Background
Liver failure—acute or acute-on-chronic—is associated with high mortality. According to the toxin hypothesis of liver failure, extracorporeal albumin dialysis (ECAD) may provide a therapeutic option during critical care by reducing endogenous albumin-bound toxic agents such as bile acids [
1]. Currently, ECAD is mainly conducted using the Molecular Adsorbents Recirculating System (MARS). Several clinical trials have certified MARS as a feasible tool in reducing patients’ bilirubin levels, improving haemodynamic status and hepatic encephalopathy (HE) [
2‐
4]. However, large clinical trials have failed to demonstrate a survival benefit for patients treated with MARS [
2,
4]. Another ECAD technique, single-pass albumin dialysis (SPAD), has been proposed. This technique enables easy access to extracorporeal liver support by using standard dialysis devices [
5]. In case reports and small clinical studies, researchers have reported the feasibility of SPAD in reducing patients’ bilirubin levels and improving their clinical condition (i.e., HE) [
6]. However, while MARS is routinely prepared with 600 ml of a 20 % albumin solution according to manufacturers’ instructions, SPAD applications were reported to run on different albumin concentrations and dialysate flow rates [
7,
8]. Thus, comparison of both techniques requires clear definitions of SPAD construction. Recently, our study group retrospectively compared the effects of MARS and 4 % SPAD with a dialysis flow rate of 700 ml/h on clinical and laboratory parameters in critically ill patients [
9]. Similar settings have already been used in an in vitro comparison of MARS and SPAD that resulted in equivalent detoxification [
8]. Among others, in our retrospective study, both devices were comparable in reducing bilirubin levels, but MARS treatment resulted in higher renal dialysis capacity, reflected by lower creatinine and urea levels after treatment [
9]. Moreover, the equivalence of both devices in clinical routine was questioned, as chemical stabilizers added to albumin solutions may reduce performance, especially during SPAD, while they are partially removed by recirculation and purification in the MARS device [
10,
11].
Thus, the aim of this study was to prospectively compare the performance of MARS and 4 % SPAD treatment in patients with severe liver failure using a single-centre crossover design, with particular focus on clinical and laboratory parameters. As a primary endpoint, we hypothesised non-inferiority in bilirubin reduction of SPAD in comparison to MARS. Changes in laboratory and clinical liver-specific parameters in both devices constituted secondary study endpoints.
Discussion
Following the autointoxication hypothesis of liver failure [
1], the aim of using liver support systems is to reduce endogenous liver toxins. They carry the potential to reduce not only water-soluble toxic substances (e.g., by means of conventional haemodialysis) but also hydrophobic agents (e.g., by using albumin as a dialysis solution). Due to its biochemical structure, albumin is able to bind numerous endo- and exogenous substrates, such as bilirubin and bile acids [
25], that can thus be cleared via ECAD (i.e., MARS or SPAD).
In this prospective crossover study, we compared these two extracorporeal liver support systems with a particular focus on their effect on bilirubin levels and further clinical and laboratory parameters. We found equivalent significant reductions in plasma bilirubin levels in both MARS and 4 % SPAD, without differences between the methods. Measurement of bilirubin, both unconjugated and conjugated, is common in intensive care units to detect serious disorders (e.g., cholestasis or haemolysis). Bilirubin levels were found to be associated with prognosis in the critical care setting [
26] and particularly in liver failure. With respect to extracorporeal liver support therapy, bilirubin is often used as one of several parameters to indicate therapy and, in addition, to assess effective elimination of albumin-bound substances. Bilirubin can easily pass the haemodialysis filter; however, due to its lipophilic structure and its low water solubility, a carrier is necessary to eliminate bilirubin via haemodialysis. Human serum albumin, with its three bilirubin binding sites [
27], has a high capacity to remove bilirubin during albumin dialysis. In a retrospective analysis, we had found similar reduction rates of bilirubin in patients treated with either MARS or SPAD [
9]. We used these results for a sample size calculation and bilirubin thereby became the primary endpoint in our present study.
Although not routinely assessed, other parameters, such as bile acids or ABiC, might be of superior clinical significance in defining excretory liver dysfunction compared with bilirubin levels. For example, toxic bile acids can induce hepatocyte and biliary epithelial cell necrosis [
28] and are therefore of potential prognostic significance [
29,
30]. Furthermore, they might contribute to remote organ dysfunction such as cirrhotic cardiomyopathy [
31] or reduced systemic vascular resistance [
32]. In contrast to bilirubin levels, TBA concentrations were significantly reduced only during MARS and not during SPAD treatment. MARS was previously shown to reduce bile acid concentrations in vivo [
33‐
35]. In the case of SPAD, two in vitro studies [
8,
36] and one small uncontrolled study [
7] addressing the clearance of bile acids have been published. Interestingly, in the study of Sauer et al. [
8], MARS and SPAD were comparable in bile acid clearance; in the study of Benyoub et al. [
7], bile acids were significantly removed during a 10-h SPAD (3.2 % albumin concentration) procedure with a dialysate flow rate of 1000 ml/h. These results are in clear contrast to our findings, the latter possibly being due to higher dialysate flow rate and longer treatment period, resulting in a much higher dialysis dose. One potential explanation for the discrepancy in the in vitro studies could be the different dialysis modes used (continuous venovenous haemodiafiltration in vitro vs. continuous venovenous haemodialysis [CVVHD] in our study). However, with a molecular weight of about 500 Da, bile acids should be removed more effectively in the dialysis mode; indeed, Benyoub et al. used CVVHD as well [
7]. The more likely explanation is an inherent limitation of the in vitro setting representing a one-compartment model (i.e., the intravascular compartment). In vivo elimination of bile acids is complicated by diffusion of bile acids from tissue and by ongoing bile acid production. Differences in the effectiveness of MARS and SPAD in removing bile acids could thereby be unmasked.
With respect to potential toxicity of bile acids and the higher removal rate of bile acids during MARS treatment in our study, the reduction of GGT levels during MARS could suggest bile duct regeneration, as high levels of GGT correlate to bile duct lesions [
37]. Thus, it can be speculated that reduction and/or expression pattern change of bile acids (i.e., allocation of conjugated to unconjugated) could improve the disease course. However, recent large randomised clinical trials of extracorporeal liver support, in either acute or acute-on-chronic liver failure [
2,
4], did not focus on bile acid removal, so further study would be needed to directly address this possibility.
Human serum albumin is the major carrier protein for a multitude of endogenous and exogenous substances. Many accumulating substances in liver failure are transported bound to albumin. ABiC was developed to assess the binding site II–specific binding capacity of human serum albumin, the binding site of, for example, bile acids [
38]. ABiC correlated inversely with severity and 30-day mortality in patients with decompensated cirrhosis [
18]. ABiC was exclusively increased during MARS treatment in the present study. Such an improvement of ABiC in MARS treatment was demonstrated previously [
39]. In the present study, SPAD was not able to improve ABiC. These findings might be at least in part an accompanying effect of the results seen in bile acids, as higher bile acid levels with higher binding at binding site II might lead to reduced ABiC. Another potential explanation for this finding may be in the stabilizers added to pharmaceutical albumin preparations. These stabilizers may occupy binding sites of albumin (e.g., octanoate, which also binds to albumin binding site II). On the one hand, these stabilizers can be bound to the adsorption columns in the albumin circuit in MARS, thereby improving capacity to remove albumin-bound substances such as bile acids from the blood circuit. On the other hand, the demonstrated higher increase of octanoate concentrations in blood of patients during SPAD treatment may contribute to the less increased ABiC in comparison to MARS treatment [
11]. Of note, the albumin solution used in the present study contained the stabilizer octanoate equivalent to caprylate. Whether removing these stabilizers (e.g., by charcoal filters) before albumin dialysis might overcome these shortcomings in SPAD therapy remains elusive.
As acute or chronic renal insufficiency displays a common complication in liver dysfunction, removal of water-soluble substances, such as urea and creatinine, is often required during ECAD. However, assessment of renal dysfunction by traditional scoring systems (i.e., AKIN criteria) is not always applicable [
40], and some patients require haemodialysis for reasons not related to creatinine levels and urine output (e.g., hypervolemia or acidosis). In the present study, the majority of the critically ill patient cohort required renal replacement therapy before study inclusion. Thus, ECAD treatment was coupled to conventional haemodialysis circuits. The ability to decrease water-soluble substances in our study was higher in MARS than in SPAD treatment, most due to differing dialysate flow rates among the ECAD devices (MARS 2000 ml/h vs. SPAD 700 ml/h). In the clinical setting, this might be of minor relevance, as it could easily be compensated for if a 7-h SPAD treatment were followed by conventional continuous renal replacement therapy using the same dialysing machine with conventional dialysing solution and a higher dialysate flow rate. However, it must be noted that variances in the reduction of water-soluble substances may also be altered by residual endogenous renal function, possibly influencing the efficacy of ECAD therapy. Nevertheless, due to the crossover study design, the influence of either MARS or SPAD on alteration of kidney retention parameters in the present study may overcome these shortcomings.
Pro-inflammatory cytokines are suspected to induce and/or augment hepatocellular damage and cholestasis, thereby contributing to the course of liver failure [
41]. However, there is evidence that growth factors and pro-inflammatory cytokines are involved in liver regeneration and proliferation of hepatocytes. Data regarding the influence of ECAD on systemic cytokine levels in patients with acute or acute-on-chronic liver failure are conflicting [
42‐
44]. Removal of cytokines via albumin dialysis has been demonstrated. In addition, circulating cytokine levels could be affected by either removal by dialysis or induced changes in the rate of production. Moreover, different anti-coagulation strategies could possibly alter cytokine removal. Thus, elevated IL-8 levels were shown to be removed via CVVHD under regional citrate anti-coagulation, but not using heparin anti-coagulation. Interestingly, IL-6 levels were altered by neither citrate nor heparin anti-coagulation [
45]. However, researchers in other clinical studies could not find any differences regarding cytokine removal during conventional haemodialysis using different anti-coagulation strategies [
46,
47]. In this respect, combining albumin dialysis with a cytokine adsorption filter may be superior regarding cytokine removal [
48]. However, the clinical relevance of elevated cytokine levels in patients with liver failure is not fully understood, as cytokines could enhance cell damage as well as induce liver regeneration [
43,
49,
50]. In our study, elevated systemic cytokine levels were lowered neither during MARS nor during SPAD application. Moreover, neither citrate nor heparin anti-coagulation altered cytokine levels in the present study.
Both systems were safe in providing extracorporeal liver support in critically ill patients, particularly in view of bleeding complications, transfusion rates and haemodynamic stability. Clinically non-significant changes in thermal balance may be explained by more pronounced warming of the dialysate solution during MARS (two heating devices: dialysis machine and MARS monitor) compared with SPAD (one single heating during haemodialysis). In SPAD treatment, we found higher rates of metabolic derangement (increase in pH, BE and lactate values) and electrolyte disturbances (decreasing calcium levels and increasing sodium levels), resulting in osmolality displacements. These changes were limited to patients receiving citrate anti-coagulation, hinting at a relative overdosing of citrate [
51], most likely due to the low dialysate flow rate of 700 ml/h in SPAD treatment and the use of commercially available citrate solutions designed for conventional haemodialysis. To compensate for this apparent disadvantage (1) dialysate flow rates could be increased or (2) lower concentrated sodium citrate solutions during SPAD procedures running on regional citrate anti-coagulation could be used. On one hand, increasing the dialysate flow rate would reduce the treatment duration or it must be accounted for by reducing the albumin content. Increasing the albumin concentration as well as the dialysate flow rate, on the other hand, would enhance costing of SPAD, abolishing the savings of around €1500 per treatment cycle using the described setting of our study in comparison to MARS (only material costing, not including personnel for device build-up).
Regarding HE, patients scored a median of grade III on the HESA scale of four grades and 14 on the Glasgow Coma Scale, indicating presence of HE in the patient cohort. Albumin dialysis was not able to significantly affect HE grade among the study population, although this was shown before by Hassanein et al. [
3]. Competitive sedative medication allowing invasive mechanical ventilation in this critically ill patient cohort may represent one explanation for this negative finding. Moreover, the study was not designed to improve HE. We evaluated HE scores after each ECAD cycle, which was not expected to provide enough detoxification capacity to alter the HE course. This is consistent with the findings of the study of Hassanein et al., in which most patients needed at least two treatments before HE improved.
Two limitations of our study are its single-centre design and the relatively small number of patients. However, we made a careful sample size calculation based on retrospective data from our institution. As liver failure is a rare disease and patients who require extracorporeal liver support must be selected carefully, the number of potential study patients is limited. Therefore, we implemented the possibility for multiple applications of up to four crossover cycles of albumin dialysis per patient in the study design. This was accounted for in the sample size analysis. Moreover, the crossover study design revealed some major advantages in the present analysis: (1) there are no confounding factors with respect to comparability between the control- and test-group, as both were represented by the same patient/cycle, and (2) requirements of type I or type II errors are generally comparable in cross-crossover design studies compared with parallel group trials, at a lower sample size [
52]. In addition, we assessed only a limited number of potential outcome-relevant factors (i.e. cytokine, bile acid or bilirubin levels and ABiC). We cannot exclude that there are significant differences regarding further parameters potentially influenced by ECAD, such as plasma levels of other toxic substances or other albumin-related factors (e.g., heavy metal binding capacity). Furthermore, the detoxification efficiency of the closed albumin circuit (MARS) and the open albumin-enriched dialysis system (SPAD) are dependent on some confounding factors; for example, high solute concentrations of albumin-bound substances might result in early saturation of adsorbers in MARS, while low concentrations might leave residual adsorption capacity. However, in SPAD, higher solute concentrations potentially will not lead to saturation of the dialysis process. In addition, longer treatment times or higher dialysate flow rates in principle could result in higher clearance of substances, albeit accompanied by higher treatment costs. The decision to run MARS in the described setting was based on the manufacturer’s recommendation for duration of MARS [
1]. This approach was previously used in larger multicentre trials on acute and acute-on-chronic liver failure, respectively [
2,
4]. Treatment settings of SPAD were based on a previous in vitro study and on our own retrospective in vivo analysis [
8,
9]. To preclude further confounders, blood flow rates were equal during respective MARS and SPAD treatments. Moreover, the crossover design itself limits bias in comparing an open against a closed system, as all patients receive both treatments.
Conclusions
This prospective, randomised, controlled crossover study demonstrated the investigated albumin dialysis procedures to be safe for temporary extracorporeal liver support. Overall, both devices displayed comparable results for most clinical and paraclinical parameters in our critically ill patient cohort, especially in light of bilirubin reduction. However, particular focus should be placed on metabolic derangements and electrolyte disturbances in case of regional citrate anti-coagulation and reduced renal haemodialysis efficacy (i.e., creatinine and urea levels) caused by reduced dialysis flow rates during SPAD application. Increasing dialysis efficacy of SPAD by increasing dialysate flow rate and extending treatment duration might compensate for this disadvantage. However, this needs to be investigated further. In addition, it would increase treatment costs of SPAD, thereby decreasing the cost savings in comparison to MARS. Moreover, in view of recent aspects regarding the pathophysiology of liver failure and consecutive remote organ failure, it remains speculative if MARS may provide advantages over SPAD by reducing bile acid concentration and improving ABiC. Thus, further studies addressing the timing and duration of ECAD, particularly in view of altering clinical and paraclinical parameters beyond bilirubin reduction (e.g., HE, cytokine removal, ABiC enhancement and bile acid changes), are warranted. Furthermore, meaningful clinical outcomes, such as patient survival, time on mechanical ventilation, vasopressor support or course of HE, should be addressed.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
CS, AK, and MB designed and monitored the study, evaluated the data, and wrote the manuscript. KM, DR, and WB were responsible for data and sample collection and storage and reviewed the manuscript. SK and SM provided laboratory assistance for ABiC measurement and reviewed the manuscript. DK provided laboratory assistance for bile acid and cytokine measurement and reviewed the manuscript. AB and US were involved in study design, patient recruitment, and manuscript revision. MK provided statistical analysis and wrote the statistical analysis part of the manuscript. MGC was involved in study monitoring, language editing, and manuscript revision. All authors reviewed the manuscript critically for important intellectual content, and all authors read and approved the final manuscript.