Convection-enhanced drug delivery (CED) enables achieving a drug concentration within brain tissue and brain tumors that is orders of magnitude higher than by systemic administration. Previous phase I/II clinical trials using intratumoral convection of interleukin-4 Pseudomonas exotoxin (PRX321) have demonstrated an acceptable safety and toxicity profile with promising signs of therapeutic activity. The present study was designed to assess the distribution efficiency and toxicity of this PRX321 using magnetic resonance imaging (MRI) and to test whether reformulation with increased viscosity could enhance drug distribution. Convection of low- [0.02% human serum albumin (HSA)] and high-viscosity (3% HSA) infusates mixed with gadolinium-diethylenetriamine pentaacetic acid and PRX321 were compared with low- and high-viscosity infusates without the drug, in normal rat brains. MRI was used for assessment of drug distribution and detection of early and late toxicity. Representative brain samples were subjected to histological examination. Distribution volumes calculated from the magnetic resonance images showed that the average distribution of 0.02% HSA was larger than that of 0.02% HSA with PRX321 by a factor of 1.98 (p < 0.02). CED of 3.0% HSA, with or without PRX321, tripled the volume of distribution compared with 0.02% HSA with PRX321 (p < 0.015). No drug-related toxicity was detected. These results suggest that the impeded convection of the PRX321 infusate used in previous clinical trials can be reversed by increasing infusate viscosity and lead to tripling of the volume of distribution. This effect was not associated with any detectable toxicity. A similar capability to reverse impeded convection was also demonstrated in a CED model using acetic acid. These results will be implemented in an upcoming phase IIb PRX321 CED trial with a high-viscosity infusate.