Objective: To determine whether the use of customized curves (CC) allows better detection of large- (LGA) or small-for-gestational age (SGA) infants at risk of adverse perinatal morbidity than non-CC in women with diabetes mellitus (DM). Material and Methods: A model of CC was applied to all infants of diabetic mothers (IDM) who attended the Hospital Universitario Materno Infantil de Canarias between 2008 and 2011. We compared perinatal outcomes of IDM classified as LGA or SGA by non-CC versus CC. Results: One of 4 LGA was appropriate for gestational age (AGA) by CC (false-positive rate: 25%) and 30% of SGA by CC were not identified by non-CC (false-negative rate). False-positive LGA and SGA showed similar perinatal outcomes to AGA infants. The rates of cesarean section, cephalopelvic disproportion, total fetal distress and shoulder dystocia were significantly higher in false-negative LGA than in AGA by CC (p < 0.004, p < 0.02, p < 0.04 and p < 0.04, respectively). Fetal distress was higher in false-negative SGA than in AGA by CC (p < 0.03). Discussion: In pregnancies complicated by DM, the use of CC allowed more accurate identification of LGA and SGA infants at high risk of perinatal morbidity than non-CC.