Effect of Increased Body Mass Index on the Accuracy of Estimated Fetal Weight by Sonography in Twins

Transcription

1 Article Effect of Increased Body Mass Index on the Accuracy of Estimated Fetal Weight by Sonography in Twins Manisha Gandhi, MD, Lauren Ferrara, MD, Victoria Belogolovkin, MD, Erin Moshier, MS, Andrei Rebaber, MD Objective. The purpose of this study was to determine whether increasing body mass index (BMI) decreases the accuracy of sonographic estimations of fetal weight in twin gestations. Methods. A chart review was conducted, in which 361 charts of patients with twin gestations over a 2-year period were reviewed. A total of 194 patients had sonographic examinations for fetal weight within 6 days of delivery and were included in the analysis. The difference between the sonographically estimated fetal weight was compared with the actual birth weight for each twin and stratified for the patient s BMI. Results. There was a significant increasing trend in mean absolute percent errors with increasing BMIs in both twins (P <.05). The mean absolute percent errors for twin A were 6% for patients with a BMI of less than 25 and 9% for those with a BMI of greater than 30. The mean absolute percent errors for twin B were 6.7% for patients with a BMI of less than 25 and almost 11.7% for those with a BMI of greater than 30. There was a significantly increasing trend in mean absolute differences in grams for both twins with increasing gestational age, with almost a 4-fold increase from less than 28 weeks to greater than 36 weeks in both twins (P <.05). Conclusions. Increasing maternal obesity decreases the accuracy of sonographically determined fetal weight in twin gestations, particularly for twin B. Key words: body mass index; estimated fetal weight; obesity; twins. Abbreviations ANOVA, analysis of variance; BMI, body mass index; NPV, negative predictive value; PPV, positive predictive value Received July 28, 2008, from the Departments of Obstetrics and Gynecology (M.G., L.F., A.R.) and Community and Preventive Medicine (E.M.), Mount Sinai School of Medicine, New York, New York USA; Carnegie Hill Imaging, New York, New York USA (A.R.); and Department of Obstetrics and Gynecology, University of Florida, Tampa, Florida USA (V.B.). Revision requested August 27, Revised manuscript accepted for publication October 11, Address correspondence to Manisha Gandhi, MD, 5 E 98th St, Second Floor, New York, NY USA. manisha.gandhi@mssm.edu Sonographic and clinical estimations of fetal weights are frequently used in daily obstetric management. Estimated fetal weight is often used when considering the mode of delivery in singleton and twin pregnancies. In addition, estimated fetal weight is used to manage twin gestations in the antepartum period, notably with regard to intertwin discordancy. Multiple studies have addressed the accuracy of sonography with respect to predicting fetal weight. 1 3 Benacerraf et al 3 studied one of the largest singleton populations at a single institution and estimated that 74% of neonates born were within 10% of the sonographically predicted estimated fetal weight. Multiple studies have addressed the accuracy of sonography with respect to predicting fetal weight and have noted that it can estimate fetal weight accurately to within an absolute percentage error ranging from ±8% to ±19%. 4 5 These findings have also been observed in twin gestations by the American Institute of Ultrasound in Medicine J Ultrasound Med 2009; 28: /09/$3.50

2 Increased Body Mass Index and Estimated Fetal Weight in Twins There are a growing number of patients with a body mass index (BMI) qualified as overweight (BMI ) or obese (BMI 30) in the United States. Data from 2 National Health and Nutrition Examination Surveys show that among adults aged 20 to 74 years, the prevalence of obesity increased from 15% (in the survey) to 32.9% (in the survey). 7 This suggests that one-third of the adult population are obese. Previous studies have noted the adverse effects maternal obesity can have on the accuracy of the sonographic fetal anatomic survey. 8,9 Field et al 10 studied 998 singleton pregnancies and found that increasing maternal obesity did not affect the accuracy of estimated fetal weight, and there was no increase in the mean absolute percent error with increasing maternal size. Another study of women with a BMI of less than 32 versus those with a BMI of greater than 32 provided similar conclusions. 11 The number of twin pregnancies has increased because of delayed childbearing and greater reliance on fertility treatments. Our study proposed to determine the effect of obesity on sonographically predicted estimated fetal weight because these predictions are frequently used in the management of twin gestations antepartum and with regard to the mode of delivery. In addition, we aimed to identify a difference in accuracy with regard to twin A versus twin B with increasing BMI as well as the accuracy of identifying intertwin discordancy as BMI increases. Materials and Methods A chart review was conducted by electronically generating a list of twin deliveries using the code for twin gestations from the International Classification of Diseases, Ninth Revision, at Mount Sinai Hospital over a 2-year period from January 2005 to February This study was approved by the Institutional Review Board at Mount Sinai Hospital. The review identified 369 patients. Inclusion criteria consisted of patients who gave birth to viable twins after 24 weeks and who had a sonographic evaluation of fetal weight within 6 days of delivery. We chose those who gave birth within 6 days between sonography and delivery to avoid having to mathematically correct the sonographically predicted weight for expected fetal growth with larger sonography-todelivery intervals. A total of 194 patients (54%) met our inclusion criteria from the charts reviewed. Images were acquired, and measurements were obtained by 1 of 5 American Registry for Diagnostic Medical Sonography certified sonographers or 3 maternal-fetal medicine fellows. All biometric parameters were reviewed by board-certified maternal-fetal medicine attending physicians at the institution for final reporting. The outcome variables collected included maternal height and weight at the time of sonography or delivery, maternal age, parity, chorionicity, gestational age at the time of sonography, gestational age at the time of delivery, estimated fetal weight for each twin at the time of sonography, and actual birth weight for each twin at the time of delivery. The BMI was calculated from the patient s reported height and maternal weight measured either at the time of sonography or at the time of delivery. The BMI was calculated as weight in kilograms divided by height in square meters. Fetal weight estimations were calculated by the Hadlock regression formula, which incorporates the biparietal diameter, head circumference, abdominal circumference, and femur length. 5 Absolute percent errors were computed by dividing the absolute difference between sonographically estimated and actual birth weights by the actual birth weight; this result was expressed as a percentage. Absolute differences in grams were computed by taking the absolute difference between sonographically estimated and actual birth weights to provide a true indication of the mean gram error that could be expected for individual fetal weight predictions. The means of the differences, absolute differences and absolute percent errors were estimated and compared across categories of BMI and gestational age. Linear regression and analysis of variance (ANOVA) P values were used to determine whether sonographic prediction of the actual birth weight varied by BMI or gestational age. Sonographically estimated intertwin discordance was calculated by taking the difference between the sonographically estimated birth weights of twins A and B and dividing by the estimated birth weight of the larger twin; this result was expressed as a percentage. Actual intertwin 302 J Ultrasound Med 2009; 28:

3 Gandhi et al discordance was calculated by using the same method, except we used the actual birth weights of each twin. At our institution, we define intertwin discordance as 25%. 12 Results The demographic data collected are presented in Table 1. The average maternal age was 33.4 years, and the average BMI was The mean birth weight ± SD for twin A was ± g, and the mean sonographically estimated fetal weight was ± g. Sonography overestimated the birth weight by an average of g in twin A (95% confidence interval, g). The mean birth weight for twin B was ± g, and the mean sonographically estimated fetal weight was ± g. Sonography overestimated the birth weight by an average of g in twin B (95% confidence interval, ). Figure 1 presents a regression line for the mean absolute percent errors for sonographic predictions of birth weight as a function of increasing BMI in twin A. The regression line has a positive slope of 0.16, indicating that the mean absolute percent error of sonographic prediction of birth weight in twin A increased 0.16 percentage points per unit increase in BMI. This did not reach statistical significance (P =.1278). Figure 2 represents twin B with a similar positive regression line for the absolute percent errors of sonographic predictions of birth weight as a function of increasing BMI. The regression line has a positive slope of 0.36, indicating that the mean absolute percent error of sonographic prediction of birth weight in twin B increased 0.36 percentage points per unit increase in BMI. This did reach statistical significance (P =.0037). We calculated a sample size as a post hoc analysis. In retrospect, with 194 women, we had 80% power to detect a 0.28 percentage point increase in the absolute percent error per unit increase in BMI. Tables 2 and 3 present the mean differences in grams, mean absolute percent errors, and mean absolute errors in grams for each BMI category for twins A and B, respectively. The mean absolute percent errors for twin A were 6% for patients with a BMI of less than 25 and 9% for those with a BMI of greater than 30. The mean absolute percent errors for twin B were 6.7% for patients with a BMI of less than 25 and almost 11.7% for those with a BMI of greater than 30. There was a statistically significant increasing trend for the mean absolute percent error and mean absolute difference in grams with increasing BMI for both twin A and twin B measurements. In addition, when ANOVA testing was used to look at the differences between each BMI category, the following was noted. There was a statistically significant pair-wise comparison for mean absolute differences between BMI categories of less than 25 and greater than 30 for twin A (P =.0133). For twin B, there was a statistically significant pair-wise comparison for mean abso- Figure 1. Accuracy of sonography for predicting birth weight as a function of BMI in twin A. Table 1. Maternal Demographic Data Collected (n = 194) Parameter Value Average maternal age ± SD, y ± 5.72 Gestational age at delivery ± SD, wk 34.7 ± 2.63 Average BMI ± SD ± 4.85 BMI <25, n 25 BMI , n 87 BMI 30, n 82 Chorionicity, n Dichorionic twin gestation 175 Monochorionic twin gestation 19 J Ultrasound Med 2009; 28:

4 Increased Body Mass Index and Estimated Fetal Weight in Twins Figure 2. Accuracy of sonography for predicting birth weight as a function of BMI in twin B. lute percent errors and mean absolute differences between BMI categories of less than 25 and 25 to 29.9 (P =.0413 and.015, respectively) and categories of less than 25 and greater than 30 (P =.0097 and ). Because of the possibility that the accuracy of sonographic estimation of fetal weights may have been affected by gestational age, we analyzed the data on the basis of gestational age groups for each twin. Although our study found no significant trend in the mean absolute percent error for increasing gestational age, we did find that the mean absolute gram difference did increase significantly for each twin with increasing gestational age, which can be seen in Tables 4 and 5. Therefore, for twin A, the mean absolute difference in grams increased from 45 g at less than 28 weeks to greater than 200 g at greater than 36 weeks, which was statistically significant. For twin B, the mean absolute difference in grams increased from 62 g at less than 28 weeks to 258 g at greater than 36 weeks, which was also statistically significant. For ANOVA testing between the Table 2. Sonographic Prediction of Fetal Weight by Increasing BMI Category for Twin A Mean Difference Mean Absolute Error Mean Absolute Difference BMI Category n (SD) [±2 SD], g (SD) [± 2 SD], % (SD) [± 2 SD], g < (169.82) 6.04 (4.8) (121.44) [ 308.2, ] [0, 15.63] [0, ] (210.45) 7.63 (6.76) 165 (142.46) [ 361.7, ] [0, 21.15] [0, ] (263.13) 9.12 (7.63) (171.18) [ 465.1, ] [0, 24.68] [0, ] P, test for linear trend a.0075 a P, overall ANOVA a a Statistically significant (P <.05). Table 3. Sonographic Prediction of Fetal Weight by Increasing BMI Category for Twin B Mean Difference Mean Absolute Error Mean Absolute Difference BMI Category n (SD) [±2 SD], g (SD) [± 2 SD], % (SD) [± 2 SD], g < (152.75) 6.7 (5.92) (116.8) [ , ] [0, 18.53] [0, ] (253.83) (8.2) (185.71) [ , ] [0, 26.99] [0, 590.9] (280.5) (9.11) (183.44) [ , ] [0, 29.91] [0, ] P, test for liner trend a.0057 a P, overall ANOVA a.0095 a a Statistically significant (P <.05). 304 J Ultrasound Med 2009; 28:

5 Gandhi et al different gestational categories, there were statistically significant mean absolute differences for both twins in the following categories: less than 28 versus greater than 36 weeks (twin A, P =.148; twin B, P =.0179), 28 to 32 versus 32 to 36 weeks (twin A, P =.0233; twin B, P =.0123), and 28 to 32 versus greater than 36 weeks (twin A, P =.0017; twin B, P =.002). In addition, twin B had a statistically significant mean absolute difference for less than 28 versus 28 to 32 weeks (P =.0459). Figures 3 and 4 show the percentages of cases in which the fetal weight estimations were within 5%, 10%, 15%, and greater than 15% of the actual birth weights of the twins. Almost 90% of twin A weight estimates in the BMI category of greater than 30 were within 15% of the actual birth weights, and almost two-thirds were within 10%. In twin B, approximately two-thirds of the weight estimates in the BMI category of greater than 30 were within 15% of the actual birth weights, and almost half of the estimates were within 10%. We also investigated whether there was a difference in accuracy depending on the time of the sonographic examination and the day of delivery because we used up to a 6-day sonography-to-delivery interval and found there was no significant difference for either twin on the basis of the number of days between sonography and delivery in our cohort. In addition, we also calculated intertwin discordance. The prevalence of intertwin discordance in our sonography cohort was 15% (n = 30) with sensitivity of 40%, specificity of 98%, a positive predictive value (PPV) of 82%, and a negative predictive value (NPV) of 88%. We found no appreciable change between the sensitivity, specificity, PPV, and NPV after dividing the patients into their respective BMI groups, although the interpretations of these calculations were limited because the numbers of intertwin discordance decreased substantially when the groups were divided up. Finally, we investigated whether there was any difference in intertwin discordance accuracy based on chorionicity in twin gestations. These results showed no significant difference between dichorionic and monochorionic twins, although this was likely due to the small sample size of 19 monochorionic twins. Our data show that sonography estimated the mean absolute intertwin discordance within 4% to 6% when compared with actual the twin discordance. Figure 3. Comparison of the distribution of the absolute percent error according to maternal BMI category for twin A. Discussion Twin gestations represent 3% of all live births in the United States today, which is a 70% increase since This study evaluated the effect of increasing BMI on the accuracy of sonographically predicted estimated fetal weight in twin gestations. Obstetric providers frequently use sonographically predicted estimated fetal weight to direct management of twin gestations in the antepartum period in addition to determining the best mode of delivery for these patients. Figure 4. Comparison of the distribution of the absolute percent error according to maternal BMI category for twin B. J Ultrasound Med 2009; 28:

6 Increased Body Mass Index and Estimated Fetal Weight in Twins Our study shows that sonography remains an accurate method for estimating fetal weight in all twin pregnancies. We did find a significant decrease in the accuracy of birth weight predictions as maternal BMI increased, but the mean absolute percent error of birth weight estimations remained below 15% for most pairs of twins. We found that sonography was more accurate at predicting estimated fetal weight for twin A versus twin B as BMI increased, with an average mean average percent error of 9.1% for twin A versus 11.6% for twin B at a BMI of greater than 30. Both of these mean absolute percent errors were significantly higher than the mean absolute percent errors of 6% to 7% for both twins in patients with a BMI of less than 25. Our study did show that the percentage of fetuses with actual birth weights within 5% to 10% of the sonographically predicted weight decreased as the BMI increased. In twin A, the percentage of fetal estimates within 10% of actual birth weights decreased from 80% to 65%. This was similar to the findings of Field et al, 10 who noted that twothirds of predictions for singleton fetuses in patients with a BMI of greater than 29 were within 10%. In twin B, the percentage of fetal estimates within 10% of actual birth weights decreased from 76% to 48%. Both twins had 70% or greater accuracy for fetal estimates within 15% of actual birth weights. It is, however, clinically important to note that our study indicated that there was a decrease in accuracy with increased BMI in twin gestations, especially where twin B is concerned. In addition, the difference between sonographically estimated fetal weight and actual birth weight increases with increasing gestational age for both twins regardless of BMI. This may particularly be relevant with regard to antepartum management decisions, primarily regarding the mode of delivery of the nonvertex second twin. Table 4. Sonographic Prediction of Fetal Weight Based on Gestational Age for Twin A Gestational Age Mean Difference Mean Absolute Error Mean Absolute Difference at Sonography, wk n (SD) [±2 SD], g (SD) [± 2 SD], % (SD) [± 2 SD], g (55.13) 5.88 (4.22) 45 (38.39) [ 82.36, ] [0, 14.32] [0, ] (106.77) 6.99 (4.41) (63.67) [ , ] [0, 15.81] [0, ] (229.42) 8.37 (7.62) (150.54) [ , ] [0, 23.61] [0, ] > (271.86) 8.08 (6.78) (178.23) [ , ] [0, 21.64] [0, 574.5] P, test for linear trend a P, overall ANOVA a a Statistically significant (P <.05). Table 5. Sonographic Prediction of Fetal Weight Based on Gestational Age for Twin B Gestational Age Mean Difference Mean Absolute Error Mean Absolute Difference at Sonography, wk n (SD) [±2 SD], g (SD) [± 2 SD], % (SD) [± 2 SD], g (91.17) 10.6 (12.31) 62.2 (63.14) [ 205.2, 245.6] [0, 35.22] [0, ] (129.26) 8.87 (5.84) (76.53) [ 38.14, ] [50, 20.56] [0, ] (259.42) (8.73) (185.23) [83.99, ] [0, 28.34] [0, ] > (290.8) (8.62) (190.81) [72.72, ] [0, 27.82] [0, ] P, test for linear trend a P, overall ANOVA a a Statistically significant (P <.05). 306 J Ultrasound Med 2009; 28:

7 Gandhi et al Several previous studies looking at intertwin discordance used a value of 25% secondary to the increased perinatal morbidity and mortality associated with this discordance. 14 Prior studies investigating intertwin discordance have shown various degrees of sensitivity and specificity. 12,14 16 Gernt et al 12 calculated the specificity, sensitivity, PPV, and NPV for intertwin birth weight discordance of 25% or greater in a cohort of 192 patients with twin gestations. Their calculations were similar to ours (ie, sensitivity of 55%, specificity of 97%, PPV of 82%, and NPV of 91%), and their prevalence was similar at 17% for actual intertwin discordance of greater than 25%. 12 Of note, their cohort of patients included those who had a sonographic examination within 16 days of delivery, of which 113 patients had an examination within 7 days, whereas our cohort of 194 patients all had their sonographic examinations within 6 days of delivery. Limitations to this study included the retrospective analysis used to obtain our data. Potential confounding variables included the different degrees of experience between sonographers. However, the physicians reading the images for quality were all board-certified maternal-fetal medicine specialists. Because most physician practices today involve multiple sonographers, our data may in fact be more consistent with current medical practice. Several previous articles have hypothesized why there is no significant decrease in accuracy for sonographically estimated fetal weight in singleton pregnancies in patients with an increased BMI. Previous studies have suggested that the BMI does not affect the accuracy of sonographically predicted estimated fetal weight secondary to the fact that fetal measurements (biparietal diameter, abdominal circumference, and femur length) rely on bony markers, which are bright and echogenic and can usually be seen through layers of fat. 10 In our study we found that the estimations for twin B were less accurate than those for twin A. We hypothesize that as the presenting twin, twin A can sometimes be more clearly visualized because it is lower in the pelvis; therefore, a large pannus can sometimes be lifted to allow clearer visualization of biometric markers. As the superior twin, twin B lies beneath a larger density of fat, which cannot be lifted and may obscure visualization to a greater degree. It may be of interest to measure the distance between the ultrasound probe and the uterus to see whether there is a difference in the accuracy based on this measurement instead of the BMI. With the advent of modern fertility technology, the rise in twin gestations in routine obstetric practices has increased substantially. We also know that obesity is on the rise as well. On the basis of our data, we conclude that the accuracy of sonographically predicted estimated fetal weight does decrease significantly in twin gestations with increasing BMI, and the impact of this effect is magnified in twin B. References 1. Chien PFW, Owen P, Khan K. Validity of ultrasound estimation of fetal weight. Obstet Gynecol 2000; 95: Shamley KT, Landon MB. Accuracy and modifying factors for ultrasonographic determination of fetal weight at term. Obstet Gynecol 1994; 84: Benacerraf BR, Gelman R, Frigoletto FD Jr. Sonographically estimated fetal weights: accuracy and limitation. Am J Obstet Gynecol 1988; 159: Pollack RN, Hauer-Pollack G, Divon MY. Macrosomia in postdates pregnancies: the accuracy of routine ultrasonographic screening. Am J Obstet Gynecol 1992; 167: Nahum GG, Stanislaw H. Ultrasonographic prediction of term birth weight: how accurate is it? Am J Obstet Gynecol 2002; 188: Jensen OH, Jenssen H. Prediction of fetal weights in twins. Acta Obstet Gynecol Scand 1995; 74: Centers for Disease Control and Prevention. Overweight and Obesity. Atlanta, GA: Centers for Disease Control and Prevention. index.htm. Accessed October Wolfe HM, Sokol RJ, Matier SM, Zador IF. Maternal obesity: a potential source of error in sonographic prenatal diagnosis. Obstet Gynecol 1990; 76: Catanzarite V, Quirk JG. Second-trimester ultrasonography: determinants of visualization of fetal anatomic structures. Am J Obstet Gynecol 1990; 163: Field NT, Piper JM, Langer O. The effect of maternal obesity on the accuracy of fetal weight estimation. Obstet Gynecol 1995; 86: Farrell T, Holmes R, Stone P. The effect of body mass index on three methods of fetal weight estimation. BJOG 2002; 109: Gernt OR, Mauldin JG, Newman RB, Durkalski VL. Sonographic prediction of twin birth weight discordance. Obstet Gynecol 2001; 97: J Ultrasound Med 2009; 28:

8 Increased Body Mass Index and Estimated Fetal Weight in Twins 13. Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacher F, Kirmeyer S. Births: Final Data for National Vital Statistics Report. Vol 55. No 1. Atlanta, GA: Centers for Disease Control and Prevention; Blickstein I, Manor M, Levi R, Goldschmidt R. Is intertwin birth weight discordance predictable? Gynecol Obstet Invest 1996; 42: Chamberlain P, Murphy M, Comerford FR. How accurate is antenatal sonographic identification of discordant birth weight in twins? Eur J Obstet Gynecol Reprod Biol 1991; 40: Caravello JW, Chauhaun SP, Morrison JC, Magann EF, Martin JN, Devoe LD. Sonographic examination does not predict twin growth discordance accurately. Obstet Gynecol 1997; 89: J Ultrasound Med 2009; 28: