We included four groups of children and young adults (aged 0–28 years) with syndromes that have high ASD prevalence: FXS, AS, NF1, and TSC, see Figure 1. ASD symptoms and cognitive functioning were assessed as part of routine clinical care that is performed in all children seen at the ENCORE expertise center for genetic neurocognitive developmental disorders within the Erasmus MC Sophia Children's Hospital in Rotterdam, the Netherlands. All children with these syndromes were included, regardless of their ASD symptomatology.

In addition, we included an nsASD group as a frame of reference. For this group, we used data collected as part of the Social Spectrum Study, a clinical cohort study that aimed to identify children at risk for ASD by screening children who had been referred to six large child and adolescent mental health services in the South-West of the Netherlands. A detailed description of the sample characteristics and data collection procedure of this group can be found in the study design paper of the Social Spectrum Study (70). We included children that participated in the in-depth assessment (T1), which consisted of all children with an ASD classification on the Social Responsiveness Scale (SRS) (71) at initial screening (T0) (N = 235) and a random selection of children that screened negative at T0 (N = 100). From this group, we included the children with an ASD classification according to the Autism Diagnostic Observation Schedule (ADOS) (72) and/or the SRS in our subgroup analyses. None of these children were diagnosed with a genetic syndrome.

For the syndrome groups we included individuals with a complete ADOS and/or SRS assessment between May 2009 to March 2021. In case complete data from multiple time points was available for an individual, we selected the most recent time point. This resulted in 507 cases for the ADOS and 375 cases for the SRS. The nsASD group data consisted of 134 ADOS assessments and 235 SRS questionnaires, collected between May 2011 and December 2013.

As each instrument has a unique focus, separate analyses were performed for the ADOS and the SRS data. First, we compared the syndrome groups on the prevalence of ADOS and SRS ASD classifications using Chi square tests. Next, we compared the main ASD severity scores (ADOS: CSS_TOT, CSS_SA, and CSS_RRB; SRS: T_TOT, T_SCI, T_RRB) between the syndrome groups using MANCOVA's with developmental level (IQ/DQ) and gender as covariates. We first did this for the syndrome groups as a whole. Then, we compared the syndrome groups again while only including individuals with an ASD classification on the ADOS or SRS, respectively, in order to include all individuals with a clinical score on that particular instrument irrespective of their classification status on the other instrument, with the nsASD group added as reference. We did this to reduce the expected within-group variability in ASD severity. Finally, for our ASD profile analysis, we also compared the ASD subgroups on subscales of the ADOS and SRS.

Data analyses were performed in IBM SPSS Statistics version 25 (86). We use an alpha level of 0.05 for all analyses and applied a Bonferroni correction for multiple comparisons in our post-hoc comparisons. In all tables the uncorrected-mean-scores, standard deviations, F-statistics, Bonferroni-corrected-p-values and the effect sizes (partial η²) are provided. Without a suitable alternative, MANCOVA's were used despite the skewed data (see Supplementary Figure 1) and correlation between the covariates and the predictor, so our results should be interpreted cautiously.

A non-response analysis for missing IQ/DQ scores in the syndrome groups revealed no differences in ADOS or SRS ASD severity scores between individuals with and without available IQ/DQ scores. For the nsASD group however, severity scores were higher for those without an available IQ/DQ score than for those with an IQ/DQ score for the ADOS [t₍₂₆₆₎ = −2.087, p = 0.038]. Because we expected the effect of IQ/DQ to be substantial in the syndrome groups we included IQ as a covariate nonetheless.

Sample characteristics are presented in Table 1. We found significant differences in ASD classification prevalence between the syndrome groups for both the ADOS and the SRS. According to both instruments, approximately a quarter of individuals in the NF1 group received an ASD classification, which was lower compared to the FXS, TSC, and AS groups in which at least half of individuals received an ASD classification. The ASD classification prevalence was highest in the FXS and AS groups. An overview of ASD classifications in the syndrome groups is provided in Supplementary Figure 2.

In line with the literature, for both instruments, the ASD classification prevalence was highest in the FXS group and lowest in the NF1 group. The prevalence of ADOS and SRS ASD classifications we found for the FXS group (around 80–90%), and the AS group (70–90%) fell on the high end of the range found in the literature. Overall, ASD classification prevalence seemed to be higher for the SRS compared to the ADOS. This difference was especially high for the TSC (26% difference) and AS groups (29% difference), although we did not test this statistically as the samples contained different individuals. The difference in ASD classification prevalence between the ADOS and SRS is likely the result of the fundamental differences between the instruments. First, as a screening instrument the SRS is required to be highly sensitive by design. Secondly, the caregiver's perspective provided by the SRS may be more subjective. Thirdly, the setting and time period that is covered by these instruments may affect the scores, as some children may have learned to present more socially desirable behavior in unfamiliar surroundings than they would do in the comfort of their own home. Alternatively, they may experience more anxiety in the presence of an unfamiliar individual.

An important factor that may explain the higher prevalence of ADOS and SRS ASD classifications in children with FXS and AS is their developmental level. Generally, individuals with FXS and AS experience more severe intellectual disability or developmental delay compared to NF1. Cognitive impairment is known to affect scores on ASD screening instruments (75), and has been related to reduced specificity of the ADOS, especially for young children (73). Cognitive impairment additionally impacts whether individuals qualify for a DSM-5 classification, which impacts their access to health care services (33). The DSM-5 requires that difficulties in social communication must be lower than could be expected based on the developmental level of the individual, so that the deficiencies in social skills cannot be attributed to developmental delay. Therefore, it is likely that difficulties in social communication in individuals with FXS and AS are more frequently attributed to developmental delay rather than to the presence of ASD. This means that children with intellectual disability or developmental delay can score in the clinical range on screening or diagnostic instruments but not qualify for a formal DSM-5 ASD classification, even though they may experience the same symptom severity. At the same time, children with intellectual disability or developmental delay who qualify for a formal DSM-5 ASD classification are likely to show more severe symptoms compared to children with a formal DSM-5 ASD classification without intellectual disability or developmental delay. Even though we controlled for IQ/DQ in our analyses, it remains unclear whether the symptoms we see in our sample are beyond their expected developmental level.

Given the wide variety of symptom profiles that exist, in both syndromic and non-syndromic populations, some researchers have raised the question whether an all-or-nothing ASD classification should be replaced by a multimodal symptom-based assessment of health care needs (68). Based on our findings, we would argue that, at least in children with a syndromic form of ASD, a symptom-based approach might be favorable. This would ensure that individuals who experience ASD symptoms are eligible for treatment of these symptoms regardless of their diagnostic status. However, further research is required to better understand the effect of developmental delay on social skills measured by ASD screening instruments. While an all-or-nothing diagnosis is still the standard, the differences between these instruments in our study highlight the importance of a multimodal assessment of ASD symptoms to diagnose ASD reliably in these children.

In line with our hypothesis, and in accordance with the ASD classification prevalence, ASD symptom severity overall was highest in the FXS group and lowest in the NF1 group when we compared the main subscales of the ADOS and SRS. Also in line with our expectations, the scores of the TSC and AS groups mostly fell between the FXS and NF1 scores. The TSC group showed higher symptom severity compared to the AS and NF1 group for the ADOS social affect domain, and higher scores than the NF1 group on the restricted and repetitive behavior domain of the SRS. Based on the similarities in the underlying pathways of FXS and AS, and of TSC and NF1, we hypothesized that we might also find similarities in symptom presentation between these groups. Contrary to our expectations, the overall ASD severity scores of the TSC group seemed to be more similar to those of the FXS group whereas the AS group scores appeared more similar to those of the NF1 group.

By comparing the syndrome groups only including individuals with an instrument-based ASD classification we reduced the within-group variability and we were able to compare our syndrome groups to a non-syndromic ASD group. The severity scores of the syndrome groups differed from those of the non-syndromic ASD group on several domains. Especially on the SRS, the syndrome groups showed less ASD symptom severity compared to non-syndromic ASD in the social domain. On the restricted and repetitive behavior domains on the other hand, the nsASD severity scores were lower than those of the FXS group, and higher than those of the TSC and NF1 groups. Despite some slight differences, the results of the subgroup comparisons were broadly similar for the ADOS and the SRS. On the main scores of the ADOS and SRS, the scores that determine whether someone qualifies for an ASD classification, the syndrome groups seem to perform relatively well in terms of social interaction and communication compared to the non-syndromic ASD group. As we discussed earlier, the DSM-5 requires that social communication difficulties are more impaired than would be expected based on the developmental level of an individual. Our findings suggest that a combination of severe developmental delay and relatively good social communication skills may prevent individuals with these syndromes from receiving an ASD classification or even a clinical diagnosis, which influences their access to health care services.

Our comparison of the ASD symptom profiles revealed several syndrome-specific strengths and challenges. While the FXS group scored highest overall, our results demonstrated a specific relative challenge for this group in the restricted interests and repetitive behavior domain. On the other end of the spectrum, the NF1 group had the least severe ASD symptoms overall, and had a specific relative strength in the restricted and repetitive behavior domain. The low severity scores in NF1 could cause ASD symptoms to be more easily overlooked in clinical practice in this group compared to the other syndrome groups. The TSC group also showed more challenges in the restricted and repetitive behavior domain, but their scores in this domain were lower than the FXS group and the nsASD group. For the AS group we found a relative strength on the social communication and social motivation domains, and a relative weakness in the reciprocal social interaction and in the creativity and play domains of the ADOS. Despite the fact that for all children with AS the lowest module was selected, this effect is likely due to a discrepancy between the developmental level of the AS group and the demands that some ADOS items put on the children. Therefore, in clinical practice, all scores that require a certain cognitive level must be interpreted cautiously for individuals with severe developmental delay.

While the genetic variations in FXS and AS, and in NF1 and TSC, respectively, affect neurodevelopmental pathways in a similar manner, the syndromes are fundamentally different in their origin as well as their phenotypes. Therefore, it would be highly unlikely that these syndromes would show an identical pattern of ASD symptoms on all subscales. Despite several differences, the syndrome pairs did show similarities on a majority of the ASD symptom subscales. To further explore the relationship between the observed similarities in the ASD profiles and the genetic pathways affected in these syndromes deep phenotyping studies are recommended that, for example, include (neural) biomarkers.

A strength of our study is that our sample was relatively free of selection bias. All assessments were carried out in the context of regular clinical care and all children were assessed regardless of whether ASD was suspected. This was supported by the fact that we did not find a difference in ASD severity between the children with and without an IQ assessment. Nevertheless, a selection bias may still exist for the FXS group. As the somatic problems are often mild or absent in FXS—as opposed to NF1, TSC and AS—it is possible that only children with more severe (behavioral) problems choose to visit the ENCORE center of expertise.

Even though our sample is relatively large for a cross-disorder comparison between rare disorders, and the sample is relatively free of bias, we could not control for neurobiological variability within the syndrome groups (e.g., the deletion status in AS, the mutation type in TSC, the locations of neuronal tubers or abnormal tissue growth in NF1 and TSC, or the genetic mosaicism and gender differences related to the X-linked nature of FXS). We also did not account for other comorbidity within our sample, such as epilepsy, ADHD or anxiety disorders. While ASD, social anxiety and ADHD are distinct neurodevelopmental conditions, their symptoms do overlap and therefore comorbidity of ADHD and/or anxiety with ASD may influence scores on both the ADOS and the SRS. Especially in FXS, social anxiety is very common, in children with and without an ASD diagnosis (87). Future in-depth studies should explore these potentially contributing factors in these individual syndromes and study their effect on the development and presentation of ASD symptoms. In the subgroup analyses the sample sizes were even smaller as we included only individuals with an ADOS or SRS ASD classification. We chose to use an instrument-based classification for both the SRS and ADOS separately, instead of DSM-5 criteria, or a combined ADOS and SRS classification, for our subgroup selection because we wanted to examine the broad spectrum of ASD symptoms regardless of whether individuals qualified for a DSM-5 classification, or whether they qualified on both instruments. Although conclusions on the presence of clinical ASD diagnoses within these syndrome groups could not be drawn with this approach, it enabled us to examine both clinical and sub-clinical symptoms present in these syndromes. While a reduction of sample sizes in our subgroup analyses was expected, given the prevalence of ASD in these syndromes, the results of the subgroup analyses should be interpreted cautiously. Therefore, replication of this study in a larger sample, while accounting for syndrome specific features, would be necessary to validate the results of this study. This may be achieved by combining data from large natural history studies, such as the FORWARD study on FXS (88) and the TSC Natural History Database (TSC Alliance) for example. Data from the FORWARD study shows that 87% of children with FXS would classify as having ASD on the SRS, irrespective of ASD diagnosis (89), which supports the findings of this study. However, because ASD is not as prevalent in all syndromes, ASD screening is not standard practice for all syndromes, so selection bias should be accounted for.

Regarding gender differences, males in general are more likely to receive an ASD diagnosis than females (90). In addition, we know that FXS is more prevalent in males, and that males with FXS are more severely affected, in terms of developmental level as well as ASD symptomatology. Therefore, we added gender as a covariate in our analyses. The proportion of males vs. females in our FXS group was high at ~75%. Because the prevalence of ASD in males is higher than in females, in FXS (33) as well as in general (33, 91), including more males may have resulted in a higher ASD classification prevalence in our FXS sample compared to other studies, and a lower developmental level. Even though our sample size was relatively large for a cross-disorder comparison between rare disorders, the number of included females was not sufficient for stratification into gender subgroups. From studies in non-syndromic ASD as well as syndrome groups, we know that gender affects the pattern of ASD symptom presentation (90). Future studies should stratify their data in order to reveal the impact of gender on the presentation of ASD characteristics in males vs. females.

Another limitation of this study is the difference in developmental level between our groups, especially between the syndrome groups and the nsASD group. As our FXS group contained more males than expected, and the IQ of males with FXS is generally lower than that of females with FXS, the IQ of our FXS group may be relatively low compared to more balanced samples. It is known that the developmental level of children and young adults can influence the measurement of ASD symptom severity (77, 78). The ADOS allows clinicians to choose a module based on the developmental level of the participant, which is estimated based on the level of spoken language. However, the cognitive profiles of syndrome groups also show syndrome specific strengths and challenges. Individuals with similar levels of spoken language may vary greatly in other areas of cognitive and motor development. There are no specific norms available for the ADOS for samples with developmental delay or complex behavioral problems. In our study we compared symptom severity in the syndrome groups to a non-syndromic ASD sample in order to better interpret the group differences that we have found. However, it is important to keep in mind that the developmental level of this non-syndromic ASD sample was higher than that of our syndrome groups. We have attempted to reduce the effect of developmental levels on our results by adding IQ or DQ scores as a covariate in our analyses, but as IQ/DQ levels are likely to vary within and between our groups, our results should be interpreted cautiously.