Cytokine Hemoadsorption During Cardiac Surgery Versus Standard Surgical Care for Infective Endocarditis (REMOVE): Results From a Multicenter Randomized Controlled Trial

Infective endocarditis (IE) affects 1 to 10 of 100 000 persons per year worldwide and is associated with up to 40% in-hospital mortality.^1–3 Approximately 50% of these patients require surgical treatment.^1,4 Cardiac surgery is often the only means of eradicating infected cardiac tissues and reconstructing or replacing dysfunctional or infected valves. Cardiopulmonary bypass (CPB) can trigger a sudden release of inflammatory mediators, such as cytokines and vasoactive peptides, into the cardiovascular circulation,⁵ and the manipulation of infected cardiac tissues during surgery can exacerbate this inflammatory response. Such a mechanism may be similar to sepsis models in which bacterial products were injected into animals⁶ and human volunteers.⁷ These investigations revealed a short but intense elevation of cytokine levels (often referred to as a cytokine storm⁸) that was associated with disorders in microcirculation⁹ and ultimately organ failure.^10,11 Interleukin (IL)–1, IL-6, tumor necrosis factor (TNF), and IL-18 have been considered key cytokines both in sepsis and during surgical procedures and are thought to play central immunopathologic roles.^10,12 We previously investigated the release profiles of cytokines and vasoactive mediators before, during, and after cardiac surgery in patients with IE as well as in patients with noninfectious valvular heart disease (VHD).¹⁷ This pilot study to our randomized trial demonstrated increased levels of cytokines and vasoactive mediators after starting CPB in both groups. The increase was much more prominent in patients with IE and the greatest differences versus the patients without IE were noticed during CPB. In addition, a significant correlation between plasma levels of IL-6, PCT (procalcitonin), and midregional pro-adrenomedullin (MR-proADM) during CPB and the severity of postoperative organ dysfunction has been demonstrated.¹⁷ It is reasonable to speculate that hemoadsorption during CPB aiming to dampen the sudden increase in circulating inflammatory mediators might reduce organ dysfunction after cardiac surgery in patients with IE.

CytoSorb (CytoSorbents Europe GmbH) is a hemoadsorption device designed for the extracorporeal reduction of cytokines in the circulating blood and has been widely used as adjuvant therapy in patients with sepsis.^18–20 This hemoadsorption device has the ability to rapidly reduce many key cytokines in experimental settings of endotoxemia²¹ and has been associated with fewer organ injuries and longer survival in animal models.^21–23 Although several studies have been conducted with CytoSorb in patients undergoing cardiac surgery, the results have been inconsistent,^24–28 and clinical outcomes have not yet been assessed in the scarce randomized evidence.^17,29–31

We performed a prospective, randomized, nonblinded, controlled trial to investigate whether the use of intraoperative hemoadsorption in patients undergoing cardiac surgery for IE reduces the severity of postoperative organ dysfunction, as measured by changes in the sequential organ failure assessment (SOFA) score. The efficacy of hemoadsorption in reducing plasma levels of inflammatory mediators was documented in a run-in phase of this trial.

Methods

Data Disclosure Statement

Access to the data will be supervised and granted by the Center for Clinical Studies of Jena University. Data from this study will be archived for 10 years on the servers of the Center for Clinical Studies to be available for use by other investigators, as long as there is no conflict with the copyrights of the publisher. Requests should be sent to Mahmoud Diab, Department of Cardiac and Thoracic Surgery, Am Klinikum 1, 07747 Jena, Germany (Email mahmoud.diab@med.uni-jena.de).

Study Design

REMOVE (Revealing Mechanisms and Investigating Efficacy of Hemoadsorption for Prevention of Vasodilatory Shock in Cardiac Surgery Patients With Infective Endocarditis) is an investigator-initiated, multicenter, randomized, nonblinded, controlled trial with 2 groups designed for assessing superiority. The trial was conducted in 14 cardiac surgery centers in Germany. The trial protocol has been described previously³² and is provided as Supplemental Material. The study was approved in September 2017 by the institutional review board and ethics committee of the Friedrich-Schiller University of Jena (5240-08/17) and by the ethics committee at each participating center.

Study Participants

Patients were eligible for enrollment if they were 18 years or older and had IE according to the modified Duke criteria³³ with an indication for surgery in accordance with the European guidelines for IE treatment.³⁴ If the patient was fully competent, written consent was sought. If capacity was impaired and a legal representative was present, consent was sought from the representative. Obtaining informed consent was carried out according to §28 of the Declaration of Helsinki.

Exclusion criteria were a low surgery-associated mortality risk (European System for Cardiac Operative Risk Evaluation Score [EuroSCORE II³⁵] ≤3), pregnancy, or immunosuppressive or immunomodulatory therapy.

Randomization

After screening, patients were assigned in a 1:1 ratio using web-based central randomization into 1 of the 2 treatment groups. To achieve baseline covariate balance across centers, randomization was stratified by center. Patients undergoing any of the compared treatments were part of the full analysis set (a modified intention-to-treat analysis set). Figure 1 shows the study flowchart.

• Download figure
• Download PowerPoint

Procedures

Eligible patients were randomly assigned to receive hemoadsorption during CPB using CytoSorb or to the control group. The hemoadsorption device was integrated into the CPB as in Figure S1. Patients in the control group were operated on according to the standard of care (ie, no hemoadsorption was connected to the CPB circuit).

The baseline SOFA score was assessed within 24 hours before surgery and the mean total postoperative SOFA score was calculated from the first postoperative day until discharge from the intensive care unit (ICU) or intermediate care (up to the 9th postoperative day). Clinical follow-up of patients was obtained on the 30th postoperative day.

In a run-in phase (the first 25 patients in each study arm), cytokine measurements were taken to assess the efficacy of hemoadsorption in reducing cytokines. Blood samples were collected at skin incision, 30 and 60 minutes after starting CPB, at the end of CPB, and 24 hours after surgery. Samples were assessed for IL-1β, IL-6, IL-10, and TNF-α with the Invitrogen ProcartaPlex high-sensitivity assay (Bender MedSystem, part of Thermo Fisher Scientific); IL-18 with the Invitrogen IL-18 ProcartaPlex assay (Bender MedSystems, part of Thermo Fisher Scientific); and PCT, C-reactive protein (CRP), C-terminal proendothelin-1 (CT-proET-1), MR-proADM, copeptin pro-arginine vasopressin (CT-proAVP), and midregional pro-atrial natriuretic peptide (MR-proANP) with BRAHMS Kryptor assay (BRAHMS GmbH, part of Thermo Fisher Scientific). In addition, samples for cell-free DNA (cfDNA) were collected at skin incision, 60 minutes after starting CPB, and 24 hours after surgery for the first 25 patients in each study arm (cfDNA isolation with QIAsymphony Circulating DNA Kit; single-read sequencing of whole genome shotgun [WGS] DNA library with Illumina HiSeq 2500 [≈30 Mio reads/sample, possible deviation up to ±5%]; and quality control with fragment analyzer and qubit).

Primary Outcomes

The primary outcome was ΔSOFA, defined as the difference between the mean total postoperative SOFA score and baseline score within 24 hours of surgery. The primary outcome and adverse events were centrally assessed. The SOFA score was measured on a scale ranging from 0 to 4 for each of 6 organ systems, with an aggregate score from 0 to 24, with higher scores indicating more severe organ dysfunction.³⁶ The mean total postoperative SOFA score was assessed from the first postoperative day (24 hours after surgery) until discharge from the ICU or intermediate care unit (maximally to the ninth postoperative day). ΔSOFA has served as a clinically relevant outcome in several clinical trials of sepsis^37–39 and can be used as a prognostic score to assess short-term mortality as recommended by Sepsis-3 (The Third International Consensus Definitions for Sepsis and Septic Shock).⁴⁰

Secondary Outcomes

The secondary outcomes of the study were as follows: 30-day mortality rate, defined as mortality within 30 days postoperatively in the hospital or anywhere after discharge; ΔSOFA subscores; cumulative incidence of stroke within 30 days of surgery; duration of mechanical ventilation; vasopressor and renal replacement therapy within 30 days after surgery; length of in-hospital and ICU stay; changes in plasma levels of cfDNA; and inflammatory mediators at the predefined time points.

Cytokines and vasoactive peptides were measured centrally in the laboratory of the external cooperation partner (BRAHMS [part of Thermo Fisher Scientific]). cfDNA was analyzed at the Department of Molecular Biotechnology/Functional Genomics, Fraunhofer Institute for Interfacial Engineering and Biotechnology, Stuttgart, Germany.

Safety assessments included the documentation of perioperative complications, adverse events, and serious adverse events.

Statistical Analyses

All statistical analyses were performed with SAS 9.4 (SAS Institute). Continuous secondary outcomes were compared by 2-sided Wilcoxon Mann-Whitney tests; median and 25th/75th percentile were reported for each group. In particular, Wilcoxon Mann-Whitney tests were applied for all cytokines at each time point (at skin incision, 30 and 60 minutes after starting CPB, at the end of CPB, and 24 hours after surgery). Fisher exact test (2-sided) was applied to compare binary secondary outcomes between groups and absolute and relative frequencies are provided for each group.

Sample size calculations were performed for the primary outcome of the mean ΔSOFA until day 9 postoperatively. Sample sizes were calculated for a Pocock-type group sequential design with a planned interim analysis for 2×25 patients (20% of the total sample size) and a final analysis. Previous studies demonstrated a 1.4-point lower SOFA score in the intervention group to be clinically relevant.^37,41 A common SD of 3.8 points was assumed, which results in a standardized effect size of 0.368 SD.⁴¹ To achieve 80% power at an overall 2-sided significance level of α=5% while including an interim analysis after 2×25 patients (α1=0.0147) and a final analysis (α2=0.0378), 125 patients needed to be included per study arm (2-sided group-sequential z test; nQuery Advisor V.7.0 Statistical Solutions). In a previous pilot study, the Δ SOFA for patients with IE was 5 (3–10).¹⁷ On the basis of an earlier study,⁴¹ we expected a 15% dropout rate, leading to a total of 296 (2×148) patients to be randomized. All analyses were performed on a modified intention-to-treat population including all randomized patients who received any of the treatments. In the primary analysis, the experimental group with CytoSorb hemoadsorption and the control group with standard care were compared regarding differences in mean total SOFA scores until discharge from the ICU or intermediate care unit (maximally to the ninth postoperative day). A 2-sided test with α=0.05 (α₁=0.0147 [2×25 patients] and α₂=0.0378 [final analysis]) was considered, because both the superiority of the experimental and that of the control group are of interest (ie, potential harm of the device).

The primary analysis was performed by a predefined linear mixed model including surgeon and baseline SOFA as fixed effect covariates and surgical center as random effect. The null hypothesis was rejected if the P value related to the Wald test statistic for the treatment effect was ≤α₁ or α₂. The safety of participants was overseen by an independent data and safety monitoring board.

All analyses of the secondary outcomes were considered exploratory (ie, we made no adjustment for multiplicity).

Kaplan-Meier analysis was performed for survival up to 30 days and survival curves were compared by the log-rank test. Hazard ratios with 95% CIs were reported to estimate the treatment effect.

Results

Between January 17, 2018, and January 31, 2020, 740 patients were assessed for eligibility. Among them, 288 patients were randomly assigned to hemoadsorption (n=142) or control (n=146).

Figure 1 shows the study flowchart. Among the 288 randomized patients, 6 patients did not undergo surgery (4 in the hemoadsorption group and 2 in the control group) and had to be excluded from the study for lack of data on primary outcome (ΔSOFA). Thus, 282 patients (138 patients in the hemoadsorption group and 144 patients in the control group) were included in the modified intention-to-treat analysis.

As shown in Table 1, the 2 groups’ baseline characteristics were well balanced. The mean baseline SOFA score within 24 hours before surgery, on which we relied to calculate the primary outcome, did not differ between groups (5.6±2.5 in the hemoadsorption group and 5.6±2.4 in the control group). Microbiological data did not differ between groups except for more coagulase-negative Staphylococcus in the control group than in the hemoadsorption group (25% versus 11.6%; P=0.01).

Figure 2 shows the primary outcome (ΔSOFA) for the 2 study groups. ΔSOFA did not differ between groups (1.79±3.75 for the hemoadsorption group and 1.93±3.53 for the control group; 95% CI, –1.30 to 0.83; P=0.666).

• Download figure
• Download PowerPoint

Table 2 shows the secondary outcomes for the 2 study groups. There was no evidence of a group difference in any secondary outcome.

Kaplan-Meier curves for survival at 30 days are shown in Figure S3. There was no difference between the hemoadsorption and control groups (log-rank, 0.078; hazard ratio, 0.931 [95% CI, 0.564–1.539]).

Figure 3 shows the intraoperative and postoperative plasma levels of IL-1β, IL-18, CT-proET-1, and IL-6 of the first 50 patients in the trial. Baseline plasma level values of the 4 inflammatory mediators did not differ significantly between groups. Plasma levels of IL-1β, IL-18, and CT-proET-1 at the end of CPB and of IL-6 24 hours postoperatively were significantly lower in the hemoadsorption group than in the control group.

• Download figure
• Download PowerPoint

Plasma levels of other inflammatory mediators as well as the cfDNA of these 50 patients are shown in Table S1. There was no evidence for a group difference in the plasma levels of any of these inflammatory mediators (IL-10, TNF-α, PCT, CRP, MR-proADM, CT-proAVP, MR-proANP) at any time point. The mean cfDNA level at 60 minutes after starting CPB was higher in the hemoadsorption group.

Table S2 shows changes to baseline plasma levels of the inflammatory mediators during CPB and 24 hours postoperatively. At the end of CPB, median changes to baseline levels were greater in the hemoadsorption group than in the control group for the following inflammatory mediators: CT-proET-1 (–95.5 versus –26.8; P=0.01), MR-proADM (–0.5 versus 0.1; P<0.01), MR-proANP (6.7 versus 150.4; P<0.01), and PCT (–0.06 versus 0.02; P<0.01).

Table 3 lists the postoperative adverse events in both groups. More than two-thirds of the patients had distributive shock, more than half experienced acute kidney injury, more than one-third had respiratory insufficiency, and 11% required reexploration for bleeding. However, there was no evidence of group differences. Table S2 shows details of the protocol violations in each group. There were 51 patients with protocol violations (27 in the hemoadsorption group and 24 in the control group). The Duke criteria for definite IE were not fulfilled by 18 patients at the time of inclusion. However, in 14 of those 18 patients, the IE diagnosis was confirmed later intraoperatively, microbiologically, or histologically. Two patients in each group did not receive the assigned treatment (crossover). CytoSorb was applied against protocol in the ICU in 9 patients (in 5 patients of the hemoadsorption group and 4 of the control group).

Discussion

This randomized controlled trial on patients undergoing cardiac surgery for IE demonstrates that hemoadsorption applied during CPB failed to reduce the severity of postoperative organ dysfunction. Although hemoadsorption reduced plasma cytokines at the end of CPB, there was no difference in any of the clinically relevant end points, including 30-day mortality. These results therefore question a direct link between reducing plasma cytokine levels and preventing organ dysfunction.

In sepsis, high levels of proinflammatory cytokines (ie, IL-6, IL-8, IL-18, and IL-1β) have been linked to higher mortality.⁸ Patients with IE are prone to an excessive release of cytokines and other inflammatory mediators during surgery.¹⁷ This release has been attributed to CPB use and to the exacerbation of inflammation through the direct manipulation of infected material while removing vegetations and resecting infected tissues. There is ample evidence that this rise in levels of cytokines during CPB correlates with the severity of postoperative organ dysfunction in patients with IE.¹⁷ The aim of this trial was to test the hypothesis that dampening such an increase in cytokines would reduce the severity of organ dysfunction in patients with IE.

Previous studies proposed the extracorporeal elimination of cytokines through hemoadsorption as a potential strategy to modulate the inflammatory response.^21,22 CytoSorb is specially designed for the nonselective removal of cytokines. CytoSorb hemoadsorption has been used in >100 000 patients in more than 67 countries and >200 case reports or observational studies have been published on its application in different settings.^24,26,30,42 However, to date, there is not a single randomized study investigating CytoSorb while also assessing a clinically relevant primary end point. A few small randomized studies investigated the efficacy of CytoSorb in reducing plasma levels of cytokines in sepsis^27,28 or cardiac surgery,^26,29–31 but the results are inconsistent.

This study is the first multicenter randomized controlled trial investigating the efficacy of CytoSorb in reducing the severity of postoperative organ dysfunction in an adequately powered population of patients undergoing surgery for IE. The primary outcome, ΔSOFA, has been used in large sepsis studies.^40,41 ΔSOFA is considered clinically relevant because it represents the main score used to assess the degree of organ dysfunction in the ICU and reflects the Sepsis-3 recommendations. A SOFA score ≥2 points is associated with an in-hospital mortality rate of >10%.⁴⁰ This is consistent with this trial’s findings, in which a preoperative SOFA score ≈6 was associated with a mortality rate >20%.

This trial verified the efficacy of CytoSorb in reducing cytokines and other inflammatory mediators in the first 50 patients enrolled. Despite demonstrating that CytoSorb was able to dampen surgery-induced increases in plasma levels of many key proinflammatory cytokines at the end of CPB, its use was not associated with any reduction in the severity of postoperative organ dysfunction or with any of our clinically relevant secondary outcomes (ie, 30-day mortality; duration of postoperative hemodialysis, ventilation, and vasopressor therapy; or length of ICU and hospital stay).

This study was designed to assess the effect of intraoperative hemoadsorption on the severity of postoperative organ dysfunction on the basis of the rationale that the surgery-induced surge in plasma cytokine levels may be causally linked to the extent of organ dysfunction.¹⁷ The total duration of hemoadsorption in this treatment group was 2.31±1.45 hours. This short duration may be considered a limitation and longer durations of hemoadsorption may exert additional effects. Although we cannot draw any conclusions on the treatment effects of longer hemoadsorption on the basis of this trial’s findings, the evidence available does not support the speculation that longer durations of hemoadsorption may exert additional effects.^27,43 Studies have investigated the cytokine-reducing effect of hemoadsorption lasting 42 hours in patients with severe sepsis²⁷ or lasting 72 hours in patients with COVID-19 pneumonia requiring extracorporeal membrane oxygenation.⁴³ Both failed to demonstrate a CytoSorb-induced reduction in plasma IL-6 level. One of these studies suggested a negative effect on survival.⁴³

Another potential limitation of this trial is the lack of blinding. However, the lack of blinding is unlikely to create a relevant bias because the primary outcome (ΔSOFA) is calculated on the basis of objective clinical and laboratory data routinely collected in the ICU, where the treatment was identical between groups. An additional trial limitation is our lack of information on postoperative right ventricular dysfunction.

Our trial failed to provide any evidence that hemoadsorption during CPB decreased the severity of postoperative organ dysfunction in patients undergoing cardiac surgery for IE. Although hemoadsorption seemed to reduce CT-proET1, IL-18, and IL-1β plasma levels at the end of CPB, it did not lower 30-day mortality or shorten the duration of postoperative hemodialysis, ventilation, or vasopressor therapy or the lengths of ICU and hospital stay.

Article Information

Acknowledgments

The authors thank the members of the data safety monitoring board (Matthias Loebe, Miami Transplant Institute, Memorial Jackson Health System, University of Miami, Florida; John A. Kellum, Department of Critical Care Medicine, University of Pittsburgh, Pennsylvania; Peter U. Heuschmann, Institute for Clinical Epidemiology and Biometry, University of Würzburg, Germany) and their partners for the laboratory examinations (BRAHMS [part of Thermo Fisher Scientific]; Department of Molecular Biotechnology/Functional Genomics, Fraunhofer Institute for Interfacial Engineering and Biotechnology; and Institute for Medicinal Microbiology, University Hospital Jena, Germany).

Supplemental Material

Supplemental Methods and Results

Tables S1–S7

Figures S1–S3

List of Definitions

REMOVE Trial Investigators

Footnotes