A. Research Question, Goal, or Specific Aims
Where in the brain (on a voxel-by-voxel basis) is the gray and white matter volume different in patients with PTSD compared to a healthy control group, from data acquired by the ENIGMA/PGC PTSD group?
ENIGMA has been very successful in identifying regional brain volume reductions in anatomically defined regions by using the FreeSurfer pipeline. This project is distinct because it uses a voxel-based morphometry (VBM) framework. VBM is a complementary technique and does not use a region of interest approach, rather it examines differences in brain volume in a large number of voxels across the brain. We have developed the ENIGMA VBM tool (https://sites.google.com/view/enigmavbm ) which is a MATLAB program that can be used by each research group to process their data (or in one place if all data is held centrally by the PTSD workgroup). Group data can then be sent to us which we will combine using a voxel-wise meta-analytical framework (Seed-based d-mapping; https://www.sdmproject.com/). VBM data in PTSD has been successfully combined before in a meta-analysis by our group (Bromis et al. 2018). We hope to be able to significantly increase the sample size further by including further samples within the ENIGMA PTSD group.
B. Analytic Plan
The main analysis involves SPM-T maps which consist of 3D images of T statistical values reflecting group differences in gray and white matter volume or contrast values at each voxel in the brain. This is produced automatically at each site from our ENIGMA VBM program. The SPM-T group maps and other QC data and statistical group images would be transferred to us. This is analysed in a meta-analytical framework using the software SDM https://www.sdmproject.com/ which applies and a random effects meta-analysis to the data to determine regions which are reduced in brain volume in PTSD.
We would split the analysis into adult, pediatric, military, and civilian samples. The main independent variable will be based on the diagnosis (PTSD or Control).
The dependent variable will be voxel-wise gray and white matter data from the structural MRI data. Covariates will include gender, age, and ICV. Other variables of interest include childhood trauma, comorbidity, and alcohol use.For quality control (QC) the program produces screen shots of individual brains during the pipeline, but this is a low resolution 2D image and is used to check the data correctly processed and is not used in the analysis. The tool has built-in automated quality checks, and we will also conduct a manual quality check on the group data files received.
C. Analytic Personnel
Matthew Kempton, Cheryl See
D. Resources Needed
In terms of the individual sites, all that would be required is MATLAB (nearly all universities and research institutions have licenses for this and we provide a simple way of checking in our instructions) the MRI files in Analyse or NiFTI format and a covariates CSV file that contains clinical data and is similar to the one used in the FreeSurfer analysis.
We do not require individual patient level data to be sent to us as the ENIGMA VBM tool will summarize sites into group level data for the analysis. The data that is produced by the program consists of voxel-wise group-level statistical maps such as SPM-T maps and beta maps. We also would require mean global volumes (e.g., mean and SDs of total gray/white/ICV) for the patient and control group which is produced by the program.
We have experience in running this software in other ENIGMA workgroups and it has been extensively tested. We are also experienced in providing support for any centres encountering problems with the software.
In terms of variables the following will be required:
Dependent variable – voxel wise gray and white matter data from structural MRI data (we would include both 1.5T and 3T MRI data and perform a sensitivity analysis to examine the effect of scanner field strength)
Main independent variable – Diagnosis (PTSD or Control)
Other variables of interest – sample type (civilian, military, pediatric, non-pediatric), comorbidity, childhood trauma, alcohol use
Covariates to adjust for: gender, age, ICV, scanner
