Comparison of removal capacity of two consecutive generations of high-flux dialysers during different treatment modalities

Transcription

1 Nephrol Dial Transplant (2011) 26: doi: /ndt/gfq803 Advance Access publication 10 February 2011 Comparison of removal capacity of two consecutive generations of high-flux dialysers during different treatment modalities Natalie Meert 1, Sunny Eloot 1, Eva Schepers 1, Horst-Dieter Lemke 2, Annemieke Dhondt 1, Griet Glorieux 1, Maria Van Landschoot 1, Marie-Anne Waterloos 1 and Raymond Vanholder 1 1 Renal Division, University Hospital Gent, Gent, Belgium and 2 EXcorLab GmbH, Obernburg, Germany Correspondence and offprint requests to: Natalie Meert; natalie.meert@ugent.be Abstract Background. Innovative modifications have been introduced in several types of dialyser membranes to improve adequacy and permselectivity. Which aspects of removal are modified and how this relates to different diffusive or convective strategies has, however, been insufficiently investigated. Methods. In a prospective cross-over study, 14 chronic kidney disease (Stage 5D) patients were dialysed with a second-generation high-flux dialyser (Polynephronä) in comparison to a first-generation type (DIAPESÒ-HF800). Both dialysers were assessed in haemodialysis, in online pre-dilution and in post-dilution haemodiafiltration. Reduction ratio (RR, %) of small water-soluble compounds (urea and uric acid), low-molecular weight proteins (LMWPs) (b 2 -microglobulin, cystatin C, myoglobin and retinol-binding protein) and protein-bound solutes (hippuric acid, indole acetic acid, indoxylsulphate and p-cresylsulphate) was assessed, together with albumin losses into the dialysate. Results. Comparing the two types of membranes, the second-generation dialyser demonstrated a higher RR for LMWPs, whilst at the same time exhibiting lower albumin losses but only during post-dilution haemodiafiltration. No differences in RR were detected for both the small watersoluble and the protein-bound compounds. Comparing dialysis strategies, convection removed the same amount of solute or more as compared to diffusion. Conclusions. The second-generation membrane resulted in a higher removal of LMWPs compared to the first-generation membrane, but for the other solutes, differences were less prominent. Convection was superior in removal of a broad range of uraemic retention solutes especially with the firstgeneration membrane. Keywords: haemodiafiltration; haemodialysis; removal; uraemic toxins Introduction In end-stage renal failure [Stage 5D chronic kidney disease (CKD)], high-flux dialysis strategies are used to mimic as adequately as possible removal by the natural glomeruli, which clear solutes approximately up to the molecular weight (MW) range of albumin (67 kda). Consequently, only compounds with a MW smaller than albumin represent a target for direct elimination by non-selective extracorporeal renal replacement therapies. The removal of low-molecular weight proteins (LMWPs) (peptides >500 Da but smaller than albumin, formerly named middle molecules ) has become one of the targets of modern highefficacy dialysis strategies. A current trend in dialysis membrane engineering is to maximize the permeability for larger LMWPs while retaining albumin. Membranes that leak substantial amounts of albumin do not meet these requirements. Particularly in convective procedures, such as haemodiafiltration, their albumin leakage is high [1]. Ó The Author Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved. For Permissions, please journals.permissions@oup.com

2 Removal study of new-generation high-flux dialyser 2625 In function of this awareness, innovative approaches have been introduced in the manufacturing of dialysis membranes [2, 3], e.g. the application of advanced spinning procedures, which enables that the innermost skin layer of the membrane can be modulated at the nanoscale level, resulting in uniformly sized, larger but homogenously distributed pores with a sharp cut-off. Their effect on solute sieving is a steeper sieving curve for LMWPs in absence of increased albumin leakage. Polynephronä (Nipro Corp., Osaka, Japan), one of the membranes under evaluation in the present study, is an example of such a new-generation dialysis membrane manufactured based on the above principles. Several questions regarding the removal capacity of these new-generation membranes can be raised: (i) how do these new membranes behave as compared to their equivalents of the previous generation; (ii) what is the relative gain in efficacy for the different physico-chemical types of uraemic solute (small water-soluble, proteinbound, LMWPs) [4] and (iii) is there a difference in behaviour in function of the applied dialysis strategy? The present study was undertaken to answer the above questions, applying in the same patients two membranes made from the same polyethersulfone backbone, either the first-generation DIAPESÒ-HF800 membrane or the secondgeneration Polynephronä membrane in a haemodialysis (HD), a pre-dilution haemodiafiltration (pre-hdf) and a post-dilution haemodiafiltration (post-hdf) setting. The reduction ratio (RR) of a broad range of uraemic retention solutes was evaluated. of dialyser type but not that of the dialysis mode was randomized. Before the start of the study and before changing the dialyser type a wash-out period of two weeks consisting of low-flux dialysis (FX 10; Fresenius Medical Care, Bad Homburg, Germany) was performed. Between the reallocation of dialysis modes, one wash-out week was always included. The study was approved by the local ethics committee and was registered in clinicaltrials.gov (NCT ). Written informed consent was obtained from all participants. Systems and treatment strategies All treatments were performed with Fresenius 4008H and 5008H (Fresenius Medical Care) dialysis machines. Treatment and patient characteristics are illustrated in Table 2. Sample collection and laboratory analysis Samples were collected during the third (mid-week) session of each respective period. Pre- and post-dialysis blood samples were drawn from the inlet line before starting the blood pump (pre-dialysis) and exactly 30 s after setting the blood pump at 50 ml/min (post-dialysis). Continuous partial sampling of spent dialysate was carried out with a collection pump inserted into the dialysate outlet line via a special connector. At the end of the dialysis session a sample was collected after stirring. Blood and dialysate samples were collected on ice. Blood samples were centrifuged (3000 r.p.m. for 10 min) and all samples were stored at 80 C until analysis. Urea (MW: 60.1 Da) was measured by a standard clinical analyser. b 2 - microglobulin (b 2 m) (MW: 11.8 kda), cystatin C (MW: 13.3 kda), myoglobin (MW: 17.6 kda), free retinol-binding protein (MW: 21.2 kda) and albumin were determined using an immune nephelometric assay (BN ProSpec; Siemens Dade-Behring, Marburg, Germany). Uric acid (MW: Da), hippuric acid (MW: Da), indole-3-acetic acid (MW: Da), indoxylsulphate (MW: Da) and p-cresylsulphate (MW: Da) were analysed with reversed phase high-performance liquid chromatography, as previously reported [5]. Only the total concentration of the protein-bound solutes (hippuric acid, indole acetic acid, indoxylsulphate and p-cresylsulphate) was determined. Materials and methods Patients Fourteen stable, adult Stage 5D kidney disease patients (10 males, 4 females, mean age years), who had been on thrice weekly maintenance dialysis for at least 6 months were enrolled in the study. Mean dialysis duration was months. The primary renal diagnoses were diabetic nephropathy (n ¼ 5), chronic glomerulonephritis (n ¼ 3), renal vascular disease (n ¼ 2), renal cortical necrosis (n ¼ 1), post-partum shock (n ¼ 1) and cause unknown (n ¼ 2). Study design The study was performed with a prospective, cross-over, randomized open-label design, which is illustrated in Figure 1. Patients were randomized for two dialysers containing two different membrane types based on the same polyethersulfone polymer: PES-170DS (containing the DIAPESÒ-HF800 membrane; Membrana GmbH, Wuppertal, Germany) and ELISIOÒ-170H 1 (containing the Polynephronä membrane), both produced by Nipro Corp. Membrane characteristics are illustrated in Table 1. Each patient received one week of three consecutive treatments with HD, pre-hdf and post-hdf with each of the two dialysers. The order Calculations and statistics For the protein-bound compounds and LMWPs, concentration postdialysis (C post(c) ) was corrected for extracellular volume changes based on differences in the patient s pre- (BW pre ) and post-dialysis body weight (BW post ) C post/corr ¼ C post /(1 1 ((BW pre BW post )/0.2 BW post )) [6]. RR was calculated according to the formula RR (%) ¼ ((C 0 C post(c) )/ C 0 ) 3 100, where C is concentration [either before or after treatment (C 0 and C post )]. Albumin losses into dialysate were calculated according to the equation: albumin losses (g) ¼ C Alb (Q D 1 Q UF ) 3 t, where C Alb is albumin concentration in a representative sample of spent dialysate, where (Q D 1 Q UF ) is the sum of dialysate flow, infusion flow and ultrafiltration flow (correction of interdialytic weight gain). Statistical comparisons were made among the two membrane types and among the three dialysis strategies. Statistical evaluation was performed with GraphPad Prism 4 (Graph- Pad software, Inc., USA, CA). Data were checked for normality with Kolmogorov Smirnov test. Data are represented as mean SD. Comparative statistical analysis of differences between treatment modes was performed by one-way analysis of variance with correction for multiple Fig. 1. Flow chart of the study.

3 2626 N. Meert et al. comparisons. Comparison between dialysers was performed with paired t-test. A P-value of <0.05 was considered statistically significant. DIAPESÒ-HF800 only in the HD mode and superiority of convection to diffusion only for DIAPESÒ-HF800. Results For all compounds determined, RRs were evaluated comparing (i) dialyser type and (ii) dialysis strategy per solute group. Water-soluble compounds RRs of small water-soluble compounds (urea and uric acid) ranged from 71 to 80%. Comparing the two dialysers, urea RR (Figure 2A) was only significantly higher with the Polynephronä in the HD mode. For uric acid (Figure 2B), the RRs were similar for the two dialysers in all three treatment modes. Comparing the dialysis strategies, only for the DIAPESÒ-HF800 dialyser significant differences were found which were for urea in the order post-hdf > pre- HDF > HD (Figure 2A). For uric acid, only the difference between HD (71%) and post-hdf (75%) reached significance (Figure 2B). In summary, there are only subtle differences for the small water-soluble compounds, with slightly but significantly more removal of urea with Polynephronä versus Table 1. Membrane characteristics a DIAPESÒ-HF800 Polynephronä Effective length (mm) Surface area (m 2 ) UF coeff (ml/h/mmhg) Membrane thickness (lm) SC b 2 m b SC myoglobin b SC albumin b Average pore diameter (nm) a UF coeff, ultrafiltration coefficient; SC, sieving coefficient. b EN1283 human blood (haematocrit 32%, total protein 60 g/l); blood flow, 300 ml/min; filtration flow, 60 ml/min. Pore size determined from sieving curves. Protein-bound compounds RRs of the protein-bound compounds [hippuric acid (50% bound), indole acetic acid (65% bound), indoxylsulphate (90% bound) and p-cresylsulphate (95% bound)] were in the range of 35 to 69%. Comparing the two dialysers, there were no differences between the two membranes (Figure 2C F). When comparing the dialysis strategies, for hippuric acid and indole acetic acid, there were also no differences noted (Figure 2C and D). However, for indoxylsulphate as well as for p-cresylsulphate, RR with both dialysers was higher in HDF compared to HD without difference among convective strategies (Figure 2E and F). In summary, for protein-bound compounds, the secondgeneration membrane offers no advantage. However, when comparing strategies, convection is superior to diffusion with both membranes, without difference between both convective strategies. Low-molecular weight proteins RRs of the four LMWPs were as follows: b 2 m (MW: 11.8 kda) 59 to 75%, cystatin C (MW: 13.3 kda) 55 to 77%, myoglobin (MW: 17.6 kda) 38 to 76% and retinol-binding protein (MW: 21.2 kda) 12 to 31%. Comparing the two dialysers, the RR of LMWPs was higher with Polynephronä for all treatment modes as compared to DIAPESÒ-HF800 (Figure 2G J). Comparing the dialysis strategies, for b 2 m and cystatin C (Figure 2G and H), the RR for Polynephronä in both HDF treatment modes was higher as compared to HD. Both HDF modes were not different from each other. For DIAPESÒ- HF800, the differences between all treatment modes were significant and in the order post-hdf > pre-hdf > HD. For myoglobin (Figure 2I), the RR for both dialysers in post-hdf was higher as compared to the respective RRs in pre-hdf and HD. Pre-HDF and HD, however, were not different from each other. For retinol-binding protein, there were no differences between the three treatment modes for both filters (Figure 2J). Table 2. Dialysis and patient characteristics a DIAPESÒ-HF800 Polynephronä HD Pre-HDF Post-HDF HD Pre-HDF Post-HDF Qb,eff (ml/min) Q D (ml/min) Qinf (ml/min) t (min) UFV (L) BW pre (kg) BW post (kg) Hct pre (%) a Qb,eff, effective blood flow; Q D, dialysate flow; Qinf, infusion flow; t, treatment duration; UFV, ultrafiltration volume; BW pre, body weight pre-dialysis; BW post, body weight post-dialysis; Hct pre, haematocrit pre-dialysis.

4 Removal study of new-generation high-flux dialyser 2627 Fig. 2. RR (%, mean 6 SD, n ¼ 14) of the water-soluble compounds (A: urea, B: uric acid), protein-bound solutes (C: hippuric acid, D: indole acetic acid, E: indoxylsulphate, F: p-cresylsulphate) and LMWPs (G: b 2 -microglobulin, H: cystatin C, I: myoglobin, J: retinol-binding protein) in HD, pre-hdf and post-hdf. DIAPESÒ-HF800 data are illustrated by the white bars and Polynephronä by the grey bars. Comparison of dialysers: *P < 0.05, **P < 0.01, ***P < versus DIAPESÒ-HF800. Comparison of dialysis strategies: with DIAPESÒ-HF800: P < 0.05, P < 0.01, P < versus HD; P < 0.01, P < versus pre-hdf; with Polynephronä: # P < 0.05, ## P < 0.01, ### P < versus HD; cxcx P < 0.01 versus pre-hdf. In summary, there is a distinct superiority of the secondgeneration membrane in removal of LMWPs. Comparing dialysis strategies, convection is superior to diffusion. For the first-generation membrane, post-dilution is superior to pre-dilution. For the second-generation membrane, this superiority of post-hdf reached statistical significance only for the solute myoglobin. Albumin The mean loss of albumin in dialysate per session varied between and g (Figure 3). Between both dialyser types, there were no differences, unless in the post-dilution HDF mode where Polynephronä caused less albumin losses in comparison with DIAPESÒ- HF800. When comparing strategies, post-dilution HDF

5 2628 N. Meert et al. Fig. 3. Albumin losses into dialysate (mean 6 SD, n ¼ 14). DIAPESÒ- HF800 data are illustrated by the white bars and Polynephronä by the grey bars. Comparison of dialysers: *P < 0.05 versus DIAPESÒ-HF800. Comparison of dialysis strategies: with DIAPESÒ-HF800: P < versus HD; P < versus pre-hdf; with Polynephronä: ## P < 0.01versus HD; cxcx P < 0.01 versus pre-hdf. induced more albumin losses than the two other modalities. Discussion This study evaluates the impact of up-to-date polymer chemistry and membrane manufacturing on the removal of uraemic solutes during diffusive and convective dialysis strategies. On one hand, two high-flux membranes that are composed of the same polymer but manufactured with different spinning techniques were compared. As compared to the first-generation membrane (DIAPESÒ-HF800), the second generation (Polynephronä) has thicker capillary walls (40 versus 30 lm), a larger pore size (78 versus 75 Å) and narrower distribution of pore size, altogether aiming at a better elimination of large molecules without higher losses of albumin. On the other hand, both dialysers are tested in three different dialysis strategies (HD, pre-hdf and post-hdf). Our main findings are (i) removal is superior with the second-generation versus first-generation high-flux dialyser for the removal of LMWPs irrespective of the dialysis strategy (Figure 2G J), whereas no marked differences in removal were found for small water-soluble or proteinbound compounds (Figure 2A F); (ii) albumin losses are smaller with the second-generation dialyser during post- HDF and similar for the other strategies (Figure 3); (iii) convective strategies are slightly superior to HD for the protein-bound compounds with the highest binding (Figure 2E and F) and the LMWPs (Figure 2G J). The superiority of the second-generation dialysers in removing LMWPs highlights that not all large pore membranes should be considered the same, even if pore size or the polymers from which they are made are identical; this was suggested also by a previous study, which was, however, based on filters from two different manufacturers [7, 8]. Based on our results, we can in addition stress that the new-generation dialyser results in a higher LMWPs removal (steeper sieving curve), without substantially affecting albumin losses. Comparisons with previous studies calculating RR of LMWPs are rather difficult because of the often considerably differing treatment parameters and should be interpreted with care. In the literature, we found several studies but we selected two representative studies to compare our data [9, 10]. In comparison to the study of Krieter et al., we can conclude that the RR of b 2 -microglobulin and cystatin C were similar with the RRs obtained with the Polynephronä in our study. However, for the larger LMWPs (myoglobin and retinol-binding protein), the Polynephronä resulted in 5 16% higher RRs. Compared to the study of Maduell et al., we can conclude that the RR of b 2 -microglobulin is only slightly higher (4 8%) with Polynephronä. The RR of myoglobin was markedly higher (13 37%) with Polynephronä. For the LMWPs, differences among convective strategies (post-dilution > pre-dilution) are more pronounced for the first-generation dialyser (Figure 2G I). For the second-generation dialyser, post-hdf is superior compared to the other techniques only for myoglobin. For retinol-binding protein, the differences among dialysis strategies did not reach significance, probably due to the fact that this molecule is not well removed during the treatments since it forms a complex with transthyretin in plasma, resulting in a MW of 76 kda [11]. It should be noted that the effective length of the newgeneration dialyser is longer than that of the older one, which conceivably could result in increased internal filtration and back filtration. This could partly explain the superior RR of LMWPs with Polynephronä compared to DIAPESÒ-HF800, even in a HD setting. For the protein-bound compounds no benefit of the newgeneration dialyser could be demonstrated. Based on these findings, it might be assumed that only the free fraction of these solutes is eliminated essentially by diffusion. If this is indeed the mechanism at play, the distribution and structure of pores would have little impact on removal of these solutes which all have a low MW. One could hypothesize that loss of protein-bound solutes into dialysate is to a large extent related to losses together with albumin. Hence, since DIAPESÒ-HF800 was linked to more substantial albumin losses during post-hdf, one might conclude that this effect might have been responsible at least in part for the lack of difference between both membranes. However, in a previous study, no correlation was found between the amount of albumin and that of several protein-bound solutes lost into the dialysate [12]. In addition, the calculated amount of solute lost bound to albumin was <2% of the total amount of each molecule lost into dialysate (unpublished data, Rita De Smet). Another mechanism that could play a role in removal is solute adsorption to the membrane. The study design, however, does not allow the differentiation in the importance of different aspects of removal. On the other hand, we found an increase in removal with convection compared to diffusion. These findings are in accordance with the results of a previous study from our group where we demonstrated superiority of post-hdf compared to high-flux dialysis in lowering the pre-dialysis concentration of the protein-bound compounds with the highest binding, such as p-cresylsulphate [13]. In contrast, Krieter et al. [9] could not detect a difference in removal of

6 Removal study of new-generation high-flux dialyser 2629 the protein-bound solutes, p-cresylsulphate and indoxylsulphate, between post-hdf and high-flux HD. We do not have a direct explanation for this discrepancy, although it may be that the study of Krieter et al. was under powered (n ¼ 8) to detect a difference. Again, in agreement with the results of a previous study [5], we could not demonstrate a difference in removal of protein-bound solutes between pre-dilution and post-dilution HDF. Since the main elimination mechanism of the small water-soluble compounds is diffusion, neither a reorganization of the pore structure nor the introduction of convection could cause a substantial increase in their removal. It should be noted that the Q D applied in HD (500 ml/ min) is lower in comparison to the Q D in HDF (800 ml/ min). However, this is likely to be essentially important for the clearance of small solutes [14]. The results should also be interpreted with care since the calculated RR may overestimate true removal for solutes with multicompartmental kinetics. The question could be raised here of to what extent would the increased removal of the study solutes impact on clinical outcomes. Outcome superiority of large pore (high-flux) membranes in more effectively removing LMWPs than membranes with smaller pores has been demonstrated in a number of secondary analyses of large controlled trials [15 17] and in the patient group with low serum albumin (<4 g/dl) of the membrane permeability outcomes study [18, 19]. In a subanalysis of the HEMOstudy, pre-dialysis b 2 -microglobulin averaged over the entire observation period was inversely correlated to outcomes [20, 21]. For this study, it is difficult to predict the clinical relevance since it was developed as an acute study focussing on the reduction of serum concentrations during a dialysis session. Whether the new-generation dialysers have a positive impact on clinical outcome can only be demonstrated by long-term clinical studies using a similar set-up. Transmembrane albumin losses were similar, or for post-hdf, even inferior to those with the first-generation membrane, corroborating the alleged steep decline of the sieving curve once the size of albumin is reached. The pathophysiologic impact of transmembrane albumin losses during dialysis remains currently unclear. In any case, it can be assumed that they contribute to protein depletion and hence may be linked to malnutrition, a current problem encountered in a large proportion of the dialysis population [22]. Nevertheless, the question can be raised whether hypoalbuminaemia in CKD is really related to albumin losses, rather than to defects in albumin synthesis [23]. Conclusions Second-generation large pore dialysers have a superior removal capacity for the LMWPs without increased albumin loss; independent of the dialyser used (first or second generation), post-dilution haemodiafiltration is the most adequate dialysis strategy with regard to the removal of a broad range of uraemic retention solutes. Acknowledgements. The authors thank the nursing staff for their continuing assistance in the management of the patients. Transparency declarations. This study was supported by the Nipro Corporation, Japan. This work was partially funded by a governmental research grant from the Bijzonder Onderzoeksfonds (BOF, grant no ) and by the European Uraemic Toxin (EUTox) Work Group. EUTox is a Consortium of European researchers involved in studies and reviews related to uraemic toxicity. It was created under the auspices of the European Society for Artificial Organs and is a work group of the European Renal Association European Dialysis and Transplantation Association (ERA-EDTA) and is composed by 24 research groups throughout Europe. The research group of R.V. has been supported by research grants from Fresenius Medical Care, Baxter, Gambro and Bellco. S.E. is working as post-doctoral fellow for the Belgian Fund for Scientific Research Flanders (FWO). The results presented in this paper have not been published previously in whole or part, except in abstract format. References 1. Ward RA. Protein-leaking membranes for hemodialysis: a new class of membranes in search of an application? J Am Soc Nephrol 2005; 16: Ronco C, Nissenson AR. Does nanotechnology apply to dialysis? Blood Purif 2001; 19: Ronco C, Bowry SK, Brendolan A et al. Hemodialyzer: from macrodesign to membrane nanostructure; the case of the FX-class of hemodialyzers. Kidney Int Suppl 2002; 80: Vanholder R, De Smet R, Glorieux G et al. Review on uremic toxins: classification, concentration, and interindividual variability. Kidney Int 2003; 63: Meert N, Eloot S, Waterloos MA et al. Effective removal of proteinbound uraemic solutes by different convective strategies: a prospective trial. Nephrol Dial Transplant 2009; 24: Bergstrom J, Wehle B. No change in corrected beta 2-microglobulin concentration after cuprophane haemodialysis. Lancet 1987; 1: Ouseph R, Hutchison CA, Ward RA. Differences in solute removal by two high-flux membranes of nominally similar synthetic polymers. Nephrol Dial Transplant 2008; 23: Vanholder R, Pedrini LA. All high-flux membranes are equal but some high-flux membranes are less equal than others. Nephrol Dial Transplant 2008; 23: Krieter DH, Hackl A, Rodriguez A et al. Protein-bound uraemic toxin removal in haemodialysis and post-dilution haemodiafiltration. Nephrol Dial Transplant 2010; 25: Maduell F, Navarro V, Cruz MC et al. Osteocalcin and myoglobin removal in on-line hemodiafiltration versus low- and high-flux hemodialysis. Am J Kidney Dis 2002; 40: Raghu P, Sivakumar B. Interactions amongst plasma retinol-binding protein, transthyretin and their ligands: implications in vitamin A homeostasis and transthyretin amyloidosis. Biochim Biophys Acta 2004; 1703: De Smet R, Dhondt A, Eloot S et al. Effect of the super-flux cellulose triacetate dialyser membrane on the removal of non-protein-bound and protein-bound uraemic solutes. Nephrol Dial Transplant 2007; 22: Meert N, Waterloos MA, Van Landschoot M et al. Prospective evaluation of the change of pre-dialysis protein-bound uremic solute concentration with post-dilution on-line hemodiafiltration. Artif Organs 2010; 34: Bhimani JP, Ouseph R, Ward RA. Effect of increasing dialysate flow rate on diffusive mass transfer of urea, phosphate and {beta}2- microglobulin during clinical haemodialysis. Nephrol Dial Transplant 2010; 25(12): Chauveau P, Nguyen H, Combe C et al. Dialyzer membrane permeability and survival in hemodialysis patients. Am J Kidney Dis 2005; 45: Cheung AK, Levin NW, Greene T et al. Effects of high-flux hemodialysis on clinical outcomes: results of the HEMO study. J Am Soc Nephrol 2003; 14: Krane V, Krieter DH, Olschewski M et al. Dialyzer membrane characteristics and outcome of patients with type 2 diabetes on maintenance hemodialysis. Am J Kidney Dis 2007; 49:

7 2630 N. Meert et al. 18. Locatelli F, Hannedouche T, Jacobson S et al. The effect of membrane permeability on ESRD: design of a prospective randomised multicentre trial. J Nephrol 1999; 12: Locatelli F, Martin-Malo A, Hannedouche T et al. Effect of membrane permeability on survival of hemodialysis patients. J Am Soc Nephrol 2009; 20: Cheung AK, Rocco MV, Yan G et al. Serum beta-2 microglobulin levels predict mortality in dialysis patients: results of the HEMO study. J Am Soc Nephrol 2006; 17: Cheung AK, Greene T, Leypoldt JK et al. Association between serum 2-microglobulin level and infectious mortality in hemodialysis patients. Clin J Am Soc Nephrol 2008; 3: Stenvinkel P. Malnutrition and chronic inflammation as risk factors for cardiovascular disease in chronic renal failure. Blood Purif 2001; 19: Kaysen GA, Rathore V, Shearer GC et al. Mechanisms of hypoalbuminemia in hemodialysis patients. Kidney Int 1995; 48: Received for publication: ; Accepted in revised form: