Contents - Previous - Next
Diagnosis and misclassification
Standardization of growth charts
Symmetric versus asymmetric growth retardation
Correspondence: Dr LS Bakketeig
National Institute of Public Health, Department of Population Health Sciences, P.O. Box 4404 Torshov, N-0403 Oslo, Norway
Intrauterine growth retardation (IUGR) implies that intrauterine growth has been inhibited and that the fetus has not attained its growth potential. IUGR is a clinical term, and the diagnosis is usually based on small size for gestational age at birth (SGA). However, IUGR is not equal to SGA. Women seem to be programmed for having births of a certain size; some SGA babies are not IUGR, and some larger babies are still IUGR. A large number of growth charts, based on populations with different inclusion criteria and constructed according to different methods, have been developed and used; this complicates or invalidates comparisons between studies and populations. Growth charts (in percentiles) ought to be standardized and population-specific. In addition, the charts, when used as diagnostic tools, should allow for controlling for factors like sex and parity and, if possible, previous reproduction experience.
Intrauterine growth retardation (IUGR) implies that fetal growth is being inhibited and that the fetus does not attain its growth potential. Thus an IUGR newborn should have grown bigger if growth-inhibiting factors had not been operating in utero.
There is no standard definition of IUGR and the condition is diagnosed clinically. Modern technology, however, has made it possible to monitor intrauterine growth, and ultrasonography could provide a basis for operational definitions of variations in intrauterine growth; such definitions, however, have not yet been agreed upon. Intrauterine growth is still usually being assessed when the fetus is born or expelled. The size of the newborn is related to the duration of the pregnancy, and a relatively small size for gestational age is regarded as a reflection of intrauterine growth inhibition.
for gestation has been labeled either 'small-for-date' or
'small-for-gestational-age' (SGA). The 10th weight centile has
been the most commonly used cut-off for defining SGA births. This
implies that the 10 percent of babies with the lowest weight for
gestation are regarded as SGA. Or, where two standard deviations
below the mean weight for gestation is used as the lower cut-off,
the 2.5 percent of births with the lowest weight, are considered
SGA. The weight percentile of the newborn is determined
consulting growth charts (birth weight for gestation), which is
being discussed below.
A SGA birth is not necessarily an IUGR birth, and an IUGR birth is not necessarily a SGA birth. Of second births of mothers with a previous SGA baby, for example, around 30% will be regarded as SGA births if the 10th centile is used as cutoff. On the other hand only 2% of second births of mothers with a previous large-for-gestational-age birth (LGA; birth weight above the 90th centile) will be diagnosed as SGA using standard percentile charts (Skjærven and Bakketeig, 1989; Bakketeig and Magnus, 1992). This relationship is illustrated in Figure 1, which shows the 10th, 50th and 90th percentile for second births of mothers with a previous SGA birth and a previous LGA birth, respectively. This relationship reflects the strong tendency to repeat similar weight for gestation in successive births (Skjærven and Bakketeig, 1989; Bakketeig et al, 1979; 1986). It seems as if women were 'programmed' to have births of a certain size, as if they followed their own 'norms' (Bakketeig et al, 1979). If we assume that the 10 percent with the lowest weight for gestational age are to be regarded as 'true' SGA births, taking into account the birth weight 'norm' associated with the mother, and if we focus on 100 second births of mothers with a previous SGA birth, the following classification will occur:
Of the 10 'true' SGA births all will be correctly diagnosed (sensitivity = 100%). However, among the 90 births which are not 'true' SGA births, 20 will be wrongly labeled as SGA and thus be false positives (specificity = 77.8%).
On the other hand, if we focus on second births of mothers with a previous LGA baby, then the classification will be as follows:
Of the 10
'true' SGA births according to the mothers 'norm', only 2 will be
labeled SGA births while the remaining 8 will be classified as
not SGA or false negatives (sensitivity = 20%). On the other hand
all the 90 which are not 'true' SGA births will be correctly
diagnosed as such (specificity = 100%). This implies that in
diagnosing SGA births, a considerable misclassification occurs if
previous births are not taken into account.
tendency to repeat weight in successive gestations represents not
the only difficulty in determining whether a baby is really SGA
or not. The growth charts in use are based on different
population samples and the construction of the charts is based on
varying methods. Which infants were excluded from the sample and
which variables were controlled for? And how was the gestational
age determined and measured (Goldenberg et al, 1989)? When
measuring gestational age was the length rounded to the nearest
week or to the last completed week, or was the length based on
estimated day of gestation? Was gestational age estimated by
using Naegele's rule based on the first bleeding day of the last
menstrual period (LMP)? Or was the gestation period based on
This method is based on the assumption that in early pregnancy (i.e., up to 16-18 weeks of gestation) there is little or no variation in fetal size. The most common measurement used is the biparietal diameter (BPD). In 1969 Campbell claimed that BPD of fetuses with the same gestational age showed very little if any variation. Later it has become obvious that some variation exists, due to biological variation, observer variation and a tendency to regression towards the mean for the estimated date of confinement. Ultrasound dating tends on average to reduce the estimated gestation time by 3-5 days (Grennert et al, 1978).
The ultrasound dating of the pregnancies is not only based on the measurement of the biparietal diameter, but also on other fetal parameters like crown-rump length, fetal length, abdominal circumference, or some combination of these measurements. The precision of estimates varies by type and timing of the measurements (Breart and Ringa, 1990). Several studies (Kieler et al, 1993; Waldenstrøm et al, 1990; Campbell et al, 1985) have indicated that ultrasound dating provides a more precise estimate of spontaneous delivery than expected date of confinement based on LMP. However, when the woman is sure about her LMP, there is little difference in the precision, provided that the difference between the ultrasound and the LMP-based expected date of confinement is less than 1 week. For approximately 20% of pregnant women with a certain LMP, the discrepancy between the two types of estimates exceeds 1 week, and in these cases the ultrasound dating seems to be more reliable than the LMP one.
Since the introduction of ultrasonography dating of pregnancies affects the estimated gestational age by reducing the estimated length of pregnancy by a few days, this increases the proportion of preterm births and reduces the proportion of postterm births. Also the proportion of SGA births is reduced when one applies as reference the traditional growth charts based on gestational age estimated from the last menstrual period (Goldenberg et al, 1989; Bakketeig, 1991).
Thus, when ultrasonography dating of the pregnancies was introduced in an unselected mainly black population of births in Alabama, the proportion of SGA births fell from 15.1% in 1983 to 10.4% in 1985. Over the same period the mean gestational age was reduced by five days, and the proportion of preterm births increased from 11.2% to 17.0% (Goldenberg et al, 1989). This implies that many of the pregnancies are being redated which affects the distribution of gestational ages.
A longitudinal study of fetal growth in Scandinavia provides another recent example that illustrates the effect of ultrasonography redating and its consequences on whether the births are regarded as SGA or not (Bakketeig, 1991). Among 1944 pregnant women, there were 169 women with a certain LMP, but in whom the ultrasonography dating differed from the dating based on LPM by more than two weeks. Thirty (17.8%) of these women had a SGA birth if one based the estimate of gestational age on LMP, while only fifteen (8.9%) had a SGA birth if gestational age was based on ultrasonography dating. A similar misclassification of SGA births has recently been described by Zhang and Bowes (Zhang and Bowes, 1995). They claim that the errors by ultrasound estimation of gestational age exert the opposite effects on birth weight percentiles from error by LMP dating. LMP dating increases the weights at lower gestations (preterm births) while it lowers the weights for postterm births. Ultrasound dating does the opposite; it reduces the weights among preterm births and increases the weights among postterm births (Zhang and Bowes, 1995).
attention has been directed towards this phenomenon of redating
of expected date of confinement, which certainly also has its
clinical implications, in so far as it affects the clinical
management of the pregnancy and delivery (Bakketeig, 1991).
Although there are dissenting views (Goldenberg et al, 1989), growth charts, when used in a clinical context as diagnostic tools for individuals, ought to be population specific, parity and sex specific and, if possible, taking into account the mother's previous pregnancy outcome and the tendency to repeat similar birth weight in successive births (Skjærven and Bakketeig, 1989). Growth charts based on ultrasonography dating should be developed and used where pregnancy dating is based on ultrasonography.
confusion in definitions and methodology makes comparisons across
populations very difficult. Goldenberg et al. reviewed
standards for diagnosis of IUGR in the US (Goldenberg et al, 1989)
and found considerable variation. 10th percentile values based on
38 different studies and charts varied by nearly 500 grams at
term and by more than 400 grams even at 32 and 36 weeks of
gestation. The studies were characterized by location, source of
data, sample size, exclusions, patient characteristics and the
method for determining gestational age. The authors made a plea
for developing a chart based on a standard US reference
population. Zhang and Bowes (1995) have attempted to do that.
Ideally, an international population standard ought to exist for
each geographic region, and these standards should be well
defined and documented allowing for comparisons. The purpose and
use of these charts will differ depending on whether one intends
to use them more in a public health or in a clinical context. And
these perspectives will also have a bearing on the sophistication
with which the charts should be applied.
The size of a fetus at any stage of pregnancy reflects a complex interaction between time since fertilization, the rate of fetal cell multiplication and cell growth. The effect of growth inhibition will depend on the timing of the growth-retarding factor (Villar and Belizan, 1982). In a simplified manner it may be argued that inhibiting factors which operate early in pregnancy will cause a symmetrically growth-retarded fetus, while inhibiting factors which operate late in pregnancy will cause asymmetrical growth retardation. An example of the first type is a viral infection which can affect mitoses early in pregnancy, while an example of the latter type is uteroplacental insufficiency which retards fat deposition.
A symmetrically growth-retarded fetus is characterized by normal ponderal index but length, weight, head and abdominal circumferences below the 10th percentile for a given gestational age. An asymmetrically growth retarded fetus on the other hand is characterized by a relative preservation of length and head circumference, while the body weight is low (due mainly to a lower proportion of visceral and fat tissue). Thus the ponderal index of these fetuses/infants is low.
retardation, however, may fall anywhere on the continuum from
pure symmetric to pure asymmetric and will depend on the nature
and the timing of the inhibiting factors.
growth is certainly genetically co-determined. Whether this is
mediated through the fetal genes or through maternal genes and
conditions has not been fully established. The previously
mentioned strong tendency to repeat similar birth weight,
gestational length and weight-for-gestation could be explained
either way. Studies of maternal and paternal half siblings might
help to clarify the picture. Likewise, as data sets become
available of pregnancy outcomes across generations we may be able
to better separate genetical from environmental factors. In
Norway a Medical Birth Registry has existed for 30 years, which
means that it now becomes possible to study the outcome of the
pregnancies of women included as newborns in the registry from
1967 onwards. Preliminary analysis based on these data reveals a
tendency to repeat birth weight across generations that is
similar to the tendency to repeat birth weight within sibships
(Magnus et al, 1993). This suggests that genes play a
considerable role in determining birth weight. The correlation of
gestational ages across generations is weaker in spite of that
fact that the tendency to repeat gestational ages within sibships
is nearly as strong as the tendency to repeat birth weight. This
might indicate that genetic factors play a smaller role in
determining the length of pregnancy.
Intrauterine growth retarded fetuses represent a heterogeneous group; growth retardation varies in type and in magnitude. The classification and diagnoses of growth-retarded fetuses and newborns are still to a great extent based on varying classification systems and unstandardized reference materials. There is a need for modern technology to provide tools for more appropriate assessment of intrauterine growth, where the growth is related to the individual growth potential.
- I want to express my gratitude to Mrs. Liv Knurvik for
assisting in preparing this manuscript.
Bakketeig LS & Magnus P (1992): Small-for-gestational-age (SGA). Definitions and associated risks. Int. J. Technol. Assess. Health Care 8 Suppl. 1, 139-146.
Bakketeig LS (1991): Ultrasound dating of pregnancies changes dramatically the observed rates of pre-term, post-term, and small-for-gestational-age births: A commentary. Iatrogenics 1, 174-175.
Bakketeig LS, Bjerkedal T & Hoffman HJ (1986): Small-for gestational age births in successive pregnancy outcome: Results from a longitudinal study of births in Norway. Early Hum. Dev. 14, 187-200.
Bakketeig LS, Hoffman HJ & Harley EE (1979): The tendency to repeat gestational age and birth weight in successive births. Am. J. Obstet Gynecol 135, 1086-1103.
Breart G & Ringa V (1990): Routine or selective ultrasound scanning. Bailliers Clin. Obstet. Gynaecol 4, 45-63.
Campbell S (1969): The prediction of fetal maturity by ultrasonic measurement of the biparietal diameter. J. Obstet. Gynaecol Br. Commonw 76, 605-609.
Campbell S. Warsof SL, Little D & Cooper DJ (1985): Routine ultrasound screening for the prediction of gestational age. Obstet. Gynecol 65, 613-620.
Goldenberg RL, Cutter GR, Hoffman et al (1989): Intrauterine growth retardation: Standard for diagnosis. Am. J. Obstet. Gynecol 161, 271-277.
Goldenberg RL, Davis R. Cutter et al (1989): Prematurity, postdates and growth retardation: The influence of ultrasonography on reported gestational age. Am. J. Obstet. Gynecol 160, 462-470.
Grennert L, Persson P-H & Gennser G (1978): Benefits of ultrasonic screening of a pregnant population. Acta Obstet. Gynecol Scand. Suppl. 78, 5-14.
Kieler H. Axelson O. Nilsson S & Waldenstrøm V (1993): Comparison of ultrasonic measurement of biparietal diameter and last menstrual period as a predictor of day of delivery in woman with regular 28 day-cycles. Acta Obstet. Gynecol Scand 72, 347-349.
Magnus P. Bakketeig LS & Skjærven R (1993): Correlation of birth weight and gestational age across generations. Ann. Hum. Biol 20, 231-238.
Skjærven R & Bakketeig LS (1989): Classification of small-for-gestational age births: Weight by gestation standards of second birth conditional on size of the first. Pediatr. Perinat. Epidemiol 3, 432-447.
Villar J & Belizan JM (1982): The timing factor in the pathophysiology of the intrauterine growth retardation syndrome. Obstet. Gynecol Surv. 37, 499-506.
Waldenstrøm V, Axelsson O & Nilsson S (1990): A comparison of the ability of a sonographically measured biparietal diameter and the last menstrual period to predict the spontaneous onset of labor. Obstet. Gynecol 76, 336-338.
& Bowes Jr WA (1995): Birth weight-for-gestational-age
patterns by race, sex, and parity in the United States
population. Obstet. Gynecol 86, 200-208.
Different reasons can lead to an interest in identifying IUGR babies. Bakketeig's perspective is that of a clinician in an industrialized country, primarily interested in individual case management. From that point of view it is important to diagnose IUGR early, before the infant is born, in the hope of preventing a portion of fetal deaths that are associated with growth retardation by delivering such babies early. Although it has not been possible to confirm this in randomized trials, obstetricians like Goldenberg, for instance, are of the opinion that much of the reduction in fetal deaths in the US over the last 30 years results from these efforts to identify babies undergoing IUGR and delivering them early. If the purpose is to identify populations at risk as targets for intervention, it is possible to look for growth retardation not in utero, but at birth. The desirable specificity of reference values and growth curves depends to a large extent on this distinction and on the use for which they are intended.
If the intended purpose is individual case management, adjustments of growth curves for parity of the mother and sex of the infant seem justified since firstborns and girls are on average smaller and lighter, without this having any effect on outcome. For the reasons presented by Bakketeig, adjustments for prior birth outcomes (taking into consideration the smaller size of firstborns), maternal birthweight and adult height also appear desirable, it these data are available.
Based on the observation that children of socio-economically advantaged classes in developing countries follow growth reference curves of healthy, well-nourished children in developed countries, and that children of the same genetic background show widely differing growth performance depending on the environment in which they grow up, the prevailing opinion today is that people of all races have the same growth potential, even though this growth potential may not be attained in one generation, and that country- or race-specific growth references are therefore not appropriate. Making adjustments for the height of stunted parents, for instance in South Asia, could reinforce the wrong impression that children of this region are born small for genetic reasons and that not much can be done about this. Growth curves should certainly not be adjusted for factors that may be a cause of growth retardation.
Even though they are all interested in identifying children who are not reaching their growth potential, different groups of professionals approach the problem differently and use different classification schemes. Obstetricians responsible for individual case management and epidemiologists interested in fetal growth describe fetuses and infants as SGA (usually with a cut-off at the 10th centile) or IUGR and advocate the use of ultrasound for diagnosis and documentation of growth retardation. Few of them look at body proportions, and those who do usually use ponderal index as an indicator and for categorization of newborns as proportionate or disproportionate. Pediatricians and nutritionists, who are primarily interested in postnatal growth, monitor the infants' weight and height and use primarily weight-for-age, height-for-age and weight-for-height as indicators. The most frequently used cut-off to trigger intervention is weight-for-age below - 2SD of the NCHS reference population. Low height-for-age and weight-for-height are used to classify children as stunted and wasted respectively. These indicators are primarily intended to guide intervention. Since pre- and postnatal growth are one continuous process, a harmonization of approaches and classification schemes would be desirable.
Clinicians and other practitioners use cut-off points and thereby establish binary divides between those considered growth retarded, stunted, wasted, etc. and those who are not, those who are in need of an intervention and those who are not. More epidemiologically interested scientists concede that cut-offs are useful for triggering intervention, but emphasize that growth is a continuous process and so are the risks associated with variations in growth. In developing countries, for instance, where the whole distribution curve for growth is shifted to the left by one or even two SD in comparison to reference curves from developed countries, the magnitude of the public health problem may be underestimated by only looking at the proportion of those falling below a cut-off point.
It is not yet certain whether infants of different races, born at a particular weight-for-gestational-age, are all at the same or at different risks for health outcomes. It seems plausible that not only the degree of growth retardation, but also its etiology is an important determinant of risk of various undesirable outcomes; this, however has not yet been well documented. Smoking mothers give, on average, birth to smaller babies, but a comparison of offspring of mothers who smoked during all pregnancies and women who smoked only during some did not show much difference in terms of morbidity and mortality outcome.
difficult to estimate the portion of IUGR that is genetically and
the portion that is environmentally determined. Comparisons of
siblings with half-siblings in Scandinavia should be able to
clarify the situation in developed countries to some extent. The
current consensus is that genetic influences are relatively
unimportant, accounting for only 10 to 15% of the variation in
birthweight, and that in developed countries they have a greater
influence than in developing countries in which any genetic
effects are obscured by much larger environmental influences. A
partial reflection and illustration of this is that correlations
between maternal weight gain in pregnancy and birthweight tend to
be low and not statistically significant in industrialized
countries, whereas they can be significant in developing
countries where most mothers have low body mass at conception.
Contents - Previous - Next