Clinical predictors of 2-year outcome in preschoolers with ASD were studied in a population-based group of very young children with ASD (n = 208). Children who gained the most (n = 30) and lost the most (n = 23), i.e., increased or decreased their adaptive functioning outcome according to the Vineland Composite Score between study entry (T1) and follow-up (T2), 2 years later were compared. Individual factors that differed significantly between the two outcome groups were cognitive level, age at referral, not passing expected milestones at 18 months, autistic type behavior problems and regression. However, logistic regression analysis showed that only cognitive level at T1 (dichotomized into IQ < 70 and IQ = 70) made a unique statistically significant contribution to outcome prediction (p = <0.001) with an odds ratio of 18.01. The findings have significant clinical implications in terms of information at diagnosis regarding clinical prognosis in ASD.