382: Differentiating newborns with birth weight 3 percentiles for gestational age: secondary-analysis of a randomized clinical trial (DIGITAT)

Neonatal morbidity and mortality with growth restricted fetuses is higher among those 3% for gestational age (GA). Sonographic estimated fetal weight is poor at detecting those 3% for GA. The DIGITAT (Disproportionate Intrauterine Growth Intervention Trial At Term; Boers et al BMJ 2010;341:c7087) randomized 650 women with suspected IUGR (estimate fetal weight or abdominal circumference <10% for GA) to induction of labour vs expectant monitoring. We considered 32 potential prognostic markers. Any missing data was imputed using multiple imputation. A prediction model was developed using multivariable logistic regression with backward selection strategy. Model performance was quantified with respect to discrimination (area under ROC curve, AUC). Internal validation was done by bootstrapping. Data on all 650 women randomized in DIGITAT trial are used for this secondary analysis and among them 22% (140) delivered newborns with weight <3% and 510 (78%) above it. Of the 32 biomarkers studied, 5 ultrasound makers (estimated fetal weight, biparietal diameter percentile, head circumference percentile, abdominal circumference percentile, pulsatility index of middle cerebral artery) as well as a history of miscarriage and hypertensive disease were predictive for newborns < 3% for GA. AUC of the model was 0.79 (95% confidence intervals, 0.75 to 0.84). The shrinkage factor for the model coefficients is 0.83 and the validated AUC is 0.77 which shows that the model is valid internally. In a cohort of small for gestational age children, we identified variables that can differentiate between newborns 3% for GA. Awaiting external validity, this model may improve our ability to identify IUGR at risk of morbidity and mortality.