Since 2013 and the DSM-5, the diagnostic category of Autism Spectrum Disorder (ASD) has included 2 behavioral dimensions which represent the core defining features of ASD (1): (a) communication and social interaction deficits on the one hand, and (b) repetitive behaviors and restricted interests on the other hand (1). The term “ASD” suggests that the core features of the disorder can be measured dimensionally and, that they fall along a continuum of severity. No diagnostic subtypes are listed (1); instead, specifiers are provided to indicate associated dimensions, such as intellectual and/or language impairment, as well as the severity level of core ASD symptoms (2). Further, any known genetic or medical disorders are recorded and other co-occurring neurodevelopmental, mental, or behavioral disorders are indicated (2) to characterize subgroups.

In the DSM-5 (2), whatever the level of severity, even in the mildest forms, ASD symptoms are viewed as intrinsically autistic. However, to take one example, this conceptualization does not take into account the overlap between, on the one hand, some ASD symptoms, and on the other, Intellectual disability (3, 4), language impairment (5, 6), Attention Deficit/Hyperactivity Disorder (7, 8) or schizophrenia (9, 10). Moreover, mild ASD symptoms might be more akin to a natural variant of typical development than a formal disorder (11). Boundaries issues with typical development are also longitudinal, as has been shown in studies on “optimal outcome” or “Loss of Autism Diagnosis (LAD)” (12). These studies (13) focused on a group of individuals meeting criteria for ASD in childhood who no longer met them later in development. This definition entailed clear documentation of early ASD diagnosis, not meeting current diagnostic criteria (with ADOS scores similar to neurotypical peers with no history of ASD), and having overall cognitive, language, and social functioning standardized test scores within the average range (12). Confirmation that meeting criteria for ASD is not necessarily a lifelong state appears evident in the strikingly similar proportions (about 9%) of individuals experiencing LAD in prospective and retrospective studies (12). In line with the scope of our review, it is worth noticing that one study (14) showed that among ASD children included at age 2, those with PDD-NOS were significantly more likely than those with AD to move off the spectrum by age 4. ASD symptoms might be thus less specific than suspected in the DSM-5, especially in milder forms.

Envisioned in line with Kannner’s (15) first description and seminal work, autism is relatively consensual in its typical forms (full syndrome) (16). However the margins of the spectrum and their degree of extension are controversial (17) because of overlaps with other dimensions and blurred boundaries with typical development. This question is much more than a theoretical issue, since peripheral or “marginal” forms of autism might be more prevalent than central forms (see below). Furthermore, one could question the usefulness and relevance of recommendations made for typical forms when facing atypical ones. Past conceptualizations (18) contrasted the more prototypical form of autism, called “Autistic Disorder (AD),” with the margins of autism studied within the Pervasive Developmental Disorder - Not Otherwise Specified (PDD-NOS) category. Is therefore of special interest to explore these margins through the lens of this former diagnosis.

The PDD-NOS diagnosis category was first introduced in 1987 (19) to contrast with typical forms for autism labeled “Autistic Disorder” (AD). The DSM-III-R (19) states that “this category should be used when there is a qualitative impairment in the development of reciprocal social interaction and of verbal and nonverbal communication skills, but the criteria are not met for autistic disorder, schizophrenia or schizotypal or schizoid personality disorder. Some people with this diagnosis will exhibit a markedly restricted repertoire of activities and interests, but others will not.” In DSM-IV (20) and DSM-IV-TR (18), AD and PDD-NOS were included within the Pervasive Developmental Disorders (PDDs) that prelude ASD. PDDs also included Asperger Syndrome (AS), Childhood Disintegrative Disorder and Rett Syndrome (18, 20).

In DSM-IV-TR (18), the PDD-NOS category included atypical autism, i.e., presentations that did not meet the criteria for AD because of late age of onset, atypical symptomatology, or subthreshold symptomatology, or all of these. These recommendations echoed Towbin’s conceptualization (21) described in a reference textbook section published in 1997. He recommended the use of PDD-NOS in 4 indications: (a) as a temporary diagnosis when data are lacking or when the child is too young; (b) to designate a mild form of ASD; (c) in the case of late age of onset of autistic symptoms; (d) to depict a clinical picture with early symptom onset and impairment in social reciprocity (21). In terms of diagnostic use, the PDD-NOS, often envisioned as a diagnosis of subthreshold autism, seems to cover a wide range of clinical variability, sometimes referred to as heterogeneity; it would benefit from being studied in a more positive manner, by drawing on its clinical diversity (22).

Epidemiological studies show that at least half of PDD were PDD-NOS. In 2005, Fombonne (23) found a PDD-NOS prevalence of 37.1 per 10,000 for a PDD prevalence of 63 per 10,000 and an AD prevalence of 13 per 10,000. In 2010 in the United States, the results of the Centers for Disease Control and Prevention (CDC) showed a PDD prevalence of 14.6 per 1,000 in a population of 363,749 8-year old children (24). PDD-NOS were 46% of the PDD, AD 43% and AS 11%. Noticeably, between 2006 and 2010 the distribution of AD and PDD-NOS has not undergone substantial changes (24).

The diagnosis of PDD-NOS had a low inter-rater reliability. The inter-rater reliability was good to excellent (k from 0.95 to 0.67) for the distinction between PDD and non-PDD. However, the disentanglement of different subtypes of PDD was much less reliable (k from 0.18 to 0.65) (25–27). The DSM-IV-TR criteria were therefore more successful in differentiating PDD (25) from other psychiatric disorders than in distinguishing between the three main subtypes of PDD, namely AD, PDD-NOS and AS (28). In line with these results, a study (29) tried to assess the variation between behavioral phenotypes and clinical diagnoses of different autism spectrum disorders across 12 university-based sites. They found that clinical distinctions among categorical diagnostic subtypes of PDDs (AD, PDD-NOS and AS) were not reliable even across sites with well-documented fidelity, using standardized diagnostic instruments (29).

In spite of its low reliability, the former diagnosis of PDD-NOS is very interesting to study the margins of autism because of both its conceptualization at the end of the spectrum, and its high prevalence. The aim of our systematic review is thus to investigate the clinical characteristics of PDD-NOS in comparison with AD. The objectives were therefore to study comparatively (in both diagnoses) synchronic and diachronic characteristics such as autism symptoms, cognitive function, daily impact, associated mental health or physical health condition and developmental trajectory from infancy to adolescence. The comparison is intended to address the distinction between the typical form and the margins of autism as well as the relevance of the PDD-NOS/AD distinction. Our study therefore aims to determine whether this distinction helps mental health professionals to account for diversity within the spectrum, and to identify divergent trajectories, better than the current ASD category.

This study follows the Preferred Reporting Items for Systematic Reviews and Meta-analyses reporting guidelines (PRISMA) (30).

We conducted a systematic review of PDD-NOS in comparison with AD. Cross-sectional and longitudinal studies comparing PDD-NOS and AD showed a clear trend for lower communication and socialization impairment in children with PDD-NOS. Results on restrictive and repetitive behaviors were more contradictory. Early development was less impacted in PDD-NOS as well as the quality of life of children and families. Children with PDD-NOS showed less challenging behavior than those with AD. Schizophrenia and mood disorders appeared to be higher in PDD-NOS whereas feeding and sleep disorders tend to be less prevalent.

With regard to our results, PDD-NOS appeared as category both quantitatively and qualitatively distinct from AD. In a quantitative perspective, PDD-NOS was a lesser variant of autism with respect to the trajectory of autism symptoms, some associated mental and physical health conditions, or quality of life. This perspective is consistent with the spectrum approach of the DSM-5 (2) and a possible continuum between AD and PDD-NOS within ASD. However, in a qualitative perspective, PDD-NOS differed from AD because of higher associated schizophrenia and mood disorders. These results were unexpected with respect to a spectrum approach and suggest specific links between the margins of autism and other psychiatric dimensions. Our review coheres with Tsai and Ghaziuddin’s review (109) which concluded that the literature appeared to suggest that PDDNOS can be separated from AD, and, therefore, does not seem to lie on a “continuum” with that disorder. These authors recommended systematic research to settle the issue and that PDDNOS should be regarded as a separate category to facilitate ongoing research (109). In contrast, a recent systematic review on PDD-NOS (110) concluded that in regard of the limited consistency, sensibility, and stability of this disorder, it would be more consistent to include this diagnosis within ASD. Noticeably, the authors still recommended to collect information from professionals on adults who received a PDD-NOS diagnosis in childhood.

Our search on PubMed found 510 references. On the same period, from 1987 to 2022, the term “Autistic disorder” gave 24,639 references. PDD-NOS was thus referenced 48 times less than AD, and was therefore much less studied. However, the high prevalence of the PDD-NOS category demonstrates its high clinical relevance. This clinical importance of PDD-NOS in psychiatry was also emphasized by a study comparing youth with PDDs attending a specialty clinic for autism with those attending a general psychiatry clinic (88). More autism clinic youth met criteria for AD; more of those referred to the psychiatry clinic met the criteria for PDD-NOS (88). The discrepancy between the past clinical use of the PDD-NOS category and the scarcity of research in the field (in comparison with AD) might be due to the difficulty of studying the margins of autism because of their heterogeneity and complexity - which our review aimed to underline. PDD-NOS therefore illustrated a gap between clinical and research issues and the need to address heterogeneity from a research perspective. Based on research rather than clinical consideration of the PDD-NOS, the DSM-5 missed the complexity of the margins and opted for a continuum approach and ASD.

With regard to the 4 recommendations proposed by Towbin on the use of the PDD-NOS diagnosis category (21), it appears that, firstly, although PDD-NOS is related to earlier developmental disorders, the age of parental concern does not differ between AD and PDD-NOS. The latter result suggests that the PDD-NOS is not commonly used in the case of a later age of onset of autistic symptoms. Towbin (21) and Mandy et al. (35) suggested that PDD-NOS depicted a clinical picture with early onset of symptoms, an impairment in social reciprocity and less to no restrictive and repetitive behaviors. We showed, however, the lack of clear support to the claim of lesser frequency of restrictive and repetitive behaviors in PDD-NOS. Towbin also recommended using PDD-NOS as a temporary diagnosis when data are lacking or when the child is too young (21). This recommendation concurs with less diagnostic stability in PDD-NOS in infants and pre-schoolers, as is shown by our results. Finally, our review mostly supports the recommendation referring to PDDS-NOS as a lesser variant of autism with respect to the trajectory of autism symptoms, associated mental and physical health conditions, or quality of life.

Chien et al. showed that PDD-NOS was more associated than AD with schizophrenia and mood disorder (92). This result concurs with Joshi et al.’s study comparing youth with PDDs attending a specialty clinic for autism to those attending a general psychiatry clinic (88). General psychiatry clinic youth with PDDs (with more PDD-NOS than AD) suffered from more Major Depressive Disorder and psychosis. This result echoes the results of the Sporn et al.’s retrospective study, showing that 25% of the 76 patients who reported early onset schizophrenia had a diagnosis of PDDs during their childhood, with a clear predominance of PDD-NOS (111). In a reference sample of childhood onset schizophrenia, Rapaport et al. also found a high frequency of PDD-NOS (112).

The importance of associated psychosis and mood disorder in PDDs also echoes past syndromic conceptualization such as Multiple Complex Developmental Disorder (MCDD) (113), Pathological Demand Avoidance (PDA) (114) or Multi-Dimensional Impairments (MDI) (115) which linked social impairment, emotional lability and signs of psychosis (without a diagnosis of early onset schizophrenia). These signs of psychosis referred to thought disorder (MCDD), delusions (MCDD), poor ability to distinguish fantasy from reality (MDI, PDA). It is important to stress that MCDD and PDA were initially coined to deal with heterogeneity within PDD-NOS. However, both syndromes were not limited to the PDD-NOS category (116, 117).

The results of our review highlight that PDD-NOS was of great interest because of its high prevalence and clinical relevance. This diagnosis was useful to take into account milder forms of autism. However, it was less reliable and less stable through time than AD. Moreover, a qualitative study emphasized (118) that the autistic community might not acknowledge this diagnosis. This study was a phenomenological analysis of discussion forum dialog among 76 adult participants with self-reported diagnosis of AD and AS but not PDD-NOS (118). Participants constantly agreed that PDD-NOS should be removed in the DSM-5. Several participants expressed concern that the PDD-NOS diagnosis was not specific and reliable enough. They called the diagnosis “a junk category,” “a receptacle” (118).

Our review emphasized lower socio-communication impairments in PDD-NOS than in AD. In DSM-5, ASD diagnosis required three criteria out of three in this domain, namely deficits in social–emotional reciprocity, deficits in nonverbal communicative behaviors used for social interaction and deficits in developing, maintaining, and understanding relationship (2). To diagnose PDD-NOS, only two of them was required (18). It might therefore be suspected that the social and communication deficit characteristic of PDD-NOS wasn’t important enough for it to remain in the autism spectrum. A review and meta-analysis (119) showed that 70% of PDD-NOS did not meet the criteria for ASD. Another hypothesis postulated that PDD-NOS might belong to the Social Communication Disorder (SCD) diagnosis category. However, this category must meet all three criteria for the social-communication domain, as in ASD (2). Thus, it appears that PDD-NOS has been excluded from ASD and SCD.

Our review challenged initial views of PDD-NOS and showed that this diagnosis mostly included milder forms of autism. PDD-NOS was very clinically relevant to deal with the margins of autism and the diversity within the spectrum. However, PDD-NOS encountered too many limitations (low reliability, instability through time, low acceptability) to be retained in the DSM-5. The fact is that many PDD-NOS did not join either ASD or SCD. Thus, in order to take into account past research on PDD-NOS and the margin of autism, we argue for the creation of a new category alongside ASD and SCD, based on three main dimensions related to socialization impairment, emotional lability and psychotic symptoms and coined on PDA, MCDD and MDI. Criteria for this new category could be selected from the most valid criteria for these three syndromes and tested against the ASD criteria, to ensure that they are truly distinct. In our view, this new category would provide an opportunity for mental health professionals, along with SCD and ASD, to better take into account the neurodevelopmental disorders that have an impact on the development of social communication and its heterogeneity. From a clinical point of view, this new category would trigger the search for associated psychotic symptoms or emotional lability that might currently be understudied by professionals. In addition, new developmental trajectories, specific interventions and treatments could be identified, which would in turn make it possible to better inform and help the people concerned and their families.

The first limitation of our study comes from the terms selected for the database search. Few studies directly addressed differences between PDD-NOS and AD. However, when the DSM-IV-TR was in use, some studies included participants with PDDs including those with AD, PDD-NOS or another PDD subtype often compared to controls. In the latter case, PDD-NOS was only mentioned in the abstract or in the title when there was a difference within PDDs subgroups. That is why we may have over-selected references showing a difference between PDD-NOS and AD, and under-selected negative studies that did not. The second limitation is that we studied all clinical features of PDD-NOS. For example, some characteristics such as quality of life were rarely studied and were only related to two references. The third limitation is sample overlap between studies, as 25% of the studies were from Matson et al. Finally, our results cannot be extended to females with PDD-NOS, individuals with PDD-NOS and ID, or adults with PDD-NOS.

Our systematic review shows that PDD-NOS corresponds to milder form of autism with less impact and less associated disorder, with the exception of schizophrenia and mood disorder. Our review challenges initial views of PDD-NOS, echoes past syndromic conceptualizations such as MCDD, PDA or MDI, and argues for a quantitative and qualitative distinction between AD and PDD-NOS. The PDD-NOS diagnosis was very clinically relevant to deal with the margins of autism and the diversity within the spectrum, and therefore massively used by professionals in the field. However, PDD-NOS had many limitations (low reliability, instability through time, low acceptability) and was therefore understudied, thereby generating a discrepancy between clinical and research issues. Thus, in order to overcome this gap and to take into account past research on PDD-NOS, we suggest taxonomic changes in DSM-5, through the introduction of a new category based on three main dimensions: socialization impairment, emotional lability and psychotic symptoms. Future studies are therefore needed to test relevant criteria for this new category and possibly identify developmental trajectories, specific interventions and treatments.

RC and OP designed the systematic review and selected the studies following PRISMA guidelines. OP was contributed mainly in epistemological and methodological. RC wrote the first complete draft of the manuscript. OP, NC, and CS provided substantial modification to the manuscript. All authors contributed to the article and approved the submitted version.