Contents - Previous - Next

Linear growth retardation in relation to the three phases of growth

1. Introduction
2. The three phases of linear growth
3. Measuring and monitoring linear growth in early life
4. Growth faltering in linear growth
5. Discussion

J. Karlberg 1, F. Jalil 2, B. Lam 1, L. Low 1 and C.Y. Yeung 1

1 Department of Paediatrics, Queen Mary Hospital, University of Hong Kong, Hong Kong and
2 Department of Social and Preventive Paediatrics, King Edward Medical College, Lahore, Pakistan

1. Introduction

It is well documented that infants living in developing countries have mean birth lengths close to those reported from Western Europe and North America (WHO Working Group, 1986; Martorell & Habicht, 1986). However, between 4-6 months and 18 months of age the mean lengths' curves diverge, so that by 24 months the difference is considerable (WHO Working Group, 1986; Martorell & Habicht, 1986; Manwani & Agarwal, 1973; Mata, 1978; Karlberg, Jalil & Lindblad, 1988; Black et al., 1982; Jalil et al., 1989; Degan et al., 1983; Waterlow, Ashworth & Griffiths, 1980; Zumrawi, Dimond & Waterlow, 1987; Costello, 1989; Dewey et al., 1992). During the second and third year of life, the mean growth curve of infants in developing countries again almost parallels the Western standard curve. The reason for this faltering process is not exactly known, as it can be present in infants of normal or close to normal nutritional status (Degan et al., 1983). Plausible causative mechanisms, involving such factors as feeding patterns, general under-nutrition, specific nutrient deficits, disease, serum and intracellular growth factors, and psycho-social conditions, are discussed in other chapters of this supplement.

In this chapter we discuss the three phases of linear growth-infancy, childhood and puberty in relation to growth faltering in early life. The key process is the onset of the second phase of growth, the childhood phase, that normally takes place at 6 to 12 months of age. Illustrations are taken from two longitudinal studies carried out in Lahore, Pakistan, and Hong Kong.

2. The three phases of linear growth

2.1. Hormonal regulation of linear growth

The growth process is under the control of the endocrine system. However, not only hormones are involved; hormone-binding proteins, growth factors and their binding proteins, as well as the stage of maturity and quantity of the hormone and growth factor receptors on the target cells may play a critical role too (Cianfarani & Holly, 1989; Waters et al., 1990). Furthermore, the secretion of hormones, such as growth hormone (GH), follows a pulsatile pattern with higher peaks during the nights as well as during puberty (Stanhope, Pringle & Brook, 1988). Such growth regulatory mechanisms interact and change in character over the ages (Masse, de Zegter & Vanderschueren-Lodeweyckx, 1992). A further confusing element is that some growth can take place without involvement of central steering mechanisms, as illustrated by an animal study showing that catch-up growth was regulated locally, at the tissue level, and not necessarily by the influence of circulating serum growth factors (Baron et al., 1993). For these reasons, the measurement of a hormone or growth factor in a single serum, urine or tissue sample will shed light on only a small part of this very complicated puzzle.

It is generally agreed that we have at least three distinct endocrine phases of linear growth, as indicated by the solid upper curve in Fig. 1. The pattern of postnatal growth is well documented; a high growth rate is observed from fetal life, with a rapid deceleration up to about 3 years of age. This is followed by a period with lower, slowly decelerating velocity up to puberty. Puberty starts with an increased rate of growth, and after the age of peak height velocity has been reached a deceleration is noted until growth ceases.

Fig. 1. The ICP growth model for height for boys (Karlberg, 1989a). The mean functions are plotted for each of the three components as well as the combined growth. The average at onset of the childhood and puberty components were used. The interrupted curve at 3 years reflects the change in measuring position from lying to standing up (could have been given at 1 or 2 years, as well). The total gains given by each component are given in Table 1. The key hormones involved in the regulation of growth are also given.

How fetal linear growth is regulated is not precisely defined and no key circulating hormone has so far been identified (Gluckman, 1989; Milner & Hill, 1987; Hill, 1989). Uterus size, nutritional support and oxygen level in conjunction with insulin-like growth factors and insulin are believed to be involved in regulating fetal growth (Gluckman, 1989; Milner & Hill, 1987; Hill, 1989).

During fetal life the serum GH level is high, and GH receptors have also been detected (Hill et al., 1988; Werther, Haynes & Waters, 1991). Fetal linear growth, however, is known to be almost independent of GH (Gluckman, 1989; Milner & Hill, 1987). A lack of growth response to GH during fetal life may be due to immature GH-specific receptors in the growth plate, as noted in the rabbit (Barnard et al., 1988). GH-deficient children are on average 1-2 cm, or 2-4% shorter than normal infants at birth (Karlberg & Albertsson-Wikland, 1988; Albertsson-Wikland, Niklasson & Karlberg, 1990; Tse, Hindmarsh & Brook, 1989; Gluckman et al., 1992). Whether this minor deviation is a secondary effect due to the lack of the influential metabolic action of GH or to the lack of a direct effect of GH on the cartilage, is still a matter of debate.

It is more generally accepted that GH is responsible for growth during childhood provided that thyroid hormone secretion is normal. The exact age at which GH begins to control linear growth in humans is still uncertain (vice infra). The majority of children with isolated GH deficiency grow more or less normally during the first 6 months of life, but not thereafter (Karlberg & Albertsson-Wikland, 1988; Albertsson-Wikland, Niklasson & Karlberg, 1990; Tse, Hindmarsh & Brook, 1989; Gluckman et al., 1992).

Growth during adolescence is related both to GH and sex steroids - testosterone in males and oestrogens in females (Copeland, Paunier & Sizonenko; 1977; Keenan et al., 1993). Both GH and sex hormones are needed for normal pubertal growth, although the presence of only one of them is associated with some growth during this period. It is not clear whether GH and sex steroids interact or act independently of each other (Pescovitz, 1990).

It is thus reasonable to conclude that linear growth from birth to maturity is regulated by at least three different growth-promoting systems. Two simultaneously active, superimposed, systems are known to be involved in the adolescent growth spurt. Similarly, a postnatal continuation of the nutritionally driven fetal growth in conjunction with the GH-dependent phase of childhood growth characterizes the growth in the first year of life (Tse, Hindmarsh & Brook, 1989).

2.2. The infancy-childhood-puberty (ICP) growth model

The ICP growth model breaks down linear growth mathematically into three additive and partly superimposed components - infancy, childhood and puberty (Karlberg, 1987, 1989a,b; Karlberg et al., 1987a,b). Fig. 1 shows the shape, size and timing of each of the three components, together with their additive effects in boys.

The ICP model represents linear growth during the first three years of life by a combination of a sharply decelerating infancy component and a slowly decelerating childhood component, the latter acting from the second half of the first postnatal year. From about 3 years of age to maturity, linear growth is represented by the sum of the infancy and childhood components and a sigmoid shaped puberty component operating throughout adolescence (Fig. 1).

2.3. Curve fitting procedure

The ICP-model was sequentially fitted to the individual growth curves from birth to adulthood for the children included in the Swedish longitudinal study (Karlberg, 1987; 1989a). In view of the equilibrium of the growth process it seemed rational to represent the period of slowly decelerating growth from 3 years of age to the onset of puberty by a single component, which was called childhood phase. A simple quadratic function was found to fit growth during this period very well. After subtraction of the extrapolated values of the childhood component from the observed values during the periods before and after this phase, two additional components were extracted and modelled. As a result of this modelling process, three separate components could be isolated (Karlberg, 1987; 1989a).

1. Infancy: This constantly decelerating component begins before birth and tails off by 3-4 years of age. It can be represented by an exponential function:


2. Childhood: This phase begins during the first year of life at age tC, decelerates slowly and advances until maturity at age tE A simple second degree polynomial provides a sufficient model for this component:


3. Puberty: This phase depicts the additional growth induced by puberty and accelerates up to age at peak velocity (tV). It then decelerates till growth finishes (tE). It can be modelled using a logistic function:


In these functions, Y denotes attained body size for the relevant component at time t in years from birth and the a's, b's and c's are constants. Age at end of growth (tE) was specified as the middle of the first one-year interval after age at peak velocity, when the overall gain was less than that in the childhood component alone.

2.4. Biological interpretation of the modelled components

Empirical observations have shown that the three components of the ICP model can be observed in isolation, and that they are additive (Karlberg et al., 1987a,b; Karlberg 1989a,b, 1990). Each component of this model is therefore assumed to represent a separate biological phase of the growth process (Karlberg, 1989a). The infancy component, tentatively starting in mid-gestation and continuing, with a rapidly decelerating influence, up to 3-4 years of age, is claimed to represent the postnatal continuation of fetal growth. It is assumed that the childhood component, slowly decelerating during childhood and adolescence, corresponds basically to the effect of GH. The sigmoid shaped puberty component, the size of which is found to be independent of its timing, most likely describes the part of adolescent linear growth stimulated by sex steroids.

In the following presentation we will focus primarily on the age at onset of the childhood phase of growth. In normal infants, the onset occurs between 6 and 12 months of age and is typically abrupt (Karlberg, 1987; 1989b; 1990; Karlberg et al., 1987a), as illustrated, for instance, by the growth of a Pakistani girl plotted out in Figs 2a-f. What mediates the onset of the childhood phase of growth has not been elucidated, because information about the serum level of growth factors, hormones, the serum globulins involvement and the hormone and growth factor receptor activity and their interaction is still lacking during this critical period. However, the most plausible explanation is that it represents the age at which GH begins to influence normal human linear growth significantly. There are two major observations supporting this idea. Firstly, most (84%) of the increased gain in total body length is localised in the lower segment of the body (Karlberg, 1990). It is known that the growth of the long bones, as represented in the legs, is more sensitive to GH than other bone structures, such as the short bones in the vertebra (Frasier, 1983). Secondly, in children with isolated GH deficiency who receive no hormonal therapy, this tempo change of early linear growth is completely absent (Karlberg & Albertson-Wikland, 1988; Tse, Hindmarsh & Brook, 1989). Some other empirical observations point in the same direction (Karlberg, 1990).

Fig. 2. Length and length velocity for a normal Pakistani belonging to an upper middle class family; the monthly recorded values have been used in (a), every second month values in (b), etc. The growth rate has been expressed in mm/month in all graphs.

Figure (2a)

Figure (2b)

Figure (2c)

Figure (2d)

Figure (2e)

Figure (2f)

The final adult height of an individual is the additive result of the three underlying components (Fig. 1). This graph gives the mean path of each component, but there is an individual variation around these mean values, and the SD values for each component are 2-4 cm. It is not only the magnitude of each component that is important for the final height, but also the duration of the childhood phase; a late onset reduces height, and a late end point increases height (Karlberg, 1989a).

Table 1 includes the mean total contribution of each component to height, sitting height and leg length for the two sexes of the Swedish study. The infancy component is contributing more to sitting height than the childhood component, whereas the opposite is true for leg length. Different body tissues are thus unequally influenced by each of the three phases of growth. There is also a sex difference in the influence of the three components; the sex difference in the infancy and puberty components is related to their magnitude, whereas the childhood sex difference is basically due to a disparity in duration (2 years longer in boys than in girls; Karlberg, 1989a).

3. Measuring and monitoring linear growth in early life

3.1. Growth rate estimates and interval lengths

Growth rates usually covary with factors that can affect health, such as socio-economic status, season, disease, intervention, feeding patterns, nutritional status, serum nutritional factors or serum growth factors. Growth rate measurements, however, also include measuring errors. Therefore, we have to select intervals for growth rate estimations that are both short enough to be related to the indicator under observation and long enough to be sure that measuring errors are of minor importance.

Early human linear growth is commonly measured in terms of total body length or lower leg length, the latter being measured by the modified knemometer for infants (Michaelson et al., 1991). Both measures include measuring errors, although knemometer data include much less technical error than measures of supine length or height (Michaelson et al., 1991; Valk et al., 1983; Hermanussen et al., 1988; Wales & Milner, 1987; Wit et al., 1987; Dean, Schentag & Winter, 1990). Both measures include soft tissues and not only the skeleton; changes in the thickness of the soft tissues over time will therefore influence growth rate calculations (Hermanussen et al., 1988). The minimum time interval for calculations of lower leg length growth rates has been set at one month (Michaelson et al., 1991; Wit et al., 1987). The reason for selecting this rather long interval is to avoid an increased variation in the growth rate measure; true differences between groups will be less easy to detect when the variation is large.

Table 1. Mean total gain in height, sitting height and leg length for each component of the ICP growth model. The effect on sitting height and height due to the change in measuring position at 3 years of age has been corrected in proportion to attained size at 3 years of age for the infancy and childhood components



Infancy component (cm)

Childhood component (cm)

Puberty component (cm)

Total (cm)

Sitting height











Leg length











Total height











As far as we know, the minimum interval between supine length measurements has not yet been clearly defined, although some have recommended monthly intervals (WHO Working Group, 1986). Using very short intervals may produce growth rate patterns that are difficult to interpret. For instance, daily or weekly length measurements in early life seem to suggest a pulsatile or saltatory growth pattern (Lampl, Veldhuis & Johnson, 1992). Whether this pattern is true or just reflects measuring errors or soft tissue changes cannot be determined when using total body length. An experimental study of the daily increase in the growth plate of the rabbit did not show any saltatory growth pattern (Oerter et al., 1993).

Supine length is commonly used in linear growth studies in early life. In our research work we are interested in defining the age at onset of the childhood phase of growth, which can be determined mathematically or visually using the ICP-based growth charts (Karlberg, 1989b). The influence of the measurement interval on the determination of this stage of maturity of human growth is illustrated by the measurements of the supine length of a healthy Pakistani girl belonging to an upper middle class family (Figs 2a-f). The monthly values from birth to 36 months are plotted in Fig. 2a. Four monthly values were missing at 15 to 34 months of age, and linear interpolation was used to estimate them. The onset of the childhood phase of growth for this girl can be determined by observing either the length or the length velocity curve: the growth curve has an exponential shape till about 10 months of age, then it becomes more linear.

Figs 2b-f show the growth of the same infant at increasingly longer intervals of observation of 2,3, 4,5 and 6 months. In all these graphs the growth rate has been expressed uniformly, i.e. in units of mm change per month. Two conclusions can be reached: (i) that the age at onset of the childhood phase - at around 10 months in this girl - can be determined only when the interval between measurements is less than 3-4 months, and (ii) that the variation in growth rates over time is highly dependent on the length of the interval.

Fig. 3. Length velocity for the same girl as described in Fig. 2 using the six different age intervals. The growth rate values have been computed as the change over the interval without dividing for the length of the interval.

In Fig. 3, the six growth rate curves given in Figs 2a-f are all shown together. The growth rates are presented here as the observed change over the specific time interval. For knemometer measurements one has accepted that the technical error should be less than 10% of the measured growth rate (Michaelson et al., 1991; Wit et al., 1987). Applying this approach to supine length, assuming a small technical error of 0.2 cm, the growth rate over a certain interval should be over 2 cm; monthly intervals can be used till 5 months of life and two months' intervals at higher ages in infancy, based on the curves given in Fig. 3. For defining the onset of the childhood phase of growth, 3 months' intervals are currently being recommended (Karlberg et al., 1987a; Karlberg, 1989b); that should, in most situations, rule out any major influence of measuring errors.

3.2. Monitoring of linear growth in early life

During the period preceding the age of onset of the childhood component, 83% of normal infants have a non-linear decelerating growth pattern, free from seasonal influences (Karlberg et al., 1987a). For the majority of the infants this phase of growth seems to be very stable over time, although the magnitude of the infancy component varies from individual to individual.

At the age of onset of the childhood component, 76% of normal infants show an abrupt increase in growth rate (Karlberg et al., 1987a). A smooth pattern is more common in infants with an early onset of the childhood component than in those with a late onset; this is due to the decelerating influence of the infancy component. A smooth pattern may also reflect an onset halfway between two observations. Children with a small infancy component have an early onset of the childhood component, which seemingly compensates for the initial low gain. Girls have an earlier onset than boys, which coincides with the lower initial velocity of their infancy component. The onset of the childhood component is thus related to the magnitude of the infancy component. Onset of the childhood component after 12 months of age did not occur in a Swedish study of healthy children (Table 2) (Karlberg et al., 1987a). Later onset is indicative of disturbances in the growth process (Table 2) (Karlberg, Jalil & Lindblad, 1988; Albertsson-Wikland, Niklasson & Karlberg, 1990; Karlberg et al., 1988; 1991; Karlberg & Wit, 1991; Karlberg & Albertsson-Wikland, 1988; Karlberg et al., 1991; Karlberg, Hägglund & Strömquist, 1991; Karlberg, Kjellmer & Kristiansson, 1991).

About two-thirds of normal children shift centiles for linear growth during the first 18 months of life after which the influence of parental stature becomes more important (Smith, 1977). This phenomenon can be explained by a gradual shift in influence from the infancy to the childhood component during this period.

In the second year of life, growth is a result of both the infancy and childhood components. The majority (75%) of normal infants displays a fairly constant growth rate with a total gain close to average during this period (Karlberg et al., 1987a), while others are more variable in their growth pattern. This variation is found to be seasonal, and to be greater for those experiencing a late onset of the childhood component (Karlberg et al., 1987a). This indicates an initial irregular effect on this component, but the fluctuations in growth rate have no clear effect on the total observed gain during the second year of life. Seasonal variations and transient irregular effects of the childhood component may thus cause fluctuations, which will be reduced by calculating the growth rate over longer intervals.

Contents - Previous - Next