Predicting small for gestational age infants: is it time to update the Hadlock model?


      Prediction of fetal growth restriction (FGR) and small for gestational age (SGA) at birth has relied on an estimated fetal weight (EFW) less than the 10th percentile for gestational age. However, previously reported work showed disappointing performance. Our hypothesis is that rather than using the Hadlock EFW growth curve (the most widely used ultrasound measure) to directly predict SGA, a derived measure could generate a better performing predictive model.

      Study Design

      We reviewed over 7,000 anatomy (76805 and 76811) ultrasounds in a single institution between 1/1/2016 and 5/1/2021. Multiples were excluded. Those with complete data for follow up growth assessment at 28 and 32 weeks and birth information were included for a total of 1,402 patients. EFW was calculated from the 28 and 32 week ultrasounds using the Hadlock growth curve. Weight Z-scores were calculated using the Fenton 2013 growth curve for newborns for the 28 and 32 week EFWs as well as for the actual birthweight. SGA at birth was defined as an actual birth weight less than the 10th percentile for gestational age (i.e., birth weight Z-score less than -1.282). The relationship between prenatal ultrasound Z-scores and actual birth weight Z-score or SGA at birth was analyzed by linear and logistic regression, respectively.


      Our overall population rate of SGA was 5.6%. EFW Z-scores calculated at both 28 and 32 weeks had a strong linear correlation with Z-score at birth (p < 0.001 for both; Figure 1). EFW Z-score at 32 weeks predicted SGA at birth with an AUC of 0.908; at 28 weeks, our model predicted SGA with an AUC of 0.863 (Figure 2).


      By using measures derived from the EFW at 28 and 32 weeks, we have generated predictive models of SGA at birth with improved performance. New measures are important as patient population changes (increased obesity, diabetes) from the time when the growth charts were created. Our models perform well to predict either a birth weight percentile (via linear regression) or SGA (via logistic regression) and will be useful to aid in clinical decision-making for both prenatal and postnatal care.
      Figure thumbnail fx1
      Figure thumbnail fx2