Nephrologists are able to select membranes manufactured from cellulose or synthetic polymers, ranging in flux, and available in a variety of devices. The choice of membrane is essential for delivery of adequate therapy because: 1) it has been recognized that the long-term accumulation of β2 microglobulin leads to bone and soft tissue damage; 2) blood membrane contact results in widespread pathway activation; and 3) epidemiologic studies have demonstrated the benefits of using membranes with improved biocompatibility profiles in the treatment of end-stage renal disease. 1–3 Although cellulose based membranes remain widely used throughout the world, there has been a move away from such membranes, particularly in the United States, where in 1997 regional use of synthetic membranes reached 86%. 4 Membranes are differentiated according to polymer, but may behave in quite a different way in the clinical setting. 5 In evaluating a membrane, small molecular solute removal is based on the diffusive permeability of the membrane, while removal of large molecular weight substances is determined by pore size and pore distribution. Biocompatibility is determined by the polymers used and their interaction with blood.
In a prospective clinical crossover study, we measured and compared membrane performance in terms of solute removal and biocompatibility characterized by immune and clotting system activation changes of a high flux synthetic membrane with a microdomain structure manufactured from a blend of polyamide, polyarylethersulfone and polyvinylpyrrolidone (Polyflux, Gambro GmbH, Hechingen, Germany) with Fresenius Polysulfone (Fresenius AG, Bad Hamburg, Germany) to evaluate the effect of base polymer and membrane structure on these parameters.
Materials and Methods
Membranes and Hemodialyzers
The Polyflux membrane is manufactured from an alloy of polyamide and polyarylethersulfone which acts as a hydrophobic base polymer, and polyvinylpyrrolidone (PVP), which acts as a hydrophilic component in the form of a hollow fiber. Scanning electron micrographs of the membrane show the membrane to be an asymmetric, three-layer structure comprised of a thin inner layer with a blood contacting surface of 0.1–0.5 μm thickness, supported by a 5 μm thick sponge structure with a similar porosity as the blood contact layer, but with increasing pore diameters toward the bulk structure which, in turn, is supported by a finger type structure. The overall thickness of the membrane is 50 μm. 6 Details regarding the structure and production of Fresenius Polysulfone have been described elsewhere. 7,8 This membrane is asymmetric with a thin (∼1 μm) inner lumen supported by a porous foam structure and has a wall thickness of 40 μm.
For the purpose of this study the membranes were used in steam sterilized hemodialyzers, manufactured in accordance with CE regulations and with effective surface areas of 1.36 m2 (Polyflux Model 14S, Gambro) and 1.3 m2 (Model F60S, Fresenius).
Informed consent and ethical committee approval was sought and granted for the study performed in accordance with the guidelines of the European Standard governing clinical investigation of medical devices in human subjects (EN 540). The study involved a group of eight patients receiving regular dialysis therapy for end-stage renal disease whose mean age was 54 years. All patients were dialyzed via an arteriovenous fistula, and flow rates during treatment ranged between 350 and 400 ml/min. Treatment times were between 200 and 240 minutes. Prior to participation in the study, all patients received hemodialysis treatment for at least 6 months using low flux polysulfone membranes in conjunction with bicarbonate buffered dialysis fluid and ultrafiltration control (Drake Willock Sytem 1000, Althin Medical Inc., Portland, OR). At the time of study no patient was suffering from infection or taking medication known to interfere with measured parameters. Patient anticoagulation regimens were unaltered during the study. Extracorporeal circuits were primed with saline containing 5,000 IU heparin; at the initiation of treatment patients received a loading dose (500–2,000 IU) followed by continuous infusion (500–2,000 IU/hr). The infusion was terminated at either 60 or 90 minutes prior to the end of dialysis.
Study Design and Statistical Analysis
Elements of the study design are shown in Table 1. It was a two period crossover design in which each study period lasted three consecutive dialysis sessions. Patients were allocated to the membranes randomly, and each membrane (or dialyzer) was used during three consecutive treatments. The dialyzers were not reused. To minimize carryover, no measurements were made during the first dialysis with either of the membranes. The solute transport characteristics were established during the second treatment and biocompatibility on the third. Study analysis was carried out using a General Linear Interaction Model (GLIM) that permitted differences between patients in the respective treatment weeks to be eliminated from the treatment comparisons and standard errors. Serial measurements were aggregated into a single summary measure (area under curve, AUC) and, where necessary, the data were suitably transformed to satisfy normality assumptions for the analysis method.
Study Parameters and Analytical Methods
No measurements were made on the first occasion that each membrane was used. The assessment of the solute removal with respect to small molecular clearance and β2 microglobulin removal was made on the second occasion that each membrane was used. The clearance of small molecules was established by sampling from the arterial and venous segments of the extracorporeal circuit 1 hour after treatment initiation. Fluid removal rate during these measurements was maintained at 0.3 kg/hr (5 ml/min). Blood flow rate was determined by a bubble transit technique using calibrated blood tubing sets. The samples collected were spun, separated within 30 minutes of sampling; and levels of urea, creatinine, and phosphate were determined using an analyser (Technicon Model DAX72 Analyser, Bayer Diagnostics, Newbury, UK).
Removal of β2 microglobulin during dialysis was established by changes in plasma β2 microglobulin levels from samples taken predialysis, and at 30, 60, and 180 minutes during treatment. In addition, each membrane sieving coefficient for β2 microglobulin was established at these times. Transmem- brane recovery of β2 microglobulin, as well as total protein and albumin, were determined by continuous collection and sampling of the dialysis fluid during treatment. Plasma and dialysate β2 microglobulin levels were determined using a commercially produced ELISA assay; total protein levels in the dialysis fluid were established using a modified Lowry method 9; and a differential protein analyser (Technicon Model DPA System, Bayer Diagnostics, Newbury, UK) was used for the determination of dialysate albumin levels.
The biocompatibility profiles of the membranes were established on the third occasion used, with sampling via specially inserted sampling ports immediately before and after the dialyzer. White cell and platelet counts were established using a Coulter counter on blood samples taken at the inlet to the dialyzer. Samples collected for complement activation (C3a Des Arg, C5a Des Arg, SC5b-9) and thrombin:antithrombin III levels were taken from blood leaving the dialyzer, stored at −70°C, and analyzed using commercially produced assay kits.
Small molecule clearance of the membranes at a dialysate flow rate of 500 ml/min was comparable, despite small differences in the dialyzer surface areas and the overall wall thickness of the membranes. (Table 2). In the case of creatinine for both membranes, the clearance values were lower than expected for such molecular weight substances. A number of studies have shown that whereas urea diffuses rapidly from the red cells into plasma during the passage of blood through the dialyzer, creatinine hardly diffuses and the solute is extracted mainly from the plasma. As a consequence, equilibration takes place in blood drawn at the dialyzer outlet, influencing the clearance value. 10–12 The mean (SD) predialysis levels of β2 microglobulin were 34.8 (6.1) for the Polyflux membrane and 40.3 (9.8) for the Fresenius Polysulfone. A reduction in plasma levels was present with both membranes (Table 3). The data collected were analyzed by consideration of both actual values and percentage reductions achieved during treatment and were corrected for changes in distribution volumes. 13 Neither analysis showed evidence of membrane differences. The mechanism of β2 microglobulin removal during dialysis is by convective and diffusive transmembrane transport and adhesion to the membrane. To assess transmembrane transport, the membrane sieving coefficient was also measured after 30, 60, and 180 minutes of treatment. Blood samples were taken from blood entering and leaving the dialyzer. The filtrate sample was collected from the dialysate pathway after disconnecting the dialysate lines from the dialyzer and draining the dialyzer by passing air through the dialysate pathway. To achieve a normal response distribution with a uniform variance, values were log transformed before fitting to the standard GLIM model. There were significant effects due to patient, membrane type, and week, but not between the observations at any of the measurement times. The sieving coefficient for the Polyflux was 29% higher than for the Fresenius Polysulfone (95% confidence interval 109–151%). This difference failed to reach statistical significance (p = 0.12). Associated with transmembrane transport was the recovery of β2 microglobulin from the dialysis fluid, which averaged 207 (84) mg for the Polyflux and 147 (29) mg for Fresenius Polysulfone membrane. The standard error of difference between the means was 58 mg (p = 0.12). In addition to β2 microglobulin, the dialysis fluid also contained 7,758 mg (Polyflux) and 7,793 mg (Fresenius Polysulfone) of protein. Protein sieving coefficients of the membranes were influenced by patient, membrane, and week; however, after adjustment for the effects of week and patient, the sieving coefficient of Polyflux were 13% higher than for Fresenius Polysulfone (95% confidence interval 103–123%) The dialysate concentrations of albumin were below the limit of detection (5 mg/L, ≅600 mg/dialysis session) for samples from both membranes.
The white blood cell counts during dialysis have been normalized to pretreatment levels and expressed as a percentage of these values (Figure 1). The white cell count declined in 15 minutes to 83% (8.1) for the Polyflux membrane and to 83% (10) for the Fresenius Polysulfone; thereafter white cell count returned to pretreatment levels. These changes were not statistically different and independent of the membrane. The platelet counts (109/l) were similar predialysis: 163 (35), Polyflux; 169 (24), Fresenius Polysulfone, but a small (< 5%) decline was noted in 15 minutes of treatment with a return to pretreatment levels for both membranes. The observed changes were independent of membrane type (Figure 2)
Changes in complement component levels are shown in Table 4. The time weighted average values established by the division of the AUC by treatment duration for C3a were 641 ng/ml for the Fresenius Polysulfone and 596 ng/ml for the Polyflux membrane. For C5a, the values were 0.64 and 1.08 ng/ml, and for SC5b-9, 286 and 140.3 ng/ml. A considerable inter-patient level variation within the membranes was noted for these parameters. Logarithmic transformation of the data to satisfy the normality assumption of the analytic method failed to demonstrate statistical differences between the membranes.
Blood material contact is associated with the triggering of hemostatic reactions. Using an ELISA assay, intravascular coagulation by the measurement of thrombin:antithrombin III complex was performed (Table 5). Pretreatment baseline levels were comparable for both membranes. Levels at 60 minutes were also similar, and a slight increase by 180 minutes was noted for both membranes; statistically there was no difference between the membranes. Heparin activity during dialysis was monitored by Factor Xa activity (IU/ml). A considerable inter-patient variability was noted for this parameter, with levels exceeding the therapeutic range necessary for coagulation (0.3–1.0 IU/ml).
The manufacture of synthetic membranes is complex and details of the manufacturing process are protected by patents. They are typically manufactured by a phase inversion process in which the polymer solution is converted to a porous framework through the exchange of solvent with the precipitating nonsolvent. Modifications of the polymers used, as well as the manufacturing technique, enable control of the solute removal and biocompatibility profile of the material to be tailored to clinical requirements. 6
Our study suggests that this approach results in the ability to match the characteristics of widely used membranes such as Fresenius Polysulfone, not only in terms of solute removal, but also with respect to blood material contact phenomena. The results obtained demonstrate minor differences between the materials that are unlikely to be of clinical significance. The study has failed to differentiate between dialyzer removal of low molecular weight waste products, despite minor differences in membrane thickness and surface area. The clearance of such molecules is also dependent upon the dialysis fluid flow rate. Measurements were made at a blood flow rate of 250 ml/min and the dialysate flow rate was maintained at 500 ml/min. High efficiency therapies utilize dialysate flow rates within the range of 700–1,000 ml/min, and, under these conditions, small molecular weight clearance is increased by between 8 and 12%. Membranes are not selective in their removal of higher molecular weight compounds, and the ability to remove β2 microglobulin is counterbalanced by removal of low molecular proteins, a known aspect of CAPD. 14 Such protein loss has also been recognized in hemodialysis, 15 and long-term clinical use of high permeability membranes may necessitate closer monitoring of plasma protein levels and, where appropriate, supplementation to compensate for losses. The ability to remove large molecular weight solutes increases the possibility of bacterial fragment or endotoxin transfer from the dialysis fluid. A study by Weber et al.16 tested the possible penetration by endotoxins containing lipid A. from Escherichia coli in tangential filtration mode using an aqueous medium, as well as LPS from Pseudomonas aeruginosa through Fresenius Polysulfone and Polyflux, and observed a high adsorptive capacity with neither membrane allowing penetration.
The trigger of proinflammatory and procoagulatory pathways during renal replacement therapy is complex and not fully understood. The molecular interaction between the C3 component of the complement system and surfaces containing hydroxyl (OH) groups is central to the changes observed when using cellulose based membranes. 17 However, membranes such as those studied do not contain such groups on their surface and other factors, such as the surface ability to bind regulatory proteins e.g., Factor H, an inhibitor, or Factor B, a promoter of complement activation, may occur while the membrane surface interaction with the cell membrane containing glycoproteins and phospholipids may also play a role. The interaction of blood with materials in the extracorporeal circuit also results in a thrombogenic response. Heparin is used to prevent clotting within the circuit and to maintain device performance during treatment. It is important to emphasize that the presence of heparin does not prevent the interaction between blood elements and the materials of the extracorporeal circuit, as reflected by the changes in thrombin:antithrombin III complex.
The trend towards the use of more biocompatible membranes is beyond dispute. However, proof of a causal relationship between clinical diseases and biologic reactions is scarce, and the impact of activated pathways on patient clinical status remains a debated point. This study has demonstrated that design of membrane material by appropriate blend of polymers, results in the ability to combine effective solute removal with minimal activation of the inflammatory and coagulation pathways.
The authors thank Professor R. Wilkinson, Dr. J. S. Tapson, and Dr. A. Brown for permitting patients under their care to be studied, and the staff of the Renal Unit, the Departments of Haematology and Clinical Biochemistry of the Newcastle upon Tyne, NHS. Supported by grants from Gambro R& D, Hechingen, Germany, the Northern Counties Kidney Research Fund Trust, and the Special Trustees of the Royal Victoria Infirmary, Newcastle upon Tyne.
1. Hakim RM, Held PJ, Stannard DC, et al: Effect of the dialysis membrane on mortality of chronic hemodialysis patients. Kidney Int 50:566–570, 1996.
2. Koda Y, Nishi S, Miyazaki S, et al: Switch from conventional to high flux membrane reduces the risk of carpal tunnel syndrome and mortality of hemodialysis patients. Kidney Int 52:1096–1101, 1997.
3. Locatelli F, Marcelli D, Conte F, Limido A, Malberti F, Spotti D: Comparison of mortality in ESRD patients in convective and diffusive extracorporeal treatments. Kidney Int 55:286–293, 1999.
4. US Renal Data System:USRDS Annual Data Report: National Institutes of Health
. Bethesda, MD, National Institute of Diabetes and Digestive and Kidney Diseases, 1998.
5. Hoenich NA, Woffindin C, Brennan A, Cox PJ, Matthews JNS, Goldfinch M: A comparison of three brands of polysulfone membranes. J Am Soc Nephrol 7:871–876, 1996.
6. Deppisch R, Gohl H, Smeby L: Microdomain structure polymeric surfaces: Potential for improving blood treatment procedures. Nephrol Dial Transplant 13:1354–1359, 1998.
7. Brachtendorf TH, Nederlof B: Fresenius polysulfone-the gold standard in dialysis: 10 years of experience. Clin Nephrol 42 (Suppl 1):S3–S12, 1994.
8. Streicher E., Schneider H: The development of a polysulfone membrane. A new perspective in dialysis. Contrib Nephrol 46:1–13, 1985.
9. Svensmark O: Determination of protein content in cerebrospinal fluid. A comment on the Lowry method. Scan J Clin Lab Invest 10:50–521, 1958.
10. Colton CK, Smith KA, Merrill EW, Reece JM: Diffusion of organic solutes in stagnant plasma and red cell suspensions. Chem Eng Prog Symp 99:85–100, 1970.
11. Simpson WG: Plasma creatinine and the effect of delay in separation of samples. Ann Clin Biochem 29:307–309, 1992.
12. Descombes E, Perriard F, Fellay G: Diffusion kinetics of urea, creatinine and uric acid in blood during hemodialysis. Clinical implications. Clin Nephrol 40:286–295, 1993.
13. Bergstrom J, Wehle B: No change in corrected β2
microglobulin concentrations after cuprophane hemodialysis. Lancet 1:628–629, 1987.
14. Kagan A, Bar-Khayim Y, Schafer Z, Fainaru M: Kinetics of peritoneal protein loss during CAPD: 1 Different characteristics for low and high molecular weight proteins. Kidney Int 37:971–979, 1990.
15. Naito H., Miyazaki T: Hemodialysis with protein permeable membranes. Nephrol Dial Transpl 4:78–83, 1989.
16. Weber C, Linsberger I, RafieeTehrani M, Falkenhagen D: Permeability and adsorption capacity of dialysis membranes to lipid A. Int J Artif Organs 20:144–152, 1997.
17. Cheung AK: Biocompatibility of hemodialysis membranes. J Am Soc Nephrol 1:150–161, 1990.