The objectives of this study were to model more homogeneous subgroups within autism spectrum disorder (ASD) based on early trajectories of core symptoms; and to further characterize these subgroups in terms of trajectories of language, cognition, co-occurring (attention-deficit/hyperactivity disorder [ADHD]–related) traits and clinical outcome diagnosis.

Children (N = 203) referred for possible ASD at ages 1 to 4 years were assessed at three time points at intervals ranging from 9 months to 3 years. Assessments included standardized measures for ASD (Autism Diagnostic Observation Schedule [ADOS]), language (ADOS-language item), nonverbal IQ (NV-IQ; different tests adequate to chronological/mental age), and parent-reported behavioral problems (Infant-Toddler Social and Emotional Assessment, Child Behavior Checklist).

Latent-class growth curve analysis with ADOS total scores led to the identification of three main stable and two small improving groups: a severe–stable group (19.5% of sample)—the only group without considerable language improvement—showed persistent low NV-IQ and marked increase in attention problems over time; a moderate–stable group (21.7%) with below-average increasing NV-IQ; and a mild–stable group (48%) with stable–average NV-IQ and the highest scores on ADHD-related traits, whose ASD outcome diagnoses increased despite stable–low ASD scores. Two groups (each 5.4%) improved: one moved from severe to moderate ASD scores, and the other moved from moderate to mild/nonspectrum scores. Both of these groups improved on language, NV-IQ, and ADHD-related traits.

Results support the high stability of ASD symptoms into various severity levels, but also highlight the significant contribution of non-ASD domains in defining and explaining the different ASD trajectories.