Renal denervation (RDN) is a treatment option for patients with hypertension resistant to conventional therapy. Clinical trials have demonstrated variable benefit. To understand the determinants of successful clinical response to this treatment, we integrated porcine and computational models of intravascular radiofrequency RDN. Controlled single-electrode denervation resulted in ablation zone geometries that varied in arc, area, and depth, depending on the composition of the adjacent tissue substructure. Computational simulations predicted that delivered power density was influenced by tissue substructure, and peaked at the conductivity discontinuities between soft fatty adventitia and water-rich tissues (media, lymph nodes, etc.), not at the electrode-tissue interface. Electrode irrigation protected arterial wall tissue adjacent to the electrode by clearing heat that diffuses from within the tissue, without altering periarterial ablation. Seven days after multielectrode treatments, renal norepinephrine and blood pressure were reduced. Blood pressure reductions were correlated with the size-weighted number of degenerative nerves, implying that the effectiveness of the treatment in decreasing hypertension depends on the extent of nerve injury and ablation, which in turn are determined by the tissue microanatomy at the electrode site. These results may explain the variable patient response to RDN and suggest a path to more robust outcomes.