We studied monoclonal antibodies (MoAbs) directed against human tumour necrosis factor (TNF) for their capacity to prevent toxic or lethal effects of TNF. Two experimental models involving recombinant human TNF (rhTNF) in mice were used: the Shwartzmann reaction, and the lethality after D-galactosamine sensitization. Two MoAbs were found to be protective in both models. These MoAbs prevented mortality when given 6 h, 4 h, or 15 min before rhTNF injection but were not effective if given after TNF. In addition, our results point out that in vitro binding and even neutralizing capacities of anti-hTNF MoAbs do not necessarily reflect their protective efficacy in vivo. Therefore, the models studied here might be useful to evaluate anti-h TNF MoAbs before clinical use.