The effect of body mass index on three methods of fetal weight estimation

Authors


*Dr T. Farrell, Department of Obstetrics and Gynaecology, Rm. 125, Level 4, New Jessop Wing, Tree Root Walk, Sheffield, S10 2SF, UK.

Abstract

Objective To assess agreement between three methods of estimation of fetal weight and determine the influence of maternal obesity.

Design Prospective observational study.

Setting A tertiary referral teaching hospital.

Population Unselected women attending for induction of labour.

Method Maternal, clinical and ultrasonic estimations of fetal weight were made prior to induction of labour in 96 consenting women. The estimations of fetal weight were performed independently by the three methods.

Main outcome measure Bland and Altman plots to show limits of agreement, and intraclass correlation coefficients.

Results Variable levels of accuracy were obtained for maternal, clinical and ultrasound estimates of fetal weight. Ultrasound estimation of fetal weight performed equally best for women of high and low body mass index (BMI). For women of low BMI, the intraclass correlation coefficient (ICC) was 0.90 (95% CI 0.83–0.94) and 0.87 (95% CI 0.77–0.93) for women with high BMI. Despite this, the limits of agreement for ultrasound were in the order of −700 to +500 g. Both maternal and clinical estimation of fetal weight under-estimated true fetal weight in women with low BMI and over-estimated the true fetal weight in women with high BMI. The largest observed mean difference was obtained with clinical palpation in both low and high BMI women.

Conclusion The accuracy of ultrasound estimation of fetal weight was better than maternal and clinical estimation of fetal weight and was not influenced significantly by maternal BMI.

Introduction

Obstetric management is often influenced by clinical and ultrasound estimations of fetal weight. Estimated fetal weight (EFW) is a consideration when planning the mode of delivery for a suspected macrosomic or small fetus.

Fetal weight estimations, based on ultrasound biometry measurements using different formulae, have been reported with systematic errors of less than 10% relative to birthweight1,2. There are few reports on the effect of maternal obesity on the validity of estimations of fetal weight. Increasing maternal size represents a major risk factor for the failure of ultrasound to diagnose fetal anomalies3,4. The effect of maternal obesity on clinical and ultrasound estimation of fetal weight is uncertain.

Previous studies1,2, have suggested that maternal obesity does not affect estimations of fetal weight, however, these studies suffer from deficiencies in design and analysis. The aim of this study was to compare maternal, clinical and ultrasound estimations of fetal weight and to determine the effect of maternal body mass index (BMI) on these estimations.

Methods

The study was conducted at a tertiary referral obstetric unit serving an Australasian population. Unselected women attending for induction of labour were recruited following informed consent. Only women with a cephalic singleton pregnancy with intact membranes were considered eligible for inclusion. Women having received a biometry ultrasound scan during the last four weeks were excluded, as were women experiencing contractions or spontaneous rupture of membranes prior to completion of the assessment.

Following consent, the women were asked to give an estimate of fetal weight. Clinical estimation of fetal weight was then undertaken by one of the authors (RH). This entailed a short clinical history and assessment of fetal weight by abdominal palpation only. The short history gave details of reason for induction of labour, past medical history and past obstetric history. The clinician was not aware of the mother's estimate of fetal weight.

All women then underwent an abdominal ultrasound examination performed by two ultrasonographers experienced in assessing fetal biometry. The ultrasound examination included standard measurements of head circumference (HC), abdominal circumference (AC) and femur length (FL). Both AC and HC measurements were obtained by tracing the outer circumference and included the normal soft tissues. EFW was calculated using Hadlock's formula5 [Log10BW = 1.326 − 0.00326(AC)(FL) + 0.0107(HC) + 0.438(AC) + 0.158(FL)].

Delivery details and actual birthweight were later obtained from the case notes and delivery summary. Maternal BMI was calculated by dividing maternal weight at time of induction of labour (kg) by maternal height squared (m2). Maternal BMI at time of induction rather than at booking was used in order to determine the effect of the current BMI on fetal weight estimation and so reflect the difficulties often encountered in clinical practice.

Agreement between the three estimates of fetal weight was assessed using limits of agreement method described by Bland and Altman6 and the intraclass correlation coefficient (ICC)7. For limits of agreement6, a plot of the difference between the true birthweight and the respective method of EFW against the mean values was used to provide a visual assessment of the agreement between the two sets of measurements. The mean difference between the true and estimated measurements is an estimate of the systematic error or bias introduced by the estimated measurements, and the 95% limits of agreement provides a measure of the random error6. Both the mean difference and limits of agreement are plotted on the same graph.

The ICCs were defined as a ratio of the variance due to the true estimates of fetal weight to the total variance. If the variance due to differences between the estimates of fetal weight and the true birthweight is small, the proportion of the total variance due to differences in the true birthweights will be large, and the ICC will be high. It is generally accepted that an arbitary cutoff of >0.75 for the ICC implies good agreement8–10. The 95% confidence interval was estimated using Fisher's transformation11.

Results

Ninety-six women were included in this analysis. All women delivered within 36 hours of assessment. Table 1 demonstrates the descriptive, obstetric and ultrasound data for the study group. The mean BMI for the study was 32.0 (SD 6.3). Forty-two women had a BMI greater than 32 (mean 38.1; SD 4.3) and 54 women had a BMI less than 32 (mean 27.7; SD 2.9). No significant differences were found between the two groups in terms of the descriptive details.

Table 1.  The descriptive maternal, obstetric and ultrasound data for the whole study group and those women with high and low BMI.
Descriptive dataWhole group (n= 96) mean [SD]BMI > 32 (n= 42) mean [SD]BMI < 32 (n= 54) mean [SD]
Age (years)31 [4.9]29.6 [5.8]32 [5.8]
Parity1 [1.2]1.3 [1.5]1 [0.85]
Weight (kg)88 [18.2]103 [14]76 [9.8]
Height (cm)165 [7.1]165 [6.6]165 [7.5]
BMI32 [6.3]38.1 [4.3]27.7 [2.9]
HC (mm)342 [15.6]344 [15.6]341 [15.6]
AC (mm)352 [34]355 [29]351 [37.5]
FL (mm)74 [4.1]75 [3.5]73 [4.4]
Amniotic fluid index (cm)7.1 [4.9]7.1 [5.3]7.1 [4.6]
Caesarean section rate (%)303824

In order to determine the possible influence of BMI on the accuracy of the three methods of estimating fetal weight, the individual discrepancy between the actual birthweight and estimated weight was calculated and the correlation coefficient between the discrepancy and BMI was calculated. The correlation coefficients were 0.049, −0.045 and 0.15 for clinical estimation, maternal estimation and ultrasound estimation of fetal weight, respectively, suggesting poor correlation.

Table 2 demonstrates the mean estimates of fetal weight, the mean discrepancy between each of the three estimates of fetal weight and the birthweight and limits of agreement for the study group as a whole as well as for high and low BMI groups. No significant differences were demonstrated between the maternal, clinical and ultrasound estimates of fetal weight.

Table 2.  The mean birthweight (SD), mean discrepancy between true birthweight and estimated birthweight (SD) and limits of agreement (2SD) for the whole study group and for the high and low BMI groups.
 Whole group (n= 96)BMI > 32 (n= 42)BMI < 32 (n= 54)
 Mean birthweight (g) [SD]Mean true birthweight−estimated weight (g) [SD]Limits of agreement (g)Mean birthweight (g) [SD]Mean true birthweight−estimated weight (g) [SD]Limits of agreement (g)Mean birthweight (g) [SD]Mean true birthweight−estimated weight (g) [SD]Limits of agreement (g)
Maternal estimate3145 [561]−105 [431]−881–7223489 [487]−1.3 [473]−779–7823359 [610]−138 [396]−921–645
Clinical estimate3502 [647]17 [423]−829–8633685 [664]217 [408]−599–10333359 [601]−138 [368]−875–590
Ultrasound estimate3619 [738]134 [314]−495–7643646 [730]178 [326]−465–8393598 [750]100 [305]−504–710

Figures 1–6 demonstrate the limits of agreement between the three methods of estimating fetal weight and birthweight for both the low and high BMI groups. In general, the agreement between ultrasound estimation of fetal weight and birthweight was higher, both in high and low BMI groups, compared with clinical and maternal estimations. Both clinical and maternal estimation tended to under-estimate the true fetal weight in low BMI women, with a mean difference (2SD) of approximately 1500 g. In high BMI women, the tendency was to over-estimate the true fetal weight with clinical estimation, with a mean difference of approximately 1600 g.

Figure 1.

Limits of agreement plot of the difference between the true birthweight and maternal estimate against the mean of their values in women with BMI < 32.

Figure 2.

Limits of agreement plot of the difference between the true birthweight and clinical estimate against the mean of their values in women with BMI < 32.

Figure 3.

Limits of agreement plot of the difference between the true birthweight and ultrasound estimate against the mean of their values in women with BMI < 32.

Figure 4.

Limits of agreement plot of the difference between the true birthweight and maternal estimate against the mean of their values in women with BMI > 32.

Figure 5.

Limits of agreement plot of the difference between the true birthweight and clinical estimate against the mean of their values in women with BMI > 32.

Figure 6.

Limits of agreement plot of the difference between the true birthweight and ultrasound estimate against the mean of their values in women with BMI > 32.

The data were analysed to determine the percentage of fetuses delivered with a birthweight within 10% of the estimated weight. For the whole study group, 61% of clinical estimates, 63% of maternal estimates and 72% of ultrasound estimates were within 10% of the birthweight. In those women with a BMI less than 32, 67% of clinical, 63% of maternal and 74% of ultrasound estimates of fetal weight were within 10% of the birthweight. For those women with a high BMI, 55% clinical, 64% maternal and 64% ultrasound estimates of fetal weight were within 10% of the true birthweight.

The ICCs were all >0.79 for all methods of fetal weight estimation in women with high and low BMI (Table 3). The ICC for ultrasound estimation was 0.9 (95% CI 0.83–0.94) and 0.87 (95% CI 0.77–0.93) in low and high BMI women, respectively, and 0.89 (95% CI 0.88–0.93) for the whole study group. For the other estimates of fetal weight, despite ICC values >0.75, the 95% confidence intervals are invariably wide, reflecting the wide limits of agreement demonstrated in Figs. 1–6 using Bland and Altman's limits of agreement.

Table 3.  The ICC (95% CI) for the whole group and those with low and high BMI.
 Whole group (n= 96) ICC (95% CI)BMI > 32 (n= 42) ICC (95% CI)BMI < 32 (n= 54) ICC (95% CI)
Maternal estimate0.8 (0.71–0.86)0.8 (0.66–0.89)0.8 (0.68–0.88)
Clinical estimate0.8 (0.71–0.86)0.79 (0.64–0.88)0.82 (0.71–0.89)
Ultrasound estimate0.89 (0.88–0.93)0.87 (0.77–0.93)0.90 (0.83–0.94)

Discussion

The internal validity of this study is dependent on the rigour of its design and analysis. We have an accurate assessment of birthweight within 36 hours of assessment and all three observers (mother, obstetrician and ultrasonographer) were blinded. This study therefore fulfils the criteria for the proper design of a study to assess validity in clinical measurements11,12.

With continuous measurements such as birthweight, there are several methods of assessing validity7. We have used Bland and Altman's limits of agreement, and ICC. Limits of agreement is a method of describing the systematic and random difference between the true measurements and the test result. However, this method measures absolute agreement and takes no account of the variance of the true measurements. The ICC, however, provides a numerical estimate of agreement13 and allows for the variance of the true measurements7.

Our study results suggest that maternal, clinical and ultrasound estimation of fetal weight are of variable accuracy. Ultrasound estimation was found to be the better method of fetal weight estimation, using Bland and Altman's limits of agreement and ICC, performing better in women with a low BMI. Despite this, the percentage difference around the mean birthweight for ultrasound estimation ranged from −14% to 20% in women with BMI < 32 and −13% to 24% in women with BMI > 32. The accuracy of all measurements of fetal weight estimation in low BMI women may have been improved if the cutoff for low BMI had been set at the normal range (<25). A BMI of 32 was chosen as the cutoff level because this was the mean BMI for the whole group reflecting the level of obesity in women attending National Women's Hospital, Auckland at the current time.

Field et al.2 demonstrated that approximately two-thirds of the predicted fetal weight were within 10% of the true birthweight and did not decrease with increasing BMI. We have shown similar findings. However, increasing BMI was associated with a reduction in the number of estimates within 10%. The criticism of this kind of analysis is that it fails to allow for the variability of the true measurements unlike the ICC. Our study suggests that the effect of maternal BMI on the three estimates of fetal weight is minimal and is likely of little clinical significance. Despite ultrasound performing best, the limits of agreement for ultrasound are of the order of −500 to 800 g for the high BMI group and −500 to 700 g for the low BMI group. This magnitude of error would seem unacceptable clinically despite the findings of the ICCs, and hence, caution should be used in interpreting the ICCs from this study.

The ability to provide an accurate fetal weight estimation assumes greater importance in those populations with a high prevalence of maternal obesity. The Australasian population studied has a prevalence of maternal obesity (booking weight >100 kg) of 17% and associated increased perinatal mortality of 15.7 and 25.6 for women whose booking weight is 100–124 and 125–149 kg, respectively14. Fetal weight estimations in such women are invariably determined by ultrasound, which therefore plays a major role in the obstetric decision-making.

In conclusion, we have shown that ultrasound fetal weight estimation is currently the most accurate method available in clinical practice for the obese and non-obese pregnant women. Despite this, errors in weight estimation of ±20% are possible and must be borne in mind when decisions regarding obstetric management are formulated.

Acknowledgements

The authors would like to thank the ultrasonographers, Vanessa Engelbrecht and Christine Sharp, for their contribution to the study and Dr P.F.W. Chien for his statistical advice.

Ancillary