During a phase I study of recombinant human tumor necrosis factor (TNF) in cancer patients, serial immune studies were performed and analyzed for effects of TNF. The TNF (specific activity 9.6 x 10(6) U/mg protein, < 5.0 endotoxin units/mg protein) was given over 2 h intravenously on days 1, 8-12, 29-33, 50-54, and 71-75 at doses of 40, 80, 160, 200, and 240 micrograms/m2. Immunologic testing was performed before therapy three times and subsequently on days 2, 8, 10, 12, 29, 33, 50, 54, 71, 75, and off-study two times. Immune parameters evaluated included cytotoxicity [natural killer (NK), spontaneous lymphokine activated killer cells (LAK), LAK, and monocyte], cytokine production [spontaneous and stimulated interferon (IFN)-gamma and interleukin (IL)-2], superoxide production [resting and stimulated polymorphonuclear (PMN) and mononuclear cells (MNC)], and phenotype of peripheral blood lymphocyte subsets (CD3, CD4, CD8, CD16, CD56, CD19). Data were analyzed for long-term effects, the effect after 1 day of treatment (day 1), and for weekly effect (change from day 1 to day 5 of a given treatment week). Significant decreases were seen in the spontaneous cytotoxicity of peripheral blood NK cells and IL-2-inducible LAK cells, whereas increases in spontaneous peripheral blood LAK activity were seen with TNF treatment. Consistent increases in superoxide production of resting PMN and MNC were demonstrated, with late increases in superoxide production by opsonized, zymosan-treated PMN. No spontaneous IFN-gamma or IL-2 were noted in sera with treatment, but production of IL-2 by MNCs rose with TNF treatment. During 5 days of TNF treatment, the percentages of circulating CD8+ and CD56+ cells decreased, whereas that of CD4+ and CD19+ cells increased significantly and consistently, as determined by a multivariate analysis. Significant changes in several independently measured parameters were observed, including a dose-related diminished production of IFN-gamma by MNC stimulated by phytohemagglutinin and increased in vitro-generated LAK activity. Because there was no clinical response in this trial, no association of immunologic change with clinical response can be made. No biologically optimal dose of TNF was evident. The data suggest that TNF may act as a trigger cytokine, initiating a broad immune/inflammatory response.