Are you really limited in siRNA or you may use shRNAs? In the second case you can use retroviral or even better lentiviral particles carrying the shRNA. In case you do not want genome integration you can use adenovirus. If you have to use siRNA, it is more empirical. It depends in which organ you want it to act. For example you can inject them intraperitoneal or in the tail vein, but you then will have to check if they arrive and act on the organ you want. In the case of viruses you also have to do the same, only it is easier to detect were the virus went since they usually carry some fluorescent protein together with the shRNA.
Sorry for my more than late response ... but here attached are two interesting articles ... and maybe a third one (to be published soon) ...
For brain cancers, you can use the experimental approach described here in Le Mercier et al. JNEN 2008, which mimic the "Omaya reservoir" delivery to treat GBM patients.
depends on what is objective of your study. Local delivery is typically more effective. Systemic is tough, and there are currently very few options that work robustly- with liver being pretty much the only organ that is "addressed"