Developmental subgroups and co-occurring conditions among 5817 minimally speaking autistic youth: a latent class analysis of the SPARK cohort

Abstract Background Approximately 30% of autistic children remain minimally speaking beyond kindergarten, using few or no spoken words. Although often studied as a single group, these children show a wide range of social, motor, and nonverbal cognitive abilities. Few studies have systematically examined this heterogeneity. Characterizing minimally speaking autistic subgroups is an important first step toward understanding the genetic and neurobiological mechanisms underlying different spoken language outcomes in autism. Methods To fill this gap, we conducted a cross-sectional study of 5817 minimally speaking autistic youth aged 5–17 years (76.8% male) who participated in the Simons Powering Autism Research for Knowledge national study. We estimated latent class models separately for children (5–11 years, n = 4682) and adolescents (12–17 years, n = 1135) using four indicators: everyday spoken language level, social interaction difficulty, number of delayed early motor milestones, and nonverbal cognitive ability. We compared classes on sociodemographic characteristics, co-occurring conditions, and total interventions received using chi-square, Fisher exact, and Welch tests with false discovery rate correction. Results We identified three child subgroups and two adolescent subgroups, ranging from lower to higher skills across all four domains. Strengths in one domain generally accompanied strengths in others. Within subgroups, early motor development was a relative strength in lower-skill groups and social interaction in higher-skill groups. Among children, lower-skill subgroups had a higher prevalence of atypical head growth, seizures or epilepsy, obsessive-compulsive disorder, sleep disorders, and feeding or eating disorders (sleep disorder, higher- vs intermediate-skill child subgroup: odds ratio 0.54, 95% confidence interval (CI) 0.46–0.63) and also received more interventions (lower- vs higher-skill child subgroup: 1.9 more, 95% CI 1.7–2.2). Limitations All latent class indicators were derived from caregiver-report questionnaires, which may be subject to recall bias. Although our results show high split-half agreement of class assignments, they still need to be replicated in an independent cohort of autistic children. Conclusions Minimally speaking autistic youth show clinically meaningful heterogeneity in their developmental characteristics, co-occurring conditions, and service use. Future studies should examine the extent to which underlying genetic etiologies contribute to these subgroup differences.