To develop optimal first-trimester algorithms for the prediction of early and late fetal growth restriction (FGR). This was a prospective cohort study of singleton pregnancies undergoing first-trimester screening. FGR was defined as an ultrasound estimated fetal weight < 10(th) percentile plus Doppler abnormalities or a birth weight < 3(rd) percentile. Logistic regression-based predictive models were developed for predicting early and late FGR (cut-off: delivery at 34 weeks). The model included the a-priori risk (maternal characteristics), mean arterial pressure (MAP), uterine artery pulsatility index (UtA-PI), placental growth factor (PlGF) and soluble fms-like tyrosine kinase-1 (sFlt-1). Of the 9150 pregnancies included, 462 (5%) fetuses were growth restricted: 59 (0.6%) early and 403 (4.4%) late. Significant contributions to the prediction of early FGR were provided by black ethnicity, chronic hypertension, previous FGR, MAP, UtA-PI, PlGF and sFlt-1. The model achieved an overall detection rate (DR) of 86.4% for a 10% false-positive rate (area under the receiver-operating characteristics curve (AUC): 0.93 (95% CI, 0.87-0.98)). The DR was 94.7% for FGR with pre-eclampsia (PE) (64%of cases) and 71.4% for FGR without PE (36%of cases). For late FGR, significant contributions were provided by chronic hypertension, autoimmune disease, previous FGR, smoking status, nulliparity, MAP, UtA-PI, PlGF and sFlt-1. The model achieved a DR of 65.8% for a 10% false-positive rate (AUC: 0.76 (95% CI, 0.73-0.80)). The DR was 70.2% for FGR with PE (12%of cases) and 63.5% for FGR without PE (88%of cases). The optimal screening algorithm was different for early vs late FGR, supporting the concept that screening for FGR is better performed separately for the two clinical forms. Copyright © 2016 ISUOG. Published by John Wiley & Sons Ltd.