Objective: To estimate the comparative efficiency of first-trimester fetal growth restriction, second-trimester fetal growth restriction, and first-to-second-trimester growth lag for predicting adverse perinatal outcomes. Methods: This is a retrospective cohort study of pregnancies with reliable dating based on last menstrual periods and first-trimester ultrasound examinations. Pregnancies with multiple fetuses, aneuploidy, and major structural anomalies were excluded. Fetal crown-rump lengths at 10-14 weeks, estimated fetal weights based on fetal biometry at 18-22 weeks, and interval growth were measured and converted to gestational age-adjusted Z-scores. The primary outcome was small for gestational age (SGA) at delivery. Secondary outcomes were low birth weight, preterm delivery, stillbirth, and preeclampsia. Receiver-operating characteristics curves were used to identify the optimal definitions of early fetal growth restriction associated with SGA and to compare screening efficiencies. Multivariable logistic regression was used to adjust for confounders. Results: Among 8,978 pregnancies meeting inclusion criteria, 551 (6.5%) neonates were SGA. Crown-rump length Z-score less than -1.0, estimated fetal weights Z-score less than -1.0, and growth Z-score less than -1.0 were identified as the optimal definitions of early fetal growth restriction associated with SGA (adjusted odds ratio 1.41 [95% confidence interval (CI) 1.13-1.74], 3.44 [95% CI 2.85-4.15] and 2.61 [95% CI 2.09-3.25], respectively). The sensitivity and specificity of first- and second-trimester fetal growth restriction for predicting SGA were 21.4% and 83.4%, and 37.2% and 85.5%, respectively. The area under the receiver-operating characteristics curve for second-trimester fetal growth restriction was greater than that for first-trimester fetal growth restriction and first-to-second-trimester growth lag (0.70 compared with 0.59 and 0.66, P<.001). Conclusion: Second-trimester fetal growth restriction is superior to first-trimester fetal growth restriction and first-to-second-trimester growth lag for predicting SGA.