This Institutional Review Board–approved study was prospective and was performed over a defined 12-month period. Women with nonanomalous singleton gestations who were scheduled for elective cesarean delivery at term (at or after 37 weeks' gestation) were eligible for enrollment if delivery was scheduled on a day when sonographers were available. Women were contacted within 1 week of their scheduled delivery, to determine their willingness to participate. After we obtained informed consent, an ultrasound examination was done within 24 hours before scheduled cesarean delivery. All ultrasound examinations were done by one of two sonographers, using either 3.5- or 5-MHz curvilinear transducers (HDI Ultramark 9; Advanced Technology Laboratories, Bothell, WA). On-screen calipers were used for all measurements, and circumference calculations were generated by the ultrasound software. Intra- and interobserver variations were calculated using findings from examinations of ten women. Interobserver variation was determined by comparing the mean values (in centimeters) obtained by each sonographer of three measurements for each technique on each woman. Intraobserver variation was calculated from three measurements for each technique by each sonographer.
Fetal BPD, FL, HC, and AC were measured. The BPD was measured from the outer to the inner skull table at the level of the thalami. The HC measurements were made at the same level as the BPD measurements. The FL was measured from the proximal end of the greater trochanter to the distal metaphysis. Circumferences of the abdomen were measured in standard transverse planes at the levels of the stomach and umbilical vein–ductus venosus complex.
Three different circumference measurement techniques were used to measure each HC and AC. No attempt was made during the examinations to compare AC and HC measurements for each technique. Calipers were used in two-diameter measurements to determine two perpendicular dimensions, the anterior-posterior diameter (D1) and the transverse diameter (D2), which then were converted to an estimated circumference by the equation (D1 + D2) × 1.57. In ellipse measurements, machine-generated calipers were used to create an on-screen ellipse that was maneuvered to approximate the outline of the outer perimeter of the fetal head or abdomen. Trace measurements involved operator-controlled tracing of the outer perimeter of the fetal head or abdomen. The order of circumference measurement techniques was randomly rotated between examinations. Separate estimated fetal weights (EFWs) were determined for each of the three circumference measurement techniques, using a formula incorporating BPD, FL, HC, and AC by Hadlock et al1: EFW = log (birth weight) = 1.5115 + 0.0436 (AC) + 0.1517 (FL) − 0.00321 (AC × FL) + 0.0006923 (BPD × HC). When more than one measurement was obtained for any given biometric measure or technique, the mean was used for fetal weight calculation and other analyses.
The accuracy of birth weight estimation for each of the circumference measurement methods was estimated by calculating the mean percent deviation ± SD from the actual birth weight, using the formula 100 × (EFW − birth weight)/birth weight. The mean percent deviation from the actual birth weight represented the systematic measurement error and the SD represented the random measurement error. The mean absolute percent deviation for each technique also was determined, as follows: 100 × | EFW − birth weight |/birth weight. The mean residual | EFW − birth weight | was calculated for each circumference measurement technique to determine the direction and magnitude of the errors. Mean percent deviations, absolute percent deviations, and residuals were compared between techniques using analysis of variance. The significance of the deviations from the actual birth weight for each circumference-specific EFW was calculated by one sample t test to determine whether percent deviations were significantly different from zero. The percentage of cases in which EFWs were within 5 and 10% of the actual birth weight was calculated for each of the circumference measurement techniques and was compared across techniques using test of proportions. Percentages for birth weight were determined using data from the enrolled individuals. Categories were created for birth weight in the 25th percentile, in the 26th through the 75th percentile, and above the 75th percentile. The absolute percent errors for estimations for each birth weight percentile category were calculated by averaging the absolute values of the percent errors and were compared between techniques using analysis of variance. Statistical significance was defined as P < .05.
Fifty-four women who met initial entry criteria were scanned before their elective cesarean deliveries. One had a fetal anomaly on ultrasound examination, and therefore 53 women were included in the analyses. The mean (± SD) maternal age was 32.3 ± 5 years and the medians (ranges) for gravidity and parity were 2 (1–7) and 1 (0–3), respectively. The mean (± SD) gestational age was 38.1 ± 0.9 weeks and the mean birth weight was 3536 ± 472 g. The mean (± SD) HC and AC measurements as well as the intraobserver and interobserver variations for each measurement technique are given in Table 1.
The accuracies for estimating BW using the formula of Hadlock et al1 for each of the circumference measurement techniques are reported in Table 2. The EFWs derived from the two-diameter and ellipse circumference measurement techniques were not significantly different from the actual birth weights. The two-diameter method underestimated and the ellipse method overestimated the birth weight by nonsignificant amounts. The trace circumference measurement method significantly overestimated the actual birth weight, by 8.2 ± 11.6% or 283 g (P < .05). The mean percent deviation and percent absolute deviation of EFWs from actual BWs were significantly different for the trace method when compared with either the two-diameter or ellipse methods (P < .001). That was also true for the measured residual weights (P < .001).
When the EFW for each circumference measurement method was examined according to the frequency with which the measurement was within 5 and 10% of the actual birth weight, the two-diameter and ellipse methods were found to be similar and consistently more accurate than the trace method (Table 3). However, the difference was statistically significant only for measurements within 10%. We also found that as the actual birth weight increased, deviations of the estimated birth weight from the actual birth weight increased for two-diameter and ellipse circumference measurement methods, determined by an increasing spread of residuals (Figure 1). The greatest error in weight estimation of 16.0 ± 10.2% occurred with the use of the trace method in infants in the 25th-percentile birth weight category of 3170 g. The accuracy of the trace method was not significantly different (P = .40) from that of the two-diameter and ellipse methods for birth weights above the 75th percentile of 3865 g (Table 4).
The accuracy of sonographic fetal weight estimates is dependent on multiple factors, including biometric measures used for estimated weight calculations. Various combinations of BPD, HC, AC, and FL measurements have been proposed for predicting birth weight, with varying accuracies. Hadlock et al2 observed that EFW formulas incorporating HC, AC, and FL achieved slightly higher accuracy compared with other formulas and also reported that infant size can influence those estimates. Circumference measurement methodology also can affect measurement results.3 Our data confirmed that the technique of measurement of AC and HC contributes significantly to the accuracy of prediction of birth weight at term. Our findings also suggested that measuring circumferences using the two-diameter or ellipse method was relatively accurate, but the trace method for estimating fetal AC or HC should be avoided.
The EFW at term is clinically relevant because management decisions often are based on the results of sonographic fetal weight estimations.4,5 Clinicians sometimes make decisions based on the EFW, including decisions regarding induction and route of delivery (vaginal versus cesarean). Chauhan et al6 suggested that sonographic estimations of fetal weight at term can be so inaccurate that their use should be limited. We found that with our chosen fetal weight–estimation equation, accuracy can be optimized by choice of circumference measurement technique.
Previous investigators2,7 indicated that the accuracy of fetal weight estimation can be affected significantly by infant birth weight. We found that the absolute difference in grams between estimated and actual birth weights increased as birth weight increased, for the two-diameter and ellipse circumference measurement techniques, but that the percent error remained relatively constant. The trace circumference measurement method had the largest error for the lower EFWs. That might be due to greater difficulty in manipulating the caliper trace around the more acute curvature of smaller circumferences. Although accuracy of fetal weight estimation with the trace technique improved when birth weight was greater than 3865 g, the overall higher percent error across all birth weight limits the usefulness of that technique.
In our study, ultrasound estimation of fetal weight was done within 24 hours of delivery. Delays of more than 7 days between ultrasound examination and birth might elevate error rates falsely, because each fetus has a variable rate of continued growth after ultrasound examination until birth.7 Thus, our method optimized the accuracy of ultrasound estimation of fetal weight by avoiding the confounder of continued fetal growth after ultrasound examination. Although we used only one fetal weight–estimation formula, that is the formula used clinically in our ultrasound unit and is a common default formula for ultrasound equipment. Given the findings presented in Table 1, that the trace technique yielded greater AC measurements, we would expect other weight-estimation formulas to be affected by circumference measurement technique, depending on the relative importance of HC and AC measurements within the specific formula.
1. Hadlock FP, Harrist RB, Carpenter RJ, Deter RL, Park SK. Sonographic estimation of fetal weight: The value of femur length in addition to head and abdomen measurements. Radiology 1984; 150:535–40.
2. Hadlock FP, Harrist RB, Sharman RS, Deter RL, Park SK. Estimation of fetal weight with the use of head, body, and femur measurements—A prospective study. Am J Obstet Gynecol 1985; 151:333–7.
3. Tamura RK, Sabbagha RE, Pan WH, Vaisrub N. Ultrasonic fetal abdominal circumference: Comparison of direct versus calculated measurement. Obstet Gynecol 1986;67:833–5.
4. Levine AB, Lockwood CJ, Brown B, Lapinski R, Berkowitz RL. Sonographic diagnosis of the large for gestational age fetus at term: Does it make a difference? Obstet Gynecol 1992;79:55–8.
5. Rouse DJ, Owen J, Goldenberg RL, Cliver SP. The effectiveness and cost of elective cesarean delivery for fetal macrosomia diagnosed by ultrasound. JAMA 1996;276:1480–6.
6. Chauhan SP, Hendrix NW, Magann EF, Morrison JC, Kenney SP, Devoe LD. Limitations of clinical and sonographic estimates of birth weight: Experience with 1034 parturients. Obstet Gynecol 1998;91:72–7.
7. Sabbagha RE, Minogue J, Tamura RK, Hungerford SA. Estimation of birth weight by use of ultrasonographic formulas targeted to large-, appropriate-, and small-for-gestational-age fetuses. Am J Obstet Gynecol 1989;160:854–62.