In animal models of disease, the roles of IL-1 and TNF have been defined by specifically inhibiting each cytokine. The "anticytokine strategies" for treatment of septic shock follow the same approach. Interleukin-1 can be blocked by reducing its synthesis [1]; infusing IL-1-receptor antagonist [2-4]; or administering soluble (extracellular) IL-1 receptors [5]. Blocking TNF can be accomplished by reducing its synthesis, infusing neutralizing antibodies, or administering TNF-binding proteins that are the soluble forms of p55 or p75 TNF receptors [6, 7]. These agents are now being assessed in clinical trials.

What triggers the synthesis and release of IL-1 and TNF? The most potent agonist is bacterial lipopolysaccharide (or endotoxin), although it is important to remember that exotoxins from gram-positive bacteria and some fungal products can also stimulate IL-1 and TNF synthesis. Human blood monocytes are exquisitely sensitive to endotoxin, producing IL-1 and TNF in vitro in response to 25 to 50 pg/mL of endotoxin, a concentration achieved in the circulation during septic shock as reported by Casey and colleagues [8] in this issue of Annals. Human volunteers receiving an injection of 3 ng/kg of Escherichia coli-derived lipopolysaccharide show a dramatic increase in circulating TNF- (from <5 pg/mL to as much as 750 pg/mL, depending on the assay method). The theoretical maximum concentration of lipopolysaccharide in the blood of these volunteers would be approximately 25 pg/mL.

Thus, it is of interest in understanding the pathogenesis of septic shock that Casey and colleagues report that circulating levels of IL-1 ß, TNF-, and IL-6 correlate with mortality in these patients. Moreover, when the data were analyzed by a "lipopolysaccharide-cytokine score," which is based on lipopolysaccharide and cytokine concentrations, the correlation with mortality was highly significant. In the analysis of these patients, IL-6 was the cytokine that was most consistently elevated and that correlated best with mortality, as other studies have shown [9]. However, because IL-6 lacks the ability to induce a shock-like syndrome in animals or humans, the plasma IL-6 level is a marker rather than a cause of the syndrome. Interleukin-6 production appears to reflect biologically active IL-1 and TNF, in that blocking either of these latter cytokines significantly reduces the IL-6 level [10-12].

The importance of IL-1 or TNF, or both, in the pathogenesis of septic shock in patients must be proved by intervention studies showing improved survival when these cytokines are specifically blocked. Clearly, correlative clinical studies of plasma cytokine levels and mortality do not establish causality. Thus, the study by Casey and colleagues deserves a careful reading for other reasons. In a recent clinical trial of an anti-TNF monoclonal antibody for septic shock, increased survival was seen only in patients with the highest levels of circulating TNF- [9]. In a phase III trial of IL-1- receptor antagonist, a statistically significant improvement in survival was observed only in patients with sepsis who were at the highest risk for death [13]. Although the levels of circulating IL-1 ß are unknown in that study, we speculate that those patients with the highest risk for death would have shown the highest levels of IL-1 ß if it was collected and measured by validated methods [8, 14, 15].

Therefore, the value of measuring IL-1 ß and TNF- in the circulation may be to identify which patients are likely to benefit from anticytokine therapy. This information has the potential to improve the design of clinical trials and thus reduce the possibility that a useful therapy for septic shock will be disregarded because of apparent failure of the trial through misapplication of the therapy. Once a new therapy is approved, information on cytokine status can also reduce costs by identifying the appropriate patients for treatment. Although these are theoretical prospects, several pitfalls need to be overcome.

First, plasma contains factors that interfere with cytokine assays (reviewed in [16]); these include nonspecific and specific binding proteins. Casey and colleagues carefully tested their assays to be sure that physiologically relevant amounts of the cytokines could be recovered. They checked samples with high immunoreactivity by diluting and retesting them, and they verified immunoassay results by bioassay. They also rapidly separated the plasma from the blood cells and thus avoided artifacts associated with clotted blood. The clinical study of cytokines is in disarray because of the proliferation of commercial assay kits that are poorly characterized by the manufacturers and are used indiscriminately by researchers. Casey and colleagues are to be commended for their careful attention to assay validation. A Danish group has recently validated an IL-1 ß enzyme-linked immunosorbent assay (ELISA) by column chromatography and bioassay procedures and showed a correlation between plasma IL-1 ß levels and mortality in patients with severe burn injury [15].

Second, validated assay systems can still yield different results. Although Casey and colleagues found that high circulating IL-1 ß levels were associated with high mortality, others have found contrasting results using different assay methods. Which method is correct? It is possible that each assay correctly answers different questions. Bioassays obviously measure biologically active forms of IL-1. These include mature 17-kd IL-1 ß and smaller fragments, regardless of whether IL-1 ß is free or bound to carrier proteins such as ₂-macroglobulin. The radioimmunoassay used in some studies [17, 18] detects the relatively inactive 31-kd pro-IL-1 ß as well as bound and free forms of mature and fragmented IL-1 ß [14]. The IL-1 ß ELISA of Cistron that was used in Casey and colleagues' study detects about 10% to 15% pro-IL-1 ß [19] and does not detect IL-1 ß fragments nor IL-1 ß bound to carrier proteins. (Information provided on request from Richard Dondero of Cistron Biotechnologies. Clearly, such information is critical to the interpretation of assay results. Investigators pay a premium price for these assay kits and they have a right to demand that the kits are well characterized. All reputable kit manufacturers should furnish such information and data.) The bioassay and radioimmunoassay are better suited to determine if a host is producing IL-1 ß at all. It is important to remember that picomolar concentrations of IL-1 enhance host defense mechanisms [20] and that only excessive amounts promote disease. Free circulating IL-1 ß may exist only after natural binding proteins are diminished by disease or become saturated by excess IL-1 ß production resulting in a pathological state.

Third, septic shock does not provide the clinician with the luxury of waiting several hours for the results of a cytokine assay before making a decision about treatment. We do not recommend routine cytokine assays for clinical decision making in septic shock. Interleukin-1 and TNF contribute to septic shock by increasing the expression of genes coding for the synthesis of small mediator molecules such as prostaglandins and nitric oxide. This increased expression can result in a 4- to 6-hour difference between peak cytokine levels in the circulation and the onset of clinical shock. We do recommend prospective studies designed to show a constellation of early clinical assessments that correlates with circulating cytokine levels and, at the same time, the risk for death in these patients. It is possible to determine in less than 4 hours the amount of IL-1 ß, TNF, and IL-6 messenger RNA in a small volume of blood using the polymerase chain reaction. If this method and cytokine assays can be applied to prospective trials, the design of anticytokine therapy trials and treatment of patients with septic shock could be improved. Studies in animals clearly show the disadvantage of late anticytokine treatment, and humans entered late into anticytokine intervention are probably also at a disadvantage.