Accuracy of Estimated Fetal Weight by Ultrasound vs. Leopold’s Maneuver

Abstract

Estimated fetal weight (EFW) is frequently used for clinical decision making in obstetrics. The goals of this study were to determine the accuracy of EFW assessments by Leopold’s and ultrasound and to investigate any associations with maternal characteristics.

Postgraduate year (PGY) 1 and 2 Obstetrics and Gynecology resident physicians from Harbor-UCLA Medical Center from 2014 to 2020 performed EFW assessments on ten preterm (<37 weeks gestational age) fetuses by ultrasound biometry and ten full-term (≥37 weeks gestational age) fetuses by ultrasound biometry and Leopold’s maneuver. Assessments were included if the patients delivered within two weeks of the assessments.

1,697 EFW assessments on 1,183 patients performed by 33 residents were analyzed. 72.6% of sonographic full-term EFWs, 69.0% of Leopold’s full-term EFWs, and 61.5% of sonographic preterm EFWs were within 10% of the neonatal birth weight (BW). The lowest estimation error in our study occurred when actual BW was 3600–3700 grams (g). After adjusting for BW, residents were found to have lower accuracy when the mother had higher BMI for full-term estimation methods (Leopold’s and ultrasound, beta=0.13 and 0.12, p=.001 and .002, respectively). Maternal BMI was not related to estimation error for preterm fetuses (beta=0.01, p=.75).

Clinical and sonographic EFW assessments performed by Obstetrics and Gynecology junior residents are within 10% of neonatal BW much of the time. In our cohort, they tended to overestimate EFWs of lower BW infants and underestimate EFWs of higher BW infants. Accuracy of full full-term EFW assessments appears to decrease with increasing maternal BMI.

INTRODUCTION

Estimated fetal weight (EFW) measurements are frequently used for clinical decision making in obstetrics. EFWs can be obtained via ultrasound biometry using the Hadlock formula¹ or using Leopold’s maneuvers, a set of four abdominal palpation maneuvers to determine position, presentation, and engagement of the fetus as well as estimating fetal weight². While clinical EFW measurements by Leopold’s maneuver are variable in accuracy^3,4, ultrasound-based EFW assessments can be accurate within 10% of the actual neonatal birth weight (BW) in up to 75% of estimates⁵.

Errors in EFW measurements can lead to unnecessary procedures or surveillance in pregnancy and potentially unnecessary Cesarean deliveries. As such, it is important to train Obstetrics and Gynecology residents on the skills of performing clinically accurate EFW measurements, both by Leopold’s maneuver and ultrasound estimation methods.

As a quality improvement initiative, Obstetrics and Gynecology junior resident physicians at Harbor UCLA Medical Center’s (HUMC) residency program were instructed to perform EFW assessments on preterm and full-term fetuses. The primary goal of this study was to determine the accuracy of junior residents’ EFW assessments by both Leopold’s and ultrasound and to investigate any associations with maternal characteristics.

METHODS

This study was conducted under an institutional review board-approved protocol for the retrospective analysis of preterm and full-term EFW assessments performed by Obstetrics and Gynecology junior resident physicians at Harbor-UCLA Medical Center from 2014 to 2020.

Postgraduate year (PGY) 1 and 2 Obstetrics and Gynecology resident physicians were instructed to perform EFW assessments on ten preterm (<37 weeks gestational age) fetuses and ten full-term (≥37 weeks gestational age) fetuses in each year. EFW assessments were included in the initiative if the gravid patients delivered within two weeks of the assessments. Preterm EFW assessments were performed by ultrasound biometry, and full-term EFW assessments were performed by both ultrasound biometry and Leopold’s maneuver. Ultrasound EFW assessments were calculated by a preprogrammed Hadlock formula.

Maternal and neonatal data were obtained from the hospital’s electronic medical record (Cerner PowerChart, North Kansas City, MO). Maternal characteristics assessed included age, body mass index (BMI), gravidity and parity, gestational age at the time of EFW assessment, and delivery route (vaginal vs Cesarean). Neonatal characteristics assessed included the newborn’s birth weight.

All EFW assessments and newborn BWs were recorded in grams. Estimation error of the EFW assessments were computed as the difference between the EFW and actual newborn BW (EFW - BW). Percent estimation error was computed as dividing the estimation error over the newborn BW ((EFW - BW)/BW). The magnitude of the estimation error and percent estimation error were computed by taking the absolute value of the respective variables.

We examined descriptive statistics for birth-related variables, which were compared between preterm and full-term births using t-tests and χ² tests. We began by examining the bivariate relationships between estimation error and each of the birth-related variables. The absolute value of the estimation error was right-skewed and a square-root transformation was subsequently applied. We assessed the accuracy of estimations across the range of actual BWs by plotting the square root absolute percent estimation error across BW, by term status (pre- vs. full-term) and by estimation method (ultrasound vs. Leopold). To determine the accuracy and significant trends of percent estimation error, a regression model was performed using the square-root of percent estimation error. The effects of various predictors on accuracy of EFW assessments was evaluated with a linear mixed effects regression model.

RESULTS

1,697 EFW assessments (either via Leopold’s or ultrasound or both) on a total of 1,183 gravid patients from 2014 to 2020 performed by 33 residents were analyzed in this study. Of the 1,183 gravid patients assessed, 486 (age 29±7, gravida 3.04±2.25, para 1.36±2.16, BMI 31±7) had preterm fetuses, and 697 (age 28±6, gravida 2.45±1.66, para 1.26±3.43, BMI 32±6) had full-term fetuses. 43% of preterm fetuses and 33% of full-term fetuses were delivered by Cesarean delivery. Maternal gravidity, gestational age, delivery route, and neonatal BWs were significantly different between preterm and full-term fetuses (Table 1).

Table 1.

Maternal characteristics of preterm vs. full-term pregnancies.

Characteristic	Preterm N = 4861	Full-Term N = 6971	p-value2
Age	29 (7)	28 (6)	0.053
Gravida	3.04 (2.25)	2.45 (1.66)	<0.001
Para	1.36 (2.16)	1.26 (3.43)	0.5
Gest. Age (weeks)	32.5 (4.3)	38.9 (2.0)	<0.001
BMI	31 (7)	32 (6)	0.064
Delivery Route			0.002
Cesarean	182 (43%)	223 (33%)
Vaginal	245 (57%)	452 (67%)
Birth weight (g)	2170 (754)	3386 (480)	<0.001

¹Statistics presented: n (%); mean (SD)

²Statistical tests performed: chi-square test of independence; t-test

The relationship between estimation error and actual BW differed depending on estimation method and preterm vs. full-term status. In terms of error as a percent of the actual birthweight, for preterm fetuses, the square-root magnitude of percent estimation error decreased with increasing BW across the span of observed BWs. For full-term fetuses using ultrasound or Leopold’s method of estimation, the square-root magnitude of percent estimation error decreased with increasing BW to a BW of 3800 grams (g), after which the square-root magnitude of percent estimation error increased with increasing BW (Figure 1).

Figure 1.

Relationship between square-root percent estimation error and birthweight, by method and preterm vs. full-term status. A 2-knot cubic spline was fit in each of the graphs.

In terms of absolute estimation error, for ultrasound estimations, a LOESS-smoothed fit of the square-root absolute estimation error vs. birthweight suggested a 2-knot linear spline model. To choose the optimal knot points for this model, we iterated through several models, varying the location of the two knot points in increments of 100g. The model with the lowest AIC value was chosen; here, we found it to be at 1600g and 3800g. Using these knot points, we saw that square root absolute estimation error increased until approximately 1600g, at which point it plateaued, and then again increased after approximately 3800g. The lowest absolute percent estimation error occurred at 3600–3700g for term infants using ultrasound (Figure 2).

Figure 2.

Relationship between square-root absolute estimation error and birthweight for sonographic estimations. A 2-knot linear spline showed that estimation error increased until 1600g, then plateaued until 3800g, then increased afterward.

We stratified Bland-Altman plots for ultrasound measurements of estimation error by BW categories. Bias (i.e. the amount by which residents overestimated BW) increased with increasing actual BW when the actual BW was less than 1600g or greater than 3800g. Between 1600g and 3800g, residents tended to overestimate fetal weight by 53.5g, and 95% of estimations were between 502.4g under the actual BW to 609.4g over the actual BW (Figure 3). As a percent of actual BW, 95% of sonographic estimations were between 19.4% underestimated weight to 22.0% overestimated weight (Figure 3).

Figure 3.

Top row: Bland-Altman plots for the estimation error vs. the mean of estimated and actual birthweight for ultrasound measurements. Bottom: Bland-Altman plots for the estimation error (as a percent of birthweight) vs. the mean of the estimated and actual birthweight for ultrasound measurement.

Bland-Altman plots for Leopold’s showed that residents tended to overestimate fetal weight by 27.6g, with 95% of estimations between 626.6g under the actual BW to 681.8g over the actual BW. As a percent of actual BW, 95% of Leopold’s estimations were between 18.6% underestimated weight to 22.0% overestimated weight (Figure 4).

Figure 4.

Left: Bland-Altman plots for the estimation error vs. the mean of estimated and actual birthweight for Leopold measurements. Right: Bland-Altman plots for the estimation error (as a percent of birthweight) vs. the mean of the estimated and actual birthweight for Leopold measurements.

Considering estimation method and preterm vs. full-term status, sonographic estimations on full-term fetuses appeared to be the most accurate. 72.6% of sonographic full-term EFWs, 69.0% of Leopold’s full-term EFWs, and 61.5% of sonographic preterm EFWs were within 10% of the neonatal BW. 95% of sonographic and Leopold’s full-term EFWs were within 20% of neonatal BW, and 95% of sonographic preterm EFWs were within 25% of neonatal BW.

Maternal BMI appeared to affect estimation error for full-term fetuses estimated by sonographic and Leopold’s maneuver (Figure 5). After accounting for estimation method, preterm vs. full-term status, and actual BW, each 10-unit increase in birthweight was associated with a 0.10–0.11 unit increase in square root estimation error for Leopold’s and ultrasound methods, respectively (p’s<0.01; Table 2).

Figure 5.

Estimation error is associated with maternal BMI for both estimation methods in full-term fetuses, but not associated with maternal BMI for preterm fetuses.

Table 2.

Results from a mixed-effects linear regression model. Estimation error (square root of absolute value) is the outcome. A random effect for resident is included. Parameter estimates (B), 95% confidence intervals, and p-values are included. A piecewise linear spline was used for the overall model, while a natural cubic spline was used for the stratified models.

	Overall Model			Stratified Models
				Full-Term: Leopold’s			Full-Term: Ultrasound			Preterm: Ultrasound
Predictors	B	CI	p	B	CI	p	B	CI	p	B	CI	p
BMI (10-unit increase)	0.08	0.03 – 0.12	0.001	0.10	0.02 – 0.17	0.01	0.11	0.03 – 0.18	0.006	0.03	−0.05 – 0.10	0.529
Estimation (Ultrasound vs. Leopold’s)	−0.87	−1.48 – -0.26	0.005
Birthweight (100g) (<1600g)	0.85	0.62 – 1.08	<0.001
Birthweight (100g) (1600–3800g)	−0.07	−0.13 – −0.01	<0.001
Birthweight (100g) (>3800g)	0.96	0.69 – 1.23	<0.001
Birthweight: Spline 1				−97.71	−117.66 – -77.77	<0.001	−6.27	−12.38 – -0.15	0.045	11.95	8.08 – 15.82	<0.001
Birthweight: Spline 2				−5.38	−8.99 – -1.77	0.003	6.18	2.29 – 10.08	0.002	4.7	1.38 – 8.02	0.006
N (residents)	33 id			33 id			33 id			33 id
N (observations)	1697			651			639			407

DISCUSSION

In this study, we aimed to investigate the accuracy of EFW assessments of preterm and full-term fetuses via ultrasound and Leopold’s maneuver as performed by junior Obstetrics and Gynecology residents in a United States residency program and their associations with maternal characteristics.

Overall, the accuracy of EFW assessments performed by Obstetrics and Gynecology residents were consistent with those presented in the literature – approximately 60 to 75% of measurements falling within 10% of the newborn’s BW^6–9. Additionally, accuracy of EFW assessment in our study is also consistent with studies evaluating EFW assessments performed by experienced sonographers and physicians⁴. This indicates that resident physicians can become proficient in both Leopold and ultrasound EFW assessments early in training.

In the current study, percent estimation error decreased as BW increased for preterm fetuses. Few prior studies have investigated the accuracy and trend of preterm EFW assessments. Most such studies evaluated the accuracy in peri-viable fetuses or infants with very low BW^10,11. Similar to the present study, these studies have found that percent estimation error in EFW assessments increases with decreasing BW. This finding has several potential clinical implications. First, with lower accuracy at lower BWs, it may be less likely that obstetricians would suspect fetal growth restriction (FGR) in a small-for-gestational age (SGA) infant. Studies have shown that false negative rates for SGA can be as high as 78%¹². In addition, accurate preterm EFW assessments facilitate pediatric counseling of neonatal prognosis and maternal decision making, especially in regard to fetal surveillance and interventions in cases of peri-viable preterm premature rupture of membranes (PPROM) or preterm labor.

For full-term EFW assessments, we found that percent estimation error is the smallest at a BW of approximately 3600–3700g with increasing percent estimation error as BW increases or decreases from that value. We also found that for lower BW full-term fetuses, residents were more likely to overestimate the fetal weights, and for higher BW full-term fetuses, residents were more likely to underestimate the fetal weights. This result likely represents a “regression towards the mean” – that is, residents may tend to overestimate low BW fetuses and underestimate high BW fetuses because the most common estimates are in the 3600–3700g range and residents may not want to deviate too far from that value. Interestingly, a study evaluating EFW assessments for fetuses above 34 weeks gestational age in Sri Lanka found a similar trend to that of this study¹³. No other studies have reported such a trend. Most other studies evaluating sonographic full-term EFW assessments have found EFW assessments typically overestimate the neonatal BW^14–17. A potential explanation for this trend, especially in trainees with less experience, may be an unintentional tendency by the performing provider to avoid diagnosing FGR or macrosomia. Our results for EFW estimation error can be directly translated to errors in fetal weight percentile which are used to diagnose FGR and macrosomia when less than the 10^th percentile or greater than the 90^th percentile, respectively. True FGR versus a false positive or negative FGR result can be determined by comparing the EFW percentile to the actual neonatal BW. As noted above, unrecognized FGR is not uncommon as difficulty measuring fetal parts using third trimester ultrasound increases^15,17,18. Unrecognized FGR may preclude patients from receiving the appropriate antenatal surveillance and potentially increase the risk for stillbirth. Likewise, studies have shown a higher incidence of shoulder dystocia in patients with undiagnosed macrosomia¹⁹. To avoid such risks, if these findings are reproduced in future studies, Obstetrics and Gynecology training programs may consider additional training sessions on obtaining fetal biometry in preterm fetuses and in full-term fetuses with high BWs. Care should also be taken to ensure accurate dating so that EFW measurements and fetal weight percentiles are a true reflection of gestational age. Clinicians should be aware of other measurements of fetal growth and well-being including fetal abdominal wall fat thickness²⁰, deepest vertical pocket (DVP) or amniotic fluid index (AFI), umbilical cord artery Doppler examinations, biophysical profiles (BPP), and overall trend in growth curves²¹ when making clinical decisions such as delivery timing.

In our cohort, the accuracy of full-term EFW assessments by both estimation error and percent estimation error appears to decrease with increasing maternal BMI. This trend was true for both ultrasound and Leopold’s assessments of full-term fetuses. While some studies show a similar trend²², other studies have shown that there is no difference in accuracy of ultrasound EFWs at full-term when analyzed in terms of maternal BMI^23–25. In the present study, the decreased accuracy of EFW assessments with increasing maternal BMI may be due to the general lack of ultrasound and Leopold’s experience in the performing resident physicians, whereas assessments in other studies are performed by practicing physicians, sonographers, or midwives. As such, Obstetrics and Gynecology residency programs may consider extra training sessions on optimizing sonographic skills to improve accuracy of EFW assessments in patients with elevated BMI.

Other than maternal BMI, estimation error for ultrasound EFWs may be attributed to a variety of reasons including performer inexperience, incorrect dating, inconsistency in caliber placement, incorrect image planes due to probe or fetal positioning, or ultrasound calibration errors²⁶. EFW estimation error from Leopold’s maneuvers may be also attributed to performer inexperience as well as maternal conditions such as fibroids, polyhydramnios, or anterior placenta location².

Lastly, when full-term clinical EFW assessments by Leopold’s and ultrasound assessments are compared, we found that the ultrasound EFWs are more accurate. Though the literature has mixed findings on this comparison, similar to our findings, most studies report that ultrasound estimations are as accurate^6,27 or more accurate^28–30 than clinical estimations.

Our findings demonstrate the importance of training Obstetrics and Gynecology residents in how to perform accurate sonographic and Leopold’s EFW assessments to increase accuracy and reduce the number of false positive outcomes. As radiologists also perform prenatal ultrasounds, it is similarly important for radiologists and radiology resident physicians to receive good training in measuring accurate EFWs to allow for well-informed clinical decision making by the obstetrician.

Strengths of this study include a standardized protocol at one institution, thereby avoiding variability in instructions to learners. In addition, each provider performed the same number of preterm and full-term EFW assessments in both years as junior resident physicians, providing continuity of providers. Skews in the data were accounted for by performing square-root analysis of the studied variables to normalize the data.

Our study also has some limitations. Estimations were performed by multiple providers who were junior resident physicians. Our sample size of EFW assessments is relatively small. The study was performed at an urban county hospital serving an underserved population with differing racial and socioeconomic backgrounds. Our results may not be generalizable to populations that differ significantly. Because of the retrospective design, not all maternal characteristics were available for optimal statistical analysis. Lastly, the accuracy of EFW assessments may have been affected by the two-week margin. Prior studies have shown that accuracy was significantly different when greater than one week passed between examination and delivery⁴.

CONCLUSION

Both clinical and sonographic estimated fetal weight assessments performed by Obstetrics and Gynecology junior resident physicians are similar in accuracy to those performed by other providers or sonographers, falling within 10% of neonatal BW much of the time. The lowest absolute percent estimation error in full-term EFW assessments occurred at approximately 3600–3700g with overestimation at lower BW infants and underestimation at higher BW infants. In addition, accuracy of full full-term EFW assessments appears to decrease with increasing maternal BMI. Further studies should be performed to investigate and reproduce these trends.