In most gestational diabetes mellitus (GDM) studies, cohorts have included women combined into study populations without regard to whether hyperglycemia was present earlier in pregnancy. In this study we sought to compare perinatal outcomes between groups: women with early GDM (EGDM group: diagnosis before 20 weeks but no treatment until 24-28 weeks if GDM still present), with late GDM (LGDM group: present only at 24-28 weeks), and with normoglycemia at 24-28 weeks (control subjects). This is a secondary analysis of a randomized controlled treatment trial where we studied, among women with risk factors, early (<20 weeks' gestation) GDM defined according to World Health Organization 2013 criteria. Those receiving early treatment for GDM treatment were excluded. GDM was treated if present at 24-28 weeks. The primary outcome was a composite of birth before 37 weeks' gestation, birth weight ≥4,500 g, birth trauma, neonatal respiratory distress, phototherapy, stillbirth/neonatal death, and shoulder dystocia. Comparisons included adjustment for age, ethnicity, BMI, site, smoking, primigravity, and education. Women with EGDM (n = 254) and LGDM (n = 467) had shorter pregnancy duration than control subjects (n = 2,339). BMI was lowest with LGDM. The composite was increased with EGDM (odds ratio [OR] 1.59, 95% CI 1.18-2.12)) but not LGDM (OR 1.19, 95% CI 0.94-1.50). Induction of labor was higher in both GDM groups. In comparisons with control subjects there were higher birth centile, higher preterm birth rate, and higher rate of neonatal jaundice for the EGDM group (but not the LGDM group). The greatest need for insulin and/or metformin was with EGDM. Adverse perinatal outcomes were increased with EGDM despite treatment from 24-28 weeks' gestation, suggesting the need to initiate treatment early, and more aggressively, to reduce the effects of exposure to the more severe maternal hyperglycemia from early pregnancy.