The present study builds upon the findings of our previous investigations that characterized the autistic and related social behavior profiles of boys with FXS who meet DSM-IV criteria for ASD [Kau et al., 2004
; Kaufmann et al., 2004
; Budimirovic et al., 2006
], by extending into a longitudinal evaluation of ASD in FXS. Our results demonstrated that ASD, and in particular Aut, is a stable diagnosis, with approximately 70% diagnostic agreement over time. FXS+ASD also seems to be a distinct FXS subphenotype, since, for the most part, the same profile of autistic behaviors and other abnormalities in social behaviors contributed to either differentiation between the FXS+ASD and FXS+None groups or autistic behavior severity over the three-year period: relatively greater impairment in complex social interactions, predominantly peer relationships; deficits in socially-relevant communication, mainly make-believe or social imitative play; severity of social withdrawal; and delays in adaptive communication and receptive language skills. In addition, in terms of skills, delay in adaptive socialization was the most consistent correlate of autistic behavior severity. Despite this overall stability, autistic behavior scores decreased in FXS+ASD boys across time points, which, in conjunction with the worsening of autistic behaviors and social withdrawal in FXS+None boys, led to less differentiation between the two subcohorts at the last time point (T3) than at baseline (T1) and the midpoint of assessment (T2). Scores on several cognitive parameters changed over time within the FXS+None group but were mainly stable in FXS+ASD, contributing to less group distinction at T3. Prominent among these parameters were adaptive socialization and non-verbal IQ. Although some social behaviors improved in the FXS+ASD group (e.g., socioemotional reciprocity), acquisition of communication skills did not progress at a steady rate, leading to a decrease in receptive and expressive language skills over time. Consequently, the last two parameters differentiated FXS+ASD from FXS+None at T3. The finding that the two groups, FXS+ASD and FXS+None, have clearly different developmental trajectories further supports the conclusion of FXS+ASD as a unique entity in FXS.
At present, FXS is one of the main genetic causes of ASD [Muhle et al., 2004
; Schaefer et al., 2008
]. The controversy about the meaning of meeting DSM-IV criteria for an individual with a well-delineated genetic disorder, such as FXS, will probably remain well beyond the time when current diagnostic guidelines are replaced in 2012 [Kaufmann et al., 2008
]. Therefore, in-depth characterizations of FXS+ASD are needed in order to develop adequate diagnostic and therapeutic strategies for this substantial subset of mainly male FXS subjects [Hagerman et al., in press]. Stability of a diagnostic label is an important criterion for its validity. In longitudinal studies of idiopathic Aut utilizing the ADI-R, among other measures, stability of Aut diagnosis over time was reported to be between 67–80% [Charman et al., 2005
; Moss et al., 2008
; Kleinman et al., 2008
]. However, no studies, to our knowledge, have examined whether the DSM-IV classification of ASD is a stable diagnosis in the FXS population. Our data indicate that, in FXS boys aged approximately 3–8 years, the diagnosis of ASD is relatively stable. Approximately 80% of boys with FXS+Aut remained within the ASD spectrum, with about 60% maintaining the Aut diagnosis. This contrasts with the approximate 20% diagnostic agreement over time for PDD. Between T1 and T3, 30% of boys with FXS+Aut improved out of the ASD category while the proportion in boys with FXS+PDD was 75%. On the other hand, 20% of boys with FXS+None worsened into the ASD category. These figures reflect the relative improvement in autistic behavior over time in FXS+ASD and the opposite in FXS+None, which led to a slight decrease in the overall proportion of FXS+ASD subjects during the three-year observation period (43% at T1 vs. 35% at T3). The marked stability of the Aut diagnostic label in this study further supports its consideration as a distinctive entity within the FXS phenotype. Conversely, the instability of the PDD diagnosis probably represents the influence of communication and other deficits in young children that can mimic mild autistic impairment. To some extent, the situation in FXS is not substantially different from that in the general population, in which young children receive the PDD diagnosis early but are found to be out of the spectrum later in life [Turner and Stone, 2007
; Kleinman et al., 2008
]. Considering the trend towards improvement in a variety of abnormal behaviors but stagnation in the acquisition of language skills in our FXS+ASD group, a longer follow up is needed in order to fully appreciate the significance of ASD diagnosis in males with FXS.
One of the main outcomes of our previous research on ASD in FXS was the demonstration of a distinctive profile of autistic behavior in FXS+ASD [Kaufmann et al., 2004
], as well as the identification of other social behaviors, both social skills and aberrant behaviors, that are major correlates of ASD in FXS [Kau et al., 2004
; Budimirovic et al., 2006
]. The present investigation demonstrated that the profile of relatively more prominent deficits in peer relationships and socioemotional reciprocity in FXS+ASD [Kaufmann et al., 2004
] is indeed a core feature, since it was found to be the main correlate of autistic behavior severity and a distinguishing factor for ASD status at all time points. Only impairment in socioemotional reciprocity was not a determinant of ADI-R total scores or of ASD status at the last observation, because of a relative improvement in scores of FXS+ASD boys in this area. Contrary to some assumptions about the FXS neurobehavioral phenotype [Baumgartner et al., 1995; Hagerman, 2002
], basic nonverbal social behaviors (e.g., eye gaze avoidance) and stereotypic and repetitive behaviors do not have major influence on diagnosis or severity of autistic behavior. Other than social interaction, only deficits in communication, as measured by the ADI-R, contributed to autistic severity over time. Examination of skills confirmed that only adaptive socialization, and not communication skills, consistently influenced severity of autistic behavior at all time points, and ASD status, as we reported earlier [Kaufmann et al., 2004
; Budimirovic et al., 2006
], at the first two time points but not at T3. Social withdrawal was a parameter that distinguished FXS+ASD from FXS+None over time, but not a determinant of ASD status per se
. Similarly, adaptive communication and receptive language skills longitudinally differentiated the FXS+ASD group but were not determinants of ASD status.
These data support the conclusions of two of our previous studies [Kaufmann et al., 2004
; Budimirovic et al., 2006
], that FXS+ASD is characterized by a primary impairment in socialization skills and deficits in communication skills are a secondary component, as well as that severe social withdrawal is an important but not obligatory attribute. Data on adaptive socialization in idiopathic ASD and in FXS highlight the importance of this parameter in understanding and diagnosing FXS+ASD [Kraijer, 2000
; Fisch et al., 2002
; Fenton et al., 2003
; Hatton et al., 2003
; Budimirovic et al., 2006
]. While our current findings on the longitudinal ADI-R profiles are consistent with studies of autistic behavior in idiopathic ASD, which report significant improvement in scores in ADI-R’s Recs and Comm domains and less drastic, or lack of, change in ADI-R’s Reps domain [Piven et al., 1996
; Fecteau et al., 2003
; Seltzer et al., 2003
; Charman et al., 2005
; Shattuck et al., 2007
; Moss et al., 2008
], to our knowledge, no other study on ADI-R profiles in FXS equivalent to the present one has been published to date. Therefore, the significance of our findings in FXS for ASD of other causes is still unknown. Nevertheless, our longitudinal data in conjunction with the few cross-sectional studies of FXS+ASD [Rogers et al., 2001
; Kau et al., 2004
], examining a variety of cognitive and behavioral parameters, support the notion of FXS+ASD as a distinctive subphenotype in FXS that shares many features with idiopathic ASD.
A secondary set of findings in this study involved the evolution of cognitive parameters in boys with FXS. Fisch and colleagues 
demonstrated that in children with FXS+None or idiopathic Aut, declines in IQ scores were observed during the younger years in both groups, but as children aged, the idiopathic Aut group attained a plateau while the FXS+None group continued to decline. Importantly, however, the authors point out the decline in scores was not representative of loss of skills, but rather of slow development. This is in line with our current findings of marked declines in FSIQ standard scores over time, particularly in NVIQ, in our FXS+None subcohort, reflective of delayed development as raw scores either improved slightly or remained stable. In contrast, the FXS+ASD group remained cognitively stable. This unexpected profile contributed to the diminished cognitive and behavioral distinction between the FXS+ASD and FXS+None groups at T3. The findings in the FXS+ASD group are in limited agreement with the literature on idiopathic ASD, which indicates that the initial relative greater delay in VIQ (vs. NVIQ) tends to disappear by adolescence [Werner et al., 2005
]. In our FXS+ASD sample, both VIQ and NVIQ tended to remain stable and, if anything, verbal skills appeared to be more impaired in late childhood. Since the mean ages in our cohort were about 5–7 years from T1 to T3, perhaps declines in IQ within the FXS+ASD subcohort occurred prior to enrollment in our study and at a younger age than in FXS boys without ASD, who may reach a plateau at an age outside the present study’s range. Wright-Talamante and colleagues 
also studied longitudinal changes in IQ in FXS children. After comparing FXS boys with full methylation with those with either allele size or methylation mosaic patterns, they concluded that mosaic subjects, with a higher mean IQ, may be at greater risk for decline since their initial cognitive abilities were higher than the full mutation subjects. These findings are in correspondence with our findings of IQ score decline in our higher functioning FXS+None subcohort. However, in the Wright-Talamante et al. study the mean baseline age of the full mutation group was well above the mean baseline age of our cohort, while the mean baseline age of the mosaic subjects was in the middle of our cohort’s age range. Therefore, further longitudinal studies of younger and older populations with different behavioral profiles will be needed to better understand IQ changes in FXS over time.
Similar to Philofsky et al. 
, we found receptive language was a relative strength in our FXS+None subcohort, showing higher scores than in expressive language at all time points. Receptive and expressive language scores were almost identical in our FXS+ASD subcohort at each time point, strengthening the already established thought that early deficits in receptive language are an early warning sign for ASD [Philofsky et al., 2004
], although language skills did not influence autistic behavior severity or ASD status.
The present study is, to our knowledge, the first attempt at evaluating ASD in FXS over time using DSM-IV criteria. A recent investigation [Hatton et al., 2006
] of autistic behavior in FXS applying the Childhood Autism Rating Scale (CARS) found a slight, but significant, increase in CARS scores over time. However, this report did not differentiate autistic and non-autistic groups; therefore, the results may be concordant with the increase in ADI-R scores in our FXS+None group. No other investigation has addressed the ASD phenomenon in FXS children from a longitudinal perspective. In spite of the strength of our analytical approach and the use of DSM-IV criteria, we recognize our study presented several limitations. Our sample size and diagnostic composition of the ASD cohort (i.e., low N for PDD and Aut by T3) were limiting factors. The use of two different measures of cognitive function, and the incorporation of an IQ-like DQ ratio for a subset of subjects, as well as two measures of language skills, were other limitations inherent to the study of young children with lower cognitive function. A direct measure of non-verbal communication should also be incorporated in future studies, since many children in this population are non-verbal and deficits in communication were shown to have some influence on autistic behavior severity and ASD classification. Based on our results, interventions targeting the behaviors and skills found to remain impaired or delayed over time in the FXS+ASD subcohort, such as peer relationships and adaptive socialization, could be a more effective strategy in the treatment of children with FXS and ASD. Further, our data show there is general improvement in autistic behavior in FXS+ASD subjects over time and perhaps with intervention this improvement could be further enhanced. Similarly, early intervention in skills that worsened over time in FXS+None subjects, such as NVIQ and behaviors related to peer relationships, may prove useful to help prevent this decline. The strength of our data indicates FXS+ASD is clearly a distinctive and significant entity, with implications for ASD in general, which warrants further research efforts using behavioral and other approaches (e.g., neuroimaging). Of concern is the wide range of prevalence rates of Aut in FXS due to several different methods and measures employed by the various research groups, which can be misleading. We stress the caution that must be exercised when using the ADOS-G for primary or confirmatory diagnosis of ASD in children with FXS due to its susceptibility for false positives. The high prevalence of social anxiety in the FXS population [Budimirovic et al., 2006
; Sullivan et al., 2007
; Hagerman et al., in press] may be a contributing factor to these false positives; however, this issue has not been formally studied. Therefore, we suggest a standard method for diagnosing ASD in FXS that adheres to DSM-IV criteria using both the ADI-R and the ADOS-G, and, if necessary, the utilization of the VABS to settle diagnostic disagreements between the two assessments, as reported [Tomanik et al., 2007
]. Although ADI-R items were included in our analyses, due to the length and complexity of this paper, we chose to exclude these results. However, examination of impairment in specific behaviors (e.g., “play” behaviors) in FXS+ASD and their contributions to ASD status is an important aspect that should be investigated further.