Accuracy of ultrasound in estimating fetal weight in New Zealand

Abstract

Introduction

Ultrasound estimation of fetal weight is an important factor guiding antenatal management. We aimed to review the accuracy of ultrasound in predicting fetal weight and birthweight category and identify influencing factors.

Methods

We performed a retrospective study of term pregnant women who underwent ultrasound within 7 days of delivery at National Women's Health between January 2019 and January 2020. Stillbirths, major fetal anomalies and multiple pregnancies were excluded. Estimated fetal weight (EFW) was calculated using Hadlock formula and compared with birthweights. We evaluated change in weight categories due to these errors.

Results

Of 560 fetuses included, three quarters (n = 425, 76%) of EFWs were within 10% of birthweight. 135 fetuses had EFWs either less than 90% (n = 19) or greater than 110% (n = 116). Fetuses with EFW < 90% had longer times between scanning and delivery, lower EFW and higher maternal BMI. Fetuses with EFW > 110% were associated with higher EFW, later gestational age and older maternal age. US incorrectly estimated 71 (12.7%) fetal birthweight categories. Underestimated weight category (8.9%) was associated with higher maternal BMI.

Discussion

Inaccurate EFWs were more common at the extremes of fetal weight. A significant association was underestimation birthweight in mothers with increased BMI, who are at increased risk for perinatal and surgical complications.

Conclusion

Our accuracy of 76% correctly predicted EFWs compares favourably with previous studies. Clinicians and sonographers should be aware of the increased risk for inaccurate categorisation of fetuses at the extremes of EFW and in mothers with increased BMI.

Introduction

Ultrasound (US) estimation of fetal weight is an important factor guiding obstetric antenatal management and decisions around mode and timing of birth. Recognition of abnormal fetal growth identifies fetuses at increased risk of morbidity and mortality and is essential for planning appropriate care. Inaccuracies may lead to over‐medicalisation of pregnancy with associated increase in maternal and family stress and increased use of clinician time. Inappropriate labelling of fetuses as small for gestational age (SGA) could result in over utilisation of hospital resources. Fetuses not appropriately identified as SGA or large for gestational age (LGA) may be at increased risk of poor outcomes, with missed opportunity for close monitoring and early intervention. Our purpose was to examine the accuracy of ultrasound scans performed at National Women's Hospital by comparing ultrasound estimated fetal weights obtained within 7 days of delivery with actual birthweights. This timing within a week of delivery has been shown to be most accurate for predicting estimated fetal weight (EFW).¹, ² Although previous studies have shown that estimated fetal weight should be accurate to within 5% of actual birthweight, 10% is considered clinically acceptable,¹, ², ³ and so in this study, variation within 10% was regarded as acceptable in keeping with international standards.

We aimed to identify maternal and technical factors that may influence the accuracy of ultrasound either in overestimating or underestimating fetal weights by more than 10%. Maternal and fetal characteristics and sonographer experience were evaluated.⁴ Characteristics included body mass index (BMI), ethnicity and diabetic status.

Methods

We performed a single centre retrospective study of all singleton pregnancies equal to or greater than 37 weeks. This included all patients who had ultrasound scans at National Women’s Health, Auckland City Hospital between January 2019 and January 2020. Ethics approval was granted by the Research Governance Group for Women and Neonatal Health at Auckland District Health Board. Ultrasound had to have been performed within 7 days of birth¹, ³ and include the national accepted standards of fetal biometry measurements as specified by the Australasian Society for Ultrasound in Medicine using the Hadlock formula for calculating estimated fetal weight. This formula incorporates all four fetal biometric parameters; abdominal circumference (AC), biparietal diameter (BPD), head circumference (HC) and femoral length (FL) and is routinely applied to estimate fetal weight on all machines used at National Women’s Health. Exclusion criteria included fetal demise at time of ultrasound, multiple pregnancies and major fetal anomalies that would potentially affect any of the four biometry measurements (AC, HC, BPD and FL) such as anencephaly, achondroplasia or gastroschisis. Demographic data included maternal age, BMI, ethnicity, diabetes and smoking status. Pregnancy outcome data including gestational age at birth, days between ultrasound and birth, and birthweight were recorded, as was level of sonographer experience.

US fetal weight estimations were compared with infants’ actual recorded weight at delivery. The percentage difference between estimated fetal weight and real birthweight was calculated with the following formula: relative difference % = [(EFW – birthweight)/birthweight] × 100.

Patients were divided into three groups: estimated fetal weights that were < 90% underestimation of birthweight, estimated fetal weights that were accurate to within 10% of birthweight, and those EFWs that were > 110% overestimation of birthweight. Demographic, sonographic and pregnancy factors in the < 90% and > 110% groups were compared with those fetuses with accurate EFWs, to identify factors that may distinguish these patients and to identify features that may have contributed to ultrasound inaccuracy. This included identifying any relationship between estimated fetal weight variance and ethnicity, BMI and comorbidities such as diabetes or smoking status. In addition, fetuses whose actual birthweights were classified as small for gestational age (<2500 g) and large for gestational age (>4000 g) were assessed to identify how many of these were incorrectly categorised due to inaccurately estimated fetal weights.

Demographics and ultrasound data are presented as mean ± standard deviation, median (interquartile range) and number (percentage). Statistical analyses were performed with Stata 13 (StataCorp LP, College Station, TX, USA) and GraphPad (GraphPad Software, San Diego, California, USA). Groups were compared using Student’s t‐test, ANOVA and chi^‐square. Multivariate analysis was performed by logistic regression with stepwise inclusion of the variables with the strongest association and the outcome of interest. P < 0.05 was considered statistically significant.

Results

Of 719 fetuses scanned within a 12 month period from January 2019 to January 2020, 159 were excluded (158 were less than 37 weeks, and one fetal demise was diagnosed at time of ultrasound), leaving 560 fetuses included in the analysis.

Patient demographics and gestational characteristics are shown in Table 1. The mean gestational age was 38.8 weeks, with an average of 4 days between ultrasound scan and birth date. Mean maternal age was 31 years, with an average BMI of 27 kg/m². The predominant ethnicity was Asian (40%), with NZ European/European just under one third (29.5%), and Maori 8%. Most women were non‐smokers. Sonographer experience varied, with 58% identified as experienced.

Table 1.

Maternal demographics and gestational information.

	N = 560
Maternal age (years)	31 ± 5.4
Maternal BMIa	27 ± 7
Maternal ethnicity
Asian	226 (40%)
NZ European/European	165 (29.5%)
Pacific Peoples	107 (19%)
Maori	45 (8%)
Other	17 (3%)
Maternal diabetes (pre‐existing and gestational)
Present	133 (23.8%)
Absent	427 (76.2%)
Smoking status
No	519 (92.7%)
Yes	41 (7.3%)
Sonographer experience
Inexperienced (<5 years)	233 (41.6%)
Experienced (>5 years)	327 (58.4%)
Time between scan and delivery (days)	4 ± 1.9
Gestational age (weeks)	38.8 ± 1.2
Scan weight (grams ± SD)	3365 ± 565
Birthweight (grams ± SD)	3240 ± 541
Difference EFWb vs BWc (%)	4 ± 8

^aBody mass index.

^bEstimated fetal weight.

^cBirthweight.

There were 135 fetuses (24.1%) for whom birthweight was overestimated or underestimated by more than 10% by fetal ultrasound (Table 2).

Table 2.

Factors associated with error outside ± 10% of estimated fetal weights.

	<90% of BW N = 19	Within ± 10% N = 425	>110% of BW N = 116
Birth weight (g)	3358 ± 523	3271 ± 551	3106 ± 484*
Time from scan to delivery (days)	5 ± 2*	4 ± 1.9	4 ± 1.8
EFWa (g)	2904 ± 514*	3327 ± 551	3575 ± 556*
Gestational age (weeks)	38.4 ± 1.2	38.7 ± 1.2	39 ± 1.2*
Maternal age (years)	30.7 ± 5.1	30.9 ± 5.4	32.4 ± 5.3*
Maternal BMIb	32 ± 9*	27 ± 7	26 ± 7
Sonographer experience (years)	13 ± 10	12 ± 10	11 ± 10
Inexperienced (<5years)	7 (37%)	176 (41%)	50 (43%)
Experienced (>5years)	12 (63%)	249 (59%)	66 (57%)
Ethnicity
Asian	4 (1.8%)	171 (75.7%)	51 (22.6%)
Maori	1 (2.2%)	35 (77.8%)	9 (20.0%)
NZ Euro	6 (3.6%)	124 (75.2%)	35 (21.2%)
Other	2 (11.8%)	13 (76.5%)	2 (11.8%)
Pacific	6 (5.6%)	82 (76.6%)	19 (17.8%)
Change in predicted birthweight category		**
Up	4 (21.1%)	17 (79.0%)	0 (0%)
No change	15 (79.0%)	384 (77.6%)	90 (77.6%)
Down	0 (0%)	24 (90.4%)	26 (22.4%)

^aEstimated fetal weight.

^bBody mass index.

^*P < 0.05 vs. EFW within 10%; **P < 0.0001 for trend.

Overall, there were 71 out of 560 fetuses (12.7%) for whom birthweight categories were inaccurately estimated by ultrasound (Table 3). This included 21 with overestimated birth weights (3.75%). In this group, fetuses that should have been categorised as SGA were incorrectly labelled normal (n = 8, 1.4%) or LGA (n = 1, 0.2%), or those with normal birthweights were labelled LGA (n = 12, 2.1%). In 50 fetuses (8.9%), the EFW was underestimated, either for fetuses that should have been categorised as LGA labelled normal (n = 32, 5.7%) or SGA (n = 1, 0.2%), or normal weight fetuses labelled as SGA (n = 17, 3.0%). One fetus who was SGA at birth was labelled LGA at ultrasound, and one fetus who was LGA at birth was labelled SGA by EFW. Inaccurate weight classification by US was more than twice as likely to be underestimated (8.9%) as overestimated (3.8%).

Table 3.

Sonographically predicted weight category versus birth weight category.

		Birth weight classification
		SGA N = 43	Normal N = 464	LGA N = 53	Total
Sonographic EFWa classification	SGAb	25 (58.1%)	8	1	34
	Normal	17	424 (91.4%)	12	453
	LGAc	1	32	40 (75.5%)	73
	Total	43	464	53	560

Grey‐shaded cells correspond to fetuses with concordant sonographic EFW and birthweight categories. Blue‐shaded cells correspond to fetuses where the sonographic EFW underestimated the birthweight category. Red‐shaded cells correspond to fetuses where the sonographic EFW overestimated the birthweight category.

^aEstimated fetal weight.

^bSmall for gestational age (<2500 g).

^cLarge for gestational age (>4000 g).

Features associated with sonographic EFW

Variables associated with inaccurate sonographic measurements and the impact on birth weight categories are presented in Tables 2, 4 and 5.

Table 4.

Factors associated with errors in change of birthweight category.

	Underestimate of category N = 50	No change N = 489	Overestimate of category N = 21
Birthweight (g)	3410 ± 747	3230 ± 517	3270 ± 658
Time from scan to delivery (days)	4.7 ± 2.2	4.0 ± 2.0	4.1 ± 1.8
EFWa (g)	3633 ± 763	3345 ± 525	3182 ± 748*
Gestational age (weeks)	38.1 ± 1.0	38.8 ± 1.2	39.1 ± 1.3
Maternal age (years)	31.1 ± 5.5	31.2 ± 5.4	31.4 ± 5.4
Maternal BMIb	32.5 ± 10.4*	27.2 ± 7.3	26.9 ± 6.7
Sonographer experience (years)	13.4 ± 12.4	11.7 ± 10.3	12.4 ± 11.2
Unexperienced (<5 years)	10 (47.6%)	201 (41.1%)	22 (44.0%)
Experienced (>5 years)	11 (52.4%)	288 (58.9%)	28 (56.0%)

^aEstimated fetal weight.

^bBody mass index.

^*P < 0.001; ^†P < 0.05.

Table 5.

Factors influencing under‐ and overestimate of estimated fetal weight.

	Unadjusted OR (95% CI)	P value	Adjusted OR (95% CI)	P value
Predictors of < 90% underestimation of estimated fetal weight
Time from scan to delivery (days)	1.35 (1.05–1.73)	0.021	1.26 (0.98–1.62)	0.077
EFWa, per 100 g	0.86 (0.78–0.94)	0.001	0.85 (0.77–0.93)	0.001
BMIb, per 5 kg/m2	1.37 (1.06–1.77)	0.018	1.51 (1.14–2.01)	0.004
Predictors of > 110% overestimation of estimated fetal weight
Later GAc, per week	1.29 (1.09–1.52)	0.003	1.06 (0.86–1.31)	0.55
EFW, per 100 g	1.08 (1.04–1.12)	<0.001	1.08 (1.03–1.13)	<0.001
Maternal DMd	0.58 (0.34–0.9)	0.049	0.55 (0.30–1.00)	0.051
Maternal age, per 5 years	1.30 (1.07–1.58)	0.009	1.43 (1.16–1.76)	0.001
Predictors of birthweight category underestimate
BMI, per 5 kg/m2	1.45 (1.14–1.83)	0.002	N/A	N/A
Predictors of birthweight category overestimate
EFW, per 100 g	1.10 (1.04–1.15)	0.001	N/A	N/A

The multivariate models for each contain the variables which were significantly associated with each outcome at univariate analysis.

^aEstimated fetal weight.

^bBody mass index.

^cGestational age.

^dDiabetes mellitus.

Fetuses with EFW < 90% of actual birthweights on average, had longer times between scanning and delivery. This underestimation was also more common in those fetuses with higher EFW and in mothers with higher BMI. Overestimation of fetal weight was more likely to occur in fetuses with smaller average EFW and birthweights, later gestational age and older maternal age. We found no relationship with maternal diabetes.

Impact of inaccurate EFW on birthweight categories

Underestimating birthweight category was most commonly associated with higher maternal BMI. Overestimating birthweight category was more common in fetuses at the upper end of the EFW spectrum and was not associated with BMI or gestational age. There was no significant difference in ethnicity, in particular Maori and Pacific Island groups were comparably represented within both the over and underestimated subgroups both in terms of estimated fetal weight and category change (Table 2).

Multivariate analysis showed independent predictors for underestimating fetal weight were lower EFW and higher maternal BMI (Table 5). Independent predictors of overestimating fetal weight were higher EFW and older maternal age.

For under and overestimation of birthweight category, only a single factor was identified for each, so a multivariate analysis was not performed.

The area under receiver operating curves for the multivariate models predicting under‐ and overestimation of estimated fetal weight were 0.77 (95% CI 0.65 to 0.88) and 0.67 (95% CI 0.62 to 0.73), respectively. The area under the receiver operating curves for the multivariate model predicting class underestimate was 0.66 (95% CI 0.51 to 0.81), and for the univariate model predicting class overestimate was 0.63 (95% CI 0.52 to 0.75).

Discussion

Although the accuracy of ultrasound fetal weight estimations has been examined in other countries,¹, ², ³ there has been limited assessment for the Australasian population. The cohort of patients in this study were from National Women’s Health department which is a tertiary referral centre for more than a third of the population of New Zealand.

The proportion of fetal weight estimations at our institution that were outside the acceptable 10% variation was similar to previous studies with just over three out of four (75.9%) being within 10% of the recorded birth weight at delivery.¹, ², ⁵, ⁶, ⁷, ⁸ A retrospective audit from Wellington Hospital in 2006 also reported an accuracy of 75%.⁷ They demonstrated similar rates of under‐ and overestimation (12.7% and 12.6%) but their population demographics were not included which limits comparison with our study. Other studies prior to 2006 reported fetal weight estimation accuracies that varied between 23 and 78%,⁶ but few of these distinguished between overestimated and underestimated group features. More recent studies from other countries have similar degrees of accuracy to ours but they have different population demographics such as lower mean maternal ages, weights and normal range BMI,⁵, ⁹ compared with our average BMI of 27kg/m² which is classified as overweight.

We found that accuracy decreased at the extremes of fetal weight in both SGA and LGA babies, as other authors have described.², ⁸ Specifically, US tended to overestimate the weight of SGA fetuses and underestimate the birthweights predicted for LGA fetuses, findings which agree with those of a systematic review by Dudley et al. who also found systematic and random errors may be larger for small fetuses and that higher birthweight is underestimated.² Such errors can result in increased adverse outcomes at birth by under recognition of both SGA and LGA fetuses. Underestimating birthweight category leads to under recognition of LGA age babies who are at higher risk for perinatal morbidity, with the increased risk of asphyxia due to prolonged delivery, shoulder dystocia, brachial plexus injury as well as maternal adverse outcomes such as increased rates of emergency caesarean section, perineal injuries and post‐partum haemorrhage associated with birth. Our group of SGA fetuses in whom US overestimated both the EFWs and categories in 18 (40% of SGA by birthweight) raises even more concern, a trend also demonstrated by Sharma et al.⁵ This kind of overestimation can result in failure to recognise SGA babies who are at increased risk of still birth. These fetuses can miss out on appropriate referral to the multidisciplinary team which would consider care involving closer monitoring with regular Doppler US, the need for steroid preparation, consideration for early delivery that may be necessary for survival and neonatal specialist support at delivery. The relationship between maternal age and overestimating fetal weight is challenging to explain. This may reflect the relatively small size of this group.

We found that higher maternal BMI and longer times between scanning and delivery were the most common risk factors for underestimating fetal weight. It is understandable that increased time between scanning and delivery would cause EFWs to underestimate birth weight given the rate at which the fetus can grow in the last week of gestation. The influence of increased BMI on accuracy has been studied with variable results. Recent studies performed within 7 days of delivery such as ours have shown decreasing accuracy with increasing maternal BMI as we have demonstrated.⁹, ¹⁰, ¹¹, ¹² Aksoy et al.¹² studied 347 full‐term patients and demonstrated progressive decrease in accuracy as BMI increased. The average BMI of our population would have been categorised as obese in their study. They postulated that decreased accuracy in patients with greater BMIs was due to compromised visualisation of fetal anatomy with subsequent effect on fetal biometry. However, they did not separate factors associated with over and underestimation of birthweights as we have done. Other studies that found no relationship with BMI included fetuses of all gestations which probably influenced their findings.¹³, ¹⁴

Accurate estimation of fetal weight and weight category is even more critical in obese women to guide antenatal management and delivery, as they are at a significantly increased risk of perioperative morbidity with caesarean section, including bleeding, infection and thromboembolism, which needs to be balanced with the risk of shoulder dystocia and vaginal delivery. The impact of obesity on individual biometric measurements needs further assessment.

One of the limitations of our study is that although the study cohort was relatively large, the groups with EFWs outside 10% were relatively small, especially that of fetuses whose birthweight category was overestimated. These smaller sizes limit the statistical power for detecting significant relationships. A second possible limitation is that we only assessed accuracy using the Hadlock formula (AC, BPD, HC, FL) with no comparison with other formulae. Westerway et al evaluated the accuracy of a variety of formulae in 121 pregnancies and warned that BPD in late pregnancy may be underestimated due to moulding of the skull¹⁵; however, a more recent review by Milner et al.³ found that the Hadlock formula (AC, BPD, HC, FL) produced the most consistent mean systematic error and lowest random error. Furthermore, because this was a retrospective study, there were some limitations in the sonographic data which we could evaluate. Additional features such as fetal position at time of ultrasound, liquor volume and placental position were not recorded in the database but may be important in determining sonographic image quality and hence reliability of biometric measurements. We also did not examine delivery notes to assess adverse consequences that may have related to US inaccuracy of EFW categorisation. We propose further prospective analysis to include features such as individual biometric parameters, fetal position and liquor volume, any of which may influence the accuracy of US in predicting estimated fetal weight in the final gestational week.

Conclusion

We found ultrasound prediction of EFWs to be more than 10% outside actual birthweights in almost one quarter of this population, resulting in inaccurate birthweight category estimation in just over 10% of fetuses. This inaccuracy should alert our sonographers and obstetricians to potential errors at the extremes of fetal birthweights, particularly in mothers with increased BMI.

Authorship declaration

We acknowledge that the authorship listing conforms with the journal’s authorship policy, and that all authors are in agreement with the content of the submitted manuscript.

Conflict of interest

No financial support or conflict of interest relationships to declare.

Funding

No funding information is provided.

Author Contribution

Sarah Benson‐Cooper: Conceptualization (supporting); Data curation (equal); Formal analysis (supporting); Methodology (supporting); Project administration (lead); Resources (lead); Writing‐original draft (lead). Gregory P Tarr: Formal analysis (lead); Methodology (equal); Software (lead); Writing‐original draft (supporting). Colleen J Bergin: Conceptualization (lead); Data curation (equal); Formal analysis (supporting); Project administration (supporting); Supervision (lead); Writing‐original draft (supporting). Joanne Kelly: Conceptualization (equal); Data curation (equal); Methodology (supporting).