- Top of page
- Supporting Information
Estimating fetal weight at term helps to determine management at the end of pregnancy. Suspected macrosomia may, for example, be an indication for Cesarean delivery in women with gestational diabetes or breech presentation1, 2. On the other hand, small-for-gestational age fetuses must be identified in utero if obstetricians are to provide close monitoring and plan their delivery to reduce perinatal risks3, 4.
A woman whose fetus has a high or low weight at term will receive the most appropriate clinical management when the most accurate diagnostic techniques are used. The two principal methods for estimating birth weight at term are clinical examination and fetal ultrasound examination. Clinical estimation of fetal weight can be based either on a subjective clinical examination, with abdominal palpation and clinical background and history, or on the measurement of fundal height. In many countries, fundal height is measured, sometimes in combination with other clinical prediction methods, to estimate fetal weight5, 6. Clinical estimation of fetal weight at term is usually considered as accurate as sonographic fetal weight estimates5, 7. Nonetheless, no large study has so far compared the use of fundal height with that of ultrasound examination to predict abnormally high or low fetal weight at term in routine practice.
The objective of this study was therefore to compare the diagnostic values of these examinations in screening for macrosomia and low birth weight in a large cohort receiving routine care.
- Top of page
- Supporting Information
We used prospectively collected data from the PREMODA study, which was a prospective observational study on the delivery of breech infants at term compared with a control group in cephalic presentation2.
Data were collected from 138 French and 36 Belgian maternity units that participated in the PREMODA study between 1 January 2002 and 31 December 2002 (Appendix S1 online). The National Data Protection Authority in Paris approved the study on 9 May 2001. The population in the PREMODA study was divided in two groups: all women who gave birth in a participating hospital to a singleton fetus in breech presentation at term (≥37 weeks' gestation), alive or not, and a reference population comprising a random selection of 5% of all women in those hospitals giving birth to singleton term infants in cephalic presentation. A local investigator in each center was responsible for prospective data collection and monitoring data quality. This person forwarded data regularly to the regional and then national coordination offices, which also monitored the data prospectively. Ultrasonographic evaluation was performed routinely by certified sonographers in each center or in an imaging institute.
Two populations were defined from the PREMODA population (n = 19 415) (Figure 1). Population A, at low risk, served as a reference population for determining the equations for estimating birth weight, based either on fundal height alone or associated with clinical indicators. Inclusion criteria for Population A were: a fundal height measurement taken no more than 8 days before labor but no sonographic data obtained in that time period, no Cesarean section before labor, and no induction of labor for maternal or fetal disorders. Population B included women for whom both ultrasound data and a fundal height measurement were available, both taken no more than 8 days before delivery, allowing comparison of the two in predicting birth weight.
Gestational age was established from the date of the last menstrual period and confirmed by first-trimester ultrasonography. The midwife or physician responsible for the patient before labor measured fundal height. Although no particular procedure was recommended for this measurement, most French-language obstetrics textbooks teach the same technique; fundal height is measured from the upper border of the pubic symphysis along the longitudinal axis of the uterus to the highest point of the uterus, whether within or outside the midline.
To define the formula for the estimated fetal weight from fundal height (EFWFH) in Population A we initially used a second-degree polynomial equation for our statistical model. However, graphical examination revealed a linear relationship between fundal height and birth weight in Population A and, as we found that a second-degree polynomial did not perform significantly better than a linear model, we chose to use the latter. Gestational age at fundal height measurement, maternal parity, rupture of membranes, fetal sex, presentation, and maternal age and ethnicity were then introduced into a multiple linear regression model to determine the coefficients of the equation to predict fetal weight using fundal height and clinical factors (EFWFH+). Fetal presentation (cephalic 1, breech 2) and sex (male 1, female 2) and maternal parity (nulliparous 0, parous 1) were coded as dichotomous variables, gestational age was expressed in weeks and fundal height in centimeters. Fetal presentation, gestational age at fundal height measurement, fetal sex and maternal parity significantly improved the model (P < 0.05); these factors were retained and included in the equation to predict fetal weight from fundal height and clinical parameters.
We used Campbell and Wilkins' equation in Population B to estimate fetal weight based on sonographic measurement of abdominal circumference (EFWAC)8. We selected abdominal circumference as the basis for predicting fetal weight because it has been found to be as accurate as other more complex formulae for predicting fetal weight or macrosomia9, 10.
The accuracy of the different formulae for estimating fetal weight was then evaluated in Population B by calculating the correlation coefficient between the estimated fetal weight obtained using each equation and birth weight. The absolute values of the percentage errors in estimated fetal weight compared with birth weight were also calculated for each formula:
After calculation of the percentage of predicted fetal weights that were accurate within 10% of the birth weight, a chi-square test for matched samples was used to evaluate differences between the formulae. A Fisher's Z normal transformation was applied to allow comparison of the correlation coefficients by t-tests, and t-tests were used to compare mean absolute errors. The diagnostic values for the prediction of a birth weight ≤2500 g or ≥4000 g were then studied by constructing receiver–operating characteristics (ROC) curves as a function of fetal weight estimated by fundal height, by fundal height and clinical factors, and by abdominal circumference. Areas under the curve were calculated with the Delong and Clark–Pearson algorithm. The sensitivity of the three methods for predicting a birth weight ≥4000 g or ≤2500 g was compared with McNemar's test, with specificity set at 95%. The positive and negative predictive values and the likelihood ratios corresponding to each of these cut-offs were calculated.
Statistical analyses were performed using STATA 9.2 software (StataCorp, College Station, TX, USA).
- Top of page
- Supporting Information
The reference population (Population A) comprised 7138 women and Population B 1689 women (Figure 1). In Population A, 2.6% of newborns weighed ≤2500 g and 5.8% ≥4000 g. We calculated the parameters of the formula estimating fetal weight from fundal height to be as follows:
The correlation coefficient of EFWFH and birth weight in Population B was 0.61 (P < 0.01).
Fundal height, gestational age at fundal height measurement, maternal parity (nulliparous 0, parous 1), presentation (cephalic 1, breech 2), and fetal sex (male 1, female 2) were significantly associated with birth weight in a multiple linear regression model. This allowed us to construct an equation to estimate fetal weight by fundal height adjusted for clinical indicators:
The correlation coefficient of EFWFH+ and birth weight in Population B was 0.67 (P < 0.01).
Table 1 describes the demographic and obstetric characteristics of the women and the distribution of birth weight of the newborns in Population B. The accuracy of the clinical and sonographic estimates of fetal weight in relation to birth weight are compared in Table 2. EFWAC correlated better with birth weight than either EFWFH or EFWFH+ (P < 0.05). The absolute value of the percentage of the mean error for EFWAC was lower than that for EFWFH (P < 0.05), but not significantly different from that for EFWFH+ (P = 0.07). However, EFWFH+ was within ±10% of birth weight in a greater proportion of cases than either EFWFH or EFWAC (P < 0.05).
Table 1. Demographic and obstetric characteristics of Population B (n = 1689)
|Maternal age (years, mean ± SD)||29.2 ± 5.1|
|Maternal national origin (n (%))|
| Europe||1312 (77.7)|
| French overseas departments and territories||12 (0.7)|
| North Africa||141 (8.3)|
| Sub-Saharan Africa||63 (3.7)|
| Asia||30 (1.8)|
| Other||29 (1.7)|
| Unknown||102 (6.0)|
|GA at birth (weeks, median (IQR))||39 (38–40)|
|Cephalic presentation (n (%))||650 (38.5)|
|Birth weight (n (%))|| |
| ≤2500 g||136 (8.1)|
| 2501–3999 g||1429 (84.6)|
| ≥4000 g||124 (7.3)|
Table 2. Comparison of the three methods for estimating fetal weight (EFW) in relation to birth weight in Population B
|EFW||Correlation with birth weight (r)||Mean absolute error (%, 95% CI)||Predictions within 10% (%)|
Figures 2 and 3 show the ROC curves corresponding to the diagnostic value of each of the three equations for predicting a birth weight of ≤2500 g or ≥4000 g, respectively. For diagnosis of a birth weight ≤2500 g, the area under the ROC curve for EFWAC was significantly greater than that for EFWFH (P = 0.03), but not that for EFWFH+ (P = 0.11). For diagnosis of a birth weight ≥4000 g, the area under the ROC curve for EFWAC was again significantly better than that for EFWFH (P = 0.001) but not that for EFWFH+ (P = 0.12).
Figure 2. Receiver–operating characteristics curves for the prediction of birth weight ≤2500 g according to estimated fetal weight calculated using fundal height (a), fundal height in combination with other clinical factors (b) and fetal abdominal circumference (c). Area under the curve = 0.85, 0.87 and 0.91 in (a), (b) and (c), respectively.
Download figure to PowerPoint
Figure 3. Receiver–operating characteristics curves for the prediction of birth weight ≥4000 g according to estimated fetal weight calculated using fundal height (a), fundal height in combination with other clinical factors (b) and fetal abdominal circumference (c). Area under the curve = 0.86, 0.88 and 0.91 in (a), (b) and (c), respectively.
Download figure to PowerPoint
When a threshold corresponding to a specificity close to 95% was set, the sensitivity of EFWAC in screening for newborns weighing ≤2500 g was significantly higher than that of EFWFH (50.7% vs. 41.2%, P < 0.05) or EFWFH+ (50.7% vs. 40.4%, P < 0.05). Similarly, the sensitivity of EFWAC in screening for newborns weighing ≥4000 g was significantly higher than that of EFWFH (54.0% vs. 37.1%, P < 0.05) or EFWFH+ (54.0% vs. 45.1%, P < 0.05) (McNemar test, P < 0.05) (Table 3).
Table 3. Diagnostic performance of the three estimated fetal weight (EFW) equations in the prediction of birth weight ≤2500 g or ≥4000 g, with specificity set at around 95%
|Equation||Cut-off (g)||Sensitivity (%, 95% CI)||Specificity (%, 95% CI)||PPV (%)||NPV (%)||LR+||LR−|
|Birth weight ≤2500 g|| |
| EFWFH||2949 (FH = 29 cm)||41.2 (38.8–43.5)||94.2 (93.0–95.3)||38.4||94.8||7.1||0.6|
| EFWFH+||2772||40.4 (38.1–42.8)||95.1 (94.1–96.1)||42.0||94.8||8.3||0.6|
| EFWAC||2335 (AC = 292 mm)||50.7 (48.4–53.1)*†||95.0 (94.0–96.1)||47.3||97.0||9.9||0.5|
|Birth weight ≥4000 g|| |
| EFWFH||3818 (FH = 37 cm)||37.1 (34.8–39.4)||95.0 (93.9–96.0)||36.8||95.0||7.3||0.7|
| EFWFH+||3710||45.1 (42.8–47.5)*||95.0 (93.9–96.0)||41.5||94.6||8.9||0.6|
| EFWAC||3710 (AC = 363 mm)||54.0 (51.7–56.4)*†||95.2 (94.2–96.2)||47.2||96.3||11.3||0.5|
- Top of page
- Supporting Information
This study shows that the accuracy of clinical examination based on fundal height measurement and of sonographic measurement of fetal abdominal circumference in the prediction of birth weight (i.e. mean absolute error and prediction within 10%) are similar. Ultrasound examination, however, identifies fetuses with a low or high birth weight at term better than clinical examination based on fundal height does in routine practice.
Fundal height is measured in numerous countries to detect abnormal fetal weight5, 6, 11, 12. Several reference curves of fundal height according to gestational age or birth weight have been proposed12–14. Reference charts provide a normal fundal height range between 37 and 41 weeks' gestation that varies from 30–35 cm (‘normal’ range in France and Belgium) to 33–38 cm6, 13, 14. These discrepancies may be due to differences in population morphometric characteristics and in methods for measuring fundal height. Therefore, we defined a reference curve relating fundal height to birth weight according to the population study and current routine practice in France and Belgium. Numerous studies have shown that fundal height has a poor predictive value for identifying small-for-gestational age fetuses14. We therefore improved the accuracy of fundal height by adding clinical factors to the initial model. Several studies have demonstrated, as also shown here, that the accuracy of weight prediction based on fundal height is improved by the simultaneous use of other clinical indicators, such as maternal morphological characteristics, parity, fetal sex and gestational age5, 6. Maternal weight and height could not be included in the model, however, because these data were not collected for the PREMODA study. Although significant, the slight increase in diagnostic value with the addition of clinical indicators to fundal height for the diagnosis of macrosomia in our study is probably too small to be of clinical interest.
Sonography is widely used for fetal weight estimation because it is objective, reproducible, and involves a well defined measurement procedure. Its accuracy has been compared with clinical estimates of fetal weight, which are subjective and poorly defined. Most of these studies had small sample sizes, had the same examiner obtain the ultrasound and clinical estimates, or had only one investigator involved in the study15–17. Although one large study did have different observers conducting the clinical and ultrasound examinations, all of the clinical examinations and 80% of the ultrasound examinations came from a single center18. By contrast, in our study, ultrasound and clinical measurements were made by several examiners with different qualifications (midwife, resident, senior staff) at 174 centers with different levels of care in France and Belgium. Our study thus provides the diagnostic value of these measures for predicting low or high birth weight with a good external validity.
One previous study showed that ultrasound measurement of fetal abdominal circumference was superior to fundal height for detecting small-for-gestational age fetuses19. No other large study, however, has compared the diagnostic value of fundal height and ultrasound measurements for the detection of high birth weight. Two large studies compared the prediction of fetal weight by abdominal palpation and ultrasonography7, 18. Sherman et al. found that the clinical estimation of birth weight in early labor was as accurate as routine ultrasound estimation the preceding week18. In the lower range of birth weight (<2500 g) ultrasound estimation was more accurate, in the 2500–4000-g range clinical estimation was more accurate, and in the higher range of birth weight (>4000 g) the accuracy of the two methods was similar. Similarly, Chauhan et al. found that ultrasound estimation of fetal weight was more accurate than clinical estimation in preterm but not in term or post-term pregnancies7. Conversely, we found the accuracy of fundal height was similar to that of abdominal circumference in predicting birth weight, but that ultrasound imaging had a better diagnostic value than fundal height for the prediction of high and low birth weights.
Three possible biases should, however, be discussed. First, 27% of the women in the PREMODA study who were otherwise eligible for Population A were excluded because their fundal height was not available in the database. It is possible that fundal height measurements were recorded more frequently when the fetus was suspected to be large or small for its gestational age. We do not believe that is an important bias, because the mean birth weight of the infants born to women excluded from or included in Population A did not differ significantly (data not shown). Second, only 8.7% of the women included in the PREMODA study had an ultrasound examination within 8 days of delivery and were eligible for inclusion in Population B. These women were probably selected for this examination because their fetuses were in breech presentation or were suspected of being large or small for their gestational age. Thus, the validity of formulae constructed from a low-risk reference population (clinical as well as ultrasound) may be questionable. That is, however, precisely what usually occurs in ‘real practice’ when the practitioner tries to identify fetal growth abnormalities in term pregnancies. Third, it might be argued that the knowledge of ultrasound measurement by the people measuring fundal height, or of the fundal height measurement by the sonographer, would bias the later measurement. However, the correlation coefficients relating ultrasound and fundal height measurement to birth weight did not depend on the order in which these measurements were made (data not shown). This suggests that the order of measurement (ultrasound or fundal height) did not have a significant impact on its accuracy in our study.
Although ultrasound examination has a better diagnostic value than fundal height for the detection of birth weight abnormalities, the diagnostic performance of both methods is quite low. Other authors have found similar results for both the identification of low birth weight and for macrosomic fetuses9, 19. The use of estimated fetal weight in clinical practice must take these limitations into account to avoid iatrogenic results associated with ‘overmanagement’ when the estimates indicate abnormal weight, or with excess confidence and inappropriate management when the examination results appear normal.
In conclusion, this study shows that ultrasound measurement of fetal abdominal circumference detects high and low fetal weight better than does clinical examination based on fundal height in routine practice.