Discussion

The main reason for trying to quantify the relative risks of mortality and morbidity by birth weight for this review was to provide a method for estimating the potential public health impact of interventions to reduce IUGR. If the potential impact can be ascertained, it can then be compared with estimated impacts of other potential interventions to reduce mortality and morbidity in infants and young children. This helps health planners set priorities for action and identifies areas where additional knowledge is required. Such procedures promote effective decision making. Most IUGR infants are born in developing countries. These mostly lack birth-weight-specific mortality data, the main reasons being:

i) birth weight measurements are limited because of home deliveries and cultural sensitivities (eg evil eye),

ii) cost and logistics, since large numbers of infants have to be followed from birth for many months.

In contrast in developed countries, a computerized system of linked birth and death records is often feasible. Socioeconomic and demographic data may also be available in linked records, and their effects can be controlled as in Puerto Rico (Becerra et al, 1993).

The first approach in this review to obtain relative risks of mortality for IUGR infants, was to locate birthweight-specific data for countries with a high prevalence of LBW and to assume these reflect IUGR births. Their relative risks of neonatal and postneonatal death within each 500 g birthweight-interval (Tables 2 and 4) were less consistent than the levels of risk seen for countries where most LBW births can be presumed preterm (Tables 1 and 3). There are several possible reasons for this. First, methodological weaknesses, particularly in sample size and representativeness, could have masked the true level of risk. Second, the differing aetiologies of IUGR among countries may be associated with different postnatal risks. Evidence to support this possibility is seen in the apparently differential mortality risks of wasted vs stunted LBW infants. If IUGR infants are more heterogeneous than preterm infants, a less consistent level of risk among populations might be expected. Third, patterns of disease may vary among countries. For example, if the peak age for diarrhoea is 1-5 months in country A, but 60-11 months in country B, then the relative risk of postneonatal diarrhoeal death among IUGR infants is likely to be higher in country A (as younger, lighter infants are more likely to die than older, heavier infants). Different patterns of disease having different case-fatalities could also give rise to varied relative risks. The race differentials seen in the USA serve as an illustration. Black LBW infants have lower birthweight-specific neonatal mortality rates than white LBW infants because they are less likely to have congenital anomalies and respiratory distress syndrome (Sappenfield et al, 1987).

The second approach to ascertain birth-weight-specific mortalities for IUGR infants by identifying studies where gestational ages were specified, revealed levels of risk similar to those obtained by the first approach. All the studies were from the USA. One might have anticipated different levels of risk between developed and developing countries as congenital anomalies are frequently associated with IUGR in the former (McCormick, 1985), in contrast to the situation in developing countries (Rao and Inbaraj, 1978). Since the two approaches gave similar estimates of risk, the data sets were combined and median risk ratios for IUGR infants derived (Table 10). These may be used to estimate the impact on mortality of interventions to reduce IUGR.

Although eight studies from developing countries were located regarding LBW and morbidity, in only one were preterm infants excluded and confounding variables controlled (Lira et al, 1996). In this study, an increased risk of diarrhoea, but not cough, was found in LBW infants compared to infants weighing 3000-3499 g. The sample size had insufficient power to detect a difference in pneumonia morbidity. The other studies located are consistent with there being increased risks of diarrhoea and lower respiratory tract infections in IUGR infants. An increased risk of mortality and morbidity in LBW infants compared to ABW infants is biologically plausible because their immune function is severely compromised (Chandra, 1981). This impairment is particularly marked in IUGR infants.

Few studies have examined the differential effects of birth dimensions on mortality and morbidity. Indications are that wasted LBW newborns experience more morbidity than stunted newborns in the first weeks of life, whereas stunted LBW infants experience higher mortality. Caution is required in interpreting the mortality data, however, as infants with congenital anomalies are often included. No data were located for the postneonatal period. It is quite plausible for wasted infants to have higher risks than stunted infants in neonatal morbidity, but lower risks later because wasted infants tend to correct their body fat deficits etc whereas stunted infants tend not to 'catch-up' and have continued impaired immunocompetence. Stunted infants might therefore be expected to have higher rates of infectious disease than wasted infants.

Because of the possible differential effects of birth dimensions on mortality and morbidity, infants should be characterised by birth length z-score and ponderal index in future studies involving LBW or IUGR infants. Further studies, using rigorous methodologies, are required, particularly of postneonatal mortality and morbidity. Such studies will require large numbers of infants to be recruited, especially for relatively infrequent outcomes such as mortality or pneumonia morbidity. Intensive follow-up is required for common morbidities such as diarrhoea. Studies are required in different settings to determine whether there is a consistent level of risk across populations among stunted infants, and whether this differs for infants who are wasted.

Table 10. Estimated relative risks of neonatal and postneonatal mortality for IUGR infants

^a median of 10 studies (Tables 2 and 5)
^b median of 5 studies (Tables 4 and 6)