Combining Diffusion, Convection and Absorption: A Pilot Study of Polymethylmethacrylate versus Polysulfone Membranes in the Removal of P-Cresyl Sulfate by Postdilution On-Line Hemodiaﬁltration

: Dialytic clearance of p-cresyl sulfate (pCS) and other protein-bound toxins is limited by diffusive and convective therapies, and only a few studies have examined how to improve their removal by adsorptive membranes. This study tested the hypothesis that high-ﬂux polymethylmethacrylate (PMMA) dialysis membranes with adsorptive capacity increase pCS removal compared to polysulfone membranes, in a postdilution on-line hemodiaﬁltration (OL-HDF) session. Thirty-ﬁve stable hemodialysis patients randomly completed a single study of 4 h OL-HDF with PMMA (BG2.1U, Toray ® , Tokyo, Japan) and polysulfone (TS2.1, Toray ® ) membranes. The primary endpoint was serum pCS reduction ratios (RRs) obtained with each dialyzer. Secondary outcomes included RRs of other solutes such as β 2-microglobulin, the convective volume obtained after each dialysis session, and the dialysis dose estimated by ionic dialysance (Kt) and urea kinetics (Kt/V). The RRs for pCS were higher with the PMMA membrane than those obtained with polysulfone membrane (88.9% vs. 58.9%; p < 0.001), whereas the β 2-microglobulin RRs (67.5% vs. 81.0%; p < 0.001), Kt (60.2 ± 8.7 vs. 65.5 ± 9.4 L; p = 0.01), Kt/V (1.9 ± 0.4 vs. 2.0 ± 0.5; p = 0.03), and the convection volume (18.8 ± 2.8 vs. 30.3 ± 7.8 L/session; p < 0.001) were signiﬁcantly higher with polysulfone membrane. In conclusion, pCS removal by OL-HDF was superior with high-ﬂux PMMA membranes, appearing to be a good dialysis strategy for improving dialytic clearance of pCS, enabling an acceptable clearance of β 2-microglobulin and small solutes.


Introduction
Chronic kidney disease (CKD) is characterized by the progressive accumulation of multiple chemical compounds that are normally excreted into the urine in healthy people [1][2][3][4][5]. These chemical compounds are globally known under the name of uremic toxins and are conventionally classified into three groups based on their physicochemical properties [2]. These major groups include small water-soluble compounds with molecular weight less than 500 Da, larger molecules with molecular weight greater than 500 Da

Hemodialysis Procedures
Each patient underwent two OL-HDF sessions with usual dialysis parameters: dialysis buffer with bicarbonate, dialysate flow rate (Qd) 500 mL/min, blood flow rate (Qb) between 350 to 450 mL/min, and dialysis time between 4.0 and 4.5 h. All patients received postdilution OL-HDF with automatic adjustment of the substitution fluid flow rate, to maximize substitution volume while simultaneously avoiding hemoconcentration and filter clotting [26]. All treatments were performed with the 5008 HD system (Fresenius Medical Care), and with ultrapure dialysis fluid, containing <0.1 colony-forming unit/mL and <0.03 endotoxin unit/mL. Treatment parameters, including blood and dialysate flow rates, length of the dialysis session, and ultrafiltration rate, remained unchanged during both sessions. The only difference among the two dialysis sessions in each patient was the dialyzer: high-flux PMMA BG2.1U (Toray ® , Tokyo, Japan) and high-flux PS TS2.1SL (Toray ® , Tokyo, Japan). Differences and similarities of both membranes are shown in Table 1. All the sessions were performed in the intermediate period of Wednesday or Thursday, with a 4-week interval between study sessions. During this wash-out period, patients remained in their usual HD treatment plan with no changes; all received postdilution OL-HDF with high-flux, PS FX-100 (Fresenius Medical Care ® , Bad Homburg vor der Höhe, Germany). The order of the two different treatment sessions was randomly assigned to the patients.

Blood Sampling
Predialytic blood samples were collected after insertion of the access needle, ensuring that the blood was not diluted by saline or heparin. The postdialytic sample was drawn from the arterial needle after slowing the blood pump to 50 mL/min [29]. Blood samples were collected in serum gel tubes and then were left to stand for a minimum of 50 to 60 min before centrifuging for 10 min at 3500 rpm. Serum was then separated, analyzed for small water-soluble compounds and β2-microglobulin, and finally frozen at −80 • C until analysis of pCS. Total pCS levels were analyzed by HPLC using an Agilent Technologies 1100 liquid chromatograph with a quaternary pump, a diode array detector, a thermostatted column compartment, an autosampler, and an HP Compaq computer equipped with Agilent-Chemstation software (Agilent Technologies, Santa Clara, CA, USA). Detailed information regarding serum sample preparation and HPLC analytical methodology for pCS assessment is depicted in Appendix B. The validation of the chromatographic method used is shown in Appendix C.
Urea, creatinine, uric acid, and phosphate were also measured by automated molecular absorption spectrometry methods with the C16000 Architect (Abbot Diagnostics, Abbott Park, IL, USA). The normal range is 15-50 mg/dL for urea, 0.60-1.20 mg/dL for creatinine, 3-7 mg/dL for uric acid, and 2.3-4.7 mg/dL and for phosphate. β 2 -microglobulin concentrations were determined by a solid phase chemiluminescent immunoassay with the Siemens Inmulite 2500 Immunology Analyzer. The normal range is from 0.7 to 3.4 mg/L. Other laboratory measurements were performed using standard techniques at our hospital laboratory.

Statistical Analyses
Sample size calculation was estimated according to the expected different effect of the two dialyzers in the removal of pCS. Based on published data, the reduction rates of pCS by postdilution OL-HDF are around 40.0% with high-flux PS membranes [32]. Assuming a reduction rate difference of 15% at a standard deviation of 25%, and considering an error of beta = 0.8, a sample size of at least 29 patients was estimated. Finally, 35 patients were recruited.
Descriptive statistical results are presented as mean ± SD, median and interquartile range, and as a percentage of all patients as appropriate. For treatment comparisons of RR of each solute, linear mixed models were employed with covariate adjustment for the baseline level of the solute, treatment and sequence as fixed effects, and patient as a random term. The resulting p values were based on differences in least square means for the factor treatment and a significance level of 5% was employed. Comparison of other dialysis features (real length dialysis session, Qb, arterial pressure, venous pressure, initial and final body weight, ultrafiltration volume, volume of blood processed, convective volume, Kt, and Kt/V) between the two dialysis sessions was assessed using a paired sample t-test. Data collection of dialysis parameters was carried out using Nefrosoft software, version 7.0.1 (Visual-limes, Valencia, Spain). All analyses were conducted using R statistical software, version 4.0.3 (The R Foundation for Statistical Computing, Vienna, Austria), using the "lme4" and "RCommander" packages.

Patient Characteristics
Thirty-five patients accepted to participate and were enrolled in the study. Patient characteristics are summarized in Table 2. All patients were Caucasian.

Dialysis Features
Most dialysis parameters were similar in both dialysis sessions, including duration, Qb, initial and final weight, ultrafiltration volume, arterial and venous pressures, and blood processed. Due to its higher permeability, the replacement fluid volume in postdilution OL-HDF was significantly greater with PS than that obtained with the PMMA dialyzer (Table 3). Both dialysis study sessions were performed without relevant clinical incidents (data not shown).

Discussion
This is the first controlled study evaluating the effect of one of the last PMMA dialyzers suitable for HDF use, the BG-U membrane, on pCS removal by postdilution OL-HDF in prevalent HD patients. PBUT removal remains a challenge in the treatment of HD patients and strategies to decrease levels and hence toxicity, aiming to reduce the cardiovascular burden of these patients, are needed [23,[33][34][35][36]. Whereas in the healthy kidney, PBUT clearance mostly depends on tubular secretion, in dialysis therapies the removal of these toxins is limited to the unbound fraction, not being affected by the pore size of the dialyzer [37], and only slightly by convective transport [38][39][40]. Conversely, PBUTs may be removed by using the adsorptive properties of certain biomaterials, including resins [41,42] and PMMA membranes [43]. Results suggest that BG-U dialyzers, compared to high-flux PS, are highly effective for reducing pCS levels. We also demonstrated that PMMA BG-U series achieved an acceptable convective volume for routine use, confirming a much higher permeability capacity than previous PMMA dialyzers.
With a slightly anionic PMMA membrane, the BG-U series were designed to offer higher biocompatibility to patients, with a controlled pore radius around 70 Å and a uniform distribution of pore size that guarantee high water permeability and porosity [44]. These modifications might enable the use of the BG-U dialyzers in OL-HDF with appropriate convective volume and acceptable albumin loss [25]. They combine the three mechanisms of diffusion, convection, and adsorption in a simple way, improving permeability and adsorption of not only low molecular weight proteins but also of higher molecular weight proteins up to 160,000 Da [45]. Although no other clinical study assessing the effect of BG-U dialyzers in pCS is available, we speculate that the higher efficacy on pCS removal observed in our study with BG-U series may be due to the adsorption properties of PMMA membranes [43]. Several studies have previously demonstrated the efficient removal of other PBUTs such as furancarboxylic acid and pentosidine [46,47], as well as inflammatory markers such as TNF-α, IL-1β, IL-6, and C-reactive protein, by other PMMA dialyzers [48]. The high protein adsorption capacity of these membranes is due to their symmetrical pore structure which provides a large specific surface area [28,49]. Whereas in PMMA membranes the whole membrane thickness is involved in the separation process allowing toxin adsorption, in PS membranes (with asymmetrical pores) only a fine layer of ≈1 µm is responsible for the separation process, while the remaining membrane thickness has structural functions only [20]. These differences in membrane structure may explain the distinct pCS removal profile obtained with the two tested dialyzers, which constitutes the most original finding of this study. Although the results of short-term studies such as this one may not adequately reflect long-term trends and patient outcomes, we think that the reduction of pCS observed with the PMMA BG-U dialyzer could be clinically relevant. The elevated cardiovascular morbidity and mortality risk of HD patients has been repeatedly associated with levels of pCS and other PBUTs [33][34][35][36]. Moreover, recent research suggests that these PBUTs accelerate the progression of CV disease, bone disorders, and neurological complications among CKD patients [50,51]. Results from the annual survey of the Japanese Nationwide Dialysis Registry suggest that the use of PMMA membranes may reduce mortality in HD patients [52]. However, further long-term prospective studies are needed to clarify these findings.
In parallel, we also observed that PMMA BG-U series may be appropriately used by postdilution HDF, confirming the higher permeability capacity observed with new designed PMMA dialyzers [24]. The mean convective volume obtained in our study was close to the 21 L threshold which has been associated with better survival in large randomized clinical trials [53]. However, compared with PS dialyzer, which reached higher replacement volume, the convective efficacy estimated by β2-microglobulin RRs was 14 percentage points lower. These differences are comparable to those recently obtained by Maduell et al. in a safety and efficacy evaluation of PMMA NF-U series [24]. This latest generation of PMMA dialyzers may allow the achievement of high convective volume with no significant albumin loss. All these data suggest that the indication of new high-flux PMMA dialyzers in postdilution HDF may represent a practical compromise between efficient convective and adsorptive dialysis treatment.

Strengths and Limitations
Strengths of this study were its cross-sectional design, with each patient as their own control, and the use of a chromatographic method for the assessment of pCS levels, which was validated according to the European Medicines Agency (EMA) and the Federal Drug Administration (FDA) [54,55]. To eliminate confounding, the same dialysis features were applied to both HD sessions. There are additional limitations, starting with its short-term nature plus the relatively small sample size that leads us to consider this as a pilot trial in need of verification. We did not collect the dialysis fluid to quantify the elimination of toxins. We also did not assess the albumin loss in dialysate. However, available data with BG-U series suggest this PMMA dialyzer as highly adsorptive but with the same cut-off as PS dialyzers [25,56,57], and consequently, its indication in HDF seems safe and appropriate. Moreover, albumin loss is only one of many factors contributing to the risk of hypoalbuminemia in dialysis patients [58]. Additionally, strategies for reducing the risk of malnutrition in this population include improving systemic inflammation by increasing uremic toxin removal and optimizing the biocompatibility of the dialysis procedure [59]. Although these factors may be further improved by new PMMA dialyzers, we acknowledge the lack of information on albumin dialysate loss in our study, which makes our previous statement speculative.

Conclusion and Clinical Implications
This study suggests that OL-HDF with PMMA BG-U series is highly effective for the removal of pCS, enabling an acceptable clearance of β2-microglobulin and small solutes. These results support the continuing use of hydrophobic and cationic adsorptive PMMA membranes as a good alternative in HD treatment, which could potentially enhance the clinical benefits in patients on renal replacement therapies. With an increasing number of dialyzer options, there is a need to further examine the clinical effects of removal of PBUTs on quality of life and survival in HD patients, whose life expectancy continues to be unacceptably low.

Appendix B.3. Serum Sample Preparation
Methanol deproteinization of blood samples was used to avoid hydrolysis of pCS by acids [61]. Serum samples and standard solutions were treated equally. In all cases, an aliquot of 500 µL was added to 1 mL of methanol and was then incubated at room temperature for 20 min. After that, the mixture was centrifugated for 10 min at 3500 rpm and the supernatant was collected. Cyano bonding cartridges (Discovery ® DSC-CN SPE Tube, bed weight 500 mg, volume 3 mL; Supelco, Bellefonte, PA, USA) were used for solid phase extraction. Before the extraction, cartridges were conditioned with 2 mL of methanol, centrifuged subsequently for 1 min at 3500 rpm, and followed by the addition of 2 mL of water and centrifuged 1 min at 3500 rpm. After the conditioning procedure, supernatant of the deproteinized samples or standard solutions was loaded into a cartridge and was centrifuged for 1 min at 3500 rpm. The cartridge was put inside a clean borosilicate tube with 1 mL of methanol. After being centrifuged for 1 min at 3500 rpm, the eluted fraction

Appendix B.3. Serum Sample Preparation
Methanol deproteinization of blood samples was used to avoid hydrolysis of pCS by acids [61]. Serum samples and standard solutions were treated equally. In all cases, an aliquot of 500 µL was added to 1 mL of methanol and was then incubated at room temperature for 20 min. After that, the mixture was centrifugated for 10 min at 3500 rpm and the supernatant was collected. Cyano bonding cartridges (Discovery ® DSC-CN SPE Tube, bed weight 500 mg, volume 3 mL; Supelco, Bellefonte, PA, USA) were used for solid phase extraction. Before the extraction, cartridges were conditioned with 2 mL of methanol, centrifuged subsequently for 1 min at 3500 rpm, and followed by the addition of 2 mL of water and centrifuged 1 min at 3500 rpm. After the conditioning procedure, supernatant of the deproteinized samples or standard solutions was loaded into a cartridge and was centrifuged for 1 min at 3500 rpm. The cartridge was put inside a clean borosilicate tube with 1 mL of methanol. After being centrifuged for 1 min at 3500 rpm, the eluted fraction was collected and was evaporated at 37 • C under a vacuum of 600 mm Hg for 60 min in a Heidolph Synthesis 1 Multi-evaporator (Heidolph Instruments GmbH & Co.KG, Schwabach, Germany). Dry residue was reconstituted with 150 µL of mobile phase and was finally transferred into an HPLC vial for analysis.
Appendix B.4. HPLC Analytical Methodology pCS was analyzed by HPLC using an Agilent Technologies 1100 liquid chromatograph with a quaternary pump, a diode array detector, a thermostatted column compartment, an autosampler, and an HP Compaq computer equipped with Agilent-Chemstation software (Agilent Technologies, Santa Clara, CA, USA). The chromatographic separations were performed on a Kromasil ® RP C18 analytical column (150 mm length × 4.6 mm i.d., 5 µm particle diameter; Análisis Vínicos, Spain). The samples (20 µL each) were injected through a Rheodyne valve (Rheodyne, Cotati, CA, USA). The flow rate was set to 1 mL/min, temperature to 25 • C, and fluorescence detection with 214 nm for excitation and 306 nm for emission and detection [60]. Mobile phase was composed of 50 mM formic acid and methanol. An elution gradient was necessary: t = 0 min, formic acid/methanol (65:35, v/v); t = 15 min, formic acid/methanol (25:75, v/v); t = 19 min, formic acid/methanol (65:35, v/v). The column was equilibrated for 30 min prior to injection of samples. The peak area of pCS was measured in each chromatogram. Retention time of pCS was 13 min.
Formic acid and methanol solutions were vacuum filtered through 0.45 µm nylon membranes (Micron Separations, Westboro, MA, USA) and sonicated prior to HPLC analysis. An SC2 analytical microbalance (Sartorius Mechatronics, S.A., Madrid, Spain) was used to weigh pCS.

Appendix B.5. Chromatographic Method Validation
The chromatographic method was validated according to the EMA [54] and FDA [55]. For each drug, linearity, accuracy, repeatability, intermediate precision, recovery, specificity, limit of detection and quantification, and system suitability were evaluated [62].
Linearity was demonstrated by analyzing the pCS standard solutions over the range 0.05-6.25 mg/mL; a calibration curve was performed by plotting peak area against drug concentration; the coefficient of determination (r2) was calculated. The selected concentrations covered the range of expected pCS serum concentrations in patients on dialysis, according to [63] and to our preliminary studies. Accuracy was determined by comparing mean estimated concentration with the nominal value at four pCS concentration levels (0.05, 0.52, 2.60, and 6.25 mg/mL). Relative errors (REs) were also calculated. Repeatability (intra-day assay precision) was determined by analyzing four pCS standards (0.05, 0.52, 2.60, and 6.25 mg/mL) twice and calculating the RSD for each concentration level. Intermediate precision (inter-day assay precision) was determined by analyzing four pCS standards (0.05, 0.52, 2.60, and 6.25 mg/mL) daily for two days and calculating the RSD for each concentration level. Specificity of the method was ascertained by evaluating the presence of interferences at the retention time of pCS. Limit of detection (LOD) and limit of quantification (LOQ): LOD and LOQ were calculated using the following equations: LOD = 3·σ/S and LOQ = 10·σ/S; where σ is the standard deviation of y-intercepts of regression lines and S is the slope of the calibration curve. System suitability specifications and tests (SSTs) were determined from ten replicate injections of pCS standard solutions of 0.05, 0.52, 2.60, and 6.25 mg/mL. Theoretical plates (N), tailing factor (T), resolution (Rs), and repeatability (RSD of retention time and area) were determined as the mean of the ten values obtained for each parameter.