To construct new reference charts and equations for fetal biparietal diameter (BPD), head circumference (HC), abdominal circumference (AC) and femur length (FL), using a large sample of fetuses examined at 15–40 weeks in France, and to compare them with previous references.
The study data were obtained over a continuous 1-year period from a population of pregnant women undergoing ultrasound examination. Excluded were those with a known abnormal karyotype or congenital malformation, multiple pregnancies, and those with no first-trimester dating based on crown–rump length. No fetuses were excluded on the basis of abnormal biometry or birth weight. For each measurement, separate regression models were fitted to estimate both the mean and the SD at each gestational age.
Full biometric measurements were obtained for 19 647 fetuses. New charts and reference equations are reported for BPD, HC, AC and FL. Prediction intervals for the new reference charts were similar to those of previous ones, whereas there were some differences in predicted centiles.
Fetal biometry is an important part of routine examinations during the second and third trimesters of pregnancy. The use of cross-sectional reference charts and equations from a population as similar as possible to the screened population remains the gold standard. It has been demonstrated recently that the choice of reference charts and equations for fetal size has a significant impact on the quality of fetal biometry1.
In France, most operators trained in fetal sonography belong to the College Français d'Echographie Fœtale (CFEF). ‘Trained operators’ are obstetricians, radiologists or midwives who have passed a national ultrasound examination. A few years ago, CFEF undertook to establish reference charts for fetal size (Biometry 2000 project). These charts have been published elsewhere2 and are now used widely in France. However, the analytical method used in this previous report did not allow the charts to be published along with the corresponding equations. Yet such equations, and adequate modeling of the variability of fetal measurements across gestational age (GA), are mandatory for quality control3 and for comparison of fetal centiles across different reference charts1.
This study was therefore undertaken to establish French reference charts and equations in full accordance with the recommended method of analysis4, and to compare these charts with other previously published ones5–10.
This study involved a population of fetuses undergoing ultrasound examination over a continuous 1-year period. All measurements (to the nearest mm) were performed with no time constraints, by 16 trained operators, all of whom were performing more than 2000 ultrasound examinations per year at the time. Excluded were: any cases with a known abnormal karyotype or congenital malformation, multiple pregnancies, and those with no first-trimester dating based on crown–rump length (CRL)11. No fetuses were excluded on the basis of abnormal biometry or birth weight. GA and biparietal diameter (BPD), head circumference (HC), abdominal circumference (AC) and femoral diaphysis length (FL) measurements were collected in every case. Only fetuses for which all measurements were available were included in the statistical analysis. For GA, fractions of weeks were computed to the nearest week, with fractions of < 4 days and ≥ 4 days being assigned to the lower and higher weeks, respectively.
All biometric measurements were made as recommended by CFEF. BPD and HC were measured on a transverse view of the fetal head in an axial plane, at the level where the continuous midline echo is broken by the septum pellucidum in the anterior third, as described by Campbell and Thoms12. BPD was measured with the calipers placed in the middle of the echo of each side of the fetal skull and HC was derived from the measurements of the occipital–frontal diameter and the BPD, using the formula π(d1 + d2)/2, where d is diameter. AC was measured on a transverse circular plane of the fetal abdomen, just above the level of the cord insertion, as described by Campbell and Wilkin13, and was also derived from the two maximum diameters of the circumference. FL was measured on a plane showing the entire femoral diaphysis, with both ends clearly visible and an angle of < 45° to the horizontal. During the third trimester, particular care was taken not to include the epiphysis.
Statistical analyses were performed using Statistica (StatSoft, Inc. (2001) STATISTICA (data analysis software system), vers. 6, Tulsa, OK, USA) and the data were analyzed as recommended previously4, 14. The normality of measurements at each week of gestation was assessed using the Kolmogorov–Smirnov and Shapiro–Wilk's W-tests. Given the large sample size, statistically significant non-normality was accepted unless the normal plot showed clear deviation from a straight line4. Least-square regression analysis was used to model the mean, by fitting a polynomial equation. A cubic polynomial model was used (y = a + b × GA + c × GA2 + d × GA3) unless the R2 statistic and/or the fitted curve was not satisfactory4, 14. The variability in measurements was modeled by computing the SD at each week of gestation and SD values were regressed on GA using a simple linear equation (y = a + b × GA).
From these predictive mean and SD equations, it was then possible to calculate any required centile using the formula: centile = mean + K × SD, where K is the corresponding centile of the Gaussian distribution (for example, determination of the 10th and 90th centiles requires K to be ±1.28; determination of the 5th and 95th centiles requires K to be ±1.645, and determination of the 2.5th and 97.5th centiles requires K to be ±1.96). Charts were computed by plotting predicted means and 3rd, 10th, 90th and 97th centiles against GA.
In order to compare our new reference equations with previously published ones, we used the following method: the median and 5th and 95th centiles from three previously published fetal size equations5–10 were calculated at each week of gestation from 15 to 40 weeks. These centiles were then expressed as Z-scores calculated with our new equations. In all cases, Z-scores were calculated using the formula: Z-score = (XGA − MGA)/SDGA, where XGA is the measured value at a known CRL-based GA, MGA is the mean value obtained with the reference equation used at this GA, and SDGA is the SD associated with the mean value at this GA obtained with the reference equation. Results were presented graphically across GA. Such figures allow one to compare visually the predicted mean, the 5th and 95th centiles, and the prediction intervals of each reference equation.
Full biometric measurements (BPD, HC, AC and FL) were obtained for 19 647 fetuses. The median (interquartile range) number of examinations performed at each week of gestation was 238 (103, 1008). At each gestational week over 50 fetuses were always examined, with the exception of weeks 15 and 39, when 39 and 34 fetuses, respectively, were examined. The mean (± SD) number of examinations performed by each operator was 1228 ± 635.
Raw data were fitted satisfactorily with a cubic polynomial model for all biometric parameters as follows (all measurements in mm and GA in exact weeks):
SDs across GA were fitted using a simple linear fit. Figure 1 illustrates the increase in variability of the measurements with GA, together with the linear model of this variability. Fits for SDs were as follows (all SD in mm and GA in exact weeks):
For BPD: SD = 1.5022 + 0.0636 × GA(R2 = 70.81);
For HC: SD = 2.7945 + 0.345 × GA(R2 = 86.62);
For AC: SD = −2.3658 + 0.6459 × GA(R2 = 92.81);
For FL: SD = 1.0809 + 0.0609 × GA(R2 = 80.52).
Figure 2 illustrates the goodness of fit of our model by showing raw data for each measurement, with the fitted 5th, 50th and 95th centiles. Figure 3 gives charts for daily practice, with 3rd, 10th, 50th, 90th and 97th centiles fitted for each biometric parameter.
Using three previously published sets of reference equations5–10, we then calculated the 5th, 50th and 95th centiles for each biometric parameter and these values were expressed as Z-scores, based on our new equations (Figure 4).
Our study was undertaken in order to provide sonographers with new reference equations compatible with their practice. These 15–40-week charts and reference equations are of great importance to all fetal sonographers, and, as fetal biometric values differ among ethnic groups15, they are mandatory for French sonographers. In addition, although the charts can be used in daily practice to plot measurements on GA-based reference curves in order to visualize individual measurements on the population distribution, the reference equations for the mean and SD provide sonographers with the formulae needed to calculate any percentile.
Z-scores have been used increasingly in recent years, and are the World Health Organization (WHO)-recommended system for comparing individual anthropometric measurements with the reference population16. A major advantage of the Z-score system is that a group of Z-scores can be used as input for summary statistics. This system can therefore be used by sonographers as a powerful quality-control tool, allowing them to choose appropriate reference equations and to audit their daily practice1, 3.
We paid particular attention to the methodology used to construct these new ranges, doing our best to follow the recommendations made by the authors of previous methodological reviews4, 14. Step by step, we modeled the mean and then the variability, checked the model, and derived centiles and reference charts. However, although the analytical method followed standard recommendations strictly, the same did not apply to the selection of our population sample. As the Biometry 2000 project aimed to involve as many sonographers and as many measurements as possible, in order to depict French sonography practice as a whole, we had to deal with a very large sample (almost 20 000 fetuses, examined by 16 sonographers). Moreover, because it reflected daily practice, the 20–24-week and 30–34-week examinations were more frequent than were others. This sampling method could have introduced a major bias. However, it did not increase the variability of our results: our SDs are in keeping with those of other references5–10. Furthermore, variability based on measurements by several examiners is probably more relevant to clinical practice.
This paper also provides an original and practical graphical means for comparing a new reference with existing references. As illustrated in Figure 4, it can be seen rapidly how one reference differs from another: if the 5th, 50th and 95th centiles of an existing reference fit those of the new reference, then the three curves should form three lines at y = −1.645, y = 0 and y = +1.645. This also allows the prediction intervals of the various references to be compared, by comparing the [−1.645 to +1.645] interval with the [5th centile to 95th centile] interval of other references. As mentioned earlier, these figures show that our new references have prediction intervals similar to those of previously published references5–10 as the [−1.645 to +1.645] range and the [5th centile to 95th centile] range of the three other references are similar. FL values were roughly similar among the different references across GA. In contrast, there were some significant differences (up to 1.5 SD) in the predicted centiles of head measurements among the references. This once again underlines the important impact of the choice of reference charts and equations in fetal biometry1. Overall, head measurements in the other three studies were larger compared with our study. This might be explained, at least in part, by differences in caliper placement: in our study, the calipers were placed in the middle of the echo of each side of the fetal skull (the preferred method in France), whereas they were placed outer to outer in the other references. Chitty et al.6 collected data with calipers placed both outer to outer and outer to inner. The difference between the two methods ranged from 1.5 to 3 mm. This underlines the need for sonographers to check the methodology used to construct the reference. Finally, although there were some differences in AC centiles among the references, there was no particular trend in one direction, as French fetuses seem to have larger abdomens until the beginning of the third trimester, when this difference starts to reverse.
In conclusion, this study provides new charts and equations for fetal size in France, based on a very large sample of fetuses, as well as a method with which to compare new references and existing ones. It demonstrates that satisfactory prediction intervals can be obtained even with very large samples. Finally, it provides sonographers with the tools needed to choose the reference that best fits their practice, and allows them to use Z-scores for biometry and to better control the quality of their work.