Advertisement
Poster session II Clinical obstetrics, diabetes, labor, medical-surgical-disease, physiology/endocrinology, prematurity: Abstracts 237 - 386| Volume 208, ISSUE 1, SUPPLEMENT , S111, January 01, 2013

240: Ultrasound prediction of birthweight in diabetic pregnancies: 3D volumes vs 2D biometry?

      Objective

      Risk of macrosomia and shoulder dystocia in diabetic pregnancies place a premium on prediction of birthweight (BW) in late pregnancy. 2D ultrasound has historically performed poorly. We aimed to estimate if 3D humeral and femoral volumes were superior to, or could improve upon the 2D estimated fetal weight (EFW) prediction of BW.

      Study Design

      We performed a 2-year prospective cohort study of consecutive pregnancies complicated by gestational (GDM) and pre-gestational Type II diabetes (DM-II). Between 33-37 weeks, they underwent 2D ultrasound for estimated weight (EFW) by biometry, as well 3D volumes of the humerus and femur. BW and detailed clinical history was collected. Macrosomia was defined by absolute weight, using ≥ 4000g and ≥ 4500g, as well as BW percentile, using ≥ 90th and ≥ 95th %iles. 2D EFW and 3D volumes were considered both continuously as percentiles using published nomograms, and dichotomously. Multivariable regression was used to adjust for diabetes type and time in weeks between ultrasound and delivery. Receiver operator characteristic curves were used to estimate the predictive ability of the models, and were compared with the c-statistic.

      Results

      Of 176 women, 27 (15.3%) delivered an infant ≥ 4000g, 36 (20.5%) ≥ 90th%ile, and 26 (14.8%) ≥ 95%ile. 2D EFW was most predictive of BW ≥ 95%ile (AUC 0.82), followed by humerus volume (AUC 0.80), and then femur volume (AUC 0.67). The addition of 3D volumes to the 2D predictive model did not improve upon its ability to predict BW.

      Conclusion

      While humoral and femoral volumes are easily obtained during 2D ultrasounds for growth, they do not improve prediction of macrosomia in diabetic pregnancies; 2D biometry alone remains the most predictive tool for BW in diabetic pregnancies.
      Figure thumbnail grr4
      *P values compare AUC for specified model to base model AUC.