The first aim of this study was to examine the prevalence of autistic behaviors in individuals with FXS. To accomplish this objective, we administered two well-known measures of autistic behaviors to a large sample of individuals with FXS. Both of these instruments are widely used in the autism literature - the SCQ is a checklist that parallels ADI-R content, while the ADOS is an observational measure of autistic behaviors that is often considered the “gold-standard” in autism research. Results showed that 35.1% of boys and 4.3% of girls scored in the autism category on both instruments, similar to the prevalence rates of autism found in previous studies of individuals with FXS 11, 18, 19
. However, agreement on diagnostic category was poor. The low diagnostic agreement would be expected given that one measure involved parent report (the SCQ) and the other measure involved direct observations of the child under controlled conditions by an expert rater (the ADOS). In boys with FXS, 51.4% scored in the autism category on the ADOS, and 58.7% scored in the autism category on the SCQ, yet only 35.1% scored in the autism category on both instruments. These findings once again highlight the difficulty of applying dichotomous, phenomenologically defined diagnoses (i.e., using cut-off points) to symptoms and behaviors that may exist as continuous variables.
Another explanation for the discrepancy between the prevalence rate of autism observed on the two instruments concerns differences in the content of the items employed in each instrument. For some behaviors (e.g., eye gaze, facial expressions, quality of social overtures, repetitive speech), the content of the items could be considered comparable across instruments (e.g., “stereotyped utterances” on the SCQ vs “stereotyped use of words” on the ADOS). For other items, however, the content between the two instruments appears to be quite different. For example, the SCQ item “inappropriate questions” has no comparable item on the ADOS, while the ADOS items “insight”, “empathy on others' emotions”, “quality of rapport”, and “response to joint attention have no comparable items on the SCQ. Another major difference between the two scales concerns the fact that the repetitive behaviors domain is not included in the scoring algorithm of the ADOS 41
. Interestingly, in their longitudinal investigation of autism in FXS, Hernandez and colleagues 10
reported that the ADOS appeared to generate a high number of “false-positive” cases of autism in their sample of boys with FXS (i.e., 21 of 22 boys with FXS were classified in the autism category on the ADOS while only 12 of these children were classified as autistic using a combination of ADI-R and DSM-IV criteria). The discrepancy between the measures in the present study (and in the Hernandez et al study) could therefore result from the different numbers of items and domains being included in the scoring algorithms of the two measures. Future studies could employ alternate measures of autistic symptoms and/or social behaviors and language function to further assess differences between FXS and autism groups.
The second aim of the study was to determine whether the behaviors shown by individuals with FXS are similar to those diagnosed in individuals with idiopathic autism. Analysis of the items on the two measures indicated that boys with FXS showed particular forms of behaviors (e.g., stereotyped behaviors, repetitive vocalizations and eye gaze avoidance) that appeared at similar rates to those diagnosed with autism. However, impairments in a large number of social behaviors (e.g., social smiling, range of social expressions, quality of social overtures, joint attention) as well as impairments in communicative behaviors (gestures, pointing, imitation), occurred at significantly lower rates than that of individuals typically diagnosed with autism. While girls with FXS also showed a few behaviors that were similar to the autism samples (e.g., attachment to objects, inappropriate questions, gestures, and inappropriate facial expressions), the majority of items on the measures occurred at significantly lower rates than the reference samples. Taken together, these data support previous observations that while individuals with FXS exhibit high levels of social avoidance and repetitive behaviors and language, their reciprocal social interaction skills and communication skills may be qualitatively (and quantitatively) less impaired than in some samples of individuals diagnosed with idiopathic autism 29, 31
A third aim of the study was to determine whether the autistic behaviors of individuals with FXS were associated with IQ levels. A multiple regression analysis indicated that, when controlling for age and FMRP levels, IQ was significantly negatively associated with autistic behaviors in both boys and girls with FXS. Assuming that the negative association between IQ and autistic behavior does not result from item overlap in the measures of autism and IQ, there are several possible explanations for the observed association. First, decreased levels of IQ, or some variable that is correlated with IQ, could trigger or amplify autistic behaviors. Second, autistic symptoms, or some variable that is correlated with autistic symptoms, could impair performance on intelligence testing. Third, IQ and autistic behaviors could result from the effects of an unidentified “third” variable with simultaneous causal effects on both IQ and on autistic symptoms.
Can we conclude from these analyses that the impairment in IQ in children with FXS actually causes autistic symptoms? This would not be a valid conclusion. Although FMRP levels were not predictive of autistic behaviors, this does not preclude the possibility that there is some other unmeasured variable, such as generalized brain dysfunction, that leads to the impaired intellectual development as well as the autistic symptoms in individuals with FXS. In the regression model, the amount of variance in autistic symptoms explained by IQ, age, FMRP and psychoactive medication use was 51% in girls and 28% in boys with FXS. While the low percentage of variance explained in boys with FXS was most likely due to severe range restriction in IQ scores and FMRP levels in this group, there are clearly other important factors, such as the family environment or other biological factors that could be involved in the appearance of autistic behaviors in FXS 47
We found that there was no effect of FMRP levels on autistic behaviors when controlling for age, IQ, and psychoactive medication use. Bailey and colleagues 4
also found no association between FMRP levels and autistic behaviors, as assessed on the Childhood Autism Rating Scale (CARS). By contrast, in a longitudinal study of 83 young children with FXS (mean initial age = 4.5 years), Hatton and colleagues 8
found that FMRP levels and autistic behavior on the CARS were significantly negatively associated. However, intellectual ability levels were not included in their analysis.
Several studies have shown that the caudate nucleus, a region of the brain thought to be involved in learning and memory, is significantly larger in children with FXS compared to typically developing controls 48, 49
. Intriguingly, studies have also shown that the caudate nucleus and amygdala are significantly larger in children with autism, suggesting that there may be common underlying brain pathology in the two disorders 50, 51
. However, in a recent brain imaging study, Hazlett and colleagues 52
reported that children with FXS had a significantly larger caudate nucleus and a smaller amygdala than children with autism. These authors suggested that this “double dissociation” in brain pathology points to differing pathogenic mechanisms underlying FXS and autism.
There are a number of limitations associated with this study. First, only a proportion of the study sample (70 out of 120 participants) received the ADOS assessment. Thus, agreement between the measures on diagnostic category was limited to the 70 participants who had received both instruments. Second, for the 33 girls who received the ADOS assessment, the majority (31 out of 33 participants) had received either Module 3 or 4 because they had fluent expressive language skills. Consequently, data were not available in that group for items on the ADOS that appear only in Modules 1 and 2. A third limitation concerns the potential differences between the current sample of children with FXS and the reference samples. For example, the mean age of the children in the present study was approximately 13 years, while the mean age of the reference samples on the ADOS and SCQ were approximately 9 and 17 years respectively. While we did not find age to be associated to autistic behaviors in this study, there is some evidence from longitudinal studies that autistic behaviors may decrease with age 8
. To circumvent this limitation, we could have included an age- and gender-matched autism control group in our study. This would facilitate minimizing differences between subject groups (e.g., gender and age) and would also ensure that the ADOS assessments were performed in a similar manner in the two groups, if possible, conducted blind to subject diagnosis. Future studies could therefore compare FXS and idiopathic autism groups recruited at the same study site and assessed by the same study team. However, had we recruited our own sample of children with autism at Stanford, they would likely have had some characteristics that differed from the autism reference groups contained in the manuals of the measures (and, incidentally, also from other autism control groups recruited by investigators at other institutions). We therefore felt that comparing our sample of individuals with FXS to individuals with “established” autism (defined by the particular autism measure) was a sound strategy for this initial analysis.
The findings of this study may have important implications for interventions. Currently, children with FXS who show autistic-like behaviors may be referred to intervention services designed for children with autism. However, the data in this study suggests that continuing to implement interventions designed for individuals with autism may not be optimal since individuals with FXS do not necessarily show the same profile or severity of behaviors associated with idiopathic autism. While some subjects with FXS generated summary scores that reached criteria for autism, the scores on many individual symptoms were relatively mild. It is possible that interventions designed to address the specific phenotypic features of FXS (e.g., with an emphasis on repetitive-stereotypic motor behavior and language), rather than the broad symptoms of autism, may therefore be more beneficial to individuals with FXS. Future studies are needed to test this premise. A number of investigators, for example, have begun to evaluate medications targeted to the specific underlying pathology in FXS i.e., the downstream effects of reduced FMRP (for reviews see 53, 54
). Interventions have also been conducted to address specific cognitive and behavioral symptoms shown by individuals with FXS55, 56
. If these FXS-specific interventions are found to be more successful than standard autism interventions, these treatments should be prioritized in comparison to those developed (and relatively non-specifically) for children with autism. Service agencies would also need to recognize FXS as its own entity, with services being made available to all individuals with FXS irrespective of whether or not they have also been assigned a diagnosis of autism. Until more FXS-specific interventions are developed however, it is possible that individuals with FXS who have an additional diagnosis of “autism” may receive more appropriate school services relative to an “other health impaired” diagnosis.
While we could not rule out other factors that might have contribute to the appearance of autistic behaviors in FXS, it seems likely that degree of intellectual ability is at least a proxy for some of those factors. Given this result, and the finding that the profile of behaviors associated with FXS appears to differ from individuals with idiopathic autism, the practice of diagnosing children with FXS as autistic may become increasingly obsolete in the future. In his book “The Concept of Mind”, the philosopher Gilbert Ryle 57
identified several types of “category mistake” that pervade scientific discourse. A category mistake is an instance in which items are suggested to be components or equal members of the same category despite the fact that they exist at different levels of explanation (e.g., “brain” and “mind”). It is possible that the grouping together of FXS (a biological disease) with autism (a phenomenologically-defined behavioral disorder) may be another such type of category mistake. We acknowledge however, that the issue of whether to give an autism diagnosis to individuals with a known genetic syndrome (e.g., FXS, Rett's disorder, Down syndrome and others) is a complex one. We believe that further study of this issue may facilitate the implementation of more specific and effective treatments for individuals with FXS in the future.