Hi Andrea,
Others may differ on their opinions to this. The approach that you suggest is what I have used in the past, which showed good correspondence between the VBM8, FSL_Anat, and FreeSurfer pipelines (within reason that is).
0. Make sure you have smoothed the data in either SPM or AFNI.
1. Use 3dFWHMx to estimate the smoothness of each anatomical brain produced from VBM8 toolbox (m0wrp1*.nii). I would recommend not using -automask, and instead creating a group average from the grey matter (3dMean) and then creating a mask of that (3dAutomask). You might need to play with the -clfrac and -dilate options for masking a bit! Then use that mask for 3dFWHMx.
2. For 3dClustSim, I would just supply it the mask from before (which provides voxel and image sizes) and the -fwhmxyz values. I tend to tack on the -LOTS and -niml options.
For final analyses I ended up using 3dMVM and 3dttest++.