Dear FreeSurfer Developers,

I am using FreeSurfer 5.3 to analyze whether a risk factor explains variation in brain structures related to surface area, mean thickness, and subcortical volumes. 

Documentation suggests using estimated Total Intracranial Volume (eTIV) as a covariate to adjust for brain size in analyses of surface area and volumes:  (https://surfer.nmr.mgh.harvard.edu/fswiki/eTIV). Previous posts have also suggested using BrainSegNotVent or Total Gray Matter Volume as covariates as a "whole brain" covariate, with the choice of covariates representing different hypotheses regarding local vs. global atrophy. 

However, it seems to me that using BrainSegNotVent or Total Gray Matter Volume as covariates could lead to a significant loss in power in detecting effects on area/thickness measures from a risk factor associated with atrophy across many regions. On the other hand, eTIV may suffer from not accounting accounting for local atrophy (in addition to heightened measurement error, as frequently alluded into previous posts). But still, a normalization covariate is needed to account for differences in brain size need to be accounted

In the analysis of cortical surface area or mean cortical thickness among normal populations, could it make sense to use cortical white matter volume as the "whole brain" covariate? In normal populations, cortical white matter volume is significantly correlated with surface area, mean thickness, and gray matter volumes but its value is not derived from the gray matter structures. My understanding is that FreeSurfer sets cortical surface area as the area at the boundary of white and gray matter.  As a covariate in models of cortical surface area or mean thickness, white matter volume could side-step the loss in power in detecting effects lead to both global atrophy in gray matter volume while also partly adjusting for total brain differences. 

Cortical white matter volume is not a standard "whole brain" covariate in the literature (in the way eTIV or Total Gray Matter Volume are). However, am I misunderstanding how FreeSurfer computes these Cortical White Matter that makes it unsuitable as a "whole brain" covariate?

Thank you.

Best,

Chintan