Objective: To find a multivariate model for predicting small-for-gestational age newborns at 36 weeks' gestation by using clinical, biochemical and ultrasound measurements.
Materials and methods: We evaluated 564 low-risk pregnant women and recorded maternal age, maternal body mass index, maternal mean blood pressure, soluble fms-like tyrosine kinase-1 (multiples of the median), placental growth factor (multiples of the median), soluble fms-like tyrosine kinase-1/placental growth factor ratio, estimated fetal weight centile and mean uterine artery pulsatility index at 36 weeks. Binary logistic regression was used. Statistical significance was set at 95% level (p < 0.05).
Results: We found three multivariate models showing relatively small differences in predictive capability. Model 1 only included estimated fetal weight centiles (area under the curve [AUC] 0.86; R2 = 0.42; p < 0.0001), Model 2 estimated fetal weight centiles and placental growth factor (multiples of the median) (AUC 0.87; R2 = 0.44; p < 0.0001) and Model 3 estimated fetal weight centiles, placental growth factor (multiples of the median) and mean uterine artery pulsatility index (AUC 0.88; R2 = 0.45; p < 0.0001).
Conclusion: Small-for-gestational age at delivery may be predicted by using a multivariate formula. The inclusion of parameters other than estimated fetal weight centile at 36 weeks' gestation modestly improves the predictive capability of the model. Clinical decisions should consider whether or not these slight differences deserve a change in current strategies.
Keywords: Small-for-gestational age; mean uterine artery pulsatility index; placental growth factor; soluble fms-like tyrosine kinase-1; third trimester scan.