Additional related information may be found at:
Neuropsychopharmacology: The Fifth Generation of Progress

Back to Psychopharmacology - The Fourth Generation of Progress

Autism and Pervasive Developmental Disorders

Linda J. Lotspeich


Pervasive developmental disorders (PDD), which include autism, are neurobiological conditions characterized by severe disturbances in the three areas of: 1) social relatedness, 2) communication, and 3) routines and interests. PDD has intrigued professionals since the first description of autism by Leon Kanner in 1943 (40).  According to the DSM IV (5), the PDD category includes five disorders, and by far the most prevalent of these are autistic disorder, Asperger’s disorder and pervasive developmental disorder - not otherwise specified (PDD-NOS).  PDD presents with either many or few symptoms, and symptom intensity ranges from mild to severe. Thus, individuals with PDD exhibit a spectrum of symptoms.  Generally speaking, more severely affected individuals are given the diagnosis of autistic disorder, and the less severely affected are given the diagnosis of PDD-NOS or Asperger’s Disorder.  For example, in a severe PDD case, a typical autistic child might be mute, mentally retarded, preoccupied with spinning objects, and actively avoid all social interactions. In a milder PDD case, a child with Asperger’s disorder might have normal intelligence, speak in full sentences, but be unable to carry on a socially appropriate conversation, and be preoccupied with memorizing schedules or sports statistics. 

The prevalence of autistic disorder in the general population is approximately  5 per 10,000, with a 3:1 predominance of males to females (71).  Although the pathophysiological picture is incomplete, it appears that PDD is a comprehensive category of behavioral and cognitive impairments arising from diverse neurological abnormalities.

This chapter will  summarize what is known today about the neurobiology of PDD.  Four major topics will be reviewed: diagnostic issues, associated neurological disorders, neurobiological research, and psychopharmacological treatment. Diagnostic issues are considered first: which disorders are included in the spectrum of PDD, and what are the specific diagnostic criteria for these disorders? Second, fragile X syndrome and Rett Disorder, the two PDD-associated neurogenetic disorders, are reviewed.  This is followed by a review of recent developments in neurobiological research that includes genetic, neuropathology, neuroimaging and neurochemical studies of PDD. Finally, the effectiveness of medications used in the treatment of specific PDD symptoms will be presented.



To better understand current diagnostic approaches to PDD, it is helpful to review the history of the diagnosis of autistic disorder.  For the past 50 years autism has been diagnosed from its clinical presentation, and even today there are no  biological markers to underpin a clinical diagnosis. Autism cannot be diagnosed at birth but is usually noted or suspected between the first and third years of life.  Although diagnosis is based on clinical behaviors, for many years there were no agreed upon formal criteria.  During the 1970's two diagnostic paradigms were developed for diagnosing autism (63, 66).  However, there still remained a large group of individuals with definite autistic characteristics who did not meet the full criteria for autism. Thus, when autism was included in the International Classification of Diseases, 9th Revision (79) and the Diagnostic and Statistical Manual of Mental Disorders III, (3)  an effort was made to include individuals who met an extended definition of autism, as well as those who met the full diagnostic criteria.

The ICD 9 established a category of disorders - psychoses with origin specific to childhood - which included the four disorders of 1) infantile autism, 2) disintegrative psychosis, 3) other atypical childhood psychosis and, 4) unspecified. In the DSM III, the term pervasive developmental disorders  (PDD) was introduced to reflect the developmental nature and pervasive deficits of these conditions. The DSM III subdivided the PDD category into five disorders: 1) infantile autism; 2) residual infantile autism, 3) child-onset PDD, 4) residual child-onset PDD and, 5) atypical autism.  These groupings subdivided children according to both the age of symptom presentation and symptom severity.

Both the ICD 9 and the DSM III defined infantile autism by narrow criteria, reflecting the more severe and classical Kanner-type of autism. With the revision of the DSM III, (3) the criteria for autistic disorder became less restrictive, and in consequence many more children and adults were given the diagnosis of autism.  In the DSM III-R (4), PDD was simplified to only two disorders: 1) autistic disorder and 2) pervasive developmental disorder - not otherwise specified (PDD-NOS). Autistic disorder in turn was defined by a set of 16 criteria, grouped into the three areas of impaired socialization, impaired communication, and restricted range of interests and behaviors. The broader term PDD-NOS implied qualitative impairment in reciprocal social interactions and communication skills with or without restricted range of interests and behaviors.  This diagnosis was applied to persons with only a few of the autism criteria, who had previously been referred to as "autistic like" or "having autistic features."

The recognition of milder forms of PDD (such as Asperger’s disorder) and the realization that PDD symptoms are common to other, non-autistic childhood disorders (such as Rett disorder) is reflected in the more recently published ICD 10 (80) and DSM-IV (5).  Both the ICD 10 and DSM-IV retained the category of pervasive developmental disorders; the ICD 10 includes seven diagnostic disorders and the DSM-IV includes five (Tables 1 and 2).  The specific criteria for autistic disorder are almost identical between the ICD-10 and the DSM-IV.  In these diagnostic systems there are approximately 12 criteria divided into the three symptom areas of impaired socialization, impaired communication, and restricted range of behaviors, activities and interests. 

Both the ICD 10 and the DSM-IV included a new diagnosis of Asperger’s disorder. Persons with Asperger's disorder (44) typically have a normal IQ, but are socially awkward,  pedantic, and preoccupied with narrow interests, such as memorizing maps or schedules. They also have no delay or minimal delay in language development. Asperger’s disorder was first described by Hans Asperger in 1944 (6), only one year after Kanner's initial description of autism (40). Thirty-five years later, Wing (77) revived the concept of Asperger's syndrome and recommended that it be considered along with autism, since the two disorders shared the symptoms of impaired development of social interaction, communication and imagination.

It appears that the PDD diagnoses, though better differentiated than in the past, still lack unambiguous definitions;  diagnosis must still be made according to behavioral criteria rather than by specific biological markers. Explicit diagnostic criteria for autistic disorder now exist, but  diagnostic criteria for the other forms of PDD, such as PDD-NOS and Asperger's disorder, remain vague (75). Thus, by necessity, most neurobiological research studies of PDD are restricted to subjects who meet the full criteria of autistic disorder.  To date, there are few neurobiological studies of subjects diagnosed as PDD-NOS or Asperger's disorder.



Autism is associated with a variety of neurological disorders , and these have been reviewed in the previous edition of this ACNP publication (82) and by Lotspeich and Ciaranello (47). Autism has been associated with  genetic disorders such as fragile X syndrome, phenylketonuria, tuberous sclerosis, neurofibromatosis and Rett disorder. Chromosomal abnormalities have been associated with autism including partial trisomy 15, a long Y chromosome, trisomy 21, and XYY.   Less commonly reported in the literature are the non-genetic conditions associated with autism.   These include infectious illnesses such as congenital rubella, acute encephalopathy, and cytomegalovirus.  This chapter will examine fragile X syndrome, the most common specific neurological disorder associated with autism, and Rett disorder, which is most likely a specific neurogenetic disorder. 

Fragile X Syndrome

Fragile X syndrome (fra X) is a genetic disorder known to be due to defects in a single gene on the X chromosome. The estimated prevalence of fra X  is 1 per 2,500 to 1 in 4,0000 (34).  Affected individuals have a characteristic cytological marker on the X chromosome (a fragile site at Xq27.3), accompanied by a clinical picture of variable cognitive and physical features.  A DNA sequence spanning the fra X site has been cloned, and the FMR-1 gene (familial mental retardation-1) was found to be located at the fragile site (53). This FMR-1 gene is known to be expressed in the developing brain, as well as in other organs. The product of the FMR-1 gene is a protein (FMR protein) with 2 binding regions for RNA. It may function as a regulator of the expression of a variety of different messenger RNAs (1). In individuals suffering from fragile X syndrome, the FMR-1 gene is not expressed, because of the (CGG)n nucleotide triplet expansion described below. It has been suggested that in the absence of FMR protein, the normal regulation of certain messenger RNAs by the protein cannot occur; thus this genetic defect may be responsible for the pleiotropic phenotype associated with the fragile X syndrome (1). 

Fra X is recognized for its unusual inheritance pattern that is characteristic of triplet repeat expansion disorders.   For instance, although the affected gene is on the X chromosome, fra X does not follow the usual pattern of recessive X-linked inheritance, with only boys being affected. Instead, some females are also affected, and the degree of  expression of the disorder increases with successive generations, in both males and females. This phenomenon of increasing expression from one generation to the next has been referred to as the Sherman paradox (70),  or "genetic anticipation". Sherman noted that when an abnormal X chromosome is passed from a carrier male (negative cytogenetically, normal phenotype) through his daughter to her children, the children are likely to develop the symptoms of fra X, with boys at greater risk than girls. 

Genetic anticipation is explained in terms of the behavior of the fra X gene (FMR-1). Within this gene is a nucleotide triplet,  CGG,  which is repeated 5 to 50 times in normal individuals.  The vast majority of fra X subjects have a mutant gene in which the number of CGG repeats has been amplified to a number much greater than normal, typically 200 or more. Males in whom the number of (CGG)n repeats is greater than 200 almost invariably show clinical and cytogenetic signs of the disease, and are said to have the full fra X mutation.  Girls with the full mutation are also symptomatic but to a lesser degree than boys.  Carriers of the fra X gene actually have premutations (50 - 200 repeats) and are typically asymptomatic.   The phenotypic expression of fra X is extensively reviewed in Feinstein and Reiss (26).

The phenomenon of "genetic anticipation" is now reasonably well understood at a molecular level. In families with fra X positive children, the maternal grandfather can be demonstrated to have FMR-1 (CGG)n repeats at a level of 50 - 200. These carrier grandfathers are asymptomatic, and their moderately long (CGG)n repeats are known as premutations. The proband's mother inherits the 50 - 200 (CGG)n repeats, and in most cases she is asymptomatic. During oogenesis in the mother-to-be and/or early embryogenesis of the fetus, the premutation of 50 - 200 (CGG)n repeats can be amplified to more than 200 repeats.  Male children born to such a mother have a 50:50 chance of inheriting the defective X chromosome; those who do inherit the chromosome with the amplified full mutation are almost always symptomatic.  Female children also have a 50:50 chance of inheriting the defective gene. Those who do inherit the defective gene will have both a fra X chromosome from their mother and a normal paternal X chromosome. Only one X chromosome is active in a given cell; thus the greater the fraction of cells expressing the mutant X chromosome, the more likely a daughter is to suffer the stigmata of fra X. 

Autistic characteristics are observed in some children with fragile X syndrome.  A number of studies have reported the percentages of children/adults who meet full diagnostic criteria for autism and who also have fra X. The incidence of fra X in autistic populations ranges from 0% to 16% with a mean of 4% (24). This  wide range is probably due to the variety of sites in which patients are studied (residential versus outpatient),  to variable diagnostic criteria for autism, and to differences in the percentage of fra X-positive cells previously used as a diagnostic criterion.  For a discussion regarding the phenotypic similarities between autism and fra X the reader is again referred to Feinstein and Reiss (26).  It is clear that fra X puts individuals at increased risk for autistic behaviors.

Rett Syndrome

The neuro-degenerative disorder known as Rett disorder (RD), first described by Rett in 1966 (62), is also associated with autism.  To date, RD has been seen almost exclusively among females; incidence is estimated to be about 1 in 10,000 to 22,000 (33).  Infants with RD develop normally until approximately 6 months of age, when developmental delays and regression occur, and  microcephaly begin to appear. A striking characteristic of RD is the loss of purposeful hand movement, concurrent with the develop of stereotypic hand movements such as hand wringing.  Between the ages of 1 and 4 years children develop gait ataxia; as they get older, some develop spasticity and lose ambulation. Rett disorder is also characterized by a variety of additional symptoms including severe to profound mental retardation, autistic behaviors, seizures, and hyperventilation.

Despite a number of genetic and neurobiological studies, the etiology of this syndrome is not understood, and the genetic basis is uncertain. In a comprehensive review of RD (76), the recent developments and proposed etiologies are discussed.  These findings will be briefly summerized in this chapter. Based on patterns of inheritance, RD is considered to be a sex-linked dominant condition, lethal in males, with sporadic new mutations.   Many chromosomal abnormalities have been detected in individuals with RD, such as mosaicism for X chromosome terminal deletion.  Twin studies reveal 100% concordance in MZ twins and 0% concordance in DZ twins. A handful of families appear to be multiplex for RD; the syndrome may occur in 1st and 2nd degree relatives of unaffected mothers of RD girls. In these familial cases an explanation based on sporadic mutation would seem unlikely. To account for situations like this, other patterns of inheritance have been proposed, for example, non-random X chromosome inactivation.

A number of lines of evidence now argue strongly that a major gene for RD is located on the long arm of the X chromosome, and a number of candidate genes in this region have been examined. Xiang et al. (81) presented haplotype analysis of 9 families with at least 2 closely related females affected by classic RD. They concluded that the RD locus is likely to lie within Xq28, close to the DNA marker DXS15. They suggested two GABA receptor genes, GABRE and GABRA3, as candidate genes for Rett syndrome.

Co-morbidity of autism and RS has been noted (76).  During the early stages of the disorder (ages 2 - 5 years),  many children meet the diagnostic criteria for autism. Witt-Engerstrom and Gillberg (78) found that a majority of 47 children with RS had been diagnosed with autistic disorder, or with autistic features, prior to the diagnosis of RS.  Since many girls with RS also meet the diagnosis of autism, it is essential that RS be considered in the differential diagnosis of girls with autistic symptoms.

Autistic disorder is associated with several known genetic disorders as well as with a few non-genetic disorders.  An understanding of these disorders,  particularly those with a  genetic basis, should aid our understanding of the etiology of autism. This chapter has reviewed two such disorders,  each representing a different genetic mechanism.  Although the gene responsible for the fragile X syndrome is located on the X chromosome, the disease phenotype does not observe the classical pattern of recessive X-linked inheritance. Fragile X  thus  provides a new model for behavioral genetic disorders based on triplet amplification and genetic anticipation.  Rett syndrome provides a very different model; it occurs only in girls has an unusual mode of inheritance that may be due to the differential X-inactivation process occurring in female cells.   Although these two genetic disorders account for only a small fraction of the total cases of autism, the appearance of autistic characteristics in these genetic disorders should help us in unraveling some of the underlying neurobiological mechanisms of autism.



Genetic Studies

Multiple cases of autism within a single family have been noted, suggesting a genetic etiology.  As a consequence,  investigators have increasingly focused their efforts on genetic approaches in an attempt to identify the genes that cause autism.. These efforts now include twin studies, family studies, segregation analysis and linkage analysis.  As with almost all neuropsychiatric disorders, the genetics of autism does not follow simple mendelian inheritance patterns and appears to be quite complex.  This section reviews family studies, and discusses both segregation and linkage analysis.

Family Studies   In a recent review of autism family studies, Szatmari (74) reported recurrence risks of autism and other PDD diagnoses in first and second degree relatives.  The sibling risk for autism ranges from 2% to 6% (average 2.2%).   When other PDD diagnoses are included the sibling risk increases to 5.5%. Averaged over several studies the risk of autism for second degree relatives is 0.18%, and 0.12% for third degree relatives.

Twin studies strongly suggest that autism is a heritable disorder. In a recent study, Bailey et al. (7) examined same-sex twins ascertained from the school age population of Great Britain. Sixty-nine percent (11/16) of the monozygotic (MZ) twins were concordant for autism, whereas none (0/11) of the dyzygotic (DZ) twins were concordant.  When cognitive and social disabilities were included in the definition of the phenotype concordance increased to 87% for MZ and to 9% for DZ twins.  The other population based twin studies (reviewed, 74) confirm that the concordance rate for MZ twins is much higher than for DZ twins.  The MZ concordance rate for autism ranges from 36 - 91% and the DZ concordance rate is consistently 0%.  

Results from these family studies provide evidence that autism may be a polygenic disorder.  The sharp decline of autism risk from first to second/third degree relatives is inconsistent with a single gene disorder and more likely due to a polygenic pattern of inheritance.  The MZ/DZ ratio from the twin studies is also inconsistent with a simple mendelian inheritance pattern.  The expected MZ/DZ ratio is 2 for a single dominant gene and 4 for a single recessive gene.  The number of DZ twins available to study is small, thus its reasonable to use the sibling risk rate of 2 - 3% as an estimate of the DZ concordance rate.  If we do this, then the MZ/DZ ratio for autism is much greater than that expected for a single gene inheritance pattern and thus, more consistent with a polygenic inheritance pattern.

Family studies have also explored the presence of a milder phenotype among siblings, parents and second degree relatives of persons with autism.  A recent review of these studies can be found in Bailey et al. (10).  These studies have indicated that there may be a mild phenotype of autism found in first and second degree relatives that consists of impairments of cognition, communication, and socialization abilities and repetitive abnormalities.

Segregation Analysis:  Two studies of segregation analysis in autism have been published.  In the first Ritvo et al. (64) recruited 46 families with multiple incidence of autism. Families were drawn from several sources and not from a single population. The pattern of inheritance revealed in this analysis supported the hypothesis of an autosomal recessive pattern of inheritance for autism, and rejected a polygenic model. A second segregation analysis by Jorde et al. (39) surveyed 185 autism families ascertained in the Utah-UCLA Epidemiology study of autism. This survey suggested a multifactorial threshold model.

Linkage Analysis:   Several genetic linkage studies of autism have been published, and these include both candidate gene approaches and full genome screens.  In a linkage analysis of a psychiatric disease, one looks for linkage between the disease phenotype and specific DNA markers mapped to specific chromosomes. The DNA markers are polymorphic - that is, a given marker will have several DNA sequence variants that are distinguishable in the lab. It is this property of DNA polymorphism that makes the markers informative.

In a candidate gene approach a gene of known location is used as a DNA marker in an effort to establish association with a disease.  Candidate genes are chosen because they have functions that suggest they might be involved in a particular disease.  For example certain neurotransmitter receptor genes make good “candidate genes” for many psychiatric disorders.  The association of autism with fragile X syndrome has prompted researchers to test the FMR-1 gene as a candidate gene for autism. Hallmayer et al. (35) used DNA probes to examine the degree of amplification of (CGG)n repeats in a group of autism multiplex families.  No obvious abnormalities were detected in the size of the (CGG)n repeat region.  In addition, two microsatellite markers close to the fragile X gene were used in linkage analysis and no linkage with autism was detected.  Thus, idiopathic autism is not associated with amplification of (CGG)n repeats in the FMR-1 gene, nor is the autism phenotype positionally linked to this gene.  However, since the FMR-1 gene product binds to many RNAs, the association between fragile X disorder and autism may be due to the altered regulation of some of the still unidentified autism genes by the FMR-1 protein.

The much-studied hyperserotonemia of autism, and the positive response of some autistic patients to SSRIs (discussed later in this chapter) both suggest involvement of the serotonin system in autism, and thus serotonin-related genes are of interest as candidate genes. Two groups have conducted transmission disequilibrium tests  in the region of the serotonin transporter (5-HTT) gene on chromosome 17, with conflicting results.  Both groups took advantage of a reported polymorphism of the 5-HTT gene, involving short and long forms of the promoter region, with the longer form associated with greater expression of the gene. Cook et al. (17) found the shorter variant to be associated with autism. In contrast, Klauck et al. (43) found preferential transmission of the longer variant to autistic subjects. Since the two studies did not replicate each other, there was no confirmation of a candidate gene role for the 5-HTT gene. 

A specific candidate chromosomal region has also been tested for its possible linkage to autism. Chromosome 15 in particular was suspect, due to the association of autism with the cytogenetic abnormality of duplications within the 15q11-13 region (30). Using transmission disequilibrium testing in 140 autism singleton subjects, Cook et al. (16)  reported linkage between this chromosome 15 region and a marker located in the GABA receptor subunit gene (GABRB3). In a full genome screen using sib-pair analysis in 37 autism multiplex families, Pericak-Vance et al. (55) reported linkage between autism and another marker 5 cM distal to the GABRB3 marker on chromosome 15.  However, Salmon et al. (submitted) were unable to detect linkage in this chromosomal 15 region using both sib-pair linkage analysis and transmission disequilibrium testing of 90 autism multiplex families.  

Genome Screens: In a genome screen several hundred markers are used so that every chromosome has multiple markers evenly spaced along its length. One then looks for linkage of specific marker polymorphisms with disease phenotype in the population studied.  If linkage is seen between the disease phenotype and a specific marker or markers, a segment of the chromosome is identified as the potential neighborhood of a disease gene. Since there are roughly 50,000 - 100,000 human genes, but only a few hundred markers in most genome screens, localization is at best only approximate. The results of a genome screen are reported as lod scores, which are simply the likelihood ratios in favor of linkage at a particular chromosomal location. Lod scores are calculated in logarithms of base 10; so a lod score of 3.0 is a likelihood ratio of 1,000:1 in favor of linkage, while a lod score of -2.0 is a likelihood ratio of 100:1 against linkage.  The first such study in autism failed to show linkage between autism and 30 chromosomal markers (72).   A recent study by the International Molecular Genetic Study of Autism Consortium (38) identified a possible marker for autism on chromosome 7q; the Consortium reported a lod score of 2.5 for this region in 87 sib-pair families.  In a genome screen of 90 sib-pair families using approximately 400 markers, Risch et al. (in preparation) identified one region on chromosome 1 with a lod score greater than 2.  The results of this study suggest that there are no genes of major effect linked to autism; more likely autism is the results of interaction between many genes of modest effect. Risch et al. (in preparation),  failed to replicate the chromosome 7q findings (38), nor was linkage seen with the 15 q11 region (16, 55) or the 5-HTT marker (17, 43). 

In summary, twin and family studies strongly indicate a genetic basis for autism and point towards a polygenic model.  Genome screens have been carried out, but thus far no genes of major effect have been found, and it is beginning to appear that, when autism-associated genes are found, there will be many and each will account for only a modest part of the autism risk.  Thus, the genetics of autism is complex; multiple genes interact with each other and possibly also interact with environmental risk factors.  There is a suggestion from autism family studies that non-PDD family members have an increased frequency of cognitive and social differences, as might be expected if these families are genetically loaded with some of the multiple genes for autism.


Neuropathological Studies 

There have been relatively few autopsy studies of autism, but from these studies a neuropathology of autism is beginning to emerge, though not without controversy. Over a 15 year period Kemper and Bauman (42) have completed detailed autopsy studies on the brains of 9 autistic cases. They have now examined 8 males and 1 female, varying in age from 6 to 54, and have compared them to identically processed age- and sex-matched controls. These authors summarize their findings as follows: “their are 3 different neuropathologies in (autistic) brains: a curtailment of the normal development of neurons in the forebrain limbic system; an apparent congenital decrease in the number of Purkinje cells (cerebellum); and age-related changes in cell size and number of neurons in the nucleus of the diagonal band of Broca, in the cerebellar nuclei, and in the inferior olive.” The age related changes are particularly interesting, and point to “a prolonged process that extends from neuronal hypertrophy in childhood to atrophy and, in some nuclei, to cell loss in later life.”  In the cerebellum, a reduction of Purkinje cells was particularly noted in the posterolateral neocerebellum and adjacent archicerebellar cortex, while cells in the anterolateral cerebellum and vermis were relatively spared.

More recently, Bailey et al. (9) reported results of neuropathological examination of 6 mentally handicapped autistic subjects.  Four of the 6 brains were megencephalic, and most demonstrated developmental abnormalities in the brainstem. In the cerebellum, their results agree with those of Kemper and Bauman (42) in finding decreased numbers of Purkinje cells in most cases; breaks were also seen in the dentate ribbon. In the cerebral hemispheres, Bailey et al. (9) found a variety of developmental abnormalities in individual cases, including areas of increased cortical thickness and high neuronal density, and neuronal disorganization with focally increased numbers of single neurons in the white matter. The hippocampi appeared normal except for increased cell density in one case. In contrast, Kemper and Bauman (42) reported hippocampal abnormalities, including a decrease in cell size and an increased cell density in the hippocampi of their 9 subjects.

Bailey et al. (9) emphasize that as yet, no single pathology common to all cases of autism has been identified. As abnormalities have now been reported in brainstem, cerebellum, limbic system and neocortex, it seems likely that a combination of diverse developmental abnormalities, varying from case to case, may be responsible for the symptomatology of autism.   The significance of neuroanatomical findings in explaining either etiology or behavioral changes is uncertain; however they seem to suggest that the neuropathology of autism has its origins in early development of the brain, with pathological process in certain regions continuing into adult life.

Neuroimaging Studies:

Radiological imaging studies have been performed on autistic subjects using the full spectrum of imaging methodologies.  Although it is now clear that brain abnormalities are common in autism, no single abnormality is present in all subjects.  Indeed, many autistic subjects appear to have grossly normal brains.  Earlier studies using computerized tomography (CT) were reviewed in a previous ACNP edition (82).  The more recent studies, including magnetic resonance imaging (MRI) and positron emission tomography (PET), are briefly reviewed here. An appreciable number of structural MRI studies have been completed in the past few years, thus only a few will be reviewed here; for a more comprehensive review see Minshew et al. (52).

Structural Imaging:  The pioneering MRI investigations of autism (for review see ref.47) revealed nonspecific findings of enlarged fourth ventricle, increased fourth ventricle/posterior fossa ratio, decreased cerebellum/posterior fossa ratio, and decreased cerebellum/total brain ratio.  More recent MRI scans have focused on measures of total brain area and volume, as well as of specific structures of the cerebellum and the cerebrum.

Cerebellum:  The neuropathology studies of Kemper and Bauman (42) describing cerebellum Purkinje cell loss in autism called attention to the cerebellum as a possible site of anatomic abnormality in autism.  There have been a number of MRI studies of the cerebellum, with particular attention paid to the cerebellar vermi.  The studies of Baumann and Kemper do not, however, suggest involvement of the cerebellar vermi.  In several studies reviewed by Courchesne (22) investigators found that the vermal lobules VI and VII were on average smaller in autistic subjects than in normal controls.  Investigators who used control subjects matched chiefly for IQ failed to replicate these findings (reviewed, 27).  The disagreement between these studies over the relative size of the cerebellar vermal lobules is apparently due to the differences in the control groups used, and the current debate focuses on which subjects constitute the most appropriate control group (22, 27).  It appears that some autistic subjects have smaller cerebellar vermi VI and VII than do normal controls but these vermi do not differ significantly from those of IQ-matched controls.  To further complicate the picture, Courchesne has recently reported a subgroup of autistic subjects who have larger than normal vermal lobules VI and VII (21).

Cerebrum:  Piven et al. (58) found that autistic subjects had greater total brain volume, total tissue and total lateral ventricle volumes than controls matched for age, height and nonverbal IQ. This is consistent with enlarged head circumference in many persons with autism (8).  Follow-up MRI studies (57, 61) revealed that the increase in total brain size was accounted for by increased regional brain enlargement of the parietal, temporal and occipital lobes, as well as the cerebellum, but not of the frontal lobes.  Courchesne et al. (20) reported that a significant fraction of autistic subjects had parietal lobe abnormalities.  These abnormalities included loss of cortical volume, loss of white matter, and thinning of the corpus callosum.

Structures of the Cerebrum:  Increasingly, as MRI resolution has improved, investigators have turned their attention to the structures of the cerebrum.  Gaffney et al. (28) for instance, reviewed axial scans for potential abnormalities in the size of the thalamus and other structures of the basal ganglia; all were unremarkable except for the right lenticular nucleus, which was, on average, significantly larger in the autistic group.  Two studies revealed no differences in hippocampal volumes between autism and controls (60, 67).  Several MRI studies have reported reduced size of the middle and posterior portion of the corpus callosum in autistic subjects (25, 59). Since there is a concomitant increase in the size of the parietal, temporal and occipital lobes in these same subjects, there is a suggestion of abnormal development of neural connectivity between the cerebral hemispheres in autism. 

In a recent study that combines genetic and imaging approaches, Kates et al. (41) studied a pair of 7 year-old MZ twins discordant for autism; one fulfilled the criteria for classical autism, while the other had constrictions in social awareness and play, which are characteristic of the broader phenotype for autism. The twins were compared to 5 age- and sex-matched unaffected peers. Both twins had reduced volume of the superior temporal gyrus and frontal lobe relative to the control sample, while the classically autistic twin had, in addition, markedly smaller caudate, amygdaloid and hippocampal volumes, and smaller cerebellar vermis lobules VI and VII. These authors suggest that the abnormalities of subcortical structures that differentiate one twin from the other may underlie classical autism, while the shared cortical abnormalities may be characteristic of the broader autism phenotype.  

Functional neuroimaging:   Increasingly, the living brain is being examined with techniques that allow detection of localized regions of altered metabolism and blood flow. In autism, a few PET studies have been completed, but results are inconclusive, due to differences in subject selection, sensory state, and methodology used. For these reasons several groups have reported interesting results that have yet to be confirmed by others. For instance, in a study combining MRI and PET, Haznedar et al. (36)  found that, in 7 adult autistic subjects, a region of the right anterior cingulate cortex is metabolically less active and smaller than in controls. This area of the cortex may be involved in executive functioning. In another study, pronounced asymmetries have been reported  in brain images of a small number of autistic boys and their siblings.  Chugani et al. (15) used PET procedures to image the uptake of a serotonin precursor in the brain. They found marked asymmetries in uptake in the dentatothalamocortical pathway of their autistic subjects, compared with relatively symmetrical uptake in a sibling control group. The structures involved - that is the dentate nucleus of the cerebellum, thalamus, and frontal cortex - are important for language production and sensory integration. This is an interesting finding that needs to be replicated.

In summary, MRI studies are revealing abnormalities in the size of specific structures in more that one area of the brain.  There does not appear to be a single neuroanatomical abnormality associated with all cases of autism.  As is the pattern in neuropathological studies, abnormalities are seen in various parts of the autistic brain and investigators are not in complete agreement.  At present there is MRI evidence for the following structural abnormalities in autism: small cerebellar vermal lobules VI and VII, small posterior corpus callosum, and a generally enlarged brain volume.  These differences need to be studied further before any conclusions can be drawn about their involvement in the neurological mechanisms of autism. In the future, with improved functional imaging technologies (fMRI), it should be possible to build a consensus on which functional changes are relevant to autism symptomatology.

Neurochemical  Studies

Neurochemical studies of autism were begun in the early 1960s. Although there are clear abnormalities in levels of a number of neurotransmitters, as yet no one has yet put forward a unifying hypothesis of autism based on a specific defect in a single neurotransmitter system.  The large body of work on the neurochemistry and immunochemistry of autism is reviewed in the previous edition of this ACNP publication (82), and serotonin studies have been reviewed by Cook et al. (18).  Thus, only a selected group of more recent studies will be reviewed here. These studies have focused on serotonin and the opioids.

Serotonin  Serotonin (5-HT) has received the most attention in autism. It has been noted in a number of studies over the past 30 years that approximately 30% of autistic subjects have abnormally high levels of whole blood serotonin (80). Particular attention has been paid to 5-HT alterations because inhibitors of the serotonin transporter like fluoxetine (19, 23) have a beneficial effect on repetitive symptoms and aggressive behaviors in autism. Recent serotonin studies have focused on families, and several have found evidence that high blood levels of 5-HT are familial. Kuperman et al. (45) measured platelet-rich plasma 5-HT in families with an autistic proband, and found a significant correlation of 5-HT levels between parents and children. They reported elevated whole blood 5-HT in a majority of the first degree relatives of autistic probands with elevated 5-HT.  In contrast autistic probands with normal levels of 5-HT tended to have relatives with normal 5-HT levels (2, 18).   Autism multiplex families appear to have the highest levels of blood serotonin.  Piven et al. (56), for instance, compared 5-HT blood levels in 5 autistic multiplex children with levels in two control groups (23 singleton autistic children and 10 normal controls). Mean 5-HT levels were highest in the multiplex group and lowest in the normal group. The data thus indicate that high 5-HT blood levels are familial and may be associated with genetic liability to autism.

More recently McBride et al. (48) studied a group 77 autistic and 87 control subjects who varied by age, race, and diagnosis.  They found that race and age have a profound effect on whole blood 5-HT levels.  While prepubertal autistic subjects were hyperserotonemic compared to controls, race and pubertal status were more predictive of 5-HT levels than was the diagnosis of autism. Platelet 5-HT levels begin to drop at puberty, and postpubertal subjects were found to have lower levels of 5-HT than prepubertal subjects. In the postpubertal population the difference between autistic and control 5-HT levels narrowed considerably, compared to prepubertal subjects. In this study serotonin levels were not elevated in control subjects with mental retardation. The effect of race on blood 5-HT levels was marked, with white children having significantly lower 5-HT levels than black or Latino children. The authors suggest that, although hyperserotonemia is clearly a characteristic of a subgroup of autistic children, differences noted in earlier studies may have been overestimated through failure to control for race and pubertal status.

Opioids   Many autistic children have a very high pain threshold and exhibit self-abusive behaviors. A parallel has been noted between these behaviors and the behavior of animals addicted to opiates, where there is also a  decreased response to pain and an increase in self-injurious behavior (69, 54). Thus, there has been considerable interest in determining the levels of endogenous opioids in autism.  There have in fact been two reports of elevated endogenous opioids in autism; Gillberg et al. (31) noted higher CSF endorphin fraction II levels in  20 autistic children  compared to controls; Ross et al. (65) also noted high CSF beta-endorphin levels in autistic children (6 subjects, 8 controls).  These numbers are small, and confounding factors such as age, race and gender have not been carefully studied.

In summary,  although there are some abnormalities in measured levels of neurochemicals in autism,  a unifying theory remains elusive. The elevation of serotonin in blood found in some persons with autism is the most consistent finding, and it may be significant that drugs able to influence CNS levels of serotonin (clomipramine and fluoxetine) also appear to have therapeutic value in autism.



The treatment of autism has focused on extinguishing specific aberrant behaviors (e.g. hyperactivity, temper tantrums, stereotypies, and self-injurious behaviors) and improving social relatedness, use of language and overall cognitive abilities.  There are no cures for PDD, but treatment modalities have been developed to modify specific behaviors. Effective treatment requires an interdisciplinary approach that includes education, speech and language therapy,  occupational therapy, behavior modification and medications. The reader is referred to Berkell (11)  for a reviews of behavioral forms of treatment. This chapter will focus on the psychopharmacology of autism and PDD. 

During the last 25 years there have been a number of clinical studies of medications in autism.  The most studied and most useful agent historically has been haloperidol (13), which is known to improve many of the aberrant behaviors of autism. Recent studies have also focused on the effects of opioid blockers and serotonin reuptake inhibitors.  This review will summarize the systematic studies on effects of these psychopharmacologic agents in autism that have appeared since the publication of Campbell's review (13).  The Campbell review covered the period of the 70s and early 80s, when haldolperidol and fenfluramine were primarily being studied.  The reader is also referred to another recent comprehensive review by McDougle (49).

Neuroleptics   Campbell (13) has conducted several well designed double-blind placebo-controlled studies of haloperidol.  The behavioral improvements observed included significant decreases in social withdrawal, stereotypies, hyperactivity, abnormal object relationships, fidgetiness, negativism, angry affect, and lability.  However, recently published longitudinal study reported up to 33.9% of 118 children treated with haloperidol developed neuroleptic-induced dyskinesias (14).  Risk factors for development of neuroleptic-induced dyskinesias appeared to include greater cumulative dose, female gender and pre- and perinatal complications.  A single double-blind study on the new atypical neuroleptic risperidone has been published (50).  The study revealed that 57% of a group of autistic adults receiving risperidone responded favorably versus none of the placebo group.   The subjects of this study did not experience extrapyramidal effects or seizures, although the risk of tardive dyskinesia will not be determined until longer term treatment studies have been conducted. 

Opioid blockers   Opioid blockers such as naloxone and naltrexone have been examined for their potential effectiveness in modifying self-injurious behavior (SIB) in autism, and this field has recently been reviewed by Gillberg et al. (29).  They concluded that the opioid hypothesis for autism has produced little of clinical relevance and that naloxone, naltrexone and other opioid antagonists should not be in the first line of treatment for SIB. Though opioid antagonists have a positive effect on self-injurious behavior in several studies, the effects did not always rise to statistical significance. Naltrexone seems, however, to be free of serious negative side effects and thus may be of some benefit in patients with moderate or severe SIB.

Serotonergic agents   Motivated in part by biochemical evidence of hyperserotonemia in autism, researchers have evaluated a variety of serotonergic agents for their effectiveness in treatment of autism.  These studies first focused on fenfluramine, an agent that lowers both peripheral and central serotonin levels.  The results of these studies are contradictory, but  generally speaking, there appears to be only weak evidence that fenfluramine is an effective treatment for autistic behaviors (13). More recent studies have examined the role of serotonin reuptake blockers, including fluoxetine, clomipramine, fluvoxamine and sertraline. It seems reasonable that serotonin medications effective in the treatment of obsessive compulsive disorder might also be effective in moderating the perseverative and ritualistic behaviors seen in autism.

Clomipramine is a nonslective tricyclic medication that blockes the uptake of serotonin, norepinephrine and dopamine.  Nevertheless, clomipramine is the most potent serotonin reuptake inhibitor among the tricylic antidepressants.  Several studies of clomipramine in autistic subjects have appeared in the past few years.  Gordon and colleagues have conducted a double-blind crossover study using clomipramine, desipramine, and placebo (32). In their study of 24 subjects, they found clomipramine to be superior to both desipramine and placebo in reducing repetitive behaviors, compulsive behaviors, and anger.  Clomipramine and desipramine were equally effective in treating hyperactivity.  The second study focused on autistic adults.  Brodkin et al. (12)  conducted a prospective, open-label design in 33 autistic adults and reported a positive clinical response in 55% of the subjects.  The clinical improvements were seen in repetitive thoughts and behaviors, aggressive behaviors, as well as some improvement in social relatedness and verbal responsiveness.  A third study did not find any significant improvements in a small group of young autistic children, and some of the children suffered seizures during clomipramine treatment (68).

Investigators are also studying the effects of the more selective serotonin reuptake inhibitors.  In an open-label study, a majority of 23 autistic subjects (ages 8 to 27 years) demonstrated global improvement when treated with fluoxetine (19), although several subjects developed agitation requiring discontinuation of fluoxetine.  These positive findings were confirmed in another open-label study with young autistic children (ages 2 through 7 years) (23).  Twenty-two of the 37 subjects in this study experienced positive benefit that was sustained during the course of treatment (a mean of 21 months) without untoward side effects.  Treatment failure was related to development of hyperactivity, agitation and/or aggressiveness.  A double-blind study has been reported on fuvoxamine in autistic adults (51).  This study found that 53% of a group of autistic adults receiving fluvoxamine responded favorably compared to none of the placebo group.  Positive effects included decreased disruptive behaviors, such as repetitive behaviors and aggression, and improvements in social relatedness and use of language.  Side effects of fluvoxamine consisted of mild sedation and nausea.  Two groups have looked at the clinical effects of sertraline in small numbers of autistic subjects and found evidence of improvements in self-injury, aggression, and adverse behavioral reactions associated with environmental changes (73, 37). 

In summary, studies available to date indicate that there are beneficial effects of psychopharmacologic agents particularly in treating the more disturbing behaviors of aggression, agitation, hyperactivity and repetitive behaviors.  There is as yet little evidence that medications are effective in treating the core autistic symptoms of poor socialization and impaired language development, with the exception of some very preliminary findings with the serotonin reuptake inhibitors (23, 51).  There are still too few double-blind placebo-controlled medication studies in autism to permit more definitive conclusions.  As with other child psychiatric conditions, there are fewer studies in children due to a general reluctance to study medications in minors.  On an optimistic note, federal guidelines have recently changed to encourage psychopharmacological studies in children both by the pharmaceutical industry and by academic clinical researchers.  Since off-label applications are increasing in the PDD population, systematic studies of these medications is urgently needed.



Since my last review of autism for this text (46), the research landscape has changed somewhat.  Four years ago autism was considered to be a heterogeneous group of disorders with overlapping phenotypes.  Today, chiefly because of advances in genetics, it is seems likely that most cases of autism are the result of a polygenic disorder, in which many genes play a role.  A minority of cases are clearly discrete medical diseases (such as fragile X syndrome) that may be caused by single genes. According to this view, most cases of autism would involve slightly different but overlapping sets of susceptibility genes.   If this model of polygenic etiology is accurate, it would go a long way towards explaining many of the research findings discussed in this review.  It also would account for the variation in imaging and neuropathology results, in which abnormalities vary from individual to individual.  It would help explain why some abnormalities, like loss of Purkinje cells, are common to many persons with autism, but there is no universal brain abnormality.  The polygenic model would also explain why certain biochemical abnormalities, like hyperserotonemia, are present in only some cases of autism.

What does the future hold for autism research?  Hopefully, some of the genes responsible for autism susceptibility will be identified during the next few years.  These may be identified by more extensive and elaborate genome screens, but it is equally probable that success will come from candidate gene approaches.  The human genome project will provide an increasingly complete set of genes and alleles for testing.  There is also hope that functional brain imaging will highlight the neuropathophysiology of autism.  Functional descriptions of brain abnormalities common in autism should allow substantial improvements in both behavioral and pharmacological therapies.


Work from the author's laboratory was supported by grants from the National Alliance for Research on Schizophrenia and Depression (NARSAD), and from the Scottish Rite Benevolent Foundation's Schizophrenia Research Program, N.M.J., U.S.A.

published 2000