Predictive value of sonographic screening to detect small for gestational age neonates

Introduction: Antenatal screening of small for gestational age (SGA) neonates is a clinical challenge, with many cases going undetected. We aimed to evaluate the predictive value of sonographic screening at second and early third trimester in predicting SGA neonates.

Methods: This was a retrospective cohort study including women with singleton pregnancies who delivered at term in a single referral center in 2014. Two cohorts were included: 1) sonographic estimation of fetal weight (EFW) performed at routine anatomy scan (18-24 weeks) and 2) sonographic EFW at third trimester (28-34 weeks). EFW percentiles were tabulated using Hadlock IV and Kramer formulas. Birth weight (BW) percentiles were determined by Kramer curve. Linear regression was used to generate a prediction model for BW percentiles based on EFW percentiles. Predictive accuracy of multiple EFW percentiles for BW <10th and <5th percentiles were calculated.

Results: In total, 698 and 1039 women were included in the second and third trimester cohorts, respectively. At the second trimester, EFW percentile was significantly yet weakly correlated with BW percentile (r=0.27, p<0.001). The area under the ROC curve for the prediction of BW <10th and <5th percentiles were 0.64 and 0.65, respectively. Negative and positive predictive values of different percentile cut-offs are presented in Table 1. At the third-trimester, EFW percentiles according to both Hadlock and Kramer curves were moderately correlated with BW percentile (r=0.56 and r=0.56, respectively). The area under the ROC curve for prediction of BW <10th and < 5th percentiles was similar, 0.79 for both Hadlock and Kramer curves. The 10th percentile cut-off yielded positive and negative predictive values of 75% and 88.4%, respectively, for the prediction of SGA (Table 2).

Conclusion: While second trimester sonographic screening poorly predicts SGA neonates, early third trimester screening has a moderate predictive accuracy. The traditional cut-off of the 10th percentile is the most appropriate.