ARTEMIS: terminal complement blockade failed cardiac surgery kidney injury a second time

What was tried

Alexion ran ARTEMIS, registered as NCT05746559 and ALXN1210-CSA-AKI-318, a phase 3, randomized, quadruple-masked, placebo-controlled, multicenter study. Participants received a single weight-based intravenous dose of ravulizumab or placebo one to seven days before non-emergent sternotomy with cardiopulmonary bypass. Doses ranged from 2700 mg for participants of 30 to 40 kg up to 3600 mg at 100 kg or above, and model-based simulation predicted complete C5 inhibition for at least 18 days from that single dose.

Eligibility required known chronic kidney disease of stage 3A, 3B, or 4 for at least 90 days, plus planned multi-vessel coronary artery bypass grafting, valve replacement or repair, or a combined procedure. Emergency surgery, off-pump surgery, prior transplantation, and cardiogenic shock were excluded. The primary endpoint was the number of participants experiencing major adverse kidney events at day 90 after bypass, defined as initiation of kidney replacement therapy, death from any cause, or a fall in estimated glomerular filtration rate of at least 25 percent from baseline. The estimate was based on serum cystatin C rather than creatinine, because skeletal muscle loss after cardiac surgery can inflate creatinine-based filtration estimates.

The trial started on 2023-04-06. Planned enrollment was approximately 736 participants, 368 per arm, sized for at least 90 percent power to detect a 10 percentage-point absolute difference in the primary endpoint at a two-sided alpha of 0.05. Actual enrollment reached 555. Primary completion was recorded as 2025-10-07 and study completion as 2026-05-07. The posted reason for stopping reads: "The ALXN1210-CSA-AKI-318 trial was discontinued on 29-Apr-2026 due to lack of efficacy in this overall study universe / patient population."

The biological hypothesis

Cardiopulmonary bypass activates complement through several routes at once. Blood contacts the bypass circuit, hemolysis releases free heme, and the resulting oxidative stress drives cleavage of C5 into C5a and C5b. C5a is a potent anaphylatoxin and C5b initiates assembly of the membrane attack complex. In the kidney, that terminal pathway has direct experimental support: work in the Journal of Clinical Investigation in 2000 showed a predominant role for C5b-9 in renal ischemia and reperfusion injury.

Ravulizumab binds C5 and blocks its cleavage, producing immediate and sustained terminal pathway inhibition. It is approved in paroxysmal nocturnal hemoglobinuria, atypical hemolytic uremic syndrome, and generalized myasthenia gravis, so its pharmacology and safety profile were well characterized before ARTEMIS opened. Patients with pre-existing chronic kidney disease have limited renal reserve, which made them the plausible enrichment population for a peri-operative renal protection effect.

Open Targets scores the C5 association with acute kidney injury at 0.3302, driven by clinical evidence (datatype score 0.533) and literature (0.210). C5 is a well-drugged target: Open Targets lists 13 distinct clinical or approved agents against it, including eculizumab, crovalimab, pozelimab, zilucoplan, cemdisiran, nomacopan, and pexelizumab.

What actually happened

ARTEMIS stopped for lack of efficacy, and no endpoint-level result has been posted to the registry. The registry does not report the interim event rates, the observed treatment difference, or which analysis triggered the decision. What is on the record is that 555 of a planned 736 participants were randomized, that primary completion had already been reached in October 2025, and that the sponsor called the failure on 2026-04-29.

The class precedent is public and predates the trial by nearly two decades. Pexelizumab, a single-chain antibody fragment against C5, was tested in PRIMO-CABG and reported in JAMA in 2004. Among 2,746 patients undergoing coronary artery bypass grafting alone, death or myocardial infarction within 30 days occurred in 134 of 1,373 pexelizumab patients (9.8%) and 161 of 1,359 placebo patients (11.8%), a relative risk of 0.82 with a 95 percent confidence interval of 0.66 to 1.02 and p = 0.07. The trial missed its primary endpoint. In the broader intention-to-treat population of 3,099 patients including valve procedures, the same relative risk of 0.82 reached significance, with a confidence interval of 0.68 to 0.99 and p = 0.03. A 2008 meta-analysis in the Journal of Thoracic and Cardiovascular Surgery pooled 15,196 patients across the pexelizumab program.

Failure mechanism, best guess

The archetype is efficacy_failure, stated by the sponsor. The likely mechanism is a mismatch between the breadth of injury in cardiac surgery associated kidney injury and the narrowness of terminal complement blockade.

Bypass-related kidney injury has multiple parallel drivers: hemodynamic instability and reduced renal perfusion, hemolysis and free hemoglobin toxicity, microembolism, nephrotoxic drug exposure, and inflammation. Terminal complement is one contributor within that set. Blocking C5 completely still leaves the upstream C3 pathway, the contact and coagulation systems, and the non-immune insults intact. PRIMO-CABG had already shown what the ceiling looks like: a point estimate near a 20 percent relative reduction in a cardiac composite, with confidence intervals that touched or crossed no effect. ARTEMIS was powered for a 10 percentage-point absolute difference in a 90-day kidney composite, which is a considerably larger demand than the class had ever demonstrated in this surgical setting.

How to prevent this next time

No endpoint-level ARTEMIS data are available, so a Bayesian reanalysis cannot be built here. The levers that were available are the qualitative ones, and the strongest of them is base-rate adjustment against a directly analogous prior trial.

Base rates. PRIMO-CABG tested terminal C5 blockade in cardiopulmonary bypass at a scale ARTEMIS could not match and did not clear its primary endpoint. Any prior for a peri-operative C5 effect in this population should have been anchored on that result rather than on the mechanistic strength of the C5b-9 ischemia-reperfusion literature.

Biomarker enrichment. Chronic kidney disease stage identifies patients at high risk of the outcome. It does not identify patients whose injury is complement-driven. Peri-operative complement activation markers, for example C5a or soluble C5b-9 measured across the bypass run, would separate a mechanism-enriched population from a merely high-risk one.

Effect-size discipline. A 10 percentage-point absolute reduction in a 90-day composite was an assumption, not an observation. Sizing to a smaller and better-supported effect changes both the trial and the decision to run it.

The single highest leverage change would have been to run a complement-activation-enriched phase 2 with a biomarker-anchored effect estimate before committing 555 patients to a phase 3 powered for an effect no C5 inhibitor had shown in cardiac surgery.

What this means for similar programs

C5 is now a heavily worked target with 13 clinical or approved agents on the Open Targets record, and its wins cluster in diseases where complement is the disease rather than one input among many. Programs extending C5 inhibition into multifactorial acute organ injury, whether in cardiac surgery, transplantation, or critical care, should treat ARTEMIS and PRIMO-CABG as the relevant reference class. The saturation component of the modelled risk score for C5 is 13.89 out of 15, the highest of its five inputs.

Open questions

What were the observed MAKE90 rates in each arm, and how close did the point estimate come to the assumed 10 percentage-point difference? Did any prespecified subgroup, for example participants with higher baseline complement activation or longer bypass time, show separation? Was the 2026-04-29 decision driven by a formal interim futility boundary or by an unblinded sponsor review? Will Alexion publish the dataset?

Sources

1. • ClinicalTrials.gov v2 record, NCT05746559, ARTEMIS. https://clinicaltrials.gov/study/NCT05746559 - A phase 3 study of ravulizumab to protect patients with chronic kidney disease from cardiac surgery-associated acute kidney injury and major adverse kidney events (ARTEMIS). Trials, 2025. https://- Terminal complement blockade with pexelizumab during coronary artery bypass graft surgery requiring cardiopulmonary bypass (PRIMO-CABG). JAMA, 2004. https://- Pexelizumab in ischemic heart disease: a systematic review and meta-analysis on 15,196 patients. Journal of Thoracic and Cardiovascular Surgery, 2008. https://- Predominant role for C5b-9 in renal ischemia/reperfusion injury. Journal of Clinical Investigation, 2000. https://- Kidney injury following cardiac surgery: a review of our current understanding. American Journal of Cardiovascular Drugs, 2025. https://- Ravulizumab pharmacokinetics and pharmacodynamics in patients with generalized myasthenia gravis. Journal of Neurology, 2023. https://- Research progress on the complement system in ischemia-reperfusion injury of organ transplantation. Transplantation Reviews, 2026. https://- Open Targets Platform, target C5 (ENSG00000106804), association with acute kidney injury (MONDO_0002492), accessed 2026-07-26. https://platform.opentargets.org/target/ENSG00000106804 - ChEMBL molecule record CHEMBL3989986 (ravulizumab). https://www.ebi.ac.uk/chembl/compound_report_card/CHEMBL3989986/ - openFDA drug adverse event API, ravulizumab and ULTOMIRIS query, accessed 2026-07-26. https://api.fda.gov/drug/event.json.