Prepared by Michael B. First, M.D., DSM Consultant to the American Psychiatric Institute for Research and Education (APIRE), a subsidiary of the American Psychiatric Association
The APA, in collaboration with the WHO, NIH, and the M.I.N.D. Institute at the University of California, Davis, convened a diagnosis-related research planning conference focusing on autism and other pervasive developmental disorders at the M.I. N.D. Institute in Sacramento California, on February 3-5, 2008. The conference was the last in a series of NIH-funded conferences on “The Future of Psychiatric Diagnosis: Refining the Research Agenda” that is administered by APA’s American Psychiatric Institute for Research and Education (APIRE). The conference co-chairs were Susan Swedo, M.D., National Institute of Mental Health (Bethesda, MD), Poul Thorsen, M.D., Ph.D., University of Aarhus (Aarhus, Denmark), and Daniel Pine, M.D., National Institute of Mental Health (Bethesda, MD.) Twenty six invited scientists from around the world participated.
Susan Swedo, M.D. (Bethesda, MD), co-chair of the conference and chair of the DSM-V Autism Work Group, began by providing an overview of the plans for the DSM-V work group. The job of the work group is to identify criteria that will be more accurate and precise and to determine what data are needed to recommend a change in the criteria. Given that the work group does not have the resources to collect new data; recommendations will have to be based on a review of the literature and on secondary analyses of existing data. She emphasized that all changes must be warranted by data and by experience in the field and warned against just tinkering with the diagnostic criteria. Proposed new criteria for autism will need to be reliable, valid, and developmentally sensitive and will need to be tested out in field trials. She noted that the audience for the DSM is diverse: primary care clinicians are the main users, so criteria must be clinically useful and manageable, given that clinicians often have only five minutes to do their evaluations. The agenda of the conference is divided into a series of nine panels, each one focusing on a question that was raised during the first meeting of the DSM-V Autism Work Group.
The first panel addressed the question what are the core symptom domains in autism? Three alternatives have been proposed: social deficits only (leading to communication deficits and repetitive/fixated behavior), social and communication deficits (but not repetitive behaviors/fixated interests), or all three symptom domains (i.e., social deficits, communication deficits, and repetitive behaviors/fixated interests), as currently defined in DSM-IV. In her introduction to the panel, Amy Wetherby, Ph.D., (Tallahassee, Florida) noted that currently the core symptoms of autism are divided into three domains: impairment in social interaction (e.g., impairment in the use of nonverbal behavior; lack of spontaneous sharing; lack of social/emotional reciprocity; failure to develop peer relationships), impairment in communication (e.g., delay in or lack of development of spoken language and gestures; impairment in the ability to initiate or maintain conversation; repetitive and idiosyncratic use of language; lack of pretend play), and repetitive behaviors and fixated interests (e.g., preoccupation with restricted patterns of interest; inflexible adherence to routines; repetitive movements; preoccupation with parts of objects). A prospective study of the general population which looked at 29 features to see which features at age 2 predicted the later development of an Autistic Spectrum Disorder (ASD) found nine features from the three domains, supporting the notion of a triad of domains. Looking at the ability of these features to distinguish between ASD, other developmental disorders, and typical development both early and late during the second year, repetitive movements of objects (e.g., swiping, rubbing, squeezing objects, lining up objects, collecting objects) were better able to distinguish between ASD and developmental disorder both early and late during the second year of life than were repetitive movements of the body (e.g., flapping arms or hands, rubbing, tapping, or pressing body parts, stiffening fingers, hands, or arms) or social communication problems, which can only distinguish between these groups later during the second year, suggesting that there may be separate strands underpinning ASD early on which become intertwined by late in the second year. In her presentation, Catherine Lord, Ph.D., (Ann Arbor, MI) recommended caution when doing factor analyses on ASD data sets, given that subject characteristics and instrument characteristics affect results. All studies have found considerable correlations between factors (with correlations between 0.37- 0.66, at least for restricted, repetitive behaviors) and have required oblique factors, but that the number of factors has differed greatly across studies. Most of the studies have suggested that social items and non-verbal items are hard to separate but, on the whole, non-verbal communication items are the elements that go together to create social deficits. Dr. Lord proposed that the diagnostic criteria for autism include social communication deficits and restricted, repetitive behaviors and that chronological age, current expressive language level, nonverbal IQ, and resulting impairment also be taken into account. John Constantino, M.D., (St. Louis, MO), in his presentation, noted that, 1) autism represents the severe end of a continuum of inherited social deficiencies that occur in nature and it is arbitrary where to draw the line between affected and normal states.; 2) there are no data that support Asperger’s disorder breeding true; familial idiopathic PDDs share genetic origins; for example, in a Danish epidemiological sibling study, both autism and Asperger’s disorder increase the risk of autism in siblings (although the risk is higher for probands with autism); 3) DSM-IV criterion domains for autism are NOT empirically-derivable by factor, cluster, or latent class analyses, suggesting a unitary underlying factor structure; 4) that although ADHD and PDD share some secondary symptoms, they are inherited largely independently, can co-occur, have been shown to respond independently to some of the same treatments and therefore should be allowed to be diagnosed simultaneously (i.e. NOT mutually exclusive for assignment of diagnosis); and 5) an important goal for DSM-V is to develop norms that provide the structure for a system of dimensional characterization of developmental psychopathology (for example, how much deficiency in impulse control should be expected for a child with an IQ of 75?). Finally, Francesca G. Happé, Ph.D., (London, UK), presented data regarding whether the familiar triad of impairments in autism (i.e., social impairments, communication impairments, and restricted repetitive behavior and interests) are all due to a single underlying cause or whether they are influenced by separate causes. In studies of a population-based sample of twins (at ages 7, 8, and 10), correlations between trait measures of the three triad areas ranged from 0.3 to 0.4. In addition, cross-twin cross-trait correlations suggested that at least half the genes acting on these (highly heritable) traits act specifically on one part of the triad. While the data on children with extreme (impaired) scores in each domain do suggest that the three triad impairments cluster above chance, there are large numbers of children with parent-reported impairments in just one triad area. The results from these studies indicate that aspects of the triad can be seen in isolation. In terms of cognitive theories and brain imaging, there is also no single explanation for the three aspects of the triad. Dr. Happé noted, however, that when impairments in all three domains occur together, there may be a qualitative shift, perhaps reflecting a stripping away of the individual’s ability to compensate. She concluded by noting the need to consider the separability of the domains of the triad when considering a dimensional approach to the diagnosis of autism spectrum disorders.
The second panel addressed the question how are fixated interests and stereotypes related to each other, to autism, and to obsessions and compulsions? In his introduction to the panel, Edwin H. Cook, Jr., M.D., (Chicago, IL) noted that restricted and repetitive behaviors are currently de-emphasized relative to social impairment. Why should a child with mild social and communicative impairment whose main impairment is disabling restrictive and repetitive behaviors be considered to have less likelihood of autistic disorder? Is our lack of emphasis on severity of this area justified by the data or is it based on our “preoccupation” with social and communication function? Why do we have use of stereotyped language in the communication domain rather than the restricted and repetitive behavior domain? Dr. Cook raised the question of whether repetitive and restricted behaviors would be better split into two facets (i.e., insistence on sameness and repetitive sensory motor behavior), noting that they might have different behavioral and treatment implications (e.g., elements of insistence on sameness are responsive to SSRIs in some patients; it may make more sense to treat insistence on sameness as a single entity since it does get in the way of other interventions, in comparison to repetitive sensory motor behavior [RSMB] which often is reduced with optimal educational interventions, even though the interventions are not directed specifically at RSMB). James Bodfish, Ph.D., (Chapel Hill, NC), in his presentation, noted that according to factor analyses, there is likely more than one variety of repetitive behavior, with some analyses suggesting that insistence on sameness and compulsions represent a higher order factor and repetitive sensory motor behavior a lower order factor. He also questioned which variety is relatively more specific to autism (repetitive motor behaviors including repetitive use of objects, or sameness, rituals, compulsions, or fixated interests; different studies have conflicting results); which variety has the earliest onset and persists with age (all varieties maintain with age although severity tends to diminish). Finally, regarding the role of repetitive behaviors in the diagnostic criteria for autism, multiple studies have shown that severity of sameness/rituals (but not repetitive motor behavior) is independent of social language deficits, that repetitive behaviors vary continuously within autism (i.e., they do not form taxa or subtypes); that increased repetitive behaviors are associated with increased social and/or language deficits (i.e., severity of the two domains runs together); and that early repetitive behaviors predict later social / language deficits. Susan Swedo, M.D., (Bethesda, MD), in her presentation, noted how little data has been reported regarding the difference between fixated interests in autism and obsessive-compulsive symptoms in obsessive-compulsive disorder (OCD), noting that there may be a clear difference between childhood onset obsessions and adult-onset obsessions. Differences in treatment response, patterns of comorbidity and neuroimaging findings suggest that in adults, obsessions seem to fit best with anxiety disorders and in children are more closely linked to attention deficit hyperactivity disorder and tic disorders. Regarding the difference between stereotypies in ASD and compulsions in OCD, the two symptoms are more similar than different in presentation. For example, when the repetitive behavior is interrupted, children respond negatively, regardless of whether the behavior is a compulsion or a stereotypy. Dr. Swedo noted three possible hypotheses regarding the relationship: that symptoms in autism and OCD are the same phenomena and result from similar etiologies; that they are the same behavioral abnormality, but represent different etiologies; or that symptoms in autism and OCD only appear to be similar but differ in nature and etiopathogenesis. Comparing ASD and obsessive-compulsive personality disorder (OCPD), similarities between them include rigidity, need for sameness, and individuals being “cold” and socially isolated; differences include age at onset (earlier in ASD), degree of impairment (greater in ASD), and presence of comorbid symptoms (more in ASD).
The third panel addressed the question how does the presentation of autism change across the lifespan? In her introduction, Catherine Lord, Ph.D., (Ann Arbor, MI) commented that one of the real improvements in DSM-IV was to include criterion items that would apply to both pre-school and school-age children; however, the items still do not adequately apply to toddlers, adolescents, and adults. When assessing toddlers, one challenge is to get reliable information from a parent who may not be familiar with normal childhood development; for adolescents and adults, compensatory strategies and learning might alter the presentation. Interpreting longitudinal data can be challenging because the proportion of children who have difficulties with the ASD domains change over time, as do factor loadings. For example, repetitive behaviors start early in the course and persist over time, but the number of behaviors goes down and they tend not to be as interfering. On the other hand, insistence on sameness tends to start later and increases slightly. Amy Wetherby, Ph.D., (Tallahassee, FL) noted that prospective studies of general population samples have documented core deficits in social communication, fixated interests (e.g., sticky attention to objects) and repetitive behaviors in children with ASD in the second year. Deceleration of development may be characteristic of the unfolding of diagnostic features of ASD over the second year, making screening and early detection more challenging. Social communication and repetitive behaviors are independent constructs early in the second year and become intertwined by late in the second year, with repetitive behaviors contributing to social affect deficits at 3 years of age. These early core deficits appear to have a cascading effect that impacts the children’s learning environment by limiting what they can get out of it. Keith Widaman, Ph.D., (Davis, CA), in his presentation, looked at lifespan changes from the perspective of mental retardation. Cross-sectional studies of intelligence have shown that there are substantial improvements in many areas until the developmental period ends around age 20; after that point, stability is more likely. For mental retardation, which is considered a developmental disorder, the diagnosis is supposed to be made in the first 18 years of life. Does it therefore make sense to make a diagnosis of ASD and intellectual disabilities in someone presenting at age 50? An empirical basis for making long term predictions about outcome is lacking for individuals with autism. Harry H. Wright, M.D., (Columbia. SC), noted that despite a consensus that autism nearly always develop before 3 years of age, it is only recently that significant information on the early presentation of ASD in very young (birth to 3 years of age) children has been reported. He called for similar efforts to develop developmentally sensitive criteria and diagnostic algorithms across the lifespan, especially in early and middle adulthood. Dr. Wright recounted how his clinic has been getting referrals for evaluation of adolescent and adult family members (e.g., older siblings, uncles) of young children diagnosed with autism but there are no guidelines for making a diagnosis of ASD in such older individuals; adult examples are needed for the various autism criteria. This may be a particular issue for African-American families who are typically referred at a later age than their Caucasian counterparts.
The fourth panel addressed the question how does developmental regression (and particularly Childhood Disintegrative Disorder) fit into the autism spectrum? In her introduction, Sarah J. Spence, M.D., Ph.D., (Bethesda, MD) raised a number of questions about the construct of regression in ASD, including what is it and how commonly does it occur? Does regression represent true loss of skills versus just a plateau, and which skills should define a regression? Does it require normal development to precede it? When does it happen? What causes it? How important are medical factors? s it specific to ASDs? Regression is currently only included in the diagnostic criteria in Childhood Disintegrative Disorder (CDD), where it is required. Clinically it is commonly seen in Rett Syndrome but not included in the criteria, and is not even mentioned in Autistic Disorder, Asperger’s Disorder, or PDD-NOS, although it is known to occur fairly frequently. Childhood Disintegrative Disorder was first mentioned by Heller in 1908 and was first included in the DSM starting with DSM-IV. The available literature on CDD consists of case reports and a few case series. The paucity of reports is confirmation that CDD is a rare condition, with prevalence thought to be 1.1 to 6.4 per 100,000. It was once assumed that various presentations of CDD have a common medical etiopathogenesis, but a cause is not identified in most cases. Dr. Spence summarized some of the challenges in defining the role of regression in CDD and other ASDs as follows: 1) determining the validity of the age requirement (24 months-10 years); 2) operationalizing the definition of regression; 3) close examination of pathogenesis and clarification of the role of co-morbid medical diagnoses; and 4) determining its impact on prognosis and response to intervention. In his presentation, Hiroshi Kurita, M.D., (Tokyo, Japan) argued that although CDD is quite rare and research on CDD very scarce, there is no convincing evidence to abolish or merge it with autistic disorder, and that it should be preserved as an independent disorder. He made a number of specific suggestions to improve the CDD criteria set, including: 1) changing the requirement in criterion A for the regression history from “apparently normal development” to “no clinically significant delay or abnormality in development”; 2) deleting motor skills regression from the areas of regression in criterion B because of scarcity; and 3) in criterion D, having the diagnosis of CDD pre-empt a diagnosis of autistic disorder so that CDD with onset before age 3 is not diagnosed as autistic disorder (one-third to one-half of CDD patients show their regression between ages 2 and 3). Compared with autistic disorder, CDD has no significant difference in clinical symptomatology apart from its higher incidences of epilepsy and EEG abnormalities. Dr. Kurita concluded by noting that given the rarity of CDD, a standard format of evaluating a possible CDD case or a case with a history of regression is needed to facilitate early detection and research of CDD and other PDD with regression. Pauline A. Filipek, M.D., (Irvine, CA) showed prospective video clips of an infant aged 13 months through 26 months that illustrated apparently normal development at age 13 months, with gestures and 6 to 7 spontaneous words. One month later, he presented to the clinic with a loss of response to name, followed by loss of his 6 to 7 words, onset of highly repetitive play and loss of eye contact, which progressed to frank autistic disorder by age 24 months. This sequence raised questions about what separates autistic regression from CDD -- how often does regression occur in autistic disorder? Is CDD simply autistic regression that occurs later, after age 24 months? During the ensuing discussion, it was pointed out that regression may be a more common feature of autism than was previously thought with some prospective studies indicating that a loss of skills is the rule rather than the exception. In regressive autism, the skills are lost in the second year of life, while in “early onset” autism, skills are lost in the first year of life. There was general agreement that symptom onset is on a continuum between regression and non-regression and that defining the borders between the two can be difficult. Diagnostic certainty is particularly problematic because most parents are not going to pick up a regression of acquired skills unless the child has acquired language which then is lost.
The fifth panel addressed the question Asperger’s Disorder – is it Autism? In her introduction, Francesca G. Happé, Ph.D., (London, UK) raised some of the key questions that have arisen regarding the diagnosis of Asperger’s Disorder, which was introduced into DSM-IV in 1994. These questions include: is there an ‘Asperger’ subgroup of autism with distinct cause, course, cognitive profile, and intervention needs, and if so, what is its relation to other ASDs? Asperger’s disorder is essentially defined as meeting criteria for autism without the language impairment. Lorna Wing introduced the term in 1981 to aid recognition of the part of the autism spectrum with good IQ and language. It has increased awareness and recognition and helped to clarify the core deficits of ASD, but also increases the possibility that there may be over-diagnosis of ASD. Asperger’s disorder has also had an impact on family studies of autism with regard to what we recognize as “caseness.” Dr. Happe noted that the current criteria do not work: they do not allow for developmental change, the early language criteria do not demarcate groups with different prognoses, it is hard to apply the diagnosis for adult cases, and there is no clear conceptual basis for the diagnosis. Dr. Happe concluded that although there is a recognizable Asperger’s type and that some cases of classic autism grow into this picture, she wonders whether there may be a better classification schema. Sally Ozonoff, Ph.D., (Sacramento, CA), in her presentation, compared high functioning autism (HFA) with Asperger’s, and noted that there were few differences in their definitional DSM-IV criteria; both require two social symptoms and one repetitive/stereotyped symptom, both are in the average range intellectually and have current fluent language. The main criterion distinguishing the two disorders is the requirement in Asperger’s that onset of language occurs at the expected time, e.g., single words by age 2. Dr. Ozonoff noted that it is difficult to evaluate the literature since definitions vary across studies and that many children who are thought clinically to have Asperger’s actually meet criteria for autism (which supercedes a diagnosis of Asperger's). There is some evidence to suggest that Asperger's and HFA do not represent distinct disorders: they co-occur in the same families and do not “breed true” (i.e., family members of patients with Asperger's have HFA and family members of patients with HFA have Asperger's); children with autism who develop language have similar outcome to Asperger's; HFA and Asperger's are indistinguishable by school-age; and although studies find better language skills and/or verbal IQ in Asperger's, multiple studies have found no group differences in other neuropsychological domains.
The sixth panel addressed the question, Is Autism a Life-Long Diagnosis? Bryan H. King, M.D., Ph.D., (Seattle, WA) introduced the panel by asking if one outgrows a diagnosis or it remits, (particularly in the context of genetic underpinnings), is it still there? Residual language deficits have been demonstrated in some children with a well-documented history of autism who have gone on to have excellent outcomes. As children age, they tend to lose some of the severity of their illness so that the diagnosis can change from autistic disorder to PDD-NOS. A problem is caused by linking a diagnosis to the severity of manifestations of the symptoms. This causes difficulties for service provision and for clinical trials; for example, the FDA will not allow for an indication that was created for autism to transfer over to PDD-NOS, even though it may be the same children who have the two different diagnoses (ASD when young and PDD-NOS as they develop further). The current classification mixes categorical definition with the severity of disorder. Eric Fombonne, M.D., (Montreal, Canada) began his presentation by noting that a diagnosis of autistic disorder is usually highly stable; less than 7% fall out of the autistic spectrum on follow-up. On the other hand, children with ASD who do move out of the spectrum are most likely to start out as PDD-NOS, reflected the very mixed bag of patients that fall under this rubric. It is possible that PDD-NOS represents a lower “dose” of autism (i.e., still ASD); it could be a phenocopy (i.e., non-ASD that present with PDD symptoms, including genetic syndromes such as Rett disorder); it could be the result of environmental exposure, (such as fetal alcohol syndrome), or it could result from plain measurement error. When PDD-NOS is no longer present, is the phenotype qualitatively the same but the associated impairment has become minimal (i.e., is it only the impairment that has disappeared) or is there evidence for true remission of autism? Long term follow-up and adult studies are essential to address the question of remission. With respect to considerations for DSM-V, Dr. Fombonne concluded by noting that the heterogeneity of the PDD-NOS diagnosis is problematic as there are at least two ways to meet criteria: the pattern of symptoms is the same as in ASD (all domains affected) but of lesser severity, or the pattern may be different (one or two domains only). It may be useful in the future to measure/document the impairment associated with each domain (rather than globally). Finally, age of onset is a further problem as it often depends on parental recognition or retrospective recall. Craig Newschaffer, Ph.D., (Philadelphia, PA), in his presentation, noted that we are still at a point where the connections between pathology and the phenotype of autism are not understood. How the manifestations of underlying pathology change over time and what, if any correlations exist between pathology and clinical phenotype are still unknown. Regarding the evolution of the behavioral phenotype over the life course, there are very few cross-sectional studies in adolescents and adults, and even fewer longitudinal studies. The lack of availability of age-appropriate measures may contribute to this lack. The few small sample studies that have been published focus on the loss of diagnosis (typically in children) and presentation of previously undiagnosed adults. From the clinical perspective, should the diagnosis of autism be lost when symptom-based criteria are no longer met? This has potential impact on treatment plans and access to services for the child or young adult and their family. The diagnosis of ASD in symptomatic adults is also difficult, due to the lack of tools appropriate for use in adults and the difficulty in obtaining the developmental data required to make a diagnosis of ASD (which must present before age 3 years). The differential diagnosis is particularly challenging in adults because of the number of other disorders that affect the social domain. These limitations suggest that other features (e.g., nonverbal communication, prosody, subtle use of language, posture, etc.) might prove to be better diagnostic criteria for adults? Without such clarity, it is questionable whether the prevalence of ASD in adult populations can be validity and accurately estimated through epidemiological studies.
The seventh panel addressed the question of how does comorbidity affects symptoms of Autism? In her introduction to the panel, Sally Rogers, Ph.D., (Sacramento, CA) noted that other developmental disorders occur commonly with autism (e.g., up to 86% also have non-verbal learning disorders) as well as other psychiatric disorders (e.g., 20-30% with affective disorders, 33-75% with ADHD), and various medical conditions (10-37%, with the rate depending on the severity of the degree of intellectual disability. Dr. Rogers argued that one way to address this comorbidity is to adopt a dimensional approach which could address sociability and social capacities, communication, intellect, activity level, motor skills and movement, mood and temperament, academic abilities, adaptive behavior, maladaptive behavior/psychiatric symptoms, and health. Autism could be viewed in the same way as we do intellectual disabilities (i.e., IQ) -- as a final common behavioral pathway of many different biological etiologies: “the autisms,” as opposed to our current view of autism as primarily a familial genetically-determined disorder with comorbidities occurring more frequently in the more severe cases. Tony Charman, Ph.D., (London, UK) presented on the topic of autism and intellectual disability and noted that two recent population-based surveys assessing ASD at different levels of intellectual disability found rates of 9-12% for individuals with mild ID (IQ 50-70) and rates of 26-30% for moderate ID (IQ 35-50). Rates of ASD in those with more severe ID (IQ<35) are not known. Although in the current ASD criteria only peer relationships and conversational abilities are explicitly related to developmental and language levels, other symptoms are dependent on developmental maturation as well. Should there be different symptom thresholds depending on IQ? For example, looking at repetitive behaviors in ASD, there are difference prevalence rates in high versus low IQ patients. Comorbidity patterns are thought to differ by intellectual abilities, but this may be the result of diagnostic overshadowing, which often leads to underdiagnosis of psychiatric comorbidity, i.e., if there is already a diagnosis of ASD, why give an additional psychiatric diagnosis? Thomas Anders, M.D., (Sacramento, CA) presented on sleep in children with neurodevelopmental disorders, noting that anecdotally, sleep disorders are a major problem in ASD. In a study using an actigraph to measure sleep in children with autism or developmental disabilities, and among typically developing children, children with autism had the least amount of sleep (less 24-hour sleep time and less nap time); those with developmental disabilities were intermediate between the children with autism and the typically developing controls; however, children with developmental disabilities had the most fragmented sleep. Dr. Anders concluded that children with autism likely have a clock problem compared with children with other developmental disabilities. Walter Kaufmann, M.D., (Baltimore, MD) presented on ASD and cognition. He proposed three alternative hypotheses for the relationship between ASD and intellectual disability (ID): 1) ASD is non-specific and is simply another manifestation of severe neurobiological impairment; 2) it is selective but also a reflection of decreased communication skills; and 3) ASD in ID is a selective and specific impairment in social interaction. In Down syndrome, children who meet criteria for ASD tend to have IQs in the severe and profound range whereas in Fragile X, ASD diagnoses are concentrated in the moderate IQ range. However, low IQ per se has a minimal influence upon ASD status in ID. Dr. Kaufmann concluded that although low IQ is a predisposing factor to ASD, not every child with severe to profound ID meets criteria for ASD. Qualitative impairments of social interaction (in joint attention, in shared enjoyment, perhaps others), appear to be relatively independent of mental age and are found in individuals with severe-profound ID. Among individuals with both ID and ASD, the presence of repetitive and stereotypic behavior is more severe and complex than would be expected on the basis of the lower mental age alone. However, IQ less than 25 to 30 is a limiting factor for behavioral interpretation given that such individuals do not have the behavioral repertoire characteristic of ASD.
The eighth panel addressed the question what role should neurobiology play in the DSM-V diagnostic criteria for autism? In his introduction to the panel, Joe Piven, M.D., (Chapel Hill, NC) noted that even though autism is defined exclusively in DSM-IV in terms of behavioral criteria, evidence for the biological basis of autism is growing with studies demonstrating variable support and variable explanatory power for biological variables. An increasing number of investigations are demonstrating associations between ASD and genetic aberrations (e.g., chromosome 15 duplications, chromosome 16 deletions, familial types), as well as biological markers (e.g., neurotransmitter levels), neuroimaging results (e.g., brain volume), head circumference (e.g., macrocephaly), electrophysiological testing (e.g., ERP, EEG) and neuropsychological assessments (e.g., face processing). However, these findings are not sufficiently specific or cohesive enough to allow for the identification of clinically meaningful subgroups or to be used as risk markers. Regarding how to incorporate subgroups of ASDs with an identifiable etiology (e.g., Fragile X, tuberous sclerosis) into the diagnostic framework, options include listing the medical condition on Axis III, including a subtyping scheme on Axis I (e.g., autism, Fragile X type), or excluding it from the diagnosis altogether, as is now done with Rett’s syndrome. Finally, given the rapid pace of research on the biological basis of the ‘autisms’, it will be important to construct a flexible diagnostic system that is able to incorporate new information before the appearance of DSM-VI. Catherine Barthelemy, M.D,. Ph.D.m (Tours, France) then presented data demonstrating how electrophysiology combined with genetics can help to define new markers for better phenotypes and perhaps endophenotypes of autism. She described a series of studies focusing on event related potentials in the detection of abnormalities in the processing of social information, especially faces and voice, hypothesizing that the need for sameness in autism could be rooted in abnormal sensitivity to change. In these studies, latency of response to automatic change is significantly shorter in children with autism compared to normal children and may be a good index of the reaction to change. Moreover, the shorter the latency, the more severe the reaction to change measured by behavioral scales. Dr. Barthelemy concluded by noting that the advantages of electrophysiological markers include that they are objective measurements which are quantifiable (in terms of amplitude, latency, etc), non-invasive, low cost, and easy to apply to both children and adults. Further studies are needed on large populations to determine sensitivity, specificity (especially with regard to other disorders such as OCD) and stability over time. Edwin H. Cook, Jr., M.D., (Chicago, IL) then presented on the genetics of autism. The most common model for autism, called the common variant-common disease model, postulates that there are various risk variants that contribute to autism. For example, if there are five risk variants, one needs to have hits on all five variants for the disorder; hits on four variants would lead to ASDNOS. It is thought that the vast majority of autism cases are multifactorial. Another model is the “big hit” model, in which autism is caused by a genetic abnormality with higher penetrance; examples include chromosome 15q11-13 duplication or triplication (0.5-3%), Fragile X (0.5-3%), 16p11.2 deletion (0.5-1%), SHANK3 mutation (0.5-1%), etc. The most likely model is that these “less complex” cases represent situations where the chromosomal or single gene variant is equivalent to a number of smaller effect variants. It is important to understand, however, that the so-called etiological syndromes have substantial variability in terms of behavioral manifestations: a diagnosis of Fragile X, maternal 15q11-q13 duplication, VCFS, SHANK3 mutation, or 16p11.2 deletion is NOT equivalent to diagnosis of an ASD and even when there is a strong association, knowing the genetic defect does not provide a predictive description of that child’s behavioral manifestations. Some syndromes are associated with patterns of symptoms: for example, Fragile X and gaze aversion, maternal 15q11-q13 duplication and mood lability. Other syndromes (e.g., 16p11.2 deletions) are without obvious distinguishing features including variable cognitive function. Finally, many patients have unique submicroscopic deletions and duplications and many genetic abnormalities associated with ASD may be de novo and non-recurrent. Walter Kaufmann, M.D., (Baltimore, MD) presented on Rett Syndrome and ASD, noting that up to 75% of researchers feel that Rett should not be in a separate category in the ASD section. In Rett, autistic features are sometimes indistinguishable from autism but they are only present during a certain phase of the illness (i.e., between 1 and 3 years of age) and will get better without treatment. Those girls who do not have the marked motor features most characteristic of Rett (i.e., increased tone, dypraxic gate, hand-wringing) are more likely to be diagnosed as idiopathic autism because genetic testing does not seem warranted. Finally, he presented on two biomarkers that suggest the potential future use of imaging and molecular profiles in the diagnosis of autism spectrum disorders: abnormal size of the posterior-superior vermis (absolute hypoplasia in idiopathic ASD, relative hyperplasia in Fragile X + ASD) and lymphoid cell abnormalities in cytoplasmic FMR1 interacting protein 1 (CYFIP1) pathway expression in both Fragile X and chromosome 15 duplication associated with ASD.
The final panel addressed International, Cultural, and Gender considerations in the diagnosis of Autistic Spectrum Disorders. In his introduction, Poul Thorsen, M.D., Ph.D., (Aarhus, Denmark) focused on gender differences, noting that the male-female ratio for a diagnosis of ASD is 5:1 whereas the ratio for childhood autism is 3-4:1. The rate of recurrence in siblings of affected individuals is 2-8%, which is much greater than the risk in the general population. A history of low birth weight and/or being small-for-gestational age was more common among high-functioning girls with autism than among their unaffected female siblings whereas there were no differences in frequency of low weight between high functioning males with autism and their male siblings. Autism sex ratio (male: female) is lower in individuals with ID than in individuals with normal cognitive functioning and lower in individuals with significant dysmorphology or microcephaly than those without. Data from a Danish psychiatry registry of children born from 1990 to 1999 found that males on average met 0.63 more items than females. Males on average met 0.29 more items concerning impaired communication and 0.28 more items concerning restricted behaviors than females. For autism cases without mental retardation, males overall met on average 0.69 more items than females and on average 0.28 more items concerning restricted behaviors when adjusting for age. He concluded that since males consistently met more autism items than females, the current diagnostic criteria may favor the diagnosis of autism in males. In her presentation, Diana E. Schendel, Ph.D., (Atlanta, GA) began by discussing the difficulties in trying to connect prevalence studies and diagnostic criteria given that the connection between the diagnostic criteria and case features is not a direct line; when you start with the diagnostic criteria, even if they are being implemented directly, the implementation process will influence the outcome or features of the case group you find. For example, most studies need to rely on an agency to identify cases and often agencies differ in the characteristics of their case group, possibly due to differences in populations being served. As another example, a Danish study which compared rates of autism to other childhood psychiatric disorders has shown a similar increase in incidence for these other disorders over time, suggesting that the similar trends reflect the shared processes by which cases with these different disorders have been identified. As a third example, another administrative process that may influence case group features is the significant shift in age of diagnosis over time. In a Danish study, shifts in age of diagnosis, especially substantial acceleration at younger ages, artificially inflated the differences in observed risk for autism among young children in more recent cohorts compared to older cohorts. Autism trends based on data with incomplete follow up (i.e., they lack follow-up to the age at which all individuals have been identified) can be confounded by changes in age of diagnosis over time. To the extent that these and other administrative factors influence observed prevalence and/or distribution of behavioral phenotypes in a case group (e.g, autism in toddlers vs. older ages) observed at a specific point in time, then the case group profile of “What is autism?” may vary accordingly. In his presentation, Kang-E Michael Hong, M.D., (Seoul, South Korea) raised the question of whether it is possible that some cases of ASD are caused by environmental factors by examining three sets of data. The first set of data came from studies of Reactive Attachment Disorder (RAD) in Korea. There are big socio-cultural changes in Korea with rapid modernization that has brought changes in child-rearing. One result appears to be increased reports of RAD among children who have not suffered gross physical abuse or neglect; instead these are children who are thought to have been subjected to pure emotional and social deprivation. These so-called RAD children appear clinically very similar to autistic children and have been labeled as having a “Korean syndrome of attachment” mimicking ASD. This raises the question of whether RAD is totally separate from PDD, i.e., is it possible that some cases constitute a type of ASD? Other sets of data come from the recent studies of institutionalized children in Romania in which there is a close association between duration of deprivation and severity of attachment disorder, with many of the children described as “quasi-autistic” (at least in early childhood). A third set of data potentially supporting the premise that ASD may be related to care-giving comes from the huge increase in prevalence rates of autistic disorders. Although the main reasons for rising prevalence are likely to be improvement in ascertainment and broadened definition of ASD, it is possible that changes in child rearing practices and problems of parental emotional unavailability during early infancy may play an etiological role in modernizing societies. Dr. Hong concluded by proposing a research agenda on the role of environmental risk factors in the development of ASD, including: 1) definitive research on environmental/experiential factors, particularly quality of caretaking, as a modifier or etiological contributor in ASD; 2) continuing the need to separate classical autism and the rest of ASD in carrying out further longitudinal studies with regard to etiology, phenotype, progress patterns and intervention; 3) a need for RAD and ASD research camps to converge to decide the relationship between these diagnostic categories; and 4) application of the concept of sensitive period of social development and attachment in understanding, studying and intervening ASD. In his presentation, Craig Newschaffer, Ph.D., (Philadelphia, PA) considered the range of factors the must be taken into account when trying to explain differences in U.S. diagnostic criteria and prevalence rates compared to those found in Europe and Asia. These factors include differences in study methodology, diagnostic criteria, community diagnostic tendencies, cultural context, calendar time, and true variations in the underlying risk. Language and cultural factors are especially important considerations, making adaptation of tools to other languages and cultures particularly challenging. For example, in a Chinese pilot study, when translating diagnostic interviews into Mandarin, it was discovered that there is no distinction between singleton and plural or between past and present tense, creating great difficulties with historical assessments. Furthermore, the cultural norm in China is for gestures to be discouraged and for persistence to be highly valued. Dr. Newschaffer concluded that there will be an explosion of epidemiologic studies worldwide and that work is beginning on how these studies can be conducted in more culturally robust ways.
Upon conclusion of the panels, participants convened into three breakout groups to formulate recommendations for research and suggestions for possible changes to be considered for DSM-V and ICD-11.
The first breakout group recommended the following revisions in the diagnostic criteria: 1) examine the age of onset requirements for Childhood Disintegrative Disorder and Autistic Disorder; 2) add more descriptive text to describe how to apply the criteria to different developmental levels and ages (i.e., adult and young child examples); 3) revise criteria for imaginative play (it is too restrictive in terms of the age that it is appropriate for); 4) revise the criterion for peer relationships given that it is too broad (most psychiatric disorders affect peer relationships); and 5) revise communication criteria (item 2a, delay in or total lack of spoken language). The group also suggested the following be considered: 1) explore the possibility of taking IQ into account in the diagnostic criteria either by making distinctions between children above/below IQ 50 and/or by adding text and examples; 2) clarifying the role and importance of current versus historical symptoms (e.g., if the individual had a symptom earlier such as echolalia but not currently, does it still count?); 3) operationalizing the criteria for PDD-NOS in order to standardize the meaning of the diagnosis (currently, PDD-NOS is a “wastebasket” category which includes individuals with widely disparate symptom patterns): options include counting items contributing to diagnosis and/or identifying particular subgroups; 4) adding a requirement for impairment (which also would need to be operationalized); 5) adding a broader autism phenotype with qualitative, personality language, or behavioral features.
The second breakout group made the following recommendations. With respect to evidence supportive of the DSM-IV approach, DSM-V should continue to include aspects of the three domains (i.e., impairment in social interaction, impairment in communication; and restricted or repetitive interests or activities) although the group did not agree it should necessarily be a triad. DSM-V should also continue to require early onset before age 3; and to keep the behavioral syndromal approach. Changes in DSM-V recommended by the group include: 1) eliminating the exclusion for Rett’s disorder and Childhood Disintegrative Disorders and instead adding a specifier to indicate that it is “associated with diagnosed general medical condition; 2) eliminating Rett’s Disorder from ASD; 3) eliminating CDD as distinct category and instead modifying the age at onset in Autistic Disorder to be: “onset by age 3 (or by age 6 years in cases where there is a normal period of development followed by clinically significant regression in acquired skills)” and adding a specifier to indicate “with late onset and regression” so that such cases can be identified; 4) regression needs to be better characterized (e.g., description of patterns in the text); 5) given that children with profound ID could meet criteria for Autistic Disorder as a result of nonspecific cognitive disabilities; add a clause to criteria advising that developmental level should be taken into account when making the diagnosis; 6) make diagnostic criteria more developmentally appropriate by adding examples of how each criterion is manifest across the life span, especially for infants, toddlers, and adults; 7) eliminate Asperger’s Disorder from this section; 8) change the name of the diagnostic grouping from PDD to ASD; 9) put symptoms inside a dimensionality framework that integrates with other disorders; 10) consider adding cognitive style as an autism symptom; 11) add optional specifiers to the NOS category (which would be named ASDNOS in DSM-V) (e.g., subthreshold symptom count, atypical age at onset, subthreshold triad distribution); 12) separate symptom assessment from disability by adding a criterion that establishes caseness based on exceeding a cut-point on functioning scale (to be determined); 13) emphasize use of “partial” and “full remission” to indicate symptom improvement; and 14) eliminate exclusion of ADHD from ASD (i.e., allow comorbid diagnosis of ADHD and ASD). Finally, the group recommended that data sets be combined for meta-analysis in order to: 1) examine longitudinal changes over time; 2) determine developmentally appropriate examples; 3) look at presentations in adolescents and adults; 4) examine effects of changing criteria on prevalence; 5) determine effects of combining social impairment and communication domains; 6) examine the applicability of symptoms with respect to gender; 7) gather sensitivity and specificity data for individual items for possible weighting of items or re-ordering; 8) examine patterns of comorbidity in population-based samples; 9) evaluate symptom presentation to identify clinically meaningful subtypes of ASD; and 10) design methodologies to explore the relationship between excluded disorders (e.g., ADHD) and ASD in order to determine whether treatment response is different in comorbid cases and whether excluded symptoms are best understood as associated features of ASD (in which case the hierarchy should be retained) versus an independent disorder (in which case the hierarchy should be eliminated). Finally, the group recommended adding dimensions that should be rated after making diagnosis of AD or ASDNOS to further identify functioning and impairment.
The third breakout group made the following recommendations: 1) delete Asperger’s disorder; 2) delete CDD; and 3) create an ASD with two types: Type I would be for prototypical cases characterized by problems in social interaction, social communication, and repetitive behaviors or preoccupations, and Type II is for atypical cases. Data needed to inform such a decision include: 1) the number of criteria to be met for Type I and Type II (at least one in each category); 2) core symptomatology over various ages and developmental stages; 3) clarification of the requirements for diagnosis in females and diverse cultural groups; 4) a definition of impairment at different ages and developmental stages; and 5) consideration of effects of IQ and of comorbid diagnoses. Other specific suggestions include: 1) determining whether obsessive-compulsive symptoms occurring in ASD are part of the ASD or warrant a separate diagnosis of OCD; 2) removing the ADHD exclusion; 3) adding better examples for criterion items across the lifespan; 4) adopting a better definition for regression; 5) determining whether ASD remits and what a residual state might look like; and 6) consider genetics as a modifier versus continuing to code it on Axis III.