Beta-2 microglobulin clearance in high-flux dialysis and convective dialysis modalities: a meta-analysis of published studies

Beta-2 microglobulin clearance in high-flux dialysis and convective dialysis modalities: a... ABSTRACT Background Recent meta-analyses suggest that higher removal of beta-2 microglobulin (β2M) with either high-flux (HFD) dialysis or hemodiafiltration (HDF) may be associated with decreased total and cardiovascular mortality in dialysis patients. However, there are limited data about the performance of high flux dialyzers and/or convective therapies in removing β2M. Methods This is a random effects meta-analysis and meta-regression of data extracted from randomized controlled trials and observational studies in hemodialysis, hemofiltration and HDF regarding the efficiency of high flux dialyzers to remove β2M. Studies were searched using ProQuest in SCOPUS, EMBASE and MEDLINE. Results We included 69 studies from 1 January 2001 to 12 June 2017 on 1879 patients with 6771 available measurements. Average β2M clearance was 48.75 mL/min [95% confidence interval (CI) 42.50–55.21] for conventional HF dialysis, and 87.06 mL/min (95% CI 75.08–99.03) for convective therapies (hemofiltration and HDF) with substantial heterogeneity among studies [P (Q) ≤ 0.001]. In multivariable meta-regression analyses, we found significantly higher β2M clearance for polyarylethersulfone dialyzers when used for HFD and polysulfone membranes in convective therapies. However, the mass of β2M removed into the dialysate did not depend on membrane material. Adjusted dialysate-side (−22.279, 95% CI  −9.8 to −34.757, P < 0.001) β2M clearances were significantly lower than whole blood clearances, suggesting that adsorption contributes substantially to β2M removal. Higher Kuf, blood flow and substitution fluid rates but not dialysate flow rates were associated with statistically significant and clinically meaningful elevation in β2M clearance from the body independent of the dialysis modality. Conclusions Membrane composition and characteristics, modality (convective versus diffusive), blood flow rates and substitution fluid rates in HDF play a significant role in the efficient removal of β2M from the body in both diffusive and convective dialysis. beta-2 microglobulin, clearance, hemodiafiltration, hemofiltration, high-flux hemodialysis INTRODUCTION The accumulation of middle molecular weight solutes, such as beta-2 microglobulin (β2M), is toxic to various body tissues and has been associated with adverse cardiovascular and infectious outcomes among patients with end-stage renal disease (ESRD) [1, 2]. β2M precipitates and forms fibrillary structures and amyloid deposits in bones, periarticular tissues [3], vessel walls and internal organs, especially the heart [4–7]. Dialysis-related amyloidosis and other disorders associated with abnormal β2M accumulation and function [8] are clinically silent, develop early in the development and progression of chronic kidney disease (CKD) and may even imply a potential causal link with the highly prevalent cardiovascular disease (CVD) in ESRD patients [9, 10]. Several meta-analyses of randomized controlled trials (RCTs) in conventional dialysis suggest that high-flux dialyzers, which more efficiently remove β2M than their low-flux (LF) counterparts, are associated with improved cardiovascular outcomes [11, 12]. Convective therapies, including hemodiafiltration (HDF) and hemofiltration, achieve even higher middle molecule clearances relative to HF dialysis. These therapies may improve the chronic retention of β2M over time noted with thrice-weekly HFD [5, 11, 13]. In these modalities, clearance is a function of the total volume of solution utilized (both dialysate flow rate and replacement solution). A recent individual patient-level meta-analysis of published RCTs suggests that the higher clearance from the body achieved by these therapies may result in clinically and statistically significant improvement in total and cardiovascular mortality relative to conventional HFD [14, 15]. Nevertheless, the quality of the evidence and the putative effects of convective dialysis have been called into question by large collaborative aggregate level meta-analyses by the Cochrane Group [16, 17] and others [18, 19]. The interpretation of these contradicting analyses of data outcomes is complicated by the limited evidence synthesis of the performance and the determinants of β2M clearance by high flux dialyzers when the latter are used in conventional or convective forms of renal replacement therapies. The aforementioned meta-analyses have reported only on a limited number of studies that examined dialyzer clearance or β2M mass removal, focusing instead on reduction ratios as the sole measure of dialyzer performance. None of the aforementioned studies has attempted to analyze the impact of different dialysis configurations (e.g. membrane material, surface area, substitution fluid rate) on multiple measures of β2M body removal. This literature gap limits our ability to better understand the performance of these therapies, and how best to modify treatment parameters to optimize clearance of middle molecules, thus moving beyond urea-centric approaches that have been widely used in modern dialysis. To do so, we conducted a meta-analysis of data about the performance of HFD and/or convective dialysis therapies to remove β2M. We included studies published between 2001 and 2017, covering the period in which the landmark RCTs in HFD [13, 20] and HDF [21–23] were published. MATERIALS AND METHODS This is a meta-analysis of data collected in RCTs and observational studies in hemodialysis (HD) about the performance (ability) of HFD and convective therapies (HDF or hemofiltration, HF) to remove β2M from the body. The focus of this meta-analysis was on studies that could provide determinations of β2M ‘clearance from the body’ as the primary outcome measure of dialysis procedure performance. Search strategy The overarching search strategy for this meta-analysis was to include studies that had employed formal methods to characterize dialytic performance. Our initial focus was on studies published from 1 January 2001 to 31 December 2013. The date range was determined to capture performance of dialyzers that were likely used in the main outcomes trials in HFD and HDF. Subsequently, we extended the search for articles up to 12 June 2017. The search was based on free text and MeSH terms (see Text Query in Supplementary data). Articles were searched by using ProQuest in two databases (EMBASE and MEDLINE) for the initial query and only in MEDLINE from 1 January 2014 and onwards as we did not have access to ProQuest after that date. We used the SCOPUS database to compile a list of citations from, as well as citations to, the articles considered relevant after abstract and full text review of the initial search. Articles in this citation analysis were also subjected to abstract and full text review as detailed below. Inclusion and exclusion criteria for abstract review Eligible studies reported in vivo measurements of β2M clearance from the body (primary outcome of this meta-analysis). Second, we examined β2M reduction ratio and/or β2M mass removal from the body in human subjects receiving HFD, HDF or hemofiltration among the studies reporting β2M clearance measurements. Studies performed before 2001, in vitro studies, review studies and meta-analyses were excluded along with studies not involving extracorporeal circuits (e.g. peritoneal dialysis), mathematical simulations without experimental data, and studies on extracorporeal circuits perfused in a closed loop manner with non-blood fluid (crystalloid or colloid) or ex vivo blood. Process Two reviewers (M.-E.R. and G.T.) independently screened potentially relevant titles and abstracts to ensure that the identified studies met the inclusion criteria and none of the exclusion criteria. Then the abstract review was adjudicated by C.P.A. All adjudicated papers were selected for full text review by M.-E.R. and C.P.A. to ensure they met the full text inclusion criteria for the meta-analysis. Full text review for papers written in Chinese was performed by Y.-H.N. and Z.X. Abstract and full text criteria are provided in the Supplementary data. Citation analysis was carried out by M.-E.R. and C.P.A. using the same abstract and full text criteria as the initial search. Data extraction We did not restrict articles by language. Data for the articles in English were extracted from tables and figures by M.-E.R. and C.P.A. Information from non-English publications was extracted from the abstract and the tables in the text. Data for the articles in Chinese were extracted from tables and figures by Y.-H.N. and Z.X. All data were inserted into standardized data collection forms and imported into an Excel spreadsheet. Measurements extracted included: (i) kinetic parameters [type of therapy, flow pump parameters, membrane surface area (MSA), dialyzer material, dialysis session duration, ultrafiltration volumes, session frequency] and (ii) β2M body clearance measurements, mass removal and reduction ratios. Volumes infused and ultra-filtered were converted from L to mL/min to account for the confounding role of dialysis session duration on convective clearance. For studies for which we had individual patient-level data (i.e. HEMO), we aggregated measurements to distinct groups defined by the type of dialyzer used, prior to analysis. Dialyzer specifications (Kuf: ultrafiltration coefficient, MSA) were downloaded from the manufacturer’s brochures and if those were not available (e.g. discontinued products), from dialysis textbooks and articles in the literature. Quality assessment Quality metrics of the included studies were assessed independently by two reviewers (C.P.A. and M.-E.R.) using the Effective Public Health Project Quality Assessment Tool for Quantitative Studies (EPHPP) (see Table S1) [24]. This tool was developed by the Effective Public Health Project, Canada and was chosen because it covers any quantitative study design. The latter was a particularly desirable feature for our project, which included RCTs, non-randomized controlled and uncontrolled studies. This quality assessment tool is comprised of the following components: selection bias, study design, confounders, blinding, data collection methods, withdrawals and dropouts, intervention integrity and analyses. Each section is rated as strong, moderate or weak by each reviewer. At the end, a global rating for the meta-analysis is provided. Statistical analysis Most of the studies included, reported on multiple ‘configurations’, i.e. combinations of dialysis operational parameters (e.g. pump flow rates, infusion volume, dialyzers) in the same patient groups. For this meta-analysis, a multi-level random effects model was adopted to account for clustering of measurements within the same configurations and within the same study. Despite the computational complexity, this approach is conceptually similar to using a paired t-test for the analysis of matched sample data. One subtle feature of this approach is that it enforces a form of averaging of multiple measurements from the same study. For studies reporting instantaneous clearance values, this implies that our object of analysis is the average of the instantaneous clearances. This quantity may not be much different from the average clearance computed via other means (e.g. pooled dialysate samples or pre-post β2M measurements), even though the individual measurements averaged may be far from it, e.g. due to loss of dialyzer performance over time. We opted for this approach, because we feel that the clinically relevant quantity is the capacity of the dialyzer to remove β2M over the entire course of the treatment (average clearance) rather than at any given point in time. This modeling was conducted separately for studies of convective and diffusive therapies reporting β2M clearance and together for studies of convective and diffusive therapies reporting β2M mass removal. Clearance values, reduction ratios and mass removal of β2M were summarized and heterogeneity was assessed graphically by the use of forest plots. Meta-regression models were utilized to statistically assess heterogeneity. For these models, the same multi-level structure was used as the one that was used to generate the forest plots. Univariate meta-regressions, assessing each variable in isolation, were followed by multivariable meta-regressions adjusting for more than one study characteristics. Variables were selected by univariate meta-regression analyses at the level of P = 0.05 if >70% of the studies were available for these analyses. The Restricted Maximum Likelihood (REML) approach was used to derive unbiased point estimates of dialysis relevant parameters (themselves treated as fixed effects) but at the expense of wider confidence intervals (CIs) for these models. Analysis of variance (ANOVA) was used to assess the global statistical significance of study characteristics with more than two levels (e.g. type of dialysis procedure) by comparing models that adjusted for these characteristics versus the models that did not. ANOVA tests were carried out in models fitted with conventional Maximum Likelihood approach, since these tests cannot be applied to compare models with different fixed effects specifications when REML is used. Operational parameters of clinical interest (e.g. substitution volume flows or year of the study) were forced into the models even if not significant in univariate models. Secular trends in the performance of the dialyzers over time were assessed by including the year of the publication as a covariate in the models. In these analyses, 2001 was taken as Year 0 and the secular trend was defined as a linear change in the outcome (e.g. clearance) with each subsequent year. Outcomes explored with meta-regression models were β2M clearance, β2M mass removal and the pre-dialysis and post-dialysis β2M reduction ratio. All analyses were performed in R statistical software (version 3.1.1) with the package metaphor [25]. RESULTS Study search results Electronic searches from 1 January 2001 to 12 June 2017 identified 638 potentially relevant reports. Of these, 481 were excluded after title and abstract review. After adjudication, 150 articles were selected for full text review and 53 relevant articles were identified (52 were published before 2014). Out of these, 47 articles reported aggregate (group data) and 5 studies reported patient-level data. In addition, the HEMO study (one of the studies identified in the initial search) provided data about 984 patients with 3967 measurements in non-reused dialyzers (most dialyzers were reused in HEMO). These measurements were taken from the HEMO analytic data files distributed by the National Institutes for Digestive Diabetes and Kidney Diseases (NIDDK), made available to our group through a data use agreement. Citation analysis of these 53 papers in SCOPUS identified 673 potentially relevant studies; we screened out 622 papers based on abstract review and selected 109 for full text review. Full text review uncovered 34 papers that had been identified during the initial search and 16 papers with relevant clearance data. A summary flow diagram is shown in Figure 1. The overall final study population for this meta-analysis consisted of 69 studies of 1879 patients with 6771 available measurements. FIGURE 1 View largeDownload slide Flow diagram of the literature search. FIGURE 1 View largeDownload slide Flow diagram of the literature search. Study characteristics Table 1 presents the characteristics of the patients that participated in the included studies, such as number of patients, age, gender, time on chronic dialysis therapy and their pre-dialysis weight. The same table details characteristics of the included studies, which fell into two main categories: comparisons of different types of dialyzers (46 on HFD) and comparisons of different types of convective dialysis therapies [31 studies on post-dilution HDF (post-HDF), 6 on pre-dilution HDF (pre-HDF), 15 on mid-dilution HDF (mid-HDF), 5 on mixed HDF (mixed-HDF), 2 studies on pre-dilution hemofiltration (pre-HF) and 2 studies on post-dilution hemofiltration (post-HF)]. These studies used a wide variety of dialyzer membrane material, e.g. cellulose acetate (CA, n = 4), polysulfone (PS, n = 146), polymethylmethacrylate (PMMA, n = 2), polyacrylnitrile (PAN, n = 2) and polyarylethersulfone (PAES, n = 97). All included studies enrolled patients under chronic dialysis regimens. Participant numbers were highly variable and ranged from 5 to 52. Only one study (HEMO [20]) had 984 participants. Clearances (mL/min), reduction ratios of β2M and/or β2M mass removal (mg or g/session) were measured and reported either in the blood side (serum or plasma) or in the dialysate side at a single time point during the dialysis session (instantaneous) or as average over the course of the treatment. A wide variety of methods were used for the calculation of clearance. The formulas and the numerical aspects of these approaches are summarized in the Supplementary data. Other study characteristics such as blood and dialysate flow rate, treatment duration, substitution fluid rate and MSA are reported as average and standard errors in Table 1. Table 1 Characteristics of the studies analyzed First Author Year N N meas Female Age Vintage PreWt Modality Material MSA QB QD Qinf Duration Leto [26] 2001 15 30 40.0% 45.7 (—) 156.3 (—) (—) HFD CA/PS 1.3 (0.1) 250.0 (0.0) 600.0 (0.0) (—) 240.0 (0.0) Xu [27] 2001 10 10 40.0% 70.2 (5.6) 71.2 (37.0) 63.9 (10.6) HFD PS 1.8 (0.5) (—) (—) (—) 300.0 (0.0) Yamada [28] 2001 28 28 39.0% 58.1 (16.4) 64.0 (47.0) 49.0 (8.0) HFD PS 1.42 (0.0) 188.0 (18.0) 500.0 (0.0) (—) 237.0 (18.0) Stiller [29] 2002 15 15 73.0% 54.3 (10.2) 134.0 (100.6) (—) HFD PAES/PS 1.24 (0.1) (—) (—) (—) 240.0 (0.0) Eknoyan [20] 2002 984 3967 59.0% 58.6 (13.7) 63.1 (59.2) 71.8 (1.5) HFD PMMA/CA/PS/ PAN/PAES 1.8 (0.2) 372.4 (8.8) 671.8 (10.1) (—) 204.5 (2.7) Ding [30] 2002 12 36 33.0% 49.7 (11.3) 83.5 (76.7) (—) pre-HDF/post-HDF PS 1.3 (0.0) 250.0 (0.0) 616.7 (2.9) 92.5 (3.0) 282.5 (29.6) Klingel [31] 2002 22 22 0.0% 61.4 (—) (—) 74.6 (0.0) HFD PS 1.3 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 228.8 (0.0) Mann [32] 2003 5 5 0.0% (—) (—) (—) HFD PS 1.6 (0.05) (—) (—) (—) 240.0 (0.0) Mandolfo [33] 2003 8 16 0.0% 61.4 (—) (—) 68 (8.6) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 550.0 (0.0) 30.0 (0.0) 240.0 (0.0) Pedrini [34] 2003 20 20 35.0% 63.0 (17.0) 116.4 (86.4) 60.3 (12.6) post-HDF/mixed-HDF PS 2.1 (0.0) 403.0 (55.9) 580.2 (36.6) 219.6 (36.6) 227.0 (17.7) Ward [35] 2003 12 24 41.6% 53.0 (13.0) 63.0 (18.3) (—) HFD PS 1.8 (0.0) 410.0 (1.9) 700.0 (0.0) (—) 228.0 (11.7) Bammens [36] 2004 14 70 28.6% 66.6 (3.1) 24.8 (10.0) 62.10 (1.94) HFD/pre-HDF/post-HDF PS 1.8 (0.0) 323.9 (116.3) 500.0 (0.0) 87.0 (0.0) 230.0 (0.0) Yamashita [37] 2004 5 5 80.0% (—) (—) (—) post-HF PS 1.8 (0.0) (—) (—) 84.2 (18.8) 120.0 (0.0) Emiliani [38] 2004 10 10 20.0% 66.0 (18.0) 80.0 (36.0) 66.2 (7.5) mid-HDF PAES 2.6 (0.0) 312.0 (18.0) 500.0 (0.0.0) 43.6 (7.2) 240.0 (10.0) Leypoldt [39] 2004 22 88 37.5% 61.0 (18.0) (—) 80.3 (19.4) HFD PS 1.77 (0.0) 338.0 (49.6) 540.0 (60.0) (—) 178.5 (19.0) Lucchi [40] 2004 10 20 40.0% 61.1 (8.9) 51.8 (35.9) (—) HFD/post-HDF PS 1.6 (0.0) 300.0 (0.0) 625.0 (0.0) 20.9 (0.0) 240.0 (0.0) Pisitkun [41] 2004 9 18 22% 48.0 (6.1) 51.4 (42.0) 55.2 (8.3) HFD/mid-HDF PS 2.7 (0.9) 475.0 (36.4) 800.0 (0.0) 59.9 (9.7) 240.0 (0.0) Tonelli [42] 2004 5 15 0.0% (—) (—) (—) HFD PS 1.8 (0.0) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Santoro [43] 2005 20 40 0.0% (—) (—) (—) HFD/mid-HDF PAES 1.90 (0.0) 363.0 (39.2) (—) 58.0 (11.9) (—) Brendolan [44] 2005 3 24 0.0% (—) (—) (—) HFD/post-HDF PS 2.2 (0.4) 333.3 (12.0) 500.0 (0.0) 16.6 (1.9) 226.1 (14.5) Padrini [45] 2005 11 22 36.4% 66.8 (11.9) 80.9 (66.9) 64.1 (9.2) post-HF/pre-HF PAES 2.1 (0.0) 327.8 (22.3) (—) 186 (40.1) 240 (6) Petras [46] 2005 6 36 0.0% 56.0 (16.0) 86.0 (50.0) (—) HFD/post-HDF/Pre-HF PAES 2.1 (0.0) 350.0 (0.0) 500.0 (0.0) 95.0 (0.0) 240.0 (0.0) Krieter [47] 2005a 5 5 60.0% 52.0 (22) (—) 68.5 (27.5) mid-HDF PAES 1.9 (0.0) 400.0 (0.0) 800.0 (0.0) 200.0 (0.0) 205.0 (15.0) Krieter [48] 2005b 10 40 30.0% 57.3 (13.7) 99.6 (92.4) 66.3 (10.4) mid-HDF/post-HDF PAES/PS 1.9 (0.1) 400.0 (0.0) 550.0 (0.0) 148.3 (2.9) 240 (23.4) Evenepoel [49] 2006 20 20 25.0% 68.8 (10.9) 19.3 (31.5) 59.9 (7.9) HFD PS 1.8 (0.0) 322.7 (21.6) 500.0 (0.0) (—) 230.0 (0.0) Mandolfo [50] 2006 12 18 66.7% 69.0 (9.0) 117.6 (69.6) 65.2 (8.1) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 700.0 (0.0) 50.0 (0.0) 240.0 (0.0) Nakashima [51] 2006 12 24 0.0% 49.1 (12.1) 127.2 (73.2) 66.8 (12.7) HFD PS 2.10 (0.0) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Panich [52] 2006 10 20 50.0% 58.2 (14.7) (—) 54.2 (3.8) HFD/post-HDF PS 1.8 (0.0) 425.3 (12.6) (—) 61.6 (0.6) 240.0 (0.0) Pedrini [53] 2006 12 72 25.0% 64.2 (6.6) 45.0 (38.0) 64.9 (11.2) mixed-HDF PS 2.1 (0.0) 422.0 (37.9) 609.0 (27.9) 178.0 (20.9) 218.0 (25.9) Potier [54] 2007 6 18 0.0% (—) (—) (—) post-HDF/pre-HDF/mid-HDF PAES 1.90 (0.0) 360.0 (0.0) 500.0 (0.0) 175.0 (0.0) (—) Feliciani [55] 2007 10 30 20.0% 64.7 (8.0) 54.7 (57.7) 73.25 (12.5) mixed-HDF/mid-HDF PAES/PS 1.85 (0.1) 385.5 (18.3) 609.0 (20.7) 167.5 (14.1) 231.5 (16.8) Krieter [56] 2007 8 32 62.5% 62.1 (13.8) 76.0 (55.3) 68.5 (7.1) HFD PS/PAES 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 236.0 (10.5) Santoro [57] 2007 8 16 50.0% 56.6 (23.6) (—) (—) mid-HDF PAES 1.9 (0.0) 306.5 (10.3) (—) 100.0 (0.0) 231.0 (10.3) Tiranathanagul [58] 2007 12 48 33.0% 54.2 (13.6) 42.0 (32.3) 62.85 (9.4) post-HDF/mid-HDF PS 2.7 (0.9) 416.7 (24.1) 800.0 (0.0) 113.0 (6.0) 240.0 (0.0) Abe [59] 2008 15 45 40% 65.5 (13.2) 72.9 (63.8) (—) HFD PMMA/CA/PS 1.5 (0.1) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Eloot [60] 2008 9 27 55.5% 71.0 (10.0) 19.0 (12.0) 79.0 (11.5) HFD PS 1.8 (0.0) 260.0 (0.0) 260.0 (0.0) (—) 360.0 (0.0) Mandolfo [61] 2008 8 16 37.5% 72.2 (4.8) 62.0 (24.0) 61.7 (11) HFD/mid-HDF PAES 1.9 (0.0) 251.5 (32.4) 700.0 (0.0) 56.0 (4.8) 240.0 (0.0) Spalding [62] 2008 12 12 50.0% 65.3 (12.9) (—) (—) HFD/post-HDF (—) (—) 358.4 (84.4) 800.0 (0.0) 37.5 (12.6) 197.4 (55.3) Krieter [63] 2008a 8 40 25.0% 64.0 (16.0) 70.0 (74.0) 74.2 (10.7) HFD/post-HDF PAES/PS 1.7 (0.1) 300.0 (0.0) 460.0 (0.0) 40.0 (0.0) 240.0 (0.0) Ouseph [64] 2008a 12 48 25.0% 57.0 (4.0) 52.0 (17.0) 81.3 (4.35) HFD PS/PAES 1.65 (0.1) 382.0 (4.8) 800.0 (0.0) (—) 219.0 (5.3) Ouseph [64] 2008b 12 60 41.6% 46.0 (3.0) 48.0 (8.0) 84.2 (6.75) HFD PS/PAES 1.90 (0.2) 404.0 (1.0) 800.0 (0.0) (—) 240.0 (0.0) Krieter [65] 2008b 8 48 37.5% 63.0 (14.0) 77.5 (38.9) 76.5 (11.15) HFD PAES/PS 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Joyeux [66] 2008 20 40 35.0% 58.0 (20.9) 46.0 (46.1) 71.5 (10.2) HFD/post-HDF PAES 2.10 (0.0) 310.5 (33.3) (—) 31.8 (5.2) 235.5 (14.6) Lee [67] 2009 8 16 50.0% 68.7 (19.1) 50.3 (53.1) (—) HFD PAES 1.1 (0.0) 325.0 (24.6) 500.0 (0.0) (—) 255.0 (14.8) Meert [68] 2009 14 42 50.0% 63.5 (17.0) 30.2 (36.0) (—) pre-HDF/pre-HF/post-HDF PAES 1.8 (0.2) 312.3 (15.6) 384.7 (5.0) 185.7 (20.7) 249.3 (13.0) Pedrini [69] 2009 15 90 20.0% 67.3 (8.7) 44.1 (20.8) 76.9 (13.8) mid-HDF PAES 2.1 (0.2) 378.5 (27.4) 599.5 (6.5) 167.5 (9.7) 223.0 (21.4) Susantitaphong [70] 2009 12 36 66.6% 59.5 (13.5) 81.6 (52.8) 57.5 (11.6) Pre-HDF/mid-HDF/post-HDF PAES 2.2 (0.1) 440.3 (19.9) 554.2 (10.4) 245.9 (2.1) 240.0 (0.0) Troidle [71] 2009 8 8 0.0% 45.0 (7.0) (—) (—) HFD PS 1.8 (0.0) 400.0 (0.0) 600.0 (0.0) (—) 480.0 (0.0) Wang [72] 2009 18 54 27.8% 46.9 (9.6) 52.5 (—) (—) HFD PS 1.5 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Bhimani [73] 2010 12 144 75.0% 60.0 (4.0) 81.0 (19.0) (—) HFD PS/PAES 1.7 (0.2) 400.0 (1.8) 550.0 (0.0) (—) (—) Gasco [74] 2010 16 263 50.0% 52.7 (—) (—) (—) post-HDF PS 1.8 (0.0) 323.0 (0.0) 800.0 (0.0) (—) 242.0 (0.0) Kohn [75] 2010 5 38 60.0% (—) 142.0 (60.9) 86.0 (29.0) HFD PAES (—) 425.0 (75.0) 200.0 (0.0) (—) 174.0 (15.0) Krieter [76] 2010 8 64 12.5% 63.0 (12.0) 81.8 (144.0) 70.8 (17.5) HFD/post-HDF PAES 1.9 (0.0) 378.0 (31.1) 500.0 (0.0) 47.0 (6.0) 229.0 (20.7) Park [77] 2010 52 52 52.0% 54.0 (12.4) 112.7 (188.6) (—) HFD PS 1.3 (0.0) 237.0 (23.0) 500.0 (0.0) (—) 240.0 (12.0) Basile [78] 2011 11 22 18.2% 54.1 (17.8) 78.0 (60.2) 69.1 (9.9) HFD PS 1.8 (0.0) 270.0 (0.0) 270.0 (0.0) (—) 469.1 (2.7) Ficheux [79] 2011 18 54 0.0% 79.7 (1.7) (—) 66.1 (2.3) HFD PS 2.2 (0.1) 318.0 (2.0) 500.0 (9.8) (—) 222.0 (2.9) Pedrini [80] 2011 15 60 33.3% 67.2 (8.3) (—) 73.1 (14.0) post-HDF/mid-HDF PS/PAES 2.3 (0.1) 374.0 (34.0) 580.0 (39.7) 147.5 (11.1) 224.0 (18.5) Panichi [81, 82] 2012 30 180 33.3% 55.9 (14.0) 58.0 (59.0) (—) post-HDF PAES 2.1 (0.0) 313.5 (32.7) 600.0 (0.0) 78.1 (0.0) 235.0 (13.8) Susantitaphong [83] 2012 12 48 66.6% 57.8 (14.8) 43.2 (42.0) 55.5 (11.1) mid-HDF/mixed-HDF PAES 2.2 (0.0) 425.0 (24.5) 600.0 (0.0) 200.0 (0.0) 240.0 (0.0) Tessitore [84] 2012 26 26 53.9% 63.0 (12.0) (—) (—) HFD PP 0.7 (0.0) 297.0 (32.0) 500.0 (0.0) (—) 230.0 (13.0) von Albertini [85] 2013 12 35 0.0% (—) (—) (—) HFD/post-HDF PAES/PS 1.8 (0.0) 417.1 (0.0) 667.4 (0.0) 30.7 (0.0) 206.3 (22.5) Heaf [86] 2013 12 96 30.0% 63.1 (11.7) 78.0 (52.8) 79.2 (17.8) HFD PAES 2.0 (0.0) 276.0 (38.7) 500.0 (0.0) (—) 240.0 (0.0) Melo [87] 2014 14 28 50.0% 48.9 (14.4) (—) 76.35 (19.63) HFD/post-HDF PS 2.0 (0.0) 375.0 (8.2) 760.0 (0.0) 40.0 (0.0) 115.7 (16.8) Pedrini [88] 2014 16 32 18.8% (—) (—) 77.60 (10.78) post-HDF PAES/PS 2.20 (0.10) 388.0 (25.9) 574.5 (39.0) 121.0 (11.9) 226.5 (13.7) Cornelis [89] 2014 13 52 23.1% 53.6 (20.4) 49.0 (29.0) (—) HFD/post-HDF PS 1.8 (0.0) 286.0 (4.8) 573.7 (13.1) 30.1 (1.4) 366.3 (3.9) Potier [90] 2016 6 24 66.7% 65.4 (25.5) 68.6 (43.7) 73.9 (2.1) HFD/post-HDF/ mixed-HDF/pre-HDF PS 2.3 (0.0) 339.4 (3.4) 600.0 (0.0) 122.1 (5.1) 240.0 (0.0) Gayrard [91] 2017 12 48 50.0% 73.0 (12.0) (—) 71.0 (1.9) HFD/post-HDF PS 1.8 (0.0) 366.3 (5.1) 602.3 (1.0) 51.8 (1.1) 233.6 (2.9) Kirsch [92] 2017 39 59 28.2% 60.5 (13.6) 63.1 (43.8) 80.2 (18.4) post-HDF/HFD PS 1.9 (0.1) 368.1 (12.8) (—) 27.5 (1.4) 252.4 (11.8) First Author Year N N meas Female Age Vintage PreWt Modality Material MSA QB QD Qinf Duration Leto [26] 2001 15 30 40.0% 45.7 (—) 156.3 (—) (—) HFD CA/PS 1.3 (0.1) 250.0 (0.0) 600.0 (0.0) (—) 240.0 (0.0) Xu [27] 2001 10 10 40.0% 70.2 (5.6) 71.2 (37.0) 63.9 (10.6) HFD PS 1.8 (0.5) (—) (—) (—) 300.0 (0.0) Yamada [28] 2001 28 28 39.0% 58.1 (16.4) 64.0 (47.0) 49.0 (8.0) HFD PS 1.42 (0.0) 188.0 (18.0) 500.0 (0.0) (—) 237.0 (18.0) Stiller [29] 2002 15 15 73.0% 54.3 (10.2) 134.0 (100.6) (—) HFD PAES/PS 1.24 (0.1) (—) (—) (—) 240.0 (0.0) Eknoyan [20] 2002 984 3967 59.0% 58.6 (13.7) 63.1 (59.2) 71.8 (1.5) HFD PMMA/CA/PS/ PAN/PAES 1.8 (0.2) 372.4 (8.8) 671.8 (10.1) (—) 204.5 (2.7) Ding [30] 2002 12 36 33.0% 49.7 (11.3) 83.5 (76.7) (—) pre-HDF/post-HDF PS 1.3 (0.0) 250.0 (0.0) 616.7 (2.9) 92.5 (3.0) 282.5 (29.6) Klingel [31] 2002 22 22 0.0% 61.4 (—) (—) 74.6 (0.0) HFD PS 1.3 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 228.8 (0.0) Mann [32] 2003 5 5 0.0% (—) (—) (—) HFD PS 1.6 (0.05) (—) (—) (—) 240.0 (0.0) Mandolfo [33] 2003 8 16 0.0% 61.4 (—) (—) 68 (8.6) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 550.0 (0.0) 30.0 (0.0) 240.0 (0.0) Pedrini [34] 2003 20 20 35.0% 63.0 (17.0) 116.4 (86.4) 60.3 (12.6) post-HDF/mixed-HDF PS 2.1 (0.0) 403.0 (55.9) 580.2 (36.6) 219.6 (36.6) 227.0 (17.7) Ward [35] 2003 12 24 41.6% 53.0 (13.0) 63.0 (18.3) (—) HFD PS 1.8 (0.0) 410.0 (1.9) 700.0 (0.0) (—) 228.0 (11.7) Bammens [36] 2004 14 70 28.6% 66.6 (3.1) 24.8 (10.0) 62.10 (1.94) HFD/pre-HDF/post-HDF PS 1.8 (0.0) 323.9 (116.3) 500.0 (0.0) 87.0 (0.0) 230.0 (0.0) Yamashita [37] 2004 5 5 80.0% (—) (—) (—) post-HF PS 1.8 (0.0) (—) (—) 84.2 (18.8) 120.0 (0.0) Emiliani [38] 2004 10 10 20.0% 66.0 (18.0) 80.0 (36.0) 66.2 (7.5) mid-HDF PAES 2.6 (0.0) 312.0 (18.0) 500.0 (0.0.0) 43.6 (7.2) 240.0 (10.0) Leypoldt [39] 2004 22 88 37.5% 61.0 (18.0) (—) 80.3 (19.4) HFD PS 1.77 (0.0) 338.0 (49.6) 540.0 (60.0) (—) 178.5 (19.0) Lucchi [40] 2004 10 20 40.0% 61.1 (8.9) 51.8 (35.9) (—) HFD/post-HDF PS 1.6 (0.0) 300.0 (0.0) 625.0 (0.0) 20.9 (0.0) 240.0 (0.0) Pisitkun [41] 2004 9 18 22% 48.0 (6.1) 51.4 (42.0) 55.2 (8.3) HFD/mid-HDF PS 2.7 (0.9) 475.0 (36.4) 800.0 (0.0) 59.9 (9.7) 240.0 (0.0) Tonelli [42] 2004 5 15 0.0% (—) (—) (—) HFD PS 1.8 (0.0) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Santoro [43] 2005 20 40 0.0% (—) (—) (—) HFD/mid-HDF PAES 1.90 (0.0) 363.0 (39.2) (—) 58.0 (11.9) (—) Brendolan [44] 2005 3 24 0.0% (—) (—) (—) HFD/post-HDF PS 2.2 (0.4) 333.3 (12.0) 500.0 (0.0) 16.6 (1.9) 226.1 (14.5) Padrini [45] 2005 11 22 36.4% 66.8 (11.9) 80.9 (66.9) 64.1 (9.2) post-HF/pre-HF PAES 2.1 (0.0) 327.8 (22.3) (—) 186 (40.1) 240 (6) Petras [46] 2005 6 36 0.0% 56.0 (16.0) 86.0 (50.0) (—) HFD/post-HDF/Pre-HF PAES 2.1 (0.0) 350.0 (0.0) 500.0 (0.0) 95.0 (0.0) 240.0 (0.0) Krieter [47] 2005a 5 5 60.0% 52.0 (22) (—) 68.5 (27.5) mid-HDF PAES 1.9 (0.0) 400.0 (0.0) 800.0 (0.0) 200.0 (0.0) 205.0 (15.0) Krieter [48] 2005b 10 40 30.0% 57.3 (13.7) 99.6 (92.4) 66.3 (10.4) mid-HDF/post-HDF PAES/PS 1.9 (0.1) 400.0 (0.0) 550.0 (0.0) 148.3 (2.9) 240 (23.4) Evenepoel [49] 2006 20 20 25.0% 68.8 (10.9) 19.3 (31.5) 59.9 (7.9) HFD PS 1.8 (0.0) 322.7 (21.6) 500.0 (0.0) (—) 230.0 (0.0) Mandolfo [50] 2006 12 18 66.7% 69.0 (9.0) 117.6 (69.6) 65.2 (8.1) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 700.0 (0.0) 50.0 (0.0) 240.0 (0.0) Nakashima [51] 2006 12 24 0.0% 49.1 (12.1) 127.2 (73.2) 66.8 (12.7) HFD PS 2.10 (0.0) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Panich [52] 2006 10 20 50.0% 58.2 (14.7) (—) 54.2 (3.8) HFD/post-HDF PS 1.8 (0.0) 425.3 (12.6) (—) 61.6 (0.6) 240.0 (0.0) Pedrini [53] 2006 12 72 25.0% 64.2 (6.6) 45.0 (38.0) 64.9 (11.2) mixed-HDF PS 2.1 (0.0) 422.0 (37.9) 609.0 (27.9) 178.0 (20.9) 218.0 (25.9) Potier [54] 2007 6 18 0.0% (—) (—) (—) post-HDF/pre-HDF/mid-HDF PAES 1.90 (0.0) 360.0 (0.0) 500.0 (0.0) 175.0 (0.0) (—) Feliciani [55] 2007 10 30 20.0% 64.7 (8.0) 54.7 (57.7) 73.25 (12.5) mixed-HDF/mid-HDF PAES/PS 1.85 (0.1) 385.5 (18.3) 609.0 (20.7) 167.5 (14.1) 231.5 (16.8) Krieter [56] 2007 8 32 62.5% 62.1 (13.8) 76.0 (55.3) 68.5 (7.1) HFD PS/PAES 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 236.0 (10.5) Santoro [57] 2007 8 16 50.0% 56.6 (23.6) (—) (—) mid-HDF PAES 1.9 (0.0) 306.5 (10.3) (—) 100.0 (0.0) 231.0 (10.3) Tiranathanagul [58] 2007 12 48 33.0% 54.2 (13.6) 42.0 (32.3) 62.85 (9.4) post-HDF/mid-HDF PS 2.7 (0.9) 416.7 (24.1) 800.0 (0.0) 113.0 (6.0) 240.0 (0.0) Abe [59] 2008 15 45 40% 65.5 (13.2) 72.9 (63.8) (—) HFD PMMA/CA/PS 1.5 (0.1) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Eloot [60] 2008 9 27 55.5% 71.0 (10.0) 19.0 (12.0) 79.0 (11.5) HFD PS 1.8 (0.0) 260.0 (0.0) 260.0 (0.0) (—) 360.0 (0.0) Mandolfo [61] 2008 8 16 37.5% 72.2 (4.8) 62.0 (24.0) 61.7 (11) HFD/mid-HDF PAES 1.9 (0.0) 251.5 (32.4) 700.0 (0.0) 56.0 (4.8) 240.0 (0.0) Spalding [62] 2008 12 12 50.0% 65.3 (12.9) (—) (—) HFD/post-HDF (—) (—) 358.4 (84.4) 800.0 (0.0) 37.5 (12.6) 197.4 (55.3) Krieter [63] 2008a 8 40 25.0% 64.0 (16.0) 70.0 (74.0) 74.2 (10.7) HFD/post-HDF PAES/PS 1.7 (0.1) 300.0 (0.0) 460.0 (0.0) 40.0 (0.0) 240.0 (0.0) Ouseph [64] 2008a 12 48 25.0% 57.0 (4.0) 52.0 (17.0) 81.3 (4.35) HFD PS/PAES 1.65 (0.1) 382.0 (4.8) 800.0 (0.0) (—) 219.0 (5.3) Ouseph [64] 2008b 12 60 41.6% 46.0 (3.0) 48.0 (8.0) 84.2 (6.75) HFD PS/PAES 1.90 (0.2) 404.0 (1.0) 800.0 (0.0) (—) 240.0 (0.0) Krieter [65] 2008b 8 48 37.5% 63.0 (14.0) 77.5 (38.9) 76.5 (11.15) HFD PAES/PS 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Joyeux [66] 2008 20 40 35.0% 58.0 (20.9) 46.0 (46.1) 71.5 (10.2) HFD/post-HDF PAES 2.10 (0.0) 310.5 (33.3) (—) 31.8 (5.2) 235.5 (14.6) Lee [67] 2009 8 16 50.0% 68.7 (19.1) 50.3 (53.1) (—) HFD PAES 1.1 (0.0) 325.0 (24.6) 500.0 (0.0) (—) 255.0 (14.8) Meert [68] 2009 14 42 50.0% 63.5 (17.0) 30.2 (36.0) (—) pre-HDF/pre-HF/post-HDF PAES 1.8 (0.2) 312.3 (15.6) 384.7 (5.0) 185.7 (20.7) 249.3 (13.0) Pedrini [69] 2009 15 90 20.0% 67.3 (8.7) 44.1 (20.8) 76.9 (13.8) mid-HDF PAES 2.1 (0.2) 378.5 (27.4) 599.5 (6.5) 167.5 (9.7) 223.0 (21.4) Susantitaphong [70] 2009 12 36 66.6% 59.5 (13.5) 81.6 (52.8) 57.5 (11.6) Pre-HDF/mid-HDF/post-HDF PAES 2.2 (0.1) 440.3 (19.9) 554.2 (10.4) 245.9 (2.1) 240.0 (0.0) Troidle [71] 2009 8 8 0.0% 45.0 (7.0) (—) (—) HFD PS 1.8 (0.0) 400.0 (0.0) 600.0 (0.0) (—) 480.0 (0.0) Wang [72] 2009 18 54 27.8% 46.9 (9.6) 52.5 (—) (—) HFD PS 1.5 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Bhimani [73] 2010 12 144 75.0% 60.0 (4.0) 81.0 (19.0) (—) HFD PS/PAES 1.7 (0.2) 400.0 (1.8) 550.0 (0.0) (—) (—) Gasco [74] 2010 16 263 50.0% 52.7 (—) (—) (—) post-HDF PS 1.8 (0.0) 323.0 (0.0) 800.0 (0.0) (—) 242.0 (0.0) Kohn [75] 2010 5 38 60.0% (—) 142.0 (60.9) 86.0 (29.0) HFD PAES (—) 425.0 (75.0) 200.0 (0.0) (—) 174.0 (15.0) Krieter [76] 2010 8 64 12.5% 63.0 (12.0) 81.8 (144.0) 70.8 (17.5) HFD/post-HDF PAES 1.9 (0.0) 378.0 (31.1) 500.0 (0.0) 47.0 (6.0) 229.0 (20.7) Park [77] 2010 52 52 52.0% 54.0 (12.4) 112.7 (188.6) (—) HFD PS 1.3 (0.0) 237.0 (23.0) 500.0 (0.0) (—) 240.0 (12.0) Basile [78] 2011 11 22 18.2% 54.1 (17.8) 78.0 (60.2) 69.1 (9.9) HFD PS 1.8 (0.0) 270.0 (0.0) 270.0 (0.0) (—) 469.1 (2.7) Ficheux [79] 2011 18 54 0.0% 79.7 (1.7) (—) 66.1 (2.3) HFD PS 2.2 (0.1) 318.0 (2.0) 500.0 (9.8) (—) 222.0 (2.9) Pedrini [80] 2011 15 60 33.3% 67.2 (8.3) (—) 73.1 (14.0) post-HDF/mid-HDF PS/PAES 2.3 (0.1) 374.0 (34.0) 580.0 (39.7) 147.5 (11.1) 224.0 (18.5) Panichi [81, 82] 2012 30 180 33.3% 55.9 (14.0) 58.0 (59.0) (—) post-HDF PAES 2.1 (0.0) 313.5 (32.7) 600.0 (0.0) 78.1 (0.0) 235.0 (13.8) Susantitaphong [83] 2012 12 48 66.6% 57.8 (14.8) 43.2 (42.0) 55.5 (11.1) mid-HDF/mixed-HDF PAES 2.2 (0.0) 425.0 (24.5) 600.0 (0.0) 200.0 (0.0) 240.0 (0.0) Tessitore [84] 2012 26 26 53.9% 63.0 (12.0) (—) (—) HFD PP 0.7 (0.0) 297.0 (32.0) 500.0 (0.0) (—) 230.0 (13.0) von Albertini [85] 2013 12 35 0.0% (—) (—) (—) HFD/post-HDF PAES/PS 1.8 (0.0) 417.1 (0.0) 667.4 (0.0) 30.7 (0.0) 206.3 (22.5) Heaf [86] 2013 12 96 30.0% 63.1 (11.7) 78.0 (52.8) 79.2 (17.8) HFD PAES 2.0 (0.0) 276.0 (38.7) 500.0 (0.0) (—) 240.0 (0.0) Melo [87] 2014 14 28 50.0% 48.9 (14.4) (—) 76.35 (19.63) HFD/post-HDF PS 2.0 (0.0) 375.0 (8.2) 760.0 (0.0) 40.0 (0.0) 115.7 (16.8) Pedrini [88] 2014 16 32 18.8% (—) (—) 77.60 (10.78) post-HDF PAES/PS 2.20 (0.10) 388.0 (25.9) 574.5 (39.0) 121.0 (11.9) 226.5 (13.7) Cornelis [89] 2014 13 52 23.1% 53.6 (20.4) 49.0 (29.0) (—) HFD/post-HDF PS 1.8 (0.0) 286.0 (4.8) 573.7 (13.1) 30.1 (1.4) 366.3 (3.9) Potier [90] 2016 6 24 66.7% 65.4 (25.5) 68.6 (43.7) 73.9 (2.1) HFD/post-HDF/ mixed-HDF/pre-HDF PS 2.3 (0.0) 339.4 (3.4) 600.0 (0.0) 122.1 (5.1) 240.0 (0.0) Gayrard [91] 2017 12 48 50.0% 73.0 (12.0) (—) 71.0 (1.9) HFD/post-HDF PS 1.8 (0.0) 366.3 (5.1) 602.3 (1.0) 51.8 (1.1) 233.6 (2.9) Kirsch [92] 2017 39 59 28.2% 60.5 (13.6) 63.1 (43.8) 80.2 (18.4) post-HDF/HFD PS 1.9 (0.1) 368.1 (12.8) (—) 27.5 (1.4) 252.4 (11.8) N, number; N meas, number of measurements; Vintage, time on chronic intermittent dialysis in months; PreWt, pre-dialysis weight in kilograms; MSA, membrane surface area (in square meters); QB, blood flow rate (mL/min); QD, dialysis fluid flow rate (mL/min); Duration, the dialysis session (in min). For each parameter the table summarizes the mean and the SD over all arms in each study or a (—) if the relevant parameter could not be extracted from the paper. Table 1 Characteristics of the studies analyzed First Author Year N N meas Female Age Vintage PreWt Modality Material MSA QB QD Qinf Duration Leto [26] 2001 15 30 40.0% 45.7 (—) 156.3 (—) (—) HFD CA/PS 1.3 (0.1) 250.0 (0.0) 600.0 (0.0) (—) 240.0 (0.0) Xu [27] 2001 10 10 40.0% 70.2 (5.6) 71.2 (37.0) 63.9 (10.6) HFD PS 1.8 (0.5) (—) (—) (—) 300.0 (0.0) Yamada [28] 2001 28 28 39.0% 58.1 (16.4) 64.0 (47.0) 49.0 (8.0) HFD PS 1.42 (0.0) 188.0 (18.0) 500.0 (0.0) (—) 237.0 (18.0) Stiller [29] 2002 15 15 73.0% 54.3 (10.2) 134.0 (100.6) (—) HFD PAES/PS 1.24 (0.1) (—) (—) (—) 240.0 (0.0) Eknoyan [20] 2002 984 3967 59.0% 58.6 (13.7) 63.1 (59.2) 71.8 (1.5) HFD PMMA/CA/PS/ PAN/PAES 1.8 (0.2) 372.4 (8.8) 671.8 (10.1) (—) 204.5 (2.7) Ding [30] 2002 12 36 33.0% 49.7 (11.3) 83.5 (76.7) (—) pre-HDF/post-HDF PS 1.3 (0.0) 250.0 (0.0) 616.7 (2.9) 92.5 (3.0) 282.5 (29.6) Klingel [31] 2002 22 22 0.0% 61.4 (—) (—) 74.6 (0.0) HFD PS 1.3 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 228.8 (0.0) Mann [32] 2003 5 5 0.0% (—) (—) (—) HFD PS 1.6 (0.05) (—) (—) (—) 240.0 (0.0) Mandolfo [33] 2003 8 16 0.0% 61.4 (—) (—) 68 (8.6) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 550.0 (0.0) 30.0 (0.0) 240.0 (0.0) Pedrini [34] 2003 20 20 35.0% 63.0 (17.0) 116.4 (86.4) 60.3 (12.6) post-HDF/mixed-HDF PS 2.1 (0.0) 403.0 (55.9) 580.2 (36.6) 219.6 (36.6) 227.0 (17.7) Ward [35] 2003 12 24 41.6% 53.0 (13.0) 63.0 (18.3) (—) HFD PS 1.8 (0.0) 410.0 (1.9) 700.0 (0.0) (—) 228.0 (11.7) Bammens [36] 2004 14 70 28.6% 66.6 (3.1) 24.8 (10.0) 62.10 (1.94) HFD/pre-HDF/post-HDF PS 1.8 (0.0) 323.9 (116.3) 500.0 (0.0) 87.0 (0.0) 230.0 (0.0) Yamashita [37] 2004 5 5 80.0% (—) (—) (—) post-HF PS 1.8 (0.0) (—) (—) 84.2 (18.8) 120.0 (0.0) Emiliani [38] 2004 10 10 20.0% 66.0 (18.0) 80.0 (36.0) 66.2 (7.5) mid-HDF PAES 2.6 (0.0) 312.0 (18.0) 500.0 (0.0.0) 43.6 (7.2) 240.0 (10.0) Leypoldt [39] 2004 22 88 37.5% 61.0 (18.0) (—) 80.3 (19.4) HFD PS 1.77 (0.0) 338.0 (49.6) 540.0 (60.0) (—) 178.5 (19.0) Lucchi [40] 2004 10 20 40.0% 61.1 (8.9) 51.8 (35.9) (—) HFD/post-HDF PS 1.6 (0.0) 300.0 (0.0) 625.0 (0.0) 20.9 (0.0) 240.0 (0.0) Pisitkun [41] 2004 9 18 22% 48.0 (6.1) 51.4 (42.0) 55.2 (8.3) HFD/mid-HDF PS 2.7 (0.9) 475.0 (36.4) 800.0 (0.0) 59.9 (9.7) 240.0 (0.0) Tonelli [42] 2004 5 15 0.0% (—) (—) (—) HFD PS 1.8 (0.0) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Santoro [43] 2005 20 40 0.0% (—) (—) (—) HFD/mid-HDF PAES 1.90 (0.0) 363.0 (39.2) (—) 58.0 (11.9) (—) Brendolan [44] 2005 3 24 0.0% (—) (—) (—) HFD/post-HDF PS 2.2 (0.4) 333.3 (12.0) 500.0 (0.0) 16.6 (1.9) 226.1 (14.5) Padrini [45] 2005 11 22 36.4% 66.8 (11.9) 80.9 (66.9) 64.1 (9.2) post-HF/pre-HF PAES 2.1 (0.0) 327.8 (22.3) (—) 186 (40.1) 240 (6) Petras [46] 2005 6 36 0.0% 56.0 (16.0) 86.0 (50.0) (—) HFD/post-HDF/Pre-HF PAES 2.1 (0.0) 350.0 (0.0) 500.0 (0.0) 95.0 (0.0) 240.0 (0.0) Krieter [47] 2005a 5 5 60.0% 52.0 (22) (—) 68.5 (27.5) mid-HDF PAES 1.9 (0.0) 400.0 (0.0) 800.0 (0.0) 200.0 (0.0) 205.0 (15.0) Krieter [48] 2005b 10 40 30.0% 57.3 (13.7) 99.6 (92.4) 66.3 (10.4) mid-HDF/post-HDF PAES/PS 1.9 (0.1) 400.0 (0.0) 550.0 (0.0) 148.3 (2.9) 240 (23.4) Evenepoel [49] 2006 20 20 25.0% 68.8 (10.9) 19.3 (31.5) 59.9 (7.9) HFD PS 1.8 (0.0) 322.7 (21.6) 500.0 (0.0) (—) 230.0 (0.0) Mandolfo [50] 2006 12 18 66.7% 69.0 (9.0) 117.6 (69.6) 65.2 (8.1) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 700.0 (0.0) 50.0 (0.0) 240.0 (0.0) Nakashima [51] 2006 12 24 0.0% 49.1 (12.1) 127.2 (73.2) 66.8 (12.7) HFD PS 2.10 (0.0) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Panich [52] 2006 10 20 50.0% 58.2 (14.7) (—) 54.2 (3.8) HFD/post-HDF PS 1.8 (0.0) 425.3 (12.6) (—) 61.6 (0.6) 240.0 (0.0) Pedrini [53] 2006 12 72 25.0% 64.2 (6.6) 45.0 (38.0) 64.9 (11.2) mixed-HDF PS 2.1 (0.0) 422.0 (37.9) 609.0 (27.9) 178.0 (20.9) 218.0 (25.9) Potier [54] 2007 6 18 0.0% (—) (—) (—) post-HDF/pre-HDF/mid-HDF PAES 1.90 (0.0) 360.0 (0.0) 500.0 (0.0) 175.0 (0.0) (—) Feliciani [55] 2007 10 30 20.0% 64.7 (8.0) 54.7 (57.7) 73.25 (12.5) mixed-HDF/mid-HDF PAES/PS 1.85 (0.1) 385.5 (18.3) 609.0 (20.7) 167.5 (14.1) 231.5 (16.8) Krieter [56] 2007 8 32 62.5% 62.1 (13.8) 76.0 (55.3) 68.5 (7.1) HFD PS/PAES 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 236.0 (10.5) Santoro [57] 2007 8 16 50.0% 56.6 (23.6) (—) (—) mid-HDF PAES 1.9 (0.0) 306.5 (10.3) (—) 100.0 (0.0) 231.0 (10.3) Tiranathanagul [58] 2007 12 48 33.0% 54.2 (13.6) 42.0 (32.3) 62.85 (9.4) post-HDF/mid-HDF PS 2.7 (0.9) 416.7 (24.1) 800.0 (0.0) 113.0 (6.0) 240.0 (0.0) Abe [59] 2008 15 45 40% 65.5 (13.2) 72.9 (63.8) (—) HFD PMMA/CA/PS 1.5 (0.1) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Eloot [60] 2008 9 27 55.5% 71.0 (10.0) 19.0 (12.0) 79.0 (11.5) HFD PS 1.8 (0.0) 260.0 (0.0) 260.0 (0.0) (—) 360.0 (0.0) Mandolfo [61] 2008 8 16 37.5% 72.2 (4.8) 62.0 (24.0) 61.7 (11) HFD/mid-HDF PAES 1.9 (0.0) 251.5 (32.4) 700.0 (0.0) 56.0 (4.8) 240.0 (0.0) Spalding [62] 2008 12 12 50.0% 65.3 (12.9) (—) (—) HFD/post-HDF (—) (—) 358.4 (84.4) 800.0 (0.0) 37.5 (12.6) 197.4 (55.3) Krieter [63] 2008a 8 40 25.0% 64.0 (16.0) 70.0 (74.0) 74.2 (10.7) HFD/post-HDF PAES/PS 1.7 (0.1) 300.0 (0.0) 460.0 (0.0) 40.0 (0.0) 240.0 (0.0) Ouseph [64] 2008a 12 48 25.0% 57.0 (4.0) 52.0 (17.0) 81.3 (4.35) HFD PS/PAES 1.65 (0.1) 382.0 (4.8) 800.0 (0.0) (—) 219.0 (5.3) Ouseph [64] 2008b 12 60 41.6% 46.0 (3.0) 48.0 (8.0) 84.2 (6.75) HFD PS/PAES 1.90 (0.2) 404.0 (1.0) 800.0 (0.0) (—) 240.0 (0.0) Krieter [65] 2008b 8 48 37.5% 63.0 (14.0) 77.5 (38.9) 76.5 (11.15) HFD PAES/PS 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Joyeux [66] 2008 20 40 35.0% 58.0 (20.9) 46.0 (46.1) 71.5 (10.2) HFD/post-HDF PAES 2.10 (0.0) 310.5 (33.3) (—) 31.8 (5.2) 235.5 (14.6) Lee [67] 2009 8 16 50.0% 68.7 (19.1) 50.3 (53.1) (—) HFD PAES 1.1 (0.0) 325.0 (24.6) 500.0 (0.0) (—) 255.0 (14.8) Meert [68] 2009 14 42 50.0% 63.5 (17.0) 30.2 (36.0) (—) pre-HDF/pre-HF/post-HDF PAES 1.8 (0.2) 312.3 (15.6) 384.7 (5.0) 185.7 (20.7) 249.3 (13.0) Pedrini [69] 2009 15 90 20.0% 67.3 (8.7) 44.1 (20.8) 76.9 (13.8) mid-HDF PAES 2.1 (0.2) 378.5 (27.4) 599.5 (6.5) 167.5 (9.7) 223.0 (21.4) Susantitaphong [70] 2009 12 36 66.6% 59.5 (13.5) 81.6 (52.8) 57.5 (11.6) Pre-HDF/mid-HDF/post-HDF PAES 2.2 (0.1) 440.3 (19.9) 554.2 (10.4) 245.9 (2.1) 240.0 (0.0) Troidle [71] 2009 8 8 0.0% 45.0 (7.0) (—) (—) HFD PS 1.8 (0.0) 400.0 (0.0) 600.0 (0.0) (—) 480.0 (0.0) Wang [72] 2009 18 54 27.8% 46.9 (9.6) 52.5 (—) (—) HFD PS 1.5 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Bhimani [73] 2010 12 144 75.0% 60.0 (4.0) 81.0 (19.0) (—) HFD PS/PAES 1.7 (0.2) 400.0 (1.8) 550.0 (0.0) (—) (—) Gasco [74] 2010 16 263 50.0% 52.7 (—) (—) (—) post-HDF PS 1.8 (0.0) 323.0 (0.0) 800.0 (0.0) (—) 242.0 (0.0) Kohn [75] 2010 5 38 60.0% (—) 142.0 (60.9) 86.0 (29.0) HFD PAES (—) 425.0 (75.0) 200.0 (0.0) (—) 174.0 (15.0) Krieter [76] 2010 8 64 12.5% 63.0 (12.0) 81.8 (144.0) 70.8 (17.5) HFD/post-HDF PAES 1.9 (0.0) 378.0 (31.1) 500.0 (0.0) 47.0 (6.0) 229.0 (20.7) Park [77] 2010 52 52 52.0% 54.0 (12.4) 112.7 (188.6) (—) HFD PS 1.3 (0.0) 237.0 (23.0) 500.0 (0.0) (—) 240.0 (12.0) Basile [78] 2011 11 22 18.2% 54.1 (17.8) 78.0 (60.2) 69.1 (9.9) HFD PS 1.8 (0.0) 270.0 (0.0) 270.0 (0.0) (—) 469.1 (2.7) Ficheux [79] 2011 18 54 0.0% 79.7 (1.7) (—) 66.1 (2.3) HFD PS 2.2 (0.1) 318.0 (2.0) 500.0 (9.8) (—) 222.0 (2.9) Pedrini [80] 2011 15 60 33.3% 67.2 (8.3) (—) 73.1 (14.0) post-HDF/mid-HDF PS/PAES 2.3 (0.1) 374.0 (34.0) 580.0 (39.7) 147.5 (11.1) 224.0 (18.5) Panichi [81, 82] 2012 30 180 33.3% 55.9 (14.0) 58.0 (59.0) (—) post-HDF PAES 2.1 (0.0) 313.5 (32.7) 600.0 (0.0) 78.1 (0.0) 235.0 (13.8) Susantitaphong [83] 2012 12 48 66.6% 57.8 (14.8) 43.2 (42.0) 55.5 (11.1) mid-HDF/mixed-HDF PAES 2.2 (0.0) 425.0 (24.5) 600.0 (0.0) 200.0 (0.0) 240.0 (0.0) Tessitore [84] 2012 26 26 53.9% 63.0 (12.0) (—) (—) HFD PP 0.7 (0.0) 297.0 (32.0) 500.0 (0.0) (—) 230.0 (13.0) von Albertini [85] 2013 12 35 0.0% (—) (—) (—) HFD/post-HDF PAES/PS 1.8 (0.0) 417.1 (0.0) 667.4 (0.0) 30.7 (0.0) 206.3 (22.5) Heaf [86] 2013 12 96 30.0% 63.1 (11.7) 78.0 (52.8) 79.2 (17.8) HFD PAES 2.0 (0.0) 276.0 (38.7) 500.0 (0.0) (—) 240.0 (0.0) Melo [87] 2014 14 28 50.0% 48.9 (14.4) (—) 76.35 (19.63) HFD/post-HDF PS 2.0 (0.0) 375.0 (8.2) 760.0 (0.0) 40.0 (0.0) 115.7 (16.8) Pedrini [88] 2014 16 32 18.8% (—) (—) 77.60 (10.78) post-HDF PAES/PS 2.20 (0.10) 388.0 (25.9) 574.5 (39.0) 121.0 (11.9) 226.5 (13.7) Cornelis [89] 2014 13 52 23.1% 53.6 (20.4) 49.0 (29.0) (—) HFD/post-HDF PS 1.8 (0.0) 286.0 (4.8) 573.7 (13.1) 30.1 (1.4) 366.3 (3.9) Potier [90] 2016 6 24 66.7% 65.4 (25.5) 68.6 (43.7) 73.9 (2.1) HFD/post-HDF/ mixed-HDF/pre-HDF PS 2.3 (0.0) 339.4 (3.4) 600.0 (0.0) 122.1 (5.1) 240.0 (0.0) Gayrard [91] 2017 12 48 50.0% 73.0 (12.0) (—) 71.0 (1.9) HFD/post-HDF PS 1.8 (0.0) 366.3 (5.1) 602.3 (1.0) 51.8 (1.1) 233.6 (2.9) Kirsch [92] 2017 39 59 28.2% 60.5 (13.6) 63.1 (43.8) 80.2 (18.4) post-HDF/HFD PS 1.9 (0.1) 368.1 (12.8) (—) 27.5 (1.4) 252.4 (11.8) First Author Year N N meas Female Age Vintage PreWt Modality Material MSA QB QD Qinf Duration Leto [26] 2001 15 30 40.0% 45.7 (—) 156.3 (—) (—) HFD CA/PS 1.3 (0.1) 250.0 (0.0) 600.0 (0.0) (—) 240.0 (0.0) Xu [27] 2001 10 10 40.0% 70.2 (5.6) 71.2 (37.0) 63.9 (10.6) HFD PS 1.8 (0.5) (—) (—) (—) 300.0 (0.0) Yamada [28] 2001 28 28 39.0% 58.1 (16.4) 64.0 (47.0) 49.0 (8.0) HFD PS 1.42 (0.0) 188.0 (18.0) 500.0 (0.0) (—) 237.0 (18.0) Stiller [29] 2002 15 15 73.0% 54.3 (10.2) 134.0 (100.6) (—) HFD PAES/PS 1.24 (0.1) (—) (—) (—) 240.0 (0.0) Eknoyan [20] 2002 984 3967 59.0% 58.6 (13.7) 63.1 (59.2) 71.8 (1.5) HFD PMMA/CA/PS/ PAN/PAES 1.8 (0.2) 372.4 (8.8) 671.8 (10.1) (—) 204.5 (2.7) Ding [30] 2002 12 36 33.0% 49.7 (11.3) 83.5 (76.7) (—) pre-HDF/post-HDF PS 1.3 (0.0) 250.0 (0.0) 616.7 (2.9) 92.5 (3.0) 282.5 (29.6) Klingel [31] 2002 22 22 0.0% 61.4 (—) (—) 74.6 (0.0) HFD PS 1.3 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 228.8 (0.0) Mann [32] 2003 5 5 0.0% (—) (—) (—) HFD PS 1.6 (0.05) (—) (—) (—) 240.0 (0.0) Mandolfo [33] 2003 8 16 0.0% 61.4 (—) (—) 68 (8.6) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 550.0 (0.0) 30.0 (0.0) 240.0 (0.0) Pedrini [34] 2003 20 20 35.0% 63.0 (17.0) 116.4 (86.4) 60.3 (12.6) post-HDF/mixed-HDF PS 2.1 (0.0) 403.0 (55.9) 580.2 (36.6) 219.6 (36.6) 227.0 (17.7) Ward [35] 2003 12 24 41.6% 53.0 (13.0) 63.0 (18.3) (—) HFD PS 1.8 (0.0) 410.0 (1.9) 700.0 (0.0) (—) 228.0 (11.7) Bammens [36] 2004 14 70 28.6% 66.6 (3.1) 24.8 (10.0) 62.10 (1.94) HFD/pre-HDF/post-HDF PS 1.8 (0.0) 323.9 (116.3) 500.0 (0.0) 87.0 (0.0) 230.0 (0.0) Yamashita [37] 2004 5 5 80.0% (—) (—) (—) post-HF PS 1.8 (0.0) (—) (—) 84.2 (18.8) 120.0 (0.0) Emiliani [38] 2004 10 10 20.0% 66.0 (18.0) 80.0 (36.0) 66.2 (7.5) mid-HDF PAES 2.6 (0.0) 312.0 (18.0) 500.0 (0.0.0) 43.6 (7.2) 240.0 (10.0) Leypoldt [39] 2004 22 88 37.5% 61.0 (18.0) (—) 80.3 (19.4) HFD PS 1.77 (0.0) 338.0 (49.6) 540.0 (60.0) (—) 178.5 (19.0) Lucchi [40] 2004 10 20 40.0% 61.1 (8.9) 51.8 (35.9) (—) HFD/post-HDF PS 1.6 (0.0) 300.0 (0.0) 625.0 (0.0) 20.9 (0.0) 240.0 (0.0) Pisitkun [41] 2004 9 18 22% 48.0 (6.1) 51.4 (42.0) 55.2 (8.3) HFD/mid-HDF PS 2.7 (0.9) 475.0 (36.4) 800.0 (0.0) 59.9 (9.7) 240.0 (0.0) Tonelli [42] 2004 5 15 0.0% (—) (—) (—) HFD PS 1.8 (0.0) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Santoro [43] 2005 20 40 0.0% (—) (—) (—) HFD/mid-HDF PAES 1.90 (0.0) 363.0 (39.2) (—) 58.0 (11.9) (—) Brendolan [44] 2005 3 24 0.0% (—) (—) (—) HFD/post-HDF PS 2.2 (0.4) 333.3 (12.0) 500.0 (0.0) 16.6 (1.9) 226.1 (14.5) Padrini [45] 2005 11 22 36.4% 66.8 (11.9) 80.9 (66.9) 64.1 (9.2) post-HF/pre-HF PAES 2.1 (0.0) 327.8 (22.3) (—) 186 (40.1) 240 (6) Petras [46] 2005 6 36 0.0% 56.0 (16.0) 86.0 (50.0) (—) HFD/post-HDF/Pre-HF PAES 2.1 (0.0) 350.0 (0.0) 500.0 (0.0) 95.0 (0.0) 240.0 (0.0) Krieter [47] 2005a 5 5 60.0% 52.0 (22) (—) 68.5 (27.5) mid-HDF PAES 1.9 (0.0) 400.0 (0.0) 800.0 (0.0) 200.0 (0.0) 205.0 (15.0) Krieter [48] 2005b 10 40 30.0% 57.3 (13.7) 99.6 (92.4) 66.3 (10.4) mid-HDF/post-HDF PAES/PS 1.9 (0.1) 400.0 (0.0) 550.0 (0.0) 148.3 (2.9) 240 (23.4) Evenepoel [49] 2006 20 20 25.0% 68.8 (10.9) 19.3 (31.5) 59.9 (7.9) HFD PS 1.8 (0.0) 322.7 (21.6) 500.0 (0.0) (—) 230.0 (0.0) Mandolfo [50] 2006 12 18 66.7% 69.0 (9.0) 117.6 (69.6) 65.2 (8.1) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 700.0 (0.0) 50.0 (0.0) 240.0 (0.0) Nakashima [51] 2006 12 24 0.0% 49.1 (12.1) 127.2 (73.2) 66.8 (12.7) HFD PS 2.10 (0.0) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Panich [52] 2006 10 20 50.0% 58.2 (14.7) (—) 54.2 (3.8) HFD/post-HDF PS 1.8 (0.0) 425.3 (12.6) (—) 61.6 (0.6) 240.0 (0.0) Pedrini [53] 2006 12 72 25.0% 64.2 (6.6) 45.0 (38.0) 64.9 (11.2) mixed-HDF PS 2.1 (0.0) 422.0 (37.9) 609.0 (27.9) 178.0 (20.9) 218.0 (25.9) Potier [54] 2007 6 18 0.0% (—) (—) (—) post-HDF/pre-HDF/mid-HDF PAES 1.90 (0.0) 360.0 (0.0) 500.0 (0.0) 175.0 (0.0) (—) Feliciani [55] 2007 10 30 20.0% 64.7 (8.0) 54.7 (57.7) 73.25 (12.5) mixed-HDF/mid-HDF PAES/PS 1.85 (0.1) 385.5 (18.3) 609.0 (20.7) 167.5 (14.1) 231.5 (16.8) Krieter [56] 2007 8 32 62.5% 62.1 (13.8) 76.0 (55.3) 68.5 (7.1) HFD PS/PAES 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 236.0 (10.5) Santoro [57] 2007 8 16 50.0% 56.6 (23.6) (—) (—) mid-HDF PAES 1.9 (0.0) 306.5 (10.3) (—) 100.0 (0.0) 231.0 (10.3) Tiranathanagul [58] 2007 12 48 33.0% 54.2 (13.6) 42.0 (32.3) 62.85 (9.4) post-HDF/mid-HDF PS 2.7 (0.9) 416.7 (24.1) 800.0 (0.0) 113.0 (6.0) 240.0 (0.0) Abe [59] 2008 15 45 40% 65.5 (13.2) 72.9 (63.8) (—) HFD PMMA/CA/PS 1.5 (0.1) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Eloot [60] 2008 9 27 55.5% 71.0 (10.0) 19.0 (12.0) 79.0 (11.5) HFD PS 1.8 (0.0) 260.0 (0.0) 260.0 (0.0) (—) 360.0 (0.0) Mandolfo [61] 2008 8 16 37.5% 72.2 (4.8) 62.0 (24.0) 61.7 (11) HFD/mid-HDF PAES 1.9 (0.0) 251.5 (32.4) 700.0 (0.0) 56.0 (4.8) 240.0 (0.0) Spalding [62] 2008 12 12 50.0% 65.3 (12.9) (—) (—) HFD/post-HDF (—) (—) 358.4 (84.4) 800.0 (0.0) 37.5 (12.6) 197.4 (55.3) Krieter [63] 2008a 8 40 25.0% 64.0 (16.0) 70.0 (74.0) 74.2 (10.7) HFD/post-HDF PAES/PS 1.7 (0.1) 300.0 (0.0) 460.0 (0.0) 40.0 (0.0) 240.0 (0.0) Ouseph [64] 2008a 12 48 25.0% 57.0 (4.0) 52.0 (17.0) 81.3 (4.35) HFD PS/PAES 1.65 (0.1) 382.0 (4.8) 800.0 (0.0) (—) 219.0 (5.3) Ouseph [64] 2008b 12 60 41.6% 46.0 (3.0) 48.0 (8.0) 84.2 (6.75) HFD PS/PAES 1.90 (0.2) 404.0 (1.0) 800.0 (0.0) (—) 240.0 (0.0) Krieter [65] 2008b 8 48 37.5% 63.0 (14.0) 77.5 (38.9) 76.5 (11.15) HFD PAES/PS 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Joyeux [66] 2008 20 40 35.0% 58.0 (20.9) 46.0 (46.1) 71.5 (10.2) HFD/post-HDF PAES 2.10 (0.0) 310.5 (33.3) (—) 31.8 (5.2) 235.5 (14.6) Lee [67] 2009 8 16 50.0% 68.7 (19.1) 50.3 (53.1) (—) HFD PAES 1.1 (0.0) 325.0 (24.6) 500.0 (0.0) (—) 255.0 (14.8) Meert [68] 2009 14 42 50.0% 63.5 (17.0) 30.2 (36.0) (—) pre-HDF/pre-HF/post-HDF PAES 1.8 (0.2) 312.3 (15.6) 384.7 (5.0) 185.7 (20.7) 249.3 (13.0) Pedrini [69] 2009 15 90 20.0% 67.3 (8.7) 44.1 (20.8) 76.9 (13.8) mid-HDF PAES 2.1 (0.2) 378.5 (27.4) 599.5 (6.5) 167.5 (9.7) 223.0 (21.4) Susantitaphong [70] 2009 12 36 66.6% 59.5 (13.5) 81.6 (52.8) 57.5 (11.6) Pre-HDF/mid-HDF/post-HDF PAES 2.2 (0.1) 440.3 (19.9) 554.2 (10.4) 245.9 (2.1) 240.0 (0.0) Troidle [71] 2009 8 8 0.0% 45.0 (7.0) (—) (—) HFD PS 1.8 (0.0) 400.0 (0.0) 600.0 (0.0) (—) 480.0 (0.0) Wang [72] 2009 18 54 27.8% 46.9 (9.6) 52.5 (—) (—) HFD PS 1.5 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Bhimani [73] 2010 12 144 75.0% 60.0 (4.0) 81.0 (19.0) (—) HFD PS/PAES 1.7 (0.2) 400.0 (1.8) 550.0 (0.0) (—) (—) Gasco [74] 2010 16 263 50.0% 52.7 (—) (—) (—) post-HDF PS 1.8 (0.0) 323.0 (0.0) 800.0 (0.0) (—) 242.0 (0.0) Kohn [75] 2010 5 38 60.0% (—) 142.0 (60.9) 86.0 (29.0) HFD PAES (—) 425.0 (75.0) 200.0 (0.0) (—) 174.0 (15.0) Krieter [76] 2010 8 64 12.5% 63.0 (12.0) 81.8 (144.0) 70.8 (17.5) HFD/post-HDF PAES 1.9 (0.0) 378.0 (31.1) 500.0 (0.0) 47.0 (6.0) 229.0 (20.7) Park [77] 2010 52 52 52.0% 54.0 (12.4) 112.7 (188.6) (—) HFD PS 1.3 (0.0) 237.0 (23.0) 500.0 (0.0) (—) 240.0 (12.0) Basile [78] 2011 11 22 18.2% 54.1 (17.8) 78.0 (60.2) 69.1 (9.9) HFD PS 1.8 (0.0) 270.0 (0.0) 270.0 (0.0) (—) 469.1 (2.7) Ficheux [79] 2011 18 54 0.0% 79.7 (1.7) (—) 66.1 (2.3) HFD PS 2.2 (0.1) 318.0 (2.0) 500.0 (9.8) (—) 222.0 (2.9) Pedrini [80] 2011 15 60 33.3% 67.2 (8.3) (—) 73.1 (14.0) post-HDF/mid-HDF PS/PAES 2.3 (0.1) 374.0 (34.0) 580.0 (39.7) 147.5 (11.1) 224.0 (18.5) Panichi [81, 82] 2012 30 180 33.3% 55.9 (14.0) 58.0 (59.0) (—) post-HDF PAES 2.1 (0.0) 313.5 (32.7) 600.0 (0.0) 78.1 (0.0) 235.0 (13.8) Susantitaphong [83] 2012 12 48 66.6% 57.8 (14.8) 43.2 (42.0) 55.5 (11.1) mid-HDF/mixed-HDF PAES 2.2 (0.0) 425.0 (24.5) 600.0 (0.0) 200.0 (0.0) 240.0 (0.0) Tessitore [84] 2012 26 26 53.9% 63.0 (12.0) (—) (—) HFD PP 0.7 (0.0) 297.0 (32.0) 500.0 (0.0) (—) 230.0 (13.0) von Albertini [85] 2013 12 35 0.0% (—) (—) (—) HFD/post-HDF PAES/PS 1.8 (0.0) 417.1 (0.0) 667.4 (0.0) 30.7 (0.0) 206.3 (22.5) Heaf [86] 2013 12 96 30.0% 63.1 (11.7) 78.0 (52.8) 79.2 (17.8) HFD PAES 2.0 (0.0) 276.0 (38.7) 500.0 (0.0) (—) 240.0 (0.0) Melo [87] 2014 14 28 50.0% 48.9 (14.4) (—) 76.35 (19.63) HFD/post-HDF PS 2.0 (0.0) 375.0 (8.2) 760.0 (0.0) 40.0 (0.0) 115.7 (16.8) Pedrini [88] 2014 16 32 18.8% (—) (—) 77.60 (10.78) post-HDF PAES/PS 2.20 (0.10) 388.0 (25.9) 574.5 (39.0) 121.0 (11.9) 226.5 (13.7) Cornelis [89] 2014 13 52 23.1% 53.6 (20.4) 49.0 (29.0) (—) HFD/post-HDF PS 1.8 (0.0) 286.0 (4.8) 573.7 (13.1) 30.1 (1.4) 366.3 (3.9) Potier [90] 2016 6 24 66.7% 65.4 (25.5) 68.6 (43.7) 73.9 (2.1) HFD/post-HDF/ mixed-HDF/pre-HDF PS 2.3 (0.0) 339.4 (3.4) 600.0 (0.0) 122.1 (5.1) 240.0 (0.0) Gayrard [91] 2017 12 48 50.0% 73.0 (12.0) (—) 71.0 (1.9) HFD/post-HDF PS 1.8 (0.0) 366.3 (5.1) 602.3 (1.0) 51.8 (1.1) 233.6 (2.9) Kirsch [92] 2017 39 59 28.2% 60.5 (13.6) 63.1 (43.8) 80.2 (18.4) post-HDF/HFD PS 1.9 (0.1) 368.1 (12.8) (—) 27.5 (1.4) 252.4 (11.8) N, number; N meas, number of measurements; Vintage, time on chronic intermittent dialysis in months; PreWt, pre-dialysis weight in kilograms; MSA, membrane surface area (in square meters); QB, blood flow rate (mL/min); QD, dialysis fluid flow rate (mL/min); Duration, the dialysis session (in min). For each parameter the table summarizes the mean and the SD over all arms in each study or a (—) if the relevant parameter could not be extracted from the paper. Study quality Quality of the included studies varied widely based on each of the five components of the EPHPP (Table S1). For our meta-analysis, the global rating was characterized to be of moderate quality for most of the included studies [93], strong for 20 and weak for 16 studies. Both reviewers discussed the ratings and there was no discrepancy between them with respect to the components’ ratings and the final global scoring and rating. This high inter-rate agreement was in line with a previous evaluation of the EPHPP [94]. Main determinants of moderate quality were selection bias, study design and blinding procedures (methodologic heterogeneity), whereas data collection, study confounders and withdrawals/dropouts provided strong quality to the included studies. β2M dialyzer clearance in diffusive, high flux dialysis This meta-analysis also included 49 studies on HF dialysis, which evaluated 147 configurations of dialyzers and operational characteristics of treatment (e.g. blood or dialysate blood flows). Average (over the course of treatment) β2M clearance was 48.75 mL/min (95% CI 42.50–55.21) with substantial heterogeneity among studies [P (Q) ≤ 0.001] (Figure 2). Instantaneous β2M clearance was 52.09  mL/min (95% CI 41.39–62.78) with substantial heterogeneity among studies [P (Q) < 0.001] (Figure S1). There were no differences between instantaneous and average (over the course of the treatment) β2M clearances in univariate meta-regressions (difference of 1.88 mL/min, 95% CI −6.58 to 10.34, P =  0.66). Therefore, we combined instantaneous and average β2M clearances together for meta-regression analyses. First, we explored the sources of heterogeneity through ‘univariate’ meta-regressions examining only one study characteristic. Kuf (and Kuf scaled to MSA), clearance calculation formula, MSA, indexing clearance to the plasma (rather than blood) volume compartment, blood pump flow rate and dialysis membrane material were statistically significant predictors of variation in β2M clearance by diffusive, HF dialysis in these analyses (Table S2). Interestingly, there was no evidence of a secular trend of improving dialytic clearance over the last 17 years. Subsequently, we carried out ‘multivariable’ meta-regression to simultaneously adjust for multiple study characteristics. In these analyses shown in Table 2, we forced the type of measurement (instantaneous versus average) and the secular trend into the models. We found a significantly higher β2M clearance for PAES dialyzers (higher by 12.25 mL/min, 95% CI 5.472–19.028, P < 0.0001) relative to PS dialyzers. A significantly higher β2M clearance was found for higher blood flow rates in HF dialysis, i.e. an increase of 0.091 mL/min per 1 mL/min blood flow rate, 95% CI 0.024–0.159, P = 0.007). Adjusted dialysate side clearances were significantly lower than blood clearances (by 22.279 mL/min, 95% CI 9.8–34.757, P < 0.001). Other significant predictors were Kuf of the dialyzer (scaled to the MSA), while the MSA was of borderline significance (P = 0.057). In these multivariable analyses, there was no evidence for improving dialyzer performance over calendar time (P = 0.854). Similarly, there was no statistically significant difference in sensitivity analysis that compared the HEMO measurements against all the other measurements, or when we ran the multivariate regression, excluding the HEMO study (data not shown). Table 2 Metaregression of β2M clearance for high flux dialysis Variable Effect size (mL/min) CI P (Wald) Blood pump flow (per mL/min) 0.091 (0.024 to 0.159) 0.007 Kuf (scaled to MSA) 0.803 (0.373 to 1.232) <0.001 MSA (per m2) 10.923 (−0.327 to 22.173) 0.057 Dialysis membrane (relative to PS) PAES 12.25 (5.472 to 19.028) <0.001 CA 5.025 (−7.01 to 17.061) 0.413 PAN 3.571 (−10.378 to 17.519) 0.616 PMMA 9.15 (−2.501 to 20.8) 0.124 Compartment  Blood (versus plasma) 8.876 (−3.999 to 21.75) 0.177 Clearance side  Dialysate (versus blood) −22.279 (−34.757 to −9.8) <0.001 Type of measurement  Instantaneous (versus average) 6.589 (−3.422 to 16.6) 0.197 Secular trenda 0.178 (−1.716 to 2.072) 0.854 Variable Effect size (mL/min) CI P (Wald) Blood pump flow (per mL/min) 0.091 (0.024 to 0.159) 0.007 Kuf (scaled to MSA) 0.803 (0.373 to 1.232) <0.001 MSA (per m2) 10.923 (−0.327 to 22.173) 0.057 Dialysis membrane (relative to PS) PAES 12.25 (5.472 to 19.028) <0.001 CA 5.025 (−7.01 to 17.061) 0.413 PAN 3.571 (−10.378 to 17.519) 0.616 PMMA 9.15 (−2.501 to 20.8) 0.124 Compartment  Blood (versus plasma) 8.876 (−3.999 to 21.75) 0.177 Clearance side  Dialysate (versus blood) −22.279 (−34.757 to −9.8) <0.001 Type of measurement  Instantaneous (versus average) 6.589 (−3.422 to 16.6) 0.197 Secular trenda 0.178 (−1.716 to 2.072) 0.854 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). Inclusion of these variables decreased the apparent degree of heterogeneity by more than half (Q statistic of unadjusted model 4694.5603 versus 1947.1515 for the fully adjusted model), but significant heterogeneity did remain (P-value of QE statistic <0.001). Results based on 123 distinct configurations of dialyzer and dialysis procedure operational parameters. See text for other abbreviations. Table 2 Metaregression of β2M clearance for high flux dialysis Variable Effect size (mL/min) CI P (Wald) Blood pump flow (per mL/min) 0.091 (0.024 to 0.159) 0.007 Kuf (scaled to MSA) 0.803 (0.373 to 1.232) <0.001 MSA (per m2) 10.923 (−0.327 to 22.173) 0.057 Dialysis membrane (relative to PS) PAES 12.25 (5.472 to 19.028) <0.001 CA 5.025 (−7.01 to 17.061) 0.413 PAN 3.571 (−10.378 to 17.519) 0.616 PMMA 9.15 (−2.501 to 20.8) 0.124 Compartment  Blood (versus plasma) 8.876 (−3.999 to 21.75) 0.177 Clearance side  Dialysate (versus blood) −22.279 (−34.757 to −9.8) <0.001 Type of measurement  Instantaneous (versus average) 6.589 (−3.422 to 16.6) 0.197 Secular trenda 0.178 (−1.716 to 2.072) 0.854 Variable Effect size (mL/min) CI P (Wald) Blood pump flow (per mL/min) 0.091 (0.024 to 0.159) 0.007 Kuf (scaled to MSA) 0.803 (0.373 to 1.232) <0.001 MSA (per m2) 10.923 (−0.327 to 22.173) 0.057 Dialysis membrane (relative to PS) PAES 12.25 (5.472 to 19.028) <0.001 CA 5.025 (−7.01 to 17.061) 0.413 PAN 3.571 (−10.378 to 17.519) 0.616 PMMA 9.15 (−2.501 to 20.8) 0.124 Compartment  Blood (versus plasma) 8.876 (−3.999 to 21.75) 0.177 Clearance side  Dialysate (versus blood) −22.279 (−34.757 to −9.8) <0.001 Type of measurement  Instantaneous (versus average) 6.589 (−3.422 to 16.6) 0.197 Secular trenda 0.178 (−1.716 to 2.072) 0.854 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). Inclusion of these variables decreased the apparent degree of heterogeneity by more than half (Q statistic of unadjusted model 4694.5603 versus 1947.1515 for the fully adjusted model), but significant heterogeneity did remain (P-value of QE statistic <0.001). Results based on 123 distinct configurations of dialyzer and dialysis procedure operational parameters. See text for other abbreviations. FIGURE 2 View largeDownload slide Forest plot of average (over the course of the treatment) β2M dialyzer clearance in HFD. Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). FIGURE 2 View largeDownload slide Forest plot of average (over the course of the treatment) β2M dialyzer clearance in HFD. Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). β2M clearance in convective dialysis therapies This meta-analysis included 63 papers on HDF and 5 hemofiltration studies that examined 132 unique configurations of dialyzers, infusion volumes and patient cohorts. Average β2M clearance (over the course of treatment) was 8706 mL/min (95% CI 75.08–99.03) with substantial heterogeneity among studies [P (Q) ≤ 0.001] (Figure 3). Instantaneous β2M clearance was 125.26 mL/min (95% CI 103.92–146.59) with substantial heterogeneity among studies [P (Q) ≤ 0.001] (Figure S2). Kuf, blood pump flow rate, blood (versus plasma) compartment clearance, the side of the clearance (blood versus dialysate) were significant predictors in ‘univariate’ meta-regressions (Table S3). MSA, membrane material and substitution fluid infusion rates were not significant predictors in these univariate analyses. In ‘multivariable’ meta-regression analyses (Table 3) we found a significantly higher β2M clearance from the body when this calculation was indexed to whole blood versus plasma, while dialysate side body clearance was substantially lower than plasma by −41.523 mL/min (95% CI −54.525 to −28.52, P < 0.0001). Higher blood flow (0.188 mL/min per 1 mL/min blood flow, 95% CI 0.046–0.330, P = 0.01), membrane material (PS higher than PAES or PMMA) and certain forms of modality (e.g. pre-dilution HDF versus pre-dilution hemofiltration) but not substitution fluid infusion rates were significantly associated with higher β2M clearances. ANOVA tests suggested that both membrane material (P = 0.0033) and any convective modality (P = 0.0013) were significant predictors of dialytic body clearance of β2M. Table 3 Metaregression of β2M clearance for convective therapies (HF/HDF) Variable Effect size (per mL/min) CI P (Wald) Kuf (scaled to MSA) 1.691 (0.609 to 2.773) 0.002 MSA (per m2) −1.336 (−19.017 to 16.346) 0.882 Compartment  Blood (versus plasma) 49.868 (34.794 to 64.942) <0.001 Clearance side  Dialysate (versus blood) −41.523 (−54.525 to −28.52) <0.001 Blood pump flow (per mL/min) 0.188 (0.046 to 0.33) 0.01 Dialysis membrane  PAES −23.524 (−40.635 to −6.412) 0.007  PMMA −22.421 (−41.627 to −3.215) 0.022 Type of measurement  Instantaneous (versus average) 4.719 (−7.401 to 16.84) 0.445 Substitution fluid rate (per mL/min) 0.046 (−0.045 to 0.137) 0.321 Modality (relative to pre-hemofiltration)  post-hemofiltration 42.719 (−1.957 to 87.395) 0.061  post-HDF −7.764 (−27.834 to 12.306) 0.448  mid-HDF 5.614 (−16.493 to 27.721) 0.619  mixed-HDF −12.972 (−36.947 to 11.002) 0.289  pre-HDF −25.464 (−45.137 to −5.792) 0.011 Secular trenda −0.925 (−3.31 to 1.46) 0.447 Variable Effect size (per mL/min) CI P (Wald) Kuf (scaled to MSA) 1.691 (0.609 to 2.773) 0.002 MSA (per m2) −1.336 (−19.017 to 16.346) 0.882 Compartment  Blood (versus plasma) 49.868 (34.794 to 64.942) <0.001 Clearance side  Dialysate (versus blood) −41.523 (−54.525 to −28.52) <0.001 Blood pump flow (per mL/min) 0.188 (0.046 to 0.33) 0.01 Dialysis membrane  PAES −23.524 (−40.635 to −6.412) 0.007  PMMA −22.421 (−41.627 to −3.215) 0.022 Type of measurement  Instantaneous (versus average) 4.719 (−7.401 to 16.84) 0.445 Substitution fluid rate (per mL/min) 0.046 (−0.045 to 0.137) 0.321 Modality (relative to pre-hemofiltration)  post-hemofiltration 42.719 (−1.957 to 87.395) 0.061  post-HDF −7.764 (−27.834 to 12.306) 0.448  mid-HDF 5.614 (−16.493 to 27.721) 0.619  mixed-HDF −12.972 (−36.947 to 11.002) 0.289  pre-HDF −25.464 (−45.137 to −5.792) 0.011 Secular trenda −0.925 (−3.31 to 1.46) 0.447 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). Inclusion of these variables decreased the apparent degree of heterogeneity by >70% (Q statistic of unadjusted model 2361.8089 versus 675.8222 for the fully adjusted model), but significant heterogeneity did remain (P-value of QE statistic <0.001). Results are based on 127 distinct configurations of dialyzer and dialysis procedure operational parameters. See text for the abbreviations. Table 3 Metaregression of β2M clearance for convective therapies (HF/HDF) Variable Effect size (per mL/min) CI P (Wald) Kuf (scaled to MSA) 1.691 (0.609 to 2.773) 0.002 MSA (per m2) −1.336 (−19.017 to 16.346) 0.882 Compartment  Blood (versus plasma) 49.868 (34.794 to 64.942) <0.001 Clearance side  Dialysate (versus blood) −41.523 (−54.525 to −28.52) <0.001 Blood pump flow (per mL/min) 0.188 (0.046 to 0.33) 0.01 Dialysis membrane  PAES −23.524 (−40.635 to −6.412) 0.007  PMMA −22.421 (−41.627 to −3.215) 0.022 Type of measurement  Instantaneous (versus average) 4.719 (−7.401 to 16.84) 0.445 Substitution fluid rate (per mL/min) 0.046 (−0.045 to 0.137) 0.321 Modality (relative to pre-hemofiltration)  post-hemofiltration 42.719 (−1.957 to 87.395) 0.061  post-HDF −7.764 (−27.834 to 12.306) 0.448  mid-HDF 5.614 (−16.493 to 27.721) 0.619  mixed-HDF −12.972 (−36.947 to 11.002) 0.289  pre-HDF −25.464 (−45.137 to −5.792) 0.011 Secular trenda −0.925 (−3.31 to 1.46) 0.447 Variable Effect size (per mL/min) CI P (Wald) Kuf (scaled to MSA) 1.691 (0.609 to 2.773) 0.002 MSA (per m2) −1.336 (−19.017 to 16.346) 0.882 Compartment  Blood (versus plasma) 49.868 (34.794 to 64.942) <0.001 Clearance side  Dialysate (versus blood) −41.523 (−54.525 to −28.52) <0.001 Blood pump flow (per mL/min) 0.188 (0.046 to 0.33) 0.01 Dialysis membrane  PAES −23.524 (−40.635 to −6.412) 0.007  PMMA −22.421 (−41.627 to −3.215) 0.022 Type of measurement  Instantaneous (versus average) 4.719 (−7.401 to 16.84) 0.445 Substitution fluid rate (per mL/min) 0.046 (−0.045 to 0.137) 0.321 Modality (relative to pre-hemofiltration)  post-hemofiltration 42.719 (−1.957 to 87.395) 0.061  post-HDF −7.764 (−27.834 to 12.306) 0.448  mid-HDF 5.614 (−16.493 to 27.721) 0.619  mixed-HDF −12.972 (−36.947 to 11.002) 0.289  pre-HDF −25.464 (−45.137 to −5.792) 0.011 Secular trenda −0.925 (−3.31 to 1.46) 0.447 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). Inclusion of these variables decreased the apparent degree of heterogeneity by >70% (Q statistic of unadjusted model 2361.8089 versus 675.8222 for the fully adjusted model), but significant heterogeneity did remain (P-value of QE statistic <0.001). Results are based on 127 distinct configurations of dialyzer and dialysis procedure operational parameters. See text for the abbreviations. FIGURE 3 View largeDownload slide Forest plot of average (over the course of treatment) β2M dialyzer clearance in convective dialysis (HF/HDF). Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). FIGURE 3 View largeDownload slide Forest plot of average (over the course of treatment) β2M dialyzer clearance in convective dialysis (HF/HDF). Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). In our dataset, there were 73 distinct configurations in post-HDF, which allowed us to better clarify the effects of different parameters upon dialytic clearance. Significant predictors of dialytic clearance in post-HDF were the substitution fluid infusion rate: increase by 0.297 mL/min for each mL/min increase in infusion rate (95% CI 0.200–0.394, P < 0.001) and Kuf: increase 1.346 mL/min for each mL/min/mmHg/m2 (95% CI 0.271–2.420, P = 0.014), while dialysis with a PAES dialyzer was associated with reduced clearance by −18.480 mL/min (95% CI −34.86 to −2.101, P = 0.027). Dialysis with a membrane with a higher surface area was associated with a numerically higher β2M clearance of 37.040 mL/min/m2 (95% CI −1.487 to 75.566); this association was of borderline statistical significance (P = 0.06). Interestingly, higher blood pump flow rates were not associated with enhanced dialytic clearance in post-HDF (0.042 mL/min for each mL/min increase in blood pump flow rates, 95% CI −0.045 to 0.128, P = 0.345), while other factors (side of clearance, blood versus plasma compartment calculations, instantaneous versus average clearance and secular trends) were numerically like the patterns noted in Table 3 (data not shown). β2M reduction ratios are higher in convective versus diffusive dialysis therapies For this meta-analysis, we identified a total of 140 configurations (with covariate information) that reported reduction ratios of β2M in either HFD (n = 81) and convective dialysis therapies (n = 59) for multivariable adjustments. In univariate analysis, convective dialysis (taken as a group) afforded greater β2M reduction ratios by 14.300% (95% CI 10.756–17.845%, P < 0.0001) relative to HF dialysis (estimate of 59.169%, 95% CI 55.484–62.854%). In multivariable meta-regression analyses (Table 4), higher membrane Kuf was a significant predictor of higher β2M reduction ratio in both diffusive and convective dialysis. In HFD, β2M reduction ratios were significantly higher for PAES (8.367%, 95% CI 2.913–13.822%, P = 0.003) compared with PS dialyzers. There was a strong secular trend in the reduction ratio afforded by HF dialysis, i.e. an increase of 1.443% per year since 2001 (95% CI 0.363–2.523, P =0.009). There were no differences by membrane material or type of modality in convective therapies, yet higher substitution flow rates were associated with higher β2M reduction ratios. Table 4 Metaregression of β2M reduction ratios for high flux dialysis and convective therapies HF dialysis Convective therapies Effect size (per mL/min) CI P (Wald) Effect size (per mL/min) CI P (Wald) Blood pump flow (per mL/min) −0.01 (−0.051 to 0.03) 0.619 −0.017 (−0.051 to 0.018) 0.344 Kuf (scaled to MSA) 0.388 (0.073 to 0.703) 0.016 0.326 (0.046 to 0.606) 0.023 MSA 7.44 (−1.987 to 16.867) 0.122 5.068 (−2.155 to 12.291) 0.169 Membrane material (ref: PS)  PAES 8.367 (2.913 to 13.822) 0.003 −0.836 (−4.792 to 3.121) 0.679  PMMA 12.403 (4.737 to 20.07) 0.002 −2.491 (−9.733 to 4.751) 0.5  CA 0.262 (−9.675 to 10.199) 0.959 — — —  PAN 3.525 (−6.677 to 13.727) 0.498 — — — Correction of post dialysis β2M value  Corrected for hemoconcentration −2.04 (−11.171 to 7.091) 0.661 −4.29 (−12.542 to 3.962) 0.308 Secular trend 1.443 (0.363 to 2.523) 0.009 0.438 (−0.198 to 1.075) 0.177 Substitution fluid rate (per mL/min) 0.077 (0.001 to 0.152) 0.047 Modality (relative to pre-hemofiltration)  post-hemofiltration — — — 15.931 (−6.334 to 38.195) 0.161  post-HDF — — — 19.583 (−0.727 to 39.893) 0.059  mid-HDF — — — 16.235 (−2.687 to 35.156) 0.093  mixed-HDF — — — 14.587 (−4.303 to 33.477) 0.13  pre-HDF — — — 6.891 (−9.21 to 22.992) 0.402 HF dialysis Convective therapies Effect size (per mL/min) CI P (Wald) Effect size (per mL/min) CI P (Wald) Blood pump flow (per mL/min) −0.01 (−0.051 to 0.03) 0.619 −0.017 (−0.051 to 0.018) 0.344 Kuf (scaled to MSA) 0.388 (0.073 to 0.703) 0.016 0.326 (0.046 to 0.606) 0.023 MSA 7.44 (−1.987 to 16.867) 0.122 5.068 (−2.155 to 12.291) 0.169 Membrane material (ref: PS)  PAES 8.367 (2.913 to 13.822) 0.003 −0.836 (−4.792 to 3.121) 0.679  PMMA 12.403 (4.737 to 20.07) 0.002 −2.491 (−9.733 to 4.751) 0.5  CA 0.262 (−9.675 to 10.199) 0.959 — — —  PAN 3.525 (−6.677 to 13.727) 0.498 — — — Correction of post dialysis β2M value  Corrected for hemoconcentration −2.04 (−11.171 to 7.091) 0.661 −4.29 (−12.542 to 3.962) 0.308 Secular trend 1.443 (0.363 to 2.523) 0.009 0.438 (−0.198 to 1.075) 0.177 Substitution fluid rate (per mL/min) 0.077 (0.001 to 0.152) 0.047 Modality (relative to pre-hemofiltration)  post-hemofiltration — — — 15.931 (−6.334 to 38.195) 0.161  post-HDF — — — 19.583 (−0.727 to 39.893) 0.059  mid-HDF — — — 16.235 (−2.687 to 35.156) 0.093  mixed-HDF — — — 14.587 (−4.303 to 33.477) 0.13  pre-HDF — — — 6.891 (−9.21 to 22.992) 0.402 Kuf, ultrafiltration coefficient of a dialyzer. Table 4 Metaregression of β2M reduction ratios for high flux dialysis and convective therapies HF dialysis Convective therapies Effect size (per mL/min) CI P (Wald) Effect size (per mL/min) CI P (Wald) Blood pump flow (per mL/min) −0.01 (−0.051 to 0.03) 0.619 −0.017 (−0.051 to 0.018) 0.344 Kuf (scaled to MSA) 0.388 (0.073 to 0.703) 0.016 0.326 (0.046 to 0.606) 0.023 MSA 7.44 (−1.987 to 16.867) 0.122 5.068 (−2.155 to 12.291) 0.169 Membrane material (ref: PS)  PAES 8.367 (2.913 to 13.822) 0.003 −0.836 (−4.792 to 3.121) 0.679  PMMA 12.403 (4.737 to 20.07) 0.002 −2.491 (−9.733 to 4.751) 0.5  CA 0.262 (−9.675 to 10.199) 0.959 — — —  PAN 3.525 (−6.677 to 13.727) 0.498 — — — Correction of post dialysis β2M value  Corrected for hemoconcentration −2.04 (−11.171 to 7.091) 0.661 −4.29 (−12.542 to 3.962) 0.308 Secular trend 1.443 (0.363 to 2.523) 0.009 0.438 (−0.198 to 1.075) 0.177 Substitution fluid rate (per mL/min) 0.077 (0.001 to 0.152) 0.047 Modality (relative to pre-hemofiltration)  post-hemofiltration — — — 15.931 (−6.334 to 38.195) 0.161  post-HDF — — — 19.583 (−0.727 to 39.893) 0.059  mid-HDF — — — 16.235 (−2.687 to 35.156) 0.093  mixed-HDF — — — 14.587 (−4.303 to 33.477) 0.13  pre-HDF — — — 6.891 (−9.21 to 22.992) 0.402 HF dialysis Convective therapies Effect size (per mL/min) CI P (Wald) Effect size (per mL/min) CI P (Wald) Blood pump flow (per mL/min) −0.01 (−0.051 to 0.03) 0.619 −0.017 (−0.051 to 0.018) 0.344 Kuf (scaled to MSA) 0.388 (0.073 to 0.703) 0.016 0.326 (0.046 to 0.606) 0.023 MSA 7.44 (−1.987 to 16.867) 0.122 5.068 (−2.155 to 12.291) 0.169 Membrane material (ref: PS)  PAES 8.367 (2.913 to 13.822) 0.003 −0.836 (−4.792 to 3.121) 0.679  PMMA 12.403 (4.737 to 20.07) 0.002 −2.491 (−9.733 to 4.751) 0.5  CA 0.262 (−9.675 to 10.199) 0.959 — — —  PAN 3.525 (−6.677 to 13.727) 0.498 — — — Correction of post dialysis β2M value  Corrected for hemoconcentration −2.04 (−11.171 to 7.091) 0.661 −4.29 (−12.542 to 3.962) 0.308 Secular trend 1.443 (0.363 to 2.523) 0.009 0.438 (−0.198 to 1.075) 0.177 Substitution fluid rate (per mL/min) 0.077 (0.001 to 0.152) 0.047 Modality (relative to pre-hemofiltration)  post-hemofiltration — — — 15.931 (−6.334 to 38.195) 0.161  post-HDF — — — 19.583 (−0.727 to 39.893) 0.059  mid-HDF — — — 16.235 (−2.687 to 35.156) 0.093  mixed-HDF — — — 14.587 (−4.303 to 33.477) 0.13  pre-HDF — — — 6.891 (−9.21 to 22.992) 0.402 Kuf, ultrafiltration coefficient of a dialyzer. β2M mass removal is higher in convective versus diffusive dialysis therapies For this meta-analysis, we identified 60 configurations reporting mass removal data (mg/session) of β2M. β2M mass removal (mg/session) was 151.66 mg/session (95% CI 126.98–176.34, P < 0.001) with substantial heterogeneity among studies [P (Q) < 0.001] (Figure 4). Kuf and type of modality were significant predictors of higher dialytic mass removal of β2M (data not shown) in univariate metaregression analyses. In multivariable meta-regressions (Table 5), Kuf and convective (relative to HF dialysis) were associated with higher removal of β2M into the dialysate (P < 0.001 in ANOVA). Removal of β2M was numerically higher with pure filtration therapies rather than HDF. However, when we restricted the analyses to convective techniques (n = 31), there was no statistically significant difference among the different techniques in terms of their ability to remove β2M from the body (P = 0.892). Furthermore, there was no evidence for heterogeneity in this analysis (residual heterogeneity, P = 0.08). More extensive analysis of the role of the substitution volume on β2M mass removal by post-HDF was limited by the small number of configurations (n = 12) that reported dialytic mass removal of β2M. Table 5 Metaregression of β2M removal for high flux dialysis and convective therapies Effect size (mg/session) CI P (Wald) Blood flow (per mL/min) −0.157 (−0.398 to 0.084) 0.202 Kuf scaled to MSA 2.229 (0.316 to 4.142) 0.022 MSA (per m2) −0.206 (−66.052 to 65.64) 0.995 Membrane material (relative to polysulfone)  PAES −1.874 (−34.069 to 30.321) 0.909  CA 22.983 (−61.608 to 107.573) 0.594 Modality (relative to HF dialysis)  mid-HDF 56.138 (−1.787 to 114.063) 0.057  mixed-HDF 97.522 (41.638 to 153.405) <0.001  post-HDF 54.714 (22.879 to 86.549) <0.001  post-hemofiltration 151.036 (−17.467 to 319.538) 0.079  pre-HDF 41.564 (1.7 to 81.427) 0.041  pre-hemofiltration 163.451 (−71.28 to 398.182) 0.172 Secular trenda 1.783 (−2.718 to 6.283) 0.438 Effect size (mg/session) CI P (Wald) Blood flow (per mL/min) −0.157 (−0.398 to 0.084) 0.202 Kuf scaled to MSA 2.229 (0.316 to 4.142) 0.022 MSA (per m2) −0.206 (−66.052 to 65.64) 0.995 Membrane material (relative to polysulfone)  PAES −1.874 (−34.069 to 30.321) 0.909  CA 22.983 (−61.608 to 107.573) 0.594 Modality (relative to HF dialysis)  mid-HDF 56.138 (−1.787 to 114.063) 0.057  mixed-HDF 97.522 (41.638 to 153.405) <0.001  post-HDF 54.714 (22.879 to 86.549) <0.001  post-hemofiltration 151.036 (−17.467 to 319.538) 0.079  pre-HDF 41.564 (1.7 to 81.427) 0.041  pre-hemofiltration 163.451 (−71.28 to 398.182) 0.172 Secular trenda 1.783 (−2.718 to 6.283) 0.438 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). See text for other abbreviations. Table 5 Metaregression of β2M removal for high flux dialysis and convective therapies Effect size (mg/session) CI P (Wald) Blood flow (per mL/min) −0.157 (−0.398 to 0.084) 0.202 Kuf scaled to MSA 2.229 (0.316 to 4.142) 0.022 MSA (per m2) −0.206 (−66.052 to 65.64) 0.995 Membrane material (relative to polysulfone)  PAES −1.874 (−34.069 to 30.321) 0.909  CA 22.983 (−61.608 to 107.573) 0.594 Modality (relative to HF dialysis)  mid-HDF 56.138 (−1.787 to 114.063) 0.057  mixed-HDF 97.522 (41.638 to 153.405) <0.001  post-HDF 54.714 (22.879 to 86.549) <0.001  post-hemofiltration 151.036 (−17.467 to 319.538) 0.079  pre-HDF 41.564 (1.7 to 81.427) 0.041  pre-hemofiltration 163.451 (−71.28 to 398.182) 0.172 Secular trenda 1.783 (−2.718 to 6.283) 0.438 Effect size (mg/session) CI P (Wald) Blood flow (per mL/min) −0.157 (−0.398 to 0.084) 0.202 Kuf scaled to MSA 2.229 (0.316 to 4.142) 0.022 MSA (per m2) −0.206 (−66.052 to 65.64) 0.995 Membrane material (relative to polysulfone)  PAES −1.874 (−34.069 to 30.321) 0.909  CA 22.983 (−61.608 to 107.573) 0.594 Modality (relative to HF dialysis)  mid-HDF 56.138 (−1.787 to 114.063) 0.057  mixed-HDF 97.522 (41.638 to 153.405) <0.001  post-HDF 54.714 (22.879 to 86.549) <0.001  post-hemofiltration 151.036 (−17.467 to 319.538) 0.079  pre-HDF 41.564 (1.7 to 81.427) 0.041  pre-hemofiltration 163.451 (−71.28 to 398.182) 0.172 Secular trenda 1.783 (−2.718 to 6.283) 0.438 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). See text for other abbreviations. FIGURE 4 View largeDownload slide Forest plot of β2M mass removal (mg/session) in diffusive HFD and convective (HDF/HF) dialysis. Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). FIGURE 4 View largeDownload slide Forest plot of β2M mass removal (mg/session) in diffusive HFD and convective (HDF/HF) dialysis. Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). DISCUSSION This meta-analysis, combining 69 studies and including 1879 patients with 6771 clearance measurements, shows that membrane composition, modality (convective versus diffusive), blood flow rates and substitution fluid infusion rates independent of the dialysis modality are significant determinants of HF dialyzer performance in removing β2M. Our analysis is timely, as it provides quantitative information to aid the interpretation of a number of meta-analyses and secondary analyses of HD [11, 12, 14]. The significance of this work lies in our analysis of nearly 8-fold higher number of studies than previous reports by the Cochrane Group [16, 17] and others [18, 19]. Furthermore, our access to the primary study records of the HEMO trial allowed us to assess dialyzer performance using patient-level information from non-reused membranes thus overcoming a major limitation of a previous report [18]. One of the main and novel results of this study was that membrane material proved to be an important determinant of β2M clearance. Higher β2M clearances were noted with dialyzers made from PAES in respect to PS when applied in HF dialysis, the opposite of when applied in HDF. This is probably related to the chemical composition of the membranes as well as the 3D structure of the membranes and the different pressure profiles in these two modalities. The influence of membrane material on β2M clearance of HF dialysis was first reported 30 years ago [95]. Of relevance to our report, this early investigation showed that some dialysis membranes, such as cellulose acetate dialyzers, appear to induce β2M production during dialysis, whereas others, such as PS, do not. In the same study volume-controlled dialysis with HF membranes (PS 0.65 m2 and PS 1.25 m2) lowered β2M; clearance values, however, were significantly higher when these dialyzers were used in a HDF procedure. In another study [96] among patients receiving conventional HD using CA membranes, β2M levels increased 25.4% after HD, whereas in patients receiving HF HD using PS membrane, β2M levels decreased significantly (43.0%) after HD. Our results are also in accordance to a prospective, randomized, crossover study showing that the clearance of β2M was higher with PAES than PS [97]. Interestingly enough, β2M clearance during HDF was related to membrane material but in the inverse direction than in HF dialysis. We hypothesize that this is due to differential adsorption of β2M on membranes under the different transmembrane pressure (TMP) profiles of dialysis and HDF. Application of the higher TMP during HDF may result in a disproportionate increase in β2M adsorption in PS relative to PAES, so that the difference in clearance between the two membranes seen in HF is nearly reversed. An alternative explanation invokes a more efficient convection in membranes without adsorption versus those with more adsorption e.g. as a result of membrane clogging. Regardless of the explanation, this observation should be corroborated in future prospective, head to head comparisons given the substantial heterogeneity of methodologies for the measurement of β2M clearance employed by the different studies. Notwithstanding the effects of membrane material on β2M reduction ratio, it should be noted that recovery of β2M into the dialysate, was not affected by membrane material. This is consistent with a landmark prospective RCT [97], showing that the higher β2M clearance of PAES did not translate into more efficient mass removal of β2M. In that study, it was postulated that the higher mass removal of β2M by PAES arises from transmembrane transport augmented by adsorption within the membrane matrix. Membrane adsorption was experimentally demonstrated >20 years ago [98, 99] and the propensity of different membranes to differentially adsorb low molecular weight proteins was recently characterized with proteomic techniques [100]. Our analysis recapitulates previous findings that despite the higher clearance, β2M removal in the dialysate is not higher with any of the currently available membranes. This suggests that adsorption to the membrane, rather than convective or diffusive elimination of β2M in the dialysate, underlines the differences between dialyzers of different membrane material. The a priori plausibility of differential adsorption of β2M in membranes according to the dialysis mode is high. There are reports using proteomic techniques that demonstrate differential absorption of β2M in PS versus triacetate membranes [93], PS versus PMMA membranes [101] or even the same PS when exposed to the different pressure profiles associated with HF versus low flux dialysis [102]. An interesting report also showed a change of contribution of the different forms of clearance when the same dialyzer used in post- versus pre-HDF mode (adsorption is lower in post) [45]. Hence, the available data do point to differential adsorption patterns by material, permeability and even mode of HDF. The only credible way to test our hypothesis that PAES and PS adsorb β2M differently under HF dialysis and HDF is by properly designed head to head comparisons using standardized collection methods, blood and dialysate clearances and possibly proteomic techniques. An interesting direction for future innovations in dialyzer development that builds on this hypothesis would explore the properties of different membranes to optimize clearance for convective versus diffusive forms of dialysis. There have been reports in the literature about dialyzers (some of them already in the market) that are specifically targeted for convective therapies [103, 104], while safety considerations about albumin loss suggest that not all HF dialyzers may be used in high-volume convective therapies [105]. Such considerations should be taken into account during the design of follow-up studies in convective therapies. Our results suggest that dialyzers introduced in the last 15 years do not have substantially larger β2M clearance than those used during the landmark HEMO study in the late 1990s and early 2000s when used for conventional (diffusive) dialysis. Nevertheless, large secular trends consistent with improving dialyzer performance were observed when reduction ratios, rather than measured clearance or mass removal, were analyzed. Collectively, our analysis suggests not only that the basic mechanisms of middle molecule elimination by HF dialyzers has remained unchanged over the years, but the quantitative aspects of middle molecule centric HF dialysis have largely remained unchanged since HEMO was published. We should point out that these assessments do not apply to the emerging class of middle cut-off dialyzers, which not only have substantially higher middle molecule clearance than high flux membranes, but may even narrow the gap between high flux dialysis and HDF [92]. Despite the apparent lack of improvements in dialyzer performance, higher clearance (by up to 44%) may be attained by using the same dialyzers in convective therapies (HF or HDF). This was also noted when alternative, simple measures of middle molecule elimination, i.e. the reduction ratios, were utilized to compare diffusive and convective forms of dialysis. There are two mechanisms by which higher (pump) blood flow rates may increase β2M clearance in convective therapies: directly by increasing the amount of β2M available for diffusive clearance and indirectly by allowing higher rates of substitution fluid to be used, boosting the convective clearance. The latter mechanism is underscored by our finding that higher fluid substitution rates were significantly associated with higher β2M clearances in post-HDF therapies. This finding is supported by early studies on online HDF [106, 107] comparing the reduction ratios and the clearances of β2M, BUN, creatinine and phosphorus between HD and online HDF with 40–120 mL/min substitution fluid. The maximum benefit was achieved in HDF 100 (i.e. with 24 L substitution volume per 4-h treatment) versus classical HD. Another study of 2293 incident patients treated by post-dilution online HDF determined the convection volume threshold and range associated with survival advantage [108]. The relative adjusted survival rate was found to increase at about 55 L/week of convection volume and to stay increased up to about 75 L/week. The same paper found a nearly linear decrease in pre-dialysis β2M concentration by 0.6 mg/L for every 10 L/week of additional convection volume as the latter increased from 40 to 75 L/week. However, this mode can only be achieved with a permanent effective blood flow rate of at least 300 mL/min, since less than a third of this value can be accepted as the flow rate of the substitution fluid to avoid too high a TMP causing damage to the membrane. In the modern era, technical developments such as the adoption of variable ultrafiltration rates adapted to the level of the TMP during the treatment can be applied to achieve such high convection rates [109]. In fact there was a direct linear relationship between blood pump and dialysate flow rates in all the studies we analyzed, so that higher blood flow rates were associated with higher substitution fluid flow rates. The net result is that patients whose access could support high blood pump flow rates were the ones who received higher substitution fluid rate (>100 mL/min) and experienced the largest dialytic β2M removal. This pattern may be clinically significant, since a recent meta-analysis [14] of the large online HDF trials [21–23] and post hoc analyses published by the investigator teams in the last 5 years suggest an overall and cardiovascular survival advantage for these high-fluid rates. Treatment center policies about blood flow, treatment time, filter size and even hemoglobin level can be used in conjunction with the aforementioned technical innovations to achieve high convection volumes despite non-modifiable factors such as dialysis access that limit the achievement of higher blood and substitution fluid flow rates [109]. A surprising finding of our analysis was the lack of a meaningful effect of higher dialysate flow rates on improving diffusive or convective middle molecule clearance. This observation, which seems to go against classical teachings, is however fully in line with recent experiments about contemporary dialyzers for both small [110–112] and middle molecule clearance [113]. Design innovations such as spacer yarns in the fiber bundle, fiber undulations and changes in fiber-packing density have reduced the dependence of clearance on dialysate flow rates because of improved flow distribution in the dialyzer. Theoretical analysis based on the Weryński [114] and Michaels [115] equations relating diffusive clearance, sieving coefficient, Membrane Transfer Area Coefficient, blood and dialysate flow rates suggests that for dialyzers used in modern HF dialysis (sieving coefficient S = 0.65) and HDF (S = 0.75), increasing the dialysate flow by 60% from 500 to 800 mL/min will have a very small effect (∼0.4 mL/min) on middle molecule body clearance. Some limitations of this meta-analysis need to be acknowledged. First, the studies included differed in study design, methodologically (methods used for the calculation of clearance, dialysis modalities) and operationally (different dialyzers, different blood and dialysate flow rates, etc.). In particular, different approaches to calculate clearance will systematically overestimate (e.g. whole blood versus plasma) or underestimate (e.g. dialysate versus plasma) the dialytic clearance. We attempted to account for these systematic differences in our analyses through statistical modeling. However, residual confounding cannot be excluded. Such confounding may particularly apply to the apparent lack of an improvement of convective dialyzer performance with time, during a period in which many manufacturers released dialyzers with higher sieving coefficients for β2M and thus greater capacity for convective clearance. These dialyzers may also be more likely to remove β2M through adsorption in the inner layers of the dialyzer, so that studies relying on dialysate side measurements may have missed this finding. It should be noted that despite the lack of a statistically significant effect, the magnitude of the temporal trend for all dialyzer performance measures considered, is in the direction of more efficient removal with time. As further studies become available, our finding may notwithstand the passage of time. Second, most of the included studies recruited chronic HD patients on a thrice-weekly 4-h treatment schedule. Third, the apparent lack of an effect of higher dialysate flows may not apply to short, frequent, slow flow dialysis for membranes that do not exhibit enhanced dialytic removal at higher flows in conventional thrice-weekly dialysis [116]. Fourth, the limited sample size, selection of sampling points in the source data and analytical methodology of mixed models may have limited our ability to detect a statistically significant difference between instantaneous and average dialyzer clearances. Finally, this work is limited to adult patients and cannot be generalized to the pediatric dialysis population. CONCLUSIONS Dialysis prescription parameters (e.g. blood and dialysate flow rates in HD and infusion volume in HDF), as well as membrane material (HD), are major determinants of β2M clearance from the body in renal replacement therapies. Future prospective studies should standardize methodology for these measurements and investigate a wide variety of dialysis configurations to directly account for variability within and between patients and dialysis units. Such experimental studies are better suited than our statistical analyses to highlight clinically important differences related to the differential effects of β2M body removal seen with membranes of different material to inform their use in clinical HD and HDF. SUPPLEMENTARY DATA Supplementary data are available at ndt online. AUTHORS’ CONTRIBUTIONS The study was conceived and data were analyzed by the corresponding author. Data were generated by M.-E.R., G.T., Y.-H.N., Z.X., A.A. and R.F. Significant intellectual content was contributed by T.D.N. and M.L.U. All authors contributed to the interpretation of the data, drafting and revision of the manuscript. All authors have approved the final version of the article that was uploaded to the journal website. ACKNOWLEDGEMENTS A preliminary version of this work was presented in abstract form (THPO958, FRPO894) at Kidney Week 2016, Annual Meeting of the American Society of Nephrology, Chicago on November 2016. FUNDING No entity, private, public, commercial or non-for-profit funded directly or indirectly to this work. CONFLICT OF INTEREST STATEMENT None declared. REFERENCES 1 Cheung AK , Greene T , Leypoldt JK et al. . Association between serum β2-microglobulin level and infectious mortality in hemodialysis patients . Clin J Am Soc Nephrol 2008 ; 3 : 69 – 77 Google Scholar CrossRef Search ADS PubMed 2 Okuno S , Ishimura E , Kohno K et al. . Serum beta2-microglobulin level is a significant predictor of mortality in maintenance haemodialysis patients . Nephrol Dial Transplant 2008 ; 24 : 571 – 577 Google Scholar CrossRef Search ADS PubMed 3 Dember LM , Jaber BL. Unresolved issues in dialysis: dialysis-related amyloidosis: late finding or hidden epidemic? Semin Dial 2006 ; 19 : 105 – 109 Google Scholar CrossRef Search ADS PubMed 4 Jadoul M , Garbar C , van Ypersele de Strihou C. Pathological aspects of beta(2)-microglobulin amyloidosis . Semin Dial 2001 ; 14 : 86 – 89 Google Scholar CrossRef Search ADS PubMed 5 Thomas G , Jaber BL. Innovation in the treatment of uremia: Proceedings from the Cleveland Clinic Workshop: Convective Therapies for Removal of Middle Molecular Weight Uremic Toxins in End-Stage Renal Disease: a review of the evidence . Semin Dial 2009 ; 22 : 610 – 614 Google Scholar CrossRef Search ADS PubMed 6 Jadoul M , Garbar C , Noël H et al. . Histological prevalence of β2-microglobulin amyloidosis in hemodialysis: a prospective post-mortem study . Kidney Int 1997 ; 51 : 1928 – 1932 Google Scholar CrossRef Search ADS PubMed 7 Jadoul M , Garbar C , Vanholder R et al. . Prevalence of histological beta2-microglobulin amyloidosis in CAPD patients compared with hemodialysis patients . Kidney Int 1998 ; 54 : 956 – 959 Google Scholar CrossRef Search ADS PubMed 8 Argyropoulos CP , Chen SS , Ng Y-H et al. . Rediscovering beta-2 microglobulin as a biomarker across the spectrum of kidney diseases . Front Med 2017 ; 4 : 1 – 25 Google Scholar CrossRef Search ADS 9 Gal R , Korzets A , Schwartz A et al. . Systemic distribution of beta 2-microglobulin-derived amyloidosis in patients who undergo long-term hemodialysis. Report of seven cases and review of the literature . Arch Pathol Lab Med 1994 ; 118 : 718 – 721 Google Scholar PubMed 10 Takayama F , Miyazaki S , Morita T et al. . Dialysis-related amyloidosis of the heart in long-term hemodialysis patients . Kidney Int 2001 ; 59 : S172 – S176 Google Scholar CrossRef Search ADS 11 Argyropoulos C , Roumelioti M-E , Sattar A et al. . Dialyzer reuse and outcomes of high flux dialysis . PLoS One 2015 ; 10 : e0129575 Google Scholar CrossRef Search ADS PubMed 12 Palmer SC , Rabindranath KS , Craig JC et al. . High-flux versus low-flux membranes for end-stage kidney disease . Cochrane Database Syst Rev 2012 ; http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD005016.pub2/abstract (10 November 2017, date last accessed) 13 Locatelli F , Martin-Malo A , Hannedouche T et al. . Effect of membrane permeability on survival of hemodialysis patients . J Am Soc Nephrol 2009 ; 20 : 645 – 654 Google Scholar CrossRef Search ADS PubMed 14 Peters SAE , Bots ML , Canaud B et al. . Haemodiafiltration and mortality in end-stage kidney disease patients: a pooled individual participant data analysis from four randomized controlled trials . Nephrol Dial Transplant 2016 ; 31 : 978 – 984 Google Scholar CrossRef Search ADS PubMed 15 Nubé MJ , Peters SAE , Blankestijn PJ et al. . Mortality reduction by post-dilution online-haemodiafiltration: a cause-specific analysis . Nephrol Dial Transplant 2017 ; 32 : 548 – 555 Google Scholar PubMed 16 Nistor I , Palmer SC , Craig JC et al. . Convective versus diffusive dialysis therapies for chronic kidney failure: an updated systematic review of randomized controlled trials . Am J Kidney Dis 2014 ; 63 : 954 – 967 Google Scholar CrossRef Search ADS PubMed 17 Nistor I , Palmer SC , Craig JC et al. . Haemodiafiltration, haemofiltration and haemodialysis for end-stage kidney disease . Cochrane Database Syst Rev 2015 ; doi: 10.1002/14651858.CD006258.pub2 18 Susantitaphong P , Siribamrungwong M , Jaber BL. Convective therapies versus low-flux hemodialysis for chronic kidney failure: A meta-analysis of randomized controlled trials . Nephrol Dial Transplant 2013 ; 28 : 2859 – 2874 Google Scholar CrossRef Search ADS PubMed 19 Wang AY , Ninomiya T , Al-Kahwa A et al. . Effect of hemodiafiltration or hemofiltration compared with hemodialysis on mortality and cardiovascular disease in chronic kidney failure: a systematic review and meta-analysis of randomized trials . Am J Kidney Dis 2014 ; 63 : 968 – 978 Google Scholar CrossRef Search ADS PubMed 20 Eknoyan G , Beck GJ , Cheung AK et al. . Effect of dialysis dose and membrane flux in maintenance hemodialysis . N Engl J Med 2002 ; 347 : 2010 – 2019 Google Scholar CrossRef Search ADS PubMed 21 Grooteman MPC , van den Dorpel MA , Bots ML et al. . Effect of online hemodiafiltration on all-cause mortality and cardiovascular outcomes . J Am Soc Nephrol 2012 ; 23 : 1087 – 1096 Google Scholar CrossRef Search ADS PubMed 22 Maduell F , Moreso F , Pons M et al. . High-efficiency postdilution online hemodiafiltration reduces all-cause mortality in hemodialysis patients . J Am Soc Nephrol 2013 ; 24 : 487 – 497 Google Scholar CrossRef Search ADS PubMed 23 Ok E , Asci G , Toz H et al. . Mortality and cardiovascular events in online haemodiafiltration (OL-HDF) compared with high-flux dialysis: results from the Turkish OL-HDF Study . Nephrol Dial Transplant 2013 ; 28 : 192 – 202 Google Scholar CrossRef Search ADS PubMed 24 Thomas BH , Ciliska D , Dobbins M et al. . A process for systematically reviewing the literature: providing the research evidence for public health nursing interventions . Worldviews Evid Based Nurs 2004 ; 1 : 176 – 184 Google Scholar CrossRef Search ADS PubMed 25 Viechtbauer W. Conducting meta-analyses in R with the metafor package . J Stat Softw 2010 ; 36 : 1 – 48 Google Scholar CrossRef Search ADS 26 Leto E , Bilal F , Osmic I. Efficiency of high-flow dialyzers in removal of beta-2-microglobulin . Medicinski Arh 2001 ; 55 : 225 – 226 27 Xu XQ , Gruner N , Al-Bashir A et al. . Determination of extra renal clearance and generation rate of beta2-microglobulin in hemodialysis patients using a kinetic model . ASAIO J 2001 ; 47 : 623 – 627 Google Scholar CrossRef Search ADS PubMed 28 Yamada T , Akiba T , Sasaki S. Analysis of beta(2)-microglobulin kinetics in hemodialysis by a modified variable-volume one-compartment model . Blood Purif 2001 ; 19 : 388 – 394 Google Scholar CrossRef Search ADS PubMed 29 Stiller S , Xu XQ , Gruner N et al. . Validation of a two-pool model for the kinetics of beta2-microglobulin . Int J Artif Organs 2002 ; 25 : 411 – 420 Google Scholar CrossRef Search ADS PubMed 30 Ding F , Ahrenholz P , Winkler RE et al. . Online hemodiafiltration versus acetate-free biofiltration: A prospective crossover study . Artif Organs 2002 ; 26 : 169 – 180 Google Scholar CrossRef Search ADS PubMed 31 Klingel R , Ahrenholz P , Schwarting A et al. . Enhanced functional performance characteristics of a new polysulfone membrane for high-flux hemodialysis . Blood Purif 2002 ; 20 : 325 – 333 Google Scholar CrossRef Search ADS PubMed 32 Mann H , Al-Bashir A , Melzer H et al. . Diacap® α-polysulfone HI PS: A new dialysis membrane with optimum β2-microglobulin elimination . Int J Artif Organs 2003 ; 26 : 461 – 466 Google Scholar CrossRef Search ADS PubMed 33 Mandolfo S , Malberti F , Imbasciati E et al. . Impact of blood and dialysate flow and surface on performance of new polysulfone hemodialysis dialyzers . Int J Artif Organs 2003 ; 26 : 113 – 120 Google Scholar CrossRef Search ADS PubMed 34 Pedrini LA , De Cristofaro V. On-line mixed hemodiafiltration with a feedback for ultrafiltration control: effect on middle-molecule removal . Kidney Int 2003 ; 64 : 1505 – 1513 Google Scholar CrossRef Search ADS PubMed 35 Ward RA , Ouseph R. Impact of bleach cleaning on the performance of dialyzers with polysulfone membranes processed for reuse using peracetic Acid . Artif Organs 2003 ; 27 : 1029 – 1034 Google Scholar CrossRef Search ADS PubMed 36 Bammens B , Evenepoel P , Verbeke K et al. . Removal of the protein-bound solute p-cresol by convective transport: A randomized crossover study . Am J Kidney Dis 2004 ; 44 : 278 – 285 Google Scholar CrossRef Search ADS PubMed 37 Yamashita AC , Kawanishi H. Kinetics and dose of daily hemofiltration . Blood Purif 2004 ; 22 (Suppl 2) : 14 – 19 Google Scholar CrossRef Search ADS PubMed 38 Emiliani G , Briganti M , Montanari A et al. . [On-line haemodiafiltration with sorbent-regenerated ultrafiltrate as replacement fluid: Beta2-microglobulin removal versus filtration fraction] . G Ital Nefrol 2004 ; 21 (Suppl 30) : S80 – S84 Google Scholar PubMed 39 Leypoldt JK , Cheung AK , Deeter RB et al. . Kinetics of urea and beta-microglobulin during and after short hemodialysis treatments . Kidney Int 2004 ; 66 : 1669 – 1676 Google Scholar CrossRef Search ADS PubMed 40 Lucchi L , Fiore GB , Guadagni G et al. . Clinical evaluation of internal hemodiafiltration (iHDF): A diffusive-convective technique performed with internal filtration enhanced high-flux dialyzers . Int J Artif Organs 2004 ; 27 : 414 – 419 Google Scholar CrossRef Search ADS PubMed 41 Pisitkun T , Eiam-Ong S , Tiranathanagul K et al. . Convective-controlled double high flux hemodiafiltration: A novel blood purification modality . Int J Artif Organs 2004 ; 27 : 195 – 204 Google Scholar CrossRef Search ADS PubMed 42 Tonelli M , Dymond C , Gourishankar S et al. . Extended reuse of polysulfone hemodialysis membranes using citric acid and heat . ASAIO J 2004 ; 50 : 98 – 101 Google Scholar CrossRef Search ADS PubMed 43 Santoro A , Conz PA , De Cristofaro V et al. . Mid-dilution: The perfect balance between convection and diffusion. Contrib Nephrol 2005 ; 149 : 107 – 114 44 Brendolan A , Nalesso F , Fortunato A et al. . Dialytic performance evaluation of RexeedTM: A new polysulfone-based dialyzer with improved flow distributions . Int J Artif Organs 2005 ; 28 : 966 – 975 Google Scholar CrossRef Search ADS PubMed 45 Padrini R , Canova C , Conz P et al. . Convective and adsorptive removal of beta2-microglobulin during predilutional and postdilutional hemofiltration . Kidney Int 2005 ; 68 : 2331 – 2337 Google Scholar CrossRef Search ADS PubMed 46 Petras D , Fortunato A , Soffiati G et al. . Sequential convective therapies (SCT): a prospective study on feasibility, safety, adequacy and tolerance of on-line hemofiltration and hemodiafiltration in sequence . Int J Artif Organs 2005 ; 28 : 482 – 488 Google Scholar CrossRef Search ADS PubMed 47 Krieter DH , Collins G , Summerton J et al. . Mid-dilution on-line haemodiafiltration in a standard dialyser configuration . Nephrol Dial Transplant 2005 ; 20 : 155 – 160 Google Scholar CrossRef Search ADS PubMed 48 Krieter DH , Falkenhain S , Chalabi L et al. . Clinical cross-over comparison of mid-dilution hemodiafiltration using a novel dialyzer concept and post-dilution hemodiafiltration . Kidney Int 2005 ; 67 : 349 – 356 Google Scholar CrossRef Search ADS PubMed 49 Evenepoel P , Bammens B , Verbeke K et al. . Superior dialytic clearance of β2-microglobulin and p-cresol by high-flux hemodialysis as compared to peritoneal dialysis . Kidney Int 2006 ; 70 : 794 – 799 Google Scholar CrossRef Search ADS PubMed 50 Mandolfo S , Borlandelli S , Imbasciati E. Leptin and β2-microglobulin kinetics with three different dialysis modalities . Int J Artif Organs 2006 ; 29 : 949 – 955 Google Scholar CrossRef Search ADS PubMed 51 Nakashima A , Ogata S , Doi S et al. . Performance of polysulfone membrane dialyzers and dialysate flow pattern . Clin Exp Nephrol 2006 ; 10 : 210 – 215 Google Scholar CrossRef Search ADS PubMed 52 Panich A , Tiranathanagul K , Praditpornsilpa K et al. . The effectiveness of on-line hemodiafiltration on beta-2 microglobulin clearance in end stage renal disease . J Med Assoc Thai 2006 ; 89 (Suppl 2) : S1 – S8 Google Scholar PubMed 53 Pedrini LA , Cozzi G , Faranna P et al. . Transmembrane pressure modulation in high-volume mixed hemodiafiltration to optimize efficiency and minimize protein loss . Kidney Int 2006 ; 69 : 573 – 579 Google Scholar CrossRef Search ADS PubMed 54 Potier J. Mid-dilution: an innovative high-quality and safe haemodiafiltration approach. Contrib Nephrol 2007 ; 158 : 153 – 160 55 Feliciani A , Riva MA , Zerbi S et al. . New strategies in haemodiafiltration (HDF): prospective comparative analysis between on-line mixed HDF and mid-dilution HDF . Nephrol Dial Transplant 2007 ; 22 : 1672 – 1679 Google Scholar CrossRef Search ADS PubMed 56 Krieter DH , Morgenroth A , Barasinski AA et al. . Effects of a polyelectrolyte additive on the selective dialysis membrane permeability for low-molecular-weight proteins . Nephrol Dial Transplant 2007 ; 22 : 491 – 499 Google Scholar CrossRef Search ADS PubMed 57 Santoro A , Ferramosca E , Mancini E et al. . Reverse mid-dilution: New way to remove small and middle molecules as well as phosphate with high intrafilter convective clearance . Nephrol Dial Transplant 2007 ; 22 : 2000 – 2005 Google Scholar CrossRef Search ADS PubMed 58 Tiranathanagul K , Yossundharakul C , Techawathanawanna N et al. . Comparison of middle-molecule clearance between convective control double high-flux hemodiafiltration and on-line hemodiafiltration . Int J Artif Organs 2007 ; 30 : 1090 – 1097 Google Scholar CrossRef Search ADS PubMed 59 Abe M , Okada K , Maruyama T et al. . Clinical evaluation of plasma insulin and C-peptide levels with 3 different high-flux dialyzers in diabetic patients on hemodialysis . Int J Artif Organs 2008 ; 31 : 898 – 904 Google Scholar CrossRef Search ADS PubMed 60 Eloot S , Van Biesen W , Dhondt A et al. . Impact of hemodialysis duration on the removal of uremic retention solutes . Kidney Int 2008 ; 73 : 765 – 770 Google Scholar CrossRef Search ADS PubMed 61 Mandolfo S , Borlandelli S , Imbasciati E et al. . Pilot study to assess increased dialysis efficiency in patients with limited blood flow rates due to vascular access problems . Hemodial Int 2008 ; 12 : 55 – 61 Google Scholar CrossRef Search ADS PubMed 62 Spalding EM , Pandya P , Farrington K. Effect of high haematocrit on the efficiency of high-flux dialysis therapies . Nephron Clin Pract 2008 ; 110 : c86 – c92 Google Scholar CrossRef Search ADS PubMed 63 Krieter DH , Hunn E , Morgenroth A et al. . Matching efficacy of online hemodiafiltration in simple hemodialysis mode . Artif Organs 2008 ; 32 : 903 – 909 Google Scholar CrossRef Search ADS PubMed 64 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 : 1704 – 1712 Google Scholar CrossRef Search ADS PubMed 65 Krieter DH , Lemke H-D , Wanner C. A new synthetic dialyzer with advanced permselectivity for enhanced low-molecular weight protein removal . Artif Organs 2008 ; 32 : 547 – 554 Google Scholar CrossRef Search ADS PubMed 66 Joyeux V , Sijpkens Y , Haddj-Elmrabet A et al. . Optimized convective transport with automated pressure control in on-line postdilution hemodiafiltration . Int J Artif Organs 2008 ; 31 : 928 – 936 Google Scholar CrossRef Search ADS PubMed 67 Lee D , Haase M , Haase-Fielitz A et al. . A pilot, randomized, double-blind, cross-over study of high cut-off versus high-flux dialysis membranes . Blood Purif 2009 ; 28 : 365 – 372 Google Scholar CrossRef Search ADS PubMed 68 Meert N , Eloot S , Waterloos M-A et al. . Effective removal of protein-bound uraemic solutes by different convective strategies: a prospective trial . Nephrol Dial Transplant 2009 ; 24 : 562 – 570 Google Scholar CrossRef Search ADS PubMed 69 Pedrini LA , Feliciani A , Zerbi S et al. . Optimization of mid-dilution haemodiafiltration: technique and performance . Nephrol Dial Transplant 2009 ; 24 : 2816 – 2824 Google Scholar CrossRef Search ADS PubMed 70 Susantitaphong P , Tiranathanagul K , Hanvivatvong O et al. . A simple efficient technique of “mid-dilution” on-line hemodiafiltration . Blood Purif 2009 ; 28 : 93 – 101 Google Scholar CrossRef Search ADS PubMed 71 Troidle L , Finkelstein F , Hotchkiss M et al. . Enhanced solute removal with intermittent, in-center, 8-hour nocturnal hemodialysis . Hemodial Int 2009 ; 13 : 487 – 491 Google Scholar CrossRef Search ADS PubMed 72 Wang Y-M , Zhang W-M , Wang B-S et al. . Solute clearance characteristics of REXEEDTM series dialyzer during high-flux dialysis . J Shanghai Jiaotong Univ 2009 ; 29 : 858 – 861 73 Bhimani JP , Ouseph R , Ward RA. Effect of increasing dialysate flow rate on diffusive mass transfer of urea, phosphate and β2-microglobulin during clinical haemodialysis . Nephrol Dial Transplant 2010 ; 25 : 3990 – 3995 Google Scholar CrossRef Search ADS PubMed 74 Gascó J , Iñigo V , Mascarós V et al. . Kinetics of beta2-microglobulin and urea in high-efficiency hemodiafiltration . Blood Purif 2010 ; 30 : 224 – 225 75 Kohn OF , Coe FL , Ing TS. Solute kinetics with short-daily home hemodialysis using slow dialysate flow rate . Hemodial Int 2010 ; 14 : 39 – 46 Google Scholar CrossRef Search ADS PubMed 76 Krieter DH , Hackl A , Rodriguez A et al. . Protein-bound uraemic toxin removal in haemodialysis and post-dilution haemodiafiltration . Nephrol Dial Transplant 2010 ; 25 : 212 – 218 Google Scholar CrossRef Search ADS PubMed 77 Park J-S , Kim G-H , Kang CM et al. . Application of cystatin C reduction ratio to high-flux hemodialysis as an alternative indicator of the clearance of middle molecules . Korean J Intern Med 2010 ; 25 : 77 – 81 Google Scholar CrossRef Search ADS PubMed 78 Basile C , Libutti P , Di Turo AL et al. . Removal of uraemic retention solutes in standard bicarbonate haemodialysis and long-hour slow-flow bicarbonate haemodialysis . Nephrol Dial Transplant 2011 ; 26 : 1296 – 1303 Google Scholar CrossRef Search ADS PubMed 79 Ficheux A , Gayrard N , Szwarc I et al. . The use of SDSPAGE scanning of spent dialysate to assess uraemic toxin removal by dialysis . Nephrol Dial Transplant 2011 ; 26 : 2281 – 2289 Google Scholar CrossRef Search ADS PubMed 80 Pedrini LA , Gmerek A , Wagner J. Efficiency of post-dilution hemodiafiltration with a high-flux α-polysulfone dialyzer . Int J Artif Organs 2011 ; 34 : 397 – 404 Google Scholar CrossRef Search ADS PubMed 81 Panichi V , De Ferrari G , Saffiotti S et al. . Comparison of on-line HDF modes automated TMP control vs. Volume control on achieved convective volume and middle molecule clearance . Nephrol Dial Transplant 2012 ; 27 : ii205 – ii206 Google Scholar CrossRef Search ADS 82 Panichi V , De Ferrari G , Saffioti S et al. . Divert to ultra: differences in infused volumes and clearance in two on-line hemodiafiltration treatments . Int J Artif Organs 2012 ; 35 : 435 – 443 Google Scholar CrossRef Search ADS PubMed 83 Susantitaphong P , Tiranathanagul K , Katavetin P et al. . Efficacy comparison between simple mixed-dilution and simple mid-dilution on-line hemodiafiltration techniques: a crossover study . Artif Organs 2012 ; 36 : 1059 – 1065 Google Scholar CrossRef Search ADS PubMed 84 Tessitore N , Bedogna V , Girelli D et al. . Effect of a single hemodialysis session by HFR and low-flux bicarbonate dialysis on serum hepcidin-25 levels: a randomized cross-over study . Nephrol Dial Transplant 2012 ; 27 (Suppl 2): i207 85 von Albertini B , Mathieu C , Cherpillod A et al. . In-vivo 2 microglobulin clerance in high-flux HD & HDF . Nephrol Dial Transplant 2013 ; 28 (Supplement 1) : i422 86 Heaf JG , Axelsen M , Pedersen RS. Multipass haemodialysis: A novel dialysis modality . Nephrol Dial Transplant 2013 ; 28 : 1255 – 1264 Google Scholar CrossRef Search ADS PubMed 87 Melo NCV , Moyses RMA , Elias RM et al. . Reprocessing high-flux polysulfone dialyzers does not negatively impact solute removal in short-daily online hemodiafiltration . Hemodial Int 2014 ; 18 : 473 – 480 Google Scholar CrossRef Search ADS PubMed 88 Pedrini LA , Krisp C , Gmerek A et al. . Patterns of proteins removed with high-flux membranes on high-volume hemodiafiltration detected with a multidimensional LC-MS/MS strategy . Blood Purif 2014 ; 38 : 115 – 126 Google Scholar CrossRef Search ADS PubMed 89 Cornelis T , Van Der Sande FM , Eloot S et al. . Acute hemodynamic response and uremic toxin removal in conventional and extended hemodialysis and hemodiafiltration: A randomized crossover study . Am J Kidney Dis 2014 ; 64 : 247 – 256 Google Scholar CrossRef Search ADS PubMed 90 Potier J , Bowry S , Canaud B. Clinical performance assessment of CorDiax filters in hemodialysis and hemodiafiltration. Contrib Nephrol 2017 ; 189 : 237 – 245 91 Gayrard N , Ficheux A , Duranton F et al. . Consequences of increasing convection onto patient care and protein removal in hemodialysis . PloS One 2017 ; 12 : e0171179 Google Scholar CrossRef Search ADS PubMed 92 Kirsch AH , Lyko R , Nilsson L-G et al. . Performance of hemodialysis with novel medium cut-off dialyzers . Nephrol Dial Transplant 2017 ; 32 : 165 – 172 Google Scholar PubMed 93 Pieroni L , Mortera SL , Greco V et al. . Biocompatibility assessment of haemodialysis membrane materials by proteomic investigations . Mol BioSyst 2015 ; 11 : 1633 – 1643 Google Scholar CrossRef Search ADS PubMed 94 Armijo-Olivo S , Stiles CR , Hagen NA et al. . Assessment of study quality for systematic reviews: a comparison of the Cochrane Collaboration Risk of Bias Tool and the Effective Public Health Practice Project Quality Assessment Tool: methodological research . J Eval Clin Pract 2012 ; 18 : 12 – 18 Google Scholar CrossRef Search ADS PubMed 95 Mayer G , Thum J , Woloszczuk W et al. . Beta-2-microglobulin in hemodialysis patients. Effects of different dialyzers and different dialysis procedures . Am J Nephrol 1988 ; 8 : 280 – 284 Google Scholar CrossRef Search ADS PubMed 96 Acchiardo S , Kraus AP Jr , Jennings BR. Beta 2-microglobulin levels in patients with renal insufficiency . Am J Kidney Dis 1989 ; 13 : 70 – 74 Google Scholar CrossRef Search ADS PubMed 97 Hoenich NA , Stamp S. Clinical performance of a new high-flux synthetic membrane . Am J Kidney Dis 2000 ; 36 : 345 – 352 Google Scholar CrossRef Search ADS PubMed 98 Clark WR , Macias WL , Molitoris BA et al. . Membrane adsorption of β2-microglobulin: equilibrium and kinetic characterization . Kidney Int 1994 ; 46 : 1140 – 1146 Google Scholar CrossRef Search ADS PubMed 99 Clark WR , Macias WL , Molitoris BA et al. . Plasma protein adsorption to highly permeable hemodialysis membranes . Kidney Int 1995 ; 48 : 481 – 488 Google Scholar CrossRef Search ADS PubMed 100 Donadio C , Kanaki A , Sami N et al. . Proteomic evaluation of low- and high-flux hemodialysis . Nephrol Dial Transplant 2013 ; 28 (Suppl 1): i421 Google Scholar CrossRef Search ADS 101 Ishikawa I , Chikazawa Y , Sato K et al. . Proteomic analysis of serum, outflow dialysate and adsorbed protein onto dialysis membranes (polysulfone and PMMA) during hemodialysis treatment using SELDI-TOF-MS . Am J Nephrol 2006 ; 26 : 372 – 380 Google Scholar CrossRef Search ADS PubMed 102 Han S , Yang K , Sun J et al. . Proteomics investigations into serum proteins adsorbed by high-flux and low-flux dialysis membranes . Proteomics Clin Appl 2017 ; doi: 10.1002/prca.201700079 103 Cuoghi A , Caiazzo M , Monari E et al. . New horizon in dialysis depuration: characterization of a polysulfone membrane able to break the ‘albumin wall’ . J Biomater Appl 2015 ; 29 : 1363 – 1371 Google Scholar CrossRef Search ADS PubMed 104 Maduell F , Arias-Guillen M , Fontseré N et al. . Elimination of large uremic toxins by a dialyzer specifically designed for high-volume convective therapies . Blood Purif 2014 ; 37 : 125 – 130 Google Scholar CrossRef Search ADS PubMed 105 Potier J , Queffeulou G , Bouet J. Are all dialyzers compatible with the convective volumes suggested for postdilution online hemodiafiltration? Int J Artif Organs 2016 ; 39 : 460 – 470 Google Scholar CrossRef Search ADS PubMed 106 Lornoy W , Becaus I , Billiouw JM et al. . Remarkable removal of beta-2-microglobulin by on-line hemodiafiltration . Am J Nephrol 1998 ; 18 : 105 – 108 Google Scholar CrossRef Search ADS PubMed 107 Lornoy W , Becaus I , Billiouw JM et al. . On-line haemodiafiltration. Remarkable removal of beta2-microglobulin. Long-term clinical observations . Nephrol Dial Transplant 2000 ; 15 : 49 – 54 Google Scholar CrossRef Search ADS PubMed 108 Canaud B , Barbieri C , Marcelli D et al. . Optimal convection volume for improving patient outcomes in an international incident dialysis cohort treated with online hemodiafiltration . Kidney Int 2015 ; 88 : 1108 – 1116 Google Scholar CrossRef Search ADS PubMed 109 Marcelli D , Kopperschmidt P , Bayh I et al. . Modifiable factors associated with achievement of high-volume post-dilution hemodiafiltration: results from an international study . Int J Artif Organs 2015 ; 38 : 244 – 250 Google Scholar CrossRef Search ADS PubMed 110 Albalate M , Pérez-García R , De Sequera P et al. . Is it useful to increase dialysate flow rate to improve the delivered Kt? BMC Nephrol 2015 ; 16 : 1 – 6 Google Scholar CrossRef Search ADS 111 Albalate Ramón M , de Sequera Ortiz P , Pérez-García R et al. . What is the optimum dialysate flow in post-dilution online haemodiafiltration? Nefrologia 2015 ; 35 : 533 – 538 Google Scholar CrossRef Search ADS PubMed 112 Ward RA , Idoux JW , Hamdan H et al. . Dialysate flow rate and delivered kt/v urea for dialyzers with enhanced dialysate flow distribution . Clin J Am Soc Nephrol 2011 ; 6 : 2235 – 2239 Google Scholar CrossRef Search ADS PubMed 113 Maduell F , Ojeda R , Arias-Guillén M et al. . Optimisation of dialysate flow in on-line hemodiafiltration . Nefrologia 2015 ; 35 : 473 – 478 Google Scholar CrossRef Search ADS PubMed 114 Weryński A. Evaluation of the impact of ultrafiltration on dialyzer clearance . Artif Organs 1979 ; 3 : 140 – 142 Google Scholar CrossRef Search ADS PubMed 115 Michaels AS. Analysis of membrane-transport devices . ASAIO Trans 1966 ; 12 : 387 – 392 116 Leypoldt JK , Kamerath CD , Gilson JF et al. . Dialyzer clearances and mass transfer-area coefficients for small solutes at low dialysate flow rates . ASAIO J 2006 ; 52 : 404 – 409 Google Scholar CrossRef Search ADS PubMed © The Author 2017. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nephrology Dialysis Transplantation Oxford University Press

Beta-2 microglobulin clearance in high-flux dialysis and convective dialysis modalities: a meta-analysis of published studies

Loading next page...
 
/lp/ou_press/beta-2-microglobulin-clearance-in-high-flux-dialysis-and-convective-iUbil0adti
Publisher
Oxford University Press
Copyright
© The Author 2017. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved.
ISSN
0931-0509
eISSN
1460-2385
D.O.I.
10.1093/ndt/gfx311
Publisher site
See Article on Publisher Site

Abstract

ABSTRACT Background Recent meta-analyses suggest that higher removal of beta-2 microglobulin (β2M) with either high-flux (HFD) dialysis or hemodiafiltration (HDF) may be associated with decreased total and cardiovascular mortality in dialysis patients. However, there are limited data about the performance of high flux dialyzers and/or convective therapies in removing β2M. Methods This is a random effects meta-analysis and meta-regression of data extracted from randomized controlled trials and observational studies in hemodialysis, hemofiltration and HDF regarding the efficiency of high flux dialyzers to remove β2M. Studies were searched using ProQuest in SCOPUS, EMBASE and MEDLINE. Results We included 69 studies from 1 January 2001 to 12 June 2017 on 1879 patients with 6771 available measurements. Average β2M clearance was 48.75 mL/min [95% confidence interval (CI) 42.50–55.21] for conventional HF dialysis, and 87.06 mL/min (95% CI 75.08–99.03) for convective therapies (hemofiltration and HDF) with substantial heterogeneity among studies [P (Q) ≤ 0.001]. In multivariable meta-regression analyses, we found significantly higher β2M clearance for polyarylethersulfone dialyzers when used for HFD and polysulfone membranes in convective therapies. However, the mass of β2M removed into the dialysate did not depend on membrane material. Adjusted dialysate-side (−22.279, 95% CI  −9.8 to −34.757, P < 0.001) β2M clearances were significantly lower than whole blood clearances, suggesting that adsorption contributes substantially to β2M removal. Higher Kuf, blood flow and substitution fluid rates but not dialysate flow rates were associated with statistically significant and clinically meaningful elevation in β2M clearance from the body independent of the dialysis modality. Conclusions Membrane composition and characteristics, modality (convective versus diffusive), blood flow rates and substitution fluid rates in HDF play a significant role in the efficient removal of β2M from the body in both diffusive and convective dialysis. beta-2 microglobulin, clearance, hemodiafiltration, hemofiltration, high-flux hemodialysis INTRODUCTION The accumulation of middle molecular weight solutes, such as beta-2 microglobulin (β2M), is toxic to various body tissues and has been associated with adverse cardiovascular and infectious outcomes among patients with end-stage renal disease (ESRD) [1, 2]. β2M precipitates and forms fibrillary structures and amyloid deposits in bones, periarticular tissues [3], vessel walls and internal organs, especially the heart [4–7]. Dialysis-related amyloidosis and other disorders associated with abnormal β2M accumulation and function [8] are clinically silent, develop early in the development and progression of chronic kidney disease (CKD) and may even imply a potential causal link with the highly prevalent cardiovascular disease (CVD) in ESRD patients [9, 10]. Several meta-analyses of randomized controlled trials (RCTs) in conventional dialysis suggest that high-flux dialyzers, which more efficiently remove β2M than their low-flux (LF) counterparts, are associated with improved cardiovascular outcomes [11, 12]. Convective therapies, including hemodiafiltration (HDF) and hemofiltration, achieve even higher middle molecule clearances relative to HF dialysis. These therapies may improve the chronic retention of β2M over time noted with thrice-weekly HFD [5, 11, 13]. In these modalities, clearance is a function of the total volume of solution utilized (both dialysate flow rate and replacement solution). A recent individual patient-level meta-analysis of published RCTs suggests that the higher clearance from the body achieved by these therapies may result in clinically and statistically significant improvement in total and cardiovascular mortality relative to conventional HFD [14, 15]. Nevertheless, the quality of the evidence and the putative effects of convective dialysis have been called into question by large collaborative aggregate level meta-analyses by the Cochrane Group [16, 17] and others [18, 19]. The interpretation of these contradicting analyses of data outcomes is complicated by the limited evidence synthesis of the performance and the determinants of β2M clearance by high flux dialyzers when the latter are used in conventional or convective forms of renal replacement therapies. The aforementioned meta-analyses have reported only on a limited number of studies that examined dialyzer clearance or β2M mass removal, focusing instead on reduction ratios as the sole measure of dialyzer performance. None of the aforementioned studies has attempted to analyze the impact of different dialysis configurations (e.g. membrane material, surface area, substitution fluid rate) on multiple measures of β2M body removal. This literature gap limits our ability to better understand the performance of these therapies, and how best to modify treatment parameters to optimize clearance of middle molecules, thus moving beyond urea-centric approaches that have been widely used in modern dialysis. To do so, we conducted a meta-analysis of data about the performance of HFD and/or convective dialysis therapies to remove β2M. We included studies published between 2001 and 2017, covering the period in which the landmark RCTs in HFD [13, 20] and HDF [21–23] were published. MATERIALS AND METHODS This is a meta-analysis of data collected in RCTs and observational studies in hemodialysis (HD) about the performance (ability) of HFD and convective therapies (HDF or hemofiltration, HF) to remove β2M from the body. The focus of this meta-analysis was on studies that could provide determinations of β2M ‘clearance from the body’ as the primary outcome measure of dialysis procedure performance. Search strategy The overarching search strategy for this meta-analysis was to include studies that had employed formal methods to characterize dialytic performance. Our initial focus was on studies published from 1 January 2001 to 31 December 2013. The date range was determined to capture performance of dialyzers that were likely used in the main outcomes trials in HFD and HDF. Subsequently, we extended the search for articles up to 12 June 2017. The search was based on free text and MeSH terms (see Text Query in Supplementary data). Articles were searched by using ProQuest in two databases (EMBASE and MEDLINE) for the initial query and only in MEDLINE from 1 January 2014 and onwards as we did not have access to ProQuest after that date. We used the SCOPUS database to compile a list of citations from, as well as citations to, the articles considered relevant after abstract and full text review of the initial search. Articles in this citation analysis were also subjected to abstract and full text review as detailed below. Inclusion and exclusion criteria for abstract review Eligible studies reported in vivo measurements of β2M clearance from the body (primary outcome of this meta-analysis). Second, we examined β2M reduction ratio and/or β2M mass removal from the body in human subjects receiving HFD, HDF or hemofiltration among the studies reporting β2M clearance measurements. Studies performed before 2001, in vitro studies, review studies and meta-analyses were excluded along with studies not involving extracorporeal circuits (e.g. peritoneal dialysis), mathematical simulations without experimental data, and studies on extracorporeal circuits perfused in a closed loop manner with non-blood fluid (crystalloid or colloid) or ex vivo blood. Process Two reviewers (M.-E.R. and G.T.) independently screened potentially relevant titles and abstracts to ensure that the identified studies met the inclusion criteria and none of the exclusion criteria. Then the abstract review was adjudicated by C.P.A. All adjudicated papers were selected for full text review by M.-E.R. and C.P.A. to ensure they met the full text inclusion criteria for the meta-analysis. Full text review for papers written in Chinese was performed by Y.-H.N. and Z.X. Abstract and full text criteria are provided in the Supplementary data. Citation analysis was carried out by M.-E.R. and C.P.A. using the same abstract and full text criteria as the initial search. Data extraction We did not restrict articles by language. Data for the articles in English were extracted from tables and figures by M.-E.R. and C.P.A. Information from non-English publications was extracted from the abstract and the tables in the text. Data for the articles in Chinese were extracted from tables and figures by Y.-H.N. and Z.X. All data were inserted into standardized data collection forms and imported into an Excel spreadsheet. Measurements extracted included: (i) kinetic parameters [type of therapy, flow pump parameters, membrane surface area (MSA), dialyzer material, dialysis session duration, ultrafiltration volumes, session frequency] and (ii) β2M body clearance measurements, mass removal and reduction ratios. Volumes infused and ultra-filtered were converted from L to mL/min to account for the confounding role of dialysis session duration on convective clearance. For studies for which we had individual patient-level data (i.e. HEMO), we aggregated measurements to distinct groups defined by the type of dialyzer used, prior to analysis. Dialyzer specifications (Kuf: ultrafiltration coefficient, MSA) were downloaded from the manufacturer’s brochures and if those were not available (e.g. discontinued products), from dialysis textbooks and articles in the literature. Quality assessment Quality metrics of the included studies were assessed independently by two reviewers (C.P.A. and M.-E.R.) using the Effective Public Health Project Quality Assessment Tool for Quantitative Studies (EPHPP) (see Table S1) [24]. This tool was developed by the Effective Public Health Project, Canada and was chosen because it covers any quantitative study design. The latter was a particularly desirable feature for our project, which included RCTs, non-randomized controlled and uncontrolled studies. This quality assessment tool is comprised of the following components: selection bias, study design, confounders, blinding, data collection methods, withdrawals and dropouts, intervention integrity and analyses. Each section is rated as strong, moderate or weak by each reviewer. At the end, a global rating for the meta-analysis is provided. Statistical analysis Most of the studies included, reported on multiple ‘configurations’, i.e. combinations of dialysis operational parameters (e.g. pump flow rates, infusion volume, dialyzers) in the same patient groups. For this meta-analysis, a multi-level random effects model was adopted to account for clustering of measurements within the same configurations and within the same study. Despite the computational complexity, this approach is conceptually similar to using a paired t-test for the analysis of matched sample data. One subtle feature of this approach is that it enforces a form of averaging of multiple measurements from the same study. For studies reporting instantaneous clearance values, this implies that our object of analysis is the average of the instantaneous clearances. This quantity may not be much different from the average clearance computed via other means (e.g. pooled dialysate samples or pre-post β2M measurements), even though the individual measurements averaged may be far from it, e.g. due to loss of dialyzer performance over time. We opted for this approach, because we feel that the clinically relevant quantity is the capacity of the dialyzer to remove β2M over the entire course of the treatment (average clearance) rather than at any given point in time. This modeling was conducted separately for studies of convective and diffusive therapies reporting β2M clearance and together for studies of convective and diffusive therapies reporting β2M mass removal. Clearance values, reduction ratios and mass removal of β2M were summarized and heterogeneity was assessed graphically by the use of forest plots. Meta-regression models were utilized to statistically assess heterogeneity. For these models, the same multi-level structure was used as the one that was used to generate the forest plots. Univariate meta-regressions, assessing each variable in isolation, were followed by multivariable meta-regressions adjusting for more than one study characteristics. Variables were selected by univariate meta-regression analyses at the level of P = 0.05 if >70% of the studies were available for these analyses. The Restricted Maximum Likelihood (REML) approach was used to derive unbiased point estimates of dialysis relevant parameters (themselves treated as fixed effects) but at the expense of wider confidence intervals (CIs) for these models. Analysis of variance (ANOVA) was used to assess the global statistical significance of study characteristics with more than two levels (e.g. type of dialysis procedure) by comparing models that adjusted for these characteristics versus the models that did not. ANOVA tests were carried out in models fitted with conventional Maximum Likelihood approach, since these tests cannot be applied to compare models with different fixed effects specifications when REML is used. Operational parameters of clinical interest (e.g. substitution volume flows or year of the study) were forced into the models even if not significant in univariate models. Secular trends in the performance of the dialyzers over time were assessed by including the year of the publication as a covariate in the models. In these analyses, 2001 was taken as Year 0 and the secular trend was defined as a linear change in the outcome (e.g. clearance) with each subsequent year. Outcomes explored with meta-regression models were β2M clearance, β2M mass removal and the pre-dialysis and post-dialysis β2M reduction ratio. All analyses were performed in R statistical software (version 3.1.1) with the package metaphor [25]. RESULTS Study search results Electronic searches from 1 January 2001 to 12 June 2017 identified 638 potentially relevant reports. Of these, 481 were excluded after title and abstract review. After adjudication, 150 articles were selected for full text review and 53 relevant articles were identified (52 were published before 2014). Out of these, 47 articles reported aggregate (group data) and 5 studies reported patient-level data. In addition, the HEMO study (one of the studies identified in the initial search) provided data about 984 patients with 3967 measurements in non-reused dialyzers (most dialyzers were reused in HEMO). These measurements were taken from the HEMO analytic data files distributed by the National Institutes for Digestive Diabetes and Kidney Diseases (NIDDK), made available to our group through a data use agreement. Citation analysis of these 53 papers in SCOPUS identified 673 potentially relevant studies; we screened out 622 papers based on abstract review and selected 109 for full text review. Full text review uncovered 34 papers that had been identified during the initial search and 16 papers with relevant clearance data. A summary flow diagram is shown in Figure 1. The overall final study population for this meta-analysis consisted of 69 studies of 1879 patients with 6771 available measurements. FIGURE 1 View largeDownload slide Flow diagram of the literature search. FIGURE 1 View largeDownload slide Flow diagram of the literature search. Study characteristics Table 1 presents the characteristics of the patients that participated in the included studies, such as number of patients, age, gender, time on chronic dialysis therapy and their pre-dialysis weight. The same table details characteristics of the included studies, which fell into two main categories: comparisons of different types of dialyzers (46 on HFD) and comparisons of different types of convective dialysis therapies [31 studies on post-dilution HDF (post-HDF), 6 on pre-dilution HDF (pre-HDF), 15 on mid-dilution HDF (mid-HDF), 5 on mixed HDF (mixed-HDF), 2 studies on pre-dilution hemofiltration (pre-HF) and 2 studies on post-dilution hemofiltration (post-HF)]. These studies used a wide variety of dialyzer membrane material, e.g. cellulose acetate (CA, n = 4), polysulfone (PS, n = 146), polymethylmethacrylate (PMMA, n = 2), polyacrylnitrile (PAN, n = 2) and polyarylethersulfone (PAES, n = 97). All included studies enrolled patients under chronic dialysis regimens. Participant numbers were highly variable and ranged from 5 to 52. Only one study (HEMO [20]) had 984 participants. Clearances (mL/min), reduction ratios of β2M and/or β2M mass removal (mg or g/session) were measured and reported either in the blood side (serum or plasma) or in the dialysate side at a single time point during the dialysis session (instantaneous) or as average over the course of the treatment. A wide variety of methods were used for the calculation of clearance. The formulas and the numerical aspects of these approaches are summarized in the Supplementary data. Other study characteristics such as blood and dialysate flow rate, treatment duration, substitution fluid rate and MSA are reported as average and standard errors in Table 1. Table 1 Characteristics of the studies analyzed First Author Year N N meas Female Age Vintage PreWt Modality Material MSA QB QD Qinf Duration Leto [26] 2001 15 30 40.0% 45.7 (—) 156.3 (—) (—) HFD CA/PS 1.3 (0.1) 250.0 (0.0) 600.0 (0.0) (—) 240.0 (0.0) Xu [27] 2001 10 10 40.0% 70.2 (5.6) 71.2 (37.0) 63.9 (10.6) HFD PS 1.8 (0.5) (—) (—) (—) 300.0 (0.0) Yamada [28] 2001 28 28 39.0% 58.1 (16.4) 64.0 (47.0) 49.0 (8.0) HFD PS 1.42 (0.0) 188.0 (18.0) 500.0 (0.0) (—) 237.0 (18.0) Stiller [29] 2002 15 15 73.0% 54.3 (10.2) 134.0 (100.6) (—) HFD PAES/PS 1.24 (0.1) (—) (—) (—) 240.0 (0.0) Eknoyan [20] 2002 984 3967 59.0% 58.6 (13.7) 63.1 (59.2) 71.8 (1.5) HFD PMMA/CA/PS/ PAN/PAES 1.8 (0.2) 372.4 (8.8) 671.8 (10.1) (—) 204.5 (2.7) Ding [30] 2002 12 36 33.0% 49.7 (11.3) 83.5 (76.7) (—) pre-HDF/post-HDF PS 1.3 (0.0) 250.0 (0.0) 616.7 (2.9) 92.5 (3.0) 282.5 (29.6) Klingel [31] 2002 22 22 0.0% 61.4 (—) (—) 74.6 (0.0) HFD PS 1.3 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 228.8 (0.0) Mann [32] 2003 5 5 0.0% (—) (—) (—) HFD PS 1.6 (0.05) (—) (—) (—) 240.0 (0.0) Mandolfo [33] 2003 8 16 0.0% 61.4 (—) (—) 68 (8.6) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 550.0 (0.0) 30.0 (0.0) 240.0 (0.0) Pedrini [34] 2003 20 20 35.0% 63.0 (17.0) 116.4 (86.4) 60.3 (12.6) post-HDF/mixed-HDF PS 2.1 (0.0) 403.0 (55.9) 580.2 (36.6) 219.6 (36.6) 227.0 (17.7) Ward [35] 2003 12 24 41.6% 53.0 (13.0) 63.0 (18.3) (—) HFD PS 1.8 (0.0) 410.0 (1.9) 700.0 (0.0) (—) 228.0 (11.7) Bammens [36] 2004 14 70 28.6% 66.6 (3.1) 24.8 (10.0) 62.10 (1.94) HFD/pre-HDF/post-HDF PS 1.8 (0.0) 323.9 (116.3) 500.0 (0.0) 87.0 (0.0) 230.0 (0.0) Yamashita [37] 2004 5 5 80.0% (—) (—) (—) post-HF PS 1.8 (0.0) (—) (—) 84.2 (18.8) 120.0 (0.0) Emiliani [38] 2004 10 10 20.0% 66.0 (18.0) 80.0 (36.0) 66.2 (7.5) mid-HDF PAES 2.6 (0.0) 312.0 (18.0) 500.0 (0.0.0) 43.6 (7.2) 240.0 (10.0) Leypoldt [39] 2004 22 88 37.5% 61.0 (18.0) (—) 80.3 (19.4) HFD PS 1.77 (0.0) 338.0 (49.6) 540.0 (60.0) (—) 178.5 (19.0) Lucchi [40] 2004 10 20 40.0% 61.1 (8.9) 51.8 (35.9) (—) HFD/post-HDF PS 1.6 (0.0) 300.0 (0.0) 625.0 (0.0) 20.9 (0.0) 240.0 (0.0) Pisitkun [41] 2004 9 18 22% 48.0 (6.1) 51.4 (42.0) 55.2 (8.3) HFD/mid-HDF PS 2.7 (0.9) 475.0 (36.4) 800.0 (0.0) 59.9 (9.7) 240.0 (0.0) Tonelli [42] 2004 5 15 0.0% (—) (—) (—) HFD PS 1.8 (0.0) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Santoro [43] 2005 20 40 0.0% (—) (—) (—) HFD/mid-HDF PAES 1.90 (0.0) 363.0 (39.2) (—) 58.0 (11.9) (—) Brendolan [44] 2005 3 24 0.0% (—) (—) (—) HFD/post-HDF PS 2.2 (0.4) 333.3 (12.0) 500.0 (0.0) 16.6 (1.9) 226.1 (14.5) Padrini [45] 2005 11 22 36.4% 66.8 (11.9) 80.9 (66.9) 64.1 (9.2) post-HF/pre-HF PAES 2.1 (0.0) 327.8 (22.3) (—) 186 (40.1) 240 (6) Petras [46] 2005 6 36 0.0% 56.0 (16.0) 86.0 (50.0) (—) HFD/post-HDF/Pre-HF PAES 2.1 (0.0) 350.0 (0.0) 500.0 (0.0) 95.0 (0.0) 240.0 (0.0) Krieter [47] 2005a 5 5 60.0% 52.0 (22) (—) 68.5 (27.5) mid-HDF PAES 1.9 (0.0) 400.0 (0.0) 800.0 (0.0) 200.0 (0.0) 205.0 (15.0) Krieter [48] 2005b 10 40 30.0% 57.3 (13.7) 99.6 (92.4) 66.3 (10.4) mid-HDF/post-HDF PAES/PS 1.9 (0.1) 400.0 (0.0) 550.0 (0.0) 148.3 (2.9) 240 (23.4) Evenepoel [49] 2006 20 20 25.0% 68.8 (10.9) 19.3 (31.5) 59.9 (7.9) HFD PS 1.8 (0.0) 322.7 (21.6) 500.0 (0.0) (—) 230.0 (0.0) Mandolfo [50] 2006 12 18 66.7% 69.0 (9.0) 117.6 (69.6) 65.2 (8.1) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 700.0 (0.0) 50.0 (0.0) 240.0 (0.0) Nakashima [51] 2006 12 24 0.0% 49.1 (12.1) 127.2 (73.2) 66.8 (12.7) HFD PS 2.10 (0.0) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Panich [52] 2006 10 20 50.0% 58.2 (14.7) (—) 54.2 (3.8) HFD/post-HDF PS 1.8 (0.0) 425.3 (12.6) (—) 61.6 (0.6) 240.0 (0.0) Pedrini [53] 2006 12 72 25.0% 64.2 (6.6) 45.0 (38.0) 64.9 (11.2) mixed-HDF PS 2.1 (0.0) 422.0 (37.9) 609.0 (27.9) 178.0 (20.9) 218.0 (25.9) Potier [54] 2007 6 18 0.0% (—) (—) (—) post-HDF/pre-HDF/mid-HDF PAES 1.90 (0.0) 360.0 (0.0) 500.0 (0.0) 175.0 (0.0) (—) Feliciani [55] 2007 10 30 20.0% 64.7 (8.0) 54.7 (57.7) 73.25 (12.5) mixed-HDF/mid-HDF PAES/PS 1.85 (0.1) 385.5 (18.3) 609.0 (20.7) 167.5 (14.1) 231.5 (16.8) Krieter [56] 2007 8 32 62.5% 62.1 (13.8) 76.0 (55.3) 68.5 (7.1) HFD PS/PAES 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 236.0 (10.5) Santoro [57] 2007 8 16 50.0% 56.6 (23.6) (—) (—) mid-HDF PAES 1.9 (0.0) 306.5 (10.3) (—) 100.0 (0.0) 231.0 (10.3) Tiranathanagul [58] 2007 12 48 33.0% 54.2 (13.6) 42.0 (32.3) 62.85 (9.4) post-HDF/mid-HDF PS 2.7 (0.9) 416.7 (24.1) 800.0 (0.0) 113.0 (6.0) 240.0 (0.0) Abe [59] 2008 15 45 40% 65.5 (13.2) 72.9 (63.8) (—) HFD PMMA/CA/PS 1.5 (0.1) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Eloot [60] 2008 9 27 55.5% 71.0 (10.0) 19.0 (12.0) 79.0 (11.5) HFD PS 1.8 (0.0) 260.0 (0.0) 260.0 (0.0) (—) 360.0 (0.0) Mandolfo [61] 2008 8 16 37.5% 72.2 (4.8) 62.0 (24.0) 61.7 (11) HFD/mid-HDF PAES 1.9 (0.0) 251.5 (32.4) 700.0 (0.0) 56.0 (4.8) 240.0 (0.0) Spalding [62] 2008 12 12 50.0% 65.3 (12.9) (—) (—) HFD/post-HDF (—) (—) 358.4 (84.4) 800.0 (0.0) 37.5 (12.6) 197.4 (55.3) Krieter [63] 2008a 8 40 25.0% 64.0 (16.0) 70.0 (74.0) 74.2 (10.7) HFD/post-HDF PAES/PS 1.7 (0.1) 300.0 (0.0) 460.0 (0.0) 40.0 (0.0) 240.0 (0.0) Ouseph [64] 2008a 12 48 25.0% 57.0 (4.0) 52.0 (17.0) 81.3 (4.35) HFD PS/PAES 1.65 (0.1) 382.0 (4.8) 800.0 (0.0) (—) 219.0 (5.3) Ouseph [64] 2008b 12 60 41.6% 46.0 (3.0) 48.0 (8.0) 84.2 (6.75) HFD PS/PAES 1.90 (0.2) 404.0 (1.0) 800.0 (0.0) (—) 240.0 (0.0) Krieter [65] 2008b 8 48 37.5% 63.0 (14.0) 77.5 (38.9) 76.5 (11.15) HFD PAES/PS 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Joyeux [66] 2008 20 40 35.0% 58.0 (20.9) 46.0 (46.1) 71.5 (10.2) HFD/post-HDF PAES 2.10 (0.0) 310.5 (33.3) (—) 31.8 (5.2) 235.5 (14.6) Lee [67] 2009 8 16 50.0% 68.7 (19.1) 50.3 (53.1) (—) HFD PAES 1.1 (0.0) 325.0 (24.6) 500.0 (0.0) (—) 255.0 (14.8) Meert [68] 2009 14 42 50.0% 63.5 (17.0) 30.2 (36.0) (—) pre-HDF/pre-HF/post-HDF PAES 1.8 (0.2) 312.3 (15.6) 384.7 (5.0) 185.7 (20.7) 249.3 (13.0) Pedrini [69] 2009 15 90 20.0% 67.3 (8.7) 44.1 (20.8) 76.9 (13.8) mid-HDF PAES 2.1 (0.2) 378.5 (27.4) 599.5 (6.5) 167.5 (9.7) 223.0 (21.4) Susantitaphong [70] 2009 12 36 66.6% 59.5 (13.5) 81.6 (52.8) 57.5 (11.6) Pre-HDF/mid-HDF/post-HDF PAES 2.2 (0.1) 440.3 (19.9) 554.2 (10.4) 245.9 (2.1) 240.0 (0.0) Troidle [71] 2009 8 8 0.0% 45.0 (7.0) (—) (—) HFD PS 1.8 (0.0) 400.0 (0.0) 600.0 (0.0) (—) 480.0 (0.0) Wang [72] 2009 18 54 27.8% 46.9 (9.6) 52.5 (—) (—) HFD PS 1.5 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Bhimani [73] 2010 12 144 75.0% 60.0 (4.0) 81.0 (19.0) (—) HFD PS/PAES 1.7 (0.2) 400.0 (1.8) 550.0 (0.0) (—) (—) Gasco [74] 2010 16 263 50.0% 52.7 (—) (—) (—) post-HDF PS 1.8 (0.0) 323.0 (0.0) 800.0 (0.0) (—) 242.0 (0.0) Kohn [75] 2010 5 38 60.0% (—) 142.0 (60.9) 86.0 (29.0) HFD PAES (—) 425.0 (75.0) 200.0 (0.0) (—) 174.0 (15.0) Krieter [76] 2010 8 64 12.5% 63.0 (12.0) 81.8 (144.0) 70.8 (17.5) HFD/post-HDF PAES 1.9 (0.0) 378.0 (31.1) 500.0 (0.0) 47.0 (6.0) 229.0 (20.7) Park [77] 2010 52 52 52.0% 54.0 (12.4) 112.7 (188.6) (—) HFD PS 1.3 (0.0) 237.0 (23.0) 500.0 (0.0) (—) 240.0 (12.0) Basile [78] 2011 11 22 18.2% 54.1 (17.8) 78.0 (60.2) 69.1 (9.9) HFD PS 1.8 (0.0) 270.0 (0.0) 270.0 (0.0) (—) 469.1 (2.7) Ficheux [79] 2011 18 54 0.0% 79.7 (1.7) (—) 66.1 (2.3) HFD PS 2.2 (0.1) 318.0 (2.0) 500.0 (9.8) (—) 222.0 (2.9) Pedrini [80] 2011 15 60 33.3% 67.2 (8.3) (—) 73.1 (14.0) post-HDF/mid-HDF PS/PAES 2.3 (0.1) 374.0 (34.0) 580.0 (39.7) 147.5 (11.1) 224.0 (18.5) Panichi [81, 82] 2012 30 180 33.3% 55.9 (14.0) 58.0 (59.0) (—) post-HDF PAES 2.1 (0.0) 313.5 (32.7) 600.0 (0.0) 78.1 (0.0) 235.0 (13.8) Susantitaphong [83] 2012 12 48 66.6% 57.8 (14.8) 43.2 (42.0) 55.5 (11.1) mid-HDF/mixed-HDF PAES 2.2 (0.0) 425.0 (24.5) 600.0 (0.0) 200.0 (0.0) 240.0 (0.0) Tessitore [84] 2012 26 26 53.9% 63.0 (12.0) (—) (—) HFD PP 0.7 (0.0) 297.0 (32.0) 500.0 (0.0) (—) 230.0 (13.0) von Albertini [85] 2013 12 35 0.0% (—) (—) (—) HFD/post-HDF PAES/PS 1.8 (0.0) 417.1 (0.0) 667.4 (0.0) 30.7 (0.0) 206.3 (22.5) Heaf [86] 2013 12 96 30.0% 63.1 (11.7) 78.0 (52.8) 79.2 (17.8) HFD PAES 2.0 (0.0) 276.0 (38.7) 500.0 (0.0) (—) 240.0 (0.0) Melo [87] 2014 14 28 50.0% 48.9 (14.4) (—) 76.35 (19.63) HFD/post-HDF PS 2.0 (0.0) 375.0 (8.2) 760.0 (0.0) 40.0 (0.0) 115.7 (16.8) Pedrini [88] 2014 16 32 18.8% (—) (—) 77.60 (10.78) post-HDF PAES/PS 2.20 (0.10) 388.0 (25.9) 574.5 (39.0) 121.0 (11.9) 226.5 (13.7) Cornelis [89] 2014 13 52 23.1% 53.6 (20.4) 49.0 (29.0) (—) HFD/post-HDF PS 1.8 (0.0) 286.0 (4.8) 573.7 (13.1) 30.1 (1.4) 366.3 (3.9) Potier [90] 2016 6 24 66.7% 65.4 (25.5) 68.6 (43.7) 73.9 (2.1) HFD/post-HDF/ mixed-HDF/pre-HDF PS 2.3 (0.0) 339.4 (3.4) 600.0 (0.0) 122.1 (5.1) 240.0 (0.0) Gayrard [91] 2017 12 48 50.0% 73.0 (12.0) (—) 71.0 (1.9) HFD/post-HDF PS 1.8 (0.0) 366.3 (5.1) 602.3 (1.0) 51.8 (1.1) 233.6 (2.9) Kirsch [92] 2017 39 59 28.2% 60.5 (13.6) 63.1 (43.8) 80.2 (18.4) post-HDF/HFD PS 1.9 (0.1) 368.1 (12.8) (—) 27.5 (1.4) 252.4 (11.8) First Author Year N N meas Female Age Vintage PreWt Modality Material MSA QB QD Qinf Duration Leto [26] 2001 15 30 40.0% 45.7 (—) 156.3 (—) (—) HFD CA/PS 1.3 (0.1) 250.0 (0.0) 600.0 (0.0) (—) 240.0 (0.0) Xu [27] 2001 10 10 40.0% 70.2 (5.6) 71.2 (37.0) 63.9 (10.6) HFD PS 1.8 (0.5) (—) (—) (—) 300.0 (0.0) Yamada [28] 2001 28 28 39.0% 58.1 (16.4) 64.0 (47.0) 49.0 (8.0) HFD PS 1.42 (0.0) 188.0 (18.0) 500.0 (0.0) (—) 237.0 (18.0) Stiller [29] 2002 15 15 73.0% 54.3 (10.2) 134.0 (100.6) (—) HFD PAES/PS 1.24 (0.1) (—) (—) (—) 240.0 (0.0) Eknoyan [20] 2002 984 3967 59.0% 58.6 (13.7) 63.1 (59.2) 71.8 (1.5) HFD PMMA/CA/PS/ PAN/PAES 1.8 (0.2) 372.4 (8.8) 671.8 (10.1) (—) 204.5 (2.7) Ding [30] 2002 12 36 33.0% 49.7 (11.3) 83.5 (76.7) (—) pre-HDF/post-HDF PS 1.3 (0.0) 250.0 (0.0) 616.7 (2.9) 92.5 (3.0) 282.5 (29.6) Klingel [31] 2002 22 22 0.0% 61.4 (—) (—) 74.6 (0.0) HFD PS 1.3 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 228.8 (0.0) Mann [32] 2003 5 5 0.0% (—) (—) (—) HFD PS 1.6 (0.05) (—) (—) (—) 240.0 (0.0) Mandolfo [33] 2003 8 16 0.0% 61.4 (—) (—) 68 (8.6) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 550.0 (0.0) 30.0 (0.0) 240.0 (0.0) Pedrini [34] 2003 20 20 35.0% 63.0 (17.0) 116.4 (86.4) 60.3 (12.6) post-HDF/mixed-HDF PS 2.1 (0.0) 403.0 (55.9) 580.2 (36.6) 219.6 (36.6) 227.0 (17.7) Ward [35] 2003 12 24 41.6% 53.0 (13.0) 63.0 (18.3) (—) HFD PS 1.8 (0.0) 410.0 (1.9) 700.0 (0.0) (—) 228.0 (11.7) Bammens [36] 2004 14 70 28.6% 66.6 (3.1) 24.8 (10.0) 62.10 (1.94) HFD/pre-HDF/post-HDF PS 1.8 (0.0) 323.9 (116.3) 500.0 (0.0) 87.0 (0.0) 230.0 (0.0) Yamashita [37] 2004 5 5 80.0% (—) (—) (—) post-HF PS 1.8 (0.0) (—) (—) 84.2 (18.8) 120.0 (0.0) Emiliani [38] 2004 10 10 20.0% 66.0 (18.0) 80.0 (36.0) 66.2 (7.5) mid-HDF PAES 2.6 (0.0) 312.0 (18.0) 500.0 (0.0.0) 43.6 (7.2) 240.0 (10.0) Leypoldt [39] 2004 22 88 37.5% 61.0 (18.0) (—) 80.3 (19.4) HFD PS 1.77 (0.0) 338.0 (49.6) 540.0 (60.0) (—) 178.5 (19.0) Lucchi [40] 2004 10 20 40.0% 61.1 (8.9) 51.8 (35.9) (—) HFD/post-HDF PS 1.6 (0.0) 300.0 (0.0) 625.0 (0.0) 20.9 (0.0) 240.0 (0.0) Pisitkun [41] 2004 9 18 22% 48.0 (6.1) 51.4 (42.0) 55.2 (8.3) HFD/mid-HDF PS 2.7 (0.9) 475.0 (36.4) 800.0 (0.0) 59.9 (9.7) 240.0 (0.0) Tonelli [42] 2004 5 15 0.0% (—) (—) (—) HFD PS 1.8 (0.0) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Santoro [43] 2005 20 40 0.0% (—) (—) (—) HFD/mid-HDF PAES 1.90 (0.0) 363.0 (39.2) (—) 58.0 (11.9) (—) Brendolan [44] 2005 3 24 0.0% (—) (—) (—) HFD/post-HDF PS 2.2 (0.4) 333.3 (12.0) 500.0 (0.0) 16.6 (1.9) 226.1 (14.5) Padrini [45] 2005 11 22 36.4% 66.8 (11.9) 80.9 (66.9) 64.1 (9.2) post-HF/pre-HF PAES 2.1 (0.0) 327.8 (22.3) (—) 186 (40.1) 240 (6) Petras [46] 2005 6 36 0.0% 56.0 (16.0) 86.0 (50.0) (—) HFD/post-HDF/Pre-HF PAES 2.1 (0.0) 350.0 (0.0) 500.0 (0.0) 95.0 (0.0) 240.0 (0.0) Krieter [47] 2005a 5 5 60.0% 52.0 (22) (—) 68.5 (27.5) mid-HDF PAES 1.9 (0.0) 400.0 (0.0) 800.0 (0.0) 200.0 (0.0) 205.0 (15.0) Krieter [48] 2005b 10 40 30.0% 57.3 (13.7) 99.6 (92.4) 66.3 (10.4) mid-HDF/post-HDF PAES/PS 1.9 (0.1) 400.0 (0.0) 550.0 (0.0) 148.3 (2.9) 240 (23.4) Evenepoel [49] 2006 20 20 25.0% 68.8 (10.9) 19.3 (31.5) 59.9 (7.9) HFD PS 1.8 (0.0) 322.7 (21.6) 500.0 (0.0) (—) 230.0 (0.0) Mandolfo [50] 2006 12 18 66.7% 69.0 (9.0) 117.6 (69.6) 65.2 (8.1) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 700.0 (0.0) 50.0 (0.0) 240.0 (0.0) Nakashima [51] 2006 12 24 0.0% 49.1 (12.1) 127.2 (73.2) 66.8 (12.7) HFD PS 2.10 (0.0) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Panich [52] 2006 10 20 50.0% 58.2 (14.7) (—) 54.2 (3.8) HFD/post-HDF PS 1.8 (0.0) 425.3 (12.6) (—) 61.6 (0.6) 240.0 (0.0) Pedrini [53] 2006 12 72 25.0% 64.2 (6.6) 45.0 (38.0) 64.9 (11.2) mixed-HDF PS 2.1 (0.0) 422.0 (37.9) 609.0 (27.9) 178.0 (20.9) 218.0 (25.9) Potier [54] 2007 6 18 0.0% (—) (—) (—) post-HDF/pre-HDF/mid-HDF PAES 1.90 (0.0) 360.0 (0.0) 500.0 (0.0) 175.0 (0.0) (—) Feliciani [55] 2007 10 30 20.0% 64.7 (8.0) 54.7 (57.7) 73.25 (12.5) mixed-HDF/mid-HDF PAES/PS 1.85 (0.1) 385.5 (18.3) 609.0 (20.7) 167.5 (14.1) 231.5 (16.8) Krieter [56] 2007 8 32 62.5% 62.1 (13.8) 76.0 (55.3) 68.5 (7.1) HFD PS/PAES 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 236.0 (10.5) Santoro [57] 2007 8 16 50.0% 56.6 (23.6) (—) (—) mid-HDF PAES 1.9 (0.0) 306.5 (10.3) (—) 100.0 (0.0) 231.0 (10.3) Tiranathanagul [58] 2007 12 48 33.0% 54.2 (13.6) 42.0 (32.3) 62.85 (9.4) post-HDF/mid-HDF PS 2.7 (0.9) 416.7 (24.1) 800.0 (0.0) 113.0 (6.0) 240.0 (0.0) Abe [59] 2008 15 45 40% 65.5 (13.2) 72.9 (63.8) (—) HFD PMMA/CA/PS 1.5 (0.1) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Eloot [60] 2008 9 27 55.5% 71.0 (10.0) 19.0 (12.0) 79.0 (11.5) HFD PS 1.8 (0.0) 260.0 (0.0) 260.0 (0.0) (—) 360.0 (0.0) Mandolfo [61] 2008 8 16 37.5% 72.2 (4.8) 62.0 (24.0) 61.7 (11) HFD/mid-HDF PAES 1.9 (0.0) 251.5 (32.4) 700.0 (0.0) 56.0 (4.8) 240.0 (0.0) Spalding [62] 2008 12 12 50.0% 65.3 (12.9) (—) (—) HFD/post-HDF (—) (—) 358.4 (84.4) 800.0 (0.0) 37.5 (12.6) 197.4 (55.3) Krieter [63] 2008a 8 40 25.0% 64.0 (16.0) 70.0 (74.0) 74.2 (10.7) HFD/post-HDF PAES/PS 1.7 (0.1) 300.0 (0.0) 460.0 (0.0) 40.0 (0.0) 240.0 (0.0) Ouseph [64] 2008a 12 48 25.0% 57.0 (4.0) 52.0 (17.0) 81.3 (4.35) HFD PS/PAES 1.65 (0.1) 382.0 (4.8) 800.0 (0.0) (—) 219.0 (5.3) Ouseph [64] 2008b 12 60 41.6% 46.0 (3.0) 48.0 (8.0) 84.2 (6.75) HFD PS/PAES 1.90 (0.2) 404.0 (1.0) 800.0 (0.0) (—) 240.0 (0.0) Krieter [65] 2008b 8 48 37.5% 63.0 (14.0) 77.5 (38.9) 76.5 (11.15) HFD PAES/PS 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Joyeux [66] 2008 20 40 35.0% 58.0 (20.9) 46.0 (46.1) 71.5 (10.2) HFD/post-HDF PAES 2.10 (0.0) 310.5 (33.3) (—) 31.8 (5.2) 235.5 (14.6) Lee [67] 2009 8 16 50.0% 68.7 (19.1) 50.3 (53.1) (—) HFD PAES 1.1 (0.0) 325.0 (24.6) 500.0 (0.0) (—) 255.0 (14.8) Meert [68] 2009 14 42 50.0% 63.5 (17.0) 30.2 (36.0) (—) pre-HDF/pre-HF/post-HDF PAES 1.8 (0.2) 312.3 (15.6) 384.7 (5.0) 185.7 (20.7) 249.3 (13.0) Pedrini [69] 2009 15 90 20.0% 67.3 (8.7) 44.1 (20.8) 76.9 (13.8) mid-HDF PAES 2.1 (0.2) 378.5 (27.4) 599.5 (6.5) 167.5 (9.7) 223.0 (21.4) Susantitaphong [70] 2009 12 36 66.6% 59.5 (13.5) 81.6 (52.8) 57.5 (11.6) Pre-HDF/mid-HDF/post-HDF PAES 2.2 (0.1) 440.3 (19.9) 554.2 (10.4) 245.9 (2.1) 240.0 (0.0) Troidle [71] 2009 8 8 0.0% 45.0 (7.0) (—) (—) HFD PS 1.8 (0.0) 400.0 (0.0) 600.0 (0.0) (—) 480.0 (0.0) Wang [72] 2009 18 54 27.8% 46.9 (9.6) 52.5 (—) (—) HFD PS 1.5 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Bhimani [73] 2010 12 144 75.0% 60.0 (4.0) 81.0 (19.0) (—) HFD PS/PAES 1.7 (0.2) 400.0 (1.8) 550.0 (0.0) (—) (—) Gasco [74] 2010 16 263 50.0% 52.7 (—) (—) (—) post-HDF PS 1.8 (0.0) 323.0 (0.0) 800.0 (0.0) (—) 242.0 (0.0) Kohn [75] 2010 5 38 60.0% (—) 142.0 (60.9) 86.0 (29.0) HFD PAES (—) 425.0 (75.0) 200.0 (0.0) (—) 174.0 (15.0) Krieter [76] 2010 8 64 12.5% 63.0 (12.0) 81.8 (144.0) 70.8 (17.5) HFD/post-HDF PAES 1.9 (0.0) 378.0 (31.1) 500.0 (0.0) 47.0 (6.0) 229.0 (20.7) Park [77] 2010 52 52 52.0% 54.0 (12.4) 112.7 (188.6) (—) HFD PS 1.3 (0.0) 237.0 (23.0) 500.0 (0.0) (—) 240.0 (12.0) Basile [78] 2011 11 22 18.2% 54.1 (17.8) 78.0 (60.2) 69.1 (9.9) HFD PS 1.8 (0.0) 270.0 (0.0) 270.0 (0.0) (—) 469.1 (2.7) Ficheux [79] 2011 18 54 0.0% 79.7 (1.7) (—) 66.1 (2.3) HFD PS 2.2 (0.1) 318.0 (2.0) 500.0 (9.8) (—) 222.0 (2.9) Pedrini [80] 2011 15 60 33.3% 67.2 (8.3) (—) 73.1 (14.0) post-HDF/mid-HDF PS/PAES 2.3 (0.1) 374.0 (34.0) 580.0 (39.7) 147.5 (11.1) 224.0 (18.5) Panichi [81, 82] 2012 30 180 33.3% 55.9 (14.0) 58.0 (59.0) (—) post-HDF PAES 2.1 (0.0) 313.5 (32.7) 600.0 (0.0) 78.1 (0.0) 235.0 (13.8) Susantitaphong [83] 2012 12 48 66.6% 57.8 (14.8) 43.2 (42.0) 55.5 (11.1) mid-HDF/mixed-HDF PAES 2.2 (0.0) 425.0 (24.5) 600.0 (0.0) 200.0 (0.0) 240.0 (0.0) Tessitore [84] 2012 26 26 53.9% 63.0 (12.0) (—) (—) HFD PP 0.7 (0.0) 297.0 (32.0) 500.0 (0.0) (—) 230.0 (13.0) von Albertini [85] 2013 12 35 0.0% (—) (—) (—) HFD/post-HDF PAES/PS 1.8 (0.0) 417.1 (0.0) 667.4 (0.0) 30.7 (0.0) 206.3 (22.5) Heaf [86] 2013 12 96 30.0% 63.1 (11.7) 78.0 (52.8) 79.2 (17.8) HFD PAES 2.0 (0.0) 276.0 (38.7) 500.0 (0.0) (—) 240.0 (0.0) Melo [87] 2014 14 28 50.0% 48.9 (14.4) (—) 76.35 (19.63) HFD/post-HDF PS 2.0 (0.0) 375.0 (8.2) 760.0 (0.0) 40.0 (0.0) 115.7 (16.8) Pedrini [88] 2014 16 32 18.8% (—) (—) 77.60 (10.78) post-HDF PAES/PS 2.20 (0.10) 388.0 (25.9) 574.5 (39.0) 121.0 (11.9) 226.5 (13.7) Cornelis [89] 2014 13 52 23.1% 53.6 (20.4) 49.0 (29.0) (—) HFD/post-HDF PS 1.8 (0.0) 286.0 (4.8) 573.7 (13.1) 30.1 (1.4) 366.3 (3.9) Potier [90] 2016 6 24 66.7% 65.4 (25.5) 68.6 (43.7) 73.9 (2.1) HFD/post-HDF/ mixed-HDF/pre-HDF PS 2.3 (0.0) 339.4 (3.4) 600.0 (0.0) 122.1 (5.1) 240.0 (0.0) Gayrard [91] 2017 12 48 50.0% 73.0 (12.0) (—) 71.0 (1.9) HFD/post-HDF PS 1.8 (0.0) 366.3 (5.1) 602.3 (1.0) 51.8 (1.1) 233.6 (2.9) Kirsch [92] 2017 39 59 28.2% 60.5 (13.6) 63.1 (43.8) 80.2 (18.4) post-HDF/HFD PS 1.9 (0.1) 368.1 (12.8) (—) 27.5 (1.4) 252.4 (11.8) N, number; N meas, number of measurements; Vintage, time on chronic intermittent dialysis in months; PreWt, pre-dialysis weight in kilograms; MSA, membrane surface area (in square meters); QB, blood flow rate (mL/min); QD, dialysis fluid flow rate (mL/min); Duration, the dialysis session (in min). For each parameter the table summarizes the mean and the SD over all arms in each study or a (—) if the relevant parameter could not be extracted from the paper. Table 1 Characteristics of the studies analyzed First Author Year N N meas Female Age Vintage PreWt Modality Material MSA QB QD Qinf Duration Leto [26] 2001 15 30 40.0% 45.7 (—) 156.3 (—) (—) HFD CA/PS 1.3 (0.1) 250.0 (0.0) 600.0 (0.0) (—) 240.0 (0.0) Xu [27] 2001 10 10 40.0% 70.2 (5.6) 71.2 (37.0) 63.9 (10.6) HFD PS 1.8 (0.5) (—) (—) (—) 300.0 (0.0) Yamada [28] 2001 28 28 39.0% 58.1 (16.4) 64.0 (47.0) 49.0 (8.0) HFD PS 1.42 (0.0) 188.0 (18.0) 500.0 (0.0) (—) 237.0 (18.0) Stiller [29] 2002 15 15 73.0% 54.3 (10.2) 134.0 (100.6) (—) HFD PAES/PS 1.24 (0.1) (—) (—) (—) 240.0 (0.0) Eknoyan [20] 2002 984 3967 59.0% 58.6 (13.7) 63.1 (59.2) 71.8 (1.5) HFD PMMA/CA/PS/ PAN/PAES 1.8 (0.2) 372.4 (8.8) 671.8 (10.1) (—) 204.5 (2.7) Ding [30] 2002 12 36 33.0% 49.7 (11.3) 83.5 (76.7) (—) pre-HDF/post-HDF PS 1.3 (0.0) 250.0 (0.0) 616.7 (2.9) 92.5 (3.0) 282.5 (29.6) Klingel [31] 2002 22 22 0.0% 61.4 (—) (—) 74.6 (0.0) HFD PS 1.3 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 228.8 (0.0) Mann [32] 2003 5 5 0.0% (—) (—) (—) HFD PS 1.6 (0.05) (—) (—) (—) 240.0 (0.0) Mandolfo [33] 2003 8 16 0.0% 61.4 (—) (—) 68 (8.6) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 550.0 (0.0) 30.0 (0.0) 240.0 (0.0) Pedrini [34] 2003 20 20 35.0% 63.0 (17.0) 116.4 (86.4) 60.3 (12.6) post-HDF/mixed-HDF PS 2.1 (0.0) 403.0 (55.9) 580.2 (36.6) 219.6 (36.6) 227.0 (17.7) Ward [35] 2003 12 24 41.6% 53.0 (13.0) 63.0 (18.3) (—) HFD PS 1.8 (0.0) 410.0 (1.9) 700.0 (0.0) (—) 228.0 (11.7) Bammens [36] 2004 14 70 28.6% 66.6 (3.1) 24.8 (10.0) 62.10 (1.94) HFD/pre-HDF/post-HDF PS 1.8 (0.0) 323.9 (116.3) 500.0 (0.0) 87.0 (0.0) 230.0 (0.0) Yamashita [37] 2004 5 5 80.0% (—) (—) (—) post-HF PS 1.8 (0.0) (—) (—) 84.2 (18.8) 120.0 (0.0) Emiliani [38] 2004 10 10 20.0% 66.0 (18.0) 80.0 (36.0) 66.2 (7.5) mid-HDF PAES 2.6 (0.0) 312.0 (18.0) 500.0 (0.0.0) 43.6 (7.2) 240.0 (10.0) Leypoldt [39] 2004 22 88 37.5% 61.0 (18.0) (—) 80.3 (19.4) HFD PS 1.77 (0.0) 338.0 (49.6) 540.0 (60.0) (—) 178.5 (19.0) Lucchi [40] 2004 10 20 40.0% 61.1 (8.9) 51.8 (35.9) (—) HFD/post-HDF PS 1.6 (0.0) 300.0 (0.0) 625.0 (0.0) 20.9 (0.0) 240.0 (0.0) Pisitkun [41] 2004 9 18 22% 48.0 (6.1) 51.4 (42.0) 55.2 (8.3) HFD/mid-HDF PS 2.7 (0.9) 475.0 (36.4) 800.0 (0.0) 59.9 (9.7) 240.0 (0.0) Tonelli [42] 2004 5 15 0.0% (—) (—) (—) HFD PS 1.8 (0.0) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Santoro [43] 2005 20 40 0.0% (—) (—) (—) HFD/mid-HDF PAES 1.90 (0.0) 363.0 (39.2) (—) 58.0 (11.9) (—) Brendolan [44] 2005 3 24 0.0% (—) (—) (—) HFD/post-HDF PS 2.2 (0.4) 333.3 (12.0) 500.0 (0.0) 16.6 (1.9) 226.1 (14.5) Padrini [45] 2005 11 22 36.4% 66.8 (11.9) 80.9 (66.9) 64.1 (9.2) post-HF/pre-HF PAES 2.1 (0.0) 327.8 (22.3) (—) 186 (40.1) 240 (6) Petras [46] 2005 6 36 0.0% 56.0 (16.0) 86.0 (50.0) (—) HFD/post-HDF/Pre-HF PAES 2.1 (0.0) 350.0 (0.0) 500.0 (0.0) 95.0 (0.0) 240.0 (0.0) Krieter [47] 2005a 5 5 60.0% 52.0 (22) (—) 68.5 (27.5) mid-HDF PAES 1.9 (0.0) 400.0 (0.0) 800.0 (0.0) 200.0 (0.0) 205.0 (15.0) Krieter [48] 2005b 10 40 30.0% 57.3 (13.7) 99.6 (92.4) 66.3 (10.4) mid-HDF/post-HDF PAES/PS 1.9 (0.1) 400.0 (0.0) 550.0 (0.0) 148.3 (2.9) 240 (23.4) Evenepoel [49] 2006 20 20 25.0% 68.8 (10.9) 19.3 (31.5) 59.9 (7.9) HFD PS 1.8 (0.0) 322.7 (21.6) 500.0 (0.0) (—) 230.0 (0.0) Mandolfo [50] 2006 12 18 66.7% 69.0 (9.0) 117.6 (69.6) 65.2 (8.1) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 700.0 (0.0) 50.0 (0.0) 240.0 (0.0) Nakashima [51] 2006 12 24 0.0% 49.1 (12.1) 127.2 (73.2) 66.8 (12.7) HFD PS 2.10 (0.0) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Panich [52] 2006 10 20 50.0% 58.2 (14.7) (—) 54.2 (3.8) HFD/post-HDF PS 1.8 (0.0) 425.3 (12.6) (—) 61.6 (0.6) 240.0 (0.0) Pedrini [53] 2006 12 72 25.0% 64.2 (6.6) 45.0 (38.0) 64.9 (11.2) mixed-HDF PS 2.1 (0.0) 422.0 (37.9) 609.0 (27.9) 178.0 (20.9) 218.0 (25.9) Potier [54] 2007 6 18 0.0% (—) (—) (—) post-HDF/pre-HDF/mid-HDF PAES 1.90 (0.0) 360.0 (0.0) 500.0 (0.0) 175.0 (0.0) (—) Feliciani [55] 2007 10 30 20.0% 64.7 (8.0) 54.7 (57.7) 73.25 (12.5) mixed-HDF/mid-HDF PAES/PS 1.85 (0.1) 385.5 (18.3) 609.0 (20.7) 167.5 (14.1) 231.5 (16.8) Krieter [56] 2007 8 32 62.5% 62.1 (13.8) 76.0 (55.3) 68.5 (7.1) HFD PS/PAES 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 236.0 (10.5) Santoro [57] 2007 8 16 50.0% 56.6 (23.6) (—) (—) mid-HDF PAES 1.9 (0.0) 306.5 (10.3) (—) 100.0 (0.0) 231.0 (10.3) Tiranathanagul [58] 2007 12 48 33.0% 54.2 (13.6) 42.0 (32.3) 62.85 (9.4) post-HDF/mid-HDF PS 2.7 (0.9) 416.7 (24.1) 800.0 (0.0) 113.0 (6.0) 240.0 (0.0) Abe [59] 2008 15 45 40% 65.5 (13.2) 72.9 (63.8) (—) HFD PMMA/CA/PS 1.5 (0.1) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Eloot [60] 2008 9 27 55.5% 71.0 (10.0) 19.0 (12.0) 79.0 (11.5) HFD PS 1.8 (0.0) 260.0 (0.0) 260.0 (0.0) (—) 360.0 (0.0) Mandolfo [61] 2008 8 16 37.5% 72.2 (4.8) 62.0 (24.0) 61.7 (11) HFD/mid-HDF PAES 1.9 (0.0) 251.5 (32.4) 700.0 (0.0) 56.0 (4.8) 240.0 (0.0) Spalding [62] 2008 12 12 50.0% 65.3 (12.9) (—) (—) HFD/post-HDF (—) (—) 358.4 (84.4) 800.0 (0.0) 37.5 (12.6) 197.4 (55.3) Krieter [63] 2008a 8 40 25.0% 64.0 (16.0) 70.0 (74.0) 74.2 (10.7) HFD/post-HDF PAES/PS 1.7 (0.1) 300.0 (0.0) 460.0 (0.0) 40.0 (0.0) 240.0 (0.0) Ouseph [64] 2008a 12 48 25.0% 57.0 (4.0) 52.0 (17.0) 81.3 (4.35) HFD PS/PAES 1.65 (0.1) 382.0 (4.8) 800.0 (0.0) (—) 219.0 (5.3) Ouseph [64] 2008b 12 60 41.6% 46.0 (3.0) 48.0 (8.0) 84.2 (6.75) HFD PS/PAES 1.90 (0.2) 404.0 (1.0) 800.0 (0.0) (—) 240.0 (0.0) Krieter [65] 2008b 8 48 37.5% 63.0 (14.0) 77.5 (38.9) 76.5 (11.15) HFD PAES/PS 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Joyeux [66] 2008 20 40 35.0% 58.0 (20.9) 46.0 (46.1) 71.5 (10.2) HFD/post-HDF PAES 2.10 (0.0) 310.5 (33.3) (—) 31.8 (5.2) 235.5 (14.6) Lee [67] 2009 8 16 50.0% 68.7 (19.1) 50.3 (53.1) (—) HFD PAES 1.1 (0.0) 325.0 (24.6) 500.0 (0.0) (—) 255.0 (14.8) Meert [68] 2009 14 42 50.0% 63.5 (17.0) 30.2 (36.0) (—) pre-HDF/pre-HF/post-HDF PAES 1.8 (0.2) 312.3 (15.6) 384.7 (5.0) 185.7 (20.7) 249.3 (13.0) Pedrini [69] 2009 15 90 20.0% 67.3 (8.7) 44.1 (20.8) 76.9 (13.8) mid-HDF PAES 2.1 (0.2) 378.5 (27.4) 599.5 (6.5) 167.5 (9.7) 223.0 (21.4) Susantitaphong [70] 2009 12 36 66.6% 59.5 (13.5) 81.6 (52.8) 57.5 (11.6) Pre-HDF/mid-HDF/post-HDF PAES 2.2 (0.1) 440.3 (19.9) 554.2 (10.4) 245.9 (2.1) 240.0 (0.0) Troidle [71] 2009 8 8 0.0% 45.0 (7.0) (—) (—) HFD PS 1.8 (0.0) 400.0 (0.0) 600.0 (0.0) (—) 480.0 (0.0) Wang [72] 2009 18 54 27.8% 46.9 (9.6) 52.5 (—) (—) HFD PS 1.5 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Bhimani [73] 2010 12 144 75.0% 60.0 (4.0) 81.0 (19.0) (—) HFD PS/PAES 1.7 (0.2) 400.0 (1.8) 550.0 (0.0) (—) (—) Gasco [74] 2010 16 263 50.0% 52.7 (—) (—) (—) post-HDF PS 1.8 (0.0) 323.0 (0.0) 800.0 (0.0) (—) 242.0 (0.0) Kohn [75] 2010 5 38 60.0% (—) 142.0 (60.9) 86.0 (29.0) HFD PAES (—) 425.0 (75.0) 200.0 (0.0) (—) 174.0 (15.0) Krieter [76] 2010 8 64 12.5% 63.0 (12.0) 81.8 (144.0) 70.8 (17.5) HFD/post-HDF PAES 1.9 (0.0) 378.0 (31.1) 500.0 (0.0) 47.0 (6.0) 229.0 (20.7) Park [77] 2010 52 52 52.0% 54.0 (12.4) 112.7 (188.6) (—) HFD PS 1.3 (0.0) 237.0 (23.0) 500.0 (0.0) (—) 240.0 (12.0) Basile [78] 2011 11 22 18.2% 54.1 (17.8) 78.0 (60.2) 69.1 (9.9) HFD PS 1.8 (0.0) 270.0 (0.0) 270.0 (0.0) (—) 469.1 (2.7) Ficheux [79] 2011 18 54 0.0% 79.7 (1.7) (—) 66.1 (2.3) HFD PS 2.2 (0.1) 318.0 (2.0) 500.0 (9.8) (—) 222.0 (2.9) Pedrini [80] 2011 15 60 33.3% 67.2 (8.3) (—) 73.1 (14.0) post-HDF/mid-HDF PS/PAES 2.3 (0.1) 374.0 (34.0) 580.0 (39.7) 147.5 (11.1) 224.0 (18.5) Panichi [81, 82] 2012 30 180 33.3% 55.9 (14.0) 58.0 (59.0) (—) post-HDF PAES 2.1 (0.0) 313.5 (32.7) 600.0 (0.0) 78.1 (0.0) 235.0 (13.8) Susantitaphong [83] 2012 12 48 66.6% 57.8 (14.8) 43.2 (42.0) 55.5 (11.1) mid-HDF/mixed-HDF PAES 2.2 (0.0) 425.0 (24.5) 600.0 (0.0) 200.0 (0.0) 240.0 (0.0) Tessitore [84] 2012 26 26 53.9% 63.0 (12.0) (—) (—) HFD PP 0.7 (0.0) 297.0 (32.0) 500.0 (0.0) (—) 230.0 (13.0) von Albertini [85] 2013 12 35 0.0% (—) (—) (—) HFD/post-HDF PAES/PS 1.8 (0.0) 417.1 (0.0) 667.4 (0.0) 30.7 (0.0) 206.3 (22.5) Heaf [86] 2013 12 96 30.0% 63.1 (11.7) 78.0 (52.8) 79.2 (17.8) HFD PAES 2.0 (0.0) 276.0 (38.7) 500.0 (0.0) (—) 240.0 (0.0) Melo [87] 2014 14 28 50.0% 48.9 (14.4) (—) 76.35 (19.63) HFD/post-HDF PS 2.0 (0.0) 375.0 (8.2) 760.0 (0.0) 40.0 (0.0) 115.7 (16.8) Pedrini [88] 2014 16 32 18.8% (—) (—) 77.60 (10.78) post-HDF PAES/PS 2.20 (0.10) 388.0 (25.9) 574.5 (39.0) 121.0 (11.9) 226.5 (13.7) Cornelis [89] 2014 13 52 23.1% 53.6 (20.4) 49.0 (29.0) (—) HFD/post-HDF PS 1.8 (0.0) 286.0 (4.8) 573.7 (13.1) 30.1 (1.4) 366.3 (3.9) Potier [90] 2016 6 24 66.7% 65.4 (25.5) 68.6 (43.7) 73.9 (2.1) HFD/post-HDF/ mixed-HDF/pre-HDF PS 2.3 (0.0) 339.4 (3.4) 600.0 (0.0) 122.1 (5.1) 240.0 (0.0) Gayrard [91] 2017 12 48 50.0% 73.0 (12.0) (—) 71.0 (1.9) HFD/post-HDF PS 1.8 (0.0) 366.3 (5.1) 602.3 (1.0) 51.8 (1.1) 233.6 (2.9) Kirsch [92] 2017 39 59 28.2% 60.5 (13.6) 63.1 (43.8) 80.2 (18.4) post-HDF/HFD PS 1.9 (0.1) 368.1 (12.8) (—) 27.5 (1.4) 252.4 (11.8) First Author Year N N meas Female Age Vintage PreWt Modality Material MSA QB QD Qinf Duration Leto [26] 2001 15 30 40.0% 45.7 (—) 156.3 (—) (—) HFD CA/PS 1.3 (0.1) 250.0 (0.0) 600.0 (0.0) (—) 240.0 (0.0) Xu [27] 2001 10 10 40.0% 70.2 (5.6) 71.2 (37.0) 63.9 (10.6) HFD PS 1.8 (0.5) (—) (—) (—) 300.0 (0.0) Yamada [28] 2001 28 28 39.0% 58.1 (16.4) 64.0 (47.0) 49.0 (8.0) HFD PS 1.42 (0.0) 188.0 (18.0) 500.0 (0.0) (—) 237.0 (18.0) Stiller [29] 2002 15 15 73.0% 54.3 (10.2) 134.0 (100.6) (—) HFD PAES/PS 1.24 (0.1) (—) (—) (—) 240.0 (0.0) Eknoyan [20] 2002 984 3967 59.0% 58.6 (13.7) 63.1 (59.2) 71.8 (1.5) HFD PMMA/CA/PS/ PAN/PAES 1.8 (0.2) 372.4 (8.8) 671.8 (10.1) (—) 204.5 (2.7) Ding [30] 2002 12 36 33.0% 49.7 (11.3) 83.5 (76.7) (—) pre-HDF/post-HDF PS 1.3 (0.0) 250.0 (0.0) 616.7 (2.9) 92.5 (3.0) 282.5 (29.6) Klingel [31] 2002 22 22 0.0% 61.4 (—) (—) 74.6 (0.0) HFD PS 1.3 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 228.8 (0.0) Mann [32] 2003 5 5 0.0% (—) (—) (—) HFD PS 1.6 (0.05) (—) (—) (—) 240.0 (0.0) Mandolfo [33] 2003 8 16 0.0% 61.4 (—) (—) 68 (8.6) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 550.0 (0.0) 30.0 (0.0) 240.0 (0.0) Pedrini [34] 2003 20 20 35.0% 63.0 (17.0) 116.4 (86.4) 60.3 (12.6) post-HDF/mixed-HDF PS 2.1 (0.0) 403.0 (55.9) 580.2 (36.6) 219.6 (36.6) 227.0 (17.7) Ward [35] 2003 12 24 41.6% 53.0 (13.0) 63.0 (18.3) (—) HFD PS 1.8 (0.0) 410.0 (1.9) 700.0 (0.0) (—) 228.0 (11.7) Bammens [36] 2004 14 70 28.6% 66.6 (3.1) 24.8 (10.0) 62.10 (1.94) HFD/pre-HDF/post-HDF PS 1.8 (0.0) 323.9 (116.3) 500.0 (0.0) 87.0 (0.0) 230.0 (0.0) Yamashita [37] 2004 5 5 80.0% (—) (—) (—) post-HF PS 1.8 (0.0) (—) (—) 84.2 (18.8) 120.0 (0.0) Emiliani [38] 2004 10 10 20.0% 66.0 (18.0) 80.0 (36.0) 66.2 (7.5) mid-HDF PAES 2.6 (0.0) 312.0 (18.0) 500.0 (0.0.0) 43.6 (7.2) 240.0 (10.0) Leypoldt [39] 2004 22 88 37.5% 61.0 (18.0) (—) 80.3 (19.4) HFD PS 1.77 (0.0) 338.0 (49.6) 540.0 (60.0) (—) 178.5 (19.0) Lucchi [40] 2004 10 20 40.0% 61.1 (8.9) 51.8 (35.9) (—) HFD/post-HDF PS 1.6 (0.0) 300.0 (0.0) 625.0 (0.0) 20.9 (0.0) 240.0 (0.0) Pisitkun [41] 2004 9 18 22% 48.0 (6.1) 51.4 (42.0) 55.2 (8.3) HFD/mid-HDF PS 2.7 (0.9) 475.0 (36.4) 800.0 (0.0) 59.9 (9.7) 240.0 (0.0) Tonelli [42] 2004 5 15 0.0% (—) (—) (—) HFD PS 1.8 (0.0) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Santoro [43] 2005 20 40 0.0% (—) (—) (—) HFD/mid-HDF PAES 1.90 (0.0) 363.0 (39.2) (—) 58.0 (11.9) (—) Brendolan [44] 2005 3 24 0.0% (—) (—) (—) HFD/post-HDF PS 2.2 (0.4) 333.3 (12.0) 500.0 (0.0) 16.6 (1.9) 226.1 (14.5) Padrini [45] 2005 11 22 36.4% 66.8 (11.9) 80.9 (66.9) 64.1 (9.2) post-HF/pre-HF PAES 2.1 (0.0) 327.8 (22.3) (—) 186 (40.1) 240 (6) Petras [46] 2005 6 36 0.0% 56.0 (16.0) 86.0 (50.0) (—) HFD/post-HDF/Pre-HF PAES 2.1 (0.0) 350.0 (0.0) 500.0 (0.0) 95.0 (0.0) 240.0 (0.0) Krieter [47] 2005a 5 5 60.0% 52.0 (22) (—) 68.5 (27.5) mid-HDF PAES 1.9 (0.0) 400.0 (0.0) 800.0 (0.0) 200.0 (0.0) 205.0 (15.0) Krieter [48] 2005b 10 40 30.0% 57.3 (13.7) 99.6 (92.4) 66.3 (10.4) mid-HDF/post-HDF PAES/PS 1.9 (0.1) 400.0 (0.0) 550.0 (0.0) 148.3 (2.9) 240 (23.4) Evenepoel [49] 2006 20 20 25.0% 68.8 (10.9) 19.3 (31.5) 59.9 (7.9) HFD PS 1.8 (0.0) 322.7 (21.6) 500.0 (0.0) (—) 230.0 (0.0) Mandolfo [50] 2006 12 18 66.7% 69.0 (9.0) 117.6 (69.6) 65.2 (8.1) HFD/post-HDF PS 1.8 (0.0) 350.0 (0.0) 700.0 (0.0) 50.0 (0.0) 240.0 (0.0) Nakashima [51] 2006 12 24 0.0% 49.1 (12.1) 127.2 (73.2) 66.8 (12.7) HFD PS 2.10 (0.0) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Panich [52] 2006 10 20 50.0% 58.2 (14.7) (—) 54.2 (3.8) HFD/post-HDF PS 1.8 (0.0) 425.3 (12.6) (—) 61.6 (0.6) 240.0 (0.0) Pedrini [53] 2006 12 72 25.0% 64.2 (6.6) 45.0 (38.0) 64.9 (11.2) mixed-HDF PS 2.1 (0.0) 422.0 (37.9) 609.0 (27.9) 178.0 (20.9) 218.0 (25.9) Potier [54] 2007 6 18 0.0% (—) (—) (—) post-HDF/pre-HDF/mid-HDF PAES 1.90 (0.0) 360.0 (0.0) 500.0 (0.0) 175.0 (0.0) (—) Feliciani [55] 2007 10 30 20.0% 64.7 (8.0) 54.7 (57.7) 73.25 (12.5) mixed-HDF/mid-HDF PAES/PS 1.85 (0.1) 385.5 (18.3) 609.0 (20.7) 167.5 (14.1) 231.5 (16.8) Krieter [56] 2007 8 32 62.5% 62.1 (13.8) 76.0 (55.3) 68.5 (7.1) HFD PS/PAES 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 236.0 (10.5) Santoro [57] 2007 8 16 50.0% 56.6 (23.6) (—) (—) mid-HDF PAES 1.9 (0.0) 306.5 (10.3) (—) 100.0 (0.0) 231.0 (10.3) Tiranathanagul [58] 2007 12 48 33.0% 54.2 (13.6) 42.0 (32.3) 62.85 (9.4) post-HDF/mid-HDF PS 2.7 (0.9) 416.7 (24.1) 800.0 (0.0) 113.0 (6.0) 240.0 (0.0) Abe [59] 2008 15 45 40% 65.5 (13.2) 72.9 (63.8) (—) HFD PMMA/CA/PS 1.5 (0.1) 200.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Eloot [60] 2008 9 27 55.5% 71.0 (10.0) 19.0 (12.0) 79.0 (11.5) HFD PS 1.8 (0.0) 260.0 (0.0) 260.0 (0.0) (—) 360.0 (0.0) Mandolfo [61] 2008 8 16 37.5% 72.2 (4.8) 62.0 (24.0) 61.7 (11) HFD/mid-HDF PAES 1.9 (0.0) 251.5 (32.4) 700.0 (0.0) 56.0 (4.8) 240.0 (0.0) Spalding [62] 2008 12 12 50.0% 65.3 (12.9) (—) (—) HFD/post-HDF (—) (—) 358.4 (84.4) 800.0 (0.0) 37.5 (12.6) 197.4 (55.3) Krieter [63] 2008a 8 40 25.0% 64.0 (16.0) 70.0 (74.0) 74.2 (10.7) HFD/post-HDF PAES/PS 1.7 (0.1) 300.0 (0.0) 460.0 (0.0) 40.0 (0.0) 240.0 (0.0) Ouseph [64] 2008a 12 48 25.0% 57.0 (4.0) 52.0 (17.0) 81.3 (4.35) HFD PS/PAES 1.65 (0.1) 382.0 (4.8) 800.0 (0.0) (—) 219.0 (5.3) Ouseph [64] 2008b 12 60 41.6% 46.0 (3.0) 48.0 (8.0) 84.2 (6.75) HFD PS/PAES 1.90 (0.2) 404.0 (1.0) 800.0 (0.0) (—) 240.0 (0.0) Krieter [65] 2008b 8 48 37.5% 63.0 (14.0) 77.5 (38.9) 76.5 (11.15) HFD PAES/PS 1.7 (0.1) 300.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Joyeux [66] 2008 20 40 35.0% 58.0 (20.9) 46.0 (46.1) 71.5 (10.2) HFD/post-HDF PAES 2.10 (0.0) 310.5 (33.3) (—) 31.8 (5.2) 235.5 (14.6) Lee [67] 2009 8 16 50.0% 68.7 (19.1) 50.3 (53.1) (—) HFD PAES 1.1 (0.0) 325.0 (24.6) 500.0 (0.0) (—) 255.0 (14.8) Meert [68] 2009 14 42 50.0% 63.5 (17.0) 30.2 (36.0) (—) pre-HDF/pre-HF/post-HDF PAES 1.8 (0.2) 312.3 (15.6) 384.7 (5.0) 185.7 (20.7) 249.3 (13.0) Pedrini [69] 2009 15 90 20.0% 67.3 (8.7) 44.1 (20.8) 76.9 (13.8) mid-HDF PAES 2.1 (0.2) 378.5 (27.4) 599.5 (6.5) 167.5 (9.7) 223.0 (21.4) Susantitaphong [70] 2009 12 36 66.6% 59.5 (13.5) 81.6 (52.8) 57.5 (11.6) Pre-HDF/mid-HDF/post-HDF PAES 2.2 (0.1) 440.3 (19.9) 554.2 (10.4) 245.9 (2.1) 240.0 (0.0) Troidle [71] 2009 8 8 0.0% 45.0 (7.0) (—) (—) HFD PS 1.8 (0.0) 400.0 (0.0) 600.0 (0.0) (—) 480.0 (0.0) Wang [72] 2009 18 54 27.8% 46.9 (9.6) 52.5 (—) (—) HFD PS 1.5 (0.0) 250.0 (0.0) 500.0 (0.0) (—) 240.0 (0.0) Bhimani [73] 2010 12 144 75.0% 60.0 (4.0) 81.0 (19.0) (—) HFD PS/PAES 1.7 (0.2) 400.0 (1.8) 550.0 (0.0) (—) (—) Gasco [74] 2010 16 263 50.0% 52.7 (—) (—) (—) post-HDF PS 1.8 (0.0) 323.0 (0.0) 800.0 (0.0) (—) 242.0 (0.0) Kohn [75] 2010 5 38 60.0% (—) 142.0 (60.9) 86.0 (29.0) HFD PAES (—) 425.0 (75.0) 200.0 (0.0) (—) 174.0 (15.0) Krieter [76] 2010 8 64 12.5% 63.0 (12.0) 81.8 (144.0) 70.8 (17.5) HFD/post-HDF PAES 1.9 (0.0) 378.0 (31.1) 500.0 (0.0) 47.0 (6.0) 229.0 (20.7) Park [77] 2010 52 52 52.0% 54.0 (12.4) 112.7 (188.6) (—) HFD PS 1.3 (0.0) 237.0 (23.0) 500.0 (0.0) (—) 240.0 (12.0) Basile [78] 2011 11 22 18.2% 54.1 (17.8) 78.0 (60.2) 69.1 (9.9) HFD PS 1.8 (0.0) 270.0 (0.0) 270.0 (0.0) (—) 469.1 (2.7) Ficheux [79] 2011 18 54 0.0% 79.7 (1.7) (—) 66.1 (2.3) HFD PS 2.2 (0.1) 318.0 (2.0) 500.0 (9.8) (—) 222.0 (2.9) Pedrini [80] 2011 15 60 33.3% 67.2 (8.3) (—) 73.1 (14.0) post-HDF/mid-HDF PS/PAES 2.3 (0.1) 374.0 (34.0) 580.0 (39.7) 147.5 (11.1) 224.0 (18.5) Panichi [81, 82] 2012 30 180 33.3% 55.9 (14.0) 58.0 (59.0) (—) post-HDF PAES 2.1 (0.0) 313.5 (32.7) 600.0 (0.0) 78.1 (0.0) 235.0 (13.8) Susantitaphong [83] 2012 12 48 66.6% 57.8 (14.8) 43.2 (42.0) 55.5 (11.1) mid-HDF/mixed-HDF PAES 2.2 (0.0) 425.0 (24.5) 600.0 (0.0) 200.0 (0.0) 240.0 (0.0) Tessitore [84] 2012 26 26 53.9% 63.0 (12.0) (—) (—) HFD PP 0.7 (0.0) 297.0 (32.0) 500.0 (0.0) (—) 230.0 (13.0) von Albertini [85] 2013 12 35 0.0% (—) (—) (—) HFD/post-HDF PAES/PS 1.8 (0.0) 417.1 (0.0) 667.4 (0.0) 30.7 (0.0) 206.3 (22.5) Heaf [86] 2013 12 96 30.0% 63.1 (11.7) 78.0 (52.8) 79.2 (17.8) HFD PAES 2.0 (0.0) 276.0 (38.7) 500.0 (0.0) (—) 240.0 (0.0) Melo [87] 2014 14 28 50.0% 48.9 (14.4) (—) 76.35 (19.63) HFD/post-HDF PS 2.0 (0.0) 375.0 (8.2) 760.0 (0.0) 40.0 (0.0) 115.7 (16.8) Pedrini [88] 2014 16 32 18.8% (—) (—) 77.60 (10.78) post-HDF PAES/PS 2.20 (0.10) 388.0 (25.9) 574.5 (39.0) 121.0 (11.9) 226.5 (13.7) Cornelis [89] 2014 13 52 23.1% 53.6 (20.4) 49.0 (29.0) (—) HFD/post-HDF PS 1.8 (0.0) 286.0 (4.8) 573.7 (13.1) 30.1 (1.4) 366.3 (3.9) Potier [90] 2016 6 24 66.7% 65.4 (25.5) 68.6 (43.7) 73.9 (2.1) HFD/post-HDF/ mixed-HDF/pre-HDF PS 2.3 (0.0) 339.4 (3.4) 600.0 (0.0) 122.1 (5.1) 240.0 (0.0) Gayrard [91] 2017 12 48 50.0% 73.0 (12.0) (—) 71.0 (1.9) HFD/post-HDF PS 1.8 (0.0) 366.3 (5.1) 602.3 (1.0) 51.8 (1.1) 233.6 (2.9) Kirsch [92] 2017 39 59 28.2% 60.5 (13.6) 63.1 (43.8) 80.2 (18.4) post-HDF/HFD PS 1.9 (0.1) 368.1 (12.8) (—) 27.5 (1.4) 252.4 (11.8) N, number; N meas, number of measurements; Vintage, time on chronic intermittent dialysis in months; PreWt, pre-dialysis weight in kilograms; MSA, membrane surface area (in square meters); QB, blood flow rate (mL/min); QD, dialysis fluid flow rate (mL/min); Duration, the dialysis session (in min). For each parameter the table summarizes the mean and the SD over all arms in each study or a (—) if the relevant parameter could not be extracted from the paper. Study quality Quality of the included studies varied widely based on each of the five components of the EPHPP (Table S1). For our meta-analysis, the global rating was characterized to be of moderate quality for most of the included studies [93], strong for 20 and weak for 16 studies. Both reviewers discussed the ratings and there was no discrepancy between them with respect to the components’ ratings and the final global scoring and rating. This high inter-rate agreement was in line with a previous evaluation of the EPHPP [94]. Main determinants of moderate quality were selection bias, study design and blinding procedures (methodologic heterogeneity), whereas data collection, study confounders and withdrawals/dropouts provided strong quality to the included studies. β2M dialyzer clearance in diffusive, high flux dialysis This meta-analysis also included 49 studies on HF dialysis, which evaluated 147 configurations of dialyzers and operational characteristics of treatment (e.g. blood or dialysate blood flows). Average (over the course of treatment) β2M clearance was 48.75 mL/min (95% CI 42.50–55.21) with substantial heterogeneity among studies [P (Q) ≤ 0.001] (Figure 2). Instantaneous β2M clearance was 52.09  mL/min (95% CI 41.39–62.78) with substantial heterogeneity among studies [P (Q) < 0.001] (Figure S1). There were no differences between instantaneous and average (over the course of the treatment) β2M clearances in univariate meta-regressions (difference of 1.88 mL/min, 95% CI −6.58 to 10.34, P =  0.66). Therefore, we combined instantaneous and average β2M clearances together for meta-regression analyses. First, we explored the sources of heterogeneity through ‘univariate’ meta-regressions examining only one study characteristic. Kuf (and Kuf scaled to MSA), clearance calculation formula, MSA, indexing clearance to the plasma (rather than blood) volume compartment, blood pump flow rate and dialysis membrane material were statistically significant predictors of variation in β2M clearance by diffusive, HF dialysis in these analyses (Table S2). Interestingly, there was no evidence of a secular trend of improving dialytic clearance over the last 17 years. Subsequently, we carried out ‘multivariable’ meta-regression to simultaneously adjust for multiple study characteristics. In these analyses shown in Table 2, we forced the type of measurement (instantaneous versus average) and the secular trend into the models. We found a significantly higher β2M clearance for PAES dialyzers (higher by 12.25 mL/min, 95% CI 5.472–19.028, P < 0.0001) relative to PS dialyzers. A significantly higher β2M clearance was found for higher blood flow rates in HF dialysis, i.e. an increase of 0.091 mL/min per 1 mL/min blood flow rate, 95% CI 0.024–0.159, P = 0.007). Adjusted dialysate side clearances were significantly lower than blood clearances (by 22.279 mL/min, 95% CI 9.8–34.757, P < 0.001). Other significant predictors were Kuf of the dialyzer (scaled to the MSA), while the MSA was of borderline significance (P = 0.057). In these multivariable analyses, there was no evidence for improving dialyzer performance over calendar time (P = 0.854). Similarly, there was no statistically significant difference in sensitivity analysis that compared the HEMO measurements against all the other measurements, or when we ran the multivariate regression, excluding the HEMO study (data not shown). Table 2 Metaregression of β2M clearance for high flux dialysis Variable Effect size (mL/min) CI P (Wald) Blood pump flow (per mL/min) 0.091 (0.024 to 0.159) 0.007 Kuf (scaled to MSA) 0.803 (0.373 to 1.232) <0.001 MSA (per m2) 10.923 (−0.327 to 22.173) 0.057 Dialysis membrane (relative to PS) PAES 12.25 (5.472 to 19.028) <0.001 CA 5.025 (−7.01 to 17.061) 0.413 PAN 3.571 (−10.378 to 17.519) 0.616 PMMA 9.15 (−2.501 to 20.8) 0.124 Compartment  Blood (versus plasma) 8.876 (−3.999 to 21.75) 0.177 Clearance side  Dialysate (versus blood) −22.279 (−34.757 to −9.8) <0.001 Type of measurement  Instantaneous (versus average) 6.589 (−3.422 to 16.6) 0.197 Secular trenda 0.178 (−1.716 to 2.072) 0.854 Variable Effect size (mL/min) CI P (Wald) Blood pump flow (per mL/min) 0.091 (0.024 to 0.159) 0.007 Kuf (scaled to MSA) 0.803 (0.373 to 1.232) <0.001 MSA (per m2) 10.923 (−0.327 to 22.173) 0.057 Dialysis membrane (relative to PS) PAES 12.25 (5.472 to 19.028) <0.001 CA 5.025 (−7.01 to 17.061) 0.413 PAN 3.571 (−10.378 to 17.519) 0.616 PMMA 9.15 (−2.501 to 20.8) 0.124 Compartment  Blood (versus plasma) 8.876 (−3.999 to 21.75) 0.177 Clearance side  Dialysate (versus blood) −22.279 (−34.757 to −9.8) <0.001 Type of measurement  Instantaneous (versus average) 6.589 (−3.422 to 16.6) 0.197 Secular trenda 0.178 (−1.716 to 2.072) 0.854 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). Inclusion of these variables decreased the apparent degree of heterogeneity by more than half (Q statistic of unadjusted model 4694.5603 versus 1947.1515 for the fully adjusted model), but significant heterogeneity did remain (P-value of QE statistic <0.001). Results based on 123 distinct configurations of dialyzer and dialysis procedure operational parameters. See text for other abbreviations. Table 2 Metaregression of β2M clearance for high flux dialysis Variable Effect size (mL/min) CI P (Wald) Blood pump flow (per mL/min) 0.091 (0.024 to 0.159) 0.007 Kuf (scaled to MSA) 0.803 (0.373 to 1.232) <0.001 MSA (per m2) 10.923 (−0.327 to 22.173) 0.057 Dialysis membrane (relative to PS) PAES 12.25 (5.472 to 19.028) <0.001 CA 5.025 (−7.01 to 17.061) 0.413 PAN 3.571 (−10.378 to 17.519) 0.616 PMMA 9.15 (−2.501 to 20.8) 0.124 Compartment  Blood (versus plasma) 8.876 (−3.999 to 21.75) 0.177 Clearance side  Dialysate (versus blood) −22.279 (−34.757 to −9.8) <0.001 Type of measurement  Instantaneous (versus average) 6.589 (−3.422 to 16.6) 0.197 Secular trenda 0.178 (−1.716 to 2.072) 0.854 Variable Effect size (mL/min) CI P (Wald) Blood pump flow (per mL/min) 0.091 (0.024 to 0.159) 0.007 Kuf (scaled to MSA) 0.803 (0.373 to 1.232) <0.001 MSA (per m2) 10.923 (−0.327 to 22.173) 0.057 Dialysis membrane (relative to PS) PAES 12.25 (5.472 to 19.028) <0.001 CA 5.025 (−7.01 to 17.061) 0.413 PAN 3.571 (−10.378 to 17.519) 0.616 PMMA 9.15 (−2.501 to 20.8) 0.124 Compartment  Blood (versus plasma) 8.876 (−3.999 to 21.75) 0.177 Clearance side  Dialysate (versus blood) −22.279 (−34.757 to −9.8) <0.001 Type of measurement  Instantaneous (versus average) 6.589 (−3.422 to 16.6) 0.197 Secular trenda 0.178 (−1.716 to 2.072) 0.854 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). Inclusion of these variables decreased the apparent degree of heterogeneity by more than half (Q statistic of unadjusted model 4694.5603 versus 1947.1515 for the fully adjusted model), but significant heterogeneity did remain (P-value of QE statistic <0.001). Results based on 123 distinct configurations of dialyzer and dialysis procedure operational parameters. See text for other abbreviations. FIGURE 2 View largeDownload slide Forest plot of average (over the course of the treatment) β2M dialyzer clearance in HFD. Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). FIGURE 2 View largeDownload slide Forest plot of average (over the course of the treatment) β2M dialyzer clearance in HFD. Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). β2M clearance in convective dialysis therapies This meta-analysis included 63 papers on HDF and 5 hemofiltration studies that examined 132 unique configurations of dialyzers, infusion volumes and patient cohorts. Average β2M clearance (over the course of treatment) was 8706 mL/min (95% CI 75.08–99.03) with substantial heterogeneity among studies [P (Q) ≤ 0.001] (Figure 3). Instantaneous β2M clearance was 125.26 mL/min (95% CI 103.92–146.59) with substantial heterogeneity among studies [P (Q) ≤ 0.001] (Figure S2). Kuf, blood pump flow rate, blood (versus plasma) compartment clearance, the side of the clearance (blood versus dialysate) were significant predictors in ‘univariate’ meta-regressions (Table S3). MSA, membrane material and substitution fluid infusion rates were not significant predictors in these univariate analyses. In ‘multivariable’ meta-regression analyses (Table 3) we found a significantly higher β2M clearance from the body when this calculation was indexed to whole blood versus plasma, while dialysate side body clearance was substantially lower than plasma by −41.523 mL/min (95% CI −54.525 to −28.52, P < 0.0001). Higher blood flow (0.188 mL/min per 1 mL/min blood flow, 95% CI 0.046–0.330, P = 0.01), membrane material (PS higher than PAES or PMMA) and certain forms of modality (e.g. pre-dilution HDF versus pre-dilution hemofiltration) but not substitution fluid infusion rates were significantly associated with higher β2M clearances. ANOVA tests suggested that both membrane material (P = 0.0033) and any convective modality (P = 0.0013) were significant predictors of dialytic body clearance of β2M. Table 3 Metaregression of β2M clearance for convective therapies (HF/HDF) Variable Effect size (per mL/min) CI P (Wald) Kuf (scaled to MSA) 1.691 (0.609 to 2.773) 0.002 MSA (per m2) −1.336 (−19.017 to 16.346) 0.882 Compartment  Blood (versus plasma) 49.868 (34.794 to 64.942) <0.001 Clearance side  Dialysate (versus blood) −41.523 (−54.525 to −28.52) <0.001 Blood pump flow (per mL/min) 0.188 (0.046 to 0.33) 0.01 Dialysis membrane  PAES −23.524 (−40.635 to −6.412) 0.007  PMMA −22.421 (−41.627 to −3.215) 0.022 Type of measurement  Instantaneous (versus average) 4.719 (−7.401 to 16.84) 0.445 Substitution fluid rate (per mL/min) 0.046 (−0.045 to 0.137) 0.321 Modality (relative to pre-hemofiltration)  post-hemofiltration 42.719 (−1.957 to 87.395) 0.061  post-HDF −7.764 (−27.834 to 12.306) 0.448  mid-HDF 5.614 (−16.493 to 27.721) 0.619  mixed-HDF −12.972 (−36.947 to 11.002) 0.289  pre-HDF −25.464 (−45.137 to −5.792) 0.011 Secular trenda −0.925 (−3.31 to 1.46) 0.447 Variable Effect size (per mL/min) CI P (Wald) Kuf (scaled to MSA) 1.691 (0.609 to 2.773) 0.002 MSA (per m2) −1.336 (−19.017 to 16.346) 0.882 Compartment  Blood (versus plasma) 49.868 (34.794 to 64.942) <0.001 Clearance side  Dialysate (versus blood) −41.523 (−54.525 to −28.52) <0.001 Blood pump flow (per mL/min) 0.188 (0.046 to 0.33) 0.01 Dialysis membrane  PAES −23.524 (−40.635 to −6.412) 0.007  PMMA −22.421 (−41.627 to −3.215) 0.022 Type of measurement  Instantaneous (versus average) 4.719 (−7.401 to 16.84) 0.445 Substitution fluid rate (per mL/min) 0.046 (−0.045 to 0.137) 0.321 Modality (relative to pre-hemofiltration)  post-hemofiltration 42.719 (−1.957 to 87.395) 0.061  post-HDF −7.764 (−27.834 to 12.306) 0.448  mid-HDF 5.614 (−16.493 to 27.721) 0.619  mixed-HDF −12.972 (−36.947 to 11.002) 0.289  pre-HDF −25.464 (−45.137 to −5.792) 0.011 Secular trenda −0.925 (−3.31 to 1.46) 0.447 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). Inclusion of these variables decreased the apparent degree of heterogeneity by >70% (Q statistic of unadjusted model 2361.8089 versus 675.8222 for the fully adjusted model), but significant heterogeneity did remain (P-value of QE statistic <0.001). Results are based on 127 distinct configurations of dialyzer and dialysis procedure operational parameters. See text for the abbreviations. Table 3 Metaregression of β2M clearance for convective therapies (HF/HDF) Variable Effect size (per mL/min) CI P (Wald) Kuf (scaled to MSA) 1.691 (0.609 to 2.773) 0.002 MSA (per m2) −1.336 (−19.017 to 16.346) 0.882 Compartment  Blood (versus plasma) 49.868 (34.794 to 64.942) <0.001 Clearance side  Dialysate (versus blood) −41.523 (−54.525 to −28.52) <0.001 Blood pump flow (per mL/min) 0.188 (0.046 to 0.33) 0.01 Dialysis membrane  PAES −23.524 (−40.635 to −6.412) 0.007  PMMA −22.421 (−41.627 to −3.215) 0.022 Type of measurement  Instantaneous (versus average) 4.719 (−7.401 to 16.84) 0.445 Substitution fluid rate (per mL/min) 0.046 (−0.045 to 0.137) 0.321 Modality (relative to pre-hemofiltration)  post-hemofiltration 42.719 (−1.957 to 87.395) 0.061  post-HDF −7.764 (−27.834 to 12.306) 0.448  mid-HDF 5.614 (−16.493 to 27.721) 0.619  mixed-HDF −12.972 (−36.947 to 11.002) 0.289  pre-HDF −25.464 (−45.137 to −5.792) 0.011 Secular trenda −0.925 (−3.31 to 1.46) 0.447 Variable Effect size (per mL/min) CI P (Wald) Kuf (scaled to MSA) 1.691 (0.609 to 2.773) 0.002 MSA (per m2) −1.336 (−19.017 to 16.346) 0.882 Compartment  Blood (versus plasma) 49.868 (34.794 to 64.942) <0.001 Clearance side  Dialysate (versus blood) −41.523 (−54.525 to −28.52) <0.001 Blood pump flow (per mL/min) 0.188 (0.046 to 0.33) 0.01 Dialysis membrane  PAES −23.524 (−40.635 to −6.412) 0.007  PMMA −22.421 (−41.627 to −3.215) 0.022 Type of measurement  Instantaneous (versus average) 4.719 (−7.401 to 16.84) 0.445 Substitution fluid rate (per mL/min) 0.046 (−0.045 to 0.137) 0.321 Modality (relative to pre-hemofiltration)  post-hemofiltration 42.719 (−1.957 to 87.395) 0.061  post-HDF −7.764 (−27.834 to 12.306) 0.448  mid-HDF 5.614 (−16.493 to 27.721) 0.619  mixed-HDF −12.972 (−36.947 to 11.002) 0.289  pre-HDF −25.464 (−45.137 to −5.792) 0.011 Secular trenda −0.925 (−3.31 to 1.46) 0.447 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). Inclusion of these variables decreased the apparent degree of heterogeneity by >70% (Q statistic of unadjusted model 2361.8089 versus 675.8222 for the fully adjusted model), but significant heterogeneity did remain (P-value of QE statistic <0.001). Results are based on 127 distinct configurations of dialyzer and dialysis procedure operational parameters. See text for the abbreviations. FIGURE 3 View largeDownload slide Forest plot of average (over the course of treatment) β2M dialyzer clearance in convective dialysis (HF/HDF). Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). FIGURE 3 View largeDownload slide Forest plot of average (over the course of treatment) β2M dialyzer clearance in convective dialysis (HF/HDF). Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). In our dataset, there were 73 distinct configurations in post-HDF, which allowed us to better clarify the effects of different parameters upon dialytic clearance. Significant predictors of dialytic clearance in post-HDF were the substitution fluid infusion rate: increase by 0.297 mL/min for each mL/min increase in infusion rate (95% CI 0.200–0.394, P < 0.001) and Kuf: increase 1.346 mL/min for each mL/min/mmHg/m2 (95% CI 0.271–2.420, P = 0.014), while dialysis with a PAES dialyzer was associated with reduced clearance by −18.480 mL/min (95% CI −34.86 to −2.101, P = 0.027). Dialysis with a membrane with a higher surface area was associated with a numerically higher β2M clearance of 37.040 mL/min/m2 (95% CI −1.487 to 75.566); this association was of borderline statistical significance (P = 0.06). Interestingly, higher blood pump flow rates were not associated with enhanced dialytic clearance in post-HDF (0.042 mL/min for each mL/min increase in blood pump flow rates, 95% CI −0.045 to 0.128, P = 0.345), while other factors (side of clearance, blood versus plasma compartment calculations, instantaneous versus average clearance and secular trends) were numerically like the patterns noted in Table 3 (data not shown). β2M reduction ratios are higher in convective versus diffusive dialysis therapies For this meta-analysis, we identified a total of 140 configurations (with covariate information) that reported reduction ratios of β2M in either HFD (n = 81) and convective dialysis therapies (n = 59) for multivariable adjustments. In univariate analysis, convective dialysis (taken as a group) afforded greater β2M reduction ratios by 14.300% (95% CI 10.756–17.845%, P < 0.0001) relative to HF dialysis (estimate of 59.169%, 95% CI 55.484–62.854%). In multivariable meta-regression analyses (Table 4), higher membrane Kuf was a significant predictor of higher β2M reduction ratio in both diffusive and convective dialysis. In HFD, β2M reduction ratios were significantly higher for PAES (8.367%, 95% CI 2.913–13.822%, P = 0.003) compared with PS dialyzers. There was a strong secular trend in the reduction ratio afforded by HF dialysis, i.e. an increase of 1.443% per year since 2001 (95% CI 0.363–2.523, P =0.009). There were no differences by membrane material or type of modality in convective therapies, yet higher substitution flow rates were associated with higher β2M reduction ratios. Table 4 Metaregression of β2M reduction ratios for high flux dialysis and convective therapies HF dialysis Convective therapies Effect size (per mL/min) CI P (Wald) Effect size (per mL/min) CI P (Wald) Blood pump flow (per mL/min) −0.01 (−0.051 to 0.03) 0.619 −0.017 (−0.051 to 0.018) 0.344 Kuf (scaled to MSA) 0.388 (0.073 to 0.703) 0.016 0.326 (0.046 to 0.606) 0.023 MSA 7.44 (−1.987 to 16.867) 0.122 5.068 (−2.155 to 12.291) 0.169 Membrane material (ref: PS)  PAES 8.367 (2.913 to 13.822) 0.003 −0.836 (−4.792 to 3.121) 0.679  PMMA 12.403 (4.737 to 20.07) 0.002 −2.491 (−9.733 to 4.751) 0.5  CA 0.262 (−9.675 to 10.199) 0.959 — — —  PAN 3.525 (−6.677 to 13.727) 0.498 — — — Correction of post dialysis β2M value  Corrected for hemoconcentration −2.04 (−11.171 to 7.091) 0.661 −4.29 (−12.542 to 3.962) 0.308 Secular trend 1.443 (0.363 to 2.523) 0.009 0.438 (−0.198 to 1.075) 0.177 Substitution fluid rate (per mL/min) 0.077 (0.001 to 0.152) 0.047 Modality (relative to pre-hemofiltration)  post-hemofiltration — — — 15.931 (−6.334 to 38.195) 0.161  post-HDF — — — 19.583 (−0.727 to 39.893) 0.059  mid-HDF — — — 16.235 (−2.687 to 35.156) 0.093  mixed-HDF — — — 14.587 (−4.303 to 33.477) 0.13  pre-HDF — — — 6.891 (−9.21 to 22.992) 0.402 HF dialysis Convective therapies Effect size (per mL/min) CI P (Wald) Effect size (per mL/min) CI P (Wald) Blood pump flow (per mL/min) −0.01 (−0.051 to 0.03) 0.619 −0.017 (−0.051 to 0.018) 0.344 Kuf (scaled to MSA) 0.388 (0.073 to 0.703) 0.016 0.326 (0.046 to 0.606) 0.023 MSA 7.44 (−1.987 to 16.867) 0.122 5.068 (−2.155 to 12.291) 0.169 Membrane material (ref: PS)  PAES 8.367 (2.913 to 13.822) 0.003 −0.836 (−4.792 to 3.121) 0.679  PMMA 12.403 (4.737 to 20.07) 0.002 −2.491 (−9.733 to 4.751) 0.5  CA 0.262 (−9.675 to 10.199) 0.959 — — —  PAN 3.525 (−6.677 to 13.727) 0.498 — — — Correction of post dialysis β2M value  Corrected for hemoconcentration −2.04 (−11.171 to 7.091) 0.661 −4.29 (−12.542 to 3.962) 0.308 Secular trend 1.443 (0.363 to 2.523) 0.009 0.438 (−0.198 to 1.075) 0.177 Substitution fluid rate (per mL/min) 0.077 (0.001 to 0.152) 0.047 Modality (relative to pre-hemofiltration)  post-hemofiltration — — — 15.931 (−6.334 to 38.195) 0.161  post-HDF — — — 19.583 (−0.727 to 39.893) 0.059  mid-HDF — — — 16.235 (−2.687 to 35.156) 0.093  mixed-HDF — — — 14.587 (−4.303 to 33.477) 0.13  pre-HDF — — — 6.891 (−9.21 to 22.992) 0.402 Kuf, ultrafiltration coefficient of a dialyzer. Table 4 Metaregression of β2M reduction ratios for high flux dialysis and convective therapies HF dialysis Convective therapies Effect size (per mL/min) CI P (Wald) Effect size (per mL/min) CI P (Wald) Blood pump flow (per mL/min) −0.01 (−0.051 to 0.03) 0.619 −0.017 (−0.051 to 0.018) 0.344 Kuf (scaled to MSA) 0.388 (0.073 to 0.703) 0.016 0.326 (0.046 to 0.606) 0.023 MSA 7.44 (−1.987 to 16.867) 0.122 5.068 (−2.155 to 12.291) 0.169 Membrane material (ref: PS)  PAES 8.367 (2.913 to 13.822) 0.003 −0.836 (−4.792 to 3.121) 0.679  PMMA 12.403 (4.737 to 20.07) 0.002 −2.491 (−9.733 to 4.751) 0.5  CA 0.262 (−9.675 to 10.199) 0.959 — — —  PAN 3.525 (−6.677 to 13.727) 0.498 — — — Correction of post dialysis β2M value  Corrected for hemoconcentration −2.04 (−11.171 to 7.091) 0.661 −4.29 (−12.542 to 3.962) 0.308 Secular trend 1.443 (0.363 to 2.523) 0.009 0.438 (−0.198 to 1.075) 0.177 Substitution fluid rate (per mL/min) 0.077 (0.001 to 0.152) 0.047 Modality (relative to pre-hemofiltration)  post-hemofiltration — — — 15.931 (−6.334 to 38.195) 0.161  post-HDF — — — 19.583 (−0.727 to 39.893) 0.059  mid-HDF — — — 16.235 (−2.687 to 35.156) 0.093  mixed-HDF — — — 14.587 (−4.303 to 33.477) 0.13  pre-HDF — — — 6.891 (−9.21 to 22.992) 0.402 HF dialysis Convective therapies Effect size (per mL/min) CI P (Wald) Effect size (per mL/min) CI P (Wald) Blood pump flow (per mL/min) −0.01 (−0.051 to 0.03) 0.619 −0.017 (−0.051 to 0.018) 0.344 Kuf (scaled to MSA) 0.388 (0.073 to 0.703) 0.016 0.326 (0.046 to 0.606) 0.023 MSA 7.44 (−1.987 to 16.867) 0.122 5.068 (−2.155 to 12.291) 0.169 Membrane material (ref: PS)  PAES 8.367 (2.913 to 13.822) 0.003 −0.836 (−4.792 to 3.121) 0.679  PMMA 12.403 (4.737 to 20.07) 0.002 −2.491 (−9.733 to 4.751) 0.5  CA 0.262 (−9.675 to 10.199) 0.959 — — —  PAN 3.525 (−6.677 to 13.727) 0.498 — — — Correction of post dialysis β2M value  Corrected for hemoconcentration −2.04 (−11.171 to 7.091) 0.661 −4.29 (−12.542 to 3.962) 0.308 Secular trend 1.443 (0.363 to 2.523) 0.009 0.438 (−0.198 to 1.075) 0.177 Substitution fluid rate (per mL/min) 0.077 (0.001 to 0.152) 0.047 Modality (relative to pre-hemofiltration)  post-hemofiltration — — — 15.931 (−6.334 to 38.195) 0.161  post-HDF — — — 19.583 (−0.727 to 39.893) 0.059  mid-HDF — — — 16.235 (−2.687 to 35.156) 0.093  mixed-HDF — — — 14.587 (−4.303 to 33.477) 0.13  pre-HDF — — — 6.891 (−9.21 to 22.992) 0.402 Kuf, ultrafiltration coefficient of a dialyzer. β2M mass removal is higher in convective versus diffusive dialysis therapies For this meta-analysis, we identified 60 configurations reporting mass removal data (mg/session) of β2M. β2M mass removal (mg/session) was 151.66 mg/session (95% CI 126.98–176.34, P < 0.001) with substantial heterogeneity among studies [P (Q) < 0.001] (Figure 4). Kuf and type of modality were significant predictors of higher dialytic mass removal of β2M (data not shown) in univariate metaregression analyses. In multivariable meta-regressions (Table 5), Kuf and convective (relative to HF dialysis) were associated with higher removal of β2M into the dialysate (P < 0.001 in ANOVA). Removal of β2M was numerically higher with pure filtration therapies rather than HDF. However, when we restricted the analyses to convective techniques (n = 31), there was no statistically significant difference among the different techniques in terms of their ability to remove β2M from the body (P = 0.892). Furthermore, there was no evidence for heterogeneity in this analysis (residual heterogeneity, P = 0.08). More extensive analysis of the role of the substitution volume on β2M mass removal by post-HDF was limited by the small number of configurations (n = 12) that reported dialytic mass removal of β2M. Table 5 Metaregression of β2M removal for high flux dialysis and convective therapies Effect size (mg/session) CI P (Wald) Blood flow (per mL/min) −0.157 (−0.398 to 0.084) 0.202 Kuf scaled to MSA 2.229 (0.316 to 4.142) 0.022 MSA (per m2) −0.206 (−66.052 to 65.64) 0.995 Membrane material (relative to polysulfone)  PAES −1.874 (−34.069 to 30.321) 0.909  CA 22.983 (−61.608 to 107.573) 0.594 Modality (relative to HF dialysis)  mid-HDF 56.138 (−1.787 to 114.063) 0.057  mixed-HDF 97.522 (41.638 to 153.405) <0.001  post-HDF 54.714 (22.879 to 86.549) <0.001  post-hemofiltration 151.036 (−17.467 to 319.538) 0.079  pre-HDF 41.564 (1.7 to 81.427) 0.041  pre-hemofiltration 163.451 (−71.28 to 398.182) 0.172 Secular trenda 1.783 (−2.718 to 6.283) 0.438 Effect size (mg/session) CI P (Wald) Blood flow (per mL/min) −0.157 (−0.398 to 0.084) 0.202 Kuf scaled to MSA 2.229 (0.316 to 4.142) 0.022 MSA (per m2) −0.206 (−66.052 to 65.64) 0.995 Membrane material (relative to polysulfone)  PAES −1.874 (−34.069 to 30.321) 0.909  CA 22.983 (−61.608 to 107.573) 0.594 Modality (relative to HF dialysis)  mid-HDF 56.138 (−1.787 to 114.063) 0.057  mixed-HDF 97.522 (41.638 to 153.405) <0.001  post-HDF 54.714 (22.879 to 86.549) <0.001  post-hemofiltration 151.036 (−17.467 to 319.538) 0.079  pre-HDF 41.564 (1.7 to 81.427) 0.041  pre-hemofiltration 163.451 (−71.28 to 398.182) 0.172 Secular trenda 1.783 (−2.718 to 6.283) 0.438 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). See text for other abbreviations. Table 5 Metaregression of β2M removal for high flux dialysis and convective therapies Effect size (mg/session) CI P (Wald) Blood flow (per mL/min) −0.157 (−0.398 to 0.084) 0.202 Kuf scaled to MSA 2.229 (0.316 to 4.142) 0.022 MSA (per m2) −0.206 (−66.052 to 65.64) 0.995 Membrane material (relative to polysulfone)  PAES −1.874 (−34.069 to 30.321) 0.909  CA 22.983 (−61.608 to 107.573) 0.594 Modality (relative to HF dialysis)  mid-HDF 56.138 (−1.787 to 114.063) 0.057  mixed-HDF 97.522 (41.638 to 153.405) <0.001  post-HDF 54.714 (22.879 to 86.549) <0.001  post-hemofiltration 151.036 (−17.467 to 319.538) 0.079  pre-HDF 41.564 (1.7 to 81.427) 0.041  pre-hemofiltration 163.451 (−71.28 to 398.182) 0.172 Secular trenda 1.783 (−2.718 to 6.283) 0.438 Effect size (mg/session) CI P (Wald) Blood flow (per mL/min) −0.157 (−0.398 to 0.084) 0.202 Kuf scaled to MSA 2.229 (0.316 to 4.142) 0.022 MSA (per m2) −0.206 (−66.052 to 65.64) 0.995 Membrane material (relative to polysulfone)  PAES −1.874 (−34.069 to 30.321) 0.909  CA 22.983 (−61.608 to 107.573) 0.594 Modality (relative to HF dialysis)  mid-HDF 56.138 (−1.787 to 114.063) 0.057  mixed-HDF 97.522 (41.638 to 153.405) <0.001  post-HDF 54.714 (22.879 to 86.549) <0.001  post-hemofiltration 151.036 (−17.467 to 319.538) 0.079  pre-HDF 41.564 (1.7 to 81.427) 0.041  pre-hemofiltration 163.451 (−71.28 to 398.182) 0.172 Secular trenda 1.783 (−2.718 to 6.283) 0.438 Kuf, ultrafiltration coefficient of a dialyzer. a Secular trend includes a slope to adjust for a linear trend of increasing clearance in studies published in more recent years relative to HEMO (2001). See text for other abbreviations. FIGURE 4 View largeDownload slide Forest plot of β2M mass removal (mg/session) in diffusive HFD and convective (HDF/HF) dialysis. Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). FIGURE 4 View largeDownload slide Forest plot of β2M mass removal (mg/session) in diffusive HFD and convective (HDF/HF) dialysis. Comp, compartment; Kuf, ultrafiltration coefficient of the dialyzer; n, number; N meas, number of measurements; QB, blood flow rate (mL/min); QD, dialysate flow rate (mL/min). DISCUSSION This meta-analysis, combining 69 studies and including 1879 patients with 6771 clearance measurements, shows that membrane composition, modality (convective versus diffusive), blood flow rates and substitution fluid infusion rates independent of the dialysis modality are significant determinants of HF dialyzer performance in removing β2M. Our analysis is timely, as it provides quantitative information to aid the interpretation of a number of meta-analyses and secondary analyses of HD [11, 12, 14]. The significance of this work lies in our analysis of nearly 8-fold higher number of studies than previous reports by the Cochrane Group [16, 17] and others [18, 19]. Furthermore, our access to the primary study records of the HEMO trial allowed us to assess dialyzer performance using patient-level information from non-reused membranes thus overcoming a major limitation of a previous report [18]. One of the main and novel results of this study was that membrane material proved to be an important determinant of β2M clearance. Higher β2M clearances were noted with dialyzers made from PAES in respect to PS when applied in HF dialysis, the opposite of when applied in HDF. This is probably related to the chemical composition of the membranes as well as the 3D structure of the membranes and the different pressure profiles in these two modalities. The influence of membrane material on β2M clearance of HF dialysis was first reported 30 years ago [95]. Of relevance to our report, this early investigation showed that some dialysis membranes, such as cellulose acetate dialyzers, appear to induce β2M production during dialysis, whereas others, such as PS, do not. In the same study volume-controlled dialysis with HF membranes (PS 0.65 m2 and PS 1.25 m2) lowered β2M; clearance values, however, were significantly higher when these dialyzers were used in a HDF procedure. In another study [96] among patients receiving conventional HD using CA membranes, β2M levels increased 25.4% after HD, whereas in patients receiving HF HD using PS membrane, β2M levels decreased significantly (43.0%) after HD. Our results are also in accordance to a prospective, randomized, crossover study showing that the clearance of β2M was higher with PAES than PS [97]. Interestingly enough, β2M clearance during HDF was related to membrane material but in the inverse direction than in HF dialysis. We hypothesize that this is due to differential adsorption of β2M on membranes under the different transmembrane pressure (TMP) profiles of dialysis and HDF. Application of the higher TMP during HDF may result in a disproportionate increase in β2M adsorption in PS relative to PAES, so that the difference in clearance between the two membranes seen in HF is nearly reversed. An alternative explanation invokes a more efficient convection in membranes without adsorption versus those with more adsorption e.g. as a result of membrane clogging. Regardless of the explanation, this observation should be corroborated in future prospective, head to head comparisons given the substantial heterogeneity of methodologies for the measurement of β2M clearance employed by the different studies. Notwithstanding the effects of membrane material on β2M reduction ratio, it should be noted that recovery of β2M into the dialysate, was not affected by membrane material. This is consistent with a landmark prospective RCT [97], showing that the higher β2M clearance of PAES did not translate into more efficient mass removal of β2M. In that study, it was postulated that the higher mass removal of β2M by PAES arises from transmembrane transport augmented by adsorption within the membrane matrix. Membrane adsorption was experimentally demonstrated >20 years ago [98, 99] and the propensity of different membranes to differentially adsorb low molecular weight proteins was recently characterized with proteomic techniques [100]. Our analysis recapitulates previous findings that despite the higher clearance, β2M removal in the dialysate is not higher with any of the currently available membranes. This suggests that adsorption to the membrane, rather than convective or diffusive elimination of β2M in the dialysate, underlines the differences between dialyzers of different membrane material. The a priori plausibility of differential adsorption of β2M in membranes according to the dialysis mode is high. There are reports using proteomic techniques that demonstrate differential absorption of β2M in PS versus triacetate membranes [93], PS versus PMMA membranes [101] or even the same PS when exposed to the different pressure profiles associated with HF versus low flux dialysis [102]. An interesting report also showed a change of contribution of the different forms of clearance when the same dialyzer used in post- versus pre-HDF mode (adsorption is lower in post) [45]. Hence, the available data do point to differential adsorption patterns by material, permeability and even mode of HDF. The only credible way to test our hypothesis that PAES and PS adsorb β2M differently under HF dialysis and HDF is by properly designed head to head comparisons using standardized collection methods, blood and dialysate clearances and possibly proteomic techniques. An interesting direction for future innovations in dialyzer development that builds on this hypothesis would explore the properties of different membranes to optimize clearance for convective versus diffusive forms of dialysis. There have been reports in the literature about dialyzers (some of them already in the market) that are specifically targeted for convective therapies [103, 104], while safety considerations about albumin loss suggest that not all HF dialyzers may be used in high-volume convective therapies [105]. Such considerations should be taken into account during the design of follow-up studies in convective therapies. Our results suggest that dialyzers introduced in the last 15 years do not have substantially larger β2M clearance than those used during the landmark HEMO study in the late 1990s and early 2000s when used for conventional (diffusive) dialysis. Nevertheless, large secular trends consistent with improving dialyzer performance were observed when reduction ratios, rather than measured clearance or mass removal, were analyzed. Collectively, our analysis suggests not only that the basic mechanisms of middle molecule elimination by HF dialyzers has remained unchanged over the years, but the quantitative aspects of middle molecule centric HF dialysis have largely remained unchanged since HEMO was published. We should point out that these assessments do not apply to the emerging class of middle cut-off dialyzers, which not only have substantially higher middle molecule clearance than high flux membranes, but may even narrow the gap between high flux dialysis and HDF [92]. Despite the apparent lack of improvements in dialyzer performance, higher clearance (by up to 44%) may be attained by using the same dialyzers in convective therapies (HF or HDF). This was also noted when alternative, simple measures of middle molecule elimination, i.e. the reduction ratios, were utilized to compare diffusive and convective forms of dialysis. There are two mechanisms by which higher (pump) blood flow rates may increase β2M clearance in convective therapies: directly by increasing the amount of β2M available for diffusive clearance and indirectly by allowing higher rates of substitution fluid to be used, boosting the convective clearance. The latter mechanism is underscored by our finding that higher fluid substitution rates were significantly associated with higher β2M clearances in post-HDF therapies. This finding is supported by early studies on online HDF [106, 107] comparing the reduction ratios and the clearances of β2M, BUN, creatinine and phosphorus between HD and online HDF with 40–120 mL/min substitution fluid. The maximum benefit was achieved in HDF 100 (i.e. with 24 L substitution volume per 4-h treatment) versus classical HD. Another study of 2293 incident patients treated by post-dilution online HDF determined the convection volume threshold and range associated with survival advantage [108]. The relative adjusted survival rate was found to increase at about 55 L/week of convection volume and to stay increased up to about 75 L/week. The same paper found a nearly linear decrease in pre-dialysis β2M concentration by 0.6 mg/L for every 10 L/week of additional convection volume as the latter increased from 40 to 75 L/week. However, this mode can only be achieved with a permanent effective blood flow rate of at least 300 mL/min, since less than a third of this value can be accepted as the flow rate of the substitution fluid to avoid too high a TMP causing damage to the membrane. In the modern era, technical developments such as the adoption of variable ultrafiltration rates adapted to the level of the TMP during the treatment can be applied to achieve such high convection rates [109]. In fact there was a direct linear relationship between blood pump and dialysate flow rates in all the studies we analyzed, so that higher blood flow rates were associated with higher substitution fluid flow rates. The net result is that patients whose access could support high blood pump flow rates were the ones who received higher substitution fluid rate (>100 mL/min) and experienced the largest dialytic β2M removal. This pattern may be clinically significant, since a recent meta-analysis [14] of the large online HDF trials [21–23] and post hoc analyses published by the investigator teams in the last 5 years suggest an overall and cardiovascular survival advantage for these high-fluid rates. Treatment center policies about blood flow, treatment time, filter size and even hemoglobin level can be used in conjunction with the aforementioned technical innovations to achieve high convection volumes despite non-modifiable factors such as dialysis access that limit the achievement of higher blood and substitution fluid flow rates [109]. A surprising finding of our analysis was the lack of a meaningful effect of higher dialysate flow rates on improving diffusive or convective middle molecule clearance. This observation, which seems to go against classical teachings, is however fully in line with recent experiments about contemporary dialyzers for both small [110–112] and middle molecule clearance [113]. Design innovations such as spacer yarns in the fiber bundle, fiber undulations and changes in fiber-packing density have reduced the dependence of clearance on dialysate flow rates because of improved flow distribution in the dialyzer. Theoretical analysis based on the Weryński [114] and Michaels [115] equations relating diffusive clearance, sieving coefficient, Membrane Transfer Area Coefficient, blood and dialysate flow rates suggests that for dialyzers used in modern HF dialysis (sieving coefficient S = 0.65) and HDF (S = 0.75), increasing the dialysate flow by 60% from 500 to 800 mL/min will have a very small effect (∼0.4 mL/min) on middle molecule body clearance. Some limitations of this meta-analysis need to be acknowledged. First, the studies included differed in study design, methodologically (methods used for the calculation of clearance, dialysis modalities) and operationally (different dialyzers, different blood and dialysate flow rates, etc.). In particular, different approaches to calculate clearance will systematically overestimate (e.g. whole blood versus plasma) or underestimate (e.g. dialysate versus plasma) the dialytic clearance. We attempted to account for these systematic differences in our analyses through statistical modeling. However, residual confounding cannot be excluded. Such confounding may particularly apply to the apparent lack of an improvement of convective dialyzer performance with time, during a period in which many manufacturers released dialyzers with higher sieving coefficients for β2M and thus greater capacity for convective clearance. These dialyzers may also be more likely to remove β2M through adsorption in the inner layers of the dialyzer, so that studies relying on dialysate side measurements may have missed this finding. It should be noted that despite the lack of a statistically significant effect, the magnitude of the temporal trend for all dialyzer performance measures considered, is in the direction of more efficient removal with time. As further studies become available, our finding may notwithstand the passage of time. Second, most of the included studies recruited chronic HD patients on a thrice-weekly 4-h treatment schedule. Third, the apparent lack of an effect of higher dialysate flows may not apply to short, frequent, slow flow dialysis for membranes that do not exhibit enhanced dialytic removal at higher flows in conventional thrice-weekly dialysis [116]. Fourth, the limited sample size, selection of sampling points in the source data and analytical methodology of mixed models may have limited our ability to detect a statistically significant difference between instantaneous and average dialyzer clearances. Finally, this work is limited to adult patients and cannot be generalized to the pediatric dialysis population. CONCLUSIONS Dialysis prescription parameters (e.g. blood and dialysate flow rates in HD and infusion volume in HDF), as well as membrane material (HD), are major determinants of β2M clearance from the body in renal replacement therapies. Future prospective studies should standardize methodology for these measurements and investigate a wide variety of dialysis configurations to directly account for variability within and between patients and dialysis units. Such experimental studies are better suited than our statistical analyses to highlight clinically important differences related to the differential effects of β2M body removal seen with membranes of different material to inform their use in clinical HD and HDF. SUPPLEMENTARY DATA Supplementary data are available at ndt online. AUTHORS’ CONTRIBUTIONS The study was conceived and data were analyzed by the corresponding author. Data were generated by M.-E.R., G.T., Y.-H.N., Z.X., A.A. and R.F. Significant intellectual content was contributed by T.D.N. and M.L.U. All authors contributed to the interpretation of the data, drafting and revision of the manuscript. All authors have approved the final version of the article that was uploaded to the journal website. ACKNOWLEDGEMENTS A preliminary version of this work was presented in abstract form (THPO958, FRPO894) at Kidney Week 2016, Annual Meeting of the American Society of Nephrology, Chicago on November 2016. FUNDING No entity, private, public, commercial or non-for-profit funded directly or indirectly to this work. CONFLICT OF INTEREST STATEMENT None declared. REFERENCES 1 Cheung AK , Greene T , Leypoldt JK et al. . Association between serum β2-microglobulin level and infectious mortality in hemodialysis patients . Clin J Am Soc Nephrol 2008 ; 3 : 69 – 77 Google Scholar CrossRef Search ADS PubMed 2 Okuno S , Ishimura E , Kohno K et al. . Serum beta2-microglobulin level is a significant predictor of mortality in maintenance haemodialysis patients . Nephrol Dial Transplant 2008 ; 24 : 571 – 577 Google Scholar CrossRef Search ADS PubMed 3 Dember LM , Jaber BL. Unresolved issues in dialysis: dialysis-related amyloidosis: late finding or hidden epidemic? Semin Dial 2006 ; 19 : 105 – 109 Google Scholar CrossRef Search ADS PubMed 4 Jadoul M , Garbar C , van Ypersele de Strihou C. Pathological aspects of beta(2)-microglobulin amyloidosis . Semin Dial 2001 ; 14 : 86 – 89 Google Scholar CrossRef Search ADS PubMed 5 Thomas G , Jaber BL. Innovation in the treatment of uremia: Proceedings from the Cleveland Clinic Workshop: Convective Therapies for Removal of Middle Molecular Weight Uremic Toxins in End-Stage Renal Disease: a review of the evidence . Semin Dial 2009 ; 22 : 610 – 614 Google Scholar CrossRef Search ADS PubMed 6 Jadoul M , Garbar C , Noël H et al. . Histological prevalence of β2-microglobulin amyloidosis in hemodialysis: a prospective post-mortem study . Kidney Int 1997 ; 51 : 1928 – 1932 Google Scholar CrossRef Search ADS PubMed 7 Jadoul M , Garbar C , Vanholder R et al. . Prevalence of histological beta2-microglobulin amyloidosis in CAPD patients compared with hemodialysis patients . Kidney Int 1998 ; 54 : 956 – 959 Google Scholar CrossRef Search ADS PubMed 8 Argyropoulos CP , Chen SS , Ng Y-H et al. . Rediscovering beta-2 microglobulin as a biomarker across the spectrum of kidney diseases . Front Med 2017 ; 4 : 1 – 25 Google Scholar CrossRef Search ADS 9 Gal R , Korzets A , Schwartz A et al. . Systemic distribution of beta 2-microglobulin-derived amyloidosis in patients who undergo long-term hemodialysis. Report of seven cases and review of the literature . Arch Pathol Lab Med 1994 ; 118 : 718 – 721 Google Scholar PubMed 10 Takayama F , Miyazaki S , Morita T et al. . Dialysis-related amyloidosis of the heart in long-term hemodialysis patients . Kidney Int 2001 ; 59 : S172 – S176 Google Scholar CrossRef Search ADS 11 Argyropoulos C , Roumelioti M-E , Sattar A et al. . Dialyzer reuse and outcomes of high flux dialysis . PLoS One 2015 ; 10 : e0129575 Google Scholar CrossRef Search ADS PubMed 12 Palmer SC , Rabindranath KS , Craig JC et al. . High-flux versus low-flux membranes for end-stage kidney disease . Cochrane Database Syst Rev 2012 ; http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD005016.pub2/abstract (10 November 2017, date last accessed) 13 Locatelli F , Martin-Malo A , Hannedouche T et al. . Effect of membrane permeability on survival of hemodialysis patients . J Am Soc Nephrol 2009 ; 20 : 645 – 654 Google Scholar CrossRef Search ADS PubMed 14 Peters SAE , Bots ML , Canaud B et al. . Haemodiafiltration and mortality in end-stage kidney disease patients: a pooled individual participant data analysis from four randomized controlled trials . Nephrol Dial Transplant 2016 ; 31 : 978 – 984 Google Scholar CrossRef Search ADS PubMed 15 Nubé MJ , Peters SAE , Blankestijn PJ et al. . Mortality reduction by post-dilution online-haemodiafiltration: a cause-specific analysis . Nephrol Dial Transplant 2017 ; 32 : 548 – 555 Google Scholar PubMed 16 Nistor I , Palmer SC , Craig JC et al. . Convective versus diffusive dialysis therapies for chronic kidney failure: an updated systematic review of randomized controlled trials . Am J Kidney Dis 2014 ; 63 : 954 – 967 Google Scholar CrossRef Search ADS PubMed 17 Nistor I , Palmer SC , Craig JC et al. . Haemodiafiltration, haemofiltration and haemodialysis for end-stage kidney disease . Cochrane Database Syst Rev 2015 ; doi: 10.1002/14651858.CD006258.pub2 18 Susantitaphong P , Siribamrungwong M , Jaber BL. Convective therapies versus low-flux hemodialysis for chronic kidney failure: A meta-analysis of randomized controlled trials . Nephrol Dial Transplant 2013 ; 28 : 2859 – 2874 Google Scholar CrossRef Search ADS PubMed 19 Wang AY , Ninomiya T , Al-Kahwa A et al. . Effect of hemodiafiltration or hemofiltration compared with hemodialysis on mortality and cardiovascular disease in chronic kidney failure: a systematic review and meta-analysis of randomized trials . Am J Kidney Dis 2014 ; 63 : 968 – 978 Google Scholar CrossRef Search ADS PubMed 20 Eknoyan G , Beck GJ , Cheung AK et al. . Effect of dialysis dose and membrane flux in maintenance hemodialysis . N Engl J Med 2002 ; 347 : 2010 – 2019 Google Scholar CrossRef Search ADS PubMed 21 Grooteman MPC , van den Dorpel MA , Bots ML et al. . Effect of online hemodiafiltration on all-cause mortality and cardiovascular outcomes . J Am Soc Nephrol 2012 ; 23 : 1087 – 1096 Google Scholar CrossRef Search ADS PubMed 22 Maduell F , Moreso F , Pons M et al. . High-efficiency postdilution online hemodiafiltration reduces all-cause mortality in hemodialysis patients . J Am Soc Nephrol 2013 ; 24 : 487 – 497 Google Scholar CrossRef Search ADS PubMed 23 Ok E , Asci G , Toz H et al. . Mortality and cardiovascular events in online haemodiafiltration (OL-HDF) compared with high-flux dialysis: results from the Turkish OL-HDF Study . Nephrol Dial Transplant 2013 ; 28 : 192 – 202 Google Scholar CrossRef Search ADS PubMed 24 Thomas BH , Ciliska D , Dobbins M et al. . A process for systematically reviewing the literature: providing the research evidence for public health nursing interventions . Worldviews Evid Based Nurs 2004 ; 1 : 176 – 184 Google Scholar CrossRef Search ADS PubMed 25 Viechtbauer W. Conducting meta-analyses in R with the metafor package . J Stat Softw 2010 ; 36 : 1 – 48 Google Scholar CrossRef Search ADS 26 Leto E , Bilal F , Osmic I. Efficiency of high-flow dialyzers in removal of beta-2-microglobulin . Medicinski Arh 2001 ; 55 : 225 – 226 27 Xu XQ , Gruner N , Al-Bashir A et al. . Determination of extra renal clearance and generation rate of beta2-microglobulin in hemodialysis patients using a kinetic model . ASAIO J 2001 ; 47 : 623 – 627 Google Scholar CrossRef Search ADS PubMed 28 Yamada T , Akiba T , Sasaki S. Analysis of beta(2)-microglobulin kinetics in hemodialysis by a modified variable-volume one-compartment model . Blood Purif 2001 ; 19 : 388 – 394 Google Scholar CrossRef Search ADS PubMed 29 Stiller S , Xu XQ , Gruner N et al. . Validation of a two-pool model for the kinetics of beta2-microglobulin . Int J Artif Organs 2002 ; 25 : 411 – 420 Google Scholar CrossRef Search ADS PubMed 30 Ding F , Ahrenholz P , Winkler RE et al. . Online hemodiafiltration versus acetate-free biofiltration: A prospective crossover study . Artif Organs 2002 ; 26 : 169 – 180 Google Scholar CrossRef Search ADS PubMed 31 Klingel R , Ahrenholz P , Schwarting A et al. . Enhanced functional performance characteristics of a new polysulfone membrane for high-flux hemodialysis . Blood Purif 2002 ; 20 : 325 – 333 Google Scholar CrossRef Search ADS PubMed 32 Mann H , Al-Bashir A , Melzer H et al. . Diacap® α-polysulfone HI PS: A new dialysis membrane with optimum β2-microglobulin elimination . Int J Artif Organs 2003 ; 26 : 461 – 466 Google Scholar CrossRef Search ADS PubMed 33 Mandolfo S , Malberti F , Imbasciati E et al. . Impact of blood and dialysate flow and surface on performance of new polysulfone hemodialysis dialyzers . Int J Artif Organs 2003 ; 26 : 113 – 120 Google Scholar CrossRef Search ADS PubMed 34 Pedrini LA , De Cristofaro V. On-line mixed hemodiafiltration with a feedback for ultrafiltration control: effect on middle-molecule removal . Kidney Int 2003 ; 64 : 1505 – 1513 Google Scholar CrossRef Search ADS PubMed 35 Ward RA , Ouseph R. Impact of bleach cleaning on the performance of dialyzers with polysulfone membranes processed for reuse using peracetic Acid . Artif Organs 2003 ; 27 : 1029 – 1034 Google Scholar CrossRef Search ADS PubMed 36 Bammens B , Evenepoel P , Verbeke K et al. . Removal of the protein-bound solute p-cresol by convective transport: A randomized crossover study . Am J Kidney Dis 2004 ; 44 : 278 – 285 Google Scholar CrossRef Search ADS PubMed 37 Yamashita AC , Kawanishi H. Kinetics and dose of daily hemofiltration . Blood Purif 2004 ; 22 (Suppl 2) : 14 – 19 Google Scholar CrossRef Search ADS PubMed 38 Emiliani G , Briganti M , Montanari A et al. . [On-line haemodiafiltration with sorbent-regenerated ultrafiltrate as replacement fluid: Beta2-microglobulin removal versus filtration fraction] . G Ital Nefrol 2004 ; 21 (Suppl 30) : S80 – S84 Google Scholar PubMed 39 Leypoldt JK , Cheung AK , Deeter RB et al. . Kinetics of urea and beta-microglobulin during and after short hemodialysis treatments . Kidney Int 2004 ; 66 : 1669 – 1676 Google Scholar CrossRef Search ADS PubMed 40 Lucchi L , Fiore GB , Guadagni G et al. . Clinical evaluation of internal hemodiafiltration (iHDF): A diffusive-convective technique performed with internal filtration enhanced high-flux dialyzers . Int J Artif Organs 2004 ; 27 : 414 – 419 Google Scholar CrossRef Search ADS PubMed 41 Pisitkun T , Eiam-Ong S , Tiranathanagul K et al. . Convective-controlled double high flux hemodiafiltration: A novel blood purification modality . Int J Artif Organs 2004 ; 27 : 195 – 204 Google Scholar CrossRef Search ADS PubMed 42 Tonelli M , Dymond C , Gourishankar S et al. . Extended reuse of polysulfone hemodialysis membranes using citric acid and heat . ASAIO J 2004 ; 50 : 98 – 101 Google Scholar CrossRef Search ADS PubMed 43 Santoro A , Conz PA , De Cristofaro V et al. . Mid-dilution: The perfect balance between convection and diffusion. Contrib Nephrol 2005 ; 149 : 107 – 114 44 Brendolan A , Nalesso F , Fortunato A et al. . Dialytic performance evaluation of RexeedTM: A new polysulfone-based dialyzer with improved flow distributions . Int J Artif Organs 2005 ; 28 : 966 – 975 Google Scholar CrossRef Search ADS PubMed 45 Padrini R , Canova C , Conz P et al. . Convective and adsorptive removal of beta2-microglobulin during predilutional and postdilutional hemofiltration . Kidney Int 2005 ; 68 : 2331 – 2337 Google Scholar CrossRef Search ADS PubMed 46 Petras D , Fortunato A , Soffiati G et al. . Sequential convective therapies (SCT): a prospective study on feasibility, safety, adequacy and tolerance of on-line hemofiltration and hemodiafiltration in sequence . Int J Artif Organs 2005 ; 28 : 482 – 488 Google Scholar CrossRef Search ADS PubMed 47 Krieter DH , Collins G , Summerton J et al. . Mid-dilution on-line haemodiafiltration in a standard dialyser configuration . Nephrol Dial Transplant 2005 ; 20 : 155 – 160 Google Scholar CrossRef Search ADS PubMed 48 Krieter DH , Falkenhain S , Chalabi L et al. . Clinical cross-over comparison of mid-dilution hemodiafiltration using a novel dialyzer concept and post-dilution hemodiafiltration . Kidney Int 2005 ; 67 : 349 – 356 Google Scholar CrossRef Search ADS PubMed 49 Evenepoel P , Bammens B , Verbeke K et al. . Superior dialytic clearance of β2-microglobulin and p-cresol by high-flux hemodialysis as compared to peritoneal dialysis . Kidney Int 2006 ; 70 : 794 – 799 Google Scholar CrossRef Search ADS PubMed 50 Mandolfo S , Borlandelli S , Imbasciati E. Leptin and β2-microglobulin kinetics with three different dialysis modalities . Int J Artif Organs 2006 ; 29 : 949 – 955 Google Scholar CrossRef Search ADS PubMed 51 Nakashima A , Ogata S , Doi S et al. . Performance of polysulfone membrane dialyzers and dialysate flow pattern . Clin Exp Nephrol 2006 ; 10 : 210 – 215 Google Scholar CrossRef Search ADS PubMed 52 Panich A , Tiranathanagul K , Praditpornsilpa K et al. . The effectiveness of on-line hemodiafiltration on beta-2 microglobulin clearance in end stage renal disease . J Med Assoc Thai 2006 ; 89 (Suppl 2) : S1 – S8 Google Scholar PubMed 53 Pedrini LA , Cozzi G , Faranna P et al. . Transmembrane pressure modulation in high-volume mixed hemodiafiltration to optimize efficiency and minimize protein loss . Kidney Int 2006 ; 69 : 573 – 579 Google Scholar CrossRef Search ADS PubMed 54 Potier J. Mid-dilution: an innovative high-quality and safe haemodiafiltration approach. Contrib Nephrol 2007 ; 158 : 153 – 160 55 Feliciani A , Riva MA , Zerbi S et al. . New strategies in haemodiafiltration (HDF): prospective comparative analysis between on-line mixed HDF and mid-dilution HDF . Nephrol Dial Transplant 2007 ; 22 : 1672 – 1679 Google Scholar CrossRef Search ADS PubMed 56 Krieter DH , Morgenroth A , Barasinski AA et al. . Effects of a polyelectrolyte additive on the selective dialysis membrane permeability for low-molecular-weight proteins . Nephrol Dial Transplant 2007 ; 22 : 491 – 499 Google Scholar CrossRef Search ADS PubMed 57 Santoro A , Ferramosca E , Mancini E et al. . Reverse mid-dilution: New way to remove small and middle molecules as well as phosphate with high intrafilter convective clearance . Nephrol Dial Transplant 2007 ; 22 : 2000 – 2005 Google Scholar CrossRef Search ADS PubMed 58 Tiranathanagul K , Yossundharakul C , Techawathanawanna N et al. . Comparison of middle-molecule clearance between convective control double high-flux hemodiafiltration and on-line hemodiafiltration . Int J Artif Organs 2007 ; 30 : 1090 – 1097 Google Scholar CrossRef Search ADS PubMed 59 Abe M , Okada K , Maruyama T et al. . Clinical evaluation of plasma insulin and C-peptide levels with 3 different high-flux dialyzers in diabetic patients on hemodialysis . Int J Artif Organs 2008 ; 31 : 898 – 904 Google Scholar CrossRef Search ADS PubMed 60 Eloot S , Van Biesen W , Dhondt A et al. . Impact of hemodialysis duration on the removal of uremic retention solutes . Kidney Int 2008 ; 73 : 765 – 770 Google Scholar CrossRef Search ADS PubMed 61 Mandolfo S , Borlandelli S , Imbasciati E et al. . Pilot study to assess increased dialysis efficiency in patients with limited blood flow rates due to vascular access problems . Hemodial Int 2008 ; 12 : 55 – 61 Google Scholar CrossRef Search ADS PubMed 62 Spalding EM , Pandya P , Farrington K. Effect of high haematocrit on the efficiency of high-flux dialysis therapies . Nephron Clin Pract 2008 ; 110 : c86 – c92 Google Scholar CrossRef Search ADS PubMed 63 Krieter DH , Hunn E , Morgenroth A et al. . Matching efficacy of online hemodiafiltration in simple hemodialysis mode . Artif Organs 2008 ; 32 : 903 – 909 Google Scholar CrossRef Search ADS PubMed 64 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 : 1704 – 1712 Google Scholar CrossRef Search ADS PubMed 65 Krieter DH , Lemke H-D , Wanner C. A new synthetic dialyzer with advanced permselectivity for enhanced low-molecular weight protein removal . Artif Organs 2008 ; 32 : 547 – 554 Google Scholar CrossRef Search ADS PubMed 66 Joyeux V , Sijpkens Y , Haddj-Elmrabet A et al. . Optimized convective transport with automated pressure control in on-line postdilution hemodiafiltration . Int J Artif Organs 2008 ; 31 : 928 – 936 Google Scholar CrossRef Search ADS PubMed 67 Lee D , Haase M , Haase-Fielitz A et al. . A pilot, randomized, double-blind, cross-over study of high cut-off versus high-flux dialysis membranes . Blood Purif 2009 ; 28 : 365 – 372 Google Scholar CrossRef Search ADS PubMed 68 Meert N , Eloot S , Waterloos M-A et al. . Effective removal of protein-bound uraemic solutes by different convective strategies: a prospective trial . Nephrol Dial Transplant 2009 ; 24 : 562 – 570 Google Scholar CrossRef Search ADS PubMed 69 Pedrini LA , Feliciani A , Zerbi S et al. . Optimization of mid-dilution haemodiafiltration: technique and performance . Nephrol Dial Transplant 2009 ; 24 : 2816 – 2824 Google Scholar CrossRef Search ADS PubMed 70 Susantitaphong P , Tiranathanagul K , Hanvivatvong O et al. . A simple efficient technique of “mid-dilution” on-line hemodiafiltration . Blood Purif 2009 ; 28 : 93 – 101 Google Scholar CrossRef Search ADS PubMed 71 Troidle L , Finkelstein F , Hotchkiss M et al. . Enhanced solute removal with intermittent, in-center, 8-hour nocturnal hemodialysis . Hemodial Int 2009 ; 13 : 487 – 491 Google Scholar CrossRef Search ADS PubMed 72 Wang Y-M , Zhang W-M , Wang B-S et al. . Solute clearance characteristics of REXEEDTM series dialyzer during high-flux dialysis . J Shanghai Jiaotong Univ 2009 ; 29 : 858 – 861 73 Bhimani JP , Ouseph R , Ward RA. Effect of increasing dialysate flow rate on diffusive mass transfer of urea, phosphate and β2-microglobulin during clinical haemodialysis . Nephrol Dial Transplant 2010 ; 25 : 3990 – 3995 Google Scholar CrossRef Search ADS PubMed 74 Gascó J , Iñigo V , Mascarós V et al. . Kinetics of beta2-microglobulin and urea in high-efficiency hemodiafiltration . Blood Purif 2010 ; 30 : 224 – 225 75 Kohn OF , Coe FL , Ing TS. Solute kinetics with short-daily home hemodialysis using slow dialysate flow rate . Hemodial Int 2010 ; 14 : 39 – 46 Google Scholar CrossRef Search ADS PubMed 76 Krieter DH , Hackl A , Rodriguez A et al. . Protein-bound uraemic toxin removal in haemodialysis and post-dilution haemodiafiltration . Nephrol Dial Transplant 2010 ; 25 : 212 – 218 Google Scholar CrossRef Search ADS PubMed 77 Park J-S , Kim G-H , Kang CM et al. . Application of cystatin C reduction ratio to high-flux hemodialysis as an alternative indicator of the clearance of middle molecules . Korean J Intern Med 2010 ; 25 : 77 – 81 Google Scholar CrossRef Search ADS PubMed 78 Basile C , Libutti P , Di Turo AL et al. . Removal of uraemic retention solutes in standard bicarbonate haemodialysis and long-hour slow-flow bicarbonate haemodialysis . Nephrol Dial Transplant 2011 ; 26 : 1296 – 1303 Google Scholar CrossRef Search ADS PubMed 79 Ficheux A , Gayrard N , Szwarc I et al. . The use of SDSPAGE scanning of spent dialysate to assess uraemic toxin removal by dialysis . Nephrol Dial Transplant 2011 ; 26 : 2281 – 2289 Google Scholar CrossRef Search ADS PubMed 80 Pedrini LA , Gmerek A , Wagner J. Efficiency of post-dilution hemodiafiltration with a high-flux α-polysulfone dialyzer . Int J Artif Organs 2011 ; 34 : 397 – 404 Google Scholar CrossRef Search ADS PubMed 81 Panichi V , De Ferrari G , Saffiotti S et al. . Comparison of on-line HDF modes automated TMP control vs. Volume control on achieved convective volume and middle molecule clearance . Nephrol Dial Transplant 2012 ; 27 : ii205 – ii206 Google Scholar CrossRef Search ADS 82 Panichi V , De Ferrari G , Saffioti S et al. . Divert to ultra: differences in infused volumes and clearance in two on-line hemodiafiltration treatments . Int J Artif Organs 2012 ; 35 : 435 – 443 Google Scholar CrossRef Search ADS PubMed 83 Susantitaphong P , Tiranathanagul K , Katavetin P et al. . Efficacy comparison between simple mixed-dilution and simple mid-dilution on-line hemodiafiltration techniques: a crossover study . Artif Organs 2012 ; 36 : 1059 – 1065 Google Scholar CrossRef Search ADS PubMed 84 Tessitore N , Bedogna V , Girelli D et al. . Effect of a single hemodialysis session by HFR and low-flux bicarbonate dialysis on serum hepcidin-25 levels: a randomized cross-over study . Nephrol Dial Transplant 2012 ; 27 (Suppl 2): i207 85 von Albertini B , Mathieu C , Cherpillod A et al. . In-vivo 2 microglobulin clerance in high-flux HD & HDF . Nephrol Dial Transplant 2013 ; 28 (Supplement 1) : i422 86 Heaf JG , Axelsen M , Pedersen RS. Multipass haemodialysis: A novel dialysis modality . Nephrol Dial Transplant 2013 ; 28 : 1255 – 1264 Google Scholar CrossRef Search ADS PubMed 87 Melo NCV , Moyses RMA , Elias RM et al. . Reprocessing high-flux polysulfone dialyzers does not negatively impact solute removal in short-daily online hemodiafiltration . Hemodial Int 2014 ; 18 : 473 – 480 Google Scholar CrossRef Search ADS PubMed 88 Pedrini LA , Krisp C , Gmerek A et al. . Patterns of proteins removed with high-flux membranes on high-volume hemodiafiltration detected with a multidimensional LC-MS/MS strategy . Blood Purif 2014 ; 38 : 115 – 126 Google Scholar CrossRef Search ADS PubMed 89 Cornelis T , Van Der Sande FM , Eloot S et al. . Acute hemodynamic response and uremic toxin removal in conventional and extended hemodialysis and hemodiafiltration: A randomized crossover study . Am J Kidney Dis 2014 ; 64 : 247 – 256 Google Scholar CrossRef Search ADS PubMed 90 Potier J , Bowry S , Canaud B. Clinical performance assessment of CorDiax filters in hemodialysis and hemodiafiltration. Contrib Nephrol 2017 ; 189 : 237 – 245 91 Gayrard N , Ficheux A , Duranton F et al. . Consequences of increasing convection onto patient care and protein removal in hemodialysis . PloS One 2017 ; 12 : e0171179 Google Scholar CrossRef Search ADS PubMed 92 Kirsch AH , Lyko R , Nilsson L-G et al. . Performance of hemodialysis with novel medium cut-off dialyzers . Nephrol Dial Transplant 2017 ; 32 : 165 – 172 Google Scholar PubMed 93 Pieroni L , Mortera SL , Greco V et al. . Biocompatibility assessment of haemodialysis membrane materials by proteomic investigations . Mol BioSyst 2015 ; 11 : 1633 – 1643 Google Scholar CrossRef Search ADS PubMed 94 Armijo-Olivo S , Stiles CR , Hagen NA et al. . Assessment of study quality for systematic reviews: a comparison of the Cochrane Collaboration Risk of Bias Tool and the Effective Public Health Practice Project Quality Assessment Tool: methodological research . J Eval Clin Pract 2012 ; 18 : 12 – 18 Google Scholar CrossRef Search ADS PubMed 95 Mayer G , Thum J , Woloszczuk W et al. . Beta-2-microglobulin in hemodialysis patients. Effects of different dialyzers and different dialysis procedures . Am J Nephrol 1988 ; 8 : 280 – 284 Google Scholar CrossRef Search ADS PubMed 96 Acchiardo S , Kraus AP Jr , Jennings BR. Beta 2-microglobulin levels in patients with renal insufficiency . Am J Kidney Dis 1989 ; 13 : 70 – 74 Google Scholar CrossRef Search ADS PubMed 97 Hoenich NA , Stamp S. Clinical performance of a new high-flux synthetic membrane . Am J Kidney Dis 2000 ; 36 : 345 – 352 Google Scholar CrossRef Search ADS PubMed 98 Clark WR , Macias WL , Molitoris BA et al. . Membrane adsorption of β2-microglobulin: equilibrium and kinetic characterization . Kidney Int 1994 ; 46 : 1140 – 1146 Google Scholar CrossRef Search ADS PubMed 99 Clark WR , Macias WL , Molitoris BA et al. . Plasma protein adsorption to highly permeable hemodialysis membranes . Kidney Int 1995 ; 48 : 481 – 488 Google Scholar CrossRef Search ADS PubMed 100 Donadio C , Kanaki A , Sami N et al. . Proteomic evaluation of low- and high-flux hemodialysis . Nephrol Dial Transplant 2013 ; 28 (Suppl 1): i421 Google Scholar CrossRef Search ADS 101 Ishikawa I , Chikazawa Y , Sato K et al. . Proteomic analysis of serum, outflow dialysate and adsorbed protein onto dialysis membranes (polysulfone and PMMA) during hemodialysis treatment using SELDI-TOF-MS . Am J Nephrol 2006 ; 26 : 372 – 380 Google Scholar CrossRef Search ADS PubMed 102 Han S , Yang K , Sun J et al. . Proteomics investigations into serum proteins adsorbed by high-flux and low-flux dialysis membranes . Proteomics Clin Appl 2017 ; doi: 10.1002/prca.201700079 103 Cuoghi A , Caiazzo M , Monari E et al. . New horizon in dialysis depuration: characterization of a polysulfone membrane able to break the ‘albumin wall’ . J Biomater Appl 2015 ; 29 : 1363 – 1371 Google Scholar CrossRef Search ADS PubMed 104 Maduell F , Arias-Guillen M , Fontseré N et al. . Elimination of large uremic toxins by a dialyzer specifically designed for high-volume convective therapies . Blood Purif 2014 ; 37 : 125 – 130 Google Scholar CrossRef Search ADS PubMed 105 Potier J , Queffeulou G , Bouet J. Are all dialyzers compatible with the convective volumes suggested for postdilution online hemodiafiltration? Int J Artif Organs 2016 ; 39 : 460 – 470 Google Scholar CrossRef Search ADS PubMed 106 Lornoy W , Becaus I , Billiouw JM et al. . Remarkable removal of beta-2-microglobulin by on-line hemodiafiltration . Am J Nephrol 1998 ; 18 : 105 – 108 Google Scholar CrossRef Search ADS PubMed 107 Lornoy W , Becaus I , Billiouw JM et al. . On-line haemodiafiltration. Remarkable removal of beta2-microglobulin. Long-term clinical observations . Nephrol Dial Transplant 2000 ; 15 : 49 – 54 Google Scholar CrossRef Search ADS PubMed 108 Canaud B , Barbieri C , Marcelli D et al. . Optimal convection volume for improving patient outcomes in an international incident dialysis cohort treated with online hemodiafiltration . Kidney Int 2015 ; 88 : 1108 – 1116 Google Scholar CrossRef Search ADS PubMed 109 Marcelli D , Kopperschmidt P , Bayh I et al. . Modifiable factors associated with achievement of high-volume post-dilution hemodiafiltration: results from an international study . Int J Artif Organs 2015 ; 38 : 244 – 250 Google Scholar CrossRef Search ADS PubMed 110 Albalate M , Pérez-García R , De Sequera P et al. . Is it useful to increase dialysate flow rate to improve the delivered Kt? BMC Nephrol 2015 ; 16 : 1 – 6 Google Scholar CrossRef Search ADS 111 Albalate Ramón M , de Sequera Ortiz P , Pérez-García R et al. . What is the optimum dialysate flow in post-dilution online haemodiafiltration? Nefrologia 2015 ; 35 : 533 – 538 Google Scholar CrossRef Search ADS PubMed 112 Ward RA , Idoux JW , Hamdan H et al. . Dialysate flow rate and delivered kt/v urea for dialyzers with enhanced dialysate flow distribution . Clin J Am Soc Nephrol 2011 ; 6 : 2235 – 2239 Google Scholar CrossRef Search ADS PubMed 113 Maduell F , Ojeda R , Arias-Guillén M et al. . Optimisation of dialysate flow in on-line hemodiafiltration . Nefrologia 2015 ; 35 : 473 – 478 Google Scholar CrossRef Search ADS PubMed 114 Weryński A. Evaluation of the impact of ultrafiltration on dialyzer clearance . Artif Organs 1979 ; 3 : 140 – 142 Google Scholar CrossRef Search ADS PubMed 115 Michaels AS. Analysis of membrane-transport devices . ASAIO Trans 1966 ; 12 : 387 – 392 116 Leypoldt JK , Kamerath CD , Gilson JF et al. . Dialyzer clearances and mass transfer-area coefficients for small solutes at low dialysate flow rates . ASAIO J 2006 ; 52 : 404 – 409 Google Scholar CrossRef Search ADS PubMed © The Author 2017. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Nephrology Dialysis TransplantationOxford University Press

Published: Nov 27, 2017

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off