Estimation of fetal weight in twins: a new mathematical model

Objectives Evaluation of new mathematical formula (Femur 4) derived from a twin population to estimate fetal weight in twins using ultrasound. Compari son of Femur 4 is with conventional mathematical models. Design Retrospective analysis of ultrasonic measurements of 297 twin babies from 24 to 40 weeks of gestation who were born within 10 days of ultrasoun d examination. Setting Aberdeen Maternity Hospital. Methods With ultrasonic measurements obtained from twin babies, estimated f etal weight was calculated using the mathematical models of Campbell, Shepa rd and Hadlock. The calculations were repeated for the model of Femur 4. Al l models were compared against Femur 4. Results The coefficient of determination of the linear regression between t he actual and predicted weight was highest for Femur 4 (0.852). Femur 4 had the highest proportion of babies with estimated weights within 10% of actu al birthweight (71.4%). in babies who weighed between 2000 and 3000 g, Femu r 4 had the least systematic and random error of -1.69 and 8.96, respective ly. For babies below the 10th centile for weight, Femur 4 had comparable po sitive and negative predictive values of 76.0% and 92.3%, respectively. Fem ur 4 was equally poor at predicting growth discordancy with positive and ne gative predictive values of 70.0% and 86.5% only. Conclusion Femur 4 requires measurements of femur length and abdominal circ umference only, hence avoiding the need to obtain difficult head measuremen ts which is a common problem in twins. It is a good model for estimation of fetal weight in twins. However, prediction of growth discordancy remains p roblematic.