Arch Gynecol Obstet DOI 10.1007/s00404-014-3604-y
MATERNAL-FETAL MEDICINE
Sonographic weight estimation in fetal macrosomia: influence of the time interval between estimation and delivery F. Faschingbauer • U. Dammer • E. Raabe • M. Schneider C. Faschingbauer • M. Schmid • A. Mayr • R. L. Schild • M. W. Beckmann • S. Kehl
•
Received: 21 July 2014 / Accepted: 16 December 2014 Springer-Verlag Berlin Heidelberg 2014
Abstract Purpose To evaluate the influence of the time interval between examination and delivery on the accuracy of sonographic weight estimation (WE) in fetal macrosomia. Materials and methods 896 singleton pregnancies (birth weight [ 4,000 g) with a total of 1,281 sonographic weight estimations were included in this retrospective cohort study. Fetuses were divided into six groups with regard to the time interval between estimation and delivery: group 1: scan-to-delivery interval: 0 days; group 2: scan-to-delivery interval: 1–3 days; group 3: scan-to-delivery interval: 4–7 days; group 4: scan-to-delivery interval: 8–14 days; group 5: scan-to-delivery interval: 15–21 days; group 6: scan-to-delivery interval: 22–42 days. The accuracy of WE was compared between five commonly used formulas using Electronic supplementary material The online version of this article (doi:10.1007/s00404-014-3604-y) contains supplementary material, which is available to authorized users. F. Faschingbauer (&) U. Dammer E. Raabe M. Schneider C. Faschingbauer M. W. Beckmann S. Kehl Department of Obstetrics and Gynecology, Erlangen University Hospital, Universita¨tsstrasse 21–23, 91054 Erlangen, Germany e-mail: [email protected] M. Schmid Department of Medical Informatics, Biometry and Epidemiology, Rheinische Friedrich-Wilhelms-University Bonn, Bonn, Germany A. Mayr Department of Medical Informatics, Biometry and Epidemiology, University of Erlangen-Nuremberg, Erlangen, Germany R. L. Schild Department of Obstetrics and Perinatal Medicine, DDH Frauenkliniken, Hanover, Germany
means of percentage errors (MPE), random error, medians of absolute percentage errors (MAPE), and proportions of estimates within 10 % of actual birth weight. Results Significant differences were found between the time interval groups with regard to MAPE and MPE values (p \ 0.001). All formulas showed a systematic underestimation of fetal weight (negative MPEs) (p \ 0.05). MPE values were closest to zero in time interval group 1 and 2. From group 3 to 6, a continuous decrease was observed. The lowest MAPE was found with the Merz formula in group 1 and 2. Values increased continuously from group 3 to 6. Differences between time interval group one and three did not reach statistical significance. Conclusions WE in fetal macrosomia shows the best results when examinations are performed within 7 days before delivery, using the formula of Merz et al. Accuracy significantly decreases after this time period. Keywords Ultrasound Fetal weight estimation Birth weight Time interval Macrosomia
Introduction The incidence of fetal macrosomia has risen continuously in recent years [1, 2]. Antenatal diagnosis is of major importance in the management of labor and delivery and for preventing fetal and maternal trauma during childbirth [3, 4]. Typical risks include a prolonged second stage of labor, serious maternal trauma after vaginal and surgical delivery, increased postpartum hemorrhage and shoulder dystocia with brachial plexus paralysis and/or clavicular fracture [5–7]. During the past 40 years, sonographic assessment of the fetus and estimation of its weight have become part of
123
Arch Gynecol Obstet
routine practice in obstetrics. Several formulas have been published, most of them involving combinations of several biometric parameters [8–13]. However, most of these formulas were derived from heterogeneous groups and were not specifically designed for estimation of fetal weight at the upper end of the weight scale. Several studies have assessed the accuracy of sonographic weight estimation (WE) in macrosomic fetuses and most of them found a significant underestimation of fetal weight with errors exceeding 10 % [8, 14–18]. The majority of these studies analyzed sonograms that were performed close to term within 7 days before delivery [8, 15–18]. However, in numerous patients, fetal WE has either been performed before this time period or is conducted when the patient presents during the latent or active phase of labor. While in the first case interim fetal growth may lead to a further underestimation, ultrasound performed intrapartum may be accompanied by several other problems: first, as delivery approaches, the fetal head descends into the maternal pelvis leading to inaccurate measurements of the biparietal diameter (BPD), occipitofrontal diameter (OFD) and head circumference (HC). Second, an increased risk of abdominal circumference distortion or posterior position of the femora is more frequently observed close to delivery and finally, decreasing amounts of amniotic fluid at term could limit the accuracy of all measurements. So far, no studies have evaluated the influence of the time interval between examination and delivery on the accuracy of sonographic WE in fetal macrosomia. Therefore, the aim of the present study was to compare the accuracy of sonographic fetal WE between six time interval groups ranging from an estimation on the day of delivery to a maximum of 6 weeks, analyzing more than 1,281 biometric measurements of fetuses with a birth weight of more than 4,000 g.
Materials and methods This retrospective, cross-sectional, single-center study included 896 singleton pregnancies with a total of 1,281 sonographic weight estimations over a 7-year period (2003–2009). An automatic query of our perinatal ultrasound database was conducted with the following inclusion criteria: singleton pregnancy with cephalic presentation, a birth weight of more than 4,000 g and an ultrasound examination with complete biometric parameters—BPD, OFD, HC(HC = 2.325 9 ((OFD)2 ? (BPD)2)1/2), abdominal transverse diameter (ATD), abdominal anterior– posterior diameter (APAD), AC(AC = p 9 (ATD ? APD)/2), and femur length (FL) within a time interval between zero and 42 days before delivery; and an absence
123
of any chromosomal or structural anomalies. Intrauterine fetal deaths were excluded. In our institution, a sonographic weight estimation is routinely performed in all patients registering for delivery. Thus, in the vast majority of cases, this was the indication for the examination. Fetuses were divided into six groups with regard to the time interval between estimation and delivery: • • • • • •
Group 1: weight estimation on the day of delivery n = 220. Group 2: time interval between estimation and delivery 1–3 days, n = 426. Group 3: time interval between estimation and delivery 4–7 days, n = 184. Group 4: time interval between estimation and delivery 8–14 days, n = 174. Group 5: time interval between estimation and delivery 15–21 days, n = 101. Group 6: time interval between estimation and delivery 22–42 days, n = 176.
In cases of repeated measurements within one group, only one examination was randomly included. Gestational age was calculated from the last menstrual period and was confirmed by or recalculated with biometric measurements obtained from the first fetal biometry in early pregnancy (in accordance with the recommendations of the American College of Obstetricians and Gynecologists, ACOG) [19]. The examinations were performed by a total of 44 sonographers in accordance with widely accepted quality standards [20, 21]. A variety of ultrasound machines (Siemens Elegra, Siemens Sonoline, G60 S, Siemens Sienna, Siemens Medical Solutions, Erlangen, Germany; GE Voluson 730 Expert, GE Voluson e Laptop, GE Healthcare, Solingen, Germany; Xario and Nemio systems, Toshiba Medical Systems, Neuss, Germany) were used with standard techniques. Birth weight and neonatal length were measured within 1 h after delivery by the nursing staff. For estimation of fetal weight, five widespread formulas (Hadlock et al., Shepard et al., Merz et al., Warsof et al. [9–13]) were used, including the biometric parameters HC, BPD, AC and FL (supplementary Table). Measurements are given in centimeters and BW in grams. In the department in which the study was conducted, fetal weight is routinely measured by ultrasound examination during the diagnostic work-up. Ethical approval for the study was, therefore, not sought. The accuracy of the estimated fetal weight was assessed by calculating (a) the percentage error (PE): (EFW - BW)/ BW 9 100, the mean of which reflects the systematic deviation of a model from the actual BW; (b) the random error (standard deviation of the PE)—a measure of precision that reflects the random component of the prediction error; (c) the absolute percentage error (APE): |(EFW - BW)/
Arch Gynecol Obstet
BW| 9 100, which takes both the systematic and the random error into account; and (d) the percentage of fetal WEs falling within prespecified error bounds (APE \ 10 %). MPEs for all formulas and groups were compared to zero using one-sample t tests. To compare the prediction accuracy between the interval groups, linear mixed-effect regression analysis was performed separately for the different formulas with PEs and APEs as outcome variables and the interval group as predictor variable. Additionally, subject-specific random effects were included to adjust for repeated measurements. Overall group differences were assessed via Likelihood-ratio tests, group-specific differences toward measurements at delivery were assessed via Wald tests for the effect estimates with group 1 as reference. Differences between the formulas inside the interval groups were assessed via paired t tests (PE), the Snedecor and Cochran method (random error), Wilcoxon signedrank tests (APE) and McNemar tests (APE \ 10 %). Linear mixed-effect regression was applied to analyze the combined effects of the time interval and the different formulas. This allows assessing the overall differences between the formulas, adjusted for the time interval and the effect of the time interval for each specific formula. The PEs and APEs from all formulas served as outcome variables, the formula and an interaction between the time interval and
the formula as fixed effects. A subject-specific random effect was included to adjust for repeated measurements. All p values were adjusted for multiple comparisons using the Bonferroni–Holm method.
Results The automatic query of our perinatal data base with the above-mentioned inclusion and exclusion criteria resulted in 9,385 fetuses of the whole weight range, of which 915 fetuses had a birth weight of more than 4,000 g. 19 fetuses had to be excluded due to implausible values because of incorrect documentation. The demographic and obstetric characteristics of the women and fetuses in the different groups are presented in Table 1. There were no relevant differences between the groups regarding maternal BMI, maternal age, median gestational age at delivery and median birth weight. Significant differences were found between the time interval groups with regard to the median absolute percentage error (MAPE) and the mean percentage error (MPE) for all evaluated formulas (p \ 0.001, data not shown). All equations showed a systematic underestimation of fetal weight (negative MPEs). A significant bias was found for all MPE values when compared to zero (p \ 0.05; data not shown).
Table 1 Principal clinical parameters in the different time interval groups Parameter
Group 1, n = 220
Group 2, n = 426
Group 3, n = 184
Group 4, n = 174
Group 5, n = 101
Group 6, n = 176
Interval: days, median (IQR)
0 (0)
1 (1)
5 (2)
10 (4)
18 (4)
30 (10)
Maternal age: years, median (IQR)
32.56 (6.81)
31.92 (6.96)
32.40 (7,40)
31.95 (6.56)
30.70 (7.48)
31.59 (6.81)
Maternal BMI: kg/m2, median (IQR)
24.90 (7.40)
24.55 (7.60)
24.85 (8.65)
25.45 (8.67)
25.30 (8.40)
25.90 (8.30)
Birth weight: g, median (IQR)
4,225 (290)
4,210 (248)
4,220 (302)
4,225 (265)
4,241 (300)
4,230 (275)
Gestational age: weeks, median (IQR)
40 (1)
40 (1)
40 (1)
40 (2)
40 (2)
40 (2)
Male fetus (%)
63.20
68.80
69.00
64.40
66.30
66.50
IQR interquartile range
Table 2 Comparison of the mean percentage errors (MPE) and random errors (RE) of fetal weight estimations between the different time interval groups Groups
Hadlock I MPE
Hadlock II
p value
RE
MPE
Merz
p value
RE
MPE
Shepard p value
RE
MPE
Warsof p value
RE
MPE
p value
RE
Group 1
-7.00
–
9.40
-5.30
–
9.44
-3.74
–
7.71
-1.54
–
10.50
-6.12
–
10.09
Group 2
-6.94
0.774
9.33
-5.40
0.607
0.480
8.44
-1.88
0.495
11.07
-6.44
0.504
10.67
7.98 7.56
-4.11 0.020 -8.03 \0.001
9.47
-3.97
Group 3 -9.02 0.016 Group 4 -13.18 \0.001
8.51 8.62
-7.68 0.006 -11.80 \0.001
8.74 8.69
-5.71 0.014 -8.67 \0.001
Group 5 -16.63 \0.001
8.74
-15.55 \0.001
8.81
-11.26 \0.001
7.99
-11.22 \0.001
9.60
-15.55 \0.000
9.28
Group 6 -27.32 \0.001
9.38
-26.30 \0.001
9.52
-19.64 \0.001
8.82
-22.26 \0.001
10.67
-26.13 \0.001
10.23
10.33 9.46
-8.65 0.011 -12.44 \0.001
9.87 9.14
p values refer to differences of the percentage error towards group 1 (reference) and are based on Wald tests for the effect estimates in a mixedeffect model (to adjust for repeated measurements). All p values were adjusted for multiple comparisons using the Bonferroni–Holm method
123
Arch Gynecol Obstet
Overall MPE values were closest to zero in time interval group one and two (Table 2; Fig. 1). No significant differences were found between WEs performed on the day of delivery and WEs with a scan-to-delivery interval of 1–3 days (time interval group 1 and 2). From group 3 to 6, a continuous decrease was shown with significant differences in comparison with group 1 (Table 2; Fig. 1). The best results were achieved with the Shepard formula in WEs performed on the day of delivery (MPE: -1.54 in
group 1) (Tables 2, 3; Fig. 1). In the other groups, values decreased from -1.88 in group 2 to -22.26 in group 6. No significant differences were found regarding MAPE values in time interval groups 1 and 2 (Table 4; Fig. 2). After a scan-to-delivery interval of more than 3 days (group three to six), MAPE values increased continuously (Table 4; Fig. 2). However, differences between time interval group one and three did not reach statistical significance (Table 4). The lowest MAPE was found with the Merz formula in WEs
Fig. 1 Distribution of percentage errors obtained in the different study groups
Table 3 Comparison of the mean percentage errors (MPE) and random errors (RE) of fetal weight estimations between the different formulas inside the time interval groups Regression formula
Group 1 p value (MPE)
Group 2 p value (RE)
p value (MPE)
Group 3 p value (RE)
p value (MPE)
Group 4 p value (RE)
p value (MPE)
Group 5 p value (RE)
p value (MPE)
Group 6 p value (RE)
p value (MPE)
p value (RE)
Hadlock I
\0.001
\0.001
\0.001
\0.001
\0.001
0.008
\0.001
\0.001
\0.001
0.007
\0.001
0.002
Hadlock II Merz
\0.001 \0.001
\0.001 *****
\0.001 \0.001
\0.001 *****
\0.001 \0.001
\0.001 *****
\0.001 0.006
\0.001 *****
\0.001 0.897
0.005 *****
\0.001 *****
0.001 *****
Shepard
*****
\0.001
*****
\0.001
*****
\0.001
*****
\0.001
*****
\0.001
0.001
\0.001
Warsof
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
p values refer to comparisons of the best-performing formula (*****) with all others and are based on paired t tests (MPE) and the Snedecor and Cochran method (RE). All p values were adjusted for multiple comparisons using the Bonferroni–Holm method
123
IQR interquartile range
p values refer to differences of the absolute percentage error towards group 1 (reference) and are based on Wald tests for the effect estimates in a mixed-effect model (to adjust for repeated measurements) All p values were adjusted for multiple comparisons using the Bonferroni–Holm method
4.55
26.73 15.54 (11.33) \0.001
26.09 (14.37) \0.001 13.64
44.55 11.84 (10.34) \0.001
22.32 (15.27) \0.001 13.07
45.54 \0.001 10.82 (9.01)
18.93 (12.89) \0.001 2.84
25.74
2.27
16.48 (10.78) \0.001 22.77 17.52 (11.10) \0.001
27.13 (13.30) \0.001
Group 5
Group 6
26.27 (13.69) \0.001
33.91
57.07 0.265
12.93 (11.82) \0.001
9.11 (11.07) 64.13
52.87 0.045
1.000 6.51 (9.72)
9.26 (9.52) 55.75
67.39 0.288
\0.001 8.83 (9.13)
6.32 (8.66) 57.07
37.93 \0.001
0.488 7.64 (10.11)
Group 4
12.44 (9.65)
54.89
32.76
0.532
\0.001
8.87 (10.59)
13.35 (9.67)
Group 3
59.55
61.50 0.609
– 8.07 (10.33)
7.71 (10.83) 65.02
61.36 –
1.000 7.15 (9.57)
7.00 (9.13) 73.64
73.71 0.654
– 5.71 (7.75)
5.54 (7.67) 66.43
63.64 –
0.847 7.33 (9.08)
6.70 (9.23) 60.91
62.21 0.907 7.84 (9.38) Group 2
– 7.90 (10.35) Group 1
p value MAPE (IQR) MAPE (IQR) p value
B10 % MAPE (IQR)
p value
B10 %
MAPE (IQR)
p value
B10 %
MAPE (IQR)
p value
B10 %
Warsof Shepard Merz Hadlock II Hadlock I
In this study, the influence of the time interval between estimation and delivery on the accuracy of sonographic WE in fetal macrosomia was evaluated and compared between six time interval groups ranging from an estimation on the day of delivery to a maximum of 6 weeks. Hereby, significant differences were found between the different time interval groups. Overall, the best results were found in WEs that were performed within 3 days before delivery. Similar results were shown in a study performed by Heer et al. [22]. The authors analyzed 820 singleton
Groups
Discussion
Table 4 Comparison of median absolute percentage errors (MAPE) and percentages of weight estimations within a 10 % range (B10 %) between the different time interval groups
performed within 1–3 days before delivery (MAPE: 5.54 in group 2) (Table 5, Fig. 2). Similar results were shown regarding the distribution of WEs within prespecified error bounds. Again, the best results were achieved in WEs performed within 1–3 days before delivery with the Merz formula: in group 2, 73.71 % of fetal WEs were falling within the 10 % range of the actual BW (Table 5; Fig. 3). These values decreased continuously from 67.39 % in group 3 to 13.07 % in group 6 (Table 4; Fig. 3). The random errors of WEs in the different groups are presented in Table 2. No major differences were found between the time interval groups. The best results were found with the model of Merz: values ranged from 7.56 in group 4 to 8.82 in group 6 (Tables 2, 3; Fig. 4). These results are further supported by an additional regression analysis investigating the combined effects of the formulas and the time interval: regarding the PEs, effect estimates (Table 6) indicated the best results for the Shepard formula (e.g., -1.10 compared to -6.28 for Hadlock I, p \ 0.001) with significantly better performance compared to all other equations (adjusted for the time interval). Furthermore, the model showed a significant negative effect of the time interval. The lowest effect of the time interval on the PE was observed for the Merz formula (-0.53 per day; Table 6). Regarding the APEs, the corresponding model indicated that best results can be achieved with the Merz formula (e.g., 6.20 compared to 8.68 for Warsof, p \ 0.001; Table 6) when adjusting for the time interval. The interval itself had a significant negative effect on the accuracy: every additional day between estimation and delivery resulted in a higher APE for all formulas. Again, the lowest effect of the time interval on the APE was observed for the Merz formula (0.42 per day; Table 6). These regression models can be, hence, also used to estimate the expected average error: for example, the average APE of the Merz formula at the day of delivery would be 6.20, while it increases to 10.4 (6.20 ? 10 9 0.42) if delivery was 10 days later (Table 6).
B10 %
Arch Gynecol Obstet
123
Arch Gynecol Obstet
Fig. 2 Distribution of absolute percentage errors obtained in the different study groups
Table 5 Comparison of median absolute percentage errors (MAPE) and percentages of weight estimations within a 10 % range (B10 %) between the different formulas inside the time interval groups Regression formula
Group 1
Group 2
Group 3
Group 4
Group 5
Group 6
p value (MAPE)
p value (B10 %)
p value (MAPE)
p value (B10 %)
p value (MAPE)
p value (B10 %)
p value (MAPE)
p value (B10 %)
p value (MAPE)
p value (B10 %)
p value (MAPE)
p value (B10 %)
Hadlock I
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
Hadlock II
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
Merz
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
1.000
Shepard
\0.001
\0.001
\0.001
\0.001
0.001
0.286
0.007
0.383
0.035
1.000
\0.001
*****
Warsof
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
p values refer to comparisons of the best-performing formula (*****) with all others and are based on Wilcoxon signed-rank test (MAPE) and McNemar tests (B10 %). All p values were adjusted for multiple comparisons using the Bonferroni–Holm method
pregnancies and evaluated 9 different factors that potentially influence the precision of sonographic weight estimation. Of the 9 evaluated factors, only a short interval between sonographic WE and delivery (0–7 vs. 8–14 days) had a statistically significant impact. MAPE, achieved with one of the Hadlock formulas was 8.52 in WEs performed within 1 week before delivery and rose to 10.37 in WEs with an interval of eight to 14 days.
123
In analogy with these results, Basha et al. analyzed sonographic WEs performed in 415 fetuses within 14 days before delivery with one of the Hadlock formulas [23]. They found a MAPE of 6.1 and 8.2 if the fetus was delivered within 7 days or eight to 14 days of sonography, respectively. In a study performed by Mongelli et al. [24], 276 fetal WEs within 35 days before delivery were analyzed. The weight estimation was either left unchanged or
Arch Gynecol Obstet
Fig. 3 Distribution of weight estimations with an absolute percentage error \10 % obtained in the different study groups
extrapolated to the time of delivery. MPE was -6.5 without any adjustment for the time interval and 5.9 after using the extrapolation method, showing a significant improvement of accuracy when taking the influence of the time interval between estimation and delivery into account. However, in the latter studies, fetuses of the whole weight range were included. Therefore, a direct comparison to our data is not possible. No impact on the accuracy of the WE caused by the time between investigation and delivery was reported by Benacerraf et al. [25]. They analyzed 1.301 fetuses of the whole weight range that were performed within a week before delivery. MPE values reflect the systematic deviation of a model from the actual BW. Application of all evaluated formulas showed a systematic underestimation of fetal weight. Similar results were found by Hoopmann et al., analyzing WE in 350 fetuses with a birth weight [4,000 g [16]. MPEs were ranging from -4.8 for the Merz equation to -9.3 for the Warsof formula. In the latter study, however, all estimations were performed within 1 week before delivery and the influence of the scan-to-delivery time interval was not evaluated. Our results did not show any significant differences regarding MPEs between WEs
performed on the day of delivery and WEs with a scan-todelivery interval of 1–3 days. However, after an interval of more than 3 days, values decreased continuously, indicating a rising underestimation caused by interim fetal growth. The standard deviation of the MPE demonstrates the random error of WEs. No major differences were found between WEs performed in the different time interval groups. Both the systematic and the random error are represented by the MAPE. The lowest values were found with the Merz formula in WEs performed within 1–3 days before delivery (MAPE: 5.54 in group 2). Similar results were found by Kurmanavicius et al., analyzing WE in 5,612 fetuses with a birth weight between 500 and 5,000 g [18]. In the group of fetuses with a birth weight of more than 4,000 g (n = 445), the lowest MAPE was found by the Merz formula, followed by the equations of Shepard and Hadlock. All WE were performed within 1 week before delivery and again the influence of the scan-todelivery time interval was not evaluated. In our study, the most accurate results were found in WEs performed within 1 week before delivery. Differences regarding MAPE values between time interval groups 1, 2 and 3 did not reach statistical significance. After an interval of more than 7 days, however, WEs were shown to be increasingly
123
Arch Gynecol Obstet
Fig. 4 Distribution of random errors obtained in the different study groups Table 6 Effect estimates from multivariable linear mixed-effect models for PEs and APEs representing the overall effect of the applied formula (adjusted for the time interval) and the formula-specific effect of the time interval (per day) PE Formula effect on PE
APE p value
Formula-specific interval effect on PE (per day)
p value
Formula effect on APE
p value
Formula specific interval effect on APE (per day)
p value
Hadlock I
-6.28
\0.001
-0.68
\0.001
8.32
\0.001
0.59
\0.001
Hadlock II
-4.71
\0.001
-0.70
\0.001
7.54
\0.001
0.59
\0.001
Merz Shepard
-3.43 -1.10
\0.001 –
-0.53 -0.68
\0.001 \0.001
6.20 7.48
– \0.001
0.42 0.45
\0.001 \0.001
Warsof
-5.73
\0.001
-0.66
\0.001
8.68
\0.001
0.54
\0.001
Numbers refer to the estimated effects of the corresponding formula and the interaction between formula and time interval in two separate mixedeffect models (PEs and APEs). The percentage errors from all formulas served as response variables, the formula and the interaction between formula and time interval as fixed effects. A subject-specific random intercept was included to adjust for repeated measurements. p values are based on Wald tests for comparing the effect estimates of the time interval to zero and the estimates of the formulas to the best-performing ones (Shepard and Merz for PE and APE, respectively). All p values were adjusted for multiple comparisons using the Bonferroni–Holm method
inaccurate. While up to 66 % of estimates with the Hadlock formula were within 10 % of birth weight in WEs that were performed within 1 week before delivery, these rates dropped to less than 38 % after an interval of more than 7 days and decreased down to nearly 25 % after more than 2 weeks.
123
These results illustrate the major impact of interim fetal growth on the accuracy of fetal WE in macrosomic fetuses. Potential problems accompanying fetal WE close to delivery like an increasing descend of the fetal head into the maternal pelvis or decreasing amniotic fluid are clearly outweighed by the interim fetal growth.
Arch Gynecol Obstet
In conclusion, WE in fetal macrosomia shows the best results when examinations are performed within 7 days before delivery, using the formula of Merz et al. Accuracy significantly decreases after this time period. Conflict of interest
None.
13.
14.
15.
References 16. 1. Kramer MS, Morin I, Yang H et al (2002) Why are babies getting bigger? Temporal trends in fetal growth and its determinants. J Pediatr 141:538–542 2. Surkan PJ, Hsieh CC, Johansson AL, Dickman PW, Cnattingius S (2004) Reasons for increasing trends in large for gestational age births. Obstet Gynecol 104:720–726 3. Patterson RM (1985) Estimation of fetal weight during labor. Obstet Gynecol 65:330–332 4. Dietz HP (2010) Childbirth-related Pelvic Floor Trauma. Geburtshilfe Frauenheilkd 70:969–978 5. Brieger GM, Rogers MS, Rushton AW, Mongelli M (1997) Are Hong Kong babies getting bigger? Int J Gynaecol Obstet 57:267–271 6. Orskou J, Kesmodel U, Henriksen TB, Secher NJ (2001) An increasing proportion of infants weigh more than 4,000 grams at birth. Acta Obstet Gynecol Scand 80:931–936 7. Boulet SL, Alexander GR, Salihu HM, Pass M (2003) Macrosomic births in the united states: determinants, outcomes, and proposed grades of risk. Am J Obstet Gynecol 188:1372–1378 8. Dudley NJ (2005) A systematic review of the ultrasound estimation of fetal weight. Ultrasound Obstet Gynecol 25:80–89 9. Hadlock FP, Harrist RB, Sharman RS, Deter RL, Park SK (1985) Estimation of fetal weight with the use of head, body, and femur measurements–a prospective study. Am J Obstet Gynecol 151:333–337 10. Shepard MJ, Richards VA, Berkowitz RL, Warsof SL, Hobbins JC (1982) An evaluation of two equations for predicting fetal weight by ultrasound. Am J Obstet Gynecol 142:47–54 11. Hadlock FP, Harrist RB, Carpenter RJ, Deter RL, Park SK (1984) Sonographic estimation of fetal weight. The value of femur length in addition to head and abdomen measurements. Radiology 150:535–540 12. Merz E, Lieser H, Schicketanz KH, Harle J (1988) Intrauterine fetal weight assessment using ultrasound. A comparison of
17.
18.
19. 20.
21.
22.
23.
24. 25.
several weight assessment methods and development of a new formula for the determination of fetal weight. Ultraschall Med 9:15–24 Warsof SL, Gohari P, Berkowitz RL, Hobbins JC (1977) The estimation of fetal weight by computer-assisted analysis. Am J Obstet Gynecol 128:881–892 Melamed N, Yogev Y, Meizner I et al (2009) Sonographic fetal weight estimation: which model should be used? J Ultrasound Med 28:617–629 Siemer J, Egger N, Hart N et al (2008) Fetal weight estimation by ultrasound: comparison of 11 different formulae and examiners with differing skill levels. Ultraschall Med 29:159–164 Hoopmann M, Abele H, Wagner N, Wallwiener D, Kagan KO (2010) Performance of 36 different weight estimation formulae in fetuses with macrosomia. Fetal Diagn Ther 27:204–213 Faschingbauer F, Voigt F, Goecke TW et al (2011) Fetal weight estimation in extreme macrosomia (C 4,500 g): comparison of 10 formulas. Ultraschall Med Kurmanavicius J, Burkhardt T, Wisser J, Huch R (2004) Ultrasonographic fetal weight estimation: accuracy of formulas and accuracy of examiners by birth weight from 500 to 5000 g. J Perinat Med 32:155–161 ACOG (2009) ACOG practice bulletin no. 101: ultrasonography in pregnancy. Obstet Gynecol 113:451–461 Eichhorn KH, Schramm T, Bald R, Hansmann M, Gembruch U (2006) DEGUM grade I quality standards in obstetric ultrasound diagnosis during the 19th–22nd week of pregnancy. Ultraschall Med 27:185–187 Merz E, Eichhorn KH, Hansmann M, Meinel K (2002) Quality demands on continuing differential diagnostic sonography in prenatal diagnostics (DEGUM stage II) during the 18th to 22nd weeks of gestation. Ultraschall Med 23:11–12 Heer IM, Kumper C, Vogtle N et al (2008) Analysis of factors influencing the ultrasonic fetal weight estimation. Fetal Diagn Ther 23:204–210 Basha AS, Abu-Khader IB, Qutishat RM, Amarin ZO (2012) Accuracy of sonographic fetal weight estimation within 14 days of delivery in a Jordanian population using Hadlock formula 1. Med Princ Prac Int J Kuwait Univ Health Sci Centre 21:366–369 Mongelli M, Gardosi J (1996) Gestation-adjusted projection of estimated fetal weight. Acta Obstet Gynecol Scand 75:28–31 Benacerraf BR, Gelman R, Frigoletto FD Jr (1988) Sonographically estimated fetal weights: accuracy and limitation. Am J Obstet Gynecol 159:1118–1121
123
MATERNAL-FETAL MEDICINE
Sonographic weight estimation in fetal macrosomia: influence of the time interval between estimation and delivery F. Faschingbauer • U. Dammer • E. Raabe • M. Schneider C. Faschingbauer • M. Schmid • A. Mayr • R. L. Schild • M. W. Beckmann • S. Kehl
•
Received: 21 July 2014 / Accepted: 16 December 2014 Springer-Verlag Berlin Heidelberg 2014
Abstract Purpose To evaluate the influence of the time interval between examination and delivery on the accuracy of sonographic weight estimation (WE) in fetal macrosomia. Materials and methods 896 singleton pregnancies (birth weight [ 4,000 g) with a total of 1,281 sonographic weight estimations were included in this retrospective cohort study. Fetuses were divided into six groups with regard to the time interval between estimation and delivery: group 1: scan-to-delivery interval: 0 days; group 2: scan-to-delivery interval: 1–3 days; group 3: scan-to-delivery interval: 4–7 days; group 4: scan-to-delivery interval: 8–14 days; group 5: scan-to-delivery interval: 15–21 days; group 6: scan-to-delivery interval: 22–42 days. The accuracy of WE was compared between five commonly used formulas using Electronic supplementary material The online version of this article (doi:10.1007/s00404-014-3604-y) contains supplementary material, which is available to authorized users. F. Faschingbauer (&) U. Dammer E. Raabe M. Schneider C. Faschingbauer M. W. Beckmann S. Kehl Department of Obstetrics and Gynecology, Erlangen University Hospital, Universita¨tsstrasse 21–23, 91054 Erlangen, Germany e-mail: [email protected] M. Schmid Department of Medical Informatics, Biometry and Epidemiology, Rheinische Friedrich-Wilhelms-University Bonn, Bonn, Germany A. Mayr Department of Medical Informatics, Biometry and Epidemiology, University of Erlangen-Nuremberg, Erlangen, Germany R. L. Schild Department of Obstetrics and Perinatal Medicine, DDH Frauenkliniken, Hanover, Germany
means of percentage errors (MPE), random error, medians of absolute percentage errors (MAPE), and proportions of estimates within 10 % of actual birth weight. Results Significant differences were found between the time interval groups with regard to MAPE and MPE values (p \ 0.001). All formulas showed a systematic underestimation of fetal weight (negative MPEs) (p \ 0.05). MPE values were closest to zero in time interval group 1 and 2. From group 3 to 6, a continuous decrease was observed. The lowest MAPE was found with the Merz formula in group 1 and 2. Values increased continuously from group 3 to 6. Differences between time interval group one and three did not reach statistical significance. Conclusions WE in fetal macrosomia shows the best results when examinations are performed within 7 days before delivery, using the formula of Merz et al. Accuracy significantly decreases after this time period. Keywords Ultrasound Fetal weight estimation Birth weight Time interval Macrosomia
Introduction The incidence of fetal macrosomia has risen continuously in recent years [1, 2]. Antenatal diagnosis is of major importance in the management of labor and delivery and for preventing fetal and maternal trauma during childbirth [3, 4]. Typical risks include a prolonged second stage of labor, serious maternal trauma after vaginal and surgical delivery, increased postpartum hemorrhage and shoulder dystocia with brachial plexus paralysis and/or clavicular fracture [5–7]. During the past 40 years, sonographic assessment of the fetus and estimation of its weight have become part of
123
Arch Gynecol Obstet
routine practice in obstetrics. Several formulas have been published, most of them involving combinations of several biometric parameters [8–13]. However, most of these formulas were derived from heterogeneous groups and were not specifically designed for estimation of fetal weight at the upper end of the weight scale. Several studies have assessed the accuracy of sonographic weight estimation (WE) in macrosomic fetuses and most of them found a significant underestimation of fetal weight with errors exceeding 10 % [8, 14–18]. The majority of these studies analyzed sonograms that were performed close to term within 7 days before delivery [8, 15–18]. However, in numerous patients, fetal WE has either been performed before this time period or is conducted when the patient presents during the latent or active phase of labor. While in the first case interim fetal growth may lead to a further underestimation, ultrasound performed intrapartum may be accompanied by several other problems: first, as delivery approaches, the fetal head descends into the maternal pelvis leading to inaccurate measurements of the biparietal diameter (BPD), occipitofrontal diameter (OFD) and head circumference (HC). Second, an increased risk of abdominal circumference distortion or posterior position of the femora is more frequently observed close to delivery and finally, decreasing amounts of amniotic fluid at term could limit the accuracy of all measurements. So far, no studies have evaluated the influence of the time interval between examination and delivery on the accuracy of sonographic WE in fetal macrosomia. Therefore, the aim of the present study was to compare the accuracy of sonographic fetal WE between six time interval groups ranging from an estimation on the day of delivery to a maximum of 6 weeks, analyzing more than 1,281 biometric measurements of fetuses with a birth weight of more than 4,000 g.
Materials and methods This retrospective, cross-sectional, single-center study included 896 singleton pregnancies with a total of 1,281 sonographic weight estimations over a 7-year period (2003–2009). An automatic query of our perinatal ultrasound database was conducted with the following inclusion criteria: singleton pregnancy with cephalic presentation, a birth weight of more than 4,000 g and an ultrasound examination with complete biometric parameters—BPD, OFD, HC(HC = 2.325 9 ((OFD)2 ? (BPD)2)1/2), abdominal transverse diameter (ATD), abdominal anterior– posterior diameter (APAD), AC(AC = p 9 (ATD ? APD)/2), and femur length (FL) within a time interval between zero and 42 days before delivery; and an absence
123
of any chromosomal or structural anomalies. Intrauterine fetal deaths were excluded. In our institution, a sonographic weight estimation is routinely performed in all patients registering for delivery. Thus, in the vast majority of cases, this was the indication for the examination. Fetuses were divided into six groups with regard to the time interval between estimation and delivery: • • • • • •
Group 1: weight estimation on the day of delivery n = 220. Group 2: time interval between estimation and delivery 1–3 days, n = 426. Group 3: time interval between estimation and delivery 4–7 days, n = 184. Group 4: time interval between estimation and delivery 8–14 days, n = 174. Group 5: time interval between estimation and delivery 15–21 days, n = 101. Group 6: time interval between estimation and delivery 22–42 days, n = 176.
In cases of repeated measurements within one group, only one examination was randomly included. Gestational age was calculated from the last menstrual period and was confirmed by or recalculated with biometric measurements obtained from the first fetal biometry in early pregnancy (in accordance with the recommendations of the American College of Obstetricians and Gynecologists, ACOG) [19]. The examinations were performed by a total of 44 sonographers in accordance with widely accepted quality standards [20, 21]. A variety of ultrasound machines (Siemens Elegra, Siemens Sonoline, G60 S, Siemens Sienna, Siemens Medical Solutions, Erlangen, Germany; GE Voluson 730 Expert, GE Voluson e Laptop, GE Healthcare, Solingen, Germany; Xario and Nemio systems, Toshiba Medical Systems, Neuss, Germany) were used with standard techniques. Birth weight and neonatal length were measured within 1 h after delivery by the nursing staff. For estimation of fetal weight, five widespread formulas (Hadlock et al., Shepard et al., Merz et al., Warsof et al. [9–13]) were used, including the biometric parameters HC, BPD, AC and FL (supplementary Table). Measurements are given in centimeters and BW in grams. In the department in which the study was conducted, fetal weight is routinely measured by ultrasound examination during the diagnostic work-up. Ethical approval for the study was, therefore, not sought. The accuracy of the estimated fetal weight was assessed by calculating (a) the percentage error (PE): (EFW - BW)/ BW 9 100, the mean of which reflects the systematic deviation of a model from the actual BW; (b) the random error (standard deviation of the PE)—a measure of precision that reflects the random component of the prediction error; (c) the absolute percentage error (APE): |(EFW - BW)/
Arch Gynecol Obstet
BW| 9 100, which takes both the systematic and the random error into account; and (d) the percentage of fetal WEs falling within prespecified error bounds (APE \ 10 %). MPEs for all formulas and groups were compared to zero using one-sample t tests. To compare the prediction accuracy between the interval groups, linear mixed-effect regression analysis was performed separately for the different formulas with PEs and APEs as outcome variables and the interval group as predictor variable. Additionally, subject-specific random effects were included to adjust for repeated measurements. Overall group differences were assessed via Likelihood-ratio tests, group-specific differences toward measurements at delivery were assessed via Wald tests for the effect estimates with group 1 as reference. Differences between the formulas inside the interval groups were assessed via paired t tests (PE), the Snedecor and Cochran method (random error), Wilcoxon signedrank tests (APE) and McNemar tests (APE \ 10 %). Linear mixed-effect regression was applied to analyze the combined effects of the time interval and the different formulas. This allows assessing the overall differences between the formulas, adjusted for the time interval and the effect of the time interval for each specific formula. The PEs and APEs from all formulas served as outcome variables, the formula and an interaction between the time interval and
the formula as fixed effects. A subject-specific random effect was included to adjust for repeated measurements. All p values were adjusted for multiple comparisons using the Bonferroni–Holm method.
Results The automatic query of our perinatal data base with the above-mentioned inclusion and exclusion criteria resulted in 9,385 fetuses of the whole weight range, of which 915 fetuses had a birth weight of more than 4,000 g. 19 fetuses had to be excluded due to implausible values because of incorrect documentation. The demographic and obstetric characteristics of the women and fetuses in the different groups are presented in Table 1. There were no relevant differences between the groups regarding maternal BMI, maternal age, median gestational age at delivery and median birth weight. Significant differences were found between the time interval groups with regard to the median absolute percentage error (MAPE) and the mean percentage error (MPE) for all evaluated formulas (p \ 0.001, data not shown). All equations showed a systematic underestimation of fetal weight (negative MPEs). A significant bias was found for all MPE values when compared to zero (p \ 0.05; data not shown).
Table 1 Principal clinical parameters in the different time interval groups Parameter
Group 1, n = 220
Group 2, n = 426
Group 3, n = 184
Group 4, n = 174
Group 5, n = 101
Group 6, n = 176
Interval: days, median (IQR)
0 (0)
1 (1)
5 (2)
10 (4)
18 (4)
30 (10)
Maternal age: years, median (IQR)
32.56 (6.81)
31.92 (6.96)
32.40 (7,40)
31.95 (6.56)
30.70 (7.48)
31.59 (6.81)
Maternal BMI: kg/m2, median (IQR)
24.90 (7.40)
24.55 (7.60)
24.85 (8.65)
25.45 (8.67)
25.30 (8.40)
25.90 (8.30)
Birth weight: g, median (IQR)
4,225 (290)
4,210 (248)
4,220 (302)
4,225 (265)
4,241 (300)
4,230 (275)
Gestational age: weeks, median (IQR)
40 (1)
40 (1)
40 (1)
40 (2)
40 (2)
40 (2)
Male fetus (%)
63.20
68.80
69.00
64.40
66.30
66.50
IQR interquartile range
Table 2 Comparison of the mean percentage errors (MPE) and random errors (RE) of fetal weight estimations between the different time interval groups Groups
Hadlock I MPE
Hadlock II
p value
RE
MPE
Merz
p value
RE
MPE
Shepard p value
RE
MPE
Warsof p value
RE
MPE
p value
RE
Group 1
-7.00
–
9.40
-5.30
–
9.44
-3.74
–
7.71
-1.54
–
10.50
-6.12
–
10.09
Group 2
-6.94
0.774
9.33
-5.40
0.607
0.480
8.44
-1.88
0.495
11.07
-6.44
0.504
10.67
7.98 7.56
-4.11 0.020 -8.03 \0.001
9.47
-3.97
Group 3 -9.02 0.016 Group 4 -13.18 \0.001
8.51 8.62
-7.68 0.006 -11.80 \0.001
8.74 8.69
-5.71 0.014 -8.67 \0.001
Group 5 -16.63 \0.001
8.74
-15.55 \0.001
8.81
-11.26 \0.001
7.99
-11.22 \0.001
9.60
-15.55 \0.000
9.28
Group 6 -27.32 \0.001
9.38
-26.30 \0.001
9.52
-19.64 \0.001
8.82
-22.26 \0.001
10.67
-26.13 \0.001
10.23
10.33 9.46
-8.65 0.011 -12.44 \0.001
9.87 9.14
p values refer to differences of the percentage error towards group 1 (reference) and are based on Wald tests for the effect estimates in a mixedeffect model (to adjust for repeated measurements). All p values were adjusted for multiple comparisons using the Bonferroni–Holm method
123
Arch Gynecol Obstet
Overall MPE values were closest to zero in time interval group one and two (Table 2; Fig. 1). No significant differences were found between WEs performed on the day of delivery and WEs with a scan-to-delivery interval of 1–3 days (time interval group 1 and 2). From group 3 to 6, a continuous decrease was shown with significant differences in comparison with group 1 (Table 2; Fig. 1). The best results were achieved with the Shepard formula in WEs performed on the day of delivery (MPE: -1.54 in
group 1) (Tables 2, 3; Fig. 1). In the other groups, values decreased from -1.88 in group 2 to -22.26 in group 6. No significant differences were found regarding MAPE values in time interval groups 1 and 2 (Table 4; Fig. 2). After a scan-to-delivery interval of more than 3 days (group three to six), MAPE values increased continuously (Table 4; Fig. 2). However, differences between time interval group one and three did not reach statistical significance (Table 4). The lowest MAPE was found with the Merz formula in WEs
Fig. 1 Distribution of percentage errors obtained in the different study groups
Table 3 Comparison of the mean percentage errors (MPE) and random errors (RE) of fetal weight estimations between the different formulas inside the time interval groups Regression formula
Group 1 p value (MPE)
Group 2 p value (RE)
p value (MPE)
Group 3 p value (RE)
p value (MPE)
Group 4 p value (RE)
p value (MPE)
Group 5 p value (RE)
p value (MPE)
Group 6 p value (RE)
p value (MPE)
p value (RE)
Hadlock I
\0.001
\0.001
\0.001
\0.001
\0.001
0.008
\0.001
\0.001
\0.001
0.007
\0.001
0.002
Hadlock II Merz
\0.001 \0.001
\0.001 *****
\0.001 \0.001
\0.001 *****
\0.001 \0.001
\0.001 *****
\0.001 0.006
\0.001 *****
\0.001 0.897
0.005 *****
\0.001 *****
0.001 *****
Shepard
*****
\0.001
*****
\0.001
*****
\0.001
*****
\0.001
*****
\0.001
0.001
\0.001
Warsof
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
p values refer to comparisons of the best-performing formula (*****) with all others and are based on paired t tests (MPE) and the Snedecor and Cochran method (RE). All p values were adjusted for multiple comparisons using the Bonferroni–Holm method
123
IQR interquartile range
p values refer to differences of the absolute percentage error towards group 1 (reference) and are based on Wald tests for the effect estimates in a mixed-effect model (to adjust for repeated measurements) All p values were adjusted for multiple comparisons using the Bonferroni–Holm method
4.55
26.73 15.54 (11.33) \0.001
26.09 (14.37) \0.001 13.64
44.55 11.84 (10.34) \0.001
22.32 (15.27) \0.001 13.07
45.54 \0.001 10.82 (9.01)
18.93 (12.89) \0.001 2.84
25.74
2.27
16.48 (10.78) \0.001 22.77 17.52 (11.10) \0.001
27.13 (13.30) \0.001
Group 5
Group 6
26.27 (13.69) \0.001
33.91
57.07 0.265
12.93 (11.82) \0.001
9.11 (11.07) 64.13
52.87 0.045
1.000 6.51 (9.72)
9.26 (9.52) 55.75
67.39 0.288
\0.001 8.83 (9.13)
6.32 (8.66) 57.07
37.93 \0.001
0.488 7.64 (10.11)
Group 4
12.44 (9.65)
54.89
32.76
0.532
\0.001
8.87 (10.59)
13.35 (9.67)
Group 3
59.55
61.50 0.609
– 8.07 (10.33)
7.71 (10.83) 65.02
61.36 –
1.000 7.15 (9.57)
7.00 (9.13) 73.64
73.71 0.654
– 5.71 (7.75)
5.54 (7.67) 66.43
63.64 –
0.847 7.33 (9.08)
6.70 (9.23) 60.91
62.21 0.907 7.84 (9.38) Group 2
– 7.90 (10.35) Group 1
p value MAPE (IQR) MAPE (IQR) p value
B10 % MAPE (IQR)
p value
B10 %
MAPE (IQR)
p value
B10 %
MAPE (IQR)
p value
B10 %
Warsof Shepard Merz Hadlock II Hadlock I
In this study, the influence of the time interval between estimation and delivery on the accuracy of sonographic WE in fetal macrosomia was evaluated and compared between six time interval groups ranging from an estimation on the day of delivery to a maximum of 6 weeks. Hereby, significant differences were found between the different time interval groups. Overall, the best results were found in WEs that were performed within 3 days before delivery. Similar results were shown in a study performed by Heer et al. [22]. The authors analyzed 820 singleton
Groups
Discussion
Table 4 Comparison of median absolute percentage errors (MAPE) and percentages of weight estimations within a 10 % range (B10 %) between the different time interval groups
performed within 1–3 days before delivery (MAPE: 5.54 in group 2) (Table 5, Fig. 2). Similar results were shown regarding the distribution of WEs within prespecified error bounds. Again, the best results were achieved in WEs performed within 1–3 days before delivery with the Merz formula: in group 2, 73.71 % of fetal WEs were falling within the 10 % range of the actual BW (Table 5; Fig. 3). These values decreased continuously from 67.39 % in group 3 to 13.07 % in group 6 (Table 4; Fig. 3). The random errors of WEs in the different groups are presented in Table 2. No major differences were found between the time interval groups. The best results were found with the model of Merz: values ranged from 7.56 in group 4 to 8.82 in group 6 (Tables 2, 3; Fig. 4). These results are further supported by an additional regression analysis investigating the combined effects of the formulas and the time interval: regarding the PEs, effect estimates (Table 6) indicated the best results for the Shepard formula (e.g., -1.10 compared to -6.28 for Hadlock I, p \ 0.001) with significantly better performance compared to all other equations (adjusted for the time interval). Furthermore, the model showed a significant negative effect of the time interval. The lowest effect of the time interval on the PE was observed for the Merz formula (-0.53 per day; Table 6). Regarding the APEs, the corresponding model indicated that best results can be achieved with the Merz formula (e.g., 6.20 compared to 8.68 for Warsof, p \ 0.001; Table 6) when adjusting for the time interval. The interval itself had a significant negative effect on the accuracy: every additional day between estimation and delivery resulted in a higher APE for all formulas. Again, the lowest effect of the time interval on the APE was observed for the Merz formula (0.42 per day; Table 6). These regression models can be, hence, also used to estimate the expected average error: for example, the average APE of the Merz formula at the day of delivery would be 6.20, while it increases to 10.4 (6.20 ? 10 9 0.42) if delivery was 10 days later (Table 6).
B10 %
Arch Gynecol Obstet
123
Arch Gynecol Obstet
Fig. 2 Distribution of absolute percentage errors obtained in the different study groups
Table 5 Comparison of median absolute percentage errors (MAPE) and percentages of weight estimations within a 10 % range (B10 %) between the different formulas inside the time interval groups Regression formula
Group 1
Group 2
Group 3
Group 4
Group 5
Group 6
p value (MAPE)
p value (B10 %)
p value (MAPE)
p value (B10 %)
p value (MAPE)
p value (B10 %)
p value (MAPE)
p value (B10 %)
p value (MAPE)
p value (B10 %)
p value (MAPE)
p value (B10 %)
Hadlock I
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
Hadlock II
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
Merz
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
*****
1.000
Shepard
\0.001
\0.001
\0.001
\0.001
0.001
0.286
0.007
0.383
0.035
1.000
\0.001
*****
Warsof
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
\0.001
p values refer to comparisons of the best-performing formula (*****) with all others and are based on Wilcoxon signed-rank test (MAPE) and McNemar tests (B10 %). All p values were adjusted for multiple comparisons using the Bonferroni–Holm method
pregnancies and evaluated 9 different factors that potentially influence the precision of sonographic weight estimation. Of the 9 evaluated factors, only a short interval between sonographic WE and delivery (0–7 vs. 8–14 days) had a statistically significant impact. MAPE, achieved with one of the Hadlock formulas was 8.52 in WEs performed within 1 week before delivery and rose to 10.37 in WEs with an interval of eight to 14 days.
123
In analogy with these results, Basha et al. analyzed sonographic WEs performed in 415 fetuses within 14 days before delivery with one of the Hadlock formulas [23]. They found a MAPE of 6.1 and 8.2 if the fetus was delivered within 7 days or eight to 14 days of sonography, respectively. In a study performed by Mongelli et al. [24], 276 fetal WEs within 35 days before delivery were analyzed. The weight estimation was either left unchanged or
Arch Gynecol Obstet
Fig. 3 Distribution of weight estimations with an absolute percentage error \10 % obtained in the different study groups
extrapolated to the time of delivery. MPE was -6.5 without any adjustment for the time interval and 5.9 after using the extrapolation method, showing a significant improvement of accuracy when taking the influence of the time interval between estimation and delivery into account. However, in the latter studies, fetuses of the whole weight range were included. Therefore, a direct comparison to our data is not possible. No impact on the accuracy of the WE caused by the time between investigation and delivery was reported by Benacerraf et al. [25]. They analyzed 1.301 fetuses of the whole weight range that were performed within a week before delivery. MPE values reflect the systematic deviation of a model from the actual BW. Application of all evaluated formulas showed a systematic underestimation of fetal weight. Similar results were found by Hoopmann et al., analyzing WE in 350 fetuses with a birth weight [4,000 g [16]. MPEs were ranging from -4.8 for the Merz equation to -9.3 for the Warsof formula. In the latter study, however, all estimations were performed within 1 week before delivery and the influence of the scan-to-delivery time interval was not evaluated. Our results did not show any significant differences regarding MPEs between WEs
performed on the day of delivery and WEs with a scan-todelivery interval of 1–3 days. However, after an interval of more than 3 days, values decreased continuously, indicating a rising underestimation caused by interim fetal growth. The standard deviation of the MPE demonstrates the random error of WEs. No major differences were found between WEs performed in the different time interval groups. Both the systematic and the random error are represented by the MAPE. The lowest values were found with the Merz formula in WEs performed within 1–3 days before delivery (MAPE: 5.54 in group 2). Similar results were found by Kurmanavicius et al., analyzing WE in 5,612 fetuses with a birth weight between 500 and 5,000 g [18]. In the group of fetuses with a birth weight of more than 4,000 g (n = 445), the lowest MAPE was found by the Merz formula, followed by the equations of Shepard and Hadlock. All WE were performed within 1 week before delivery and again the influence of the scan-todelivery time interval was not evaluated. In our study, the most accurate results were found in WEs performed within 1 week before delivery. Differences regarding MAPE values between time interval groups 1, 2 and 3 did not reach statistical significance. After an interval of more than 7 days, however, WEs were shown to be increasingly
123
Arch Gynecol Obstet
Fig. 4 Distribution of random errors obtained in the different study groups Table 6 Effect estimates from multivariable linear mixed-effect models for PEs and APEs representing the overall effect of the applied formula (adjusted for the time interval) and the formula-specific effect of the time interval (per day) PE Formula effect on PE
APE p value
Formula-specific interval effect on PE (per day)
p value
Formula effect on APE
p value
Formula specific interval effect on APE (per day)
p value
Hadlock I
-6.28
\0.001
-0.68
\0.001
8.32
\0.001
0.59
\0.001
Hadlock II
-4.71
\0.001
-0.70
\0.001
7.54
\0.001
0.59
\0.001
Merz Shepard
-3.43 -1.10
\0.001 –
-0.53 -0.68
\0.001 \0.001
6.20 7.48
– \0.001
0.42 0.45
\0.001 \0.001
Warsof
-5.73
\0.001
-0.66
\0.001
8.68
\0.001
0.54
\0.001
Numbers refer to the estimated effects of the corresponding formula and the interaction between formula and time interval in two separate mixedeffect models (PEs and APEs). The percentage errors from all formulas served as response variables, the formula and the interaction between formula and time interval as fixed effects. A subject-specific random intercept was included to adjust for repeated measurements. p values are based on Wald tests for comparing the effect estimates of the time interval to zero and the estimates of the formulas to the best-performing ones (Shepard and Merz for PE and APE, respectively). All p values were adjusted for multiple comparisons using the Bonferroni–Holm method
inaccurate. While up to 66 % of estimates with the Hadlock formula were within 10 % of birth weight in WEs that were performed within 1 week before delivery, these rates dropped to less than 38 % after an interval of more than 7 days and decreased down to nearly 25 % after more than 2 weeks.
123
These results illustrate the major impact of interim fetal growth on the accuracy of fetal WE in macrosomic fetuses. Potential problems accompanying fetal WE close to delivery like an increasing descend of the fetal head into the maternal pelvis or decreasing amniotic fluid are clearly outweighed by the interim fetal growth.
Arch Gynecol Obstet
In conclusion, WE in fetal macrosomia shows the best results when examinations are performed within 7 days before delivery, using the formula of Merz et al. Accuracy significantly decreases after this time period. Conflict of interest
None.
13.
14.
15.
References 16. 1. Kramer MS, Morin I, Yang H et al (2002) Why are babies getting bigger? Temporal trends in fetal growth and its determinants. J Pediatr 141:538–542 2. Surkan PJ, Hsieh CC, Johansson AL, Dickman PW, Cnattingius S (2004) Reasons for increasing trends in large for gestational age births. Obstet Gynecol 104:720–726 3. Patterson RM (1985) Estimation of fetal weight during labor. Obstet Gynecol 65:330–332 4. Dietz HP (2010) Childbirth-related Pelvic Floor Trauma. Geburtshilfe Frauenheilkd 70:969–978 5. Brieger GM, Rogers MS, Rushton AW, Mongelli M (1997) Are Hong Kong babies getting bigger? Int J Gynaecol Obstet 57:267–271 6. Orskou J, Kesmodel U, Henriksen TB, Secher NJ (2001) An increasing proportion of infants weigh more than 4,000 grams at birth. Acta Obstet Gynecol Scand 80:931–936 7. Boulet SL, Alexander GR, Salihu HM, Pass M (2003) Macrosomic births in the united states: determinants, outcomes, and proposed grades of risk. Am J Obstet Gynecol 188:1372–1378 8. Dudley NJ (2005) A systematic review of the ultrasound estimation of fetal weight. Ultrasound Obstet Gynecol 25:80–89 9. Hadlock FP, Harrist RB, Sharman RS, Deter RL, Park SK (1985) Estimation of fetal weight with the use of head, body, and femur measurements–a prospective study. Am J Obstet Gynecol 151:333–337 10. Shepard MJ, Richards VA, Berkowitz RL, Warsof SL, Hobbins JC (1982) An evaluation of two equations for predicting fetal weight by ultrasound. Am J Obstet Gynecol 142:47–54 11. Hadlock FP, Harrist RB, Carpenter RJ, Deter RL, Park SK (1984) Sonographic estimation of fetal weight. The value of femur length in addition to head and abdomen measurements. Radiology 150:535–540 12. Merz E, Lieser H, Schicketanz KH, Harle J (1988) Intrauterine fetal weight assessment using ultrasound. A comparison of
17.
18.
19. 20.
21.
22.
23.
24. 25.
several weight assessment methods and development of a new formula for the determination of fetal weight. Ultraschall Med 9:15–24 Warsof SL, Gohari P, Berkowitz RL, Hobbins JC (1977) The estimation of fetal weight by computer-assisted analysis. Am J Obstet Gynecol 128:881–892 Melamed N, Yogev Y, Meizner I et al (2009) Sonographic fetal weight estimation: which model should be used? J Ultrasound Med 28:617–629 Siemer J, Egger N, Hart N et al (2008) Fetal weight estimation by ultrasound: comparison of 11 different formulae and examiners with differing skill levels. Ultraschall Med 29:159–164 Hoopmann M, Abele H, Wagner N, Wallwiener D, Kagan KO (2010) Performance of 36 different weight estimation formulae in fetuses with macrosomia. Fetal Diagn Ther 27:204–213 Faschingbauer F, Voigt F, Goecke TW et al (2011) Fetal weight estimation in extreme macrosomia (C 4,500 g): comparison of 10 formulas. Ultraschall Med Kurmanavicius J, Burkhardt T, Wisser J, Huch R (2004) Ultrasonographic fetal weight estimation: accuracy of formulas and accuracy of examiners by birth weight from 500 to 5000 g. J Perinat Med 32:155–161 ACOG (2009) ACOG practice bulletin no. 101: ultrasonography in pregnancy. Obstet Gynecol 113:451–461 Eichhorn KH, Schramm T, Bald R, Hansmann M, Gembruch U (2006) DEGUM grade I quality standards in obstetric ultrasound diagnosis during the 19th–22nd week of pregnancy. Ultraschall Med 27:185–187 Merz E, Eichhorn KH, Hansmann M, Meinel K (2002) Quality demands on continuing differential diagnostic sonography in prenatal diagnostics (DEGUM stage II) during the 18th to 22nd weeks of gestation. Ultraschall Med 23:11–12 Heer IM, Kumper C, Vogtle N et al (2008) Analysis of factors influencing the ultrasonic fetal weight estimation. Fetal Diagn Ther 23:204–210 Basha AS, Abu-Khader IB, Qutishat RM, Amarin ZO (2012) Accuracy of sonographic fetal weight estimation within 14 days of delivery in a Jordanian population using Hadlock formula 1. Med Princ Prac Int J Kuwait Univ Health Sci Centre 21:366–369 Mongelli M, Gardosi J (1996) Gestation-adjusted projection of estimated fetal weight. Acta Obstet Gynecol Scand 75:28–31 Benacerraf BR, Gelman R, Frigoletto FD Jr (1988) Sonographically estimated fetal weights: accuracy and limitation. Am J Obstet Gynecol 159:1118–1121
123
