Comparison of White Matter Microstructure between Affected and Unaffected Siblings Accounting for the Typical Distribution of Difference between Siblings

Background:

Prior studies with diffusion tensor imaging (DTI) have demonstrated that unaffected siblings of people with ASD show similar white matter abnormalities as their affected siblings compared with typically developing controls (TDC), suggesting genetic vulnerability of unaffected siblings. However, these studies did not account for the typical distribution of difference in white matter structure between siblings.

Objectives:

In the current study, we examined differences in DTI parameters between individuals with ASD and their unaffected siblings accounting for the distribution of differences in DTI parameters between pairs of TD siblings.

Methods:

Participants

Fifteen male individuals with ASD (mean age: 28.3 yo) and their unaffected siblings (mean age: 28.0 yo) were recruited. To examine the typical distribution of differences in DTI parameters, fifteen pairs of siblings of male TDC were enrolled (mean age: 28.4 for the older siblings and 25.1 for the younger siblings). In total, we recruited 60 people.

Data acquisition

We obtained diffusion tensor imaging data at Showa University.

Preprocessing

All data were visually inspected for visible artifacts. Eddy current and head motion were corrected. Data with large head motion were excluded from the analyses. Based on the ICBM DTI-81 atlas, 48 region-of-interests (ROIs) that locate in the center of major tracts were defined. Mean fractional anisotropy (FA), mean, axial, and radial diffusivity (MD, AD, RD) were computed from 48 ROIs.

Analyses

Paired t-tests

We examined the difference in mean DTI parameters between individuals with ASD and their unaffected siblings using paired t-test. Statistically significance threshold was set at P<0.001 (=0.05/48: number of ROIs).

Bootstrapping

First, in each ROI, we computed distribution of mean DTI parameters differences between 15 pairs of unaffected TD siblings using bootstrapping with 5000 iterations. Then, mean DTI parameters differences between siblings discordance for ASD were overlayed onto the distribution of difference of DTI parameters among TD siblings to obtain quantile rank. Threshold for significance were set at either quantile above 99.95 percentile or below 0.05 percentile (equivalent to P=0.025/48: corrected for multiple comparisons).

Results:

Paired t-tests

None of the tracts showed significant difference in any DTI values between individuals with ASD and their unaffected siblings (t<1.51, P>0.153 for FA, t<2.92, P>0.011 for MD, t<2.56, P>0.023 for AD, t<2.20, P>0.045 for RD).

Bootstrapping

The analysis revealed that difference in mean AD in the left stria terminalis and right anterior corona radiata between individuals with ASD and their unaffected siblings were statistically significant (above 99.99 percentile). Difference in mean FA between siblings discordant for ASD was also marginally significantly large on typical distribution (99.16 percentile).

Conclusions:

Our novel approach accounting for the typical distribution of difference in DTI parameters captured the potential neural correlates of ASD diagnosis. Given that the stria terminalis is sexually dimorphic and the current project enrolled only males, the narrow distribution of difference of DTI parameter between TD siblings is reasonable. Large difference of the DTI parameters in this tract between individuals with ASD and their unaffected siblings may suggest atypical response to sex hormone of this tract among individuals with ASD.