Contents - Previous - Next

Between-population variation in pre-adolescent growth

1. Classifying human populations
2. Population differences in growth patterns
3. The validity of the concept of an international growth reference

S.J. Ulijaszek

Department of Biological Anthropology, University of Cambridge, Downing Street, Cambridge CB2 3DZ, UK

Between-population differences in rates of physical growth and development and attained body size are well documented, but it is difficult to determine the extent to which these differences can be attributed to genetic and environmental factors. The greatest differences are to be found between populations in industrialized and non-industrialised nations, and between well-off and poorer groups within countries. Although genetic factors cannot be discounted, such differences can largely be attributed to differences in environmental quality experienced, influencing growth largely through differentials in nutritional well-being and exposure to, and treatment of, infectious disease. Growth patterns of well-off populations and groups of high socio-economic status are less heterogeneous, but differences between major global population groupings may still exist, bringing into question the validity of the concept of an international reference for the growth of young children. In this article, information pointing to genetic differences in the growth of children of different populations is summarised, and the acceptability of the NCHS (National Center for Health Statistics, 1977) references for height by age for international use is examined.

It is concluded that the growth patterns of all major population groupings are likely to have similar genetic potential, with the exception of Asiatics. However, there are no data on either secular trends or well-off groups from populations that have until recently been genetically isolated, and it is not known whether they share the same potential for growth as the major populations that surround them. In addition, very little is known about the genetic potential for growth of Aboriginal populations in Australia, or in Pacific Islands populations. It is suggested that the growth references for height by age in current international use are at best only imperfect yardsticks for nutritional assessment.

1. Classifying human populations

Processes which have taken place in the past have shaped the human populations found in the world today. The most fundamental process has been genetic, namely the shaping of human gene-pools through natural selection (Harrison et al., 1988). In addition, human population migration has served to isolate different groups across sufficiently long time-spans to allow between-population genetic variation. Fig. 1 shows the dispersal of anatomically modern humans out of Africa, starting about 70,000 years ago.

Our knowledge of the pattern of human dispersal gives us clues about how human populations might best be classified: (1) since evidence points to humans having a common point of origin, with dispersion only coming late in evolutionary time, human populations, regardless of where they live, have tremendous genetic similarity; (2) the migratory route suggests that variation in any trait is likely to be a continuum across populations, and that any classification is to some extent arbitrary; and (3) prior to the onset of agriculture, some 12,000 years ago, humans lived as hunter-gatherers at low population densities often in isolation from each other, leading to the possibility of the development of regional population genetic differences.

Fig. 1. Dispersal of anatomically modern humans from Africa (approximate dates in thousands of years).

Migrations taking place after the onset of agriculture served to create larger, more genetically homogeneous populations across wide areas, with genetically isolated populations left in less-hospitable ecological niches. Genetically isolated populations to be found in the world today include tribal groups of hunter-gatherers in Africa, Latin America, and Asia. Large-scale colonisation of the Pacific Islands took place at this time. Later migrations, during colonial times, include the migrations of (1) Europeans to the Americas, Australasia and parts of Africa; (2) Africans, mostly of Bantu origin, to the Americas and the Caribbean; and (3) Asiatics, Chinese in particular, and Indo-Mediterraneans, largely South Asians, to most parts of the tropical world and to parts of the New World. Migrations in the post-colonial period are largely related to economics and urbanisation. Examples include the migrations of Mexicans and Hispanics to the USA, and of South Asians and Afro-Caribbeans to Britain. In addition, there is the global trend of rural to urban migration in post-colonial times.

In summary, therefore, human populations have great homogeneity for most genetically determined characteristics, between-population variation having taken place in the recent evolutionary past. Migrations and population expansions have created several larger population groupings within which genetically isolated groups may be found.

Different human populations are known to vary from each other in a large and ever-growing list of genetic markers, growth patterns, body size and composition are polygenic in character, and the genetic component of human growth in different populations is a debated but poorly resolved issue (Schmitt & Harrison, 1988; Harrison & Schmitt, 1989; Bogin, 1991). It is possible to classify populations around the world in a number of different ways; Table 1 gives a modification of the Eveleth & Tanner (1990) classification.

Table 1. Classification of population types



Other terms used

European and European origin

Living in Europe or elsewhere, of European ancestry

Caucasian, White, Anglo-American, European-American

African and African origin

Living in Africa or elsewhere, of African ancestry

Negro, Black, Black British, Afro-Caribbean, Afro-American, African-American


Living in Asia or elsewhere, of Asian ancestry

Asian, Indian (American)


Peoples of the Near East, North Africa, the Indian subcontinent and their descendants

Indian, Asian, Indo-Pakistani, Arab, South Asian

Australian Aborigines and Pacific Island peoples

Melanesian, Polynesian, Micronesian

This typology is rather rigid and simplistic. It does not comfortably include mixed populations such as Spanish-Indians in the Americas, European-Africans in the Caribbean, United States and Britain. Further, it aggregates populations that have been shown to have clear differences in growth pattern. For example, the term 'African' includes the majority of Bantu-descended populations of Africa, as well as distinctively short-statured hunter-gatherer groups such as the Mbuti of Zaire and the!Kung bushmen of Namibia. Furthermore, the term 'Australian Aborigines and Pacific Islanders' covers populations with considerable genetic heterogeneity. However clumsy the classification is, it is the best that is currently available, and is used in this article in considering broad population growth characteristics.

2. Population differences in growth patterns

The number of studies of human growth and body size carried out worldwide during the latter part of this century runs to three figures. However, many of them are of groups and populations living in poor environmental circumstances. Possible differences in growth patterns which might be attributed to genetic factors can be examined in a number of ways, including: (1) considering a body size measure such as height, of children at a given age from industrialized countries, and from the highest socio-economic groups in developing countries; and (2) examining evidence for a secular trend in body size in European and non-European populations, and whether it has reached a plateau.

A comparison of mean heights of 7-year-old-boys of different population types from industrialised countries and from the highest socioeconomic groups in developing countries is given in Fig. 2. The range of means for 28 Europeans and European-origin populations is 119.1 to 126.5 cm, similar to those for African and African-origin populations (119.6 to 126.0 cm), and Indo-Mediterranean populations (120.2 to 126.0 cm), but higher than that for Asiatic populations (118.1 to 122.6 cm). This supports Martorell's (1988) suggestion that genetic potential for growth is similar for all groups examined in this way, apart from Asiatic populations. It should be stressed, however, that there are no data available on high socioeconomic status children who are Australian Aborigines or Pacific Islanders. In addition, it is not known whether genetically isolated groups can be included within the broader population typologies, for the same reason.

Evidence for secular increases in mean height of European, European-origin, and non-European populations is shown in Figs 3, 4 and 5. Mean heights of 8-year-old girls in the Netherlands, Belgium (Brussels), Britain (London) and Sweden (Fig. 3) all show a positive secular trend, this having reached a plateau for the Netherlands, Swedish and London girls, with mean heights for all three lying between the 50th and 75th centiles of the NCHS (National Center for Health Statistics, 1977) references. For the populations of European origin, all three show a positive secular trend, having reached a plateau between the 25th and 50th centiles of the NCHS reference for Canadian and United States girls respectively, and still continuing for Australian girls. A possible explanation for this difference between Northern Europe and North America is the likelihood that the North American growth patterns represent a mixture of North and South European growth patterns. South Europeans, including Spanish and Italian, are on average shorter in childhood than their North European counterparts (Eveleth & Tanner, 1990). Thus it appears that North Europeans may have greater genetic potential for growth than North Americans.

Fig. 2. Mean statures of 7-year-old males from industrialised countries, and from well-off populations in developing countries.

European populations: Sweden (Lindgren & Strandell, 1986); Netherlands (Roede & van Weiringen, 1985); Netherlands (Gerver, 1988); United Kingdom (Ulijaszek, 1987); United Kingdom (Rona & Chinn, 1984); Norway (Waaler, 1984); United States (National Center for Health Statistics, 1977); United States (Frisancho, 1990); Hungary (Eiben & Panto, 1986); Hungary (Eiben, 1982); Ireland (Hoey et al., 1987); Denmark (Andersen et al., 1982); France (Sempé et al. 1979); Europeans in Jamaica (Ashcroft & Lovell, 1964); Italy (Kramer, 1983); German Democratic Republic (Hesse, unpublished, in Eveleth & Tanner, 1990); German Federal Republic (Danker et al., 1981); Europeans in Colombia (Spurr et al., 1982); Poland (Kurniewicz-Witczakowa et al., 1983); Canada (Shepherd et al., 1984); Spain (Hernandez et al., 1985); Switzerland (Prader & Budliger, 1977); Yugoslavia (Prebeg, 1978); Croatia (Prebeg, unpublished, in Eveleth & Tanner, 1990). African and African-origin populations: Jamaica (Ashcroft & Lovell, 1964); African-Americans, United States (Frisancho, 1990); Turkana (Little et al., 1983); Nigeria (Janes, unpublished, in Eveleth & Tanner, 1990); African-British (Ulijaszek, 1987); African-Americans, United States (National Center for Health Statistics, unpublished, in Eveleth & Tanner, 1990); Haiti (King et al., 1963).

Indo-Mediterranean populations: Chandigarh, India (Prakash & Cameron, 1981); Turks in Sweden (Mjönes, 1987); Sikhs in Britain (unpublished data, Peters & Ulijaszek); Pakistanis in Britain (unpublished data, Peters & Ulijaszek); East African Asians in Britain (unpublished data, Peters & Ulijaszek); Indian Hindus in Britain (unpublished data, Peters & Ulijaszek).

Asiatic populations: Chinese in Jamaica (Ashcroft & Lovell, 1964); Southern Chinese (Chang et al., 1963); Japan (Tanner et al., 1982); Koreans in Japan (Kim, 1982); Chinese, urban (Zhang & Huang, 1988).

Fig. 3. Secular trend in height of females at eight years of age, European populations.

(A) Netherlands: van Weiringen, 1986. (B) Sweden: Ljung et al., 1974; Cernerud & Lindgren, 1991. (C) Britain (London): Cameron, 1979. (D) Belgium (Brussels): Meredith, 1976; Vercauteren, 1984.

Fig. 4. Secular trend in height of females at 8 years of age, populations of European origin.

(A) United States: Meredith, 1976; National Center for Health Statistics, 1977; Frisancho, 1990. (B) Australia: Meredith, 1976; Pyke, 1986. (C) Canada: Shephard et al., 1984.

Evidence for the secular trend in children of non-European origin children is sparse. Such as there is, is presented in Fig. 5. A positive secular trend is shown for: (1) an African origin population (African Americans); (2) an Indo-Mediterranean population (girls in Trivandrum, Kerala, South India); (3) an Asiatic population (Japanese in Japan); and (4) a Pacific Islands group (the Bundi, of Papua New Guinea). However, in no case is there clear evidence that the secular trend has reached a plateau. Therefore, no definitive statement can yet be made about the genetic potential for growth in non-European populations.

Fig. 5. Secular trend in height of females at 8 years of age, non-European populations.

African origin: Meredith, 1976; Frisancho, 1990. Asiatic: Meredith, 1976; Tanner et al., 1982. Indo-Mediterranean: Coleman et al., 1993. Pacific Islands: Zemel & Jenkins, 1990.

3. The validity of the concept of an international growth reference

There are problems associated with the use of any growth reference internationally. Currently, the NCHS growth references are promoted for international use, and the final part of this article is a critique of such use. In nutritional assessment, growth references may be used either cross-sectionally, or longitudinally. If used cross-sectionally, there are problems with the choice of cut-off. The NCHS centiles are smoothed, using cubic spline curves (National Center for Health Statistics, 1977). This modelling procedure concentrates on the goodness of fit and smoothness of individual centiles, but pays no attention to the spacing between centiles, making no allowance for skewness in the data (Cole, 1989). Thus the highest and lowest centiles (95th and 5th, respectively) are often incorrect, and the potential for misclassification if the 5th centile is used as a cutoff may be large. It is perhaps more appropriate to use Z scores as cut-offs for screening. This is another statistical construct, but one which allows a variety of cut-offs to be chosen, including ones which are well below the 5th centile, but which may be useful in nutritional classification of populations in which stunting is both widespread and severe. However, there are problems associated with the use of Z scores. In particular, there is the possibility of misclassification, even of sections of well-off populations. Table 2 shows the proportion of 7-year-old males from various populations in industrialized nations, and populations of high socio-economic status in developing countries which fall below -2 Z scores of the NCHS references for height for age.

If the NCHS references give a perfect fit for other populations which achieve, or are close to achieving their genetic potential for growth, then it would be expected that 2.3% of each population would fall below -2 Z scores of NCHS. For the Northern European populations of Netherlands, Sweden and Britain (London), the proportions falling below -2 Z scores are considerably lower than 2.3%. For the Canadian population the proportion falling below -2 Z scores is close to this value, while for the Southern European populations of Spain and Italy the proportions are 3.1 and 4.1% respectively. Although it is not clear if the secular trend has stopped in these two countries, the overall picture supports the view that North American growth patterns are a hybrid of Northern and Southern European growth patterns.

For high socio-economic status African, and African origin populations in industrialized nations, the proportions falling below -2 Z scores of NCHS are considerably lower for four of the groups examined, similar for one of them, and higher in two of them. For Indo-Mediterranean populations, the extent of misclassification of normals is lower than for European populations, while for the Asiatic groups examined, the proportion of children falling below -2 Z scores of NCHS is close to the expected 2.3% for four of them, slightly lower for one, and much higher in two of them. Therefore, if there is the possibility of misclassifying a proportion of any group which is believed to have similar genetic potential for growth, there is also the possibility of misclassification of less well off groups, for the same reason.

Longitudinal use of the NCHS growth references is also problematic. The most important problem is that the now traditional assumption that under good environmental conditions children track along a particular centile is not strictly correct. Although it is often assumed that pre-adolescent growth in length, and subsequently height, is a smooth and continuous process, a number of studies have shown that the growth patterns of individual children are more likely to be cyclical, with measured height oscillating about a centile line, rather than tracking it. Indeed, the normal growth curve, although representing population growth phenomena quite well, is not such a good representation of any individual growth pattern. Growth phenomena observed to contradict the tracking principle include: (1) catch-up and catch-down growth in the first two years of life (Smith et al., 1976); (2) mini growth spurts (Hermanussen, 1988); (3) seasonality of growth (Cole, 1993); and (4) biennial cyclicity of growth, shown to take place in children between the ages of 3 and 11 years (Butler, McKie & Ratcliffe, 1989).

Catch-up and catch-down growth usually take place in the first two years of life. During this time, children may cross the centile lines either upwards or downwards, rather than tracking along them. Catch-up growth can occur after a period of restricted growth in utero. The mini growth spurt was first described for healthy German children, whose knee height was measured on a daily basis (Hermanussen et al., 1988). These spurts occur with a cyclicity of between 30 and 55 days. Hermanussen et al. (1988) give examples of individuals showing this phenomenon, and it appears that there is a 3-4 fold variation in the rate of skeletal growth between the fastest and slowest time of growth. For a girl aged 6.6 years, the maximum growth rate of the lower leg was 3.6 cm/year, the minimum 1.1 cm/year. For a boy aged 8.9 years, the maximum and minimum rates were 4.2 and 1.1 cm/year, respectively. Another type of deviation from the tracking principle is the seasonality of growth reported for children in Japan (Togo & Togo, 1982), the Orkneys (Marshall, 1975) and Cambridge (Cole, 1993). The broad consensus is that height velocity is greatest in the spring, weight velocity is greatest in the fall (Cole, 1993).

Table 2. Proportion of males aged 7 years below -2 Z scores of National Center for Health Statistics (1977) references of height for age, for various populations in industrialised nations, or of high socio-economic status in developing countries


% < -2 Z scores


European and European origin



Roede & van Wieringen, 1985



Lindgren & Strandell, 1986



Waaler, 1984

UK (London)


Cameron, 1979



Shephard et al., 1984



Hernandez et al., 1985



Kramer, 1983

African and African origin

Jamaica, high socio-economic status


Ashcroft & Lovell, 1964

Nigeria, high socio-economic status


Janes, unpublished, in Eveleth & Tanner, 1990

African British


Ulijaszek, 1987

African American (NHANES I & II)


Frisancho, 1990



King et al., 1963

African American (NCHS)


unpublished, in Eveleth & Tanner, 1990



Little et al., 1983


India, high socio-economic status (Chandigarh)


Prakash & Cameron, 1981

Turkish, in Sweden


Mjönes, 1987

Pakistani, Britain


Peters & Ulijaszek, unpublished data

East African Asians, Britain


Peters & Ulijaszek, unpublished data

Sikhs, Britain


Peters & Ulijaszek, unpublished data

Indian Hindus, Britain


Peters & Ulijaszek, unpublished data


Koreans in Japan


Kim, 1982



Kikuta & Takaishi, 1987

Japanese, Kyoto


Tanner et al., 1982

Japan, national sample


Tanner et al., 1982

China, urban


Zhang & Huang, 1988

Chinese in Jamaica


Ashcroft & Lovell, 1964

Southern Chinese, Hong Kong


Chang et al., 1963

Another phenomenon which is at odds with the tracking principle is that of mid-childhood cyclicity of growth. A study in Edinburgh of mid-childhood growth of 80 boys and 55 girls between the ages of 3 and 11 years showed a cyclicity of statural growth with a periodicity of 2.2 years in males, 2.1 years in females (Butler et al., 1989). Although cyclicity of growth was observed in all the children, periodicity and magnitude of peak growth rates varied.

If, in any individual, there are several cyclicities of growth operating, then it is difficult to interpret growth patterns on the basis of a small number of measurements across time. There is a need to identify deviation from the growth references as early as possible. However, under some circumstances growth cyclicity may indicate such deviation in an individual for reasons other than growth faltering.

Any use of growth references internationally should acknowledge that they can act, at best, as imperfect yardsticks, since human populations may show similar growth characteristics, but are unlikely ever to become so homogeneous that they show the same genetic potential for growth. Since the NCHS growth references do not represent the greatest possible human potential for growth, they may not be any more appropriate for international use than growth references developed in other countries. The NCHS references are poorly modelled, and there is need for the data to be reanalysed in a more sophisticated manner if they are to be of use internationally. In cross-sectional studies, the use of Z scores as cut-offs for screening is to be encouraged, since the lower centiles of NCHS are inaccurate. In longitudinal use, workers should be aware that the tracking principle is flouted even by healthy children in Western societies. Normal growth can better be described as oscillation about a centile, rather that tracking along it. Thus short-term deviation from a centile cannot be taken as evidence for pathology of any kind.

An international growth reference could be used for European and European origin populations, as well as African, African origin and Indo-Mediterranean populations. Current evidence suggests that they may not apply to Asiatic populations, but in the absence of definitive evidence of a cessation of the secular trend in any well-off Asiatic population, this assumption must remain tentative. It is not clear whether genetically isolated populations in various parts of the world including Africa, India, Latin America and Asia are likely to show the same potential for growth when placed in favourable environments. In addition, almost nothing is known about the genetic potential for growth of Aboriginal populations in Australia, or in Pacific Islands populations.

Contents - Previous - Next