Kidney International, Vol. 56 (1999), pp. 2005–2015

PERSPECTIVES IN RENAL MEDICINE

Effect of membrane composition and structure on soluteremoval and biocompatibility in hemodialysis

WILLIAM R. CLARK, RICHARD J. HAMBURGER, and MICHAEL J. LYSAGHT

Renal Division, Baxter Healthcare Corporation, McGaw Park, Illinois; Nephrology Division, Indiana University School ofMedicine, Indianapolis, Indiana; and Artificial Organs Program, Brown University, Providence, Rhode Island, USA

Effect of membrane composition and structure on solute re- and a high degree of complement activation were themoval and biocompatibility in hemodialysis. Significant changes norm. The efficiency of small solute removal by unsubsti-in extracorporeal membranes have occurred over the past five tuted cellulosic dialyzers improved markedly in the latedecades in which hemodialysis (HD) has been available as a

1960s with the introduction of modified flat-plate dialyz-therapy for both acute renal failure (ARF) and end-stage renalers [4], which significantly decreased blood compartmentdisease (ESRD). For cellulosic membranes, these changes have

included a reduction in thickness, hydroxyl group substitution, mass transfer resistance. Solute mass transfer was furtherand an increase in pore size. These modifications have resulted improved in the early 1970s when cellulosic hollow fiberin enhanced efficiency of small solute removal, a broader spec-

membranes [5] began to be offered. Also occurring intrum of overall solute removal, and an attenuation of comple-this decade was the development of synthetic mem-ment activation in comparison to the thick, unsubstituted cellu-

losic membranes of low permeability used in the early days branes, such as polyacrylonitrile (PAN) and polysulfoneof HD therapy. Synthetic membranes, originally developed [6]. These latter membranes differed fundamentally fromspecifically for use in high-flux HD and hemofiltration, have regenerated cellulose not only in their polymeric compo-also evolved during this same time period. In fact, the initially

sition but also in a number of other features, includingclear distinction between low-flux regenerated cellulosic andpore size, thickness, and hydrophobicity.high-flux synthetic membranes has become blurred, as mem-

brane formulators have developed products designed to appeal Since these early days of dialysis therapy, additionalto enthusiasts for both membrane formats. The purpose of this membranes have been introduced into both the cellulosicreview is to characterize both the solute removal and biocom- and synthetic classes. An initially clear distinction be-patibility characteristics of dialysis membranes according to

tween low-flux regenerated cellulosic and high-flux syn-their composition (that is, polymeric makeup) and structure.thetic membranes has, however, become blurred asIn this regard, the manner in which membrane biocompatibility

interacts with flux is highlighted. membrane formulators have developed products de-signed to appeal to enthusiasts for both membrane for-mats. The purpose of this article is to provide an overview

Over the past five decades, hemodialyzers used for of current dialysis membrane technology, emphasizingacute and chronic renal replacement therapy have contin- the manner in which membrane composition and struc-uously evolved. Utilization of regenerated cellulose (that ture influence both solute removal capabilities and bio-is, unsubstituted cellulosic) membranes of small appar- compatibility properties.ent pore size and large thickness in dialyzers with ineffi-cient mass transfers characteristics characterized the very

PERMEABILITY CLASSIFICATION OFearly days of hemodialysis (HD) [1–3]. In fact, coil de-DIALYSIS MEMBRANESsigns with regenerated cellulose tubes were used almost

exclusively throughout the 1960s. Therefore, inefficient Although numerous classification schemes have beensmall solute removal, negligible middle molecule removal, proposed [7], HD membranes are traditionally classified

according to water flux, a term synonymous with waterpermeability. The clinical parameter used to characterize

Key words: acute renal failure, end-stage renal disease, cellulosic dial-the water permeability of a dialyzer is the ultrafiltrationysis membranes, synthetic dialysis membranes, hemofiltration, biocom-

patibility, water flux. coefficient (KUF: ml/hr/mm Hg). The water permeabilityof a dialyzer is usually derived from in vitro experiments

Received for publication November 6, 1998in which bovine blood is ultrafiltered at varying trans-and in revised form February 12, 1999

Accepted for publication April 7, 1999 membrane pressures (TMPs). The relationship betweenplasma ultrafiltration rate and TMP is linear at relatively 1999 by the International Society of Nephrology

2005

Clark et al: HD membranes2006

low TMP values for all membranes, whereas a plateau The potential dissociation between water and solutein ultrafiltration rate occurs at relatively high TMP values flux is demonstrated specifically by a Gambro copolymer[8]. Dialyzer KUF is defined by the slope of the linear membrane containing both hydrophobic (polycarbo-portion of this ultrafiltration rate versus TMP curve. The nate) and hydrophilic (polyether) domains [9]. Althoughmembrane characteristic having the largest impact on strictly defined as a synthetic, this membrane’s uniquewater permeability is pore size. As discussed by Lysaght composition also endows it with some properties that[6], ultrafiltrate flux (ultrafiltration rate per unit area of are more characteristic of the cellulosic class. The in vitromembrane) is roughly proportional to the fourth power vitamin B12 clearance for a 1.1 m2 dialyzer of this typeof the mean membrane pore radius. As such, small (KUF 6.1 ml/hr/mm Hg) is 87 ml/min [10], and the clinicalchanges in pore size have a very large effect on water use of this dialyzer achieves a significant (approximatelypermeability. 20 to 30%) reduction in serum b2-microglobulin (b2m)

A common misconception relating to dialyzer perfor- concentration [11, 12]. In contrast, unsubstituted cellu-mance is the assumption that a membrane’s solute re- losic, cellulose acetate, and polysulfone dialyzers withmoval capabilities are necessarily correlated with its water roughly the same KUF and surface area have in vitropermeability. Diffusion, the predominant mass transfer vitamin B12 clearances of approximately 50 ml/min andmechanism in HD, can be characterized by solute flux (φ: are unable to achieve a significant reduction in serummass removal rate per unit surface area of membrane). b2m concentration in the clinical setting [11]. Ward et alBased on a model in which a membrane has N (straight) have recently suggested that the mechanism by whichcylindrical pores (per unit membrane surface area) of this membrane can remove b2m despite a low waterradius r, diffusive solute flux can be expressed as [6] permeability is either by adsorption to the hydrophobicfollows: microdomains or by diffusion through the relatively few

large pores in the hydrophilic domains [12]. Unmodifiedφ 5 lDrDC/t (Eq. 1)cellulosic and cellulose acetate membranes having this

where l is the solute partition coefficient; D is solute same dissociation between water and solute permeabilitydiffusivity; r is membrane porosity; DC is the transmem- have also been developed recently (discussed later inbrane concentration gradient, and t is membrane thick- this article).ness. (While the partition coefficient is essentially unity The only dialyzer classification scheme recognized byfor solutes such as urea and creatinine, larger solutes the United States Food and Drug Administration iswith incomplete access to the membrane pores have l based on water permeability, with low and high perme-values that are less than one.) Membrane porosity is a ability dialyzers having KUF values of ,8 and $8 ml/hr/function of both pore size and number:

mm Hg, respectively. This scheme is at odds with thatr 5 Npr2 (Eq. 2) used primarily in clinical practice in which dialyzers are

divided into low-flux, high-efficiency, and high-flux cate-Equations 1 and 2 suggest that diffusive transport isgories. This categorization is based primarily on soluterelatively favorable for low molecular weight solutes notpermeability properties. The National Institutes of Healthonly because of the inverse relationship between molecu-Hemodialysis (HEMO) Study [13] defines low-flux andlar weight and diffusivity, but also because of the greaterhigh-flux dialyzers as those having a b2m clearance ofaccess of small solutes to the membrane pore structure.less than 10 ml/min and greater than 20 ml/min duringEquation 1 also indicates that diffusive transport is en-first use, respectively. (An additional high-flux require-hanced at low values of membrane thickness.ment is a KUF .14 ml/hr/mm Hg.) The term “high effi-The dissociation between solute and water permeabil-ciency” actually had its origin as a therapy, first describedity for dialysis membranes can be explained by use ofby Keshaviah et al [14], rather than a specific type ofthe previously mentioned fundamental principles. Fordialyzer. These investigators employed large surface areaall dialysis membranes, small solutes such as urea andcellulose acetate dialyzers and relatively high blood flowcreatinine have free pore access (l 5 1). Therefore,rates to achieve sufficiently high urea clearances to allowsmall solute transport is highly dependent on membranereductions in treatment time. The use of large surfaceporosity. As equation 2 indicates, one membrane witharea dialyzers capable of achieving these high urea clear-a large number of relatively small pores and a secondances has increased to the point that a separate, albeitmembrane with a small number of relatively large poressomewhat indistinct, class of high-efficiency dialyzerscan have equivalent porosities. Although the small solutenow exists. Although this class is defined primarily bytransport properties of these two hypothetical mem-its solute removal capabilities, dialyzers in this categorybranes would be equivalent, the flux (water permeabil-generally have KUF values falling in the 8 to 15 ml/hr/ity) properties would greatly differ. This difference ismm Hg range, essentially bridging the gap between low-explained by the strong dependence of ultrafiltrate flux

on membrane pore size (described earlier in this article). flux and high-flux dialyzers.

Clark et al: HD membranes 2007

BASIC CHARACTERISTICS OF membranes because of their ability to adsorb these fac-CELLULOSIC MEMBRANES tors nonspecifically, it appears to play a role for Hemo-

phant also [24]. Finally, a vitamin E-modified cellulosicThe monomeric subunit of cellulosic membranes ismembrane has also been introduced recently [25].cellobiose, a naturally occurring saccharide found in

The SMC membrane, the polycarbonate/polyetherplants [4]. Chemically, cellobiose is a ringed structuremembrane, and certain synthetic membranes, are exam-richly endowed with hydroxyl groups. The interactionples of surfaces endowed with microdomain structures.of complement cascade products with these hydroxylAs discussed recently [21, 26], a membrane having agroups is felt to be responsible, at least partly, for thecombination of hydrophilic and hydrophobic domainsrelatively pronounced complement activation and leuko-rather than a uniformly hydrophilic or hydrophobic com-penia observed when unsubstituted cellulosic mem-position has theoretical advantages with respect to bio-branes contact blood [15]. For the past several years,compatibility. Although hydrophilic surfaces promotea major objective among manufacturers has been thecomplement activation, hydrophobic surfaces tend to bedevelopment of modified (substituted) cellulosic mem-thombogenic because of their ability to adsorb proteinsbranes in which a certain fraction of these hydroxyland activate platelets [27]. The purposeful interspersinggroups is replaced with other moieties. The substitutionof hydrophilic and hydrophobic domains in a “checker-groups diminish the degree of complement activation byboard” pattern is designed to strike a balance betweenat least three different mechanisms. One mechanism iscomplement activation and thrombogenicity and to en-the replacement of a large percentage of the hydroxylhance the overall biocompatibility of the previously men-groups with an acetate radical [5]. (Although the hollowtioned membranes.fiber membrane produced by Dow in the 1970s was ex-

The evolution in cellulosic membranes has resulted intruded as cellulose acetate, it functionally was regener-a wide spectrum of biocompatibility and flux profiles. Ifated cellulose due to a de-esterification step in the manu-complement activation and neutropenia are used as thefacturing process.) In the first truly substituted cellulosicmajor biocompatibility criteria [28, 29], regenerated cel-membrane, cellulose (di)acetate, approximately 70 tolulose is the least biocompatible, whereas cellulose triac-80% of the hydroxyl groups on the cellulosic backboneetate is the most biocompatible, with the other modifiedwere replaced with an acetate group. Most likely becausecellulosic membranes having intermediate profiles. How-this modification eliminates a large fraction of the activeever, characterization of the flux properties of these mem-surface sites for interaction with complement compo-branes is not as straightforward. For dialyzers of compa-nents, an attenuation of the intense complement activa-rable surface area, a simplistic approach is to reporttion seen with unmodified cellulosics was achieved. ThisKUF values in the following ascending order: regeneratedmembrane modification also resulted in a moderate in-cellulose , Hemophant, SMC , cellulose acetate ,crease in pore size, yielding a slightly higher water per-cellulose triacetate [10, 30]. In this simplistic scheme, ameability and broader solute removal spectrum for cellu-1.5 m2 dialyzer having a regenerated cellulose, Hemo-lose acetate in comparison to unsubstituted cellulosicphant or SMC membrane generally falls in the low-fluxmembranes of similar surface area [14]. Extrapolation ofcategory (KUF ,8 ml/hr/mm Hg), whereas comparablythis process to total replacement of the hydroxyl groupssized dialyzers having cellulose acetate and cellulose tri-resulted in the cellulose triacetate fiber characterized byacetate membranes fall in the mid-flux (KUF 10 to 20further attenuation of complement activation and higherml/hr/mm Hg) and high-flux (KUF .20 ml/hr/mm Hg)water permeability [16]. Recent data demonstrate thecategories, respectively. However, this simplistic catego-biocompatibility profiles for high-flux cellulose triacetaterization scheme breaks down in several respects. High-and polysulfone dialyzers are comparable [16–18].flux cellulose acetate membranes have now been pro-A second cellulosic substitution mechanism is the re-duced by two different HD membrane manufacturersplacement of a relatively small percentage (less than 5%)[18, 31, 32], and cellulose triacetate dialyzers of low waterof the hydroxyl groups with a bulky chemical group,permeability (KUF 9.5 ml/hr/mm Hg) are also availablewhich sterically reduces the degree of interaction be-[33]. Finally, the recent development of unmodified cel-tween complement activation products and the mem-lulosic and cellulose acetate membranes having relativelybrane. Examples for which this strategy is employed arelow water permeability but solute removal capabilitiesHemophant (tertiary amine substitution) [19, 20] andthat include b2m [34, 35] further confounds this classifi-synthetically modified cellulose (SMC; benzyl substitu-cation scheme and provides additional examples of ation group) [21]. A final mechanism by which complementdissociation between water and solute flux. Thus, as isactivation can be attenuated is modulation of factors Dthe case with synthetic membranes, neither the water norand H, which act as an up-regulator and a down-regula-solute permeability properties of a cellulosic HD mem-tor, respectively, in the complement cascade [22, 23].

Although this mechanism is most relevant for synthetic brane are defined by its specific polymeric composition.

Clark et al: HD membranes2008

BASIC CHARACTERISTICS OF polyethersulfone), a hydrophilic additive [polyvinylpyr-rolidone (PVP)] acts as a polymer alloy. PVP is used toSYNTHETIC MEMBRANESimpart sufficient hydrophilicity to the membrane to allowThe monomeric subunits of the various synthetic mem-clinical use and, as a wetting agent, modulates surfacebranes individually vary, and all differ significantly fromtension and viscosity within the pore structure duringcellobiose. The absence of surface hydroxyl groups onmembrane formulation. This latter feature explainssynthetic membranes is one factor responsible for thePVP’s importance in determining the overall pore sizereported differences in complement activation betweendistribution of synthetic membranes.synthetic membranes and either unsubstituted cellulosic

Although synthetic membranes are employed for bothmembranes or modified cellulosic membranes of low per-hemofiltration and high-flux HD, it is in the latter modemeability. Subsequent to the introduction of the AN69tthat these membranes have found their widest applica-membrane in the early 1970s, numerous synthetic mem-tion. Another synthetic membrane formulation was re-branes have been introduced for clinical use. Similar toported in the late 1980s with the introduction of low-AN69t, polysulfone and polyamide were brought to theflux versions. Low-flux polysulfone [38] and PMMA [12]market for use in both high-flux HD and hemofiltrationhave been used clinically for several years now, and[36, 37]. One obvious reason accounting for the use ofrecently, a low-flux version of a polyamide/polyethersul-these membranes in a hemofiltration mode is their sig-fone copolymer has been introduced (abstract; Kaiser etnificantly larger pore size and higher hydraulic perme-al, Blood Purif 16:236, 1998). Overall, the percentage ofability than regenerated cellulose membranes. The otherthe American HD market employing polysulfone mem-reason relates to the structural differences between thebranes has increased from approximately 10% in thesynthetic and unsubstituted cellulosic membrane groups.late 1980s to approximately 50% currently. However, inCellulosic membranes have relatively thin walls (gener-recent years in the United States, the rate of growth inally in the 6 to 15 mm range), which have a uniform (sym-the use of polysulfone has been more rapid for the low-metric) composition across their entire thickness. Althoughflux version than the high-flux version.the relative thinness of cellulosic membranes is desirable

with respect to diffusive solute transport (discussed laterin this article), this same characteristic renders many EFFECT OF MEMBRANE COMPOSITION

AND STRUCTURE ON DIALYTICcellulosic membranes unable to withstand the high ultra-SOLUTE REMOVALfiltration rates required to perform convective therapies.

The synthetic membranes have thicker walls (20 mm or One classification scheme for uremic solutes is as fol-more) that may be structurally symmetric [for example, lows: small solutes (molecular wt , 200 Da), most ofAN69t, polymethylmethacrylate (PMMA)] or asymmet- which are nitrogenous in nature, middle molecules (mo-ric (for example, polysulfone, polyamide). In the latter lecular wt 500 to 2000 Da), and low molecular wt oligo-category, a very thin “skin” (approximately 1 mm) con- peptides and proteins (molecular wt 2000 to 50,000 Da).tacting the blood compartment lumen acts primarily as

Small solute removalthe membrane’s separative element with regard to soluteremoval, whereas the remaining thickness (stroma) im- Small solute removal during HD occurs almost exclu-parts mechanical strength. In turn, the composition of sively by diffusion. To quantify a particular membrane’sthe stroma layer is quite variable for the various synthetic diffusive capabilities, its mass transfer resistance is fre-membranes [37]. For the Fresenius polysulfone membrane quently used [4]:(Fresenius Medical Care, Bad Homburg, Germany), the

RO 5 RB 1 RM 1 RD (Eq. 3)stroma is relatively homogeneous with a sponge-like struc-ture, whereas the Gambro polyamide membrane (Gam- In equation 3, the overall resistance to diffusive massbro Renal Care, Hechingen, Germany) has, adjacent to transfer of a particular solute (RO) has three components:the skin, a sponge-like stroma layer that has progres- blood compartment resistance (RB), resistance causedsively larger pores (“macrovoids” with a finger structure) by the membrane itself (RM), and dialysate compartmentin the radially outward direction. Finally, a new synthetic resistance (RD). Minimizing the mass transfer resistance(polyethersulfone) membrane developed by Membrana in the blood compartment primarily requires the use ofGmbH (formerly Akzo Nobel) (Wuppertal, Germany) relatively high flow rates (that is, shear rates) that de-has a novel configuration consisting of a sponge-like crease unstirred layers. Dialysate-side mass transfer re-stroma layer interposed between skin layers on both sistance is likewise decreased by increasing the flow rate,the inner (blood-side) and outer (dialysate-side) aspects but optimal dialysate perfusion of fiber bundles is also(abstract; Kayser et al, J Am Soc Nephrol 8:163A, 1997). a consideration. Although increasing the dialysate flow

In the production of synthetic membranes made of pri- rate may itself improve fiber bundle perfusion (discussedlater in this article), another mechanism by which thismarily hydrophobic polymers (polysulfone, polyamide,

Clark et al: HD membranes 2009

can be achieved is the inclusion of spacer yarns. These ever, because of its extensive binding to plasma proteins,devices are spacing filaments placed externally to the this compound is not useful in vivo. In fact, the removalfibers and are designed to facilitate dialysate distribution of uremic solutes having molecular weights that fall inand to reduce channeling [39]. The resistance related to the classic middle molecule category has been difficultthe membrane itself actually has two components: to quantify because of the lack of an easily measured in

vivo surrogate molecule. Because recent evidence sug-RM 5 XM/DM (Eq. 4)gests that uremic appetite suppression is mediated by

where XM is the effective diffusion path length for a the retention of a solute(s) in this size range [42], ansolute, and DM is the solute-specific membrane diffusiv- understanding of removal mechanisms for middle mole-ity. This equation indicates that a decrease in membrane cules is important. Based on dialysis practices used in theresistance can be achieved either by a decrease in mem- 1960s and early 1970s (that is, relatively low flow rates andbrane thickness or an increase in membrane diffusivity. thick, low permeability cellulosic membranes), diffusive

A recent study published by Murthy et al clearly illus- middle molecule removal was so limited that any convec-trates the concept that a membrane’s small solute re- tive removal contributed relatively substantially to totalmoval capabilities are not necessarily tied to its water removal [41]. However, the situation is vastly differentpermeability [35]. Six chronic HD patients were treated in contemporary HD, in which higher flow rates andwith a large surface area unsubstituted cellulosic dialyzer dialyzer membranes of significantly greater diffusive per-(2.2 m2; KUF, 6 ml/hr/mm Hg) and a high-flux polysulfone meability for middle molecules are employed.dialyzer (1.8 m2; KUF, 52 ml/hr/mm Hg) in a crossover Agarwal and Cronin assessed middle molecule re-manner. The mean first-use urea clearance was higher, moval by two high-efficiency dialyzers for seven patientsalthough not significantly, for the cellulosic dialyzer at studied in a crossover manner [43]. They studied genta-a blood flow rate of both 300 ml/min (246 6 6 vs. 241 6 micin, a reasonable middle molecule surrogate in light2 ml/min) and 400 ml/min (288 6 8 vs. 280 6 4 ml/min). of its appropriate molecular weight (518 Da), minimalThis equivalence was achieved despite the substantially protein binding, and small volume of distribution. Agreater pore size and higher water flux of the polysulfone variety of gentamicin removal parameters was determinedmembrane. Although the greater surface area of the cellu- for an unsubstituted cellulosic dialyzer (1.75 m2; KUF, 6.0losic dialyzer contributed to this equivalence, the same ml/hr/mm Hg; thickness, 9 mm) and polysulfone (1.8type of cellulosic dialyzer with a comparable surface area m2; KUF, 8.1 ml/hr/mm Hg; thickness, 40 mm). For both(1.7 m2) has an in vitro urea mass transfer coefficient

dialyzers, an essentially linear relationship between(permeability-area product: KoA) that is substantially

clearance and blood flow rate was observed, suggestinggreater than that of the 1.8 m2 polysulfone dialyzer stud-diffusion was the predominant removal mechanism. Atied (1030 and 945 ml/min, respectively) [10]. These find-blood and dialysate flow rates of approximately 275 andings suggest the significantly lower thickness (9 mm) of500 ml/min, respectively, gentamicin clearance (58.2 6the cellulosic dialyzer relative to that of the polysulfone8.0 vs. 41.7 6 6.9 ml/min) and gentamicin KoA (127.3 6dialyzer (40 mm) was an important factor in this study.6.3 vs. 84.6 6 4.7 ml/min) were significantly greater forAdditional recent data emphasizing the importancethe cellulosic dialyzer. Finally, for the cellulosic dialyzer,of membrane thickness in small solute removal werethese investigations produced a significant linear rela-published by Leypoldt et al [40], who determined in vitrotionship between urea Kt/V and gentamicin Kt/V (r 5urea KoA values for highly permeable cellulose triacetate0.87, P 5 0.01). Collectively, these data demonstrate(1.9 m2; KUF, 36 ml/hr/mm Hg) and polysulfone (1.8 m2;middle molecule removal by the more efficient dialyzersKUF, 55 ml/hr/mm Hg). For blood flow rates rangingused in contemporary HD is primarily achieved by diffu-between 300 and 450 ml/min and a dialysate flow ratesion. As such, the use of treatment parameters that favorof 500 ml/min, the mean urea KoA value for the cellulosediffusive solute removal, such as high flow rates andtriacetate dialyzer (1070 ml/min) was approximately 43%thin membranes, is expected to produce a concomitanthigher than that of the polysulfone dialyzer (750 ml/min).increase in middle molecule removal.An increase in the dialysate flow rate to 800 ml/min

Thalhammer et al recently measured dialytic removalresulted in a 15% increase in urea KoA for both dialyz-parameters for another drug, ofloxacin, whose molecularers, preserving the difference between the dialyzers onweight (361 Da) also falls in the middle molecule rangea percentage basis. For these dialyzers of comparable[44]. In a total of 13 patients studied in a noncrossoverwater flux, the probable explanation for this differencemanner, ofloxacin clearance and percentage reductionwas, again, the large difference in membrane thicknesswere measured for polysulfone (1.3 m2; KUF, 8.5 ml/hr/(15 vs. 40 mm, cellulose triacetate vs. polysulfone).mm Hg) and cellulose acetate dialyzers (1.5 m2; KUF, 6.0

Middle molecule removal ml/hr/mm Hg). (The measurements were made duringa period in which the drug was administered 100 mg dailyVitamin B12 (molecular weight, 1350 Da) is commonly

used for in vitro characterizations of dialyzers [41]. How- by mouth.) Mean treatment characteristics included a

Clark et al: HD membranes2010

blood flow rate of approximately 250 ml/min and dialy- These recent data demonstrate that, based on the op-erating conditions and dialyzers used in contemporarysate flow rate of 500 ml/min. Arteriovenous clearance

(108.4 6 15.8 vs. 92.4 6 39.4 ml/min; polysulfone vs. HD, middle molecule removal occurs predominantly bydiffusion. However, middle molecule removal can becellulose acetate) and percentage reduction normalized to

membrane surface area (49.6 6 5.8 vs. 45.5 6 4.8%/m2) did significantly enhanced by convective therapies, such ashemofiltration and hemodiafiltration, in which an abso-not differ significantly. However, arteriovenous clearance

normalized to the membrane surface area was signifi- lute ultrafiltration rate in excess of that required forplasma volume reduction is employed [49–51]. In gen-cantly higher for the polysulfone dialyzer (5.0 6 0.7 vs.

3.7 6 1.6 l/hr/m2, P , 0.05). eral, the degree to which convection augments total sol-ute removal is proportional to solute molecular weightWe have recently reported findings for middle mole-

cule removal by high-flux membranes [45]. The molecu- due to the inverse relationship between solute diffusivityand size [52].lar weight (1448 Da), protein binding, and volume of

distribution of vancomycin, another antibiotic commonlyLow molecular weight protein removalused in the acute renal failure (ARF) and end-stage renal

disease (ESRD) populations, enable this drug also to be a Recent interest in increasing the extracorporeal re-moval of b2m has provided insight into the general mech-uremic middle molecule surrogate [46, 47]. We measured

vancomycin removal by a cellulose triacetate dialyzer anisms mediating the removal of low molecular weightproteins. A number of studies published in the past 15(1.9 m2; KUF, 36 ml/hr/mm Hg) in eight patients who

received 1 g of vancomycin intravenously during the last years [11, 52–63] support several general conclusions.First, b2m removal by low-flux unsubstituted cellulosichour of dialysis. The mean blood and dialysate flow rates

were 423 and 600 ml/min, respectively. A significant lin- membranes is usually negligible, although certain excep-tions do exist [35]. Second, the primary mechanism byear correlation (r 5 0.88, P , 0.005) was observed be-

tween the percentage removal of vancomycin and a mea- which b2m is removed during high-flux HD varies widelyamong membranes. For certain membranes, such as sul-sure of diffusive small solute removal, urea Kt/V. In

addition, diffusion was determined to account minimally fonated PAN (AN69t) and particularly PMMA, removalis achieved predominantly or solely by adsorption. Atfor a mean of 92% of total vancomycin removal. In a

subsequent in vitro study [48], we assessed the mass the other end of the spectrum is the cellulose triacetatemembrane, for which adsorption is minimal and removaltransfer of vancomycin and another widely studied large

molecule surrogate, inulin (molecular weight, 5200 Da), occurs primarily by diffusion. High-flux polysulfone andunsulfonated PAN membranes have intermediate ad-for high-flux cellulose triacetate (1.9 m2; KUF, 36 ml/hr/

mm Hg) and polysulfone (1.8 m2; KUF, 52 ml/hr/mm Hg) sorptive characteristics and achieve transmembrane b2mremoval by a combination of diffusion and convection.dialyzers. This study corroborated the above clinical data

for vancomycin and also suggested that diffusion is the Third, at least for the high-flux synthetic membranes,enhanced convective removal by the use of high ultrafil-predominant mass transfer mechanism, even for solutes

as large as inulin, for both high-flux dialyzers. The mean tration rates (with hemofiltration or hemodiafiltration)increases b2m removal relative to standard (diffusion-first-use inulin clearance was found to be significantly

higher for cellulose triacetate versus polysulfone (76 vs. based) HD. Although many clinicians consider b2m tobe a surrogate for the low molecular weight protein class43 ml/min, respectively, P , 0.05). The lower diffusive

mass transfer resistance of the thinner cellulose triacetate of uremic solutes, this assumption has not been conclu-sively proved. Nevertheless, it is reasonable to use themembrane may account for these findings. However,

another possible explanation may relate to the pore size abundant transport data available for b2m to provideinsight into the transport characteristics of other lowdimensions of the polysulfone dialyzer specifically used

in the study. This dialyzer, which is designed for units molecular weight proteins, such as complement activa-tion products and cytokines.employing bleach-based reprocessing, has a relatively

low first-use b2m clearance, which increases significantly We recently measured the protein transport propertiesof high-flux membranes. In an initial set of studies [64],when it is reprocessed with bleach [13]. In our study

[48], inulin clearance for the polysulfone dialyzer also fragments of AN69t membranes were incubated in sa-line containing 125I-labeled b2m. Adsorption isothermsincreased significantly during bleach reprocessing, such

that no significant difference was observed between the (amount of protein adsorbed vs. concentration in solu-tion at equilibrium) [65] were generated for both poroustwo dialyzers after 10 in vitro reuses. Thus, a difference

in mean pore size may also explain the first-use differ- (commercially available) AN69t and a nonporous prep-aration of the same polymer. At equivalent solution con-ence for inulin clearance in our study. Of note, relative

to the polysulfone dialyzers used in this study, recent centrations, the b2m binding affinity of the porous PANwas approximately eight times greater than that of thedata [15] indicate the first-use b2m clearance of more

recently manufactured dialyzers is higher. nonporous PAN. These data indicate that the adsorptive

Clark et al: HD membranes 2011

Fig. 2. Generation of C3a during incubation of serum with cuprophanor PAN membranes. Symbols (h cuprophan; PAN) depict themean 6 sem (N 5 5). *The difference between cuprophan and PANis statistically significant (P , 0.001). (Reprinted with permission fromFig. 1. Isotherm curves for b2-microglobulin (b2m) adsorption to poly-Cheung et al [67] and the International Society of Nephrology.)acrylonitrile (PAN) and cellulose triacetate (CT). The relationship is

linear for both membranes. The slope of the PAN line is approximatelyfivefold greater than that of the CT line. (Reprinted with permissionfrom Clark et al and the International Society of Nephrology [66].)

INTERACTION BETWEEN BIOCOMPATIBILITYAND FLUX

Measurement of complement pathway byproducts issurface area resides primarily in the pore structure of aone technique used to assess the inflammatory responsehigh-flux HD membrane rather than the nominal surfaceelicited by exposure of blood to a dialysis membrane.

area. As such, the adsorption of a low molecular weightHowever, numerous previous studies have failed to ac-

protein is highly dependent on access of the protein count for the fact that the clinically measured comple-to a membrane’s internal pore structure. Consequently, ment components (C3a and C5a) are low molecularadsorption of low-molecular weight proteins, such as C3a weight proteins. Therefore, the concentration of theseand b2m, to low-flux membranes is not expected to be inflammatory mediators represents the net result of theclinically significant, at least in comparison to that which simultaneous processes of generation and any dialyticoccurs to high-flux membranes. removal that may occur. In this regard, complement acti-

We also compared protein removal mechanisms for vation products are similar to most uremic solutes, fordifferent high-flux membranes [66]. Based on the slopes which both generation and net removal need to be con-of the equilibrium isotherm curves, we found that the sidered [68]. The corollary of this observation is thatadsorption affinity of AN69t for b2m was approximately the permeability properties and not just the polymericfive times that of the cellulose triacetate membrane (Fig. composition of a dialysis membrane must be considered1). This finding was attributed to the greater hydropho- when evaluating complement activation data.bicity of the synthetic membrane. However, the b2m dif- The effect of simultaneous generation and removal onfusive mass transfer resistance of the AN69 membrane net complement activation was investigated in a classicwas found to be approximately 1.7-fold greater than that study published by Cheung et al [67]. Dialysis membraneof cellulose triacetate, a result that translated into an fragments composed of either (low-flux) regenerated cel-order of magnitude difference in the b2m membrane lulose (Cuprophant) or (high-flux) PAN (AN69t) werediffusion coefficients (3.25 3 1027 vs. 0.30 3 1027 cm2/ incubated in plasma containing C3 molecules previouslysec; cellulose triacetate vs. AN69t). Thus, in terms of b2m radiolabeled with 125I. After a 30-minute incubation pe-(and presumably other low molecular weight protein) riod, the generation of C3a (molecular weight, 9.5 kDa)removal, our findings corroborate previous experimental resulting from contact of the plasma with the membranesand clinical data demonstrating the high adsorptive af- was measured. For each membrane, total C3a generationfinity of the sulfonated PAN membrane [19, 67]. How- was determined by summing the amount in the supernateever, our data also suggest that this membrane is not (fluid phase) and the amount adsorbed to the membrane.very well suited for diffusive removal of uremic solutes The results (Fig. 2) indicate that the total C3a generationin this class, at least in comparison to high-flux cellulosic was actually greatest for the “biocompatible” AN69t.

However, nearly all of the C3a generated was rapidlymembranes.

Clark et al: HD membranes2012

Table 1. Inflammatory mediators both low-flux and high-flux polysulfone dialyzers [80].Peak plasma C3a concentrations were approximatelyMolecular

weight twofold higher with the low-flux than the high-flux mem-Mediator kD brane. In addition, the area under the C3a versus timeLipid A 2–4 curve, which is a time-integrated measure of complementLipopolysaccharide (LPS) fragments ,8

activation, appeared to be significantly higher for the low-C3a 8.9Granulocyte inhibitory peptide (GIP) II 9.5 flux dialyzer. A similar difference in the extent of comple-C5a 11 ment activation between low-flux and high-flux polysul-Interleukin-1 17

fone dialyzers has also been reported by Dumler et alTumor necrosis factor (monomeric) 17Factor D 23 (abstract; Dumler et al, Blood Purif 10:91–92, 1992). TheseGranulocyte inhibitory peptide (GIP) I 28 findings are most likely explained by the ability of theTumor necrosis factor (trimeric) 55

high-flux polysulfone membrane to remove a greater pro-Lipopolysaccharide (LPS) .100portion of the generated complement component, relativeReprinted [71] with permission.to the low-flux membrane. The larger mean pore size ofthe high-flux membrane allows greater pore access of thegenerated C3a and subsequent removal by a transmem-brane (diffusive or convective) or adsorptive mechanism.adsorbed to this hydrophobic membrane. Total C3a gen-

eration was significantly lower for the regenerated cellu- Similar reasoning can be applied to a recent study byHoenich et al [34]. These investigators measured serumlose membrane, but virtually all of the protein remained

in the fluid phase rather than being bound to this hydro- C3a concentrations in a group of patients treated withtwo cellulose acetate membranes of differing pore sizephilic surface. In extrapolating these data to clinical HD,

a reasonable assumption is that the fluid phase of these and a low-flux polysulfone membrane in a double cross-over design. The two cellulose acetate membranes hadexperiments corresponds to the venous (dialyzer efflu-

ent) blood line in the extracorporeal circuit. Therefore, reported mean maximum pore radii of 43 and 45 A8.Despite the similarity in membrane composition, theonly that portion of the generated C3a remaining in the

fluid phase would reach the bloodstream of the patient peak C3a concentration observed for the larger porecellulose acetate membrane was significantly less thanto act as a potential inflammatory mediator. These data

are useful in interpreting clinical studies in which com- that of its smaller pore counterpart and was not signifi-cantly different from that of low-flux polysulfone. Differ-plement protein concentrations are reported and under-

score the need to understand the permeability and solute ences in mean pore size or possibly in pore size distributionwith resultant differences in C3a removal capabilitiesremoval properties of a membrane when evaluating com-

plement activation data. may explain the differences observed for the two cellu-lose acetate membranes.It is simplistic to limit the discussion about membrane

biocompatibility to complement activation, as a number Deppisch et al have shown that the generation of an-other inflammatory marker, the terminal complementof agents have been identified as potential inflammatory

mediators in chronic HD patients. A list of these putative complex (TCC), is higher for low-flux than for high-fluxpolysulfone [81]. Because of its large molecular size,mediators appears in Table 1 [69, 70] Some of these com-

pounds, such as lipid A and lipopolysacchride fragments, removal of this mediator is not expected to be significanteven by a high-flux HD membrane. Therefore, a mecha-potentially have their origin in dialysate, a nonsterile

fluid [71–73]. Because of their relatively low molecular nism other than enhanced removal of TCC itself, possiblyone related to complement factor D, most likely accountsweight, these inflammatory mediators may undergo trans-

membrane passage and induce cytokine production in for this finding. Factor D is another example of an in-flammatory mediator that is differentially removed bythe bloodstream, either directly via an effect on mononu-

clear cells or indirectly via an effect on the alternative low-flux and high-flux HD membranes. This compound isa 23 kDa serine protease whose excretion under normalcomplement pathway [74–78]. Conversely, the majority

of the mediators that are potentially elicited in the blood, circumstances is similar to that of b2m: glomerular filtra-tion followed by proximal tubule reabsorption/catabo-such as C3a and interleukin-1, may be simultaneously

eliminated during high-flux therapies by an adsorptive lism [17, 82]. Therefore, in patients with ESRD, serumconcentrations are markedly elevated. This protein actsor transmembrane mechanism, as demonstrated in the

Cheung et al study [67]. Other investigations have con- as an up-regulator of the alternative complement path-way [83], and activation of the alternative pathway byfirmed that adsorption is also important in the removal

of other inflammatory mediators, such as factor D [19] blood–membrane interaction (with resultant C3a gener-ation) is enhanced in the presence of the high serumand cytokines [79].

Recent clinical data illustrate these concepts. Muller concentrations of factor D found in uremic patients. Re-cent data demonstrate that factor D removal is negligibleet al measured serial C3a concentrations during HD with

Clark et al: HD membranes 2013

sis of the Lundia (PRO 5) and Lundia (PRO 3) dialyzers withby low-flux HD but can be significant by hemofiltration,Gambrane polycarbonate membranes. Blood Purif 4:63–73, 1986

such that the latter can achieve a significant reduction 10. Van Stone J: Hemodialysis apparatus, in Handbook of Dialysis(2nd ed), edited by Daugirdas J, Ing T, Boston, Little, Brown,in serum factor D concentrations over time [84]. Prelimi-1994, pp 30–52nary data (abstract; Dupuy et al, Nephrol Dial Transplant

11. Kaiser J, Hagemann J, Von Herrath D, Schaefer K: Different8:986, 1993) also indicate that significant factor D re- handling of beta2-microglobulin during hemodialysis and hemofil-

tration. Nephron 48:132–135, 1988moval can be achieved with high-flux HD. Accordingly,12. Ward R, Buscaroli A, Schmidt B, Stefoni S, Gurland H, Klink-the available data suggest that high-flux therapies favor-

mann H: A comparison of dialysers with low-flux membranes:ably influence serum C3a profiles (relative to low-flux Significant differences in spite of many similarities. Nephrol Dial

Transplant 12:965–972, 1997HD) by at least two mechanisms. One mechanism is a13. Cheung A, Agodoa L, Daugirdas J, Depner T, Gotch F, Greenedampening of the alternative complement pathway by

T, Levin N, Leypoldt JK: Effects of hemodialyzer reuse on clear-enhanced factor D removal, whereas the other is removal ances of urea and b2-microglobulin. J Am Soc Nephrol 10:117–127,

1999of a significant proportion of generated C3a.14. Keshaviah P, Luehmann D, Ilstrup K, Collins A: Technical

requirements for rapid high efficiency therapies. Artif Organs110:189–194, 1986SUMMARY

15. Walton D, Cheung A: Membrane biocompatibility, in ClinicalDialysis (3rd ed), edited by Nissenson A, Fine R, Gentile D,Dialyzers used in contemporary HD are equipped withNorwalk, CT, Appleton and Lange, 1995, pp 93–120a wide variety of membranes, and within both the cellu-

16. Grooteman M, Nube M, van Limbeek J, van Houte A, Daha M,losic and synthetic classes, water and solute flux proper- van Geelen J: Biocompatibility and performance of a modified

cellulosic and a synthetic high-flux dialyzer. ASAIO J 41:215–220,ties vary widely. Although a direct correlation between1995water and solute permeability exists for some membranes,

17. Grooteman M, Nube M, Daha M, Van Limbeek J, Van Deurenthese two characteristics can be dissociated for others. M, Schoorl M, Bet P, Van Houte A: Cytokine profiles during

clinical high-flux dialysis: No evidence for cytokine generation byFor small- and middle-sized solutes, abundant clinicalcirculating monocytes. J Am Soc Nephrol 8:1745–1754, 1997data point to the importance of membrane thickness in

18. Hoenich NA, Woffindin C, Matthews JNS, Goldfinch ME,diffusive mass transfer. The removal of low molecular Turnbull J: Clinical comparison of high-flux cellulose acetate and

synthetic membranes. Nephrol Dial Transplant 9:60–66, 1994weight proteins may occur largely by adsorption for some19. Schaefer R, Horl W, Kokot K, Heidland A: Enhanced biocom-high-flux membranes, particularly those of hydrophobic

patibility with a new cellulosic membrane: Cuprophan vs hemo-synthetic composition. Because many of the mediators of phan. Blood Purif 5:262–267, 1987

20. Lucchi L, Bonucchi D, Acerbi M: Improved biocompatibilityinflammation in dialysis patients fall in this low molecularby modified cellulose membranes: The case of Hemophan. Artifweight protein category, the biocompatibility of a partic-Organs 13:417–421, 1989

ular membrane must be interpreted in conjunction with 21. Bowry S, Rintelen T: Synthetically modified cellulose: A cellulosichemodialysis membrane with minimized complement activation.its permeability properties.ASAIO J 44:M579–M583, 1998

22. Pascual M, Steiger G, Estreicher J, Macon K, Volankis J,Reprint requests to William R. Clark, M.D., Hemodialysis ResearchLaboratory, Renal Division, Baxter Healthcare Corporation, Wishard Schifferli J: Metabolism of complement factor D in renal failure.Hospital/Myers Building D711, 1001 West 10th Street, Indianapolis, Kidney Int 34:529–536, 1988Indiana 46202, USA. 23. Rauterberg E, Ritz E: Bioincompatibility of dialysis membranes:E-mail: clarkbi@baxter.com Factor H binding correlates inversely with complement activation

indicating a local imbalance of involved proteases/anti-proteases.Adv Exp Med Biol 240:365–375, 1992REFERENCES 24. Pascual M, Schifferli J: Adsorption of complement factor D bypolyacrylonitrile dialysis membranes. Kidney Int 43:903–911, 19931. Patel R, Vertes V, Bloomfield D, Levy M: Improvements in

25. Galli F, Rovidati S, Chiarantini L, Campus G, Canestrari F,use of the twin-coil kidney for chronic dialysis in a large center.Buoncristiani U: Bioreactivity and biocompatibility of a vitaminTrans Am Soc Artif Intern Organs 13:5–9, 1967E-modified multi-layer hemodialysis filter. Kidney Int 54:580–589,2. Pollard TL, Barnet BMS, Eschbach JW, Scribner BW: A tech-1998nique for the storage and multiple re-use of the Kiil dialyzer and

26. Deppisch R, Gohl H, Smeby L: Microdomain structure of poly-blood tubing. Trans Am Soc Artif Intern Organs 13:24–28, 1967meric surfaces: Potential for improving blood treatment proce-3. Tsaltas TT: Comparison of various methods of dialysis: Subjectivedures. Nephrol Dial Transplant 13:1354–1359, 1998experiences and laboratory data. Trans Am Soc Artif Intern Organs

27. Lane D, Bowry S: The scientific basis for selection of measures13:29–32, 1967of thrombogenicity. Nephrol Dial Transplant 9(Suppl 2):18–28,4. Lysaght MJ: Evolution of hemodialysis membranes. Contrib1993Nephrol 113:1–10, 1995

28. Craddock P, Fehr J, Dalmasso A, Brigham K, Jacob H: Hemodi-5. Lipps B, Stewart R, Perkins H, Holmes G, McLain E, Rolfs M,alysis leukopenia: Pulmonary vascular leukostasis resulting fromOja P: The hollow fiber artificial kidney. Trans Am Soc Artif Interncomplement activation by dialyzer cellophane membranes. J ClinOrgans 13:200–207, 1967Invest 59:879–888, 19776. Lysaght MJ: Hemodialysis membranes in transition. Contrib

29. Hakim R, Fearon D, Lazarus JM: Biocompatiblity of dialysisNephrol 61:1–17, 1988membranes: Effects of chronic complement activation. Kidney Int7. Akizawa T, Kinugasa E, Ideura T: Classification of dialysis mem-26:194–200, 1984branes by performance. Contrib Nephrol 113:25–31, 1995

30. Hoenich N, Woffindin C, Ronco C: Haemodialysers and associ-8. Henderson L: Biophysics of ultrafiltration and hemofiltration, inated devices, in Replacement of Renal Function by Dialysis (4thReplacement of Renal Function by Dialysis (4th ed), edited byed), edited by Jacobs C, Kjellstrand C, Koch K, Winchester J,Jacobs C, Kjellstrand C, Koch K, Winchester J, Dordrecht,Dordrecht, Kluwer Academic, 1996, pp 188–230Kluwer Academic Publishers, 1996, pp 114–145

9. Miller J, Shinaberger J, Henderson L, Gardner P: In vivo analy- 31. Jorstad S, Smeby L, Balstad T, Wideroe T: Removal, generation,

Clark et al: HD membranes2014

and adsorption of beta-2-microglobulin during hemofiltration with beta2-microglobulin kinetics: In vivo and in vitro studies. NephrolDial Transplant 3:284–290, 1988five different membranes. Blood Purif 6:96–105, 1988

54. Kaiser J, Hagemann J, Von Herrath D, Schaefer K: Different32. Ingram A, Parbtani A, Churchill D: Effects of two low-fluxhandling of beta2-microglobulin during hemodialysis and hemofil-cellulose acetate dialysers on plasma lipids and lipoproteins: Atration. Nephron 48:132–135, 1988cross-over trial. Nephrol Dial Transplant 13:1452–1457, 1998

55. Goldman M, Lamiche M, Dhaene M, Amraoui Z, Thayse C,33. Lonneman G, Schindler R, Luftt V, Mahiout A, Shaldon S,Vanherweghem J: Adsorption of beta2-microglobulin on dialysisKoch KM: The role of plasma coating on the permeation of cyto-membranes: Comparison of different dialyzers and effects of reuse.kine-inducing substances through dialyser membranes. NephrolInt J Artif Organs 12:373–378, 1989Dial Transplant 10:207–211, 1995

56. Klinke B, Rockel A, Abdelhamid S, Fiegel P, Walb D: Trans-34. Hoenich N, Woffindin C, Cox P, Goldfinch M, Roberts S: Clini-membrane transport and adsorption of beta2-microglobulin dur-cal characterization of Dicea: A new cellulose membrane foring hemodialysis using polysulfone, polyacrylonitrile, polymethy-haemodialysis. Clin Nephrol 48:253–259, 1997lmethacrylate, and cuprammonium rayon membranes. Int J Artif35. Murthy B, Sundaram S, Jaber B, Perrella C, Meyer K, PereiraOrgans 12:697–702, 1989B: Effect of formaldehyde/bleach reprocessing on in vivo perfor-

57. Floege J, Granolleras C, Bingel M, Deschodt G, Branger B,mances of high-efficiency cellulose and high-flux polysulfone dia-Oules R, Shaldon S, Koch K: Beta2-microglobulin kinetics duringlyzers. J Am Soc Nephrol 9:464–472, 1998hemodialysis and hemofiltration. Nephrol Dial Transplant 1:223–36. Rockel A, Hertel J, Fiegel P, Abdelhamid S, Panitz N, Walb228, 1987D: Permeability and secondary membrane formation of a high flux

58. Floege J, Granolleras C, Deschodt G, Heck M, Baudin G,polysulfone hemofilter. Kidney Int 30:429–432, 1986Branger B, Tournier O, Reinhard B, Eisenbach G, Smeby L,37. Gohl H, Buck R, Strathmann H: Basic features of polyamideKoch K, Shaldon S: High-flux synthetic vs cellulosic membranesmembranes. Contrib Nephrol 96:1–25, 1992for beta2-microglobulin removal during hemodialysis, hemodiafil-38. Piazolo P, Brech W, Niedermayer W, Albrecht J, Hennemanntration, and hemofiltration. Nephrol Dial Transplant 4:653–657,H, Witter E: Clinical multicenter study of Hemoflow F6 in compar-1989ison with different standard dialyzers. Contrib Nephrol 74:22–33,

59. Jorstad S, Smeby L, Balstad T, Wideroe J: Removal, generation,1989and adsorption of beta2-microglobulin during hemofiltration with39. Ronco C, Scabardi M, Goldoni M, Brendolan A, Crepaldi C,five different membranes. Blood Purif 6:96–105, 1988LaGreca G: Impact of spacing filaments external to hollow fibers

60. Floege J, Wilks M, Shaldon S, Koch K, Smeby L: Beta2-micro-on dialysate flow distribution and dialyzer performance. Int J Artifglobulin kinetics during hemofiltration. Nephrol Dial TransplantOrgans 20:261–266, 19973:784–789, 198840. Leypoldt JK, Cheung A, Agodoa L, Daugirdas J, Greene T,

61. Naitoh A, Tatsuguchi T, Okada M, Ohmura T, Sakai K: RemovalKeshaviah P: Hemodialyzer mass transfer-area coefficients forof beta2-microglobulin by diffusion alone is feasible using highlyurea increase at high dialysate flow rates. Kidney Int 51:2013–2017,permeable dialysis membranes. Trans Am Soc Artif Intern Organs199734:630–634, 198841. Nolph K, Nothum R, Maher J: Ultrafiltration: A mechanism for 62. Mineshama M, Hoshino T, Era K, Kitano Y, Suzuki T, Sanakaremoval of intermediate molecular weight substances in coil dialyz- T, Teraoka S, Ahishi T, Ota K: Difference in beta2-microglobulin

ers. Kidney Int 6:55–60, 1974 removal between cellulosic and synthetic polymer membrane dia-42. Anderstam B, Mamoun A, Sodersten P, Bergstrom J: Middle- lyzers. Trans Am Soc Artif Intern Organs 36:M643–M646, 1990

sized molecule fractions isolated from uremic ultrafiltrate and nor- 63. Ronco C, Heifetz A, Fox K, Curtin C, Brendolan A, Gastaldonmal urine inhibit ingestive behavior in the rat. J Am Soc Nephrol F, Crepaldi C, Fortunato A, Piertibasi G, Caberlotto A, Bru-7:2453–2460, 1996 nello A, Milan Manani S, Zanella M, La Greca G: Beta2-

43. Agarwal R, Cronin R: Heterogeneity of gentamicin clearance microglobulin removal by synthetic dialysis membranes: Mecha-between high-efficiency hemodialyzers. Am J Kidney Dis 23:47–51, nisms and kinetics of the molecule. Int J Artif Organs 20:136–143,1994 1997

44. Thalhammer F, Kletzmayr J, El Menyawi I, Kovarik J, Rosen- 64. Clark WR, Macias WL, Molitoris BA, Wang NHL: Membranekranz A, Traunmuller F, Horl W, Burgmann H: Ofloxacin adsorption of beta2-microglobulin: Equilibrium and kinetic charac-clearance during hemodialysis: A comparison of polysulfone and terization. Kidney Int 46:1140–1146, 1994cellulose acetate hemodialyzers. Am J Kidney Dis 32:642–645, 1998 65. Andrade J: Principles of protein adsorption, in Surface and Interfa-

45. Scott M, Mueller B, Clark W: Vancomycin mass transfer charac- cial Aspects of Biomedical Polymers (vol 2, Protein Adsorption),teristics of high-flux cellulosic dialysers. Nephrol Dial Transplant edited by Andrade J, New York, Plenum Press, 198512:2647–2653, 1997 66. Clark WR, Macias WL, Molitoris A, Wang NHL: Plasma pro-

46. Tan C, Lee H, Ti T, Lee E: Pharmacokinetics of intravenous tein adsorption to highly permeable hemodialysis membranes. Kid-vancomycin in patients with end-stage renal disease. Ther Drug ney Int 48:481–488, 1995Monit 12:29–34, 1990 67. Cheung AK, Parker C, Wilcox L, Janatova J: Activation of

47. Rodvold K, Blum R, Fischer J: Vancomycin pharmacokinetics in complement by hemodialysis membranes: Polyacrylonitrile bindspatients with various degrees of renal function. Antimicrob Agents more C3a than cuprophan. Kidney Int 37:1055–1059, 1990Chemother 32:848–852, 1988 68. Gotch F: Kinetic modeling in hemodialysis, in Clinical Dialysis

48. Scott M, Mueller B, Clark W: Dialyzer dependent changes in (3rd ed), edited by Nissenson A, Fine R, Gentile D, Norwalk,solute and water permeability with bleach reprocessing. Am J Appleton and Lange, 1995, pp 156–158Kidney Dis 33:87–96, 1999 69. Cohen G, Haag-Weber M, Horl W: Immune dysfunction in ure-

49. Colton C, Henderson L, Ford C, Lysaght M: Kinetics of hemodi- mia. Kidney Int 52(Suppl 62):S79–S82, 1997afiltration. I. In vitro transport characteristics of a hollow fiber 70. Haag-Weber M, Mai B, Cohen G, Horl W: Isolation of a granulo-blood ultrafilter. J Lab Clin Med 85:355–371, 1975 cyte inhibitory protein from uraemic patients with homology of

50. Henderson L, Colton C, Ford C: Kinetics of hemodiafiltration. beta2-microglobulin. Nephrol Dial Transplant 9:382–388, 1994II. Clinical characterization of a new blood cleansing modality. 71. Lonneman G: Dialysate bacteriological quality and the permeabil-J Lab Clin Med 85:372–391, 1975 ity of dialyzer membranes to pyrogens. Kidney Int 43(Suppl

51. Ledebo I: Principles and practice of hemofiltration and hemodiafil- 41):S195–S200, 1993tration. Artif Organs 22:20–25, 1998 72. Favero M, Petersen N, Boyer K, Carson L, Bond W: Microbial

52. Jindal KK, McDougall J, Woods B, Nowakowski L, Goldstein contamination of renal dialysis systems and associated health risks.MB: A study of the basic principles determining the performance Trans ASAIO 20:175–183, 1974of several high-flux dialyzers. Am J Kidney Dis 14:507–511, 1989 73. Harding G, Klein E, Pass T, Wright R, Million C: Endotoxin

53. Zingraff J, Beyne P, Urena P, Uzan M, Nguyen K, Descamps- and bacterial contamination of dialysis center water and dialysate:A cross-sectional survey. Int J Artif Organs 13:39–43, 1990Latscha B, Drueke T: Influence of haemodialysis membranes on

Clark et al: HD membranes 2015

74. Urena P, Herbelin A, Zingraff J, Lair M, Man NK, Descamps- 79. Goldfarb S, Golper T: Proinflammatory cytokines and hemofil-tration membranes. J Am Soc Nephrol 5:228–232, 1994Latscha B, Drueke T: Permeability of cellulosic and non-cellulosic

80. Muller T, Seitz M, Eckle I, Lange H, Kolb G: Biocompatibilitymembranes to endotoxin subunits and cytokine production duringdifferences with respect to the dialyzer sterilization mode. Nephronin vitro hemodialysis. Nephrol Dial Transplant 7:16–28, 199278:139–142, 199875. Lonneman G, Bingel M, Floege J, Koch K, Shaldon S, Dina- 81. Deppisch R, Schmitt V, Bommer J, Hansch G, Ritz E, Rauter-

rello C: Detection of endotoxin-like interleukin-1 inducing activ- berg E: Fluid phase generation of terminal complement complexity during in vitro dialysis. Kidney Int 33:29–35, 1988 as a novel index of biocompatibility. Kidney Int 37:696–706, 1990

76. Schindler R, Lonnemann G, Shaldon S, Koch KM, Dinarello 82. Volanakis J, Barnum S, Giddens M, Galla J: Renal filtrationand catabolism of complement protein D. N Engl J Med 312:395–C: Transcription, not synthesis, of interleukin-1 and tumor necrosis399, 1985factor by complement. Kidney Int 37:85–93, 1990

83. Deppisch R, Ritz E, Hansch G, Schols M, Rauterberg E: Bio-77. Laude-Sharp M, Caroff M, Simard L, Pusineri C, Kazatchkinecompatibility: Perspectives in 1993. Kidney Int 45(Suppl 44):S77–M, Haeffner-Cavaillon N: Induction of IL-1 during hemodialysis:S84, 1994

Transmembrane passage of intact endotoxins. Kidney Int 38:1089– 84. Kaiser J, Opperman M, Gotze O, Deppisch R, Gohl H, Asmus1094, 1990 G, Rohricht B, von Herrath D, Schaefer K: Significant reduction

78. Pereira B: Cytokine production in patients on dialysis. Blood Purif of factor D and immunosuppressive complement fragment Ba byhaemofiltration. Blood Purif 13:314–321, 199513:135–146, 1995