Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Reviews

Haemodialysis membranes
Claudio Ronco1 and William R. Clark2*
Abstract | Haemodialysis is an extracorporeal process in which the blood is cleansed via removal
of uraemic retention products by a semipermeable membrane. Traditionally, dialysis membranes
have been broadly classified on the basis of their composition (cellulosic or noncellulosic) and
water permeability (low flux or high flux). However, advances in materials technology and polymer
chemistry have led to the development of membranes with specific characteristics and refined
properties that mandate a reconsideration of traditional membrane classification systems. For
adequate characterization of these newer types of membranes, additional parameters are now
relevant, including new permeability indices, the hydrophilic or hydrophobic nature of
membranes, adsorption capacity and electrical potential. In this Review, we provide clinicians
with an updated analysis of dialysis membranes and dialysers. We discuss the basic mechanisms
that underlie solute and water removal in dialysis (that is, diffusion, convection, adsorption and
ultrafiltration) in the context of treatments that use highly permeable membranes. Specifically,
we highlight online haemodiafiltration and new therapies (for example, expanded haemodialysis)
that utilize membranes designed to produce a high degree of internal filtration. Finally, we
discuss the considerations that govern the clinically acceptable balance between large-solute
clearance and albumin loss for extracorporeal therapies.

Haemodialysis is an extracorporeal blood cleansing tech- convection), with particular emphasis on the key part
nique that is used to remove metabolic waste products played by blood–membrane interactions. We pres-
that accumulate in patients with end-stage renal disease ent an updated membrane classification scheme and a
(ESRD). Solutes and water are removed through semi- description of conventional and new parameters that
permeable membranes using different mass separation are used to characterize and quantify membrane solute
mechanisms (diffusion, convection and adsorption). and water transport. Finally, we discuss the utilization
Traditionally, dialysis membranes have been broadly clas- of membranes and dialysers in clinical practice, with an
sified on the basis of their composition (cellulosic or non- emphasis on newer and emerging applications.
cellulosic) and water permeability (low or high flux)1. The
evolution of biomaterials and improved fibre production A history of dialysis membranes
(spinning) technology have resulted in new membranes Although other approaches were previously used on a
with specific characteristics and refined properties that small scale9,10, the rotating drum kidney was the first hae-
mandate a reconsideration of traditional membrane classi- modialysis membrane configuration that was employed
fication schemes2. The incorporation of innovative manu- to treat large numbers of patients11. This device com-
facturing processes, such as polymer blending3 and surface prises a 30 m cellophane tube (inner diameter = 35 mm)
functionalization4, has led to a consideration of several that is wrapped in a spiral manner around a cylinder
1
Department of Nephrology, other parameters for membrane categorization, including that rotates in a stationary dialysate bath (Supplementary
Dialysis and Transplantation,
International Renal Research
new permeability indices5, hydrophilic versus hydrophobic Fig. 1a). Although dialysis could be provided without
Institute of Vicenza (IRRIV), balance6, adsorption capacity7 and electrical potential8. a blood pump owing to the very low resistance of the
San Bortolo Hospital, In this Review, we provide an overview of the state blood compartment, the low transmembrane pressures
Vicenza, Italy. of the art in membranes and dialysers that are used for (TMPs) that were generated by this system severely lim-
2
Davidson School of Chemical chronic haemodialysis treatment and related therapies. ited ultrafiltration capabilities. Another limitation of this
Engineering, Purdue We provide a brief historical perspective and discuss the membrane system was its large extracorporeal blood vol-
University College of
Engineering, West Lafayette,
characteristics of commonly used membranes (includ- umes (500–700 ml). A modified system in which higher
IN, USA. ing the materials technology that underlies membrane blood compartment pressures allowed for adequate
*e-mail: clarkw@purdue.edu composition) and the way in which they are assembled ultrafiltration was later developed12.
https://doi.org/10.1038/ into a dialyser. We highlight the different mechanisms The next device that was widely used, the coil dialyser
s41581-018-0002-x that mediate solute removal (mainly diffusion and (Supplementary Fig. 1a), had technological characteristics

394 | JUNE 2018 | volume 14 www.nature.com/nrneph


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

past decades to the point of effective absence from


Key points
the market. Although the performance and functional
• Traditional schemes for the classification of dialysis membranes, based simply on characteristics of some modified cellulosic membranes
composition and water permeability, are outdated and new approaches are needed. are similar to those of synthetic membranes, the use
• Dialyser utilization in clinical practice has evolved over time and is now dominated by of modified cellulosic membranes continues to fall.
devices with synthetic high-flux membranes. Therefore, the remainder of our discussion focuses
• Rational treatment prescription by clinicians requires an understanding of the basic primarily on synthetic membranes.
mechanisms underlying solute and water removal in dialysis — namely, diffusion, Synthetic membranes were originally developed for a
convection, adsorption and ultrafiltration. dialysis application >40 years ago to address the relative
• New therapies (including expanded haemodialysis) that utilize membranes designed bioincompatibility21–23 and limited permeability19,20 of
to produce a high degree of internal filtration are undergoing clinical evaluation as unmodified cellulosic membranes. Furthermore, although
potential alternatives to convective therapies, such as on-line haemodiafilitration. the structure of early cellulosic membranes (wall thick-
• The clinically acceptable amount of albumin loss for extracorporeal therapies remains ness ≤ 15 µm) facilitated diffusive transport, the inherently
to be defined. low water permeability prevented the application of cellu-
losic membranes in convective therapies. The very thick
that differed substantially from those of the rotating walls of the earliest synthetic membranes (~100 µm) were
drum13. The functional unit of this dialyser, cellophane tub- not conducive to diffusion-based therapies; thus, their
ing surrounded by a fibreglass screen, was wound together use was originally limited to haemofiltration (purely
in a single coil located in a large cylindrical drum to which convection-based). Subsequent modifications, especially
a recirculating volume of dialysate was delivered. High substantial reductions in wall thickness, allowed synthetic
blood compartment pressures could be achieved as a result membranes to be adapted for high-flux haemodialysis and
of narrow blood channels that were produced by specific haemodiafiltration. Contemporary synthetic membranes
spacing of the screen around the tubing and the addition generally have a wall thickness of 20–50 µm.
of a blood pump. However, this geometry also produced Similar to cellulosic membranes, some synthetic
high blood compartment resistances, which resulted in membranes, including sulfonated polyacrylonitrile
high TMPs and substantial obligatory ultrafiltration vol- (AN69) and poly(methyl methacrylate) (PMMA)24,25,
umes that could not be predicted or reliably controlled. The are structurally symmetric, as they have a uniform com-
commercial version of this device (the ‘twin coil’) consisted position that extends through the entire wall thickness
of two coils with a total surface area of 1.8 m2 (ref.14). (Fig. 1). However, the majority of synthetic membranes
The Kiil dialyser15 employed a parallel blood–dialysate that are used for contemporary haemodialysis have an
flow configuration in which the blood compartment was asymmetric structure in which a thin inner ‘skin’ layer
created by a series of cellophane flat sheet membranes sup- (width approximately ≤ 1 µm) at the membrane–blood
ported by plastic boards. Diffusive mass transfer efficiency interface serves as the primary size-discriminating
was aided by narrow blood channels (created by compres- element in solute removal26. The remaining wall thick-
sion of the boards) and the use of a new thin-walled cellu- ness (the ‘stroma’) acts as a support structure that also
losic membrane, cuprophan. Later adaptations that were provides a substantial surface area for the removal of
used in the 1960s and 1970s included multiple-use16 and molecules by adsorption. As opposed to the compact
disposable17 versions (Supplementary Fig. 1b). nature of the skin layer, the structure of the stroma is
Although the availability of all of these devices relatively open (‘macroporous’) and typically has a
allowed the use of clinical dialysis to grow, they had sponge-like or finger-type structure (Fig. 1).
many limitations, especially the high blood compart-
ment volumes required and the inefficient mass transfer Production of hollow-fibre dialysis membranes. The
characteristics. In the late 1960s, the hollow-fibre production of a hollow-fibre membrane is a combina-
artificial kidney (Supplementary Fig. 1c) revolutionized tion of art and science. We describe the sequence of steps
dialysis by providing improved geometry in terms of involved in the production (based on diffusion-induced
blood rheology and solute mass transfer18. The specific phase separation27,28) of an asymmetric polyethersulfone
advantages provided by this configuration included (PES) membrane, which is present in dialysers that are
an improved surface-area:volume ratio in the blood commonly prescribed in clinical practice.
compartment (that is, shorter diffusion path lengths) The process begins with the development of a pol-
and decreased boundary layer effects with acceptable ymer solution that contains the miscible components
end-to-end pressure drops, which made the hollow-fibre PES and polyvinylpyrrolidone (PVP). PVP is necessary
configuration the main choice in clinical dialysis. to impart sufficient hydrophilicity to allow wettability
At present, approximately 300 million hollow-fibre and prevent excessive protein deposition and platelet
haemodialysers are utilized annually worldwide. adhesion on exposure to blood. The PVP content and
molecular weight largely determine the rheology of the
Hollow-fibre membranes polymer solution and the ultimate structure of the stroma
General considerations. Dialysis membranes have layer. The polymer solution containing PVP is extruded
traditionally been categorized on the basis of their mate- through the outer ring of an annular nozzle (termed a
rial composition into cellulosic and synthetic membrane ‘spinneret’) that has two concentric openings, whereas
groups1,19,20 (Supplementary Table 1). Although unmod- another solution, a solvent–non-solvent mixture termed
ified cellulosic membranes were used extensively in the ‘bore liquid’, is injected through the inner tube. At the
the past, their use has dropped precipitously over the outlet of the spinneret, the bore fluid comes into contact

NATURE REvIEWS | NePHrology volume 14 | JUNE 2018 | 395


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

a Poly(methyl methacrylate) (Toray Medical) coagulation bath, resulting in a controlled precipita-


tion that converts the original homogeneous liquid
polymer solution into a two-phase system comprising
a polymer-rich phase (which forms the rigid membrane)
and a polymer-poor phase (which forms the pores). The
water content of both the luminal space and the wall of
the membrane is then reduced in a systematic manner to
b Amembris (B. Braun Medical) enable controlled shrinkage of pores, which is induced
by capillary forces. Controlled pore shrinkage is a cru-
cial step for membrane performance as it substantially
influences the pore size distribution and pore density of
the final product. The pore density is a major determi-
nant of both the hydraulic permeability (expressed as
c Polyethersulfone (Membrana 3M) the ultrafiltration coefficient, Kuf ) and the small-solute
diffusive permeability (expressed as the mass trans-
fer coefficient–area product; KoA) of the membrane.
In addition, mean pore size and pore size distribution
substantially influence Kuf and the sieving properties of
the membrane for different solutes (Fig. 2). Although the
ultrafiltration coefficient of a dialyser (ml/h/mmHg/m2)
d Helixone (Fresenius Medical Care) is a specific property that characterizes a ‘clean’ (that is,
unfouled) membrane, its effective value is influenced by
protein–membrane interactions29.
The KoA parameter is a function of blood flow rate,
dialysate flow rate and dialyser clearance, and is meas-
ured under conditions of zero net ultrafiltration. KoA can
be conceptualized as the maximum clearance obtained
e Polyethersulfone Polynephron (Nipro Corporation) for a dialyser of specific membrane surface area under
the theoretical conditions of infinite blood and dialysate
flow rates30. The condition of zero net ultrafiltration
implies that KoA is a purely diffusive parameter, an
assumption that is valid for all solute–membrane com-
binations with low-flux dialysers. However, a zero net
ultrafiltration condition is not equivalent to no fluid
f Polyethersulfone (Baxter International and Gambro) fluxes across a high-flux dialyser membrane, and a con-
dition of ‘pure diffusion’ is not readily attainable with
this type of device29. The KoA values for small solutes are
minimally influenced in the condition of zero net ultra-
filtration and remain valid for solute removal parameters
for high-flux dialysers when measured in this context.
However, diffusive KoA values for larger solutes (for
g Medisulfone (Medica) example, β2-microglobulin (β2m)) cannot be reliably esti-
mated under these conditions, as the convective clear-
ance associated with fluid fluxes (‘internal filtration’)
contributes substantially to total clearance. The concepts
underlying KoA and Kuf are examined further below.

From a single hollow fibre to a full-size haemodialyser.


Fig. 1 | Structural characteristics of some commercially In preparation for construction of a dialyser, several
available synthetic dialysis membranes. Scanning thousand (~10,000–17,000) individual hollow fibres are
electron micrographs of the fibre (left), fibre wall (middle)
assembled and wound together to form a ‘bundle’ that
and a magnified cross-sectional view of the internal skin
layer (right). For reference, the inner diameter of the fibres has defined characteristics, including a specific fibre
is ~200 μm. Different structural features of the membranes spatial arrangement and packing density31–44. These
are discernible, and the membranes have varying degrees parameters influence the dialysate flow distribution
of asymmetric configuration, ranging from minimum between fibres, which is an important determinant of
asymmetry (sponge-like; parts a–e) to maximum dialysate-side mass transfer. After this fibre bundle is
asymmetry (finger-type; parts f,g). inserted in the plastic housing of the dialyser, the fibres
are sealed in a two-step process. The first step involves
with the polymer solution and precipitation is initiated, application of a polyurethane (‘potting’) compound at
thus creating the inner lumen of the hollow fibre. either end of the bundle to encapsulate the fibres. The
With the basic hollow-fibre structure established, goals of this step are to avoid fibre lumen obstruction
the solution leaving the spinneret is immersed in a and minimize the fibre length that is exposed to the

396 | JUNE 2018 | volume 14 www.nature.com/nrneph


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

Pore density x cm2


λ ρ1
Pore density

λ
Pore size

Pore size
distribution

Pore size (Å)

b Ultrafiltration c Membrane d Membrane


UF permeability Dialysate sieving UF
coefficient
Cd Cp Cp
Blood
Blood Blood
Hollow fibre Hollow fibre Hollow fibre
2,500 300 KoA = 1,000
Qf = Kuf · TMP SC = 1 – R = CF/Cp
2,000 250 KoA = 750 1.0
Clearance (ml/min)
Ultrafiltration (ml/h)

200 KoA = 500 0.8

Solute SC
1,500
150 0.6
1,000
100 0.4
500 0.2
50
0 0 0
20 40 60 80 100 100 200 300 400 500 10 102 103 104 105
TMP (mmHg) Blood flow (ml/min) Molecular weight (Da)
Secondary membrane or gel effect TMP = ΔP – π = (Pb – Pd) – π MCO
High flux (Kuf = 30 ml/h/mmHg × m2) Pb π High flux
Low flux
Mid flux (Kuf = 20 ml/h/mmHg × m2)
Pd Oncotic
Low flux (Kuf = 8 ml/h/mmHg × m2)
Hydrostatic pressure
pressure

Fig. 2 | The manufacturing process influences both the pore size distribution and the pore density of a dialysis
membrane. a | Graph describing the relationship between pore size distribution and pore density, both of which influence
the hydraulic permeability of the membrane. b | Hydraulic permeability is clinically defined by the ultrafiltration coefficient
(Kuf), which is the slope of the linear portion of a plot between ultrafiltration rate (Qf) and transmembrane pressure (TMP).
TMP is dominated by hydrostatic pressure differences between the blood (Pb) and dialysate (Pd) compartments, with a small
contribution from plasma oncotic pressure (π). c | Pore density and size also influence diffusive permeability, which is
expressed as the mass transfer coefficient–area product (KoA). This parameter is a function of blood flow rate, dialysate
flow rate and dialyser clearance and can be conceptualized as the maximum clearance obtained for a membrane of a
specific surface area under the theoretical conditions of infinite blood and dialysate flow rates. d | Pore size distribution
also affects membrane sieving properties, which are described by a plot of the solute sieving coefficient (SC) as a function
of molecular weight and radius. Three different membrane profiles are depicted: low flux, high flux and medium cut-off
(MCO). The SC is defined as the ratio between the concentration of the solute in the ultrafiltrate (UF) (Cf) and that in
plasma water (Cp) in the absence of a gradient for diffusion and is the reciprocal of the rejection coefficient (R).
Cd, concentration of the solute in the dialysate.

potting compound as the corresponding membrane pressurizing the blood compartment and monitoring pres-
area does not contribute to mass exchange. After the sure changes in the dialysate compartment (an increase in
polyurethane undergoes curing, the bundle is cut with dialysate pressure is consistent with a fibre leak). Finally, the
a specialized blade at a defined temperature to control finished product is sterilized by different methods, includ-
surface roughness, which influences protein deposition ing steam, γ-irradiation, electron beam irradiation and eth-
and coagulation. This phase is also of great importance ylene oxide sterilization, although the latter is used rarely
for blood flow distribution within the bundle. owing to its allergenic potential. Sterilization is expected to
Once the casing is sealed with caps at either end, preserve the integrity of the dialyser housing, minimize any
the integrity of the membrane bundle is assessed by changes in membrane structure and reduce the amount

NATURE REvIEWS | NePHrology volume 14 | JUNE 2018 | 397


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

of residual chemicals to levels that can be removed with PVP is leached from synthetic membranes, especially
routine rinsing before dialysis treatment. Of note, the entire under specific sterilization conditions46,47. Furthermore,
process of hollow-fibre and dialyser manufacture occurs PVP leaching might be a potential cause of an anaphylactic
in clean rooms to ensure an environment with minimal reaction in a haemodialysis patient48, although the mech-
bacterial load and especially minimal endotoxin. anism is not clear. Finally, various reports have suggested
Scanning electron micrographs of membranes pro- that increases in membrane permeability during bleach-
duced by this procedure (wall thickness 30–35 µm) based reprocessing of high-flux polysulfone dialysers are
demonstrate the variation in structure from the skin layer due to PVP leaching49. Nevertheless, definitive data linking
(a nodular structure consisting of closely packed spheres the above findings to PVP leaching have not been reported.
of polymer material with a diameter of 50–100 nm) to the As our focus in this Review is on the determinants
more external component of the membrane. The polymer of membrane and filter performance in the clinical
density gradually decreases further away from the skin setting, we do not extensively discuss factors that influ-
layer to form the more open stroma layer, which is charac- ence biocompatibility, although a few basic principles
terized by a finger-like structure (Supplementary Fig. 2a,b) are worth highlighting. Complement activation is the
or a sponge-like structure (Supplementary Fig. 2c,d). parameter that is traditionally used for characteriza-
The production process also affects the smoothness of tion of dialysis membrane biocompatibility23. However,
the internal surface of the fibre, as analysed by atomic other indices have been studied, including those related
force microscopy (AFM) (Supplementary Fig. 2e–h). to cytokine activation, effects on inflammatory cells and
The roughness of the internal surface of the fibre coagulation1. Furthermore, certain principles generally
can influence several membrane properties, including governing the exposure of flowing blood to a biomate-
protein interactions and coagulability. One study used rial apply to haemodialysis membranes. First, both the
AFM to analyse the effect of different production meth- inflammatory potential (measured by cytokine and com-
ods on the surface roughness and pore size distributions plement activation and other indices) and the thrombo-
of two membranes of differing PES:PVP composition — genic propensity of a biomaterial are mediated by the
a PES:PVP weight percentage of 18:6 (series A) or 18:3 nature of the layer of proteins that are adsorbed instan-
(series B)45. All other aspects of the membrane prepara- taneously to the surface upon exposure to blood50–52. The
tion process were the same except for the method of heat effect of protein adsorption on membrane performance
treatment just before drying, which used either water is discussed extensively below. Second, in terms of
(95 °C for 30 min) or air (150 °C for 5 min). Changes in thrombogenicity, studies of biomaterials clearly differ-
surface roughness were quantified by Ra, which is the entiate low-flow conditions, in which clotting is medi-
mean distance between the surface and the centre of the ated predominantly by the coagulation pathway, from
reference plane of measurement. Pore-related parame- high-flow conditions, in which shear-induced platelet
ters, including molecular weight cut-off (MWCO) and effects have an important role53. Although not clearly
pore size distribution, were also measured. For both elucidated, both of these phenomena (adsorbed proteins
membrane series, heat treatment clearly attenuated sur- and flow rate) probably influence the thrombogenic
face roughness relative to the untreated membranes, and potential of a dialysis membrane. Finally, it should be
on the basis of the percentage decrease in Ra, the reduc- emphasized that the biocompatibility of a dialysis treat-
tion was more pronounced for the series B membrane. ment modality is a function not only of the membrane
This surface roughness attenuation was particularly but of the entire dialysis system, including the blood tub-
obvious after the heat treatment using air. A higher PVP ing and all other biomaterials in the extracorporeal cir-
content (series A) was associated with consolidation of cuit as well as the fluid to which the patient is exposed.
the inner surface polymer nodules into aggregates com-
pared with the more discrete nodules for the series B Hollow-fibre membrane transport
membrane. Similarly, AFM analyses of the outer sur- Blood compartment considerations. Although the
face of both of these asymmetric membranes showed a composition of hollow-fibre membranes varies con-
decrease in surface roughness from heat treatment. siderably, hollow fibres typically have an inner (blood
The PVP content of the casting solution and the compartment) diameter of ~180–220 µm and a length of
type of heat treatment also influenced the permeabil- 20–24 cm (ref.54). Different phenomena imposed by the
ity properties of the membrane. Before heat treatment, standard operating conditions of haemodialysis dictate
dextran-based experiments demonstrated that the the fairly narrow ranges that are acceptable for these
MWCO of both membranes was >200 kDa. Heat treat- structural parameters. The major incentive to reduce
ment with water increased the MWCO value some- hollow-fibre inner diameter is the enhancement of dif-
what, whereas heat treatment with air markedly reduced fusive mass transfer due to reduced diffusion path length
MWCO to a range (35–45 kDa) that is suitable for dialy- within the blood compartment. Furthermore, the inverse
sis. In general, the investigators noted a direct correlation proportionality between channel width and shear rates
between the degree of surface roughness and the degree (at constant flow rates) leads to attenuated blood-side
of pore size variability. Although the results of this inves- boundary layer effects for smaller diameters 55–57.
tigation are not necessarily directly applicable to clinical However, inspection of equation 1, which expresses the
observations, they are nevertheless useful findings that fundamental law (Hagen–Poiseuille) governing the lon-
are informative about fundamental membrane properties. gitudinal (axial) flow of blood through a cylinder (that is,
While PVP is a crucial component of synthetic mem- the hollow-fibre inner lumen), demonstrates the limits
branes, it is worth noting that some evidence suggests that on the extent to which inner diameter can be reduced58:

398 | JUNE 2018 | volume 14 www.nature.com/nrneph


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

Nevertheless, a common first-order approximation of


Q B = ΔP ∕(8μL ∕πr 4 ) (1) dialysis membrane pore structure is to assume that all
pores are parallel and have the same radius, which results
where QB is the blood flow rate, ΔP is the axial pressure in ultrafiltrate flow that is perpendicular to the flow of
drop, µ is the blood viscosity, L is the fibre length and r is blood and dialysate24. According to the Hagen–Poiseuille
the hollow-fibre radius. A more general form of equation law, the rate of ultrafiltrate flow is proportional to the
1 expresses the denominator on the right-hand side as fourth power of the pore radius (that is, r4) at constant
flow resistance as follows: TMP. Thus, the membrane parameters that have the
most substantial influence on water flux are the average
R = 8μL ∕πr 4 (2) pore size and, to a lesser extent, the pore density per unit
surface area.
An important implication of the above equations is
that an increase in flow resistance leads to a proportional Effect of membrane pore characteristics on diffusive
increase in the axial (arterial to venous) pressure drop that transport. Fick’s law of diffusion defines the diffusive
is required for attainment of a specific blood flow rate. The solute removal rate N (mass/time) as:
inverse relationship between flow resistance and the fourth
power of the radius is the most important factor that limits N = D ⋅ A(ΔC ∕Δx ) (3)
the extent to which hollow-fibre inner diameter can be
reduced. Of note, flow resistance is also directly related to where D is the solute diffusivity (area/time), A is the
blood haematocrit through an effect on viscosity59. The membrane area, ΔC is the transmembrane concentra-
above analysis is based on the assumption that blood flows tion gradient and Δx is the diffusion path length55. As
in streamlines (i.e., is laminar) in the hollow fibre and suggested previously, membrane pore density has a
behaves as a Newtonian fluid, for which viscosity defines substantial impact on diffusivity (which is specific to
the slope of the linear relationship between shear stress a particular solute–membrane combination), although
and shear rate58 (see Fig. 4 for additional details). membrane wall thickness is also an important determi-
The construction of the fibre bundle and its incor- nant of diffusive removal rates (Fig. 2). With increasing
poration into the dialyser casing have important impli- solute molecular weight, pore size limitations become
cations for blood and dialysate flow distribution and, increasingly important in restricting solute entry and
ultimately, in dialyser performance60–65. In particular, the limiting (‘hindering’) diffusion of molecules that gain
arterial blood port must be accurately designed to avoid pore entry66,67. Finally, in addition to membrane thick-
preferential flow pathways. Although blood flow velocity ness and diffusivity, the diffusive clearance for a mem-
tends to be higher in the inner fibres than in the periph- brane can differ substantially from the expected value
eral fibres of the bundle, this effect has been minimized owing to interaction with water molecules (wettability or
by improved designs of blood flow distributors, which hydrophilicity) and specific characteristics of the solute
can be conical, spiral or tangential31,40. Similar consider- at a given molecular weight (Fig. 3).
ations apply to the dialysate path, although flow veloci- Although the characteristics of the membrane have
ties in this case tend to be higher in the peripheral fibres an important influence on diffusive solute removal by
than in the inner fibres. A fibre packing density, which is a dialyser, the blood and dialysate compartments also
defined as the ratio of the cross-sectional area that com- play a substantial part. From an engineering perspective,
prises fibres to the total cross-sectional area of the dial- diffusive mass transfer in a dialyser is typically expressed
yser casing, close to 60% seems to be adequate to avoid in terms of the overall resistance to mass transfer (RO)
channelling of the dialysate flow. Additional aids to avoid and the overall mass transfer coefficient (KO), which
flow channelling include interfibre spacing filaments and are inversely related to one another. Overall resistance
fibre undulation, which provide homogeneous flow dis- to mass transfer takes into account the three major
tribution in the dialysate compartment and avoid the components of the dialyser57:
blood–dialysate flow mismatch that negatively affects
the performance of the haemodialyser32,36. In addition, R O = RB + RM + RD (4)
these features may enhance diffusive mass transfer by
attenuating dialysate-side boundary layer effects. where RB, RM and RD are resistances contributed by the
blood compartment, the membrane and the dialysate
Effect of membrane pore characteristics on hydrau- compartment, respectively (Fig. 4). As these three param-
lic flux. Hydraulic flux (water permeability) is the most eters can be viewed as resistances in series, the overall
common criterion traditionally used to classify dialysis resistance to mass transfer for a solute has a ‘controlling’
membranes29. The clinical parameter used to quantify component, which is the dialyser aspect that contributes
water permeability is Kuf, which is derived from the rela- most to the overall resistance. Whereas blood compart-
tionship between ultrafiltration rate (Qf ) and TMP over ment resistance is typically the controlling parameter
a clinically relevant range of TMP. for small solutes (irrespective of dialyser type), mem-
The Hagen–Poiseuille law is also relevant for describ- brane resistance becomes most important at a specific
ing the relationship between membrane pore structure molecular weight, with the transition occurring at a
and flux. Although minimization of pore size variation is lower molecular weight for low-flux dialysers owing to
an important priority, in practice the hollow-fibre spin- rather small mean pore sizes. However, although the
ning process produces a distribution of pore sizes (Fig. 2). membranes of high-flux dialysers have a more open pore

NATURE REvIEWS | NePHrology volume 14 | JUNE 2018 | 399


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

a Cuprophan (cellulose) Polysulfone Polyethersulfone Fig. 3 | The physical characteristics of membranes affect
Wall thickness 5–15 μm Wall thickness 75–100 μm Wall thickness 30 μm their functional properties. a | Membrane thickness and
structure influence functional characteristics. Whereas the
cellulosic cuprophan membrane has a thin wall and is
optimally utilized in diffusion-based haemodialysis, the
original synthetic membranes (for example, polysulfone)
200 μm 200 μm 200 μm had very thick walls, and despite higher permeability, the
performance in diffusion haemodialysis was poor. The
combination of thinner walls and high permeability in
modern synthetic membranes (for example,
polyethersulfone) compared with earlier membranes has
permitted diffusion and convection to be employed
• Natural polymer • Synthetic polymer- • Synthetic polymer- simultaneously. b | Diffusion-based solute clearance ideally
• Hydrophilic (hydrogel) asymmetric microporous declines with increasing solute molecular weight. The
• Low hydraulic • Hydrophobic structure • Hydrophobic–hydrophilic
observed values, however, can be different from the ideal
permeability • High hydraulic • High hydraulic
• Low sieving properties permeability permeability line owing to the characteristics of the solute, such as
• Prevalent use in diffusive • High sieving properties • High sieving properties electrical charges, binding of proteins, steric configuration
therapy (haemodialysis) • Exclusively used for • Combination diffusive- and interaction with water (hydrophilicity). For illustrative
convective therapy convective therapy purposes, water molecules in part b are not drawn to scale.
(haemofiltration) (high-flux haemodialysis
and haemodiafiltration)
c,d | The physicochemical properties of hydrophilic and
b hydrophobic membranes. Whereas there is a continuum of
the fluid phase inside the pores of hydrophilic membranes,
this is not the case for hydrophobic membranes. For this
Solute Steric
configuration
reason, substantial efforts were made to modify synthetic
δ+
membranes by the addition of hydrophilic domains for use
Diffusion-based in haemodialysis, whereas the use of hydrophobic
solute clearance High filtration rates
(convection) membranes has been limited to convective techniques. α,
δ– δ+ contact angle; δ, partial charge.

δ–
Electrical charges structure, the membrane itself rapidly becomes the con-
trolling resistance in this case also as solute molecular
weight increases55. In general, this is the case for solutes
Solute clearance

that have a molecular weight of >500 Da.


Membrane
H2O cut-off
Effect of membrane pore characteristics on convective
Hydrophilicity Protein binding transport. Diffusive transport becomes increasingly
limited as effective solute molecular weight increases
Solute molecular weight (Fig. 3b). This limitation can be overcome to some extent

c by the introduction of convective transport, which relies


Hydrophilic surface Polarity of water Hydrophobic surface on the mechanism of solvent drag68. The sieving coeffi-
δ+
cient is classically used to define the convective transport
H2O α Positive H2O properties of a membrane for a specific solute. In equa-
region α
tion 5, the sieving coefficient (SC) is the ratio between
α > 90° the solute concentration in the filtrate (Cf ) and the solute
α < 90°
Molecular orientation δ– δ+ concentration in the plasma water (Cp) in the absence of
Chaotic orientation
δ– a diffusion gradient across the membrane2,55:
Negative
region SC = C f∕C p (5)
Wetted membrane (hydrogen-bond formation) Rejection effect
The observed (measured) SC values are influenced by
interactions between the membrane and blood elements
d during dialysis. The nonspecific adsorption of a plasma
Hydrophilic membrane Hydrophobic membrane
protein layer, variously known as the secondary mem-
brane, gel or protein cake, reduces effective membrane
permeability immediately upon exposure to blood69–72 in
a process known as fouling. In the convective removal
of specific solutes, the influence of secondary membrane
formation is directly proportional to solute molecular
weight. The proteins found in the highest concentrations
in the plasma, such as albumin, fibrinogen and immu-
noglobulins, are the predominant components of the
Continuous fluid phase in pores Discontinuous fluid phase in pores secondary membrane73,74.
Hydrophilic sites Hydrophobic sites As discussed below in more detail, the phenomenon
of concentration polarization75 also causes in vitro sieving

400 | JUNE 2018 | volume 14 www.nature.com/nrneph


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

a b transfer resistance to the diffusive removal of large solutes.


100 Furthermore, fouling has a considerable effect on con-

Resistance to transport (%)


Bulk flow RO = RB + RM + RD
RB Transition layer vective solute clearances, especially for molecules >10 kDa
75
Solute transport

RM (ref.72). These constraints are particularly relevant in con-


Boundary layer
Proteins 50 ditions involving high ultrafiltration rates, which promote
RM Membrane
secondary membrane formation by more effectively deliv-
Boundary layer 25 RB ering plasma proteins to the membrane surface through
RD
RD Transition layer convection (versus lower ultrafiltration rates).
Bulk flow
0 In typical haemodialysis operating conditions,
10 102 103 104 105
log solute molecular weight (Da)
the water permeability characteristics for a standard
high-flux dialyser result in a fairly large drop in the
c Vmax d blood compartment axial pressure during treatment.
Hollow fibre Hollow fibre
The pressure drop is sufficiently large that, at some point
dV= Sh High shear rate
r dr
Blood SC = 1 – R along the length of the dialyser, the blood compartment
pressure is less than the dialysate compartment pressure
1 1

V = 0 V = ½Vmax in normal operating conditions. Thus, especially consid-


Hollow fibre Hollow fibre
ering the oncotic effects of plasma proteins in the blood
Low shear rate
r
Sh Blood
SC2 = 1 – R2
compartment, there is a point at which the ultrafiltrate
begins to be driven from the dialysate to the blood, as
dV
πr 2 T=–η
dr opposed to the ‘standard’ (blood–dialysate) direction in
the more proximal part of the dialyser (Fig. 5). In fact, this
V At wall: Shw = 4Q3 = 4Vπr = 4V
2

πr πr3 r combination of filtration and ‘backfiltration’ (refs76–80)


is considered to be the predominant mechanism by
Fig. 4 | Factors that affect diffusive and convective mass transfer. a | The overall which larger compounds are removed during standard
resistance to solute transport (RO) comprises three components: blood compartment high-flux haemodialysis81–84, as explained further below.
resistance (RB), membrane resistance (RM) and dialysate compartment resistance (RD).
The concentration of a molecule that is removed
b | The relative contribution to overall resistance for solutes of different molecular weights.
c | For small solutes, RB can be greater than RM, underlining the importance of high blood
from the blood by convection in the proximal part of a
flow rates and high velocity (V) gradients at the blood–membrane interface (wall shear high-flux dialyser is substantially reduced once it crosses
rate (Shw)). Shear rates influence not only the diffusive boundary layer but also the plasma the membrane owing to the combination of sieving
protein layer (secondary membrane) adsorbed to the surface. d | In convective modalities, and the diluting effect of dialysate flow. When a por-
high shear rates reduce the effects of both the secondary membrane (functional layer) tion of the dialysate is reinfused back into the blood
and the accumulation of partially rejected solutes (concentration polarization), with the as backfiltrate in the distal segment of the dialyser, the
sieving coefficient (SC) approximating the ideal value of the reciprocal of the rejection amount of solute re-infused by solvent drag is negligi-
coefficient of the membrane (R). Of note, R in part b is the diffusive mass transfer ble compared with that removed in the proximal part of
resistance and R in part d is the rejection coefficient. η, blood viscosity; Q, blood flow the dialyser owing to the blood–dialysate concentration
rate; r, hollow-fibre radius; T, shear stress (force per unit membrane surface area). Parts
difference, even if the filtration and backfiltration rates
b–d are reproduced from Ronco, C., Ghezzi, P. M., Brendolan, A., Crepaldi, C. & La Greca, G.
The haemodialysis system: basic mechanisms of water and solute transport in
are similar. In fact, the reinfused fluid can be considered
extracorporeal renal replacement therapies. Nephrol. Dial. Transplant. 13 (Suppl. 6), an ‘internal’ substitution fluid because the concentration
3–9 (1998), by permission of Oxford University Press (ref.55). of the solute of interest is essentially zero. As such, in
the context of high-flux haemodialysis, this mechanism
has been termed ‘internal haemodiafiltration’ or, more
profiles (obtained from test solutions that do not contain commonly, internal filtration (Fig. 5).
protein) to differ from those in the clinical setting. In the Maximizing the extent of internal filtration dur-
theoretically ideal situation, the SC is inversely related to ing high-flux haemodialysis through a combination
the rejection coefficient (R) as follows: of increased membrane permeability (increased pore
size)85 and higher axial blood compartment resistance
SC = 1− R (6) (decreased hollow-fibre inner diameter)39,86 can provide
clinically meaningful increases in large-solute clearance.
However, the above secondary membrane and con- Internal filtration rates are estimated to be as high as
centration polarization phenomena cause deviations 60 ml/min (~3.5 l/h)79, and new membrane designs may
from this ideal relationship (Fig. 4d). be able to extend this range. However, strict control of
dialysate quality is clearly of paramount importance in
Other mechanisms that influence solute removal. high-flux haemodialysis.
Conventional membranes that are used in haemodi- Membrane adsorption is also a consideration
alysis generally provide high clearance rates for small for large-solute removal during high-flux therapies.
solutes such as urea and creatinine, irrespective of flux. Membrane composition and structure (pore size)
However, membranes in current use, even those that are influence adsorptive removal, both qualitatively (that
traditionally considered to be highly permeable, provide is, the specific plasma proteins that are adsorbed)
limited clearance of compounds >10 kDa for several rea- and quantitatively73,74. The adsorptive removal of a
sons. Although these membranes have relatively large low-molecular-weight protein (for example, β2m) that
mean pore sizes (at least in comparison to unmodified is normally eliminated by the kidneys is considered
cellulosic membranes), they still offer substantial mass distinct from the more general process of secondary

NATURE REvIEWS | NePHrology volume 14 | JUNE 2018 | 401


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

membrane formation for two reasons7,87,88. First, because expresses Kuf in units of ml/h/mmHg, this parameter is
the mass of large proteins, such as albumin, fibrinogen frequently normalized to membrane surface area, as dis-
and ­immunoglobulins, adsorbed during secondary cussed below. Finally, the European Dialysis (EUDIAL)
membrane formation is a minute fraction of their cir- working group defines a high-flux dialyser as having an
culating plasma pools, this process has no relevant ultrafiltration coefficient >20 ml/h/mmHg/m2 and a β2m
effect on their plasma concentrations. Furthermore, the SC >0.6 (ref.99).
adsorption of such molecules is limited to the nominal Classification schemes that focus even more on solute
(blood-contacting) surface of the hollow fibre because permeability properties have been proposed. These new
they generally do not have access to the much larger classification systems acknowledge the importance of
surface area of the internal pore structure (as discussed larger molecules and the need to incorporate additional
below, albumin is an exception to this general rule). membrane classes that have extended removal spectra.
Second, adsorptive removal of smaller proteins that gain High-flux and ‘protein-leaking’ membranes have been
access to the large surface area of the internal pore struc- defined on the basis of a combination of water perme-
ture can be quantitatively important, thereby contributing ability, β2m removal parameters (sieving coefficient or
to clinically relevant decreases in plasma concentrations clearance) and albumin parameters (sieving coefficient
during treatment. Substantial amounts of β2m and other or amount removed)95. In this system, the high-flux class
low-molecular-weight proteins can be adsorbed by some is defined by a water permeability of 20–40 ml/h/mmHg/
membranes, contributing to a reduction in post-dialysis m2, a β2m SC of 0.7–0.8 and albumin loss of < 0.5 g (on
protein concentration (Supplementary Fig. 3). the basis of a 4 h haemodialysis treatment), whereas the
same parameters defining a protein-leaking membrane
New membrane classification are >40 m/h/mmHg/m2, 0.9–1.0, and 2–6 g, respectively.
As more knowledge has been gained over the past several Although not explicitly stated, these values correspond
years about compounds that contribute to uraemic toxicity, to ‘virgin’ membrane performance and do not reflect
it is clear that peptides, proteins and protein-bound potential diminutions during treatment as a result of
compounds are potentially important categories of secondary membrane effects.
toxins89–92. The peptide and protein category com- Two new membrane classes, medium cut-off (MCO)
prises >25 low-molecular-weight proteins that are as large and high cut-off (HCO), have been proposed, extending
as 51 kDa (ref.92). However, as discussed earlier, removal of the earlier classification scheme. The HCO class is char-
these molecules might be limited during diffusion-based acterized by a substantial increase in water permeability
therapies that utilize conventional high-flux dialys- (relative to both the high-flux and the protein-leaking
ers. In previous attempts to address this limitation, classes) and a virgin β2m SC of 1.0 (ref.100). However, the
‘protein-permeable’ membranes with larger mean pore high albumin loss rates associated with this membrane
sizes were developed93–95, although excessive albumin class generally preclude their long-term use for patients
losses were raised as a major concern95. Nevertheless, with ESRD101. In addition, this membrane has been used
these larger uraemic toxins are generally considered to be for fairly limited time periods in clinical conditions in
clinically important, and more effective dialytic removal which the potential risks due to albumin loss are con-
strategies might improve patient outcomes. sidered reasonable in relation to the potential benefits.
As new therapies are developed to address current Several studies have reported the use of an HCO mem-
limitations, membrane classification schemes also need brane for patients with myeloma-associated acute kidney
to evolve. As mentioned earlier, Kuf is the most commonly injury specifically to target augmented removal of free
used parameter for classification purposes; a value of antibody light chains to promote renal recovery102–104.
12 ml/h/mmHg differentiates low-permeability and Clinical experience with HCO membranes in critically
high-permeability dialysers according to the US Food ill patients receiving continuous renal replacement ther-
and Drug Administration96. However, this regulatory apy, specifically for the removal of inflammatory medi-
(versus clinical) definition was originally based on the ators, has also been the subject of several studies105–107,
requirement of using a dialysis machine with an auto- although the role of HCO membranes in clinical practice
mated ultrafiltration control system in conjunction with remains unclear.
a high-permeability dialyser to avoid fluid balance errors. On the basis of experiments involving determination
As ultrafiltration control is now a standard feature of all of dextran sieving coefficients over a wide molecular
dialysis machines, the clinical relevance of this definition weight range, a new solute removal parameter for the
is now debatable. Furthermore, this definition provides characterization of modern highly permeable mem-
little insight into the depuration capabilities of a dialyser; branes has been proposed100,108. This new parameter, the
for example, it does not recognize the category of ‘molecular weight retention onset’ (MWRO) index, is
high-efficiency dialysers, for which the primary criterion generated from a standard solute sieving coefficient ver-
is small-solute clearance rather than water flux97. sus molecular weight profile, as with the classic MWCO.
Although still fairly rudimentary, the revised mem- The MWRO is defined as the molecular weight at which
brane definition that was used for the HEMO Study is an the SC value first reaches 0.9 (whereas the MWCO corre-
improvement98. In this definition, the two criteria for a sponds to a SC of 0.1). The investigators rationalized this
high-flux dialyser are Kuf >14 ml/h/mmHg and first-use approach by suggesting that the MWRO index, which
β2m clearance >20 ml/min, whereas a first-use β2m clear- provides insights about pore size distribution, supple-
ance <10 ml/min defines a low-flux dialyser98. Although ments information provided by the MWCO, which is
the US Food and Drug Administration classification primarily correlated with mean pore size. The steepness

402 | JUNE 2018 | volume 14 www.nature.com/nrneph


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

a Blood in Blood out Fig. 5 | Analysis of pressure profiles in hollow-fibre


haemodialysers. a | Schematic representation of pressure
profiles inside a hollow-fibre haemodialyser. b | Study in a
60
2 m2 haemodialyser carried out with a scintigraphic analysis
of the concentration of a nondiffusible marker molecule
along the length of the fibres at zero net filtration. In this
40 Blood Dialysate condition, direct filtration and backfiltration are equal, and
pressure pressure
the change in concentration of the marker molecule along
∆P (mmHg) 20 the length of the fibre bundle permits calculation of the
cross-filtration flux in both directions. c | The same study
carried out in two haemodialysers with a different fibre
0
configuration (inner diameter is 200 µm (left) versus 180 µm
(right)) demonstrates a different degree of filtration or
–20 backfiltration and a consequent effect on large-molecule
Ultrafiltration Backfiltration clearance. The marginal effect on small molecules and the
–40 considerable effect on larger molecules (see the bar
0 L/2 L graphs) confirm the importance of convective transport
Fibre length that is determined by the internal filtration mechanism
b (which is typical of high-flux dialysis). d | Scintigraphic
studies allow the pressure profiles inside the haemodialyser
to be analysed. The point of inversion of the transmembrane
pressure moves towards the venous end of the dialyser
when net ultrafiltration increases. Despite very high
ultrafiltration rates, a small amount of backfiltration always
occurs owing to the increase in oncotic pressure generated
by increasing protein concentration in the distal portion of
the fibres. C1 and C3 represent the concentration of the
radiomarker molecule at the inlet and outlet of the
haemodialyser, respectively. C2a, C2b, and C2c represent the
peak concentrations of the radiomarker molecule defining
the point of inversion from direct filtration to backfiltration
along the length of the dialyser at different blood flows. ΔP,
pressure change; B12, vitamin B12; Crea, creatinine; L,
c dialyser length; P, phosphate; Qf, ultrafiltration rate. Part a is
adapted with permission from Canaud, B., Bosc, J. Y., Leray, H.
& Stec, F. Microbiological purity of dialysate for on-line
substitution fluid preparation. Nephrol. Dial. Transplant.
15 (Suppl. 2), 21–30 (2000), by permission of Oxford
University Press (ref.156). The graphs in part c are adapted
with permission from ref.82, American Society of Nephrology.

Inner diameter 200 μm Inner diameter 180 μm of the sieving coefficient versus molecular weight profile
300 300 is determined mostly by the proximity of the values of
n = 10 n = 10
these two parameters, as reiterated recently85. As part
Clearance (ml/min)

Clearance (ml/min)

250 250
200 200 of this work100,108, a classification scheme was proposed
P <0.01 in which the MWCO and the MWRO are utilized in
150 150
100 100 combination to define different dialyser classes.
50 50
Although extending the removal spectrum of modern
dialysis membranes beyond the capabilities of standard
0 0
high-flux devices is highly desirable, the design challenge
ea
ea

ea
ea
In 2
in

In 2
in
P

P
B1

B1
ul

ul

is to maximize the removal of large uraemic toxins while


Ur
Cr

Ur
Cr

d also maintaining albumin losses in a clinically accept-


able range for long-term treatment of patients with
Transmembrane pressure (mmHg)

C1 C2a C2b C2c C3 ESRD. The updated classification system mentioned


Qf = 150 previously includes MCO membranes that incorporate
Qf = 90 high-retention onset (HRO) properties — this mem-
Qf = 30 brane class may hold promise in addressing the challenge
of achieving acceptable albumin losses85. In comparison to
HCO membranes, the MCO class is intended to preserve
0 the β2m sieving characteristics and to improve the clear-
ance of other large-molecular-weight solutes (for example,
free antibody light chains) while demonstrating a marked
reduction in albumin permeability, which is influenced
0 4 8 12 16 20 24 28 primarily by the relative differences between MWRO and
Distance along fibre (cm)
MWCO (Supplementary Table 2). Pore size distribution

NATURE REvIEWS | NePHrology volume 14 | JUNE 2018 | 403


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

curves (Fig. 6a–c) and the corresponding sieving coefficient early time points (especially in convective treatments),
profiles (Fig. 6d) for three classes of membranes (high flux, with a subsequent ‘breakthrough’ at later time points,
MCO and HCO) reveal that as the separation between which is consistent with saturation of membrane binding
MWRO and MWCO decreases, the profile of the curve sites or reduced pore access due to fouling. Therefore,
becomes steeper, resulting in increased removal of large caution should be exercised in the interpretation of the
uraemic toxins and decreased loss of albumin. Although sieving coefficients for low-molecular-weight proteins
standard high-flux membranes made the development of for highly adsorptive membranes.
convective therapies possible, MCO membranes repre- In a 1986 study using a 1.3 m2 high-flux polysulfone
sent the basis for a new diffusion-based therapy called dialyser, researchers measured the sieving coefficients
‘expanded haemodialysis’ (refs84,85). of several low-molecular-weight proteins (11.8–55 kDa)
during post-dilution haemofiltration72. The prescribed
Clinical assessment of membranes blood flow rate and ultrafiltration rate was 200 ml/min
Following the clinical introduction of the first highly and 55 ml/min, respectively, with a mean TMP of
permeable dialysis membrane, AN69, in the late 1960s17 160 mmHg. On the basis of an assumed mean haema-
and the subsequent development of polysulfone and tocrit of 30% for the patient population, an average
other membranes for high-flux haemodialysis and filtration fraction of 40–45% can be estimated from the
haemodiafiltration109–111, clinical studies characterizing classic, plasma-based equation. Serial sieving coefficient
their performance were published, beginning in the late estimates were made during the first 20 min; peak val-
1980s87,112–122. These studies focused mostly on quanti- ues for each low-molecular-weight protein occurred in
fying the removal of β2m (molecular weight 11.8 kDa) the first 10 min. For each protein, a significant decrease
after its identification as the precursor of an amy- was observed between the peak sieving coefficient and
loid protein that is deposited specifically in patients the 20 min value (P < 0.05). Furthermore, the effect was
undergoing chronic haemodialysis123,124. These studies directly proportional to the molecular weight of the pro-
demonstrated that substantial β2m removal is possi- tein as no measurable filtration of solutes with a molecular
ble with high-flux dialysers used in the haemodialysis weight >30 kDa was observed after 20 min of treatment.
mode, which occurs through both transmembrane and These fairly unconvincing data can be mostly attributed
adsorptive mechanisms, and showed that unmodified to the combination of low blood flow rate, inadequate
cellulosic membranes are essentially impermeable to membrane surface area and high filtration fraction, all of
β2m. Furthermore, the major β2m removal mechanism which promote secondary membrane formation.
was membrane-specific, as transmembrane removal pre- For equivalent ultrafiltration rates up to ~6.5 l/h
dominated for polysulfone-based filters whereas adsorp- (with a blood flow rate of 300 ml/min), the clearance of
tion was the predominant mechanism for AN69 and low-molecular-weight proteins of 12–33 kDa is supe-
some PMMA dialysers. Finally, many of these studies rior with post-dilution haemodiafiltration compared
provided haemodiafiltration performance data, although with pre-dilution haemodiafiltration125,126. This benefit
the substitution volumes that were used were generally of the post-dilution mode has been attributed to the
less than those currently used in clinical practice. greater accumulation of the partially rejected proteins
In one study117, high-flux synthetic (AN69, polysul- at the blood–membrane interface (that is, concentra-
fone and PES) dialysers with a surface area of 1.6–1.9 m2 tion polarization)69,127 — it is this ‘submembranous’
were used for haemodialysis, haemodiafiltration and protein concentration on which the convective forces
haemofiltration in a small number of patients. A blood act. Because the degree of polarization is directly pro-
flow rate of 450 ml/min was prescribed for all therapies, portional to the extent of protein rejection, the relative
and both convective therapies were performed in the benefit of the post-dilution mode increases in proportion
post-dilution mode with online substitution fluid. The to the molecular weight of the solute. However, this same
mean ultrafiltration volume was 20.8 litres and 30.6 litres principle applies to albumin removal, making the balance
in haemodiafiltration and haemofiltration, respectively, between optimized removal of low-molecular-weight
which was derived from the reported mean ultrafil- proteins and minimized albumin losses a very important
tration rate for each modality. On the basis of effluent consideration in post-dilution haemodiafiltration.
collection, the total β2m removal was approximately High-flux haemodialysis and online post-dilution
100–125 mg, 175–200 mg and 225–250 mg for patients haemodiafiltration were compared in 45 patients with
undergoing haemodialysis, haemodiafiltration and ESRD (mean haematocrit, 0.30) who were treated over
haemofiltration, respectively (the differences between a 1-year period with the same high-flux 1.7 m2 polyary-
the three modalities were significant for all three dial- lethersulfone dialyser128. The mean delivered blood flow
ysers investigated). The β2m SC for the polysulfone and rate in the haemodialysis and haemodiafiltration groups
PES dialysers (measured at 5 min and 180 min of treat- was 286 ml/min and 269 ml/min, respectively, whereas
ment) were approximately 0.50–0.65 in the haemodia- the mean ultrafiltrate volume in the haemodiafiltration
filtration and haemofiltration modes. In addition, the group was 21 litres over a mean treatment period of
values of the β2m SC for the AN69 dialyser were low 4.1 h. On the basis of these mean values, the estimated
(~0.10) at 5 min in both the haemodiafiltration and ultrafiltration rate was approximately 85 ml/min, which
the haemofiltration modes but increased significantly results in an estimated filtration fraction of 0.43. Over
(P < 0.05) at 180 min to approximately 0.4 in haemodia- the 12-month study period, the mean pretreatment
filtration mode and 0.6 in haemofiltration mode. These serum concentration of β2m decreased modestly and to
AN69 data suggest that β2m adsorption predominates at a similar extent in both groups, although β2m clearance

404 | JUNE 2018 | volume 14 www.nature.com/nrneph


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

a b c

HF MCO HCO
Pore density x cm2

Pore density x cm2

Pore density x cm2


ea

in

in

ea

in

in

ea

in

in
m

m
B1

B1

B1
ob

ob

ob

m
Ur

Ur

Ur
β

β
2

2
bu

bu

bu
gl

gl

gl
Al

Al

Al
yo

yo

yo
M

M
Pore size (Å) Pore size (Å) Pore size (Å)

d
1.0
MWRO
HCO MCO
0.8 HF
LF

0.6
SC

0.4
Albumin
(68 kDa)
0.2 β2m
MWCO (12 kDa)

0.0
100 1,000 10,000 100,000
log molecular weight (Da)

Fig. 6 | Performance characteristics of haemodialysis membranes derived from a suggested new classification
system. Pore size distribution curves are schematically depicted for three classes of membranes, high flux (HF; part a),
medium cut-off (MCO; part b) and high cut-off (HCO; part c), as well as their sieving coefficient (SC) profiles (part d). As the
interval between molecular weight retention onset (MWRO) and molecular weight cut-off (MWCO) decreases, the profile
of the curve becomes steeper, increasing the removal of large uraemic toxins (such as β2-microglobulin (β2m)) while
decreasing the loss of albumin (part d). A low-flux (LF) membrane is shown for comparison. B12, vitamin B12. Part d is
adapted with permission from ref.85, Karger.

was significantly higher in the haemodiafiltration group 60 ml/min and 90 ml/min, respectively. The efficacy of
than in the haemodialysis group (61 ml/min versus β2m removal was assessed by the difference between
38 ml/min; P < 0.001). The researchers hypothesize that pretreatment and post-treatment serum β2m concentra-
slow intercompartment mass transfer in the body might tions, whereas albumin losses were estimated from serial
account for this finding. Furthermore, the decrease measurements of albumin concentration in the diafiltrate
in the serum concentration of complement factor D using a modified area under the curve approach.
(molecular weight 24 kDa) was significantly greater in The efficacy of β2m removal was linearly related to
the haemodiafiltration group than in the haemodialysis the substitution fluid rate for each dialyser. At a fluid
group over the study period (P < 0.001). substitution rate of 90 ml/min, the β2m removal rate for
The relationship between β2m removal and albumin most dialysers ranged between 65% and 75%. Peak val-
loss during online post-dilution haemodiafiltration ues of instantaneous albumin removal rates (in mg/min)
was assessed in a group of patients who were treated occurred at the earliest measurement time point (10 min)
with eight different high-flux dialysers of surface area and subsequently decreased in an exponential manner.
1.3–1.5 m2 (ref.129). Whereas the substitution fluid rate Furthermore, cumulative albumin loss for a given treat-
was variable (30 ml/min, 60 ml/min or 90 ml/min, ment had a strong dependence on substitution fluid rate
and a constant dialysate flow rate of 600 ml/min), the — this relationship tended to be nonlinear for dialysers
prescribed blood flow rate remained constant in each with greater albumin permeability. For a substitution
patient (mean 292 ml/min) in this population, and the fluid rate of 90 ml/min, albumin loss (normalized to a
mean haematocrit was 36%. On the basis of these mean treatment time of 4 h) was approximately 0.4–7 g. On the
values and an assumed net ultrafiltration rate of 10 ml/min, basis of these data, the investigators proposed a dialyser
the mean filtration fraction can be estimated as 0.21, membrane classification system in which low, medium
0.38 and 0.54 at substitution fluid rates of 30 ml/min, and high values for β2m removal rate (0–50%, >50–70%

NATURE REvIEWS | NePHrology volume 14 | JUNE 2018 | 405


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

and >70%, respectively) are evaluated in conjunction volumes are also associated generally with greater albu-
with similar gradations of albumin loss (0–2 g, 2–5 g min loss, without a concomitant benefit of increased
and >5 g, respectively). β2m removal139. This observation is a predictable conse-
Another study evaluated the possibility that adapta- quence of albumin polarization at the membrane surface
tions in dialyser design could ameliorate the relationship due to the combination of high ultrafiltration rates and
between β2m removal and albumin loss in post-dilution haemoconcentration, as described previously.
haemodiafiltration130. This crossover randomized trial The factors that might potentially influence albu-
involved a modified polysulfone filter that, according to min loss during post-dilution haemodiafiltration have
the manufacturer, has a more open structure in the mem- been evaluated135. These investigators assessed poly-
brane region adjacent to the skin layer than the standard sulfone dialysers with surface areas of 1.5–2.2 m2 and
high-flux polysulfone dialyser (surface area 1.4 m2 for PMMA dialysers with surface areas of 1.6–2.1 m 2.
both membranes). This adaptation is intended to increase Patients were treated with either a ‘manual’ infusion
the β2m sieving coefficient and reduce the albumin siev- technique, in which the dialysis nurse adjusts the substi-
ing coefficient (on the basis of in vitro determinations). tution fluid rate according to the TMP, or an automated
In this trial, mean blood flow rate and treatment time technique, in which the dialysis machine continuously
were approximately 405 ml/min and 4.9 h, respectively, adjusts the infusion rate on the basis of several param-
whereas total convection volume (substitution fluid eters, including membrane type, haematocrit, total
and net fluid loss) was approximately 30 litres in both protein concentration, blood flow rate and ultrafiltra-
groups. The percentage reduction in the serum concen- tion rate, with an established maximum TMP. Of note,
tration of low-molecular-weight proteins ranging from almost half of the 37 treatments that were performed
11.8 kDa (β2m) to 33 kDa (α1-microglobulin) was signif- with the automated technique had to be switched to the
icantly greater (P <0.05) for the modified dialyser than manual mode owing to high TMP values (>250 mmHg)
for the standard dialyser, as was the mass removal (on that were not adequately modulated by automated infu-
the basis of diafiltrate sampling). However, mean albu- sion rate changes. Overall, the mean albumin loss was
min loss was also significantly higher for the adapted 3.13 ± 2.5 g per session, was linearly related to TMP and
filter versus the standard filter (4.25 g versus 3.01 g was significantly higher in the automated group than
per treatment; P = 0.03). Thus, the permeability for in the manual infusion control group (3.92 ± 3.06 g ver-
low-molecular-weight proteins and albumin for these two sus 2.37 ± 1.67 g; P = 0.01). Furthermore, multivariate
dialysers correlated reasonably well in the clinical setting. regression analysis showed that TMP was the only sig-
Prescribed ultrafiltration rates that are necessary nificant predictor of albumin loss (P = 0.002), although
to achieve convection volumes >23 litres per session, the predictive value of membrane type almost reached
recently established by consensus guidelines90,131–134, are statistical significance (P = 0.054). However, no associa-
predicated on attaining fairly high blood flow rates and tion was observed between diafiltrate albumin loss and
on the dialyser achieving TMP values up to 300 mmHg. serum albumin concentration during this fairly short
However, in many patients, these conditions are asso- (3–4 month) study.
ciated with prominent haemoconcentration and sec- These albumin loss data135 are generally consistent
ondary membrane effects, which result in frequent with results from other post-dilution haemodiafiltration
pressure alarms and potential treatment interruptions. studies, and these investigators recommend that clinical
Contemporary dialysis machines utilize different concerns about albumin loss should not limit the use of
approaches in an attempt to mitigate these effects135–138, post-dilution haemodiafiltration. Nevertheless, efforts
enabling a distinction to be made between ‘maximum’ have been focused on the development of alternative
convection volumes (defined by the blood flow rate convective techniques that preserve the depuration
and TMP parameters discussed above) and convec- capabilities of post-dilution haemodiafiltration while
tion volumes that are more realistically attainable in also overcoming some of the complexities and techni-
the operating conditions of a given haemodiafiltration cal limitations of this modality141–143. In various ways,
treatment. As discussed below, although these mitigating these alternative approaches aim to reduce haemo-
technologies are designed to maximize the substitution concentration and attenuate fouling effects so that
volume, treatments employing them still frequently TMP and albumin losses can be maintained in clini-
result in operating TMP values that cannot be sustained cally acceptable ranges. However, there are insufficient
throughout a treatment. clinical data to warrant these alternative techniques
The considerable variation in Kuf, which is defined supplanting post-dilution haemodiafiltration as the
by the linear portion of the ultrafiltration rate versus standard convective modality when haemodiafiltration
TMP profile, during a post-dilution haemodiafiltra- is the chosen therapy. Furthermore, the superiority of
tion treatment can substantially influence the ability to post-dilution haemodiafiltration over standard high-flux
achieve a desired convection volume139,140. Furthermore, haemodialysis is still a matter of debate.
delivery frequently falls short for treatments in which Efforts to extend the permeability spectrum of
maximum convection volumes are prescribed, whereas high-flux synthetic membranes might enable the
treatment prescriptions that accommodate the var- presumption of a need for a convection-based therapy to
iations in the ultrafiltration rate–TMP relationship be challenged5,79,84,108,144–146. By virtue of larger pore sizes,
are much more likely to achieve the prescribed (albeit increased membrane diffusivity is one mechanism by
lower) convection volume139. In addition to treatment which the removal of large solutes is augmented with the
delivery shortfalls, prescribed maximum convection previously mentioned MCO class of dialysers (relative

406 | JUNE 2018 | volume 14 www.nature.com/nrneph


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

to standard high-flux membranes). However, although fluid administration5,144. Clinical outcome data for the new
haemodialysis using this type of dialyser is technically therapy (‘expanded haemodialysis’) that is made possible
diffusion-based, most large-solute removal still occurs by this class of dialysers will help establish its clinical role
by convection through the mechanism of internal filtra- in the future, especially in relation to haemodiafiltration.
tion, as discussed earlier. For MCO and other dialysers, The appropriate balance between low-molecular-weight
the effect of this mechanism is intentionally augmented protein removal and albumin loss can be raised a final time
through increases in the mean pore size and reductions here147. In addition to the opinion that albumin losses of up
in the inner diameter of hollow fibres. Preliminary data to 4 g per session are clinically acceptable135, a retrospective
suggest that this class of dialysers has depuration capabil- analysis of Japanese patients reported a significantly higher
ities that approach those of online post-dilution haemo- survival in patients with ESRD who sustained albumin
diafiltration without the need for (exogenous) substitution losses >3 g per treatment than in those patients with lower

Box 1 | Multidimensional classification of dialysis membranes


Use of a radar plot can provide a multidimensional classification of dialysis Composition
membranes (see the figure). The value for each parameter ranges from a Nature Structure
minimum value at the centre to a maximum value at the circumference.
The parameters are as follows.
Nature and composition Ko Kuf
The major compositional distinction is between cellulosic (natural) and
noncellulosic (synthetic) membranes. Noncellulosic synthetic membranes
(polyamide, polysulfone, polyethersulfone (PES), polyacrylonitrile (PAN),
polymethylmethacrylate (PMMA) and others) and modified cellulosic
membranes are almost exclusively used in clinical practice. In new MWRO MWCO
synthetic membranes, polymer blending enhances both biocompatibility
and performance.
Structure
Most synthetic hollow fibres have a complex structure with a finely porous Thickness Biocompatibility
internal skin layer and an external support structure. The support structure
may be sponge-like or finger-like, depending on the production method.
Ultrafiltration coefficient ζ-Potential Hydrophilicity
The ultrafiltration coefficient (Kuf; ml/h/mmHg/m2; also known as the Surface
hydraulic permeability coefficient) for a membrane is the ratio of the functionalization
ultrafiltration rate (Qf; ml/h/m2) to the transmembrane pressure (TMP;
mmHg). On the basis of water flux only, for low-flux membranes processes enable modification of the inner surface of hollow fibres by
Kuf < 10 ml/h/mmHg/m2, whereas for high-flux membranes Kuf = 20–40 ml/h/ several techniques (see the main text for examples).
mmHg/m2 and mid-flux membranes have intermediate Kuf values. However, ζ-Potential
modern classification schemes also incorporate solute removal parameters The ζ-potential is the electric potential at the blood–membrane interface
(see main text and below). due to the presence of electronegative charges located in the skin layer of
Molecular weight retention onset the membrane. The process of polymerization, the chemical composition
The molecular weight retention onset (MWRO) governs the shape of the of the membrane and polymer blending potentially contribute to the
solute sieving curve for a membrane and describes the molecular weight ζ-potential of a membrane.
and radius at which the sieving coefficient (SC) value is 0.9. Membranes Thickness
with a tight pore size distribution that were designed to have a steep The thickness of a membrane determines the distance that solutes must
sieving curve have been produced with the aims of minimizing the diffuse between the blood and the dialysate. The original cellulosic
molecular weight interval between the MWRO and the molecular weight membranes were 15 µm thick, which was subsequently reduced to 5 µm.
cut-off (MWCO) and maintaining the MWCO at a value close to the The original synthetic membranes were 70–100 µm thick, which has been
molecular weight of albumin. These membranes are described as medium reduced to 30 µm or less with a concomitant reduction of the thickness of
cut-off (MCO) membranes. the internal skin layer to approximately 1 µm or less.
Biocompatibility
Molecular weight cut-off
Although the biocompatibility of dialysis membranes can be judged by
The molecular weight cut-off (MWCO) is defined as the solute molecular
several criteria, complement activation has been the most widely studied
weight that corresponds to a SC value of 0.1. Pore size distribution
parameter. Other criteria include thrombogenicity, contact activation and
substantially influences the MWCO value of a membrane and is of critical
cytokine generation.
importance as it approaches the molecular weight of albumin owing to its
Hydrophilicity effect on unwanted albumin losses during treatment.
The material composition of a membrane affects its interaction with water.
Cellulosic membranes are hydrophilic whereas early synthetic membranes Diffusive mass transfer coefficient
were highly hydrophobic. Modifications and new polymer blending have The diffusive mass transfer coefficient (Ko) is a theoretical parameter to
resulted in synthetic membranes that are less hydrophobic so that the describe membrane performance in diffusion in ideal conditions of
combination of diffusive and convective mass transfer for solute removal is unlimited blood flow and dialysate flow. The final characteristics of a
now possible. membrane should be normalized to the membrane surface area (KoA). Ko
and KoA are important parameters to define membrane diffusive transport
Surface functionalization
capacity for a specific haemodialyser–solute combination.
The characteristics of the internal surface of the membrane are important
for the interaction with the blood. New biochemical and physical Figure adapted with permission from ref.85, Karger.

NATURE REvIEWS | NePHrology volume 14 | JUNE 2018 | 407


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

albumin losses148. The investigators attributed this finding substantial changes in membrane performance. For
to the beneficial removal of protein-bound uraemic toxins. example, functionalization of the internal surface of an
The data from this study can be viewed in the context of AN69 membrane resulted in altered coagulation prop-
peritoneal dialysis, in which albumin losses of approxi- erties and electronegativity150,151. In addition, function-
mately 4 g per day (30 g per week) are typical149. These con- alization of a PES membrane via covalent linkage with
siderations notwithstanding, the clinically acceptable limit vitamin E152–154 led to a clinically meaningful reduction
of albumin loss during extracorporeal dialysis treatments in circulating reactive oxygen species that are generated
remains to be defined more clearly. at the blood–membrane interface. These new modifica-
tions may develop further in the future, suggesting that
Membrane functionalization membrane properties should be characterized on the
The production process and polymeric composition deter- basis of not only classic parameters but also indices that
mine the final characteristics and functional properties of a are capable of assessing these new features. Finally, in the
membrane. The internal skin layer is of crucial importance future, entirely new types of membranes might be applied
because it represents the true barrier for mass separation to novel haemodialysis therapies, such as silicon-based
and is the portion of the membrane that comes in direct nanopore membranes for an implantable device155.
contact with the blood. The characteristics of the internal
surface are responsible for the blood–membrane interac- Conclusions
tions, including phenomena such as coagulation, platelet The bidirectional process of mass separation between
adhesion and protein deposition. A substantial reduction the blood and the dialysate involves several mecha-
in activation of the clotting cascade and membrane fouling nisms of interaction between the fluid phases and the
has been achieved by improvement of the inner surface membrane barrier. The nature of the fluid phases,
design and smoothness. Specific physicochemical pro- the characteristics of the solutes and the structure
cesses have been applied during membrane production of the membrane represent a combination of elements
to improve the characteristics of the membrane inner sur- that are involved in the final process of mass separation
face. These efforts have resulted in increased permeability, and dialysis. The development of a unifying classifica-
improved haemocompatibility and reduced fouling effects tion system for dialysis membranes is extremely complex
during prolonged utilization. and must be multidimensional. When overly simplistic
Several membranes that are routinely used in dialy- schemes are used, membranes that have substantial dif-
sis display characteristic surface modification through ferences of potential clinical importance cannot be differ-
oxidation, plasma treatment, pervaporation or copol- entiated. Each membrane, although belonging to a classic
ymerization. Membrane surfaces are also modified by category, has a specific area and profile that is dependent
additives, including PVP, polyethylene co-vinyl alcohol on other variables2 (Box 1). Furthermore, a uniform clas-
and polyethylene glycol. Polymer blending (for example, sification of membranes cannot be proposed unless all
polyethyleneoxide-modified polyacrylonitrile, poly- characteristics — in particular, those that may be unique
amide–PES and polyarylethersulfone), macromolecule for a specific membrane — are taken into considera-
inclusion and coating of the inner surface are other tion. For personalized (‘precision’) medicine, a clinician
mechanisms of membrane modification and adaptation can then select a dialysis membrane on the basis of the
for specific clinical uses. Finally, further modifications of specific needs of the individual patient and the desired
the membrane surface can be obtained by polymer func- clinical results while bearing in mind that cost and
tionalization via chemical reactions, molecular imprint- cost-effectiveness are important considerations.
ing technology, ion implantation and photochemical
modification. In the future, these processes might achieve Published online 05 May 2018

1. Clark, W. R., Hamburger, R. J. & Lysaght, M. J. Effect kinin-forming capacity of polyacrylonitrile dialysis 18. Lipps, B. et al. The hollow fiber artificial kidney. Trans.
of membrane structure and composition on membranes. Biomaterials29, 1139–1146 (2008). Am. Soc. Artif. Intern. Organs13, 200–207 (1967).
performance and biocompatibility in hemodialysis. 9. Haas, G. Dialysis of flowing blood in the patient. Klin. 19. Lysaght, M. J. Hemodialysis membranes in transition.
Kidney Int.56, 2005–2015 (1999). Wochenschr.70, 1888 (1923). Contrib. Nephrol.61, 1–17 (1988).
2. Ronco, C., Neri, M., Lorenzin, A., Garzotto, F. & Clark, 10. Clark, W. R. Hemodialyzer membranes and 20. Lysaght, M. J. Evolution of hemodialysis membranes.
W. R. Multidimensional classification of dialysis configurations: a historical perspective. Semin. Contrib. Nephrol.113, 1–10 (1995).
membranes. Contrib. Nephrol.191, 115–126 (2017). Dial.13, 309–311 (2000). 21. Craddock, P., Fehr, J., Dalmasso, A., Brigham, K. &
3. Bowry, S. K., Gatti, E. & Vienken, J. Contribution of 11. Kolff, W. & Berk, H. The artificial kidney: a dialyzer with Jacob, H. Hemodialysis leukopenia: pulmonary
polysulfone membranes to the success of convective a great area. Acta Med. Scand.117, 121–134 (1944). vascular leukostasis resulting from complement
dialysis therapies. Contrib. Nephrol.173, 110–118 12. Alwall, N. On the artificial kidney. I. Apparatus or activation by dialyzer cellophane membranes. J. Clin.
(2011). dialysis of blood in vivo. Acta Med. Scand.128, Invest.59, 879–888 (1977).
4. Islam, M. S. et al. Vitamin E-coated and 317–325 (1947). 22. Hakim, R., Fearon, D. & Lazarus, J. M.
heparin-coated dialyzer membranes for heparin-free 13. Kolff, W., Watschinger, B. & Vertes, B. Results in Biocompatibility of dialysis membranes: effects of
hemodialysis: a multicenter, randomized, crossover patients treated with the coil kidney. JAMA161, chronic complement activation. Kidney Int.26,
trial. Am. J. Kidney Dis.68, 752–762 (2016). 1433–1437 (1956). 194–200 (1984).
5. Zweigart, C. et al. Medium cut-off membranes — 14. Vertes, B., Aoyama, S. & Kolff, W. The twin-coil 23. Hakim, R. M. Clinical implications of hemodialysis
closer to the natural kidney removal function. Int. J. disposable artificial kidney. Trans. Am. Soc. Artif. membrane bioincompatibility. Kidney Int.44,
Artif. Organs40, 328–334 (2017). Intern. Organs3, 119–121 (1958). 484–494 (1993).
6. Boschetti-de-Fierro, A. et al. Membrane innovation in 15. Kiil, F. Development of a parallel-flow artificial kidney 24. Takeyama, T. & Sakai, Y. Polymethylmethacrylate: one
dialysis. Contrib. Nephrol.191, 100–114 (2017). in plastics. Acta Chir. Scand. Suppl.253, 142–150 biomaterial for a series of membranes. Contrib.
7. Clark, W. R., Macias, W. L., Molitoris, B. A. & Wang, (1960). Nephrol.125, 9–24 (1998).
N. H. Membrane adsorption of ß2-microglobulin: 16. Cole, J., Pollard, T. & Murray, J. Studies on the 25. Thomas, M., Moriyama, K. & Ledebo, I. AN69:
equilibrium and kinetic characterization. Kidney modified polypropylene Kiil dialyzer. Trans. Am. Soc. evolution of the world’s first high permeability
Int.46, 1140–1146 (1994). Artif. Intern. Organs9, 67–70 (1963). membrane. Contrib. Nephrol.173, 119–129 (2011).
8. Désormeaux, A., Moreau, M. E., Lepage, Y., Chanard, 17. Funck-Bretano, J. et al. A new disposable plate-kidney. 26. Clark, W. R. & Gao, D. Membranes for dialysis:
J. & Adam, A. The effect of electronegativity and Trans. Am. Soc. Artif. Intern. Organs15, 127–130 composition, structure, and function. Contrib.
angiotensin-converting enzyme inhibition on the (1969). Nephrol.137, 70–77 (2002).

408 | JUNE 2018 | volume 14 www.nature.com/nrneph


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

27. Zweigart, C. et al. in Comprehensive Membrane 52. Lambrecht, L. et al. The influence of pre-adsorbed for a new medium cut-off dialyzer. Contrib.
Science and Engineering 2nd edn (eds Drioli, E., canine von Willenbrand factor, fibronectin, and Nephrol.191, 127–141 (2017).
Giorno, L. & Fontana, E.) 215–247 (Elsevier, 2017). fibrinogen on ex vivo artificial surface-induced 80. Ronco, C., Brendolan, A., Lupi, A., Bettini, M. C. & La
28. Chung, T. S. N. in Advanced Membrane Technology thrombogenesis. Thromb. Res.41, 99–117 (1986). Greca, G. Enhancement of convective transport by
and Applications Ch. 31 (eds Li, N. N., Fane, A. G., Ho, 53. Casa, L., Deaton, D. H. & Ku, D. N. Role of high shear internal filtration in a modified experimental dialyzer.
W. S. W. & Matsuura, T.) (John Wiley and Sons, 2008). rate in thrombosis. J. Vasc. Surg.61, 1068–1080 Kidney Int.54, 979–985 (1998).
29. Clark, W. R. Quantitative characterization of (2015). 81. Ronco, C., Brendolan, A., Lupi, A., Metry, G. & Levin,
hemodialyzer solute and water transport. Semin. 54. Clark, W. R., Gao, D., Neri, M. & Ronco, C. Solute N. W. Effects of reduced inner diameter of hollow
Dial.14, 32–36 (2001). transport in hemodialysis: advances and limitations of fibers in hemodialyzers. Kidney Int.58, 809–817
30. Michaels, A. S. Operating parameters and current membrane technology. Contrib. Nephrol.191, (2000).
performance criteria for hemodialyzers and other 84–99 (2017). 82. Ronco, C., Brendolan, A., Crepaldi, C., Rodighiero, M.
membrane-separation devices. Trans. Am. Soc. Artif. 55. Ronco, C., Ghezzi, P. M., Brendolan, A., Crepaldi, C. & & Scabardi, M. Blood and dialysate flow distributions
Intern. Organs12, 387–392 (1966). La Greca, G. The haemodialysis system: basic in hollow-fiber hemodialyzers analysed by
31. Kim, J. C. et al. Effects of arterial port design on blood mechanisms of water and solute transport in computerized helical scanning technique. J. Am. Soc.
flow distribution in hemodialyzers. Blood Purif.28, extracorporeal renal replacement therapies. Nephrol. Nephrol.13, S53–S61 (2002).
260–267 (2009). Dial. Transplant.13 (Suppl. 6), 3–9 (1998). 83. Mineshima, M. New trends in HDF: validity of internal
32. Ronco, C. et al. Flow distribution analysis by helical 56. Huang, Z., Clark, W. R. & Gao, D. Determinants of filtration-enhanced hemodialysis. Blood Purif.22
scanning in polysulfone hemodialyzers: effects of fiber small solute clearance in hemodialysis. Semin. Dial.18, (Suppl. 2), 60–66 (2004).
structure and design on flow patterns and solute 30–35 (2005). 84. Ronco, C. & La Manna, G. Expanded hemodialysis: a
clearances. Hemodial. Int.10, 380–388 (2006). 57. Colton, C. K. & Lowrie, E. G. in The Kidney 2nd edn new therapy for a new class of membranes. Contrib.
33. Brendolan, A. et al. Dialytic performance evaluation of (eds Brenner, B. M. & Rector, F. C.) 2425–2489 (WB Nephrol.190, 124–133 (2017).
Rexeed: a new polysulfone-based dialyzer with Saunders, 1981). 85. Ronco, C. The rise of expanded hemodialysis. Blood
improved flow distributions. Int. J. Artif. Organs.28, 58. Bird, R. B., Stewart, W. E. & Lightfoot, E. N. in Purif.44, I–VIII (2017).
966–975 (2005). Transport Phenomena 1st edn (eds Bird, R. B., 86. Ronco, C. & Bowry, S. Nanoscale modulation of the
34. Gastaldon, F. et al. Effects of novel manufacturing Stewart, W. E. & Lightfoot, E. N.) 34–70 (John Wiley pore dimensions, size distribution and structure of a
technology on blood and dialysate flow distribution in and Sons, 1960). new polysulfone-based high-flux dialysis membrane.
a new low flux “alpha Polysulfone” hemodialyzer. 59. Merrill, E. W. Rheology of blood. Physiol. Rev.49, Int. J. Artif. Organs24, 726–735 (2001).
Int. J. Artif. Organs.26, 105–112 (2003). 863–888 (1949). 87. Goldman, M. et al. Adsorption of beta2-microglobulin
35. Ronco, C., Brendolan, A., Crepaldi, C., Gastaldon, 60. Chapdelaine, I. et al. Optimization of the convection on dialysis membranes: comparison of dialyzers and
F. & Levin, N. W. Flow distribution and cross filtration volume in online post-dilution haemodiafiltration: effect of reuse. Int. J. Artif. Organs12, 373–378
in hollow fiber hemodialyzers. Contrib. Nephrol.137, practical and technical issues. Clin. Kidney J.8, (1989).
120–128 (2002). 191–198 (2015). 88. Mares, J., Thongboonkerd, V., Tuma, Z., Moravec, J. &
36. Ronco, C. et al. Dialysate flow distribution in hollow 61. Kim, J. C. et al. Effects of dialysate flow configurations Matejovic, M. Specific adsorption of some
fiber hemodialyzers with different dialysate pathway in continuous renal replacement therapy on solute complement activation proteins to polysulfone dialysis
configurations. Int. J. Artif. Organs23, 601–609 removal: computational modeling. Blood Purif.35, membranes during hemodialysis. Kidney Int.76,
(2000). 106–111 (2013). 404–413 (2009).
37. Ronco, C. et al. Performance of DIAPES filters in CRRT. 62. Kim, J. C. et al. Computational modeling of effects of 89. Clark, W. R. & Gao, D. Low-molecular weight proteins
Contrib. Nephrol.138, 144–152 (2003). mechanical shaking on hemodynamics in hollow fibers. in end-stage renal disease: potential toxicity and
38. Brendolan, A. et al. Flow dynamic characteristics of Int. J. Artif. Organs35, 301–307 (2012). dialytic removal mechanisms. J. Am. Soc. Nephrol.13,
DIAPES hemodialyzers. Contrib. Nephrol.138, 63. Kim, J. C. et al. Enhancement of solute removal in a S41–S47 (2002).
27–36 (2003). hollow-fiber hemodialyzer by mechanical vibration. 90. Massy, Z. & Liabeuf, S. Middle-molecule uremic toxins
39. Ronco, C. et al. Hemodialyzer: from macro-design to Blood Purif.31, 227–234 (2011). and outcomes in chronic kidney disease. Contrib.
membrane nanostructure; the case of the FX-class of 64. Ronco, C. Fluid mechanics and crossfiltration in Nephrol.191, 8–17 (2017).
hemodialyzers. Kidney Int. Suppl.80, 126–142 hollow-fiber hemodialyzers. Contrib. Nephrol.158, 91. Barreto, F. C., Barreto, D. V. & Canziani, M. E. F.
(2002). 34–49 (2007). Uremia retention molecules and clinical outcomes.
40. Ronco, C. et al. Effects of hematocrit and blood flow 65. Ronco, C. & Levin, N. W. Mechanisms of solute Contrib. Nephrol.191, 18–31 (2017).
distribution on solute clearance in hollow-fiber transport in extracorporeal therapies. Contrib. 92. Eloot, S. et al. Protein-bound uremic toxin profiling as
hemodialyzers. Nephron89, 243–250 (2001). Nephrol.149, 10–17 (2005). a tool to optimize hemodialysis. PLoS ONE11,
41. Ronco, C., Ballestri, M. & Brendolan, A. New 66. Villarroel, F., Klein, E. & Holland, F. Solute flux in e0147159 (2016).
developments in hemodialyzers. Blood Purif.18, hemodialysis and hemofiltration membranes. Trans. 93. Kneis, C. et al. Elimination of middle-sized uremic
267–275 (2000). Am. Soc. Artif. Organs23, 225–232 (1977). solutes with high-flux and high-cut-off membranes: a
42. Ronco, C. et al. In vitro and in vivo evaluation of a new 67. Zydney, A. L. Bulk mass transport limitations during randomized in vivo study. Blood Purif.36, 287–294
polysulfone membrane for hemodialysis. Reference high-flux hemodialysis. Artif. Organs17, 919–924 (2013).
methodology and clinical results. (Part. 2: in vivo (1993). 94. Yu, X. The evolving patterns of uremia: unmet clinical
study). Int. J. Artif. Organs22, 616–624 (1999). 68. Ofsthun, N. J. & Zydney, A. L. Importance of needs in dialysis. Contrib. Nephrol.191, 1–7 (2017).
43. Ronco, C. et al. In vitro and in vivo evaluation of a new convection in artificial kidney treatment. Contrib. 95. Ward, R. A. Protein-leaking membranes for
polysulfone membrane for hemodialysis. Reference Nephrol.108, 53–70 (1994). hemodialysis: a new class of membranes in search of
methodology and clinical results. (Part 1: in vitro 69. Huang, Z., Gao, D., Letteri, J. J. & Clark, W. R. an application? J. Am. Soc. Nephrol.16, 2421–2430
study). Int. J. Artif. Organs22, 604–615 (1999). Blood-membrane interactions during dialysis. Semin. (2005).
44. Brendolan, A., Ronco, C., Ghezzi, P. M., Scabardi, M. Dial.22, 623–628 (2009). 96. US Food and Drug Administration. Guidance for the
& La Greca, G. Hydraulic and flow dynamic 70. Langsdorf, L. J. & Zydney, A. L. Effect of blood contact content of premarket notifications for conventional
characteristics of PMMA dialyzers. Contrib. on the transport properties of hemodialysis and high permeability hemodialyzers. FDAhttps://www.
Nephrol.125, 41–52 (1999). membranes: a two-layer model. Blood Purif.12, fda.gov/downloads/medicaldevices/
45. Barzin, J. et al. Characterization of polyethersulfone 292–307 (1994). deviceregulationandguidance/guidancedocuments/
hemodialysis membrane by ultrafiltration and 71. Morti, S. M. & Zydney, A. L. Protein-membrane ucm080166.pdf (1998).
atomic force microscopy. J. Memb. Sci.237, 77–85 interactions during hemodialysis: effects on solute 97. Keshaviah, P., Luehmann, D., Ilstrup, K. & Collins, A.
(2004). transport. ASAIO J.44, 319–326 (1998). Technical requirements for rapid high-efficiency
46. Miyata, M., Konishi, S., Shimamoto, Y., Kamada, A. & 72. Rockel, A. et al. Permeability and secondary therapies. Artif. Organs10, 189–194 (1986).
Umimoto, K. Influence of sterilization and storage membrane formation of a high flux polysulfone 98. Eknoyan, G. et al. Effect of dialysis dose and
period on elution of polyvinylpyrollidone from wet-type hemofilter. Kidney Int.30, 429–432 (1986). membrane flux in maintenance hemodialysis. N. Engl.
polysulfone membrane dialyzers. ASAIO J.61, 73. Clark, W. R., Macias, W. L., Molitoris, B. A. & Wang, J. Med.347, 2010–2019 (2002).
468–473 (2015). N. H. L. Plasma protein adsorption to highly permeable 99. Tattersall, J. E. & Ward, R. A., EUDIAL group. Online
47. Murakami, J., Kaneko, I., Kimata, N., Mineshima, M. dialysis membranes. Kidney Int.48, 481–487 (1995). hemodiafiltration: definition, dose quantification, and
& Akiba, T. Problems in the evaluation of 74. Gachon, A., Mallet, J., Trideon, A. & Deteix, P. safety revisited. Nephrol. Dial. Transplant.28, 542–
polyvinylpyrollidone elution from polysulfone Analysis of proteins eluted from hemodialysis 550 (2013).
membranes sterilized by gamma-ray radiation. Ren. membranes. J. Biomater. Sci. Polym. Ed.2, 263–276 100. Boschetti-de-Fierro, A., Voigt, M., Storr, M. & Krause,
Replac. Ther.2, 36 (2016). (1991). B. Extended characterization of a new class of
48. Marques, I., Pinheiro, K., Carmo, L., Costa, M. & 75. Colton, C. K., Henderson, L. W., Ford, C. A. & Lysaght, membranes for blood purification: the high cut-off
Abensur, H. Anaphylactic reaction induced by M. J. Kinetics of hemodiafiltration. I. In vitro transport membranes. Int. J. Artif. Organs36, 455–463 (2013).
polysulfone/polyvinylpyrrolidone membrane in the characteristics of a hollow-fiber blood ultrafilter. 101. Rousseau-Gagnon, M., Agharazii, M., De Serres, S. A.
10th session of hemodialysis with the same dialyzer. J. Lab. Clin. Med.85, 355–371 (1975). & Desmeules, S. Effectiveness of haemodiafiltration
Hemodial. Int.15, 399–403 (2011). 76. Fiore, G. B., Guadagni, G., Lupi, A., Ricci, Z. & Ronco, with heat sterilized high-flux polyphenylene HF
49. Cheung, A. K. et al. Effects of hemodialyzer reuse on C. A new semiempirical mathematical model for dialyzer in reducing free light chains in patients with
clearances of urea and β2-microglobulin. J. Am. Soc. prediction of internal filtration in hollow fiber myeloma cast nephropathy. PLoS ONE10, e0140463
Nephrol.10, 117–127 (1999). hemodialyzers. Blood Purif.24, 555–568 (2006). (2015).
50. Kuwahara, T., Markert, M. & Wauters, J. Proteins 77. Fiore, G. B. & Ronco, C. Principles and practice of 102. Hutchison, C. A. et al. Immunoglobulin free light chain
adsorbed on hemodialysis membranes modulate internal hemodiafiltration. Contrib. Nephrol.158, levels and recovery from myeloma kidney on
neutrophil activation. Artif. Organs13, 427–431 177–184 (2007). treatment with chemotherapy and high cut-off
(1989). 78. Rangel, A. V. et al. Backfiltration: past, present, and haemodialysis. Nephrol. Dial. Transplant.27,
51. Eberhart, R. et al. Influence of endogenous albumin future. Contrib. Nephrol.175, 35–45 (2011). 3823–3828 (2012).
binding on blood-material interactions. Ann. NY Acad. 79. Lorenzin, A., Neri, M., Clark, W. R. & Ronco, C. 103. Hutchison, C. A. et al. Treatment of acute renal failure
Sci.516, 78–95 (1987). Experimental measurement of internal filtration rate secondary to multiple myeloma with chemotherapy

NATURE REvIEWS | NePHrology volume 14 | JUNE 2018 | 409


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Reviews

and extended high cut-off hemodialysis. Clin. J. Am. β2-microglobulin. Biochem. Biophys. Res. 146. Heyne, N. Expanded hemodialysis therapy:
Soc. Nephrol.4, 745–754 (2009). Commun.129, 701–706 (1985). prescription and delivery. Contrib. Nephrol.191,
104. Hutchison, C. A. et al. Serum free-light chain removal 125. Ahrenholz, P., Winkler, R. E., Ramlow, W., Tiess, M. & 153–157 (2017).
by high cutoff hemodialysis: optimizing removal and Muller, W. On-line hemodiafiltration with pre- and 147. Florens, N. & Julliard, L. Large middle molecular and
supportive care. Artif. Organs32, 910–917 (2008). post-dilution: a comparison of efficacy. Int. J. Artif. albumin removal: why should we not rest on our
105. Atan, R. et al. High cut-off hemofiltration versus Organs20, 81–90 (1997). laurels? Contrib. Nephrol.191, 178–187 (2017).
standard hemofiltration: effect on plasma cytokines. 126. Ono, M., Taoka, M., Takagi, T., Ogawa, H. & Saito, A. 148. Nagai, K. et al. Implications of albumin leakage for
Int. J. Artif. Organs39, 479–486 (2016). Comparison of type of on-line hemodiafiltration from survival in maintenance hemodialysis patients: a
106. Chelazzi, C. et al. Hemodialysis with high cut-off the standpoint of low-molecular weight protein 7-year observational study. Ther. Apher. Dial.21,
hemodialyzers in patients with multi-drug resistant removal. Contrib. Nephrol.108, 38–45 (1994). 378–386 (2017).
gram-negative sepsis and acute kidney injury: a 127. Colton, C. K. Analysis of membrane processes for 149. Balafa, O., Halbesma, N., Struijk, D. G., Dekker, F. W.
retrospective, case-control study. Blood Purif.42, blood purification. Blood Purif.5, 202–251 (1987). & Krediet, R. T. Peritoneal albumin and protein losses
186–193 (2016). 128. Ward, R. A., Schmidt, B., Hullin, J., Hillebrand, G. F. & do not predict outcome in peritoneal dialysis patients.
107. Villa, G. et al. Organ dysfunction during continuous Samtleben, W. A comparison of on-line Clin. J. Am. Soc. Nephrol.6, 561–566 (2011).
veno-venous high cut-off hemodialysis in patients with hemodiafiltration and high-flux hemodialylsis: a 150. Lavaud, S. et al. Optimal anticoagulation strategy in
septic acute kidney injury: a prospective observational prospective clinical study. J. Am. Soc. Nephrol.11, haemodialysis with heparin-coated polyacrylonitrile
study. PLoS ONE12, e0172039 (2017). 2344–2350 (2000). membrane. Nephrol. Dial. Transplant.18, 2097–2104
108. Boschetti-de-Fierro, A., Voigt, M., Storr, M. & Krause, 129. Ahrenholz, P. G., Winkler, R. E., Michelsen, A., Lang, (2003).
B. MCO membranes: enhanced selectivity in high-flux D. A. & Bowry, S. K. Dialysis membrane-dependent 151. Thomas, M., Valette, P., Mausset, A. L. & Déjardin, P.
class. Sci. Rep.5, 18448 (2015). removal of middle molecules during hemodiafiltration: High molecular weight kininogen adsorption on
109. Streicher, E. & Schneider, H. The development of a the beta2-microglobulin/albumin relationship. Clin. hemodialysis membranes: influence of pH and
polysulfone membrane. A new perspective in dialysis? Nephrol.62, 21–28 (2004). relationship with contact phase activation of blood
Contrib. Nephrol.46, 1–13 (1985). 130. Maduell, F. et al. Elimination of large uremic toxins by plasma. influence of pre-treatment with
110. Schneider, H. & Streicher, E. Clinical observations of a dialyzer specifically designed for high-volume poly(ethyleneimine). Int. J. Artif. Organs23,
the polyamide hollow-fiber hemofilter in hemofiltration convective therapies. Blood Purif.37, 125–130 20–26 (2000).
systems. J. Dial.1, 737–744 (1977). (2014). 152. Panagiotou, A. et al. Antioxidant dialytic approach
111. Ota, K. et al. Short-time hemodiafiltration using 131. Maduell, F. et al. High-efficiency postdilution online with vitamin E-coated membranes. Contrib.
polymethylmethacrylate hemodiafilter. Trans. Am. Soc. hemodiafiltration reduces all-cause mortality in Nephrol.171, 101–106 (2011).
Artif. Intern. Organs24, 454–457 (1978). hemodialysis patients. J. Am. Soc. Nephrol.24, 153. Cruz, D. N. et al. Effect of vitamin E-coated dialysis
112. Kaiser, J., Hagemann, J., Von Herrath, D. & Schaefer, 487–497 (2013). membranes on anemia in patients with chronic kidney
K. Different handling of beta2-microglobulin during 132. Canaud, B. & Bowry, S. K. Emerging clinical evidence disease: an Italian multicenter study. Int. J. Artif.
hemodialysis and hemofiltration. Nephron48, on online hemodialfiltration: does volume of Organs31, 545–552 (2008).
132–135 (1988). ultrafiltration matter? Blood Purif.35, 55–62 (2013). 154. Ghezzi, P. M. & Ronco, C. Excebrane:
113. Jorstad, S., Smeby, L., Balstad, T. & Wideroe, T. E. 133. Ok, E. et al. Mortality and cardiovascular events in hemocompatibility studies by the intradialytic
Removal, generation, and adsorption of beta-2- online haemodiafiltration (OL-HDF) compared with monitoring of oxygen saturation. Contrib.
microglobulin during hemofiltration with five different high-flux dialysis: results from the Turkish OL-HDF Nephrol.127, 177–191 (1999).
membranes. Blood Purif.6, 96–105 (1988). Study. Nephrol. Dial. Transplant.28, 192–202 (2013). 155. Kim, S. et al. Diffusive silicon nanopore membranes
114. Jindal, K. K., McDougall, J., Woods, B., Nowakowski, 134. Grooteman, M. P. et al. Effect of online for hemodialysis applications. PLoS ONE11,
L. & Goldstein, M. B. A study of the basic principles hemodiafiltration on all-cause mortality and e0159526 (2016).
determining the performance of several high-flux cardiovascular outcomes. J. Am. Soc. Nephrol.23, 156. Canaud, B., Bosc, J. Y., Leray, H. & Stec, F.
dialyzers. Am. J. Kidney Dis.14, 507–511 (1989). 1087–1096 (2012). Microbiological purity of dialysate for on-line
115. Zingraff, J. et al. Influence of haemodialysis 135. Fournier, A., Birmele, B., Francois, M., Prat, L. & substitution fluid preparation. Nephrol. Dial.
membranes on beta2-microglobulin kinetics: in vivo Halimi, J. M. Factors associated with albumin loss in Transplant.15 (Suppl. 2), 21–30 (2000).
and in vitro studies. Nephrol. Dial. Transplant.3, post-dilution hemodiafiltration. Int. J. Artif. Organs38,
284–290 (1988). 76–82 (2015). Acknowledgements
116. Klinke, B., Rockel, A., Abdelhamid, S., Fiegel, P. & 136. Ronco, C. Hemodiafiltration: technical and clinical The authors thank M. Storr (Baxter International), S. Bowry
Walb, D. Transmembrane transport and adsorption of issues. Blood Purif.40 (Suppl. 1), 2–11 (2015). (Fresenius Medical Care), R. Baldini (B. Braun Medical),
beta2-microglobulin during hemodialysis using 137. Panichi, V. et al. Divert to ULTRA: differences in L. Fecondini (Medica), L. Frattini (Medtronic), W. Oshihara
polysulfone, polyacrylonitrile, polymethylmethacrylate, infused volumes and clearance in two on-line (Toray Medical), A. Simionato (Asahi Kasei Medical) and
and cuprammonium rayon membranes. Int. J. Artif. hemodiafiltration treatments. Int. J. Artif. Organs35, S. Takashi (Nipro Corporation) for their invaluable comments
Organs12, 697–702 (1989). 435–443 (2012). and the generous provision of membrane images. The
117. Floege, J. et al. High flux synthetic versus cellulosic 138. Pedrini, L. A. et al. Transmembrane pressure authors recognize the seminal contributions to end-stage
membranes for beta2-microglobulin removal during modulation in high-volume mixed hemodiafiltration to renal disease therapy made by L. Henderson, who passed
hemodialysis, hemodiafiltration, and hemofiltration. optimize efficiency and minimize protein loss. Kidney away in 2017. He was a source of inspiration for many of us,
Nephrol. Dial. Transplant.4, 653–657 (1989). Int.69, 573–579 (2006). and we owe him a debt of gratitude for his exemplary
118. Naitoh, A., Tatsuguchi, T., Okada, M., Ohmura, T. & 139. Gayrard, N. et al. Consequences of increasing leadership in the field.
Sakai, K. Removal of beta2-microglobulin by diffusion convection onto patient care and protein removal in
is feasible using highly permeable dialysis membranes. hemodialysis. PLoS ONE12, e0171179 (2017). Author contributions
Trans. Am. Soc. Artif. Intern. Organs34, 630–634 140. Ficheux, A., Ronco, C., Brunet, P. & Argiles, A. The Both authors contributed to researching data for the article
(1988). ultrafiltration coefficient: this old ‘grand inconnu’ in and writing, reviewing and editing the article before
119. Mineshima, M. et al. Difference in beta2-microglobulin dialyisis. Nephrol. Dial. Transplant.30, 204–208 submission.
removal between cellulosic and synthetic polymer (2015).
membrane dialylzers. Trans. Am. Soc. Artif. Intern. 141. Maduell, F. et al. Mid-dilution hemodiafiltration: a Competing interests
Organs36, M643–M646 (1990). comparison with pre- and postdilution modes using C.R. has received consultant or honoraria fees from Astute
120. Ronco, C. et al. Beta2-microglobulin removal by synthetic the same polyphenylene membrane. Blood Purif.28, Medical, Ortho Clinical Diagnostics, Baxter International,
dialysis membranes: mechanisms and kinetics of the 268–274 (2009). Asahi Kasei Medical, General Electric, Jafron Biomedical,
molecule. Int. J. Artif. Organs20, 136–143 (1997). 142. Krieter, D. H. et al. Clinical cross-over comparison of Estor Medical and Toray Medical. W.R.C. was formerly
121. Floege, J. et al. Beta2-microglobulin kinetics during mid-dilution hemodiafiltration using a novel dialyzer employed by Baxter International, has received consulting
hemodialysis and hemofiltration. Nephrol. Dial. concept and post-dilution hemodiafiltration. Kidney fees from Baxter International and owns Baxter International
Transplant.1, 223–228 (1987). Int.67, 349–356 (2005). stock; he is also a consultant with Medtronic, Nikkiso
122. Maeda, N. et al. Performance and mechanism of beta- 143. Shinzato, T. & Maeda, K. Push/pull hemodiafiltration. America and Astute Medical.
2-microglobulin elimination with a new PAN hollow Contrib. Nephrol.158, 169–176 (2007).
fiber membrane. Jap. J. Artif. Organs17, 3–9 (1988). 144. Kirsch, A. H. et al. Performance of hemodialysis with Publisher’s note
123. Kachel, H., Altmeyer, P., Baldamus, C. & Koch, K. novel medium cut-off dialyzers. Nephrol. Dial. Springer Nature remains neutral with regard to jurisdictional
Deposition of an amyloid-like substance as a possible Transplant.32, 165–172 (2017). claims in published maps and institutional affiliations.
complication of regular dialysis treatment. Contrib. 145. Kirsch, A. H., Rosenkranz, A. R., Lyko, R. & Krieter,
Nephrol.36, 127–132 (1983). D. H. Effects of hemodialysis therapy using medium Supplementary information
124. Gejyo, F. et al. A new form of amyloid protein cut-off membranes on middle molecules. Contrib. Supplementary information is available for this paper at
associated with chronic hemodialysis was identified as Nephrol.191, 158–167 (2017). https://doi.org/10.1038/s41581-018-0002-x.

410 | JUNE 2018 | volume 14 www.nature.com/nrneph


© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

You might also like