months old.
Fig. 1. The time necessary for a child to fall from the median height-for-age to more than 2 SD below the median, if not gaining height at all (dotted line 0%) or gaining at 30% (short dash), 50% (long dash) or 70% (continuous) of the normal rate. The upper graph (a) shows the number of months it will take for a child to become stunted, in the future, if he commences to grow more slowly than normal. The lower graph (b) shows the number of months it has taken a stunted child, during his past life, to become stunted if growing normally and then at a reduced rate.
Fig. 1b is similar except that, instead of looking forwards in time to predict when a child will fulfil the criterion for stunting, it is looking backwards in time to show for how long an already stunted child, of a certain age, has been stunting if he has not been growing at all (0%) or has been growing at a reduced rate relative to the standard.
It is quite clear that an insult has to substantially reduce the growth rate for most of the child's life for him to become stunted. Successful reversal will require a complete and permanent removal of the retarding factors. This rarely happens.
The sheer chronicity of the process means that intercurrent infections, of themselves, are unlikely to cause stunting even if they are repeated. One should not expect a few episodes of acute diarrhoea to cause stunting (Briend et al., 1989; Moy et al., 1990), although, when this was first proposed it caused a flurry of dissenting correspondence in The Lancet (Schorling & Guerrant, 1990). Even when diarrhoea is statistically associated with stunting, only a small proportion (10%) of the retardation can be ascribed to this cause (Martorell et al., 1975). On the other hand chronic dysentery and persistent diarrhoea are, as expected, associated with stunting (Henry et al., 1987; Briend et al., 1989).
This chronicity is also important
for catch-up potential. If a child is 12 months behind in his height growth, then he will
have to grow at twice the rate of a normal child of his current height-age for at least 12
months to catch up; this is faster than twice the rate of a normal child of his
chronological age, because the rate of normal height gain decreases with age. The older,
and the further behind, the longer he will have to maintain a greatly accelerated rate of
growth for full catch-up. A stage may be reached where there is simply insufficient time
remaining to make a full recovery. It is also clear that increased rates of height gain,
to be maintained over a prolonged period, will usually require a permanent change in the
subject's environment; an input for a few months is only likely to lead to the start of
By comparing elite peoples from around the world, Martorell has produced convincing evidence that the genetic component of differences in children's height is, on average, trivial in comparison to the environmental influences. Nevertheless, the height potential of an individual is related to the parental height. Before deciding how complete the catch-up growth of an individual is, we need to know that individual's potential. In impoverished communities this can lead to a circular argument. The parents were themselves short because of malnutrition; this then, in part, determines the target that the child is aiming towards (his individual potential). If he reaches this target, has he had complete catch-up? If so, is the child then normal? These questions show two different concepts of catch-up. One is a catch-up to what is expected of the child and the other is catch-up to a standard height derived from a healthy population where there is no secular trend: the two may be very different.
The classical experiments of cross breeding Shire horses with Shetland ponies (Walton & Hammond, 1938) dramatically demonstrate the effect of the mother on the subsequent growth of animals. The progeny, as adults, weighed half as much again when the Shetland was the sire rather than the dam. Similar experiments with embryos transplanted into different breeds of sheep have confirmed these findings (Hammond, 1960). The human equivalent are mono-zygotic twins of different size; the smaller twin will end up as an adult of below average height (Babson & Phillips, 1973). Several years ago Stewart, Preece & Sheppard (1975) divided a colony of rats into two. One half were given a restricted protein diet and the other half a control diet. The progeny of the restricted rats were smaller than the control rats and remained so over many generations on the restricted diet - it was as if they became a different strain of rat altogether. When the restricted rats were returned to the control diet, subsequent generations were larger, but it took three generations for them to attain the size of the control group (Stewart et al., 1980). Inherited effects that gradually wash out have been seen in other functions, apart from growth. Beach, Gershwin & Hurley (1982) subjected mice to a brief period of zinc deficiency whilst pregnant and then cross-fostered the pups at birth to normal dams. The pups were immuno-deficient as adults: their own progeny were also partially immuno-deficient; normality was virtually restored by the third generation.
The usual explanation for the Shire/Shetland phenomenon, and that observed in Stewart's rats, is that the mother's small size deprives the fetus of nutrients because of a small placenta and reduced uterine blood flow. Yet the foals are normal, except for their size - they do not demonstrate the usual stigmata of malnutrition. The 'small placenta' explanation is insufficient to explain the 'inherited' immuno-incompetence. How could an insult to the grandmother during pregnancy affect the immuno-competence of the third generation? It is now becoming clear that the maternal (and paternal) environments affect the degree and timing of expression of developmental genes - an epigenetic phenomenon known as imprinting (Hall, 1991; Solter, 1988). The mechanism of imprinting is through methylation of specific DNA bases (Surani et al., 1990) which switch the genes off, or on, during early development; this switching can persist into adulthood or beyond. For example, although we inherit a gene from both our mother and our father, the one expressed, and the timing and degree of expression, are determined by epigenic modification during early development. Thus, only the gene for insulin-like growth factor II: derived from our father is expressed, whereas only the maternal gene for the receptor of this growth factor is expressed (Haig & Graham, 1991).
We are now beginning to appreciate the importance of interaction between nutrition and genetic expression. For example when Cheviot sheep, a breed that does not have horns, are put on a zinc-deficient diet they grow horns (Mills et al., 1967). This is despite the fact that zinc deficiency severely retards growth itself. Clearly, although this breed of sheep do not normally have horns, the appropriate genes are present. They have been bred to be permanently switched off. However, a nutritional insult, zinc deficiency, changes the expression of the hidden genes.
Variation in imprinting by specific epigenomic modification, perhaps stimulated by poor nutrition, provides a satisfactory explanation for Hammond's and Stewart's experiments, as well as the studies on zinc deficiency. As oogenesis occurs in fetal life, the nutritional plane of the grandparents may indeed influence the grandchild and be as important as the nutrition of the mother. If the meiotic and early in-utero environment changes the pattern of epigenic modification to alter the potential for future somatic development over several generations, it would also offer a satisfactory explanation for the close association between height potential and familial height in societies where there are no racial differences in height. It also provides an explanation for a gradual secular trend in height which clearly transcends the genes in the germ line and yet is familialy inherited.
Clearly, control of base methylation, and the substrates and coenzymes involved in methylation reactions, warrant examination in relation to height 'potential'.
The different controls of growth at different periods of growth (Karlberg, 1989) may indeed mean that, if abnormal growth is programmed by early developmental events to occur during a particular period, there may not be the developmentally programmed stimulus to catch-up when the child enters a further period. Martorell's data on the increments in Guatemalan children from 5 to 18 years of age (Martorell, Rivera & Kaplowitz, 1990), suggest that these subjects do not have an inherent 'drive' for increased growth. As we have seen, this may not be related to post-natal experience at all.
Is this relevant to the discussion of whether catch-up growth can occur in either a person or a chronically malnourished population? If a lower potential is determined during intrauterine life, then any catch-up which occurs in an individual will be to that potential and not to a greater one. To catch up to international standards, we may have to correct the defect over several generations. Indeed, a positive secular trend is a cross-generational catch-up in height. On the other hand, if we observe a short malnourished child who surpasses his potential, defined by parental height, we should say that he has had a complete catch-up for him, even though he fails to reach the NCHS standard. This is not to say that complete individual catch-up produces an individual who will function optimally.
In the Third World there is often an
adverse prenatal development; many of the affected children later present with severe
stunting. They may have a lower growth potential. The question posed can be narrowed to
examining the potential for reversing post-natal malnutrition in a child chat is born with
a 'normal' height potential.
Of crucial importance is the relationship between retardation in height and in 'maturity'. If both are delayed to exactly the same extent, we can view this relativistically as one where time has simply run more slowly for the child; the growth yet to come, the time available for that growth and the potential may all be adequate for full catch-up to occur. If, on the other hand, maturity is more advanced with the 'biological clock' running ahead of attained growth, the time available for growth is restricted, its intensity is likely to be inferior and catch-up much less likely. Martorell has specifically addressed this point (Martorell et al., 1979). He contends chat the bone age (maturity?) of stunted children is much less delayed than height growth. However, this interpretation is largely a function of the way in which the data are expressed. They are expressed as "relative retardation", defined as the difference between the age- and sex-specific means of the standard and the stunted populations, divided by the standard deviation of the standard. This is, in effect, a Z score of the mean, rather than a mean of the Z score. Expressed in this way, the bone age of Guatemalan children at age 36 months was only -1.7 SD units, when height was -3.3 SD units. It would appear that stunting was twice as severe as the delay in bone maturation. But the curves for bone age and height have quite different shapes; the standard deviations of the normal population are much larger for bone maturity (number of wrist ossification centres) than for height. A 36-month-old child with a height of -3.3 Z has the height of a normal child of 21 months. The same child with a bone age of -1.7 Z has a maturity of between 21 months (girl) and 26 months (boy). The length of the delays are not really different, although they appear so when expressed as Z scores. In Fig. 2, I plot the standards for bone age used by Martorell (Garn & Rohman, 1960; Yarbrough et al., 1973) and the bone maturity score of similar Guatemalan children (Blanco et al., 1972). What is quite clear is that bone age is very markedly delayed in these children as well as height.
This method of assessing 'maturity', and hence potential, is open to question. The number of ossification centres in the wrist is not necessarily closely related to the maturity of the long bones and spine which determine height. Certainly, there is a poor relationship between the delay in the maturity of the long bones and that of the skull or the teeth; furthermore, the different individual ossification centres are affected by malnutrition to different degrees (Dreizen et al., 1954; Dreizen, Spirakis & Stone 1964; Garn, 1981; Gupta et al., 1988). In malnourished children there is neither a steady advance in bone maturity nor in height; indeed, the two seem to be completely out of step seasonally; in Alabama height gain is least over the summer months when bone maturity rapidly advances, whereas when the height gain is most rapid few ossification centres appear (Dreizen et al., 1959).
There is some support for the proposition that bone maturity is much less retarded than height in malnourished children (Reichman & Stein, 1968; Adams & Berridge, 1969); however, most authors report gross retardation in bone maturity (Jones & Dean, 1956; 1959; Ghosh, Varma & Bhardawaj, 1967; Dreizen et al., 1956; 1958) that is either about the same (Briers, Hoorweg & Stanfield, 1975; Graham, 1972; Keet et al., 1971) or more severe (Alvear et al., 1986) than the delay in height age.
Given the general conclusion that
the delays in maturity and in height are not significantly different when expressed in
units of time, it seems that most malnourished children retain their capacity for full
What is observed in practice? MacWilliam & Dean (1965) followed up children after discharge in Uganda; there was only a marginal improvement, from 87% of standard to 90%, after 36 months. Six to 10 years later they were still markedly stunted (Krueger, 1969), with little evidence of catch-up to either local or international standards.
Very stunted children were followed by Graham's group in Peru (Graham & Adrianzen, 1971). Their height age was only 45% of chronological age at 1 year of age; by 3 years they had caught up to 61% and by 7 years to 69%. In the longer term (Graham et al., 1982) the girls did decidedly better than the boys (they were 2 cm shorter at 1 year and 7 cm taller at 13 years). The girls were taller than their own mothers from age 14 years; the boys lagged behind their fathers at age 16 (150 vs 158 cm). This sex difference has not been frequently reported. It is clear that the girls had a very marked catch-up, probably to their height potential.
In Chile (Alvear et al., 1986), follow-up for about 6 years showed also some evidence of catch-up (87 to 92% of standard); these children had a lower birth weight, and their mothers were much smaller than local controls, so that the potential for catch-up may have been particularly low.
In Jamaica, follow-up for 2 to 6 years showed that ax-patients were not different from Jamaican control children, although most were below the Boston Standard (Garrow & Pike, 1967). Further follow-up at 6-11 years confirmed that these children were at the local standard (Richardson, 1975).
In Cape Town (Keet et al., 1971), most patients were stunted - only 8% were above the 3rd centile of the Boston Standard on admission. At 5 and 10 years, 22 and 42% were above this cut-off point, and virtually all were in a higher centile group at the 10 year than at the 5-year follow-up. This catch-up is particularly notable because it is by older children, there was no intervention, and the children had a high incidence of hypoalbuminaemia at follow-up, showing catch-up despite a degree of continuing malnutrition. At the 15-year follow-up, the catch-up had continued with annual increments in height above those expected (Bowie et al., 1980). These findings are in marked contrast to those reported from Guatemala. Bone age was retarded to about the same degree as the stunting at all the examinations, and puberty was delayed. Interestingly, both the boys and the girls were continuing to gain substantial amounts of height after 17 years when the standards approach an asymptote. The final adult height of these ax-patients has not been reported.
In Kenya, Kulin et al. (1982)
conducted a cross-sectional survey comparing three privileged schools in Nairobi with an
impoverished rural district where there is a very high prevalence of malnutrition. Table 1
gives a summary of their height data. Before puberty the rural children were 1.7 and 2.0
standard deviations behind the well-nourished. By the age of 16 the rural children had
completely caught up. Puberty was delayed by 2 to 3 years in the malnourished group. This
response, although quite different from that reported from Guatemala, has been reported
before (Dreizen, Spirakis & Stone, 1967; Bowie et al., 1980).
6.1. Immigrant studies
Mjönes (1987) studied Turkish children born in either Turkey or Sweden. The children born in Turkey were shorter on arrival in Sweden but caught up so that they were not shorter at subsequent measurements. Follow-up of immigrants from Africa, Asia or China to a poor area of Glasgow showed the immigrants to be all more advanced in height and bone age than their Scottish counterparts (Goel et al., 1981).
Schumacher, Pawson & Kretchmer (1987) reported a longitudinal study of 835 refugee children entering the USA; they were analyzed in four racial and seven age cohorts, by sex. On arrival nearly all were below the 25th NCHS centile. They gained height at an accelerated rate, often achieving mean growth rates near the 90th centile for US children. There was no indication that the younger children had relatively higher rates of gain than the older children. Catch-up was clearly occurring.
These studies show the effect of a change in physical but not family environment.
6.2. Adoption studies
Graham's group (Graham & Adrianzen, 1971) conducted a unique experiment. They kept children from very poor families in a convalescent hospital for 18 months and compared their growth with that of siblings at home. At 18 months the hospital children had a height of -1.1 Z; the children at home were -2.3 Z. Five of the children remained in the hospital and 13 were discharged; 9 months later the five were still -1 Z whereas those returned home were now -2.2 Z. Giving these infants a 'good start' did not help their growth after discharge; their growth curve from 2 to 8 years of age was almost identical to that of the children that had not been kept in hospital (Baertl, Adrianzen & Graham, 1976).
Table 1. Height (cm ±SD) of school children from rich (urban) and poor (rural) schools in Kenya. Data of Kulin et al. (1982)
The opposite effect, going from a poor to a good environment, was also studied (Graham & Adrianzen, 1972). Fig. 3 shows at admission growth retardation of about -3.5 Z in two groups of eight children, aged 8 and 9 months. At 15 months the 'control' group were discharged to their parents; they had caught up in hospital to -2.9 Z; however, at about 9 years of age they were, as expected, still grossly stunted (-2.5 Z). The second group of children were kept in hospital longer, to age 26 months, and caught up more to -2.2 Z. They also were discharged home and at 4 years of age were still severely stunted (-2.1 Z). They were then adopted. In contrast to the children that remained at home, these children exceeded the local standards and almost completely caught up to modern American standards (-0.5 Z). This study shows how children, malnourished during the first years of life, can make a complete recovery before puberty by changing their home environment within the Third World. How these children fared subsequently is not reported.
Fig. 3. Height-for-age Z scores of two groups of eight children at admission to hospital (Adm), at discharge (Dis) to home, at transfer from parental home to adoptive parents (Traps) and at follow-up (Follow). The mean age of the groups is given in parentheses. Plotted from data presented by Graham & Adrianzen (1972). Figures in parentheses = age (months).
Winick's group (Winick, Meyer & Harris, 1975; Lien, Meyer & Winick, 1977) studied female Korean children that had been adopted by families in the USA when aged 2 to 5 years. They were remeasured 4 to 15 years later. Fig. 4 shows the results according to the degree of stunting at adoption. They had all caught up to Korean standards, the taller adoptees reaching well above the Korean median.
A similar study has been conducted by Proos, Hofvander & Tuvemo (1991b). They studied Indian children adopted by Swedish parents. The children were aged from 1 to > 7 years at adoption; more than half were between 3 and 6 years. Fig. 5 shows the results from the whole group and also from those children who were stunted, and those who were not stunted, on arrival in Sweden. The children had a very pronounced catch-up within 2 years of arrival in Sweden. The children with the largest deficits caught up the most (r = 0.53). By the age of menarche the group as a whole had caught up further, to be now only 0.3 Z below the NCHS standard. Thus, by puberty (11.6 years), there had been an almost complete catch-up to a level well above local Indian standards. However, they had an early puberty (Proos, Hofvander & Tuvemo, 1991a) and their final adult stature was -1.4 Z. As adults 31% of the children were more than 2 Z below the NCHS standard. There was a close relationship between the height when adopted and this final adult height (r = 0.78).
Fig. 4. Heights of Korean children adopted by families in the USA on arrival in the USA and when tested 4 to 15 years later. The height is expressed as centiles of Korean Standards; the subjects are divided into three groups based upon their degree of stunting upon adoption. Data of Lien, Meyer & Winick (1977).
The absolute increment from adoption to adulthood is not reported, but it would appear to be about that expected from NCHS standards. This has not come about by the simple maintenance of a deficit acquired in childhood, as assumed by Martorell (Martorell, Rivera & Kaplowitz, 1990), but rather by a spectacular catch-up followed by poor pubertal/post-pubertal growth. The poor pubertal growth occurred after they had been in affluent surroundings for many years. From the data at menarche it would appear that these adoptees fulfilled their genetic potential. Is it possible that their subsequent fall-off was through genomic imprinting of a pattern appropriate to India?
6.3. Abused children
Deprivation also occurs in affluent societies. King & Taitz (1985) studied 96 such children in England, 64 of whom were discharged home, and 31 received either foster or mixed care. Fig. 6 shows the height-for-age Z scores (Tanner Standard) when the children came under care at 2 years and again at follow-up around 5 years of age. There was a considerable catch-up in the children's height. This was more marked in the children that were fostered.
Fig. 6. Heights of abused children on admission (admit), when aged 24 months, and a follow-up (follow) at age 56 months. Children were either discharged back to their home, with continuing surveillance, or to foster homes. The fostered children were initially more severely affected but were less stunted at follow-up. Plotted from data of King & Taitz (1985).
Contents - Previous - Next