Hemoperfusion with Seraph 100 Microbind Affinity Blood Filter Unlikely to Require Increased Antibiotic Dosing: A Simulations Study Using a Pharmacokinetic/Pharmacodynamic Approach

Sepsis is a leading cause of morbidity and mortality in critically ill patients [1]. Antibiotics are essential to the treatment of sepsis; however, multidrug-resistant organisms present a growing challenge to effective management [2‒4]. Therefore, modalities other than antibiotics are increasingly sought [5]. Extracorporeal blood purification is one such novel therapeutic that attempts to directly remove offending pathogens from the central circulation [6, 7]. Several hemoperfusion technologies with distinct binding profiles are currently available for clinical use in the management of sepsis. The Seraph® 100 Microbind® Affinity Blood Filter (Seraph 100) (ExThera Medical, Martinez, CA, USA) is such a hemoperfusion device that received emergency use authorization from the US Food and Drug Administration for SARS-CoV-2 and is currently being explored as an adjunct therapy to treat bacterial sepsis in critically ill patients [8, 9]. The device is comprised of polyethylene microbeads with end-point attached heparin, which interacts with pathogens similar to heparin sulfate found on the surface of cells in order to irreversibly bind and remove them from the bloodstream [10]. Seraph 100 can be configured in sequence with a dialyzer or as a stand-alone adsorption column. Seraph 100 is typically run for 4–8 h and the overall duration of therapy is left to the discretion of the prescribing physician [8].

Although Seraph 100 is a promising novel device for the removal of pathogens and treatment of sepsis, effective antibiotic therapy remains essential. There are scant clinical data providing estimates of antibiotic clearance due to the Seraph 100 device; however, an in vitro study has demonstrated Seraph 100 antibiotic clearance rates to be generally less than 0.5 L/h, regardless of antimicrobial drug class [11]. These data are supported by a case report demonstrating negligible vancomycin clearance due to Seraph 100 (0.05 L/h) [12]. Minimal Seraph 100 device clearance for remdesivir is supported by a recent PK study demonstrating median Seraph 100 clearance of 0.07 L/h [13]. However, for vancomycin and remdesivir, clinical Seraph 100 clearance estimates are based off of data from only 1 patient each. In particular for vancomycin, there were only two pre- and post-Seraph 100 device blood draws to estimate vancomycin device clearance. Therefore, given the low sample size and high residual unexplained variability observed for antibiotic concentrations in a hospital setting [14, 15], the precision of these clinically derived Seraph 100 clearance estimates is unclear. As a result, there is a possibility that Seraph 100 antibiotic clearance could be underestimated.

Given this possibility, to derive the most clinically useful information and in accordance with research recommendations from Kidney Disease: Improving Global Outcomes (KDIGO) [16], estimates of Seraph 100 antibiotic clearance should be taken in context of achievement of standard pharmacodynamic (PD) targets and probability of target attainment (PTA). The three most commonly utilized PD indices for antibiotics to determine target attainment are: (1) the percent time within a dosing interval the free drug concentration exceeds minimum inhibitory concentration (%fT>MIC), (2) the ratio of area under the curve to MIC (AUC:MIC), and (3) the ratio of maximum concentrations to MIC (C_max:MIC) [17]. We sought to better understand that the potential clinical impact Seraph 100 may have on antibiotic dosing requirements. We provide examples of drugs for each PD target. For vancomycin (PD target AUC:MIC >400) [18] and gentamicin (primary PD target C_max:MIC ≥8–10, secondary PD target AUC:MIC between 70 and 120) [17, 19, 20], we present PTA analyses with theoretical Seraph 100 device clearances ranging up to 2 L/h.

For meropenem and imipenem (PD targets fT>MIC) [17], we leverage PK data from continuous kidney replacement therapy (CKRT) studies and extrapolate dosing to the utilization of the Seraph 100. The mechanism of antibiotic removal from CKRT is convection and/or diffusion [21], in contrast to that of Seraph 100 which is adsorption. Typical rates of meropenem and imipenem clearance due to CKRT are 1–4 L/h [22‒27], where in vitro data suggest that meropenem seraph clearance is less than 0.5 L/h [11]. Although the observed device antibiotic clearances are consistent with their respective mechanisms, our extrapolation of effect on PTA is mechanism agnostic and solely relies on the empiric observations of antibiotic device clearances.

Only data available to the public from the National Library of Medicine Database were used in this study. Therefore, our work was exempt from Institutional Review Board review (Walter Reed Army Institute of Research Human Subjects Protection Board Policy #2). The term “Seraph 100” was utilized on November 16, 2021 to search the National Library of Medicine Database yielding 12 results. Of the 12 results, 2 manuscripts contained estimates of antibiotic clearance due to Seraph 100. One study was a case report providing clinical estimates of Seraph 100 vancomycin clearance [12], and the other study was an in vitro exploration of Seraph 100 clearance for many antibiotics [11]. Seraph 100 antibiotic clearance was estimated as the average of all observed Seraph 100 clearances for a given antibiotic in the respective studies. Of note, Seraph 100 in the in vitro study was primed with a filling volume of 200 mL 0.9% saline, which caused an expected decrease in postdevice antibiotic concentrations within the first 5 min. Therefore, as these initial antibiotic concentration decreases were reflective of dilution and not Seraph 100 adsorption, they were excluded from the calculation of Seraph 100 antibiotic clearance. When estimates of Seraph 100 antibiotic clearance were less than 0, 0 was imputed.

PTA simulations were not performed for meropenem or imipenem as there is adequate literature presenting PTA analysis for these antibiotics in the setting CKRT [22‒26]. These previous analyses were used to inform dosing for meropenem and imipenem while utilizing Seraph 100.

Simulations for vancomycin and gentamicin were performed in Pumas v1.1 [28]. Summary statistics and figures were generated using either Pumas v1.1 or R v4.1 with R Studio v1.4.1717. For vancomycin, AUCs from 0 to 24 h (AUC_0–24) were simulated with the formula:

where (DOSE_0–24) is the total cumulative 24-h dose and CL is the patient’s simulated inherent antibiotic clearance. For vancomycin, simulated clearances were derived from Goti et al.’s [14] PK model, which has been externally validated in a critical care population [29]. The creatinine clearance (CrCl) covariate formula from Goti et al. [14] to derive vancomycin CL in patients with renal impairment is as follows:

The simulated CLs in kidney impairment groups were based off assumed CrCL of 14 mL/min, 49 mL/min, and 89 mL/min as further explained in the discussion. Simulations did not account for hemodialysis, which was a covariate utilized in Goti et al.’s [14] model. However, reasonable extrapolations can be made from the nonhemodialysis population as explained in the discussion. Vancomycin PTA (10,000 patients per group) was performed by calculating the percentage of patients meeting the target

[18]. Theoretical Seraph 100 clearances were added to the virtual patients’ inherent vancomycin clearances. The Seraph 100 device was assumed to run for 8 h, and therefore, the theoretical Seraph 100 clearance was divided by 3 to get the average Seraph 100 clearance within a 24-h time period. Conditions for vancomycin simulations are summarized in the legend of Figure 1a and in Table 1.

For gentamicin, simulated time-concentration curves were generated using Xuan et al.’s [15] PK model. The CrCl covariate formula to derive gentamicin CL in Xuan et al. [15] is as follows:

CL = 0.047 × CrCl

Gentamicin PTA simulations (10,000 patients per group) were performed calculating the percentage of patients meeting target C_max:MIC ≥10 [20]. The Seraph 100 device was assumed to be initiated at the time of infusion with no mathematical adjustments to the theoretical Seraph 100 clearance. Conditions for gentamicin simulations are summarized in the legend of Figure 1b and in Table 2.

The mean in vitro estimate for vancomycin clearance due to Seraph 100 was 0.2 L/h. However, two of these clearance estimates were less than 0, likely due to residual unexplained variability. In contrast, the clinical case report had a mean estimate of 0.05 L/h for Seraph 100 vancomycin clearance. The overall mean vancomycin Seraph 100 clearance from these two studies (in vitro and case report) was 0.13 L/h. The mean in vitro Seraph 100 gentamycin clearance was 0.37 L/h; however, 3 of the 4 clearance estimates were less than 0. Similarly, the mean in vitro estimate of Seraph 100 meropenem clearance was 0.36 L/h, where 2 of the 4 clearance estimates were less than 0.

Figure 1a demonstrates for vancomycin, on average, that a theoretical Seraph 100 clearance up to 4 L/h would on average have a clinically insignificant influence on steady-state AUC₀₋₂₄. The exception is patients with CrCl <15 mL/min (either from chronic kidney disease or acute kidney injury) where theoretical Seraph 100 device clearances greater than 1 L/h would have a clinically significant effect on vancomycin AUC₀₋₂₄. Similarly for vancomycin, Table 1 demonstrates that theoretical Seraph 100 device clearances up to 2 L/h have a minimal effect on PTA except for patients with CrCl <15 mL/min where theoretical Seraph 100 device clearances were associated with PTA <70%.

Figure 1b demonstrates for gentamicin, on average, that theoretical Seraph 100 clearance up to 2 L/h has minimal effects on gentamicin C_max. Similarly, PTA (Table 2) demonstrates that Seraph 100 clearance up to 2 L/h has a clinically insignificant effect on achieving C_max:MIC target in patients with CrCl <60 mL/min. However, patients with CrCl = 120 mL/min and theoretical Seraph CL ≥0.5 L/h would have a clinically significant effect on achieving C_max:MIC targets.

Multiple analyses including a systematic review have demonstrated that utilization of CKRT with meropenem or imipenem CKRT clearances of 1–4 L/h would have a little effect on ability to meet a given %fT> MIC target [23‒26]. Therefore, it follows for Seraph 100 that is only run for 4–8 h a day and likely has a meropenem clearance of less than 0.5 L/h, and no dose adjustments would be necessary.

We have compiled current estimates of antibiotic clearance due to Seraph 100 and performed PTA analysis for vancomycin and gentamicin with a wide range of theoretical Seraph 100 clearances. PTA was largely unaffected for patients with theoretical Seraph 100 vancomycin and gentamicin clearances as high as 2 L/h. Importantly, these theoretical clearances are significantly higher than the mean of reported Seraph 100 vancomycin and gentamicin clearance estimates (0.13 L/h and 0.37 L/h), which suggests that on a population level, the use of Seraph 100 would have a minimal effect on the ability to achieve vancomycin or gentamicin PD targets and no standard dose increase would be necessary. Furthermore, standard of care for vancomycin and gentamicin is to perform therapeutic drug monitoring (TDM), and therefore, TDM-based tailored dosing to individuals would account for possible variability in Seraph 100 clearance.

A systematic exploration of the effect of different modalities of kidney replacement therapy in conjunction with Seraph 100 was beyond the scope of this research. As such, our simulations did not account for kidney replacement therapies. Nevertheless, reasonable extrapolations can be made to patients receiving vancomycin that are treated with both Seraph 100 and intermittent hemodialysis. Vancomycin-dosing recommendations for patients with kidney impairment are the same throughout ranges of CrCl (i.e., 50–90 mL/min, 15–50 mL/min, and <15 mL/min) [30]. Given

for a fixed dose, the smaller the CL, the higher the (AUC_0–24) and greater likelihood of achieving PTA. Therefore, we chose the highest value of CrCl for simulations from each respective kidney impairment group in order to generate the largest mean vancomycin CL for the respective group. It follows that simulations with lower values of CrCls in the respective groups at fixed doses would lead to lower vancomycin CLs, higher AUCs, and higher PTA success rates. Similarly, Goti et al. [14] reported that patients receiving hemodialysis had on average 70% of vancomycin CL as compared to nonhemodialysis patients at a given CrCl. Intermittent hemodialysis dose recommendations are similar to those for patients with CrCl <15 mL/min [18, 30]. Hence per Goti et al. [14], for a fixed CrCl vancomycin CL would on average be 70% less with resultant 30% increase in AUC and higher PTA success rates. Therefore, it is reasonable to extrapolate, as long as doses and CrCl are similar, patients receiving intermittent hemodialysis also would not require vancomycin dose adjustments based on the use of Seraph 100. However as kidney replacement therapy may be offered via many modalities, continuously or intermittently, further research via simulations or clinical PK data collection in such populations with concomitant Seraph 100 therapy may be warranted.

Regarding gentamicin, the primary PD target is C_max:MIC; however, an important secondary, but less established, PD target is AUC:MIC [17, 31]. Similar to vancomycin, on average theoretical Seraph 100 gentamicin clearance would not have a clinically significant effect on gentamicin AUC at clinically relevant targets (work not shown). Therefore, if a specific AUC target is desired for an aminoglycoside, that AUC target should drive the choice for a higher or lower dose rather than the use of Seraph 100. Of note, in patients with CrCl = 120 mL/min, gentamicin Seraph 100 clearances ≥0.5 L/h may significantly reduce PTA. Rather than empirically increasing gentamicin dose for this patient population, we suggest a high-dose extended-interval dosing approach and using the Seraph 100 device 30 min – 1 h after end of infusion. High-dose extended-interval once-daily dosing of aminoglycosides is a well-established approach and has shown similar efficacy and possibly less toxicity compared to traditional intermittent dosing [15, 32‒34]. As the Seraph 100 device is typically used for only 4–8 h each 24-h period, the device could simply be initiated 30 min to 1 h after completion of the gentamicin infusion, thereby avoiding any possible interference with C_max. Of note, this approach has precedence with high theoretical PTA achievement rates for dosing gentamycin in patients requiring prolonged intermittent kidney replacement therapy [35]. Although we did not perform PTA analyses for other aminoglycosides, in principle high-dose extended-interval dosing obviates the need for PTA analysis or dose adjustments for any aminoglycoside as there would be no possibility of C_max interference.

The main limitation to this study was heavy reliance on in vitro estimates for antibiotic Seraph 100 clearances. A given antibiotic’s physiochemical properties and its interaction with Seraph 100 should generally be consistent in vitro or in vivo, allowing for reasonable extrapolation. Regarding hypoalbuminemia, albumin has not been shown to be cleared by Seraph 100. Therefore, although hypoalbuminemia has the potential to alter antibiotic CL, this would be as a result of critical illness rather than the use of Seraph 100. However, one consideration in the critical care setting would be changes in antibiotic ionization due to severe acidosis or alkalosis, potentially making an antibiotic more susceptible to heparin binding [36]. Nevertheless, the clinical impact of this process is unclear and plasma pH could be quickly corrected if necessary [37, 38]. Further, it is standard practice for vancomycin and gentamicin (the most likely antibiotics in this study to be affected by Seraph 100), to perform TDM, which would provide individualized dosing if acid-base disorders were to play a significant role in Seraph 100 antibiotic clearance. Therefore at this time, there is not enough evidence to recommend a standard dose change based on this concern and the in vitro estimates are reasonable.

A final consideration is the dilution effect observed in vitro, where antibiotic concentrations in the first 5 min of the study were reduced by approximately 20%. This was expected as 1,000 mL of plasma containing antibiotics were run through the device including 200 mL of priming solution [11]. However, a 70 kg human has on average 14 L of extracellular fluid [39] and often antibiotics have even higher volumes of distribution (Vd) such as vancomycin, which may have a Vd as high as 105 L in a 70 kg critically ill patient [40]. Therefore, expected dilutions as a result of 200 mL priming solution would be negligible. Furthermore, to our knowledge, there is no standard recommendation to increase doses based off the concern of dilution with priming solution for any other extracorporeal clearance device, notably even for kidney replacement therapies [16]. Accordingly, we did not assume any dilution effect in our simulations and do not recommend standard dose increases of antibiotics based off a concern for dilution in clinical practice.

Vancomycin, gentamicin, meropenem, and imipenem clearances due to Seraph 100 are likely clinically insignificant. There is currently insufficient evidence to recommend standard antibiotic dose increases. For aminoglycosides, we recommend an extended interval dosing approach and only initiating Seraph 100 30 min–1 h after completion of infusion to avoid the possibility of interference with maximum concentrations.

We would like to thank Ms. Zanete Wright for her hard work and continued support of the WRAIR Clinical Pharmacology Fellowship.

Only data available to the public from the National Library of Medicine Database were used in this study. Therefore, our work was exempt from the Institutional Review Board review (Walter Reed Army Institute of Research Human Subjects Protection Board Policy #2). Informed consent was not required.

The authors have no conflicts of interest to report. This study only included the data previously published and publically available in the National Library of Medicine Database. Therefore, neither informed consent nor a trial identification number was applicable to this research. Material has been reviewed by the Walter Reed Army Institute of Research and the Uniformed Services University of the Health Sciences. There is no objection to its presentation and/or publication. The opinions and assertions expressed in this article are those of the authors and do not reflect the official policy or position of the Uniformed Services University of the Health Sciences, US Army Medical Department, Department of the Army, DoD, or the U.S. Government.

No funds were received for this research. WRAIR/USUHS Clinical Pharmacology Fellowship funds were used to cover publication expenses.

Tyler Reed performed literature review, acquired the data, and drafted the manuscript. Daniel Selig performed simulations and data analysis and drafted the manuscript. Jesse DeLuca helped conceptualize and design the study, and critically revised the manuscript. Adrian Kress helped conceptualize and design the study and critically revised the manuscript. Kevin Chung is a subject matter expert in critical care, helped interpret the data, and critically revised the manuscript. Ian Stewart is a subject matter expert in hemoperfusion, helped interpret the data, and critically revised the manuscript.

All data were extracted from the publically available literature in the National Library of Medicine Database. The search strategy used is disclosed in the methods of the manuscript. There is no applicable dataset to share.