When you specify the initial scan as the first, and the final scan as the second, then negative Jacobian differences from the Longitudinal toolbox indicate apparent volume loss. When you multiply these by the grey matter images, the this should indicate volume loss in GM.
I'm not sure exactly how or what extracting ind. Values is, although I'd guess it is the data from the plot function of SPM, which orthogonalises the raw value with the uninteresting confounds from the design matrix. I'd therefore guess that the more negative values in those who started the drug would indicate that there is more volume loss.
Because the pairwise version of the Longitudinal toolbox only gives the Jacobian difference rather than the Jacobians at each time point, you'd need to re-run the longitudinal registration with the "serial" option to generate the Jacobian maps at the two time points. You could then multiply these by the c1 image for that subject to obtain the GM at each time point. Alternatively, you could just segment the original images.
Best regards,
-John