Isolated organ perfusion of the liver or extremity with tumor necrosis factor (TNF) and melphalan results in regression of bulky tumors in the majority of patients. The efficacy of TNF in this setting is not known, although data suggest that it may exert antitumor effects primarily on tumor-associated neovasculature. We studied the effects of TNF on capillary leak in liver and tumor tissue during isolated hepatic perfusion (IHP) with melphalan. Twenty-seven patients with unresectable cancer confined to the liver underwent a 60-min hyperthermic IHP using 1.5 mg/kg melphalan alone (n = 7) or with 1.0 mg of TNF (n = 20). Complete vascular isolation was confirmed in all patients using an intraoperative leak monitoring I-131 radiolabeled albumin technique. Samples of tumor and liver were collected just prior to and immediately after IHP. There was no difference in I-131 radiolabeled cpm/g of tissue (cpm) in liver versus tumor at baseline (P2 = 0.44). After IHP, I-131 albumin cpm were higher in tumor versus liver (10,999 +/- 1,976 versus 3,821 +/- 780, respectively; P2 < 0.005). However, I-131 albumin cpm in tumor were not effected by TNF (11,636 +/- 2,518 with TNF versus 9,180 +/- 2,674 without TNF; P2 = 0.59). TNF did not affect melphalan concentrations in tumor (1,883 +/- 540 ng/g versus 1,854 +/- 861 ng/g without TNF; P2 = 0.9). Capillary leak, as reflected by diffusion of I-131 radiolabeled albumin into the interstitial space, is comparable in liver and tumor before IHP but is significantly higher in tumor after IHP. The increased diffusion in the capillary tumor bed must occur through TNF-independent mechanisms such as intrinsic features of tumor neovasculature, hyperthermia, or other unrecognized perfusion-related factors. These data indicate that TNF must continue to be critically evaluated in clinical trials before it is routinely used with melphalan in isolated organ perfusion.