Comparison of the Hadlock and Intergrowth-21^st formulae in predicting birthweight centiles for low-risk pregnancy; a secondary analysis of data

Abstract

Objective: To compare the accuracy of the Hadlock to Intergrowth-21st formulae in predicting birthweight for low-risk pregnancies.

Methods: A secondary analysis of data from the Routine Third Trimester Ultrasound (ROTTUS) trial was undertaken for 284 low-risk  pregnancies between 36 weeks and 37 weeks and 6 days gestation. The estimated fetal weight centiles using Hadlock, and  Intergrowth-21^st formulae were computed, and the accuracy in predicting birthweight centile compared using ROC curve analysis and  absolute percentage error. Logistic regression model was constructed based on ±15% as the dependent variable, to determine factors  affecting the accuracy of either method.

Results: The proportion of fetuses within 15% of the birth weight was 78.2% and 62.7%  (p<0.0001) for Hadlock and Intergrowth-21st methods, area under the curve =0.8209 (CI 0.74347-0.89838) and 0.872 (CI 0.80865-0.93544)  respectively (p<0.039). The mean absolute percentage error was -7.2 and -10.1 (p<0.0001) for Hadlock and Intergrowth-21st respectively.  The accuracy of Intergrowth-21^st method was influenced by the ultrasound-to-delivery interval.

Conclusion: Overall, the Hadlock  formula was more accurate than Intergrowth 21^st in predicting birthweight centile for low-risk pregnancies, except in large for  gestational age fetuses. Both methods demonstrated a low sensitivity and specificity.