Dear all
I'm trying to compare the cortical thickness between the neurodegenerative disease patient group and normal controls.
To match the age, I selected subjects from both groups and finally included age-matched groups into freesurfer thickness analysis (similar mean age, same age distribution from 40-73, 70 normal controls and 91 patients)
In 1st qdec analysis, direct comparison between two groups without age covariate showed thickness of temporoparietal cortex of patient group was smaller than that of normal controls (uncorrected P<0.001 or FDR corrected P<0.05).
However, when the age was included in the model as a covariate, the significant surface vertices almostly disappearred and some unexpected vertices emerged as significant, in which mean thickness were not so much different between two groups.
My question is...
1) Is there any possibility that disease-related cortical thinning area is quite similar to the age related cortical thinning area and resulted in the disappearance of significant vertices?
2) Do I have to include age covariate in that model even if I clearly matched the age distribution?
3) The area "does the average thickness differ between normal and patient?" was just same as the area "does the thickness--age correlation differ between normal and patient?", except for the color of blobs (e.g. red -> blue, blue -> red). Then, what do those question mean?
From Lyoo
류철형 wrote:
Dear all
I'm trying to compare the cortical thickness between the neurodegenerative disease patient group and normal controls.
To match the age, I selected subjects from both groups and finally included age-matched groups into freesurfer thickness analysis (similar mean age, same age distribution from 40-73, 70 normal controls and 91 patients)
In 1st qdec analysis, direct comparison between two groups without age covariate showed thickness of temporoparietal cortex of patient group was smaller than that of normal controls (uncorrected P<0.001 or FDR corrected P<0.05).
However, when the age was included in the model as a covariate, the significant surface vertices almostly disappearred and some unexpected vertices emerged as significant, in which mean thickness were not so much different between two groups.
My question is...
- Is there any possibility that disease-related cortical thinning
area is quite similar to the age related cortical thinning area and resulted in the disappearance of significant vertices?
Yes, it is possible. It looks like it might be that there is an interaction between group and age, based on what you say in #3. This can complicate looking for differences between group thickness means as the difference itself becomes age dependent.
- Do I have to include age covariate in that model even if I clearly
matched the age distribution?
Technically, you do not, but it should be a good idea to do so. You can try one of these things: (a) use DOSS (fits the same slope to both groups), or (b) stay with DODS and demean the ages. This is the same as what you are doing already, but it tests for the difference at the mean of your group instead of at 0.
- The area "does the average thickness differ between normal and
patient?" was just same as the area "does the thickness--age correlation differ between normal and patient?", except for the color of blobs (e.g. red -> blue, blue -> red). Then, what do those question mean?
This is sort of expected if there is an interaction. With DODS, you have two lines (one for each group). If they are not parallel (ie, the slopes are different, indicating an interaction) then they will cross at some age. If that age is to the right of age=0, then they will have different signs.
doug
From Lyoo
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer
freesurfer@nmr.mgh.harvard.edu
-
Douglas N Greve -
류철형