Autism in neurofibromatosis type 1: prevalence, phenotype and clinical implications

Abstract

Background

Monogenic conditions, such as neurofibromatosis type 1 (NF1), provide a unique opportunity to study autism within a known molecular pathway. This study estimated the prevalence of autism in a large international NF1 cohort and examined how autistic traits relate to cognitive, behavioral, and adaptive functioning.

Methods

In this cross-sectional study, 202 children with NF1 and 106 typically developing controls (aged 3–15 years) completed standardized cognitive testing and parent-reported behavioral measures. NF1 participants scoring ≥ 60 on the Social Responsiveness Scale-2 were evaluated using the Autism Diagnostic Observation Schedule-2 and Autism Diagnostic Interview–Revised, with DSM-5-TR autism diagnoses established by multidisciplinary consensus. Neurodevelopmental profiles were compared across NF1 subgroups (autism, broader autistic traits, no-autistic-traits) and controls.

Results

Autistic traits were elevated and continuously distributed in NF1. An estimated 28% (95%CI: 21.7%-34.3%) met DSM-5-TR criteria for autism and 22% exhibited a broader autistic phenotype, with comparable rates in males and females. Children classified as autistic and those with broader autistic traits showed similarly marked behavioral and adaptive challenges, whereas NF1 children without autistic traits showed milder difficulties. Findings establish NF1 as a condition in which autism is common and clinically significant. Autistic features in NF1 occur across the cognitive spectrum and are accompanied by overlapping behavioral and functional impairments that challenge categorical diagnostic boundaries.

Limitations

The cross-sectional design precludes conclusions about developmental trajectories. The absence of genotype data prevented exploration of genotype-phenotype correlations and the influence of polygenic risk factors outside the NF1 locus.

Conclusions

Findings provide strong support for routine, early screening for autism in NF1 and the recognition that subthreshold traits can be similarly impairing. NF1 provides a powerful model for investigating how a monogenic condition contributes to diverse neurodevelopmental profiles and informs precision-medicine approaches in autism.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13229-026-00726-z.

Background

Autism spectrum disorder (ASD; henceforth autism) is a neurodevelopmental condition characterized by challenges in social communication and interaction, occurring alongside patterns of restricted interests and repetitive behaviors that impact daily life [1]. In the general population, autism is reported in approximately 1–3% of children [2–4]. It occurs across all racial, ethnic and socioeconomic groups, and is diagnosed more than three times more frequently in boys than girls [3]. Autism is characterized by considerable heterogeneity across adaptive functioning, cognition, language, and the presence of co-occurring conditions, leading some to refer to “the autisms” rather than a single, unified entity [5, 6]. This heterogeneity reflects the complex genetic underpinnings [7], involving both low-penetrance common variants with small additive effects, and rarer high-penetrance variants typically associated with more pronounced developmental differences [8]. These genetic factors influence diverse molecular and neurodevelopmental pathways, and likely contribute to the broad spectrum of autism presentations [9].

Monogenic syndromes that confer elevated risk for autism traits provide a valuable opportunity to study autism against a clearly defined molecular background. This ‘genetics first’ approach, which begins with a known pathogenic variant and examines the resulting behavioral and cognitive profile, offers a clearer framework for investigating developmental pathways and disentangling sources of phenotypic heterogeneity. This approach has yielded important insights into autistic phenotypes within monogenic syndromes such as Fragile X syndrome, tuberous sclerosis complex, and Rett syndrome [6, 10]. Neurofibromatosis type 1 (NF1) is a natural next step for applying a genetics-first methodology. With a prevalence of 1/2700, NF1 is one of the most common autosomal dominant genetic conditions affecting the central nervous system (CNS) [11]. The condition is characterized by a range of neoplastic, cutaneous, and skeletal manifestations, including plexiform neurofibromas which affect 30–50% of individuals [12], optic pathway gliomas (15–20%) [13]; and an elevated lifetime risk of malignant peripheral nerve sheath tumors (8–13%) [14]. However, the most common morbidities associated with NF1 are the cognitive [15, 16], behavioral [17, 18], and social challenges experienced by up to 80% of children [19–21]. These neurodevelopmental difficulties arise from the neurobiological consequences of neurofibromin loss [22], and appear to be independent of tumor burden and other physical manifestations of the condition [23, 24]. Recent NF1 studies report elevated autistic trait prevalence [25, 26], with screening-based assessments indicating clinical-level autistic symptoms in 13–29% of children, and subclinical symptoms in an additional 27–31% [25, 27, 28]. However, these estimates are based on highly sensitive tools such as the Social Responsiveness Scale-2 (SRS-2) [29], which excel at detecting autistic traits, but lack specificity in populations with overlapping neurodevelopmental phenotypes such as ADHD or anxiety, potentially inflating prevalence estimates [30].

In light of this limitation, a small number of studies have used clinic-based assessments, reporting prevalence estimates ranging from 11 to 27% in school-aged children with NF1 [31–33]. However, these studies are limited by methodological constraints including small sample sizes (n range, [31–47, 31, 32]), ascertainment bias toward children with suspected behavioral concerns [32], and reliance on non-standardized developmental histories. As a result, it remains unclear whether these estimates accurately reflect autism prevalence in NF1 populations unselected for neurodevelopmental concerns. Furthermore, the cognitive, behavioral, and adaptive phenotypes of autism in NF1 are poorly characterized, limiting insights into how autistic traits manifest and vary in this population.

To address these gaps, our aim was to examine the prevalence and expression of autism in a large, international cohort of children with NF1 unselected for neurodevelopmental concerns. Children screening at risk for autism underwent gold-standard diagnostic assessment [34], along with rigorous multidisciplinary consensus to establish robust DSM-5-TR autism diagnoses. Children with NF1 were categorized by autism status: diagnosed, broader autistic phenotype (subthreshold symptoms), or no evidence of autistic traits. This stratification enabled a comprehensive analysis of autism prevalence, expression, and functional impact in NF1, informing targeted clinical assessment and support for this population.

Methods

Participants

This study includes data from 202 children with NF1 and 106 typically developing (TD) controls between 3 and 15 years of age enrolled in a prospective cross-sectional study investigating the autistic phenotype of children with NF1. Children with NF1 were recruited from multidisciplinary NF1 Clinics at three major centers: (1) The Royal Children’s Hospital/Murdoch Children’s Research Institute, Melbourne, Australia; (2) The Children’s Hospital at Westmead, Sydney, Australia; and (3) Children’s National Hospital, Washington DC, USA. Children are referred to these clinics by general practitioners and medical specialists for diagnosis, surveillance, and management of NF1. Each site services a broad clinical population, providing care to individuals representative of the broader NF1 community. NF1 diagnoses were conducted by expert clinical geneticists or neurologists according to the current diagnostic criteria at time of enrolment [35]. Exclusion criteria for all participants were: (1) insufficient English to complete the assessments; (2) significant visual or hearing impairment that would compromise the validity of psychometric testing; and (3) symptomatic CNS pathology that would confound neurobehavioral testing. Children with asymptomatic lesions, such as stable optic pathway gliomas were eligible. Additional exclusion criteria for TD controls included a positive history of any neurological or genetic condition, and developmental delay or intellectual disability. A full description of the recruitment procedure and rationale for study size determination are detailed in our published protocol [36].

Procedure

Participants underwent standardized cognitive evaluations with a site psychologist. Caregivers completed questionnaires assessing a range of behavioral and functional outcomes, including the SRS-2. NF1 participants who screened at risk for autism on the SRS-2 (Total T-score ≥ 60) were invited to complete a detailed autism evaluation comprising the Autism Diagnostic Observation Schedule–Second Edition (ADOS-2) and the Autism Diagnostic Interview–Revised (ADI-R), administered by trained psychologists with research reliability. Following these assessments, case files were independently reviewed by a multidisciplinary expert panel consisting of a child and adolescent psychiatrist (TM), clinical psychologist (AU), and speech pathologist (AB), each with training and expertise in autism assessment. All ADOS-2 examiners were trained to the level of research reliability. To verify ongoing reliability, approximately 20% of ADOS-2 administrations were cross-coded across sites, with examiners from one site independently coding video recordings from another site prior to case presentation to the panel. Mean percentage agreement was > 70% across all coded items and algorithm items. Multidisciplinary panel reviews drew principally on direct observation of the ADOS-2 video recordings and ADI-R findings, with diagnostic formulations further informed by relevant clinical, developmental, and educational histories. SRS-2 total T-scores were available to panel members but were not the primary basis for diagnostic decision-making. Final DSM-5-TR autism diagnoses were determined through multidisciplinary consensus. Participants with NF1 were classified into one of three groups: (1) diagnosed autism, (2) broad autistic phenotype (SRS-2 T-score ≥ 60 indicating elevated autistic traits but not meeting full DSM-5-TR criteria for autism as determined by the expert panel), or (3) no evidence of autistic traits (SRS-2 T-score ≤ 59). In cases of diagnostic uncertainty, consensus was reached through discussion until agreement was achieved. This approach is consistent with international guideline recommendations for autism diagnosis, which emphasize the use of standardized observational and caregiver-report instruments alongside multidisciplinary clinical consensus [37, 38].

Measures

Where available, sex-specific norms were used to generate standardized scores across all rating scales. Given the age range of the cohort (3–15 years), developmentally appropriate instruments were used for intellectual ability, executive behavior, and ADHD symptoms, with scores pooled across younger and older child versions. Pooled measures assess identical constructs on the same normative metric with comparable psychometric properties, supporting their combined use in analysis. Table S1 contains descriptions of the measures used and the scores they yield, available online.

Autism assessments

Autistic traits were assessed with the parent version of the SRS-2, a questionnaire that quantifies the severity of social communication difficulties and restricted interests/repetitive behaviors [29]. The autism evaluation consisted of a standardized observational assessment (ADOS-2) [39], and a semi-structured parent interview (ADI-R) [40].

Behavioral and functional outcomes

ADHD symptoms were assessed using the Conners-3 for 6–15-year-olds [41], and the Conners ADHD DSM-IV Rating Scale (CADS) for 3–5-year-olds [42]. The age-appropriate version of the Child Behavior Checklist (CBCL) was administered to assess internalizing problems [43, 44]. We report the DSM-oriented anxiety and affective problems scales which are ‘purer’ measures of anxiety and depression in our sample, since the CBCL internalizing composite may be confounded by the somatic manifestations associated with NF1 [11]. Executive behaviors were assessed with the Global Executive Composite score of the parent-rated Behavior Rating Inventory of Executive Function (BRIEF) for 6–15 year olds and the preschool version for those aged 3–5 years (BRIEF-P) [45, 46]. Prosocial behavior was assessed with the parent-rated Social Skills Improvement System-Rating Scale (SSIS-RS) [47]. The parent version of the Adaptive Behavior Assessment System-Third Edition (ABAS-3) was used to assess adaptive behavior across conceptual (communication and academics), social, and practical (daily living skills) domains, yielding an overall General Adaptive Composite [48].

Intelligence and academic achievement

Children aged 6–15 years completed the Wechsler Intelligence Scale for Children–Fifth Edition (WISC-V) [49], while those aged 3–5 years completed the Wechsler Preschool and Primary Scale for Children-Fourth Edition. (WPPSI-IV) [50]. All subtests required for calculating the primary indexes were administered. Academic achievement was measured using the Word Reading, Spelling and Numerical Operations subtests from the Wechsler Individual Achievement Test–Second Edition (WIAT-II) [51].

Socioeconomic risk

Children’s socioeconomic risk was stratified by maternal education level, with <12 years of schooling classified as high risk and ≥12 years classified as low risk. Maternal education was selected as the stratification variable given its strong association with child behavioral outcomes and its role as a proxy for broader socioeconomic influences [52].

Statistical analysis

Analyses were conducted in JASP version 0.95.4 [53] and Stata v19 [54]. Group differences in demographic and clinical characteristics between NF1 and TD controls were examined using chi-square tests for categorical data and independent samples t-tests for continuous measures. Assumptions for parametric analyses were checked through visual inspection of distributions, Shapiro-Wilk tests, and Levene’s tests. Data were sufficiently normal with homogeneous variances to justify parametric analyses. Autism prevalence was estimated from the evaluable NF1 sample (n = 192) using binomial 95% confidence intervals and relative risk was calculated in comparison to a weighted population prevalence estimate accounting for the geographic composition of the cohort (190 Australian, 12 US participants), derived from published Australian (1.3%) [2] and US (2.8%) [3] population prevalence figures.

To examine neurodevelopmental differences across diagnostic subgroups, NF1 participants were classified into three subgroups (autism, broader autistic phenotype, no autistic traits) and compared with TD controls. Oneway ANOVAs were used to test differences in intellectual, academic, behavioral, and adaptive outcomes, with planned post hoc comparisons to identify the source of significant main effects. Effect sizes for ANOVA main effects were expressed as partial η² and interpreted using conventional benchmarks (0.01 small, 0.06 medium, 0.14 large). Pairwise post hoc comparisons were quantified with Cohen’s d (0.20 small, 0.50 medium, 0.80 large). Type I error was controlled at the individual test level through a conservative α threshold of 0.01, two tailed testing for all analyses, and Tukey-Honestly Significant Difference (HSD) post hoc corrections within each ANOVA. No family-wise error correction was applied across neurodevelopmental outcome variables, as the domains examined share substantial variance, rendering global corrections over corrective in this context. To examine whether observed neurodevelopmental differences between groups were influenced by socioeconomic risk, all group comparisons were repeated as ANCOVAs with socioeconomic risk entered as a covariate, with pairwise comparisons based on estimated marginal means corrected using Tukey’s HSD. The funders of this study had no role in study design, data collection, data analysis, data interpretation, or the writing of this report.

Results

Sample characteristics

Demographic information of the study population is shown in Table 1. The study cohort consisted of 202 children and adolescents with NF1 and 106 TD controls who were comparable in age and sex distribution. A greater proportion of children with NF1 were classified as high socioeconomic risk compared to TD controls, χ²(2, N = 308)=19.72, p < 0.001). Across clinical measures, children with NF1 demonstrated lower intellectual and academic functioning, higher levels of ADHD symptoms, increased mood and anxiety symptoms, greater executive dysfunction, and reduced adaptive functioning relative to TD controls (all, p < 0.001).

Table 1.

Demographic and neurodevelopmental characteristics of the NF1 and typically developing (TD) control groups

Variable	NF1	TD Controls	Cohen’s d
N	202	106	-
Age, mean (SD)	8.52 (3.50)	8.51 (3.31)	0.004
Sex, n (%)
Male	109 (53.96%)	52 (49.06%)	-
Female	93 (46.06%)	54 (50.94%)	-
NF1 inheritance, n (%)
De novo	122 (60.40%)	-	-
Familial	79 (39.11%)	-	-
Unknown	1 (0.50%)	-	-
Socioeconomic Risk, n (%)
Low	161 (79.71%)	104 (98.11%)	-
High	30 (14.85%)	1 (0.94%)	-
Unknown	11 (5.45%)	1 (0.94%)	-
SRS-2 Total T-scorea	62.15 (14.29)	46.59 (5.75)	1.28**
Wechsler Intellectual Functioning, mean (SD)b
Full Scale IQ	88.89 (12.88)	106.98 (13.34)	−1.39**
Verbal Comprehension Index	93.84 (14.58)	110.02 (13.29)	−1.14**
Visual Spatial Index	88.49 (12.45)	103.22 (13.09)	−1.62**
Fluid Reasoning Index	90.76 (13.60)	106.81 (13.86)	−1.17**
WIAT-II Academic Achievement, mean (SD)b
Word Reading	90.19 (17.16)	106.99 (15.51)	−1.01**
Spelling	87.16 (15.64)	104.31 (15.58)	−1.01**
Numerical Operations	86.32 (17.75)	101.99 (16.32)	−0.91**
Conners ADHD, mean (SD)a
Inattentive symptoms	66.06 (15.62)	51.96 (11.13)	0.99**
Hyperactive/Impulsive symptoms	64.99 (16.77)	51.46 (12.51)	0.88**
CBCL DSM-5 Subscales, mean (SD)a
Depressive symptoms	61.22 (9.67)	53.69 (5.27)	0.90**
Anxiety symptoms	58.60 (9.73)	53.25 (4.89)	0.64**
BRIEF Global Executive Composite, mean (SD)a	63.04 (16.34)	47.73 (13.34)	0.93**
ABAS-3 Global Adaptive Composite, mean (SD)b	87.36 (13.68)	102.04 (11.29)	−1.14**

Note: ABAS-3 = Adaptive Behavior Assessment System – Third Edition; ADHD = Attention deficit/hyperactivity disorder; BRIEF = Behavior Rating Inventory of Executive Function; CBCL = Child Behavior Checklist; DSM = Diagnostic and Statistical Manual; NF1 = Neurofibromatosis type 1; WIAT-II = Wechsler Individual Achievement Test – Second Edition

^aT score, mean = 50, SD = 10; higher scores indicate elevated difficulties

^bStandard score, mean = 100, SD = 15; higher scores indicate better performance

**p < 0.001

Autistic traits

Autistic traits were significantly higher in the NF1 group (p < 0.001; Table 1). Distribution of SRS-2 Total T scores in the NF1 cohort was continuous, unimodal, and elevated by d = 1.28 (95%CI, 1.03–1.54) relative to the TD control group (Fig. 1). In total, 51.5% of children with NF1 exhibited clinically significant autistic traits (SRS-2 T-score ≥ 60), all of whom were invited to complete the clinical diagnostic assessment, and 19.3% were rated in the severe range (SRS-2 T-score ≥ 76). No significant sex differences were observed (mean difference = −2.46; 95%CI, −6.49–1.56), and autistic trait scores were comparable between preschool- (3–5 years) and school-aged children with NF1 (mean difference = 3.19; 95%CI, −0.52–0.07; Figure S1 available online). Children with NF1 from higher socioeconomic risk backgrounds exhibited significantly elevated SRS-2 Total T-scores compared to those from lower socioeconomic risk backgrounds (mean difference = 7.60; 95%CI, 2.18–13.02; Figure S2, available online).

Fig. 1.

SRS-2 T-score distribution. Darker blue bars indicate standardized scores within the typical range, while lighter blue represent total T-scores ≥ 60, which are associated with clinically significant autistic traits. Dotted lines mark the general population mean T-score (50) and the NF1 sample mean T-score (62.15)

Autism prevalence

Of the 107 NF1 participants with elevated SRS-2 data (T-score ≥ 60), 10 were not evaluable by the diagnostic panel due to the reasons outlined in Fig. 2. This left 192 participants in the evaluable sample for estimating autism prevalence. Within this sample:

• 54 (28.1%) met DSM-5-TR criteria for autism (95%CI 21.4%-34.4%);

• 43 (22.4%) exhibited elevated autistic traits but did not meet criteria for autism and were classified as having a broader autistic phenotype;

• 95 (49.5%) had SRS-2 Total T-scores < 60 and were classified as not having autistic traits.

Fig. 2.

Flow of participants through the NF1 autism diagnostic evaluation

Compared to a weighted population prevalence estimate of 1.39% that accounts for the geographic composition of the cohort in which 190 were Australian and 12 from the US, based on published Australian [2] and US [3] population prevalence figures, children with NF1 were 20.1 times more likely to be diagnosed with autism (95%CI, 16.1%-25.2%). Notably, autism was diagnosed in 30/104 males (28.9%) and 24/88 females (27.3%), indicating that males and females with NF1 were equally likely to meet diagnostic criteria, χ² (2, N = 192) = 0.06, p = 0.81.

There was a trend between socioeconomic risk and autism outcome for children with NF1, χ² (2, N = 181)=6.28, p = 0.04, which did not meet the prespecified significance threshold (Table S2, available online). Children from higher-risk families (mother not completing high school) were more likely to meet DSM-5-TR criteria for autism (42.9%) than those from lower risk backgrounds (24.8%), with a small-to-moderate effect size (Cramér’s V = 0.19). Rates of autism diagnosis (26.8% for de novo and 29% for familial), and trait severity were comparable across NF1 inheritance types (Table S3, available online). Exploratory analysis showed no significant differences in rates of plexiform neurofibromas, optic pathway gliomas, chemotherapy exposure, or seizure history between children with and without an autism diagnosis (all p > 0.11; Table S4, available online).

Neurodevelopmental correlates

To examine neurodevelopmental functioning across diagnostic and classification-based subgroups, we compared TD controls to NF1 participants with autism, a broader autistic phenotype, or classified as having no autistic traits, on key cognitive, behavioral, and adaptive outcomes. Neurodevelopmental outcomes across all four subgroups are displayed in Figs. 3 and 4, with full descriptive statistics and effect sizes provided in Table 2. Post hoc comparisons are provided in Tables S5–S17, and the adjusted summary statistics, covarying for socioeconomic risk are provided in Table S18, available online).

Fig. 3.

Cognitive and academic performance across NF1 subgroups and typically developing controls. Data are presented as means with 95% confidence intervals

Fig. 4.

Behavioral and adaptive functioning across NF1 subgroups and typically developing controls. Data are presented as means with 95% confidence intervals

Table 2.

Summary statistics across the three NF1 subgroups and typically developing controls across all neurodevelopmental outcomes

Outcome	NF1 Autism n = 54	NF1 BAP n = 43	NF1 No Autism n = 95	TD Control n = 106	Partial η2	Group Differences
	M (SD)	M (SD)	M (SD)	M (SD)
Verbal Comprehension Indexa	88.5 (16.0)	89.6 (12.7)	98.4 (12.7)	110.0 (13.3)	0.297	NF1 Autism (p < 0.001), NF1 BAP (p=0.003) <NF1 No Autism; ns among NF1 Autism, NF1 BAP; All NF1 < TDC (p < 0.001)
Visual Spatial Indexa	88.4 (14.4)	87.1 (13.9)	89.9 (10.8)	103.2 (13.1)	0.230	All NF1 < TDC (p < 0.001); ns among NF1 groups
Fluid Reasoning Indexa	89.2 (13.9)	88.3 (13.5)	93.8 (13.24)	106.8 (13.9)	0.246	All NF1 < TDC (p < 0.001); ns among NF1 groups
Readinga	90.5 (17.6)	89.3 (16.5)	90.6 (16.9)	107.0 (15.5)	0.199	All NF1 < TDC (p < 0.001); ns among NF1 groups
Spellinga	87.6 (18.6)	87.03 (15.0)	86.8 (14.6)	104.3 (15.6)	0.222	All NF1 < TDC (p < 0.001); ns among NF1 groups
Matha	86.9 (19.2)	81.3 (18.3)	88.2 (16.4)	102.0 (16.3)	0.181	All NF1 < TDC (p < 0.001); ns among NF1 groups
Conners ADHD – Inattentionb	75.4 (12.2)	74.2 (12.5)	56.6 (13.2)	52.0 (11.1)	0.401	NF1 Autism, NF1 BAP > NF1 No Autism, TDC (p < 0.001); ns among NF1 No Autism, TDC
Conners ADHD – Hyperactivity/Impulsivityb	74.8 (14.7)	73.4 (14.1)	55.4 (13.6)	51.5 (12.5)	0.354	NF1 Autism, NF1 BAP > NF1 No Autism, TDC (p < 0.001); ns among NF1 autism, NF1 BAP; ns among NF1 No Autism, TDC
Executive Behaviorb	73.7 (12.4)	71.9 (10.8)	52.4 (14.1)	48.7 (13.3)	0.401	NF1 Autism, NF1 BAP > NF1 No Autism, TDC (p < 0.001); ns among NF1 autism NF1 BAP; ns among NF1 No Autism, TDC
CBCL Anxiety Symptomsb	64.5 (10.5)	61.31 (9.8)	53.7 (5.9)	53.3 (4.9)	0.289	NF1 Autism, NF1 BAP > NF1 No Autism, TDC (p < 0.001); ns among NF1 autism NF1 BAP; ns among NF1 No Autism, TDC
CBCL Affective Symptomsb	65.9 (9.9)	66.9 (7.0)	55.7 (7.4)	53.7 (5.3)	0.369	NF1 Autism, NF1 BAP > NF1 No Autism, TDC (p < 0.001); ns among NF1 autism, NF1 BAP; ns among NF1 No Autism, TDC
SSIS-RS Prosocial Skillsa	78.7 (13.9)	84.0 (12.4)	99.9 (15.2)	106.0 (13.8)	0.376	NF1 Autism, NF1 BAP > NF1 No Autism, TDC (p < 0.001); ns among NF1 autism, NF1 BAP; ns among NF1 No Autism, TDC
ABAS-3 Adaptive Functioninga	78.6 (12.9)	84.0 (12.1)	94.6 (10.9)	102.0 (11.3)	0.377	NF1 Autism, NF1 BAP < NF1 No Autism, TDC (p < 0.001); NF1 No Autism < TDC; ns among NF1 autism, NF1 BAP

Note: BAP = Broader Autistic Phenotype; ns = non-significant; M = Mean; NF1 = Neurofibromatosis type 1; SD = Standard Deviation; TDC = Typically Developing Control

^a Standard Score

^b T-Score

Intelligence and academic achievement

Children with NF1 and autism, as well as those with a broader autistic phenotype, demonstrated significantly lower verbal IQ scores than NF1 participants with no autistic traits (both, p ≤ 0.003; Fig. 3). In contrast, nonverbal IQ did not differ significantly across NF1 subgroups, with equivalent performances on visual-spatial (all, p > 0.65) and fluid reasoning indexes (all, p > 0.15). All NF1 subgroups exhibited lower verbal and nonverbal IQ scores than TD controls (all, p < 0.001).

Across measures of reading, spelling, and mathematics, there were no significant differences between the three NF1 subgroups, indicating comparable academic performance regardless of autistic traits (all, p > 0.28; Fig. 3). However, all three NF1 subgroups demonstrated significantly lower academic scores than TD controls (all, p < 0.001).

Behavioral and adaptive functioning

Children with NF1 and autism or a broader autistic phenotype exhibited comparably elevated difficulties across all behavioral and functional domains (all, p ≥ 0.13 Fig. 4). Both subgroups demonstrated elevated inattentive and hyperactive/impulsive ADHD symptoms, reduced prosocial skills, greater dysexecutive behavior, elevated anxiety and depression, and lower adaptive functioning; all of which were significantly different than NF1 children with no autistic traits and TD controls (all, p < 0.001).

NF1 participants with no autistic traits scored lower than TD controls on adaptive functioning (p < 0.001), demonstrated slightly reduced prosocial skills (p = 0.021), and elevated inattentive ADHD symptoms (p = 0.045), but were otherwise comparable across behavioral outcomes (all, p > 0.17).

Table S18 presents adjusted estimates from ANCOVAs controlling for socioeconomic risk. Estimated marginal means were highly consistent to unadjusted means across all neurodevelopmental outcomes, with group difference patterns and effect sizes remaining unchanged, indicating that the observed neurodevelopmental differences between groups were not meaningfully influenced by socioeconomic risk.

Discussion

This study provides the most comprehensive characterization to date of autistic traits, cognitive functioning, and behavioral outcomes in children with NF1, raising important conceptual questions about the boundaries of autism with direct implications for diagnosis and clinical management. First, we found that 28% of children with NF1 met DSM-5-TR criteria for autism, a prevalence over 20 times higher than the general population [2]. This finding of elevated autistic traits places NF1 alongside other well-characterized neurogenetic conditions, such as Fragile X syndrome and tuberous sclerosis complex [55], as a genetic locus associated with elevated autistic traits and increased rates of autism. NF1 thus offers a powerful model for investigating the full continuum of autistic features, from subthreshold traits to diagnosable autism, within the context of a defined, genetic etiology. This creates opportunities to examine how NF1-related molecular mechanisms affect brain development and give rise to the autistic phenotype observed in this study.

Second, unlike idiopathic autism, where males are disproportionately affected [3], our NF1 cohort showed no sex difference in autism prevalence or trait severity. While population-level mechanisms such as male-linked neurobiological susceptibilities and female protective effects likely remain operative [55], the strong influence of NF1 appears sufficient to attenuate or override these sex-based modifiers. Evidence from both Nf1^± animal and human models indicates that NF1 expression plays a critical role from very early developmental stages, and continues to influence key neurobiological processes, resulting in morphological brain differences that emerge similarly in males and females [56, 57]. Clinically, this suggests that autistic features in girls with NF1 may be more readily expressed and detectable than in idiopathic autism, potentially reflecting greater biological penetrance. Mechanistically, NF1 appears to act as a potent single-gene driver capable of neutralizing sex-based differences in autism liability. [58] As such, NF1 provides a unique model for understanding how monogenic risk can interact with, and in this case eclipse, population-level modifiers that affect autism expression. The prospective consecutive recruitment design of this study, in which all children were systematically assessed regardless of clinical concern, may also have contributed by reducing the diagnostic under-identification of autistic girls that can occur in population samples [59]. Taken together, both biological and methodological factors converge to support the finding that autism prevalence is equal across sexes in NF1, with important implications for clinical practice and screening.

Third, we found little evidence that cognitive or academic abilities distinguished children with and without autism in NF1. Consistent with prior literature [31], autistic features emerged independently of cognitive functioning, contrasting with idiopathic autism, where lower IQ is often associated with greater symptom severity and poorer outcomes [60]. This pattern suggests that NF1 gene dysfunction affects neural systems critical for social and behavioral regulation, with cognitive impairments arising from overlapping but partly distinct mechanisms.

Fourth, our data reinforce the notion that autistic traits in NF1 occur along a spectrum, rather than as a dichotomous phenomenon. NF1 SRS-2 scores formed a unimodal distribution shifted to the right, with over half (51.5%) exhibiting clinically significant autistic traits, replicating prior findings from a pooled cohort of 530 individuals [25]. Importantly, we extend those findings by showing that children who met DSM-5-TR criteria for autism, and those with a broader autistic phenotype, displayed nearly identical clinical profiles across ADHD, anxiety, depression, executive dysfunction, prosocial skills, and adaptive functioning. These results suggest that behavioral, social and functional challenges are pervasive across the autism spectrum in NF1 and are not confined to those meeting diagnostic thresholds.

There is a growing body of evidence supporting the view of autism as a pluralized construct known as ‘the autisms,’ which emphasizes heterogeneity in both etiology and phenomenology [5, 6, 61]. Within idiopathic autism, the boundaries between diagnosable autism and subthreshold autistic traits have been shown to be functionally meaningful in some contexts but not others, raising ongoing debate about the utility of categorical thresholds [62]. Our NF1 findings contribute to this debate in a unique way. Because NF1 represents a genetically homogeneous population with a single defined etiology, it offers a uniquely controlled context in which to examine the functional significance of categorical diagnostic boundaries that is free from the etiological complexity that characterizes idiopathic autism. The fact that, even under these controlled genetic conditions, children above and below the formal diagnostic cutoff were functionally indistinguishable suggests that the boundary between ‘autism’ and ‘subthreshold autistic traits’ may not capture meaningful differences in clinical caseness. If categorical cutoffs fail to map onto functional reality even within a genetically defined, etiologically homogeneous condition, this raises important questions about what those boundaries are capturing. Do they reflect true discontinuities in the underlying neurodevelopmental processes, or might they represent diagnostic conventions that may obscure a more continuous phenotypic reality? These questions become all the more relevant in idiopathic autism, where etiological heterogeneity is vastly greater and the sources of phenotypic variability correspondingly more complex. Taken together, these data support the view that the diagnostic framework for autism, and particularly the threshold between diagnosable autism and subthreshold traits, should be considered as a flexible guidebook that accommodates the complexity of neurodevelopmental presentations rather than a rigid rulebook.

A major challenge for the field is to delineate the neurobiological mechanisms that converge across monogenic and polygenic forms of autism. Studying these convergence points is critical for understanding shared brain circuitry and molecular mechanisms that transcend traditional diagnostic boundaries. Monogenic conditions such as NF1 offer a genetically defined entry point into autism-relevant biology, serving as human models for dissecting pathways that may be obscured in idiopathic or polygenic form [57, 63]. Importantly, NF1 differs from many other syndromic conditions in that its neurodevelopmental phenotype is typically not associated with profound impairments. This positions NF1 as a particularly valuable genetic model for studying the mechanisms underlying common and variably expressed neurodevelopmental traits in the general population. If shared pathways, such as disruptions in synaptic signaling, cortical development, or intracellular signaling cascades can be mapped across these genetic contexts, they may provide a scaffold for interpreting polygenic risk [64]. Realizing this potential will require advanced human-centric models, such as patient-derived regionalized 3D cerebral organoid models that have the potential to capture the complexity and variability of human neurodevelopment [19].

These findings have several important clinical implications. Given the high prevalence of autism and broader autistic traits in children with NF1, clinicians should routinely evaluate social-communication and behavioral function. “Wait and see” approaches risk missing children with stronger cognitive and learning abilities but significant subthreshold symptoms who nonetheless experience substantial functional challenges. These findings reinforce the principle that intervention should be guided by functional impact rather than rigid diagnostic thresholds, recognizing that subthreshold traits confer substantial developmental risk and impairment and are equally likely to benefit from targeted supports. By focusing on functional needs across key developmental domains, clinicians can better tailor interventions and reduce missed opportunities for scaffolding. These interventions may include social skills support and behavioral strategies, which have been shown to be effective across a range of neurodevelopmental and psychiatric conditions [65], alongside environmental accommodations that reduce barriers to participation and support daily functioning. Furthermore, our data suggest a trend between higher socioeconomic risk and autism diagnosis in NF1. While not reaching statistical significance, the observed small-to-moderate effect size is consistent with broader evidence linking socioeconomic disadvantage to elevated neurodevelopmental vulnerability. Because participants in our study were assessed consecutively through a standardized protocol (regardless of socioeconomic background), ascertainment bias and differential access to diagnostic services are an unlikely explanation for this finding [66]. The consistency of neurodevelopmental outcome differences across groups before and after adjusting for socioeconomic risk (Tables S5 and S6), however, suggests that the functional profiles characterizing each subgroup are primarily driven by the neurobiological consequences of NF1 and autistic trait burden rather than socioeconomic disadvantage. Whether socioeconomic disadvantage acts as a meaningful modifier of autism expression in NF1 warrant investigation in larger, prospectively designed studies, and if confirmed, would have important implications for clinical surveillance and support for higher-risk families.

Limitations

This study has several limitations. Diagnostic autism assessments (ADOS-2 and ADI-R) were only administered to NF1 participants who screened above threshold on the SRS-2 (T-score ≥ 60). While this likely captured all autism cases given the high sensitivity of the SRS-2 [29], it remains possible that a small number of children were missed, leading to a slight underestimation of autism prevalence in NF1. The absence of genotype data prevented exploration of genotype-phenotype correlations, as well as the influence of polygenic risk factors outside the NF1 locus. Given that NF1 microdeletions are associated with more severe phenotypes, and that common variants outside of NF1 may also modulate risk, larger genetically stratified studies will be crucial to advancing our understanding of the molecular and neurobiological pathways underlying the autistic phenotype in NF1. Although the cohort was ascertained from specialist clinics, which may limit generalizability, the prevalence estimates observed here align with a prior smaller study that used a population-based recruitment strategy [31], suggesting our findings are likely to extend to broader NF1 populations. Finally, the cross-sectional nature of the study precludes conclusions about developmental trajectories. Longitudinal research is needed to understand how autistic traits and associated cognitive-behavioral profiles evolve over time in NF1, and to identify early predictors of later outcomes.

Conclusion

In summary, this study outlines a genetics-first approach to autism by highlighting NF1 as a syndrome in which autism and related traits are both common and clinically significant. We found that autism occurs at approximately 20 times the population rate, affects males and females equally, and emerges independently of cognitive ability. Children with NF1 meeting DSM-5-TR criteria for autism and those with subthreshold autistic traits were functionally indistinguishable across behavioral, social, and adaptive domains, while children with no autistic traits performed comparably to typically developing controls. This pattern supports the view that diagnostic thresholds between autism and the broader autistic phenotype in NF1 should be understood as flexible clinical guides rather than rigid categorical boundaries. These findings emphasise the importance of routine screening for autism and autistic traits in NF1, and of intervention approaches guided by functional need rather than diagnostic status. NF1 offers a powerful model for investigating the neurobiological mechanisms underlying autistic phenotypes, with insights that may extend beyond the condition to inform our understanding of autism more broadly.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

JMP, KSW, KNN, NAP, conceived and designed the study. JMP, KMH, AKC, KSW, FL, AM, HD, YR, SR, RS, NLH, YFE, NAP contributed to data collection. AU, TM, AB served on the multidisciplinary panel responsible for establishing DSM-5-TR diagnoses. JMP, KMH, SJCH performed the statistical analyses, had full access to all the data in the study and verified the data. All authors had access to aggregated results. JMP wrote the first draft of the manuscript. Study management was conducted by JMP. All authors were involved in the writing of the final manuscript and revised the manuscript. All authors were involved in the writing of the final manuscript and revised the manuscript. All authors had final responsibility for the decision to submit this work for publication.

Funding

Funding for this study was provided by an Investigator Initiated Research Award from the US Department of Defense’s Neurofibromatosis Research Program (W81XWH-15-1-0619) and a Murdoch Children’s Research Institute Clinician Scientist Fellowship (JMP).

Data availability

Data collected for the study, including de-identified individual participant data and a data dictionary defining each field, are available from the corresponding author on reasonable request for ethically approved research.

Declarations

Ethical approval and consent to participate

This research was conducted in accordance with the Declaration of Helsinki and approved on 16 September 2015 by the Human Research Ethics Committees of the Royal Children’s Hospital (HREC/16/RCHM/137) and the Sydney Children’s Hospitals Network (HREC/16/SCHN/42), and on 6 April 2016 by the Institutional Review Board of the Children’s National Hospital (Pro00007045). Written informed consent was appropriately obtained from the parents or caregivers of all participants.

Consent for publications

Not applicable.

Competing interests

DC has received honoraria from Medice, Novartis, Servier, and Takeda; royalties from Oxford University Press. All other authors declare that they have no competing interests.