Bedside Allogeneic Erythrocyte Washing with a Cell Saver to Remove Cytokines, Chemokines, and Cell-derived Microvesicles: A Clinical Feasibility Study

This was a planned substudy of the parent clinical trial, completed before the end of subject enrollment, that addressed the specific question of the feasibility and effectiveness of bedside erythrocyte washing in removing potentially injurious soluble biologic response modifiers from allogeneic erythrocyte units. This is distinct from the comparison of patient-level biomarkers and recipient clinical outcomes, powering the parent clinical trial, that we plan to report in a separate manuscript once the data and laboratory analyses are complete.

There were in vitro and prospective, observation cohort components. First, laboratory data were collected from the first 75 washed units given to a subset of patients in the intervention arm of a clinical trial evaluating bedside erythrocyte washing, with a cell saver, during cardiac surgery (NCT02094118).¹¹  Paired pre- and postwash samples from these blood unit bags were centrifuged. The supernatant was aspirated and frozen at –70°C, then subsequently batch-tested for cell-derived microvesicles, soluble CD40 ligand, chemokine ligand 5, or regulated on activation, normal T expressed and secreted (often referred to as CCL5/RANTES) and neutral lipids (previously associated with transfusion reactions) and cell-free hemoglobin (possibly increased by washing). Second, from the entire cohort randomized to the washing arm of the trial, bedside washing was defined as feasible if at least 75% of prescribed units were able to be washed per protocol. Institutional review board approval for this study protocol was obtained from both participating institutions before enrolling the first study subject in the parent trial.

All adult patients scheduled for cardiac surgery via sternotomy at one of two tertiary referral, academic heart centers (Duke University Medical Center, Durham, North Carolina; or Mayo Clinic, Rochester, Minnesota) between May 2015 and June 2019 were screened for study eligibility. Those projected to require four or more units of allogeneic erythrocytes, based on a published scoring system (supplemented by the opinion of experienced cardiac anesthesiologists, I.J.W. and W.J.M.), were further assessed for trial eligibility.¹¹  Those meeting inclusion criteria and lacking all exclusions were approached for consent to participate. Exclusion criteria were emergency surgery, history of severe recurrent transfusion reaction, refusal to receive allogeneic erythrocytes or provide informed consent, prevalent acute lung injury or congestive heart failure before randomization, expected hospital stay less than 48 h, not anticipated to survive more than 48 h, or a reoperation in a patient previously enrolled in this trial. After written informed consent, patients were randomized to either the intervention arm or the control arm for the entire day of surgery. Our laboratory and feasibility data all came from the intervention arm. A Data Safety Monitoring Board (Englewood, New Jersey) met every 6 months to determine the suitability of continuing the study for reasons of safety or futility.

For transfusions deemed necessary by the clinical team, units were handed to the study team from the four erythrocyte units typically provided in a cooler, under ice, for all cardiac cases. These were washed on-demand using a continuous automated transfusion system (CATS, Fresenius Kabi USA LLC, USA) with the Food and Drug Administration (Silver Spring, Maryland)–approved 0.9% saline wash solution, after a 4:1 dilution with 0.9% saline in the cell saver reservoir, as previously described.¹²  Predilution was included to reduce the hematocrit of packed erythrocytes to better approximate that of the salvaged blood mixed with heparin anticoagulant that cell savers were designed to process. This automated washing procedure has been shown to more efficiently remove contaminants from the supernatant.^12,13  This cell saver system was chosen as the g force applied to the cells is less for this apheresis belt system (approximately 800g) than for the Latham bowl design favored by most cell saver devices (approximately 2,000g),¹⁴  and has been reported to induce less hemolysis when compared to a standard blood bank–based cell washing device (Cobe 2991, Terumo BCT, USA).¹⁵  Staff trained to use this cell saver included research personnel, anesthesia technicians and perfusionists (Duke University Medical Center), and transfusion medicine personnel with specific expertise in cellular washing techniques (Mayo Clinic).

From the group randomized to receive allogeneic units washed with the cell saver, 6-ml samples were taken from each unit of blood before and after the washing procedures. Samples were acquired from the first 75 washed units (no more than four units from the same individual were included for the logistics of sample processing), as defined a priori in the study protocol.¹¹  These paired samples contained erythrocytes and supernatant (plasma, additive solution, and anticoagulant prewash, and mostly saline postwash). Samples were stored on ice for less than 2 h and centrifuged at 2,000g in refrigerated centrifuges with supernatant removed for a second 2,000g centrifugation before dividing into aliquots for frozen storage (–70°C). Subsequent batch analysis tested the removal of the following biologic response modifiers: chemokine ligand 5, a T cell–derived chemokine that recruits leukocytes (other than neutrophils) to sites of inflammation; soluble CD40 ligand, a platelet and T cell–derived protein that has proaggregatory effects on platelets¹⁶ ; the neutrophil priming activity of neutral lipids; and cell-derived microparticles present in the unit. The mediators have been shown to accumulate in stored blood products and have previously been associated with lung injury.^10,17–22 

Assays performed included Luminex Assays (USA) and flow cytometry (Professor Philip Norris, Vitalant Research Institute, San Francisco, California). Flow cytometry characterized the number and cellular origin of microparticles in pre- and postwash samples, with the following pairing: CD62p (P-selectin), activated endothelial cell– or platelet-derived; CD 235a (glycophorin A), erythrocyte-derived; CD108 (semaphorin-7A), activated lymphocytes or erythrocytes; and CD41a (glycoprotein IIb), megakaryocyte/platelet-derived. Due to the potential for inducing hemolysis with manipulation of the units,²³  cell-free hemoglobin was also measured in pre- and postwash samples by ELISA (Bethyl Laboratories, Inc., USA). Specific lipid priming assays were performed as previously described.^24,25  All measurements were performed in triplicate.

Due to concerns about excess sample volume being taken from the 75 patient units, additional data required to calculate percent hemolysis (pre- and postwash hematocrits, product and supernatant volumes and measures of the total amount of cell-free hemoglobin present in the supernatants) were only obtained in a post hoc quality control study using five test units that had been stored for 18 days. For percent hemolysis calculations, cell-free hemoglobin was measured by a point-of-care analyzer (Hemocue Plasma/Low Hb system, HemoCue America, USA) after immediate point-of-care processing to account for the potential for the storage on ice and the processing necessary for microparticle separation inducing hemolysis in the sample from the bags (especially compounding the stress of blood washing). The ratio of the amounts of cell-free hemoglobin to total units hemoglobin (in milligrams) was used to derive percent hemolysis.

Provision was made for the clinical teams to administer “off-protocol,” standard-issue units should the bedside washing process be expected to or be incurring an unacceptable delay based on the clinical situation. Washing erythrocyte units according to protocol determined feasibility; if an individual patient received 75% of prescribed units on protocol, then bedside washing for that patient was considered feasible.

A power and sample size calculation for the overall “Point-of-Care Washing of Allogeneic Red Blood Cells for the Prevention of Transfusion-related Respiratory Complications” (WAR-PRC) trial was reported previously,¹¹  which was based on biomarker outcomes in patients randomized to control/intervention groups. A power calculation was not performed for this specific aim a priori, but a sample size of 75 washed bags with pre- and postwash samples was thought to be feasible and within budget, based on the clearance of mediators in preliminary data. A total 75 bags with pre- and postwash paired samples provides 90% power to detect an effect size (location shift divided by the SD of the paired differences) of 0.39 using a Wilcoxon signed-rank test with a two-sided alpha level of 0.05.

For the primary prespecified analysis, pre- and postwash bag samples were compared for each of the primary outcomes using a paired Wilcoxon signed-rank test. Change in hematocrit was analyzed similarly. In secondary, post hoc analyses, we assess the association between erythrocyte storage additive solutions and outcomes. We calculated estimates that reflect the estimated multiplicative increase in geometric mean of each mediator, microvesicle subtype, and total count, as associated with a 1-day increase in storage duration. We also assessed whether there was a nonlinear relationship between storage duration and outcomes. Prewash outcomes are compared across additive solutions using Kruskal–Wallis tests. Postwash outcomes are analyzed using linear regression, adjusted for prewash values. Log transformations were applied to satisfy assumptions for normality of residuals and results transformed to report the association between additive solutions and multiplicative change in the geometric mean outcome. The association between storage age and outcomes was assessed similarly.

Applying Bonferroni correction to our nine total a priori outcomes analyzed would specify P < 0.0055 for significance. Post hoc analyses are identified as such. Statistical tests were two-tailed, and data were analyzed in using R software (version 3.6.1; http://archive.linux.duke.edu/cran/; accessed January 15, 2021).

Of 314 erythrocyte transfusions prescribed by the clinical team on the operative day among 81 patients in the washed arm, we were able to complete the bedside washing procedures per protocol on 293 units (93%). Feasibility was defined at the patient level, with patients considered feasible if at least 75% of prescribed units were washed per protocol; all transfused patients were evaluated for the feasibility of providing units washed at the bedside with a delay acceptable to the clinical team. By this definition, of 81 patients receiving a transfusion on the operative day, 80 (99%) were considered feasible. A total of 21 nonwashed units were provided off-protocol to 11 patients randomized to the washing arm of the trial. The time from decision to transfuse to start of transfusion (median [25th to 75th] percentiles) for all units transfused on the day of surgery was 5 (3 to 12) min for the standard and 31 (23 to 50) min for the washed arms. This suggests the washing process approximately incurs a 25-min delay in transfusion initiation, which was acceptable to the clinical team for 99% of patients.

Complete data from paired pre- and postwash samples, with sufficient volume to test all proposed biologic response modifiers and cell-free hemoglobin, were available from 74 of the 75 units tested (39 from the Duke University Medical Center site and 35 from Mayo Clinic). The distribution of storage ages of transfused units (standard blood bank issue) is illustrated in figure 1 with a median storage duration (median [25th to 75th] percentile) of 15.5 (12 to 23) days (maximum, 41 days); the majority of units (65%) had been stored for 20 days or less. While storage age duration was unrelated to any biomarker or cell-free hemoglobin concentrations, there were additive solution–dependent differences. Most units (n = 41, 55%) were stored in additive solution-3, with the remainder in the mannitol-containing additive solution-1 (n = 24, 32%) or additive solution-5 (n = 8, 11%); there was one citrate/phosphate/dextrose/adenine-1 unit (not analyzed). All units from the Mayo Clinic were stored in additive solution-3. After adjusting for prewash cell-free hemoglobin, there was still evidence of postwash differences by additive solution group (overall P = 0.021). Specifically, in post hoc pairwise comparisons, there is evidence that for postwash, additive solution-3 was associated with 53% higher cell-free hemoglobin compared to additive solution-1 units (estimated ratio [95% CI], 1.53 [1.06 to 2.22], P = 0.027) and 91% higher concentrations compared to additive solution-5 units (1.91 [1.11 to 3.28], P = 0.022). There were also significantly more platelet derived microvesicles postwash in the additive solution-3 units (11.19 [2.8 to 44.67], P = 0.001 and 2.08 [1.34 to 3.25], P = 0.002 respectively) and overall more microparticles in the additive solution-3 units (1.66 [1.03 to 2.69], P = 0.043) compared to additive solution-1 units.

Combining the data from both sites, we identified a wide range of concentrations of soluble mediators in the individual units before washing varying over 1,000-fold for chemokine ligand 5, over 10,000-fold for soluble CD40 ligand, and almost 100-fold for total microvesicle counts between units. These data are detailed in table 1. As shown in figures 2 and 3, there were significant reductions in the amount of each marker after washing with a clearance of 93%, 76%, and 86% for soluble CD40 ligand, chemokine ligand 5, and total microvesicles, respectively (significant P value after Bonferroni correction < 0.0055). Only 1% of the median counts of microvesicles were identified by a cell-specific marker, but microvesicles from all cell lines were significantly reduced (P < 0.004 for all), as shown in table 1. In terms of neutral lipids, there was no difference in the neutrophil priming ability of the erythrocyte supernatants before or after washing (P = 0.694).

From a safety standpoint, there was an overall increase in median cell-free hemoglobin concentration in the supernatant from approximately 70 mg/dl to approximately 210 mg/dl as measured by ELISA after two 20-min, 2,000g centrifugations. In a post hoc study of purchased additive solution-3 units (n = 20), we measured a median (IQR) increase in hematocrit from 62 (58 to 64) to 66 (60 to 70) percent (P = 0.004) after the washing procedure, and an expected decrease in product volume from 330 ml to 262 ml (n = 5 test units, not compared); after a 800g centrifugation to separate cells and supernatant, the average dose of cell-free hemoglobin from the supernatant of these five test units was less postwash than prewash (85 mg vs. 148 mg), mostly due to hemoconcentration. The percent hemolysis of these five test units ranged from 0.1 to 0.3% prewash to 0.1 to 0.2% postwash, all less than the acceptable upper limit of 0.8%. The percent hemolysis was not calculated for all 75 washed patient units, as hematocrit and volume pre- and postwash were not able to be measured with the time pressure of the clinical workflow.

The reported data support the feasibility of bedside washing of allogeneic erythrocyte units in the setting of elective major cardiac surgery both in the operating room and in the early intensive care unit course, with nearly all of the units assigned to the washed group being washed. Additionally, bedside washing of prediluted, allogeneic erythrocyte units was able to eliminate the vast majority of cytokines, chemokines, and microvesicles. This study suggests bedside washing with this cell saver is feasible, is safe, and greatly reduces the concentrations of potentially injurious factors in the erythrocyte supernatant.

Transfusion-related hyperkalemic cardiac arrest has previously been described in adult and pediatric patients,²⁶  and bedside washing has successfully reduced hyperkalemia in small case series²⁷  (useful especially in pediatric cases, recipients with renal failure, or massive transfusions). Washing cellular blood products in pediatric cardiac surgery has also been shown to reduce lactate concentrations²⁸  and proinflammatory markers²⁹  in transfusion recipients. The removal of cytokines, chemokines, and microvesicles that have previously been associated with transfusion-associated respiratory complications supports the hypothesis that bedside erythrocyte washing may provide a means for abrogating the effect of injurious mediators in the supernatant, especially in at-risk patients. In a study of adverse reactions to platelet transfusions⁴  (the product most commonly associated with reactions), Cognasse et al. identified that most patients with an adverse reaction had received a higher concentration of soluble CD40 ligand, and the “pathogenic threshold” they calculated was 6.4 ng/ml. In contrast, the concentrations in our units before washing (albeit using a different assay platform) were higher, with median (IQR) 143 (1 to 338) ng/ml prewash and 0 (0 to 0) ng/ml postwash. Based on this, we believe our washing procedure is capable of removing enough of this cytokine to possibly be able to reduce transfusion reactions with the potential to improve clinical outcomes in a larger, suitably powered study. By effectively removing the vast majority of protein,^14,30  bedside washing offers the possibility of also reducing antibody-mediated transfusion-related acute lung injury. We did not specifically test for immunoglobulins, though, so removal of these antibodies remains speculative.

The elimination of neutral lipids was ineffective with our protocol, which was surprising, as this has previously been demonstrated, albeit with a blood bank washing protocol and using blood stored for 42 days.⁸  Whether residual lipids remained after washing the stored units or were released as a result of hemolysis induced by washing is unknown. Regardless, the remaining lipids were capable of neutrophil priming and, therefore, theoretically capable of inducing lung injury or activating endothelium and predisposing to later injury.¹⁰ 

As has been previously reported,³¹  erythrocyte washing incurred a threefold increase in cell-free hemoglobin concentration in the reduced-volume supernatant, although, when calculated, percent hemolysis remained within the acceptable range (less than 0.8%).³²  Our findings regarding additive solutions were unexpected, although the mannitol in additive solution-1 may promote membrane stability³³  and, ideally, reduce hemolysis. They could also reflect variations in erythrocyte processing techniques or interdonor variability associated with the blood centers that favor additive solution-3. As mentioned, the extent of hemolysis, in the small subset we tested, was well below the acceptable level of 0.8%.³²  Similar findings were reported in the “Randomized Trial of Red Blood Cell Washing for the Prevention of Transfusion-associated Organ Injury in Cardiac Surgery” (REDWASH) study in that cell-free hemoglobin in the supernatant of washed units was higher than controls, but there was no increase detected after transfusion in vivo, likely due to the avid scavenging of free hemoglobin by haptoglobin in the plasma.³⁴  There are also some potentially important differences between the REDWASH study and our trial: we prediluted erythrocyte units containing additive solutions-1, -3, and -5 with saline, whereas they used undiluted units with a different additive solution. We also used a more rigorous washing setting on the same cell saver device. The physiologic relevance of the balance between increased cell-free hemoglobin and removal of nitric oxide scavenging microvesicles³⁵  will await future comparisons of the clinical outcomes between the washed versus standard cohorts in the parent study, although there is no reason to expect this to differ from REDWASH. Similarly, lung injury biomarkers were not included in this manuscript, as measuring the recipients of these 75 units only from the intervention arm of the parent trial was not powered or designed to detect any difference. The results of the clinical biomarkers and outcomes associated with the parent clinical trial have not yet been analyzed or published.

Regarding the microvesicle evaluations, their characterization revealed an identifiable cell lineage in only approximately 1% of detected signals, consistent with previous publications from our group and others.^36,37  Whether they lost the antigens we targeted during storage, the level of antigen per microvesicle was not high enough to detect by flow cytometry, or we should have used a broader panel to identify relevant antigens is not clear. Alternatively, it is possible that the majority of microvesicles detected represent cell debris or vesicles derived from internal multivesicular bodies rather than vesicles that bear cell surface markers, as has been shown previously.³⁸  Of note, whether the physiologic effect of cell debris versus specific cell-derived encapsulated vesicles is equivalent or not is unknown. However, total microvesicle content derived from erythrocyte units has been shown to support prothrombinase activity and thrombin generation,^39,40  so all may be biologically relevant.⁴¹  Regardless, there were substantial and significant decreases in microvesicle counts for those with identifiable cell lineage as well as for those without.

Our study has important limitations. First, our exclusive use of an apheresis belt system on a high-quality wash program setting precludes generalization of our study findings to other cell washing devices. Regarding the more common Latham bowl design, certain models have been used successfully to wash erythrocyte units,⁴²  while others appear to induce excessive hemolysis.⁴³  Second, we were limited to the Food and Drug Administration–approved saline washing solution, whereas the balanced solution Plasmalyte-A (Baxter, USA), appears to be associated with less hemolysis than saline, which has an unphysiologically high chloride level of 154 mM and an acidic pH of 5.5.⁴⁴  Future bedside washing protocols to be evaluated include simplifying the workflow by dilution with 1,000 (approximately 1:3) rather than the current 1,200 ml (1:4), using other cell saver devices, optimizing erythrocyte by diluting and washing with Plasmalyte A or other balanced, neutral solutions or even reversing aspects of the erythrocyte storage lesion by replenishing 2,3-diphosphoglycerate or removing toxic phthalates.⁴⁵  While our median postwash cell-free hemoglobin value of 210 mg/dl is similar to a mean postwash value of 212 mg/dl previously reported using a spectrophotometric technique,⁴³  the extreme values seen with our ELISA platform that required serial dilutions for most of our samples present the possibility of error. We only systematically calculated percent hemolysis in a small subset of washed units, but this would be an essential component of a future quality control process, and using spectrophotometric techniques such as that used to measure hemoglobin in our quality control subset may offer a simpler solution. Our reported median value of 210 mg/dl would result in an approximately 3 to 5 mg/dl increase in cell-free hemoglobin (normal value less than approximately 15 mg/dl), assuming a 5,000 ml circulating volume and hematocrit of 0.25, but only if the natural hemoglobin scavenger, haptoglobin, were absent. While an increase in cell-free hemoglobin above the normal limit is not expected, confirmation of the safety of this bedside washing procedure will require sampling recipients’ plasma.

A third limitation relates to the external validity of our feasibility findings. The two enrolling institutions leveraged research personnel, anesthesia technicians, perfusionists, and transfusion medicine specialists to oversee the bedside washing procedures for the purposes of the study. However, study procedures can be simplified and plausibly performed by existing personnel in the operating room. Allogeneic units can be washed in the cell saver being routinely used, and the only additional workflow would involve gravity draining the saline and erythrocyte unit into the reservoir of the cell saver using a standard “Y-spike” set. Further processing is automated, and the staff perfusionist (assisted by anesthesia personnel during bypass) could realistically perform this task. This scenario is limited to the operating room environment and our data are limited to the CATS, as we did not test other Latham bowl cell saver designs. Extending the process into the intensive care unit would be more complex and likely to require additional staffing with variable, local, economic implications.

Finally, we did not correlate the current study results with clinical outcomes, as this study was not designed or powered to do so. To be clear, this report is limited to the assessment of our ability to wash allogeneic erythrocyte units in the setting of elective cardiac surgery and to quantify the removal of soluble mediators from the allogeneic erythrocyte units. It is an a priori planned substudy of a parent clinical trial that has completed enrollment, but is pending recipient outcome assessments, data analysis, and publication. While this is an important first step in understanding the potential role of bedside allogeneic erythrocyte washing, it is insufficient evidence to support bedside washing for the prevention of transfusion-related respiratory complications. The primary aims of the parent trial, from which the current work arises, will address this question.¹¹ 

In summary, this study demonstrates that bedside washing of allogeneic erythrocytes with a familiar cell saver technology is feasible and effective in removing soluble mediators immediately before transfusion. In acute care environments such as the operative setting, where autologous cell salvage is frequently employed, this technology may offer a practical approach to removing mediators associated with adverse transfusion reactions.

The authors thank Lavonne Liedl, R.R.T., Study Coordinator, Beth Montonye, Research R.N., Clinical Specialist, Brenda Anderson, Research R.N. Study Coordinator, and Midhat Mujic, R.R.T., Study Coordinator (Department of Anesthesiology, Clinical Research Unit, Mayo Clinic, Rochester, Minnesota) for their expertise in patient recruitment, database management, quality control, and maintenance of data integrity with this study.

All authors were supported by the National Institutes of Health (Bethesda, Maryland; NIH R01HL121232; principal investigator, Dr. Kor; co-principal investigator, Dr. Welsby).

Dr. Welsby previously received material support from Fresenius Kabi (Bad Homburg, Germany; manufacturer of CATS cell saver) to generate preliminary data. Authors have indicated the following financial relationships: Dr. Welsby, CSL Behring (King of Prussia, Pennsylvania) and Biomet Biologics, LLC (Warsaw, Indiana), UpToDate (Waltham, Massachusetts); R. Silliman, Department of Defense (Washington, D.C.); and Dr. Kor, UpToDate.