Altogether, we found the autistic behavior profiles in FXS to be almost identical at T1 and T2, but different at T3, with less distinction of the FXS+ASD group from the rest of the FXS cohort. In order to ascertain what underlies this difference and how the autistic behavior profile changes over time, repeated measures ANOVAs were performed on measures of the ADI-R and then confirmed by longitudinal regression models (GEE). We found a clear difference in the nature of the change of autistic behavior over time between the groups, by which FXS+ASD subjects generally improved (i.e., decreased scores) while FXS+None subjects generally worsened (i.e., increased scores). Within the FXS+ASD subcohort, all domains improved, particularly CommNV and Recs, which showed significant and trend level improvement, respectively. At the subdomain level, Recs-4 and Comm-1 improved significantly over time. Within the FXS+None subcohort, all domains worsened, particularly Recs, which showed significant worsening. Recs-2 was the only subdomain to significantly worsen over time. Neither the FXS+ASD subjects nor the FXS+None subjects showed any significant changes over time in Reps or any of its subdomains. Figure 1 depicts the changes in Recs over time.

Analyses were also performed in order to explore the evolution of cognitive parameters in FXS+ASD, namely IQ and language skills. To be certain the BSID-II-based IQ estimates were not skewing our results, a separate analysis was performed including only subjects who were administered the SB-IV. The findings were strikingly similar to the combined SB-IV/BSID-II analysis, particularly in terms of the longitudinal progression of IQ, and so we report the results from the combined SB-IV/BSID-II cohort. As expected, FXS+ASD subjects had significantly lower (i.e., more impaired) full-scale IQ (FSIQ) than FXS+None subjects at T1 and T2. However, while the FXS+None group showed a markedly significant decline (p<0.0001) in FSIQ scores over time, the FXS+ASD group was surprisingly stable, which led to a lack of significant difference in FSIQ between the two groups at T3. Upon further analysis, we found that this drop in FSIQ within the FXS+None cohort was due to worsening in non-verbal IQ (NVIQ) scores, which markedly declined (p<0.0001) over time. Similar to FSIQ, NVIQ in the FXS+ASD group was stable, and FXS+ASD subjects had significantly greater deficits in NVIQ than FXS+None subjects at T1 and T2 but not at T3 (Fig 2). Verbal IQ (VIQ) also differentiated between groups at T1 and T2 but not at T3; however, no significant changes over time were observed in either group. Although both FSIQ and NVIQ standard scores decreased over time in the FXS+None group, raw scores either remained stable or increased slightly. Similarly, raw scores for VABScom and VABSsoc did not decrease in the FXS+None group. This suggests FXS+None subjects are not losing skills, but rather developing at a much slower rate than normal children. Consistent with the pattern seen in the autistic and social behavior profiles, the IQ profiles at T1 and T2 were nearly identical but different from T3, at which point there were no significant differences in any measure of IQ.