Contents - Previous - Next

Levels and patterns of intrauterine growth retardation in developing countries


M de Onis1, M Blössner1 and J Villar2

Correspondence: Dr Mercedes de Onis

1Nutrition Unit, World Health Organization, 20 Avenue Appia, 1211 Geneva 27, Switzerland;2Special Programme of Research, Development and Research Training in Human Reproduction, World Health Organization, 20 Avenue Appia, 1211 Geneva 27, Switzerland

The aim of this paper is to quantify the magnitude and describe the geographical distribution of intrauterine growth retardation (IUGR) in developing countries. We estimate that at least 13.7 million infants are born every year at term with low birth weight (LBW), representing 11% of all newborns in developing countries. This rate is approximately 6 times higher than in developed countries. LBW, defined as < 2500 g, affects 16.4% of all newborns, or about 20.5 million infants each year. IUGR, defined as birth weight below the 10th percentile of the birth-weight-for-gestational-age reference curve, represents 23.8%, or approximately 30 million newborns per year. Overall, nearly 75% of all affected newborns are born in Asia - mainly in South-central Asia - 20% in Africa, and about 5% in Latin America. Although some of these are healthy, small infants who merely represent the lower tail of a fetal growth distribution, in most developing countries a large proportion of newborns suffer from some degree of intrauterine growth retardation. These data demonstrate that many developing countries currently exceed the internationally recommended IUGR (> 20%) and LBW (> 15%) cut-off levels for triggering public health action, and that population-wide interventions aimed at preventing fetal growth retardation are urgently required.


Intrauterine growth retardation (IUGR) constitutes a major clinical and public health problem in developing countries. Various criteria have been used to classify an infant as having experienced normal, subnormal, or supranormal growth in utero. Most recently a WHO Expert Committee (WHO, 1995a; de Onis and Habicht, 1996) recommended the 10th percentile of a birth-weight-for-gestational-age, sex-specific, single/twins risk curve (Williams et al, 1982) for the classification of small-for-gestational-age infants (SGA). Strictly speaking, SGA and IUGR are not synonymous (Altman and Hytten, 1989). Some SGA infants (e.g. those born to short mothers) may merely represent the lower tail of the 'normal' fetal growth distribution, while other infants who have been exposed to growth-inhibiting factors may actually meet the criteria for appropriate-for-gestational-age (AGA)(e.g. those born to tall, well nourished cigarette-smokers). In individual cases, however, it is usually very difficult to determine whether or not the observed reduced birth weight is the result of true in utero growth restriction, and classification is therefore based on the established cut-off for SGA. In fact, the higher the SGA rate, the greater the likelihood that SGA is a result of IUGR (WHO, 1995a). In this paper, for the purpose of being consistent with the terminology used in the Workshop, SGA infants will be referred to as IUGR.

Historically, because valid assessment of gestational age is often unavailable in developing countries, the incidence of low birth weight (LBW) has been often used as a proxy to quantify the magnitude of IUGR in these settings. This approach, however, underestimates considerably the overall magnitude of the IUGR problem as it does not take into account those infants whose weight at birth falls below the 10th percentile but who weigh more than 2500 g; many of these infants are likely to also have IUGR. Within these constraints, we make in this paper, for the first time, an attempt to quantify the magnitude and describe the geographical distribution of intrauterine growth retardation in developing countries. We use as the basis for the analysis, the incidence of infants born at term ( 37 weeks of gestation) with low birth weight (< 2500 g), referred to as IUGR-LBW in this paper.


Source of LBW incidence rates

Data on low birth weight (defined as weight at birth < 2500g) were obtained from an updated version (September 1996) of the WHO database on Low Birth Weight compiled by the Maternal Health and Safe Motherhood Programme (WHO, 1992). This database contains information published from 1980 onward. Articles and reports are identified through a search of library databases (Medline, Healthplan, Popline and Lilacs) and by tracing references found in them. In addition, data from country studies (some unpublished) are also included. Reports containing low birth weight data are examined to ensure basic standards in data collection and reporting, before including the information into the database. For each study, the database includes information related to the geographical region, year or time-period in which the data were collected, the nature or source of the data, sample size (only those with at least 100 births are included), and information regarding gestational age at birth if available (WHO, 1992).

The present analysis has been restricted to studies identified in the database as nationally representative and carried out between 1985 and 1995. The selected country data points are based on nine different sources of information: 52 (38%) are derived from monitoring systems (i.e., Government reports for monitoring "Health for All" indicators); 23 (17%) from registration data; 17 (13%) from the Western Pacific region data bank; 13 (10%) are estimates provided by Unicef field offices; 12 (9%) from hospital data; 6 (4%) from Government reports; 4 (3%) from community-based studies; 3 (2%) are estimates (with no specific methodology mentioned); and 6 (4%) are of unknown source (i.e., data extracted from a secondary source such as Unicef reports or others).

Estimates of IUGR-LBW rates

Estimates of IUGR-LBW rates for developing countries were calculated by applying to the total incidence of LBW the linear regression equation proposed by Villar et al. (Villar et al, 1994) in which the dependent variable was IUGR-LBW and the independent variable the total rate of LBW (ß = 0.8528; SE = 0.0282; P = 0.0001; r = 0.96). This regression equation, for which we calculated the 95% confidence bands (Kleinbaum et al, 1988) (Figure 1), was derived using studies from 60 populations (WHO, 1992; Puffer and Serrano, 1987) from developing countries where gestational ages and birth weights were recorded. All 60 studies were conducted from 1965 onward, and were based on prospective research projects and surveys whose major focus was birth weight with valid gestational age assessments (Villar et al, 1994; Villar and Belizán, 1982). For developed countries, the linear relationship between the proportion of IUGR-LBW and the total incidence of LBW was not a good predictor (ß = 0.2951; SE = 0.1759; P = 0.11; r = 0.4092; n = 16) and thus, for the purpose of this analysis, only the regression model for developing countries was used.

It is important to note that the use of the regression equation only provides an estimate of the incidence of IUGR infants that are low birth weight at term (area labeled as A in Figure 2); i.e., it does not include IUGR preterm infants, or IUGR infants with birth weights above 2500 g (areas labeled as B and C, respectively, in Figure 2).

Validation of the IUGR-LBW estimation

We further tested the agreement between the IUGR-LBW estimation from the regression model and the IUGR LBW rate observed in 17 data sets included in the WHO Collaborative Study on Maternal Anthropometry and Pregnancy Outcomes (WHO, 1995b), in which information was available on both birth weight and gestational age estimated from the last menstrual period. None of these data sets had been included in the earlier analysis used to develop the linear regression equation for estimating IUGR-LBW from overall LBW rates in developing countries (Villar et al, 1994).

The agreement between the two estimations is summarized by the mean of the differences, and the precision of the estimation by the 95% confidence interval of this mean. A graphic display of the agreement between the two values is presented by plotting the difference between estimations against their mean (Martin Bland and Altman, 1986). For the purpose of these analyses, it is assumed that the differences follow an approximately normal distribution and that the differences do not vary systematically over the range of the IUGR-LBW rates.

Figure 1. Relationship between total incidence of low birth weight and incidence of low birth weight/IUGR in developing countries (adapted from Villar et al, 1994).

Figure 2. Birth weight percentiles and perinatal mortality rates (per 1000) for single male births (adapted from WHO, 1995a).

Intrauterine Growth retardation (IUGR) = A+ B +C

A = < 2500 g, 37 weeks
B = < 10th centile, < 37 weeks
C = <10th centile, 2500 g

Regional and global estimates of IUGR-LBW

Regional incidences of IUGR-LBW and LBW were estimated for each geographical area by weighting the available national incidences according to the total number of estimated live births in 1995 in each country. The numbers of IUGR-LBW and LBW newborns in each area were obtained by applying incidence estimates to the total number of live births. Global incidences in developing countries were calculated by adding the estimated number of IUGR-LBW newborns in each area and using as the denominator the 1995 total live births of all developing countries. Estimates for IUGR-LBW and LBW were obtained only for those regions where the proportion of live births covered by studies identified in the database as nationally representative was > 80%.

For this analysis, countries have been grouped according to the United Nations classification (UN 1995), dividing the world into six major areas, which are further subdivided into 20 regions. Total live births estimated for 1995 for the countries concerned were obtained from the UN Population Prospects (UN, 1995).

Contents - Previous - Next