A second-generation replication-conditional herpes simplex virus type 1 (HSV) vector defective for both ribonucleotide reductase (RR) and the neurovirulence factor γ34.5 was generated and tested for therapeutic safety and efficiency in two different experimental brain tumor models. In culture, cytotoxic activity of this double mutant HSV vector, MGH-1, for 9L gliosarcoma cells was similar to that of the HSV mutant, R3616, which is defective only for γ34.5, but was significantly weaker than that of the HSV mutant hrR3, which is defective only for RR. The diminished tumoricidal effect of the γ34.5 mutants could be accounted for by their reduced ability to replicate in 9L cells. The MGH-1 vector did not achieve significant prolongation of survival in vivo in the syngeneic 9L rat gliosarcoma model for either single brain tumor focus or multiple intracerebral and leptomeningeal tumors, when the vector was applied intratumorally or intrathecally, respectively, and with or without subsequent ganciclovir (GCV) treatment. In identical 9L brain tumor models with single and multiple foci, application of hrR3 with or without GCV was previously shown to result in marked long-term survival. Contrary to the findings with intrathecal injection of hrR3, no vector-related mortality was observed in any animals treated with MGH-1. Thus, in these rat brain tumor models, the double mutant, replication-conditional HSV vector MGH-1 showed a higher therapeutic safety than the RR-minus vector, hrR3, but had clearly decreased therapeutic efficiency compared to hrR3. The development of new HSV vectors for brain tumor gene therapy will require a balance between maximizing therapeutic efficacy and minimizing toxicity to the brain. Standardized application in brain tumor models as presented here will help to screen new HSV vectors for these requirements.