Secondary Logo

Journal Logo


Prediction of Birth Weight by Ultrasound in the Third Trimester


Author Information
  • Free

Accurate prenatal estimation of birth weight would be extremely useful in the management of labor and delivery. Clinical estimation of fetal weight using abdominal palpation has been shown to be within 500 g in 85% of cases, with more accuracy in the average, term fetus than in the preterm or macrosomic fetus.1–5 Ultrasonographic measurements have been studied extensively, with the mean absolute error of sonographically predicted birth weight ranging 6–15%. These estimations can also be significantly less accurate in infants less than 2500 g or greater than 4000 g.6–10 Direct comparisons of clinical and sonographic estimates of birth weight have found ultrasound techniques to be superior for preterm infants, clinical assessment to be superior for infants between 2500 and 4000 g, and both techniques to have similar accuracy (or inaccuracy) over 4000 g.11,12

Virtually all of the studies assessing the accuracy of sonography in estimating birth weight have used sonograms performed in early labor or within the preceding week. Although this approach eliminates the potential effects of fetal growth on the estimation of weight, it possesses several inherent flaws. First, as delivery approaches, the fetal head descends into the pelvis and becomes fixed in position, making accurate measurements of the fetal head difficult or impossible.13 Second, rupture of membranes or decreasing amniotic fluid (AF) volume at term can limit the accuracy of all measurements, particularly abdominal circumference measurements. Finally, particularly with large fetuses, inability of the transducer to capture the entire cross section of the fetal abdomen or head can lead to inaccuracies of these measurements.

Recently, Mongelli and Gardosi proposed the gestation-adjusted projection method of predicting fetal weight from sonographic measurements remote from term.14 This extrapolation technique is based on the assumption that normal fetuses do not cross percentiles on growth curves. This implies that the ratio between the estimated fetal weight at the time of the ultrasound examination and the median fetal weight at that gestational age will be the same as the ratio between estimated birth weight and the median birth weight for gestational age at delivery.

The purpose of this study was to determine the relative accuracy of remote ultrasound examination for predicting birth weight according to the gestational age at which the ultrasound was performed. Specifically, we sought to compare sonograms performed just before term (period 1, 34.0–36.9 weeks' gestation) with those performed at term (period 2, 37.0 weeks' gestation or greater) to determine which would predict birth weight more accurately using the gestation-adjusted projection method.

Materials and Methods

We reviewed the records of all patients in the perinatal ultrasound database at the Johns Hopkins Hospital between May 1994 and May 1999. Patients included in the study met the following criteria: the pregnancy was a singleton gestation; the biparietal diameter, head circumference, abdominal circumference, and femur length had been measured by sonography twice after 34 weeks' gestation (once before 37 weeks' gestation and once at or after 37 weeks' gestation); data on the outcome of delivery were available; and the infant was liveborn.

All ultrasound examinations were performed by one of three experienced ultrasonographers using standard techniques with an ATL Ultramark 9 (ATL, Bothell, WA), an Acuson 128 XP 10 (Acuson, Mountain View, CA), or a Corometrics Aloka 650 (Corometrics, Wallingford, CT). Fetal weight was calculated using Hadlock's formula and the Digisonics Fetal Growth Analysis System OB-500 (Digisonics, Houston, TX).15

Data were collected on maternal age, parity, indications for sonogram, gestational age at each sonogram, fetal biometry and estimated fetal weights, gestational age at delivery, birth weight, and the latency between sonogram and delivery. Patient records were specifically assessed for the presence of fetal anomalies, suspicion of growth restriction, suspicion of macrosomia, maternal diabetes, and the presence of oligohydramnios.

Birth weight was predicted from each sonogram using the gestation-adjusted projection method and Brenner's median fetal weights for gestational age.16 The signed error in birth weight prediction was defined as the difference between the predicted and actual birth weight, with negative values implying an underestimation of birth weight. The absolute error in birth weight prediction was defined as the absolute value of this difference. Absolute errors prevent equal overestimations and underestimations from canceling each other out when averaged together. Percent errors (signed and absolute) were obtained by dividing the error in prediction by the actual birth weight.

The Shapiro Francia test of nonnormality was used to assure a normal distribution of data. Data were then compared for each pair of sonograms from the same patient using a paired t test. This allowed each patient to serve as her own control and minimized the effects of body habitus, fetal proportions, and other anatomic considerations (presence of fibroids, placental location, fetal anomalies, etc) on the accuracy of sonographic measurements.

The percentage of predicted birth weights within 5%, 10%, and 15% of the actual birth weight were also calculated for each gestational period and compared by McNemar χ2 test. The average of the predicted birth weights for each fetus was compared with each prediction alone, using paired t tests.

The effects of latency until delivery and actual birth weight on the accuracy of birth weight prediction were assessed using Pearson correlation coefficient. Since this method requires independence among observations, data from period 1 and period 2 were analyzed separately.

A P value of < .05 was considered statistically significant. Using an α = .05 and β = .80, 24 patients would be required to detect a 200 g difference in the predicted birth weight.


The study population consisted of 138 patients undergoing 276 ultrasound examinations. The overall demographic characteristics of the patients and their infants are given in Table 1. The indications for serial third-trimester sonograms included: risk for macrosomia (maternal diabetes, increased fundal height, history of macrosomia), n = 30; risk for growth restriction (underlying maternal disease, lagging fundal height, oligohydramnios, placental abnormality), n = 61; and fetal anomaly, n = 25. Twenty-two patients had no identifiable indication for having two third-trimester ultrasound examinations.

Table 1
Table 1:
Maternal and Infant Demographics

The mean predicted birth weight errors, absolute birth weight errors, percent errors, and absolute percent errors for the entire study population are shown in Table 2. The absolute gestation-adjusted projection predicted birth weights from period 1 were superior to those from period 2.

Table 2
Table 2:
Mean Errors in Birth Weight Prediction

Separating the study population by indications for serial sonography, similar results were obtained (Table 3). Interestingly, sonograms performed after 37 weeks were more likely to overestimate the birth weight of fetuses at risk for macrosomia and underestimate the birth weight of fetuses at risk for growth restriction than studies performed between 34 and 36.9 weeks. However, these differences were not statistically or clinically significant. Similarly, there were no differences in birth weight predictions within 5%, 10%, or 15% of the actual birth weight (Table 4).

Table 3
Table 3:
Mean Errors in Birth Weight Prediction by Indication for Sonogram
Table 4
Table 4:
Percentage of Correct Birth Weight Predictions

The effect of using the information from both scans to predict birth weight accurately is shown in Table 5. Averaging birth weight predictions from both sonograms did not improve the accuracy of the predicted birth weight over that from period 1 but was superior to the prediction from period 2.

Table 5
Table 5:
Effect of Averaging Information From Both Sonograms on Prediction of Birth Weight

The absolute percent errors of gestation-adjusted projection-predicted birth weights were not significantly correlated with latency until delivery during period 1 (r = .04, P > .2) or period 2 (r = .02, P > .2). Similarly, there was no correlation in the accuracy of birth weight prediction with actual birth weight (r = − .06, P > .2 for period 1 and r =− .05, P > .2 for period 2).


The accuracy of predicting birth weight from sonographic fetal measurements varies with gestational age. This study identifies the period between 34.0 and 36.9 weeks' gestation as a better time to obtain fetal measurements to predict subsequent birth weight than later in gestation. By applying the gestation-adjusted projection method on the date of delivery, using the data obtained from an ultrasound examination at 34.0 to 36.9 weeks, birth weight can be predicted with a mean absolute error of 6.2%.

Spinnato et al17 initially proposed a mathematic model for the prediction of birth weight from ultrasound examinations remote from delivery. The model, utilizing multiple linear regression and incorporating lapse time, was developed using 245 patients who delivered within 35 days of a complete fetal biometric ultrasound examination and whose newborns weighed 1000–5000 g. The model was then validated on an additional 167 cases, confirming an 8.5% mean absolute error, which was superior to the previously described static methods of Hadlock et al,18,19 Shepard et al,20 and Ott et al21; those methods did not take latency until delivery into account. Of note, although macrosomic or growth restricted fetuses were not studied specifically, they were also not excluded from the study.

Subsequently, Mongelli and Gardosi14 evaluated 276 low-risk pregnancies that delivered within 35 days of the last ultrasound examination. The gestation-adjusted projection method was compared with Spinnato's22 method for predicting birth weight. A mean absolute error of 9.93% was found for the gestation-adjusted projection method, which was significantly lower than the 11.98% error calculated using the Spinnato method on this data set.

There are two possible limitations to our study and the use of the gestation-adjusted projection method. First, the gestation-adjusted projection method, as originally described, was only applied when the latency between sonographic measurements and delivery was 35 days or less. This was because the gestation-adjusted projection method was being compared with the Spinnato22 method, which reported unacceptable deterioration of accuracy when latency exceeded 35 days. However, our data included sonograms performed between 0 and 49 days before delivery. Since Mongelli and Gardosi14 reported no significant correlation between the prediction errors and latency interval, our assumption was that this increase in latency would not significantly affect our results. This assumption was supported by our data, which did not show a correlation between latency and the accuracy of birth weight predictions.

The second limitation in applying the gestation-adjusted projection method to our data is the inclusion of fetuses with suspected growth abnormalities (macrosomia and growth restriction). These patients comprised the majority of those who qualified for this retrospective study (91 of 138, 66%) and are precisely the population for whom predicting birth weight may be critical. Since the theoretic basis of the gestation-adjusted projection method is that normal fetuses do not cross fetal weight percentiles, including fetuses that are not growing normally could certainly affect the accuracy of this method. However, when the fetuses with suspected macrosomia or growth restriction were looked at separately, the overall mean absolute percent errors in predicting birth weight were 6.7 ± 6.4% and 6.6 ± 5.4%, respectively. These are certainly within the acceptable range of prediction errors, indicating these fetuses did not cross weight percentiles during the gestational age period studied (after 34 weeks).

The use of multiple ultrasonographic examinations in predicting birth weight was examined by Hedriana and Moore.23 That study revealed a slight improvement in birth weight prediction if the average of fetal weight percentiles of serial third-trimester observations was used, particularly in the fetuses with abnormal growth.

Our results differ from those of Hedriana and Moore in that using information from more than one ultrasound examination did not improve the accuracy of birth weight prediction. There are a number of possible explanations for this difference. First, the patients included in the single examination group in their study were a subset of the multiple examination group and were not compared in a paired fashion. Second, the single examination group in their study ranged from 32 weeks to 36 weeks at the time of the sonogram. The inclusion of studies from earlier in gestation may have affected the accuracy of birth weight prediction, particularly in abnormally grown fetuses. And finally, our study used biometry obtained by only three experienced sonographers, perhaps improving the reliability of our measurements.

Our results indicate lower absolute error and absolute percent error of birth weight prediction from sonograms performed in period 1 compared with period 2; the differences are statistically significant. However, the error identified represents a difference of less than 100 g. When the positive predictive values for predicting growth restriction and macrosomia were examined, the birth weight predictions from the two periods were similar (Table 6), indicating no clinical importance. Despite this lack of clinical importance, our study does have sufficient power to conclude that performing an additional late third-trimester sonogram offers no benefit for prediction of birth weight, whether the information used from this second ultrasound study is used alone or in conjunction with data from an earlier examination.

Table 6
Table 6:
Positive and Negative Predictive Values


1. Patterson RM. Estimation of fetal weight during labor. Obstet Gynecol 1985;65:330–2.
2. Chauhan SP, Cowan BD, Magann EF, Bradford TH, Roberts WE, Morrison JC. Intrapartum detection of a macrosomic fetus: Clinical versus eight sonographic models. Aust N Z Obstet Gynaecol 1995;35:266–70.
3. Chauhan SP, Sullivan CA, Magann EF, Perry KG, Roberts WE, Morrison JC. Estimate of birthweight among post-term pregnancy: Clinical versus sonographic. J Matern Fetal Med 1994;3:208–11.
4. Hirata GI, Medearis AL, Horenstein J, Bear MB, Platt LD. Ultrasonographic estimation of fetal weight in the clinically macrosomic fetus. Am J Obstet Gynecol 1990;162:238–42.
5. Watson WJ, Soisson AP, Harlass FE. Estimated fetal weight of the term fetus: Accuracy of ultrasound vs. clinical examination. J Reprod Med 1988;33:369–71.
6. Gallivan S, Robson SC, Chang TC, Vaughn J, Spencer JAD. An investigation of fetal growth using serial ultrasound data. Ultrasound Obstet Gynecol 1993;3:109–14.
7. Sampson MB, Thomason JL, Kelly SL, Work BA. Prediction of intrauterine fetal weight using real-time ultrasound. Am J Obstet Gynecol 1982;142:554–6.
8. Warsof SL, Wolf P, Coulehan J, Queenan JT. Comparison of fetal weight estimation formulas with and without head measurements. Obstet Gynecol 1986;67:569–73.
9. Benacerraf BR, Gelman R, Frigoletto FD. Sonographically estimated fetal weights: Accuracy and limitation. Am J Obstet Gynecol 1988;159:1118–21.
10. Pollack RN, Hauer-Pollack G, Divon MY. Macrosomia in postdates pregnancies: The accuracy of routine ultrasonographic screening. Am J Obstet Gynecol 1992;167:7–11.
11. Sherman DJ, Arieli S, Tovbin J, Siegel G, Caspi E, Bukovsky I. A comparison of clinical and ultrasonic estimation of fetal weight. Obstet Gynecol 1998;91:212–7.
12. 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.
13. Barnhard Y, Bar-Havi I, Divon MY. Accuracy of intrapartum estimates of fetal weight: Effects of oligohydramnios. J Reprod Med 1996;41:907–10.
14. Mongelli M, Gardosi J. Gestation-adjusted projection of estimated fetal weight. Acta Obstet Gynecol Scand 1996;75:28–31.
15. Hadlock FP, Harrist RB, Martinez-Poyer J. In utero analysis of fetal growth: A sonographic weight standard. Radiology 1991;181:129–33.
16. Brenner WE, Edelman DA, Hendricks CH. A standard of fetal growth for the United States of America. Am J Obstet Gynecol 1976;126:555–64.
17. Spinnato JA, Allen RD, Mendenhall HW. Birth weight prediction from remote ultrasound examination. Obstet Gynecol 1988;71:893–8.
18. 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.
19. Hadlock FP, Harrist RB, Sharman RS, Deter RL, Rossavik UK. Estimation of fetal weight with the use of head, body, and femur measurements—a prospective study. Am J Obstet Gynecol 1985; 151:333–7.
20. Shepard MJ, Richards VA, Berkowitz RL, Warsof SL, Hobbins JC. An evaluation of two equations for predicting fetal weight by ultrasound. Am J Obstet Gynecol 1982;162:47–54.
21. Ott WJ, Doyle S, Flamm S. Accurate ultrasonic estimation of fetal weight. Effect of head shape, growth patterns and amniotic fluid volume. Am J Perinatol 1986;3:193–7.
22. Spinnato JA, Allen RD, Mendenhall HW. Birth weight prediction from remote ultrasonographic examination. Am J Obstet Gynecol 1989;161:742–7.
23. Hedriana HL, Moore TR. A comparison of single versus multiple growth ultrasonographic examinations in predicting birth weight. Am J Obstet Gynecol 1994;170:1600–6.
© 2000 The American College of Obstetricians and Gynecologists