1992,
The British Journal of Radiology, 66, 987-989
Prediction of birthweight by fetal ultrasound biometry By D. A. Nzeh, MBBS, FMCR, S. Rimmer, FRCR, *W. M. 0 . Moore, FRCOG and tl_. Hunt, PhD Departments of Radiology, "Obstetrics and Gynaecology and tFaculty of Medicine and Computational Group, St. Mary's Hospital for Women and Children, Manchester, UK. (Received 7 August 1991 and in revised form 3 December 1991, accepted 3 June 1992) Keywords: Birthweight, Fetal biometry
Abstract. A total of 104 women with singleton pregnancies who were delivered between 37 and 42 weeks gestation had ultrasound scans during the fortnight before delivery. The biparietal diameter (BPD), abdominal circumference (AC) and femur length (FL) were measured in all cases. Estimation of fetal weight (EFW) was done by four different methods: using AC alone, AC/BPD, AC/FL and AC/BPD/FL. Results were compared with values of actual birthweights at delivery. There was no significant difference between the mean birthweights of the 47 boy and 57 girl fetuses studied. The EFW(Shepard) method showed the least bias overall: mean percentage error 1.7%, standard deviation (SD) 10.6%. The other three methods significantly underestimated birthweights on average: EFW(Deter), mean error 2.2%, SD 9.3%, p < 0.02; E F W ^ ^ p , ^ , mean error 5.4%, SD 9.5%, p < 0.001; EFW(Hadlock), mean error 5.6%, SD 9.3%, p < 0.001. The percentage error in each group was significantly negatively correlated (p < 0.001) with the scan-delivery interval. Two new equations were generated which gave more accurate predictions for the cases under study using AC, BPD and FL as a combination and also in addition to scan-delivery interval (SDI) in days.
Several methods are now available for estimation of birthweight using fetal measurements obtained by ultrasound. Initial approach to fetal weight estimation utilized measurements of abdominal circumference (AC) alone (Campbell & Wilkin, 1975). Different combinations of fetal parameters have also been employed: AC has been used in conjunction with biparietal diameter (BPD) (Shepard et al, 1982), AC and femur length (FL) have also been combined (Hadlock et al, 1984). The three commonly measured fetal parameters of AC, BP*D and FL have also been applied together (Hadlock et al, 1985). Accurate estimation of fetal weight (EFW) is useful in successful management of labour and care of the newborn baby in the early days of life. Birth trauma, perinatal asphyxia and maternal morbidity are some of the problems associated with vaginal delivery of the macrosomic fetus (Hirata et al, 1990). The aim of this study was to evaluate the accuracy of fetal biometry in the prediction of birthweight by means of formulae currently in use and to see whether a more reliable formula could be produced using AC, BPD and FL combined and in addition to scan-delivery interval (SDI). Patients and methods
One hundred and four women who had singleton pregnancies and were delivered between 37 and 42 weeks gestation had ultrasound scans done from day 0 to 15 prior to delivery. The age range of cases in the study was 16-42 years with a mean of 27.3 years. This paper was presented in part at the 48th annual congress of the British Institute of Radiology, Harrogate, 11-13 June 1990. Address for correspondence: Dr D. A. Nzeh, Department of Radiology, University of Ilorin, Ilorin, Nigeria. Vol. 65, No. 779
AC, BPD and FL measurements were obtained in cm in all cases. Fetal weight (EFW) was estimated using the following formulae and recorded in grams. /
(Campbell) (Sh
EFW C rC W C W
d)
_ inl-25-0.00265BPDxAC = 1U
=
io1-3598~ooo37ACxFL+OO51AC+ol844FL
^ — inl-335-0.OO34ACxFL + 0.0316BPD + 0.0457AC + (Deter) ~ 1 U
Results of EFW obtained using the above formulae were compared with actual birthweights. There were 47 boy and 57 girl fetuses in the series. Results The range of actual birthweights was 1940-4700 g, mean 3194 g, SD 499 g. There was no statistically significant difference between the mean birthweights of the male and female fetuses. EFW(Shepard) showed the least bias overall (mean error 1.7%, SD 10.6%). The other three methods significantly underestimated birthweights on average: EFW,(Deter) mean - 2 . 2 % , SD 9.3%, p < 0.02, EFW(Campbell) mean - 5 . 4 % , SD 9.5%, p < 0.001, EFW(Hadlock) mean - 5 . 6 % , SD 9.3%, p < 0.001. The percentage error in each group was significantly negatively correlated (p < 0.001) with the scan-delivery interval. For intervals up to five days, the Deter and the Shepard methods significantly overestimated the birthweight (mean 2.9%, SD 9.1%, p < 0.05 and mean 6.7%, SD 12.0%, p < 0.001 respectively) whereas the other two methods had negligible bias (Campbell: mean - 0 . 3 % , SD 9.3%; Hadlock: mean - 0 . 6 % , SD 8.4%). EFW(Shepard) was the most accurate in the 6-10 days interval (mean - 0 . 5 % , SD 7.6%) while the Hadlock, Campbell and Deter methods significantly underestimated birthweight in this range (mean —7.5%, SD 987
D. A. Nzeh, S. Rimmer, W. M. O. Moore and L. Hunt Actual
3.2
3.5 Predicted
Figure 1. Scatter plot of actual Iog10(birthweight) versus logI0(birthweight) predicted using Equation (1).
7.1%, p< 0.001; mean - 7 . 0 % , SD 7.6%, p< 0.001 and mean - 4 . 1 % , SD 7.0%, p < 0.002 respectively). All four methods significantly underestimated birthweight at intervals of 11 days or more: EFW(Shepard) mean - 3 . 8 % , SD 8.1%, p < 0.05; EFW(Deter) mean - 8 . 1 % , SD 8.3%, p< 0.001; EFW(Campbell) mean - 1 1 . 7 % , SD 7.7%, p< 0.001; and EFW(Hadlock) mean - 1 1 . 7 % , SD 9.1%, p< 0.001. Using linear regression analysis, two new equations were generated which gave closer predictions for the cases under study using AC, BPD and FL in combination and in addition to SDI as follows: Iog10(birthweight) = 0.470+ 0.4881og10BPD + 0.5541og10FL+1.3771og10AC (1) Iog10(birthweight) = 0.326+ 0.00451 (SDI) + 0.3831og,0BPD + 0.6141og10FL+1.4851og10AC (2) The two new equations for predicting birthweight gave negligible bias when subjected to the same statistical analysis as the four equations under review. Figure 1 is a scatter plot of actual Iog10(birthweight) versus predicted Iog10(birthweight) using Equation (1) and shows close correlation between predicted and actual values.
The accuracy of a given model in predicting birthweight decreases further as the population deviates further from that used to generate such a model (Hirata et al, 1990). It is therefore important that each institution should determine which formula best suits its cases, especially when no model has yet been generated locally. In the present study, EFW(Shepard) showed the least bias overall. This finding is at variance with reports from some other studies which concluded that Shepard's model, which combined AC and BPD, was less accurate than other formulae which used AC and FL (Miller, 1988). Although it has been claimed in the past that there is cessation of fetal growth at 38 weeks gestation (Rossavik et al, 1988), results from the current series did not support this observation and showed that fetal growth continued until term. When the two new equations were subjected to the same statistical analysis as the four equations under review, the newly generated equations gave more accurate results. The lapse time or scan-delivery interval (SDI) has been noted as a critical element in the accuracy of remote estimates of fetal weight (Spinnato et al, 1988). Findings from the present report corroborate this view when the newly generated equation from this study, which incorporated SDI, is applied. It is interesting that all the four formulae under review underestimated birthweigrit at intervals of 11 days or more because the daily incremental gain in fetal weight during the last two weeks of gestation was not taken into consideration in developing these equations. The formulae found most useful for prediction of birthweight among the cases studied were the newly generated equations and they are recommended for use, especially when applied in conjunction with SDI. References CAMPBELL, S. & WILKIN, D., 1975. Ultrasonic measurement of
fetal abdomen circumference in the estimation of fetal weight. British Journal of Obstetrics and Gynaecology, 82, 689-697. HADLOCK, F. P., HARRIST, R. B., CARPENTER, R. J., DETER,
R. L. & PARK, S. K., 1984. Sonographic estimation of fetal weight. Radiology, 150, 535-540. HADLOCK, F. P., HARRIST, R. B., SHARMAN, R. S., DETER, R. L.
Discussion Comparison of fetal weight models from different studies has shown that the use of multiple parameters, and in particular the combination of head, abdomen and femur length measurements, provide the most adequate estimations of fetal weight with 95% confidence interval in the range of 15-16% (Hadlock, 1990). Studying the weight estimation for fetuses small and large for their gestational age, it was observed that only formulae dependent on femur length (FL) fitted both groups well (Miller et al, 1988). Abdominal diameter measurement has been found to be less accurate in cases of oligohydramnios because the fetal skin edge may be difficult to identify when liquor volume is diminished (Valea et al, 1990). 988
& PARK, S. K., 1985. Estimation of fetal weight with the use of head, body and femur measurements—a prospective study. American Journal of Obstetrics and Gynecology, 151, 333-337. HADLOCK, F. P., 1990. Sonographic estimation of fetal weight. Radiologic Clinics of North America, 28, 47-50. HIRATA, G. I., MEDEARIS, A. L., HORENSTEIN, J., BEAR, M. B. &
PLATT, L. D., 1990. Ultrasonographic estimation of fetal weight in the clinically macrosomic fetus. American Journal of Obstetrics and Gynecology, 162, 238-242. MILLER, J. M., KISSLING, G. A., BROWN, H. L. & GABERT,
H. A., 1988. Estimated fetal weight: applicability to smalland large-for-gestational-age fetus. Journal of Clinical Ultrasound, 16, 95-97. ROSSAVIK, I. K., DETER, R. L. & WASSERSTRUM, N.,
1988.
Mathematical modeling of fetal growth v. fetal weight changes at term. Journal of Clinical Ultrasound, 16, 9-15. The British Journal of Radiology, November 1992
Prediction of birthweight by fetal ultrasound biometry SHEPARD, M. J., RICHARDS, V. A., BERKOWITZ, A. L., WARSOP,
VALEA, F. A., WATSON, W. J. & SEEDS, J. W., 1990. Accuracy of
S. L. & HOBBINS, J. C , 1982. An evaluation of two equations for predicting fetal weight by ultrasound. American Journal of Obstetrics and Gynecology, 142, 47-54.
ultrasonic weight prediction in the fetus with preterm premature ruptures of membranes. Obstetrics and Gynecology, 75, 183-185.
SPINNATO, J. A., ALLEN, E. D. & MENDENHALL, H. W.,
1988.
Birthweight prediction from remote ultrasound examination. Obstetrics and Gynecology, 71, 893-898.
Vol. 65, No. 779
989
The British Journal of Radiology, 66, 987-989
Prediction of birthweight by fetal ultrasound biometry By D. A. Nzeh, MBBS, FMCR, S. Rimmer, FRCR, *W. M. 0 . Moore, FRCOG and tl_. Hunt, PhD Departments of Radiology, "Obstetrics and Gynaecology and tFaculty of Medicine and Computational Group, St. Mary's Hospital for Women and Children, Manchester, UK. (Received 7 August 1991 and in revised form 3 December 1991, accepted 3 June 1992) Keywords: Birthweight, Fetal biometry
Abstract. A total of 104 women with singleton pregnancies who were delivered between 37 and 42 weeks gestation had ultrasound scans during the fortnight before delivery. The biparietal diameter (BPD), abdominal circumference (AC) and femur length (FL) were measured in all cases. Estimation of fetal weight (EFW) was done by four different methods: using AC alone, AC/BPD, AC/FL and AC/BPD/FL. Results were compared with values of actual birthweights at delivery. There was no significant difference between the mean birthweights of the 47 boy and 57 girl fetuses studied. The EFW(Shepard) method showed the least bias overall: mean percentage error 1.7%, standard deviation (SD) 10.6%. The other three methods significantly underestimated birthweights on average: EFW(Deter), mean error 2.2%, SD 9.3%, p < 0.02; E F W ^ ^ p , ^ , mean error 5.4%, SD 9.5%, p < 0.001; EFW(Hadlock), mean error 5.6%, SD 9.3%, p < 0.001. The percentage error in each group was significantly negatively correlated (p < 0.001) with the scan-delivery interval. Two new equations were generated which gave more accurate predictions for the cases under study using AC, BPD and FL as a combination and also in addition to scan-delivery interval (SDI) in days.
Several methods are now available for estimation of birthweight using fetal measurements obtained by ultrasound. Initial approach to fetal weight estimation utilized measurements of abdominal circumference (AC) alone (Campbell & Wilkin, 1975). Different combinations of fetal parameters have also been employed: AC has been used in conjunction with biparietal diameter (BPD) (Shepard et al, 1982), AC and femur length (FL) have also been combined (Hadlock et al, 1984). The three commonly measured fetal parameters of AC, BP*D and FL have also been applied together (Hadlock et al, 1985). Accurate estimation of fetal weight (EFW) is useful in successful management of labour and care of the newborn baby in the early days of life. Birth trauma, perinatal asphyxia and maternal morbidity are some of the problems associated with vaginal delivery of the macrosomic fetus (Hirata et al, 1990). The aim of this study was to evaluate the accuracy of fetal biometry in the prediction of birthweight by means of formulae currently in use and to see whether a more reliable formula could be produced using AC, BPD and FL combined and in addition to scan-delivery interval (SDI). Patients and methods
One hundred and four women who had singleton pregnancies and were delivered between 37 and 42 weeks gestation had ultrasound scans done from day 0 to 15 prior to delivery. The age range of cases in the study was 16-42 years with a mean of 27.3 years. This paper was presented in part at the 48th annual congress of the British Institute of Radiology, Harrogate, 11-13 June 1990. Address for correspondence: Dr D. A. Nzeh, Department of Radiology, University of Ilorin, Ilorin, Nigeria. Vol. 65, No. 779
AC, BPD and FL measurements were obtained in cm in all cases. Fetal weight (EFW) was estimated using the following formulae and recorded in grams. /
(Campbell) (Sh
EFW C rC W C W
d)
_ inl-25-0.00265BPDxAC = 1U
=
io1-3598~ooo37ACxFL+OO51AC+ol844FL
^ — inl-335-0.OO34ACxFL + 0.0316BPD + 0.0457AC + (Deter) ~ 1 U
Results of EFW obtained using the above formulae were compared with actual birthweights. There were 47 boy and 57 girl fetuses in the series. Results The range of actual birthweights was 1940-4700 g, mean 3194 g, SD 499 g. There was no statistically significant difference between the mean birthweights of the male and female fetuses. EFW(Shepard) showed the least bias overall (mean error 1.7%, SD 10.6%). The other three methods significantly underestimated birthweights on average: EFW,(Deter) mean - 2 . 2 % , SD 9.3%, p < 0.02, EFW(Campbell) mean - 5 . 4 % , SD 9.5%, p < 0.001, EFW(Hadlock) mean - 5 . 6 % , SD 9.3%, p < 0.001. The percentage error in each group was significantly negatively correlated (p < 0.001) with the scan-delivery interval. For intervals up to five days, the Deter and the Shepard methods significantly overestimated the birthweight (mean 2.9%, SD 9.1%, p < 0.05 and mean 6.7%, SD 12.0%, p < 0.001 respectively) whereas the other two methods had negligible bias (Campbell: mean - 0 . 3 % , SD 9.3%; Hadlock: mean - 0 . 6 % , SD 8.4%). EFW(Shepard) was the most accurate in the 6-10 days interval (mean - 0 . 5 % , SD 7.6%) while the Hadlock, Campbell and Deter methods significantly underestimated birthweight in this range (mean —7.5%, SD 987
D. A. Nzeh, S. Rimmer, W. M. O. Moore and L. Hunt Actual
3.2
3.5 Predicted
Figure 1. Scatter plot of actual Iog10(birthweight) versus logI0(birthweight) predicted using Equation (1).
7.1%, p< 0.001; mean - 7 . 0 % , SD 7.6%, p< 0.001 and mean - 4 . 1 % , SD 7.0%, p < 0.002 respectively). All four methods significantly underestimated birthweight at intervals of 11 days or more: EFW(Shepard) mean - 3 . 8 % , SD 8.1%, p < 0.05; EFW(Deter) mean - 8 . 1 % , SD 8.3%, p< 0.001; EFW(Campbell) mean - 1 1 . 7 % , SD 7.7%, p< 0.001; and EFW(Hadlock) mean - 1 1 . 7 % , SD 9.1%, p< 0.001. Using linear regression analysis, two new equations were generated which gave closer predictions for the cases under study using AC, BPD and FL in combination and in addition to SDI as follows: Iog10(birthweight) = 0.470+ 0.4881og10BPD + 0.5541og10FL+1.3771og10AC (1) Iog10(birthweight) = 0.326+ 0.00451 (SDI) + 0.3831og,0BPD + 0.6141og10FL+1.4851og10AC (2) The two new equations for predicting birthweight gave negligible bias when subjected to the same statistical analysis as the four equations under review. Figure 1 is a scatter plot of actual Iog10(birthweight) versus predicted Iog10(birthweight) using Equation (1) and shows close correlation between predicted and actual values.
The accuracy of a given model in predicting birthweight decreases further as the population deviates further from that used to generate such a model (Hirata et al, 1990). It is therefore important that each institution should determine which formula best suits its cases, especially when no model has yet been generated locally. In the present study, EFW(Shepard) showed the least bias overall. This finding is at variance with reports from some other studies which concluded that Shepard's model, which combined AC and BPD, was less accurate than other formulae which used AC and FL (Miller, 1988). Although it has been claimed in the past that there is cessation of fetal growth at 38 weeks gestation (Rossavik et al, 1988), results from the current series did not support this observation and showed that fetal growth continued until term. When the two new equations were subjected to the same statistical analysis as the four equations under review, the newly generated equations gave more accurate results. The lapse time or scan-delivery interval (SDI) has been noted as a critical element in the accuracy of remote estimates of fetal weight (Spinnato et al, 1988). Findings from the present report corroborate this view when the newly generated equation from this study, which incorporated SDI, is applied. It is interesting that all the four formulae under review underestimated birthweigrit at intervals of 11 days or more because the daily incremental gain in fetal weight during the last two weeks of gestation was not taken into consideration in developing these equations. The formulae found most useful for prediction of birthweight among the cases studied were the newly generated equations and they are recommended for use, especially when applied in conjunction with SDI. References CAMPBELL, S. & WILKIN, D., 1975. Ultrasonic measurement of
fetal abdomen circumference in the estimation of fetal weight. British Journal of Obstetrics and Gynaecology, 82, 689-697. HADLOCK, F. P., HARRIST, R. B., CARPENTER, R. J., DETER,
R. L. & PARK, S. K., 1984. Sonographic estimation of fetal weight. Radiology, 150, 535-540. HADLOCK, F. P., HARRIST, R. B., SHARMAN, R. S., DETER, R. L.
Discussion Comparison of fetal weight models from different studies has shown that the use of multiple parameters, and in particular the combination of head, abdomen and femur length measurements, provide the most adequate estimations of fetal weight with 95% confidence interval in the range of 15-16% (Hadlock, 1990). Studying the weight estimation for fetuses small and large for their gestational age, it was observed that only formulae dependent on femur length (FL) fitted both groups well (Miller et al, 1988). Abdominal diameter measurement has been found to be less accurate in cases of oligohydramnios because the fetal skin edge may be difficult to identify when liquor volume is diminished (Valea et al, 1990). 988
& PARK, S. K., 1985. Estimation of fetal weight with the use of head, body and femur measurements—a prospective study. American Journal of Obstetrics and Gynecology, 151, 333-337. HADLOCK, F. P., 1990. Sonographic estimation of fetal weight. Radiologic Clinics of North America, 28, 47-50. HIRATA, G. I., MEDEARIS, A. L., HORENSTEIN, J., BEAR, M. B. &
PLATT, L. D., 1990. Ultrasonographic estimation of fetal weight in the clinically macrosomic fetus. American Journal of Obstetrics and Gynecology, 162, 238-242. MILLER, J. M., KISSLING, G. A., BROWN, H. L. & GABERT,
H. A., 1988. Estimated fetal weight: applicability to smalland large-for-gestational-age fetus. Journal of Clinical Ultrasound, 16, 95-97. ROSSAVIK, I. K., DETER, R. L. & WASSERSTRUM, N.,
1988.
Mathematical modeling of fetal growth v. fetal weight changes at term. Journal of Clinical Ultrasound, 16, 9-15. The British Journal of Radiology, November 1992
Prediction of birthweight by fetal ultrasound biometry SHEPARD, M. J., RICHARDS, V. A., BERKOWITZ, A. L., WARSOP,
VALEA, F. A., WATSON, W. J. & SEEDS, J. W., 1990. Accuracy of
S. L. & HOBBINS, J. C , 1982. An evaluation of two equations for predicting fetal weight by ultrasound. American Journal of Obstetrics and Gynecology, 142, 47-54.
ultrasonic weight prediction in the fetus with preterm premature ruptures of membranes. Obstetrics and Gynecology, 75, 183-185.
SPINNATO, J. A., ALLEN, E. D. & MENDENHALL, H. W.,
1988.
Birthweight prediction from remote ultrasound examination. Obstetrics and Gynecology, 71, 893-898.
Vol. 65, No. 779
989
