Description of serial values of AC and EFW 1 Limitations of previous studies

Before serial ultrasound values of AC and EFW can be used in the diagnosis of lUGR, these serial changes need to be quantified. With the exception of the study of Deter et al. (1982a), previous studies have demonstrated the non-linearity of fetal growth towards term. If growth is confirmed to be non-linear, gestational age-independent changes in absolute measurements of AC or EFW cannot be used to quantify growth.

Few studies have addressed the quantification of serial ultrasound values of AC or EFW. One previous study evaluated fetal growth using simple increments in AC values with gestation (Divon et al. 1986). In this study of 90 fetuses, an abnormal rate of AC growth was defined as ^10 mm / 14 days. The choice of 10 mm / 14 days in this study was an arbitrary cut-off level based on the ability to distinguish between small and normal sized fetuses rather than on published standards of rate of growth of AC. The major problem with this study was the quantification of growth using a change in absolute measurement^f AC, inappropriate in view of the non-linearity of fetal growth at term.

In order to overcome the problems of quantifying serial ultrasound measurements given that fetal growth is not linear, other workers have reported the derivation of individual growth curves using the Rossavik Growth Model (Deter et al. 1984, Deter and Rossavik 1987, Deter et al. 1990, Deter and Rossavik 1992). The general equation for this model is: P = c ( t ) ^ ^ where P is the ultrasound parameter, & is a fixed coefficient determined by the anatomical parameter, c is related to genetic regulators of growth, s is an unknown regulatory system that modifies genetically determined growth and t is the duration of growth of the parameter (Rossavik and Deter 1984). The value k is suggested to be a fixed value for a specific ultrasound parameter. Appropriate values for k have been established for AC and EFW by regression analysis from serial scans obtained every 2 weeks from 12 to 26 weeks. The advantage of this model is that each fetus can act as its own control and that by performing 2 ultrasound scans before 27 weeks gestation, the coefficients c and s can be determined for that individual fetus (Simon et al. 1987). The individual growth curve for that particular fetus can then be derived and any subsequent deviation from that curve is regarded as failure to achieve its growth potential (Deter et al. 1989a, Deter et al. 1990). It also has the potential advantage that each fetus acts as its own control and that assessment of growth is therefore “individualised”.

The Rossavik model is the only statistical method that has been reported for the quantification of serial fetal measurements. However, there are marked limitations with model which severely limit its use in routine clinical practice:

1. The model depends on a constant value for the coefficient k for a particular ultrasound parameter. These values of k were derived from a small group of 20 middle-class mothers of different ethnic backgrounds (17 Caucasian, 2 Blacks, and 1 Hispanic) in Houston, Texas, USA (Deter et al. 1987). The women were chosen because they delivered at term and the infants had no abnormalities on paediatric and morphometric assessment. It is of considerable doubt as to whether such values of the coefficient k would be similar in other obstetric populations in other countries. As yet, no other group has derived separate values for k based on their own indigenous obstetric populations. The only other group of workers who have independently verified the use of this model reported a significant systematic over-prediction of AC and EFW in the third trimester using this method (Simon et al. 1987). It appears that further work needs to be done to verify the validity of this model, especially in the derivation of the coefficient k .

2. The derivation of growth curves for individual fetuses requires the data from two scans performed before 27 weeks gestation. The inherent assumption with this model is that fetal growth is normal before 27 weeks gestation, and that any growth deviation occurs subsequently (Deter et al. 1990). Whilst such a simplistic assumption may be true when using this model to detect late onset (ie. third trimester) lUGR, this will not be applicable to fetuses with early onset lUGR. Furthermore, the model requires the results of two scans performed before 27 weeks, information rarely available in clinical practice.

3. Possibly the most important limitation with the Rossavik Growth Model is the lack of any prospective data on the degree of deviation from the growth curve before fetuses are deemed to be growth retarded. To date, the model has not been validated against standard neonatal morphometry or measures of perinatal outcome. Whilst two studies (Deter et al. 1990, Ott 1990a) have used the model to predict morphometric evidence of lUGR in the neonate, neither study defined the antenatal criteria for abnormal growth nor used accepted morphometric criteria to define lUGR postnatally. Further prospective studies need to be carried out to evaluate this model against standard outcome measures of lUGR.

3.1.3.2 Need to quantifv serial values of AC and EFW

In view of the paucity of data on how best to describe serial values of AC and EFW, numerous statistical methods of quantifying change in AC / EFW with gestation have to be evaluated against neonatal morphometry, the “gold standard” for lUGR. The best statistical method of quantifying serial measurements would then be determined. This optimal measure of serial AC / EFW data would then be compared with other standard ultrasound measures, such as umbilical artery Doppler waveform indices, currently used to evaluate such fetuses.