Although the existence of childhood schizophrenia was recognized early in the twentieth century (6), the term psychosis was used so broadly in children that a spectrum of behavioral disorders and autism were grouped together under the category of childhood schizophrenia (7). The landmark studies of Kolvin (1971) first established the clinical distinction between autism and other psychotic disorders of childhood (8). However, even today high rates of initial misdiagnosis remain due to symptom overlap, particularly for mood disorders, and the presence of relatively fleeting hallucinations and delusions in nonpsychotic pediatric patients (9,10). Anxiety and stress are probably the most common causes of hallucinations in preschool children, and the prognosis of these phenomena is usually benign (11). On the other hand, psychotic phenomena in school age children generally tend to be more persistent, and are more likely to be associated with drug toxicity or significant mental illness (2,3,12,52). Moreover, recent data from large birth cohort studies suggest that self-reported psychotic symptoms at age 11 years predicted a very high risk (odds ratio 16.4, confidence interval 3.9–67.8) of schizophreniform diagnosis at age 26 years, suggesting that psychotic symptoms probably exist as a continuous phenotype rather than an all-or-none phenomenon (14).

Kolvin’s work established that children can be diagnosed with unmodified criteria for schizophrenia, although such cases are rare (15,16). Childhood-onset schizophrenia (COS) shows a pattern similar to that of poor outcome adult cases, and the psychosis of COS can usually be distinguished by its severe and pervasive nature and its nonepisodic, unremitting course (5). Additionally, these children show poorer premorbid functioning in social, motor, and language domains, learning disabilities, and disruptive behavior disorders (17,18,19), and although not reported in studies of the premorbid history of adult-onset schizophrenia (20,21), transient autistic symptoms such as hand flapping and echolalia in toddler years are common (17,22), probably reflecting more compromised early brain development.

Childhood-onset schizophrenia is rare and must be distinguished from several childhood conditions that can manifest with psychotic symptoms and/or deterioration in function:

A sizeable, heterogeneous group of children referred to the NIMH childhood onset schizophrenia study over the past 15 years had transient psychotic symptoms and multiple developmental abnormalities, but were not adequately characterized by existing DSM-IV categories (31,32,33). In the DSM nosology these patients might be considered as having either psychosis NOS or mood disorder NOS.

The MDI group, although showing similarities with childhood onset schizophrenia, has distinct features which were used as the operational diagnostic criteria by the NIMH group, as listed below (31,32):

These children somewhat resemble syndromes such as the borderline syndrome of children, or a behavioral syndrome of affective dysregulation and impairments in social behavior but with more prominent autistic spectrum characteristics labeled as the multiplex developmental disorder (MCDD) (33,34,35,52). However, these syndromes have more predominant symptoms of pervasive developmental disorder; greater evidence of a formal thought disorder; and onset before age 5 (as opposed to about age 8 for the MDI group) (33,37,38,52).

Initially, the neuropsychological test profiles, smooth pursuit eye movement abnormalities, and familial risk factors suggested that some of the MDI children fell within the schizophrenia spectrum (32), but our MDI cohort appears to have a distinct long-term clinical course (described later), while none have progressed to schizophrenia (40).

A striking phenomenological feature of COS relative to adult onset schizophrenia appears to be the higher rates of early language, social, and motor developmental abnormalities, possibly reflecting greater impairment in early brain development. In the NIMH sample, premorbid development (defined as development prior to 1 year before psychosis onset and assessed using the Cannon-Spoor Premorbid Adjustment Scale (PAS)(43) and the Hollis premorbid development scale (19)) and social and speech and language impairments were clearly impaired in COS, as has been previously observed by four other independent research centers 17,18,19,22,44,45,46,47.

Since COS represents a more severe phenotype of schizophrenia than the adult onset illness, our initial hypothesis was that most risk factors identified in adult studies would be more striking in our very early onset cases.

The diagnosis of autism or pervasive developmental disorder (PDD) has been raised early in the development in our cases (n = 19, 25%) and several studies have claimed that autism per se might be a risk factor for later psychosis (68,69,70). Premorbid PDD may be a nonspecific manifestation of impaired neurodevelopment, and also may provide an independent additive risk factor for COS. We ascertained the premorbid diagnosis of past or current autism or PDD in 75 COS probands in whom the diagnosis of PDD was made according to DSM-IV criteria (American Psychiatric Association, 1994) based on early chart reviews and clinical interviews of patients and parents. The diagnosis of PDD was made in cases when per history there were clear symptoms of autistic/PDD spectrum present prior to the onset of psychosis that were still observable at the time of the NIH evaluation.