We established a mouse model to differentiate between a lethal and non-lethal presentation of endotoxic shock. The model involved injecting different amounts of Escherichia coli LPS into C3H/HeN mice which had been 'primed' with BCG. We found that the mice receiving non-lethal and lethal doses of LPS could not be differentiated in terms of their physical symptoms for the first 8 h post-injection. Tumour necrosis factor (TNF) was detected at concentrations 2-9-fold greater in mice receiving lethal doses of LPS when compared with non-lethally injected mice. However, given that (i) the successful detection of this differential was dependent on the time of sampling and (ii) that TNF was only detected in the first 3-4 h post LPS challenge, we suggest that TNF may not be very useful as a prognostic marker in endotoxic shock. In contrast, circulating IL-6 appeared to mirror the symptoms of the endotoxic mice. The relative disappearance of IL-6 after 10 h in the non-lethally injected mice corresponded with their symptomatic recovery, while IL-6 continued to circulate up to the time of death in the lethally injected mice. Furthermore, there appeared to be a good correlation between the levels of injected LPS and the levels of IL-6 induced into the circulation. Our results suggest that IL-6, rather than TNF, may serve as a prognostic marker for endotoxic shock.