Euphrates Trial Results: Let the Dicing and Slicing Begin! by Steven P. LaRosa, M.D.

This week the much anticipated results of the Euphrates Trial were presented in Vicenza. For those of you not familiar with this acronym, this was the first blinded randomized controlled trial of the Immobilized Polymyxin (PMX) Filter in the treatment of patients with septic shock. This filter known as TORAYMYXIN has been used for years in thousands of patients in Japan and Europe. The mechanism of action is the removal of endotoxin believed to be a noxious mediator of sepsis. This clinical trial was novel in that it included a biomarker, the endotoxin activity assay (EAA), as an inclusion criteria to select patients who had the target of the therapy. Patients were included if they had septic shock and a high EAA level of > or = 0.6. Patients randomized to the treatment arm had hemoperfusion of their blood over the PMX filter during 2 2hour sessions 24 hours apart. Patients in the sham group had the usual standard of care but had a venous catheter layed over their skin with bandage in place. Hemoperfusion was performed by unblinded nephrologists behind a screen such that the clinical investigators where unaware of the patient's treatment assignment.

In all 450 patients were randomized in the study and composed the primary analysis intent to treat (ITT) population. No difference in 28 day mortality was observed in the PMX group in this primary analysis. Much was made in this presentation of an a priori analysis of a per protocol population (patients who had both PMX sessions) and had a MODS score of >9. In this population the 28 day mortality was 31.9% in the treatment arm vs. 36.9% in the sham group. The p value for this analysis however was 0.407 indicating a 40% chance that this observed 5% difference in mortality could have been due to chance alone. The data was further diced into a posthoc subgroup of this subgroup of per-protocol patients with MODS > 9 and EAA levels between 0.6 and 0.9. In this subgroup a 10.7% lower mortality was observed in the PMX group compared to the sham group with a p value of 0.0474. This analysis does not correct for the multiplicity of analysis that took place in the study i.e. a Bonferroni correction. Simply stated you must divide the p value by the number of subgroups you are testing. In this case, we know that at least 2 subgroups have been tested so the p value of 0.05  must at least be divided by 2 so that statistical significance must be less than 0.025. You are not in statistical sense allowed cut the data multiple times until you find a subgroup with a statistically significant benefit and then infer efficacy. The best that can be said about this subgroup is that it is "hypothesis- generating" and that this treatment effect would need to be confirmed in an additional adequately powered trial.

In conclusion, the efficacy of the PMX filter in patients with septic shock was not confirmed in THE EUPHRATES TRIAL. This therapy should not be approved by regulators for the treatment of patients with septic shock. If the sponsor, SPECTRAL DIAGNOSTICS, truly believes that this therapy works in patients with MODS > 9 and EAA levels between 0.6 and 0.9 then they will need to perform an additional trial to prove this.