Challenges and opportunities in early identification and intervention for children at-risk for autism spectrum disorders

Abstract

This response to both agrees and disagrees with a number of points relative to early identification and intervention for children with autism spectrum disorders (ASD). The authors, too, recognize the difficulties of identifying toddlers with ASD and the complexities of intervening with these children. It is, however, suggested that there are alternatives to choosing to wait until diagnoses for at-risk children are stable and it is believed that there are many potential benefits from intervening, even when stable diagnoses cannot be made at the time, but risk markers are present. Specifically, it is suggested that it is not necessary to differentially diagnose children in efforts to evaluate treatment effects and suggestions are provided for alternative methods. This commentary also acknowledges the importance of effective ASD screening tools, along with expert clinical opinion, to help identify these toddlers with and at-risk for ASD. Both the available literature from other researchers as well as the authors’ own work in these areas are used to make these arguments.

Keywords::

• Autism spectrum disorders
• early identification
• intervention
• screening

Introduction

Camarata (2014) provides an interesting discussion about the state of early identification and intervention in relation to children with autism spectrum disorder (ASD). In the current commentary, we discuss both points of agreement and disagreement in regard to early identification of and early intervention for children with or at-risk for ASD. Our responses are rooted in two areas, our own reading of the literature and the experiences and contributions of our research team that also involves Grace Baranek, Steve Reznick, and Lauren Turner-Brown from the University of North Carolina at Chapel Hill. Specifically, we agree with the core argument made by Camarata regarding the limited evidence in studies with rigorous treatment designs to support the effects of early intervention targeting ASD in young children. However, we also suggest additional perspectives to argue in favour of instituting early intervention. The other points where we disagree with Camarata are related to our belief that: differential diagnosis between children with receptive language disorder and those with ASD can be accomplished when multiple characteristics are used; it may not be necessary to definitively diagnose children with ASD to evaluate the effectiveness of early interventions; that it is also not necessary to differentiate between diagnoses of previously identified categories of spectrum disorders; and finally that early screening efforts are a necessary component of the early identification and intervention process. We will detail our reasons within each of the following sections.

Arguments for early intervention

Camarata's (2014) discussion on the limited evidence for early intervention for children with ASD was based on the review by Warren, McPheeters, Sathe, Foss-Feig, Glasser, and Veenstra-VanderWeele (2011) published in Pediatrics. As Camarata acknowledged, Warren et al. also noted that the studies on the UCLA/Lovaas-based interventions (see the review for individual studies cited) and the Early Start Denver Model (Dawson, Rogers, Munson, Smith, Winter, Greenson, et al., 2010) revealed a number of benefits for some children. Further, they reported the “low and insufficient strength of evidence” should not be interpreted to mean the interventions are not effective, but instead that further research is needed to verify their effectiveness.

Since Warren et al.'s (2011) review, there have been a few additional studies of early intervention showing promising results in this population of children (Carter, Messinger, Stone, Celimli, Nahmias, & Yoder, 2011; Schertz, Odom, Baggett, & Sideris, 2013). These two studies were among eight projects funded by Autism Speaks in 2006 (Autism Speaks Toddler Treatment Network, ASTTN) to test the efficacy of interventions focused on toddlers (younger than 24 months) with or at-risk for ASD, including one from our research team. Our ASTTN pilot study utilized a parent-mediated, home-based responsive teaching intervention with a community sample of children at-risk for ASD (Baranek, Watson, Turner Brown, Field, Crais, Wakeford, & Liittle, 2013a; Baranek, Watson, Turner Brown, Field, Crais, Wakeford, Little & Reznick, 2013b). At-risk status was determined by elevated scores on the First Year Inventory (FYI, Baranek, Watson, Crais, & Reznick, 2003; Reznick, Baranek, Reavis, Watson, & Crais, 2007) in both social-communication and sensory-regulatory domains. The FYI is an ASD parent-report screening tool administered at 12 months of age. Children at high risk were then assessed and randomized to either our Adaptive Responsive Teaching (ART) intervention (and referred to community services) or referred only to community services (CS). Intervention began between 14–17 months of age and lasted 6–8 months (36 1-hour sessions). In this small sample study (11 children in ART, five in CS only), the results indicated significant treatment effects for children in the ART group (compared with the CS group) on the Sensory Experiences Questionnaire (SEQ, Baranek, 1999b), and the Vineland Adaptive Behavior Scales Second Edition's Expressive and Receptive Communication and Socialization sub-scales (VABS-2, Sparrow, Cicchetti, & Balla, 2005); also, the parents in the ART group were less directive on the Maternal Behavior Rating Scales (MBRS, Mahoney, Powell, & Finger, 1986). We are currently working on an Institute of Education Sciences supported efficacy and replication study, with 80 1-year-olds at-risk for ASD currently randomized to ART and/or CS. Additionally, there are several ASTTN projects whose results will be seen in publications in the near future. Thus, from some of the studies included in Warren et al. (2011), and additional studies published since then, we see emerging evidence for the efficacy of early intervention, but we agree there is a critical need for more rigorous methods, replications, expansions, and for controlled studies of the factors that impact outcomes.

An additional reason to support early intervention is the issue of regression or plateauing of skills that some children with ASD evidence during the second year of life (Baird, Charman, Pickles, Chandler, Loucas, Meldrum, et al., 2008; Maestro, Muratori, Cesari, Pecini, Apicella, & Stern, 2006; Werner & Dawson, 2005). From both retrospective (home videos, parent report) or prospective (sibling study) methodologies, there are numerous studies indicating loss or plateauing of social skills and social interest, language and play skills, and infrequently motor skills (Ozonoff, Heung, Byrd, Hansen, & Hertz- Picciotto, 2008a; Ozonoff, Williams, & Landa, 2005; Werner & Dawson, 2005). Ozonoff, Iosif, Baguio, Cook, Moore Hill, Hutman, et al. (2010) reported declines in the frequency of gaze to faces, shared smiles, and vocalizations directed to others among infant siblings who went on to get a diagnosis of ASD. Estimates of the proportion of children with ASD who experience regression or plateauing range from 15–30% (Yirmiya & Charman, 2010); however, in a recent study using retrospective videos, Ozonoff, Iosif, Young, Hepburn, Meagan, Colombi, et al. (2011) reported that 46% of the children with ASD displayed significant loss of social-communication behaviours in the first 2 years. Further, in a paper in preparation (Siller, Turner-Brown, Morgan, & ASTTN, submitted) across the eight ASTTN intervention studies of toddlers, there were significant age effects seen on some sub-scales of both the Mullen Scales of Early Learning (MSEL, Mullen, 1995) and VABS-2, indicating that children below 18 months scored significantly higher than those from 18–23 or 24–29 months. Therefore, as Siller et al. concluded, “effective treatment may merely prevent regression rather than promote developmental gains” (p. 6). However, they also indicated that diagnostic instability may play a factor in these findings.

Finally, in terms of supporting early intervention, the work of our team also argues for the possibility of preventing regression from occurring and ameliorating symptoms that are present in children with or at-risk for ASD. In this work, we have used both retrospective home video analysis (Colgan et al., 2006; Poon, Watson, Baranek, & Poe, 2012; Watson, Crais, Baranek, Dykstra, & Wilson, 2013) and prospective studies (Baranek et al., 2013; Turner-Brown, Baranek, Reznick, Watson, & Crais, 2013). Our retrospective work has documented different gesture trajectories across 9–18 months in infants who later were diagnosed with ASD or other disabilities (Watson et al., 2013). Specifically, for toddlers with other developmental disabilities (DD) and those with typical development (TD), the trajectory across ages for social interaction (SI) and joint attention (JA) gestures is positively sloping. In contrast, the toddlers later diagnosed with ASD have a negative slope across time for SI and JA gestures. This pattern may indicate regression for some of the children with ASD. In similar fashion, our intervention results (Baranek et al., 2013) indicated significant treatment effects for children at-risk for ASD (compared with the community services group). Recall that our children were initially screened at-risk for ASD at 12 months, began intervention between 14–17 months, and were assessed after 6–8 months of intervention. The fact that they displayed improved social and sensory skills suggests one potential impact of the intervention may have been to reduce regression or plateauing of skills for some children. Thus, to delay early intervention may be a “cost” to the children who may otherwise show signs of regression or plateauing and a “cost” to the others who show early ASD markers in potentially ameliorating their symptoms. As noted by Dawson (2008), intervening with at-risk infants before the full emergence of diagnostic symptoms may be more efficacious than intervening when children are displaying the full syndrome, and clearly has implications for prevention.

Distinguishing children with receptive language disorders from those with ASD

When looking for distinguishing features for separating ASD from receptive language disorder, we agree with Camarata (2014) that non-verbal social engagement is key. However, we also suggest four other distinguishing features including the child's restrictive and repetitive behaviours, gesture use, responsiveness to language, and regression. In regard to repetitive and restrictive behaviours (RRB), some data are available to indicate that young children with ASD may be distinguished from those with other DD or TD based on the frequency and type of RRBs they use (Guthrie, Swineford, Nottke, & Wetherby, 2013; Kim & Lord, 2010; Ozonoff, Macari, Young, Goldring, Thompson, & Rogers, 2008b; Wetherby, Watts, Morgan, & Shumway, 2007; Wetherby, Woods, Allen, Cleary, Dickinson, & Lord, 2004). In addition, as documented by Guthrie et al. (2013), some children with ASD increase the number and/or severity of these behaviours over time, whereas in children with TD and those with other DD these behaviours typically decrease. Therefore, RRBs should be another area that clinicians and researchers take into account when differentially diagnosing young children.

An additional set of behaviours that may help distinguish young children with ASD from those with language disorders is gesture use. Retrospective video analysis by our research team (Watson et al., 2013) has demonstrated significant differences between infants later diagnosed with ASD or other types of DD in joint attention gestures between 9–12 months, and joint attention and social interaction gestures between 15–18 months. In addition, less variety of social interaction gestures was seen in 9–12 month olds later diagnosed with ASD (Colgan et al., 2006). The work of Wetherby et al. (2007) similarly indicates toddlers with ASD communicate less than toddlers with other DD for the function of joint attention. Therefore, amount, variety, and functional use of gestures can also help make these important distinctions.

Further, in discussing the difficulties in a differential diagnosis between ASD and language disorder in toddlers, Camarata (2014) suggests, “Both groups are likely to express frustration in the form of tantrums, but will have impaired verbal social skills and be generally unresponsive to verbal input” (p. 8). Although we agree that both groups of children are likely to show frustration and possible tantrum behaviour, our experiences with children with language disorders suggest they are rarely “unresponsive to verbal input”, even though they may not show age-expected comprehension. For example, they may not comply with a request of an adult, but they often respond in some way by looking, turning away, or making negative vocalizations or verbalizations in protest. In contrast, in our experiences, it is the children with ASD who are more likely to be unresponsive to verbal input and do not seem to notice that someone is talking to them or at times that someone else is in the room. Less response to one's own name (or requiring more prompting to respond) has been found to distinguish infants and toddlers who are later diagnosed with ASD from those with other DD or TD in several studies (Baranek, 1999a; Osterling, Dawson, & Munson, 2002; Zwaigenbaum, Bryson, Rogers, Roberts, Brian, & Szatmair, 2005). Parental and clinical reports abound of children with ASD who are unresponsive to language. This lack of responsivity to language leads some parents to report that their child acts as if they can't hear or are deaf, as shown in items endorsed by some parents on the FYI (Baranek et al., 2003) and the MCHAT (Robins, Fein, Barton, & Green, 2001). Thus, we would argue that “unresponsiveness to verbal input” is typically more characteristic of children with ASD than children with receptive language disorders.

Returning to the issue of regression, there are limited reports of children with receptive language disorders having regression. For example, Pickles, Simonoff, Conti-Ramsden, Falcaro, Simkin, Charman, et al. (2009) documented that regression occurred in only 1% of children with specific language impairment compared with 15% of children with ASD. Although this study was of children older than those discussed in this commentary, parental interview is often used to determine regression for both groups. The ASD literature has numbers of studies showing regression by both parental report and prospective study (Baird et al., 2008; Maestro et al., 2006; Ozonoff et al., 2005, 2010; Werner & Dawson, 2005). Thus, parental report of any type of regression can be another helpful factor to aid in differential diagnosis.

In sum, in addition to using social engagement, measures of child functioning including the child's restrictive and repetitive behaviours, gesture use, responsiveness to language, and regression or plateauing of skills may help make distinctions between children with ASD and those with receptive language disorders. These behaviours can be observed and measured within most assessment settings and included in parental interviews discussing the child's history.

Diagnosis of ASD: Broad or specific

We also agree with Camarata (2014) in recognizing the difficulty of diagnosing toddlers with ASD, particularly when considering behaviours that have their roots in typical development (e.g., some types of repetitive movements) or others that may overlap with intellectual disabilities (e.g., delays in play development). Further, there is a good deal of agreement in the field that differentiating between Autism and Pervasive Developmental Disorder-Not Otherwise Specified (PDD-NOS), or PDD-NOS and Asperger Syndrome is difficult (American Psychiatric Association, DSM-5, 2013), and, due to these issues, the CDC reports rates using all the PDD categories combined. Therefore, the DSM-5 argues for the use of a spectrum that does not distinguish between sub-categories of individuals with ASD. The DSM-5 further urges clinicians to focus on the severity of the ASD symptoms, the pattern and onset of the symptoms, aetiological factors, cognitive abilities, and other contributing factors (e.g., genetic or medical conditions), and to use diagnostic specifiers to detail these areas. This is the strategy our research team has taken in assessing and intervening with at-risk toddlers across a range of measures including social communication, sensory, ASD features, developmental, adaptive, social–emotional, parent–child interactional, and parental stress. Once detailed profiles of these children's scores and characteristics have been identified, we look for treatment effects across these measures at follow-up assessments. Therefore, we are not trying to distinguish types of ASD, but symptoms and features relevant to ASD, that can be compared pre- and post-treatment.

In support of the DSM-5 ASD spectrum approach, we argue that a growing body of rigorous research indicates that stable markers of ASD can be seen in young children as early as 12 months of age (Baranek, 1999a; Colgan et al., 2006; Landa, Holman, & Garret-Mayer, 2007; Osterling et al., 2002; Zwaigenbaum et al., 2005), and accurate and stable diagnosis of ASD can be made for many children by 24 months. Work by Chawarska, Klin, Paul, Macari, & Volkmar (2009) and Cox, Klein, Charman, Baird, Baron-Cohen, Swettenham, et al. (1999) in high risk samples indicated 100% stability in ASD diagnosis made before the age of 2 years. Similarly, Lord, Risi, Dilavore, Shulman, Thurm, and Pickles (2006) documented 90% stability up to 9 years of age, with most of the change related to increasing certainty of classifications moving from ASD/PDD-NOS to Autism. In addition, the recent work of Guthrie et al. (2013) following a community sample of 82 children screened at-risk at even younger ages showed high stability for a large sub-set of the children. The children were seen for a comprehensive diagnostic evaluation at a mean age of 19.39 months (range = 15–24 months) and again at 38.69 months. There was 100% stability across time in diagnosis of ASD for 68% of the children. For 15% of the children, ASD was initially ruled out (either DD or TD) and for 17% diagnosis was deferred because the clinicians could not confirm or rule out ASD. When these deferred children were evaluated at the later age, three were diagnosed with ASD, four with DD, six were identified as typically-developing, and one younger sibling of a child with ASD had diagnosis once again deferred. Thus, the overall stability of diagnosis of the participants, accounting for the sub-set of children for whom diagnoses were deferred, was 82%. Therefore, for some children a clear diagnosis can be made and will likely remain stable, whereas for others it may be wiser to wait and re-evaluate prior to making a diagnosis (but not necessarily to delay intervention in cases where clear risk markers are present).

The characteristics of the evaluation process used by Guthrie et al. (2013) also needs specification, as they used very experienced clinicians and a two-tier process that included broadband and autism screening along with a comprehensive diagnostic evaluation. Further, the evaluation included observations in both the child's home and the clinic. Although this type of labour-intensive evaluation process may not be practical for community use, there are tools and techniques that can be utilized to gain perspectives of the child at home (e.g., MacArthur-Bates Communicative Development Inventory, Fenson, Marchman, Thal, Dale, Reznick, & Bates, 2007; VABS-2); as screening tools (e.g., Communication and Symbolic Behavior Scales [CSBS] Infant Toddler Checklist, Wetherby & Prizant, 2002b) and to gain information on the child's social and play behaviours (e.g., CSBS Developmental Profile; Wetherby and Prizant, 2002a), along with the Autism Diagnostic Observation Schedule for Toddlers (ADOS-T, Luyster, Gotham, Guthrie, Coffing, Petrak, Pierce, et al., 2009).

Therefore, in line with the DSM-5 and some current literature, for young children who are 2 years of age and younger, it may be wisest to characterize their behaviours as on the ASD spectrum and then use their scores on tools such as the ADOS-T (Luyster et al., 2009) along with expert clinical opinion to characterize the certainty of the diagnosis. In addition, as suggested by the work of Guthrie et al. (2013), using multiple measures and contexts for assessments along with expert clinicians can improve the stability of the diagnosis. Further, in regards to Camarata's (2014) quandary as to whether to include what he calls “marginal cases” (p. 8) vs the “more severe cases” (p. 9) when testing intervention effects, this issue could be addressed with a measure of autism severity (e.g., standardized severity scores derived from ADOS-T scores; Hus, Gotham & Lord, 2012), as well as some rating of certainty from clinical opinion could be co-varied in analyses. In addition, if one supports the view that children with ASD are on a spectrum, then we would need children all along the spectrum as part of our intervention research if we are to truly advance the science and identify for whom certain interventions work. The final issue that our work raises is the “what do we lose if we wait to screen, diagnose, and treat” question, and that leads us to argue also for early screening.

Argument for early screening

Allaby and Sharma (2011) argue against universal screening on the grounds that it is too difficult to identify and differentially diagnosis children with ASD at 24 months of age. We, on the other hand, argue that it is too costly for the children, families, and society not to screen early. In regard to screening at young ages, there continues to be progress in refining screening tools that can identify reliably at-risk status for a number of children who should go on to receive an evaluation (Guthrie et al., 2013; Kleinman, Robins, Ventola, Pandey, Boorstein, Esser, et al. 2008; Turner-Brown et al., 2013). As noted by Guthrie et al. (2013), a community screening approach identified a number of children who may not have been otherwise screened or referred. In particular, they reported that, although 90% of the children eventually diagnosed with ASD screened positive on the Infant Toddler Checklist (Wetherby & Prizant, 2002b), only 50% of the caregivers reported any concerns about their child's development when they were screened. Therefore, half of the children may not have been identified until a later age by either their parents or primary care providers. Another key result from Guthrie et al. is that only 9% of the children who screened positive on the Infant Toddler Checklist were eventually diagnosed as typically-developing. Therefore, the other 91% of the children were diagnosed with either ASD or other developmental disabilities. Thus, there were limited false positives if the outcomes of the screening are evaluated with respect to a broader range of developmental problems, yielding a large group of children who could receive services earlier than they might have without the screening.

Similar findings have been reported by our research team (Turner-Brown et al., 2013) when we followed children who were screened early for ASD. As noted, using the FYI (Baranek et al., 2003), we screened a community sample of children at 12 months and then followed them to the age of 3 years. Of the children who screened positive at 12 months for risk for ASD and we followed to 3 years of age, 31% had a diagnosis of ASD, whereas 85% had either a confirmed diagnosis of ASD or another DD or parental concern about their development. Therefore, even children identified in early ASD screenings who are “false positives” for ASD are at very high risk for poor developmental outcomes and warrant consideration for early intervention. Camarata's (2014) point about children not receiving services as early as possible would certainly be true for our study children (Turner-Brown et al., 2013) as well as those of Guthrie et al. (2013).

The final issue of concern relative to screening is the change in ASD symptoms that may occur in some children in the first 2–3 years of life. As noted by several researchers, children's skills may improve, lessen, or plateau (Landa et al., 2007; Lord, Luyster, Guthrie, & Pickles, 2012). Although there is growing literature to show that in some children there are signs of ASD in the first 18 months, there are also a number of children who do not exhibit these kinds of symptoms (Landa et al., 2007; Werner, Dawson, Munson, & Osterling, 2005). The Landa et al. (2007) prospective study with infant siblings of children with ASD documented that 46% of the children diagnosed with ASD at 24 months did not demonstrate any obvious symptoms at 14 months. Similarly, Werner et al. (2005) reported that 35% of parents of children later diagnosed with ASD reported that their child was asymptomatic at 10–12 months of age.

These findings indicate that early screening for ASD risk in young children will very likely miss some children who later develop the disorder. Therefore, the guidelines from the Academy of Pediatrics (Johnson & Myers, 2007) recommend screening twice before the age of 24 months, specifically at 18 and 24 months (or 30 months if the child is not seen at 24 months). In keeping with the nature of these recommendations reflecting a need for serial screening, our team could envision tools such as the FYI (Baranek et al., 2003) or ITC (Wetherby & Prizant, 2002b) being used at 12 months, followed by other tools (e.g., MCHAT, ITC) used at later ages, with children referred for a complete ASD-specific diagnostic evaluation if risk is identified at any age.

In closing, as noted by Warren et al. (2011) and Camarata (2014), there is much work to be done in early identification and intervention for children with or at-risk for ASD. Better efforts to identify and characterize young children with ASD must be made so as to develop and adapt interventions that can most effectively address each child and family's unique characteristics. In addition, given the promising results emerging, more work is needed to replicate, expand, and individualize available interventions, as well as screening and diagnostic tools.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.