Fetal growth: towards an international standard

The assessment of fetal growth is an essential component of good antenatal care, and ultrasound biometry is central to clinical management which seeks to detect and prevent problems during pregnancy. Ultrasound requires adequate training, method and equipment, as well as appropriate standards. Fetal biometry for growth assessment has two principal functions. First, and most commonly, it reassures parents and clinicians when the measurements are normal. Inadequate standards can create undue anxiety and lead to unnecessary investigation, as in the case of constitutional smallness. Second, fetal biometry serves to identify abnormal size (i.e. when the measurement does not correspond to the expected value for that gestation) and abnormal growth (i.e. when serial measurements do not follow the expected curve). The measurement may be followed by further investigation (e.g. by Doppler flow), continued close surveillance and, ultimately, carefully timed intervention to deliver a baby in the best possible condition. In this instance, inaccurate standards may give false reassurance and expose the fetus to continued risk in-utero. A special challenge arises when standards derived from one country or continent are used in a different one, an issue which Merialdi and colleagues address in this issue of the Journal1. Looking at their Peruvian population, they observe that fetuses appeared to grow more slowly than commonly used reference charts from North American and European populations predict, despite the pregnancies being otherwise uncomplicated. This study is further evidence for the need for standards for fetal growth that are applicable in different populations. Ethnicity affects fetal growth and birth weight2-5, and there has long been evidence also of varying growth patterns between ethnic groups in cross-sectional6 as well as longitudinal studies7. Yet the temptation to derive standards on the basis of geography or ethnicity alone ought to be resisted2, as birth weight8 as well as fetal growth9-11 also vary with other constitutional factors. In a second paper on growth curves and ethnicity in this issue, Drooger and colleagues in Rotterdam demonstrate variation in fetal growth in various ethnic subgroups within their Dutch maternity population12. They confirm that other factors such as maternal height, weight, parity and fetal gender account for some of the variation. Such findings again highlight the need for standards that are applicable in all sections of the maternity population. However, it is uncertain whether prospective ultrasound studies would ever be able to allow for the many possible combinations of constitutional factors that can affect normal fetal growth. Instead, birth weight databases can be used to derive coefficients to calculate an optimal birthweight for each pregnancy, which can then be mathematically combined with ‘proportional’ equations to delineate individual curves for fetal weight gain. The approach is based on the rationale described in the next three sections. Variation within populations can be as large as the variation between populations. The weight standard should be individual rather than population-based, to allow for all known constitutional factors affecting fetal growth. In addition, pathological factors need to be excluded so as to define the optimal growth for that baby (i.e. its growth potential) against which the in-utero course can be monitored13. For example, a mother who smokes during pregnancy is likely to have a baby with slower growth and lower birth weight. The aim, however, is not to produce a curve which predicts this, but one which helps to detect if growth is being affected; thus the standard needs to be optimized as well as individualized. Many clinicians still plot ultrasound measurements such as biparietal diameter and/or head circumference, abdominal circumference and femur length separately on standard population charts in the third trimester, instead of just the estimated fetal weight. However, the clinical usefulness of this approach is uncertain14. Biparietal diameter or head circumference plotting in the third trimester appears of limited use in a routinely scan-dated population, and is often a cause of unnecessary anxiety as the normal limits are relatively narrow and false-positives are common. Microcephaly is usually recognized in the first or second trimester of pregnancy. According to the West Midlands Congenital Anomalies Register, which has a very high ascertainment rate, only about one case of microcephaly in 60 000 pregnancies per year is not detected until the third trimester. Similarly, the presence of short femur lengths is in most instances detected at the time of the mid-trimester scan. Traditionally, separate plots of individual ultrasound parameters have allowed assessment of the symmetry of growth failure. Asymmetrical growth restriction is conventionally considered to be due to placental failure with ‘head sparing’, while symmetrical growth restriction is associated with early-onset growth failure. However, these distinctions have wide overlaps. Symmetrical smallness can be due to misdating, constitutional smallness or early-onset growth restriction, and separate plotting of fetal head, abdomen and femur measurements will not distinguish between them. Furthermore, it is doubtful whether the distinction between symmetrical and asymmetrical growth failure adds any clinically useful information for today's practitioners. In each case, further investigations are indicated, such as biophysical assessment by Doppler ultrasonography. If growth restriction is severe, assessment for chromosomal abnormality would need to be considered in either instance. The degree of deviation is more important than the symmetry between abdominal and head measurements. Studies comparing symmetrical and asymmetrical growth restriction have failed to demonstrate differences in etiology15, fetal acid–base status at the time of cordocentesis16, neonatal morphometry17 and other indices of outcome18, 19. Using locally derived coefficients allows the computer calculation of a ‘term optimal weight’ for each pregnancy. This endpoint can be combined with a standard proportional growth curve20, derived from normal pregnancies, to delineate how the fetus is expected to reach that weight in an uncomplicated pregnancy. Thus the steepness of the curve varies with the predicted endpoint. An alternative method seeks to determine an individual growth curve for each fetus on the basis of ultrasound biometry in early pregnancy. The Rossavik model21 calculates a projected growth curve from two scans at about 18 and 24–26 weeks. However, this modeling failed to show improved prediction of fetal growth22. There are also conceptual problems with this approach. First, ultrasound error at each of the sequential scans can lead to substantial variation when forward projecting the growth curve that has been calculated from these measurements. Second, the fetus could already be affected by early-onset growth restriction, especially at the latter of the two scans, which would result in depressed values being projected as a ‘norm’. A series of studies have validated the concept of individualized standards in low-risk as well as high-risk populations for the assessment of fetal growth9, 10, 23 and birth weight24-26. The method is recommended in guidelines from the Royal College of Obstetricians and Gynaecologists27, and the use of software (which is freely available from www.gestation.net) has already been implemented in over 50 maternity units in the UK. Internationally, there is evidence that assessment against a customized, individually adjusted standard improves the distinction between normal and abnormal smallness. Smallness-for-gestational-age defined by customized limits has been found in a Dutch population to be better correlated with outcomes such as operative delivery for fetal distress and admission to neonatal intensive care than the local Dutch population standard25. Analysis of a large Swedish dataset showed that small-for-gestational-age defined by a customized birth weight centile was more closely associated with stillbirths, neonatal deaths or low Apgar scores (< 4) than unadjusted population centiles26. In fact, babies considered small by the general population standard but not by a customized standard did not have a larger risk of stillbirth, neonatal death or low Apgar scores than the average-for-gestational-age group. The inference from these findings is that ‘customized’ small-for-gestational-age represents fetal growth restriction, and that small-normal sized babies are not at greater risk than normal sized babies. Research on several fronts can further increase the international applicability of this method, focusing on fetal weight, fetal growth and birth weight. Formulae for fetal weight estimation have high validity with relatively small systematic and random error, especially when they incorporate head, abdomen and limb measurements28. The formulae of Hadlock et al.29 in particular have proven themselves in different settings14. However, because of a potential difference in anthropometric proportions in some ethnic groups, it may well be that special formulae would improve the accuracy of weight estimation. The dynamics of fetal growth may also vary for different ethnic groups, and the curves derived for predominantly American or European populations9, 11, 30 may need to be varied. This would require standardized prospective studies of fetal weight gain in different ethnic populations, to either validate the curve predicted by the customized chart, or to derive new regression equations for fetal growth dynamics in these subgroups. Birthweight coefficients for additional ethnic groups can be derived to make customized growth charts applicable to local populations. This requires an ultrasound-dated database of several thousand births, which, in addition to ethnic group, includes information about maternal height, weight and parity, and the gestational age, gender and birth weight of the baby. From experience, only a few hundred cases are needed to derive coefficients for individual ethnic groups, although this depends on heterogeneity and variation within the data. For example, recent collaboration with investigators in Auckland has defined coefficients for Maoris, Tongans and Samoans31, and the values have since been incorporated into a New Zealand version of the software program. Birthweight data from countries with other ethnic groups will allow the variation in different populations to be better understood, represented and applied in clinical practice.