FDA set to approve "blood purification" for sepsis?

What is hemoperfusion, and can it save lives in endotoxin-related septic shock?

Oct 20, 2025

Sepsis is sometimes called “blood poisoning” by laypeople, and they’re not wrong. Can the blood of patients with sepsis be removed from their bodies, detoxified, and returned, saving lives in the process?

The U.S. FDA may soon be convinced the answer is yes, thanks to the reportedly positive results of a new randomized trial. That could usher in a new therapy for sepsis, but the TIGRIS trial’s unconventional methods may leave practicing physicians skeptical of the results.

“Blood purification” for sepsis

Many of the systemic derangements and dysregulation in severe sepsis with shock occur from the uncontrolled production and circulation of inflammatory cytokines. Circulating gram-negative bacterial endotoxin (aka lipopolysaccharide) is a key contributor to the so-called cytokine storm that can result.

Excess and dysregulated circulating cytokines activate endothelium and increase nitric oxide production, causing blood vessels to dilate and leak fluid extracellularly, resulting in hypotension. (Numerous other derangements occur in the extremely complex systemic reaction to sepsis.)

Investigators have long sought therapies to filter out cytokines and endotoxin, the two primary instigators of the often deadly response by the body to sepsis, Removing them from the blood circulation, it was hoped, would improve survival and other outcomes from sepsis.

Hemadsorption ≈ Hemoperfusion for sepsis

Hemadsorption and hemoperfusion are roughly interchangeable descriptors for technologies that attempt to clear cytokines or endotoxin from the bloodstream.

A dual-lumen vascular catheter (as for dialysis) is placed in a large vein, and blood is pumped extracorporeally using a standard hemodialysis or ECMO machine. Instead of a dialysis membrane, the blood travels through a proprietary cartridge lined with a chemically adhesive material. These may be polymyxin B fibers (to clear endotoxin), porous beads made of styrene or polystyrene copolymers (for cytokines), or other substances.

As blood is pumped extracorporeally through the circuit, endotoxin and/or cytokines are sequestered in the cartridge, and the theoretically “cleaned” blood is returned to the body.

It’s an elegant idea with a strong basis in physiology, but its implementation has not yet been clearly shown to help patients with sepsis.

Cytosorb™ and Toraymyxin™

Two major purification technologies are available: non-selective cytokine removal (marketed in Europe as CytoSorb) and selective endotoxin removal (polymyxin B hemoperfusion, under investigation in the U.S. as Toraymyxin™ by Spectral Medical Inc).

CytoSorb

CytoSorb cartridges include porous beads of copolymers that non-selectively adsorb molecules ~5–60 kDa in size, including cytokines like IL-6 and TNF-alpha.

CytoSorb has been approved as an adjunctive therapy for sepsis in Asia and Europe for more than a decade, based on its technical success at removing the relevant proteins. It has not been shown to improve mortality or other outcomes in sepsis in randomized trials, but nor has it been thoroughly tested.

In one of the only randomized trials, Schadler et al (PLOSOne 2017), no mortality benefit was noted among 100 randomized patients with severe sepsis; hemadsorption-treated patients did not experience reduced cytokine levels; the treatment duration was criticized as too short.

Most or all other studies of CytoSorb in sepsis have been observational, with propensity matching used to claim benefit.

CytoSorb was also tested in severe COVID-19 in the CYCOV trial (n=34, underpowered, hemadsorption seemed to cause severe harm, did not reduce IL-6, trial stopped early) and in Stockmann et al (Crit Care Med 2022: no benefit, underpowered pilot study).

Polymyxin B endotoxin removal by hemoperfusion (Toraymyxin™)

Polymyxin B is a cationic peptide that, when arranged in fibers inside a cartridge, attracts and sequesters negatively charged lipopolysaccharide endotoxin as it circulates through an extracorporeal blood circuit, a process called hemoperfusion. Polymyxin does not enter the circulation or act as an antimicrobial per se.

Polymyxin B hemoperfusion has a larger body of randomized data for evaluation than CytoSorb, but still small.

EUPHAS trial (2009)

This multicenter Italian trial planned to enroll ~120 patients with septic shock undergoing emergency surgery for a presumed intra-abdominal source, but stopped halfway (n=64) at a prespecified interim analysis that found the 34 polymyxin B hemoperfused patients had significantly lower mortality, higher blood pressure, reduced vasopressor use, and lower organ failure scores. The small sample size at termination and unblinded assessments raised the likelihood of a chance finding, however. (Cruz et al, JAMA 2009)

ABDOMIX (2015)

A similar but larger population of septic shock patients requiring emergency surgery for peritonitis (n=243) was then enrolled at 18 French ICUs. Hemoperfusion with polymyxin-treated patients experienced numerically much higher mortality (33.6%) than conventionally treated patients (24%), which was statistically non-significant but trended toward harm (95 % CI 0.91–2.87). The hemoperfused patients had no reduction in organ failure scores (P=0.78). (Payen et al, Intensive Care Medicine 2015)

EUPHRATES (2018)

The largest, multicenter, randomized trial of 450 patients with severe sepsis and high endotoxin activity (EAA level ≥0.60) at 55 tertiary hospitals in North America, running from 2010 to 2017.

Those receiving polymyxin hemadsorption experienced no mortality benefit over those getting a sham treatment (37.7% 28-day mortality with polymyxin hemoperfusion vs. 34.5% with sham).

Nor was there a benefit in the sickest patients with high organ failure scores. Nor did patients treated with hemadsorption in EUPHRATES experience a reduction in endotoxin activity compared to sham-treated patients. (Dellinger et al, JAMA 2018)

At this point, hemadsorption with polymyxin B or copolymers might have been consigned to the growing heap of valiantly attempted sepsis therapies that were shown to be non-beneficial.

“Precision medicine” and post hoc P-hacks

In EUPHRATES, mortality was unsurprisingly very high in the patients with the highest levels of endotoxin (EAA ≥0.90), and much lower in those with the lowest levels of endotoxin. A discussion likely arose at that point among the stakeholders:

Maybe the high-endotoxin patients’ septic shock was overwhelming and sadly beyond help?

Conversely, maybe those with the lowest levels of endotoxin would not benefit from hemoperfusion, and their inclusion only made it more difficult to detect a difference between groups?

And maybe this technology that Spectral Medical licensed from Japan’s Toray Industries in 2009 could still be brought to the U.S. market somehow?

A post hoc analysis limited to the 194 patients in EUPHRATES with endotoxin levels between 0.60 and 0.89 did show a numeric reduction in 28-day mortality with polymyxin hemoperfusion (26% vs 37%). The non-significant P-value of 0.11 was reduced to 0.04 after naughtily adjusting the data for imbalances in severity of illness between the groups (this was a randomized trial, remember).

There were other hints of a benefit from hemoperfusion in the post-hoc tea leaves:

When treated with hemoperfusion, the “Goldilocks” subgroup with moderate endotoxemia had improved blood pressures and fewer ventilator days.
In contrast to the overall treated population, they did experience a measurable reduction in endotoxin.
In fact, the sub-sub-group of 87 patients with significant endotoxin clearance had almost a 50% relative mortality reduction (15.7% vs 30.6%), but that required sub-sub-grouping down to a statistically shaky 21 deaths (from the 162 deaths in the trial overall).

Because post hoc subgroup analyses are vulnerable to chance findings or even frank manipulation, they are considered hypothesis-generating only; a much larger randomized trial limited to patients with moderate endotoxemia would be required to establish any truth in the subgroup findings. But that would cost a lot of money and take a long time—luxuries that start-up theranostics companies rarely enjoy. (Why “theranostics?” We’ll get to that.)

Fortunately for Spectral Medical, a large trial wouldn’t be necessary to convince the FDA to consider approving hemoperfusion as a new treatment for sepsis in the U.S. In the new era of reliance on Bayesian statistics, the results from another underpowered and inconclusive trial could be pumped up to claim a “>99% probability” that hemoperfusion saves lives.

That trial, TIGRIS, will be the subject of part 2 of this post.