Early language milestones predict later language, but not autism symptoms in higher functioning children with autism spectrum disorders

Abstract

Language ability is a known predictor of outcome in children with autism but plays a more controversial role for higher functioning children with autism spectrum disorders (ASD). We studied the relationship between early language milestones and later structural language, adaptive functioning and autism symptoms in a sample of 76 children (mean age = 9.1 years) with an ASD, using two different language milestones: first phrase by 24 or 36 months. After controlling for age and nonverbal abilities, retrospectively reported early language milestones were predictive of later structural language abilities, measured by a sentence repetition task, and adaptive communication skills, but not autism symptoms or adaptive social skills. Acquisition of phrase speech by 24 months was sensitive to language and communication impairments in our ASD group, the majority (84%) of which had already acquired phrase speech by 36 months of age. Early available and easily collectable milestone data may be a useful marker of later language performance even in higher functioning, verbal children on the autism spectrum. When a detailed assessment of language is not possible, data on early milestones may be useful for identifying children at-risk in clinical settings and for language phenotyping in the laboratory.

Introduction

Language ability is an important predictor of outcome in children with autism (e.g., Howlin et al., 2004, Venter et al., 1992). More controversial is the role language plays for higher functioning children with autism spectrum disorders (HFASD). Language milestones and current language functioning are key elements of the diagnostic distinction between Asperger Disorder and autism (American Psychiatric Association, 2000), but investigators such as Loucas et al. (2008) have found that impaired structural language abilities are not predictive of the triad of social, communication and repetitive behavioral symptoms associated with autism. Although an essential starting point, the current nosology distinguishing autism from Asperger Disorder may have complicated our ability to determine the role of early language development in ASD because many investigators have relied on the DSM-IV-TR definitions of milestone delays (single words after 2 years and communicative phrases after 3 years), which are more lenient than commonly reported expectations for on time acquisition of speech (James, 1990, Sharp and Hillenbrand, 2008, Stott et al., 2002, Tager-Flusberg et al., 2009, Twatchtman-Cullen, 1998).

An alternative to dividing children on the autism spectrum by current DSM-IV TR diagnostic categories is to identify specific cognitive profiles or subtypes (Frith, 2001, Pennington, 2002) within ASD that may help parse the heterogeneity that can confound the search for genetic etiologies (Gottesman & Gould, 2003), and appropriately targeted interventions (Fisher and Happé, 2005, Riggs et al., 2006). For example, a common criticism of the distinction between high functioning autism and Asperger Disorder is that it does not differentiate children in a meaningful way beyond differences driven by intelligence (for reviews, see Matson et al., 2012, Witwer and Lecavalier, 2008; but see Ghazuiddin, 2008). In addition to intelligence, language functions, particularly structural language abilities, are promising candidates for subtyping children with ASD (Dawson et al., 2002, Howlin, 2005), because they are variable, and associated with both genetic differences and distinct intervention approaches (Tager-Flusberg, 2006). Furthermore, by investigating language abilities as a cognitive subtype (or language disability as a comorbidity) within ASD as opposed to embedding language in a diagnosis, we can investigate the specificity of language abilities throughout development, and test three competing hypotheses about early language abilities: they are not meaningful markers of outcome in autism; they are related to language outcome only; they are more generally related to ASD social and communication symptoms and adaptive outcomes.

Tager-Flusberg and Joseph (2003), among others, identify two language subtypes in verbal children with ASD: pragmatic deficits alone, and pragmatic deficits combined with impaired structural language (impaired phonology, vocabulary, syntax and morphology) and delays in the acquisition of language. Szatmari and colleagues (Bennett et al., 2008, Szatmari et al., 2009) found that what they termed “structural language impairment” (StrLI) in children with ASD at 6–8 years of age predicted adaptive and autistic behaviors at 17–19 years of age. Loucas et al. (2008) report that StrLI in ASD did not affect presentation of autistic symptoms during late childhood, but it did relate to adaptive communication outcomes. Thus, several groups associate the presence of StrLI in individuals with ASD with long-term risk, but there is disagreement regarding whether the risk is specific to future language functioning or related to autism symptoms and outcomes more generally.

In addition to addressing the specificity of early language development for predicting later functioning in ASD, a second unanswered question involves what constitutes an effective early marker of StrLI in ASD. Are language milestones, one element associated with StrLI, in and of themselves a predictive marker in ASD? Mayes, Calhoun, and Crites (2001) found that in ASD, acquisition of single words by 24 months and phrase speech by 36 months did not predict autism symptoms or current language abilities. On the other hand, Szatmari, Archer, Fisman, Streiner, and Wilson (1995) found that language milestones were predictive of a range of autism symptoms in children less than six years of age and social reciprocity in older children (mean age = 11 years). Their milestone criteria were spontaneous phrase speech by 36 months and an absence of marked or persistent delayed echolalia, pronoun reversal and neologisms. In a comprehensive longitudinal follow-up study, Bennett et al. (2008) used the same criteria to predict outcome through 15–17 years of age. They report that structural language skills assessed at 6–8 years of age were more predictive of adaptive outcomes and autism symptoms in later adolescence than milestones or language skills at 4–6 years of age. Milestones were, however, at least as strongly correlated with daily living and social adaptive skills as was the presence of StrLI at age 6–8 in two of their three follow-up assessments. While far from conclusive, these investigations indicate the possibility that milestones are related to later symptoms and adaptive outcome.

Further investigation of the relationship between early language milestones and outcome is warranted for many reasons. There is some evidence that easily attained milestone data has predictive power regarding ASD outcomes into adolescence, but there is disagreement regarding exactly what aspects of outcome it predicts. It is also unclear what age at milestone attainment would be most sensitive to later outcome in children with HFASD. Some groups have relied on the DSM-IV-TR milestone criteria of 36 months while 24 months is more commonly used milestone for phrase speech in the study of typical development (James, 1990, Sharp and Hillenbrand, 2008, Stott et al., 2002, Twatchtman-Cullen, 1998), and an NIH working group on language development in ASD (Tager-Flusberg et al., 2009) recommends a range of 18–30 months as the time period in which word combinations are expected. Another reason to investigate milestones is their potential to inform developmental trajectories. As argued by Volkmar, State, and Klin (2009), attempts to identify cognitive subtypes in ASD must address the developmental trajectories of those abilities, for which milestones can be a useful tool. If early language milestones are shown to meaningfully predict later outcomes, they can inform research and treatment. They might offer an easy, early-available method for parsing the heterogeneity in ASD that confounds many inquiries into its biology. For example, language milestones have previously been used as phenotypic markers in autism genetics studies (e.g., Alarcon, Yonan, Gilliam, Cantor, & Geschwind, 2005). Additionally, pediatricians and other practitioners can use milestone information when they are predicting trajectories and as an impetus for early language evaluation and intervention, even for children without intellectual disability.

In this investigation, we query whether the age of attainment of early language milestones in children with ASD predicts better outcome at school age. In an effort to explore the relationship between age of language milestone attainment and later development in higher functioning children with ASD, we compare the predictive value of relatively stringent language milestone attainment to the more lenient milestones described in DSM-IV-TR. Specifically, we investigate parent retrospective reports of early milestones in relationship to later autism symptoms, adaptive functioning, and structural language. Szatmari, Bryson, Boyle, Streiner, and Duku (2003) and Klin et al. (2007) have highlighted that outcome, or the ability to function effectively in mainstream settings, is not solely driven by severity of autism symptoms, but also by adaptive behaviors. Therefore we report on the relationship of early milestones to Vineland Adaptive Behavior scores as well as Autism Diagnostic Observation Schedule scores (ADOS; Lord, Rutter, DiLavore, & Risi, 1999). A sentence repetition task was used as a marker of structural language for two reasons: it captures core grammar, syntax, and morphology, which appear to be the greatest areas of impairment in the ASD with StrLI subtype (Loucas et al., 2008, Tager-Flusberg and Cooper, 1999); and it is described as the best overall marker of communication impairment both in typically developing children and in children with autism (Botting and Conti-Ramsden, 2003, Conti-Ramsden et al., 2001).

In this study, we investigate whether age of language milestone attainment is: (A) not related to outcome at school age; (B) narrowly related to later structural language/communication abilities; or (C) more broadly related to adaptive behavior and autism symptoms in children with HFASD. A second aim of this study is to examine the sensitivity of meeting two different language milestone criteria: first word and phrase speech milestones by 15 and 24 months, or by 24 and 36 months, respectively, in predicting outcomes at school age.