Background
Prader-Willi syndrome is a neurodevelopmental disorder that results in a complex behavioral and developmental phenotype. Caused by a lack of paternally derived imprinted genes on chromosome 15q11-q13, people with Prader-Willi syndrome (PWS) typically manifest mild to moderate intellectual disability, compulsivity, rigidity, irritability, social dysfunction, growth hormone deficiencies, and hyperphagia that can lead to life-threatening obesity [
1]. Most cases of PWS (65–75%) are caused by paternal deletions in the 15q11.2-q13 region and are further characterized by size. Type I deletions are approximately 500 mb larger than type II deletions. Some individuals have atypical deletions that do not encompass the breakpoints commonly seen in type I or II cases [
2]. Approximately 20–30% of PWS cases are due to maternal uniparental disomy (mUPD), when both copies of chromosome 15 are maternally inherited. Occasionally, individuals have paternally inherited imprinting defects (1–3%; for a review see [
3]).
Individuals with mUPD (versus deletions) are at higher risk for autism symptoms or autism spectrum disorder (ASD). They are also at heightened risk for psychotic illness, often with a depressive or affective component, which typically begins in adolescence or young adulthood [
4,
5]. High risks for both disorders presumably stem from the duplication and overexpression of maternally expressed genes in the 15q11-q13 region, including UBE3A and ATP10A [
6]. As well, persons with isodicentric 15 syndrome often show ASD or autism symptoms, and maternally inherited duplications of the 15q11-q13 region are relatively common findings in genetic studies of idiopathic autism, seen in 1–3% of these cases [
7,
8]. Although rare, point mutations in paternally imprinted genes in the PWS region, specifically in
MAGEL2 [
9] and the snoRNA region [
10], were identified in a handful of children with co-occurring ASD and PWS (or strong PWS features).
Although disruptions in the PWS 15q11-q13 region are thus often implicated in ASD, data are relatively limited on rates of ASD in PWS. Studies to date are of some concern, as most have relied on autism screeners that are completed by parents, and have not directly observed or evaluated offspring with PWS. Autism screeners identify children in need of further evaluation and are not meant to be diagnostic. Even so, such tools are widely used in research to index probable ASD. Veltman et al. [
11] administered the Autism Screening Questionnaire (later called the Social Communication Questionnaire (SCQ); [
12]) to parents of 63 offspring with PWS aged 1 to 48 years and found that 36.5% scored above the ASQ cut-point; this rate was higher (41.4%) in participants older than 4 years of age. Those with mUPD versus deletions had higher ASQ scores. Compared to controls, Descheemaeker et al. [
13] reported that 59 individuals with PWS aged 2 to 51 years had significantly elevated scores on the Pervasive Developmental Disorder Questionnaire, also completed by parents. Examining SCQ scores in 44 participants with PWS aged 3 to 37 years, Flores et al. [
14] found that 35% of those with mUPD and 16.7% of deletion cases scored above the SCQ cut-point. Finally, two comprehensive literature reviews, conducted 10 years apart, found remarkably similar rates of ASD in PWS, 25.3% [
15] and 26.7% [
16], and higher rates of ASD were noted in mUPD cases (37.7 and 35.3%) versus those with paternal deletions (18.5 and 18.5%). No study included in these two reviews used direct observations of children to establish ASD diagnoses; most relied on screeners and a few on clinical diagnoses from unspecified sources. Beyond these concerns, the studies noted above did not address methodological challenges posed by the wide age ranges of participants, from very young children to adults.
A second complication involves the salience of repetitive behaviors in the PWS phenotype. Clinically, we find that families, educators, and other professionals often use these easily observed behaviors to raise suspicions of co-occurring ASD in PWS, even though such behaviors alone are not diagnostic of ASD. Repetitive behaviors in PWS are indeed highly reminiscent of those seen in ASD, including insistence on sameness, repetitive questioning or speech, and ordering and arranging items [
17‐
19].
Comparing children with PWS versus ASD on the Childhood Routines Inventory [
20], Graeves et al. [
21] reported similarly high levels of repetitive and “just right” behaviors in both groups. In contrast, Flores et al. [
14] used the Repetitive Behavior Scale-Revised (RBS-R; [
22]) to compare 45, 3 to 37-year-olds with PWS to 207 children with ASD. The PWS group scored significantly lower than the ASD group, especially in the RBS-R’s restricted, ritualistic, and self-injurious behavior domains. It remains unclear, however, if these behaviors are differentially expressed in those PWS only versus PWS and co-occurring ASD.
The current study addresses these methodological concerns and also provides a more complete characterization of children with PWS and ASD compared to those with PWS only. We administered the Autism Diagnostic Observation Schedule (ADOS-2; [
23]), and the Social Communication Questionnaire (SCQ; [
12]) to 146 children with PWS aged 4 to 21 years. Best-estimate ASD diagnoses were made using expert clinical reviews of ADOS-2 videotapes, scores, and interview data from parents regarding their children’s current and previous educational and developmental functioning. Participants with PWS + ASD versus PWS only were subsequently compared across PWS genetic subtypes, age, gender, and their test scores from cognitive, adaptive, and repetitive behavior assessments. Finally, the study identified how well the SCQ and ADOS-2 performed in predicting ASD status as determined by the clinical review team.
Discussion
This is the first large-scale study to identify the rates and characteristics of ASD in a large cohort of children and youth with PWS based on standardized autism assessments combined with thorough clinical reviews. Findings provide new insights into how children with PWS + ASD differ from those with PWS only in cognitive and adaptive functioning, repetitive and social behaviors, gender, and PWS genetic subtypes. Further, by determining how well the SCQ screener and ADOS-2 performed in relation to ASD diagnoses, we generate new recommendations for future research.
Using best-estimate ASD diagnoses, our 12.3% rate of ASD in 146 children with PWS falls below the 25 to 41% reported in previous studies [
15,
16]. Several factors likely account for this lower rate. First, previous estimates of ASD in PWS may be inflated because studies only reported data from ASD screeners completed by parents. Although screeners are intended to identify individuals who need further evaluation, previous PWS studies did not report taking the necessary extra steps of directly evaluating those who screened positive.
Second, the clinical team may have been too conservative in making ASD diagnoses, as reflected in the 14 children who were positive on the ADOS-2 but were judged by the team to not have ASD. It is important to emphasize that the above threshold scores on the ADOS-2 are not a substitute for a diagnosis of ASD. Instead, this observational schedule provides valuable data for experienced clinicians to use in their diagnostic formulations, along with children’s developmental and medical histories and current cognitive, adaptive, and behavioral functioning [
24].
The importance of this multi-modal evaluative process is highlighted by analyses of the 14 children who were ADOS-2 positive but clinically negative. Although these children had, on average, similar cognitive and adaptive behavior scores as the PWS + ASD group, they scored significantly lower than both the PWS-only and PWS + ASD group on the Vineland’s communication domain. Following up this finding with developmental history data, the ADOS-2-positive, clinically negative group had higher rates of speech problems than their peers, especially in verbal apraxia or difficulties getting thoughts into words. Anecdotally, the research team observed that these children typically had reciprocal social intentions, but their struggles to communicate diminished the quality of their interactions. Because we did not obtain external confirmation of parental reports of verbal apraxia, or quantify the team’s clinical impressions, these results should be interpreted cautiously. Even so, socially reciprocal intentions in children with diminished communication skills offer a reasonable explanation for why these children did not receive ASD diagnoses. Our analyses of this group also aptly demonstrate the value of thorough clinical reviews of ADOS-2 and other pertinent child data in formulating ASD diagnoses.
A final possible bias is that children’s levels of cognitive functioning may have overly influenced the team. This is particularly relevant as children with PWS + ASD versus PWS only had, on average, lower IQ and adaptive behavior standard scores. Biases in relation to IQ could go in two directions. First, it is possible that the team simply did not recognize ASD in children who were higher functioning. This possibility, however, is offset by the fact that 33% of the PWS + ASD group had KBIT-2 composite IQs of 70 or higher (M = 86.88, SD = 14.09, range 72–112). A second possibility is that the team was more inclined to make ASD diagnoses in lower functioning children. However, the team’s evaluation of 14 children who were ADOS-2 positive but clinically negative argues against this possibility. These 14 children had relatively low IQ’s (M = 59.26, SD = 16.93), and 80% of them had IQ’s <70. Despite reaching ASD threshold on the ADOS-2, this group of predominantly lower functioning children was judged not to have an ASD diagnosis. The team thus appeared to adequately discriminate ASD across the IQ spectrum.
Unlike their counterparts, those with PWS + ASD had elevated stereotypies and restricted interests; both are highly characteristic of ASD. Relative to those with PWS only, the PWS + ASD group had lower verbal IQs and adaptive daily living and socialization skills. For the sample as a whole, ADOS-2 overall and social affect calibrated severity scores were negatively correlated with KBIT-2 IQ scores, especially the verbal IQ. Although calibrated scores were developed to minimize the influence of cognition and other child factors, verbal IQs still account for approximately 10–11% of the variance in the overall and social affect calibrated severity scores [
24,
25]. A recent meta-analysis of 12 neurodevelopmental syndromes (not including PWS) found higher rates of ASD risk and symptoms in syndromes characterized by low or variable IQs [
32]. Improved understandings of cognitive differences between idiopathic and syndromic ASD will help research that frames genetic syndromes as promising alternative windows into genetic or neurobiological factors associated with ASD in general [
33,
34].
Although syndromic autism may be more equitably distributed across gender [
35], we found that children with PWS + ASD were more likely to be male. It is unclear if this finding is atypical or not because previous researchers have not generally reported the gender of participants with PWS who screen positive for ASD. Until future studies can clarify this gender finding, clinicians should not use gender to help rule ASD diagnoses in or out in children with PWS.
ASD’s were predominantly seen in those with the PWS mUPD subtype; they comprised 78% of the 18 children with ASD diagnoses. Overall, 25.5% of the 55 participants with mUPD were deemed to have an ASD. mUPD is thought to stem from a rescue of trisomy 15 caused by nondisjunction of maternal chromosomes and subsequent discard of the paternal chromosome 15 [
6]. ASDs were also found in two of the seven children with imprinting defects. Individuals with imprinting defects have chromosome 15’s inherited from both parents, but the paternal chromosome contains a maternal imprint, resulting in loss of paternally expressed genes in the 15q11.2-q13 PWS region. As a result, imprinting defects are functionally similar to mUPD. With the addition of the two imprinting defect cases, the rate of ASD in mUPD increases slightly to 28.0%, but still falls below previous estimates of ASD in mUPD based on screeners.
Repetitive behaviors can be meaningfully subdivided into at least two broad domains; so-called lower order repetitive sensory and motor behaviors and “higher-order” insistence on sameness ([
36,
37], see also [
38] for evidence of a third “circumscribed interests” factor). Aside from those with PWS + ASD, stereotypies and restricted interests were neither frequent nor problematic in this cohort. Although the self-injury domain was relatively high, this was driven by a single behavior—skin picking. Stereotypical motor and senory behaviors in ASD are often negatively associated with IQ scores, while insistence on sameness is not [
36]. This same pattern of results was also found in our PWS cohort.
The most frequently occurring repetitive behaviors, seen in 76–100% of participants, involved compulsivity and insistence on sameness in routines, events, timing of events, repetitive questioning, becoming upset if interrupted, hoarding, and ordering and arranging items. Although not correlated with IQ, the RBS-R sameness and compulsive domains were robustly and negatively associated with the Vineland’s communication, daily living skills, and socialization domains. Clinically, we find that these behaviors often impede optimal adaptive functioning and are among the most difficult for parents to manage. Insistence on sameness in ASDs has been associated with specific genetic alterations [
39‐
41], including linkages to one of several GABA
A receptors located in the PWS 15q11.2-q13 region [
42]. PWS may thus serve as a promising model for understanding insistence on sameness in ASD in general.
While the PWS-only group did not meet threshold for ASD, they still had relatively frequent problems in sustaining conversations and in the quality and amount of their social interactions (see also [
43,
44]). These findings raise the intriguing possibility that deficits in social perception or cognition in most individuals with PWS do not index autism per se, but instead contribute to other emotional, behavioral, or psychiatric disorders. Indeed, both neural and genetic studies suggest similarities between PWS, ASD, schizophrenia, and psychosis [
45]. Examining evoked response potential’s (ERP) to images of faces, Key and colleagues [
46] found that individuals with PWS due to mUPD, but not deletions, showed a lack of visual ERP face-specific amplitude increase in N170, a robust pattern also seen in autism and schizophrenia [
47]. Based on structural magnetic resonance imaging, Lukoshe et al. [
48] concluded that children with PWS due to mUPD have early deviations in brain development that are reminiscent of those in ASD or schizophrenia. Finally, several genes in the PWS 15q11-q13 region are also implicated in schizophrenia or psychosis for a review see [
49]. Future work is needed that identifies specific deficits in social cognition and perception in PWS, and how these might relate to psychosis or other psychiatric disorders.
The performances of the SCQ and ADOS-2 underscore the importance of combining ASD evaluation tools with thorough clinical reviews. With the SCQ’s low positive predictive value (PPV), there was just a 29% chance (49% using the adjusted PPV) that children with PWS who screen positive will indeed have ASD. Both of these PPVs fall at or below chance levels. The ADOS-2 performed much better, with enhanced specificity and sensitivity, but it too yielded a PPV just above chance (56%), which improved to 75% using the adjusted PPV. Comparing agreement across the two measures, they did a better job agreeing on the absence versus the presence of ASD (PNP = 76.8% versus PPA = 58.8%). However, we also determined if the measures agreed or disagreed correctly, i.e., in relation to ASD diagnoses. The SCQ and ADOS-2 correctly agreed on children’s ASD status 67.8% of the time, far more than they agreed but made the wrong call (4.8%). However, when the measures disagreed, the SCQ made more wrong calls (22.6%) than the ADOS-2 (4.8%), and most of these were false positives.
On the one hand, the high false positive rate of the SCQ indicates that the SCQ is performing as it was intended—to identify children in need of further evaluation. And in clinical or educational settings, a high false positive rate is often acceptable so that more children are evaluated and none are missed who could benefit from ASD interventions. Difficulties arise, however, in PWS research that only uses ASD screeners, without follow-up evaluations of screen-positive cases. Such practices risk creating a false impression that ASDs may be quite widespread in PWS.
Our findings recommend two options for future PWS research. The fact that ADOS-2 and SCQ correctly agreed for 68% of participants suggests that the two instruments performed better together than at least the SCQ did alone. If researchers want to reduce error due to false positives, they should consider using agreements between at least two standardized indices of ASD and acknowledge that subsequent rates of ASD in PWS may be still inflated. If researchers aim to precisely identity ASD in PWS, then they need to use the ADOS-2 or other direct assessments of children. Further, these observations need to be placed in the broader context of children’s development, current functioning, and phenotypic features. Indeed, Hepburn and Moody [
50] convincingly argue that children with genetic, neurodevelopmental disorders must be evaluated for ASD in the context of their syndromic phenotype and developmental stage. In PWS, for example, it is unclear what role the syndrome’s characteristic infantile hypotonia, failure to thrive and growth hormone deficiencies might play in the expression of early indicators of ASD in joint attention, shared affect, imitation, and social attention, gestures, or responses [
51]. It is also unclear how such social deficits respond to growth hormone replacement therapy. Beyond expected improvements in linear height, body composition, and muscular strength, growth hormone therapy also boosts cognitive and adaptive skills in children with PWS [
52].
This study had several strengths, including a large, well-characterized sample and standardized, multi-modal ASD assessments that were reviewed by experts in ASD and PWS. Even so, several shortcomings need discussion. First, the study did not include a separate measure of language function and speech/language data relied solely on developmental history interviews with parents. Direct testing of language function would have been particularly helpful in characterizing the 14 ADOS-2-positive but clinically negative children. Most children with PWS have language delay, oral motor difficulties, poor articulation, slow rate of speech, flat intonation, and language skills that may fall below their level of cognitive functioning [
53,
54]. Although under some debate, recent work finds that children with ASD do not typically manifest characteristics of childhood apraxia of speech [
55]. Studies have yet to determine how or if motor speech difficulties in children with PWS map onto apraxia or are better described by other language disorders.
Another limitation is that the study excluded adults with PWS. Given the scarcity of research on co-occurring ASD in adults with genetic neurodevelopmental syndromes, we opted to focus initially on children and youth with PWS. Recently, however, Sappok at al. [
56] evaluated ADOS-2 performance in 79 adults with moderate to severe intellectual disabilities, with or without clinical diagnoses of ASD. They suggest modifications of some test stimuli to reflect more adult interests and found that ADOS-2 overall calibrated severity scores performed best in differentiating ASD from others. Further, de Bilt et al. [
57] published revised severity algorithms for Module 4 of the ADOS that discriminated between high-functioning adults with ASD, schizophrenia, sociopathy, and controls. Collectively, this work bodes well for future studies that evaluate ADOS-2 performance in adults with PWS.
A final issue relates to the analyses of SCQ performance. Other researchers have used receiver operator curves to identify SCQ cut-points that provide optimal sensitivity or specificity for their study samples. We, however, used the recommended cut-point for this measure to enable comparisons to existing literature that used this same cut-point. We also reasoned that future PWS research would not necessarily benefit from a continued reliance on screeners, even with a revised PWS-derived cut-point.