**4. Physical structure of dialysis membrane**

### **4.1. Homogeneous and asymmetry membrane**

Physical structures can be demonstrated in the following two ways, i.e., microscopic view analysis and a theoretical analysis based on mathematical models. Microscopic views are usually taken by a scanning electron microscope (SEM). Recently the microscope technology has been advancing drastically and a field-emission SEM (FE-SEM) that has much higher resolutions can be utilized widely.

**Figure 3.** A cross-sectional view of EVAL hollow fiber membrane (Asahi-Kasei, Tokyo, Japan) taken by FE-SEM.

Figure 3-1. A cross-sectional view of EVAL hollow fiber membrane (Asahi-Kasei, Tokyo, Japan) taken by FE-SEM Figure 3 is a FE-SEM of intersection of EVAL membrane (Asahi-Kasei). It is entirely a dense membrane and the entire thickness contributes to the transport resistance for solutes and water. Membranes of this kind are usually called "homogeneous." Besides EVAL, PMMA, and AN-69®, most cellulosic membranes are homogeneous. Figure 4 shows a cross-sectional view of PSf membrane (Toray). One should realize that a dense thin layer exists on the inner surface of the membrane, called "skin layer" from which the density is gradually decreasing in the radial direction. Since most part excluding the skin layer is known to have little resistance for solute and water transport, it is called the "support layer" (Figure 5). The support layer, however, has an important role for the membrane to have enough mechanical strength with little resistance for transport. Membranes of this kind are called "asymmetry." Most synthetic polymeric membranes (except for PMMA, EVAL, and AN-69®) are asymmetry. In general, although the physical thickness of synthetic polymeric membranes is thicker (approximately 35 µm) than that of cellulosic membranes (approximately 15 µm), the thickness that contributes to the separation (*Δx*) of the former is approximately 0.5-2 µm that is much thinner than the latter. As mentioned before, synthetic polymeric membranes are main stream these days because much higher solute and hydraulic permeabilities are achieved with the thinner *Δx*.

**Figure 4.** A cross-sectional view of PSf hollow fiber membrane (Toray, Tokyo, Japan) taken by FE-SEM.

**Figure 5.** Cross-sectional views of dialysis membranes.

weight of PVP are important as well as the amount of PVP used in the membrane. Moreover, PVP may be cross-linked together and to the main material of the membrane by irradiating gamma-ray in the final sterilization process. With this procedure, PVP should be tightly attached together and/or on the membrane that does not allow PVP to behave as a "cushion"

Acrylic acid is specifically chosen for polyacrylonitrile (PAN, Asahi Kasei, different from AN-69®) as a hydrophilic agent, whereas no hydrophilic agent is used in PMMA, EVAL and the original PEPA in which micro-layer separation technology plays a significant role in casting

Physical structures can be demonstrated in the following two ways, i.e., microscopic view analysis and a theoretical analysis based on mathematical models. Microscopic views are usually taken by a scanning electron microscope (SEM). Recently the microscope technology has been advancing drastically and a field-emission SEM (FE-SEM) that has much higher

> Figure 3-1. A cross-sectional view of EVAL hollow fiber membrane (Asahi-Kasei, Tokyo, Japan) taken by FE-SEM

Figure 3 is a FE-SEM of intersection of EVAL membrane (Asahi-Kasei). It is entirely a dense membrane and the entire thickness contributes to the transport resistance for solutes and water. Membranes of this kind are usually called "homogeneous." Besides EVAL, PMMA, and AN-69®, most cellulosic membranes are homogeneous. Figure 4 shows a cross-sectional view of PSf membrane (Toray). One should realize that a dense thin layer exists on the inner surface of the membrane, called "skin layer" from which the density is gradually decreasing in the radial direction. Since most part excluding the skin layer is known to have little resistance for solute and water transport, it is called the "support layer" (Figure 5). The support layer, however, has an important role for the membrane to have enough mechanical strength with

**Figure 3.** A cross-sectional view of EVAL hollow fiber membrane (Asahi-Kasei, Tokyo, Japan) taken by FE-SEM.

(cushion effect) to the blood corpuscles [5].

**4. Physical structure of dialysis membrane**

**4.1. Homogeneous and asymmetry membrane**

resolutions can be utilized widely.

procedure.

168 Updates in Hemodialysis

#### **4.2. Pore theory**

The pore theory is often used to analyze and to design physical structures of the membrane. The original pore theory was introduced by Pappenheimer *et al.* [6] to analyze the Glomerular filtration in the living kidney (Figure 6), and was later modified by Verniory *et al.* [7], intro‐ ducing steric hindrance effect. Sakai [8] further modified the model by introducing the tortuosity for transporting across the membrane. Followings are the equations for modified pore theory.

Figure 3-4. Pore theory (pore diffusion model) Assuming pores whose radius is uniformly *r*<sup>p</sup> [m] with a **Figure 6.** Pore theory (pore diffusion model). Assuming pores whose radius is uniformly *r*p [m] with a membrane thickness of *Δx* [m], through which a solute of interest whose radius is *r* s [m] is passing.

membrane thickness of

$$k\_{\rm M} = D\_{\rm w} \times f\left(q\right) \times S\_{D} \times \left(\frac{A\_{\rm k}}{\tau \times \Delta \times \tau}\right) \tag{3}$$

$$L\_{\
u} = \left(\frac{r\_p}{8\mu}\right) \times \left(\frac{A\_k}{\tau \times \Delta x}\right) \tag{4}$$

*x* [m], through which a solute of interest

$$q = \frac{r\_s}{r\_p} \tag{5}$$

$$\sigma = 1 \, \text{--} \, \text{g} \, (\text{q}) \times \text{S}\_{\text{F}} \tag{6}$$

$$S\_{\rm D} = (1 - q)^2 \tag{7}$$

$$S\_{\rm F} = 2(1 - q)^2 - (1 - q)^4 \tag{8}$$

$$f\left(q\right) = \frac{1 \cdot 2.1050q + 2.0865q^3 \cdot 1.7068q^5 + 0.72603q^6}{1 \cdot 0.75857q^5} \tag{9}$$

Dialysis Membranes — Physicochemical Structures and Features http://dx.doi.org/10.5772/59430 171

$$\log\left(q\right) = \frac{1 \cdot \left(2 \cdot 3\right) q^{\frac{2}{3}} \cdot 0.20217 q^{\frac{5}{3}}}{1 \cdot 0.75857 q^{\frac{5}{3}}}\tag{10}$$

where *k* M is the membrane permeability [m/s] (see also section 1), *D* <sup>w</sup> is the diffusion coefficient for the solute of interest in pure water [m2 /s], *A* <sup>k</sup> is the surface porosity of the membrane [-], *Δx* is the membrane thickness that contributes to the transport resistance [m], *r* s is the solute radius [m], *r* <sup>p</sup> is the pore radius of the membrane [m], *L* <sup>p</sup> is the hydraulic permeability of the membrane [m2 s/kg], *σ* is the Staverman's reflection coefficient [-], *τ* is the tortuosity of the membrane [-], *q* is the ratio of *r* s to *r* p [-], *S* D, *S* F, *f*(*q*), and *g*(*q*) are the dimensionless stereo correction factors defined as functions of *q*. The pore theory can be applied to the situation in which *q* < 0.8 is satisfied.

**4.2. Pore theory**

170 Updates in Hemodialysis

pore theory.

*r*s

Solute of interest

The pore theory is often used to analyze and to design physical structures of the membrane. The original pore theory was introduced by Pappenheimer *et al.* [6] to analyze the Glomerular filtration in the living kidney (Figure 6), and was later modified by Verniory *et al.* [7], intro‐ ducing steric hindrance effect. Sakai [8] further modified the model by introducing the tortuosity for transporting across the membrane. Followings are the equations for modified

Separation membrane

Figure 3-4. Pore theory (pore diffusion model)

thickness of *Δx* [m], through which a solute of interest whose radius is *r* s [m] is passing.

whose radius is *r*<sup>s</sup> [m] is passing.

*<sup>L</sup>* <sup>p</sup> = ( *<sup>r</sup>*<sup>p</sup> 2 <sup>8</sup>*<sup>μ</sup>* ) ×( *<sup>A</sup>*<sup>k</sup>

membrane thickness of

Assuming pores whose radius is uniformly *r*<sup>p</sup> [m] with a

**Figure 6.** Pore theory (pore diffusion model). Assuming pores whose radius is uniformly *r*p [m] with a membrane

*x* [m], through which a solute of interest

*σ* =1- *g*(*q*)×*S*<sup>F</sup> (6)

*<sup>S</sup>*<sup>D</sup> =(1 - *<sup>q</sup>*)<sup>2</sup> (7)

1 - 0.75857*<sup>q</sup>* <sup>5</sup> (9)

*<sup>S</sup>*<sup>F</sup> =2(1 - *<sup>q</sup>*)<sup>2</sup> - (1 - *<sup>q</sup>*)<sup>4</sup> (8)

*<sup>τ</sup>* <sup>×</sup> <sup>∆</sup> *<sup>x</sup>* ) (3)

*<sup>τ</sup>* <sup>×</sup> <sup>∆</sup> *<sup>x</sup>* ) (4)

*x: M*embrane thickness that contributes to the separation [m]

(5)

*r*s

*<sup>k</sup>*<sup>M</sup> <sup>=</sup>*D*<sup>w</sup> <sup>×</sup> *<sup>f</sup>* (*q*)×*<sup>S</sup> <sup>D</sup>* ×( *<sup>A</sup>*<sup>k</sup>

*<sup>q</sup>* <sup>=</sup>*r*<sup>s</sup> *r*p

*<sup>f</sup>* (*q*)= 1 - 2.1050*<sup>q</sup>* + 2.0865*<sup>q</sup>* <sup>3</sup> - 1.7068*<sup>q</sup>* <sup>5</sup> + 0.72603*<sup>q</sup>* <sup>6</sup>

*r*p

Pore

From Eqs.(3) and (4), it is clear that *A* k/(*τ Δx*) is an important factor both for solute and water transport because both *k* M and *L* p include this value. Figure 7 shows two examples of *L* [m] x *L* [m] portions of the membrane, i.e., membrane (A) with four pores with the same radius of *a* [m], and membrane (B) with one pore with a radius of 2*a*. Then the surface porosity can be calculated, respectively for membranes (A) and (B) with subscripts (A) and (B), i.e.,

$$\begin{aligned} A\_{\mathbf{k}(\mathbf{A})} &= \frac{4 \times \pi a^2}{L^{-\frac{2}{2}}} = \frac{4 \pi a^2}{L^{-\frac{2}{2}}} \\\\ A\_{\mathbf{k}(\mathbf{B})} &= \frac{\pi (2a)^2}{L^{-\frac{2}{2}}} = \frac{4 \pi a^2}{L^{-\frac{2}{2}}} \end{aligned}$$

∴ *A*k(B)=*A*k(A)

Figure 3-5. Portions of two modeled membranes with

the same surface porosity

**Figure 7.** Portions of two modeled membranes with the same surface porosity.

Then one would realize that membranes (A) and (B) have the same surface porosities, although the situations are quite different in terms of the pore diameter.

Example)

Compare the two membranes (A) and (B) that have the same surface porosity (Figure 7), tortuosity and the thickness in terms of


under the following two conditions


Solution) As stated above, *A <sup>k</sup>*, *τ*, and *Δx* are the same in two membranes, *A <sup>k</sup>*/(*τ Δx*) is just a constant.

**i.** Recalling Eq. (4) to get,

$$\begin{array}{c} L\_{\mathsf{p}(\mathsf{A})} = \left(\frac{\mathsf{a}^{2}}{8\mu}\right) \times \left(\frac{A\_{\mathsf{k}}}{\mathsf{r} \times \Delta \mathsf{x}}\right) \\\\ L\_{\mathsf{p}(\mathsf{B})} = \left(\frac{(2\mathsf{a})^{2}}{8\mu}\right) \times \left(\frac{A\_{\mathsf{k}}}{\mathsf{r} \times \Delta \mathsf{x}}\right) = \left(\frac{4\mathsf{a}^{2}}{8\mu}\right) \times \left(\frac{A\_{\mathsf{k}}}{\mathsf{r} \times \Delta \mathsf{x}}\right) \\\\ \vdots \\\\ L\_{\mathsf{p}(\mathsf{B})} = \mathsf{4} \ L\_{\mathsf{p}(\mathsf{A})} \end{array}$$

Therefore, the membrane (B) has four times higher hydraulic permeability than the membrane (A).

**ii.** Since *q*=0 may reasonably be applied in this case, recalling Eqs.(7)-(10) to get,

*S*D=*S*F=*f*(*q*)=*g*(*q*)=1

in both membranes (A) and (B). Therefore Eq.(3) may be simplified as follows,

$$k\_{\mathbf{M(A)}} = k\_{\mathbf{M(B)}} = D\_w \times (1) \times (1) \times \left(\frac{A\_k}{\tau \times \Delta \ x}\right) = D\_w \times \left(\frac{A\_k}{\tau \times \Delta \ x}\right)$$

Consequently, there is no difference between membranes (A) and (B) in terms of transport of small solutes.

**iii.** Recalling Eq.(5),

$$q\_{\rm (A)} = \frac{r\_\ast}{r\_\mathbb{P}} = \frac{a/3}{a} = \frac{1}{3}$$

$$q\_{\rm (B)} = \frac{r\_\ast}{r\_\mathbb{P}} = \frac{a/3}{2a} = \frac{1}{6}$$

Then recalling Eqs.(7) and (9) with *q* values calculated above,

$$S\_{D(\text{A})} = (1 \cdot q\_{(\text{A})})^2 = 0.8889$$

$$S\_{D(\text{B})} = (1 \cdot q\_{(\text{B})})^2 = 0.9722$$

$$f(q\_{(\text{A})}) = 0.3707$$

$$f(q\_{(\text{B})}) = 0.6587$$

Then from Eq.(3),

Then one would realize that membranes (A) and (B) have the same surface porosities, although

Compare the two membranes (A) and (B) that have the same surface porosity (Figure 7),

Solution) As stated above, *A <sup>k</sup>*, *τ*, and *Δx* are the same in two membranes, *A <sup>k</sup>*/(*τ Δx*) is just a

<sup>8</sup>*<sup>μ</sup>* ) ×( *Ak <sup>τ</sup>* <sup>×</sup> <sup>∆</sup> *<sup>x</sup>* )

*<sup>τ</sup>* <sup>×</sup> <sup>∆</sup> *<sup>x</sup>* ) <sup>=</sup> ( <sup>4</sup>*a*<sup>2</sup>

∴ *L*p(B)=4 *L*p(A)

Therefore, the membrane (B) has four times higher hydraulic permeability than the membrane

Consequently, there is no difference between membranes (A) and (B) in terms of transport of

*<sup>r</sup>*<sup>p</sup> <sup>=</sup> *<sup>a</sup>* / <sup>3</sup> *<sup>a</sup>* <sup>=</sup> <sup>1</sup> 3

*<sup>q</sup>*(A)= *<sup>r</sup>*<sup>s</sup>

**ii.** Since *q*=0 may reasonably be applied in this case, recalling Eqs.(7)-(10) to get,

in both membranes (A) and (B). Therefore Eq.(3) may be simplified as follows,

*<sup>k</sup>*M(A)=*k*M(B)=*D*<sup>w</sup> ×(1)×(1)×( *<sup>A</sup>*<sup>k</sup>

<sup>8</sup>*<sup>μ</sup>* ) ×( *Ak <sup>τ</sup>* <sup>×</sup> <sup>∆</sup> *<sup>x</sup>* )

*<sup>τ</sup>* <sup>×</sup> <sup>∆</sup> *<sup>x</sup>* ) <sup>=</sup>*Dw* ×( *<sup>A</sup>*<sup>k</sup>

*<sup>τ</sup>* <sup>×</sup> <sup>∆</sup> *<sup>x</sup>* )

*<sup>L</sup>* p(A)= ( *<sup>a</sup>* <sup>2</sup>

<sup>8</sup>*<sup>μ</sup>* ) ×( *Ak*

the situations are quite different in terms of the pore diameter.

under the following two conditions

**a.** *r*s is negligibly small compared with *a*

*<sup>L</sup>* p(B)= ( (2*a*)2

tortuosity and the thickness in terms of

**i.** hydraulic permeability

**ii.** solute permeability

**b.** *r*s*=a*/3

**i.** Recalling Eq. (4) to get,

Example)

172 Updates in Hemodialysis

constant.

(A).

*S*D=*S*F=*f*(*q*)=*g*(*q*)=1

small solutes.

**iii.** Recalling Eq.(5),

$$k\_{\text{M(A)}} = D\_{\text{w}} \times (0.3707) \times (0.8889) \times \left(\frac{A\_{\text{k}}}{\tau \times \Delta x}\right) = 0.3295 \times D\_{\text{w}} \times \left(\frac{A\_{\text{k}}}{\tau \times \Delta x}\right)$$

$$k\_{\text{M(B)}} = D\_{\text{w}} \times (0.6587) \times (0.9722) \times \left(\frac{A\_{\text{k}}}{\tau \times \Delta x}\right) = 0.6404 \times D\_{\text{w}} \times \left(\frac{A\_{\text{k}}}{\tau \times \Delta x}\right)$$

$$\therefore \quad k\_{\text{M(B)}} = 1.94 \text{ } k\_{\text{M(A)}}$$

Finally one would conclude that the membrane (B) has almost two times higher solute permeability than the membrane (A) for those solutes whose *r* <sup>s</sup> *=a*/3.

Chemical characteristic determines the hydrophilicity and hydrophobicity of the material, whereas physical structure determines the pore sizes as well as the thickness that contributes to the transport resistance. Therefore, both chemical and physical features are important for designing dialysis membrane.

#### **5. Performance of dialysis membrane**

In this section, we discuss the performances under *in vitro* ultrafiltration experiments and those under on-line HDF in clinical situations because the former is suited for evaluation of maximal performance of the membrane and the latter takes a responsibility of the real performance under advanced clinical situations.

#### **5.1. Aqueous** *in vitro* **ultrafiltration experiment**

Six filters (dialyzers), one with PSf (PS-1.6UW, Fresenius-Kawasumi Co., Tokyo, Japan) and other five with PEPA (Nikkiso Co., Tokyo, Japan) were investigated (Table 2). Since both PSf and PEPA are hydrophobic in nature, these membranes include PVP for anti-thrombosis purpose, except for one dialyzer that includes PEPA membrane with no additives (FLX). Amount of PVP used in the membrane is semi-quantitatively shown as (+++), (++), (+), and (-), respectively for "most", "much", "small" and "none".


**Table 2.** Technical specification of investigated ultrafilters

The time courses of the sieving coefficient (*s.c. <sup>4</sup>*) [9, 10] for albumin of PS-1.6UW dialyzer were shown in Figure 8. Strong time-dependent patterns were found with peak values approxi‐ mately at 10 minutes after starting experiments. The lower the albumin concentration, the higher the *s.c. <sup>4</sup>* values was found with longer time for achieving steady-state.

Figure 4-1. Time courses of the sieving coefficient for albumin under various concentrations of albumin in PS-1.6UW (PSf membrane) **Figure 8.** Time courses of the sieving coefficient for albumin under various concentrations of albumin in PS-1.6UW (PSf membrane) *Q*Bi=200 mL/min, *Q*F=10 mL/min, Volume of test sol'n=2.0 L.

*QBi* = 200 mL/min, *QF* = 10 mL/min, Volume of test sol'n = 2.0 L The time courses of *s.c. <sup>4</sup>* for albumin of three PEPA filters with albumin concentration of 3.64 mg/mL are shown in Figure 9. The *s.c. <sup>4</sup>* gradually increased in these PEPA with PVP(-) or PVP(+) and never took peak values. Membranes used in FLX and FDX basically have the same pore sizes and the only difference is that the latter contained PVP, which concludes that PVP directly influences the membrane transport of albumin. By enlarging the pore diameter by approximately 5 % in FDY with the same PVP content, the *s.c. <sup>4</sup>* increased with the enlargement accordingly.

**#**

**name of products**

174 Updates in Hemodialysis

**abbreviated names**

**Table 2.** Technical specification of investigated ultrafilters

**0.00**

**0.02**

**0.04**

**0.06**

*s.c.4* **for albumin [-]**

**0.08**

**0.10**

**Surface area [m2 ]**

**membrane**

<sup>2</sup> FLX-15GW FLX 1.5 PEPA PVP (-) standard Nikkiso Co.,

<sup>3</sup> FDX-15GW FDX 1.5 PEPA PVP (+) standard Nikkiso Co.,

<sup>4</sup> FDY-15GW FDY 1.5 PEPA PVP (+) larger Nikkiso Co.,

<sup>5</sup> FDX-150GW new FDX 1.5 PEPA PVP (++) standard Nikkiso Co.,

<sup>6</sup> FDY-150GW new FDY 1.5 PEPA PVP (++) larger Nikkiso Co.,

The time courses of the sieving coefficient (*s.c. <sup>4</sup>*) [9, 10] for albumin of PS-1.6UW dialyzer were shown in Figure 8. Strong time-dependent patterns were found with peak values approxi‐ mately at 10 minutes after starting experiments. The lower the albumin concentration, the

Figure 4-1. Time courses of the sieving coefficient for albumin under

(PSf membrane) *Q*Bi=200 mL/min, *Q*F=10 mL/min, Volume of test sol'n=2.0 L.

**Figure 8.** Time courses of the sieving coefficient for albumin under various concentrations of albumin in PS-1.6UW

The time courses of *s.c. <sup>4</sup>* for albumin of three PEPA filters with albumin concentration of 3.64 mg/mL are shown in Figure 9. The *s.c. <sup>4</sup>* gradually increased in these PEPA with PVP(-) or

various concentrations of albumin in PS-1.6UW (PSf membrane)

**0 100 200 300 400 500 600 Time [min]**

*QBi* = 200 mL/min, *QF* = 10 mL/min, Volume of test sol'n = 2.0 L

higher the *s.c. <sup>4</sup>* values was found with longer time for achieving steady-state.

1 PS-1.6UW PS 1.6 PSf PVP (+++) (Not available)

**materials hydrophilic agent pore size info**

**membrane make**

Fresenius Medical Care, Badhonburg, Germany

Tokyo, Japan

Tokyo, Japan

Tokyo, Japan

Tokyo, Japan

Tokyo, Japan

**0.81mg/mL x 1.0 1.21mg/mL x 1.5 2.42mg/mL x 3.0 3.64mg/mL x 4.5**

Figure 4-2. Time courses of the sieving coefficient for albumin under a fixed albumin concentration (3.64 mg/mL) in three PEPA membrane dialyzers Curves are different from the ones found **Figure 9.** Time courses of the sieving coefficient for albumin under a fixed albumin concentration (3.64 mg/mL) in three PEPA membrane dialyzers Curves are different from the ones found with PSf membrane. *Q*Bi=200 mL/min, *Q*F=10 mL/min, Volume of test sol'n=2.0 L.

with PSf membrane. *QBi* = 200 mL/min, *QF* = 10 mL/min, Volume of test sol'n = 2.0 L The time courses of *s.c. <sup>4</sup>* for albumin of the latest version of PEPA dialyzers are depicted in Figure 10 for albumin concentration of 3.64 mg/mL. It should be noted that the peak values were found in new PEPA membranes that included increased amount of PVP at 6 minutes after starting the experiments. Moreover, time dependent pattern of these curves are different from the ones shown in Figure 9 and are similar to those found with PSf membrane in Figure 8. Then it may be concluded that the time course of *s.c. <sup>4</sup>* for albumin is strongly dependent on the amount of PVP included in the membrane and not on the main material of the membrane.

Since the albumin concentrations of the test solutions were lower by the factor of 1/30-1/10 to the standard albumin concentration in human blood (3.6 – 4.0 g/dL), *s.c. <sup>4</sup>* values for albumin shown above do not directly correspond to the clinical results. One should, however, need to consider that the membrane separation characteristics depend on the pore diameter, amount of hydrophilic agent as well as experimental conditions [11].

#### **5.2. Clinical performance of super-high flux dialyzers/diafilters**

According to the Japanese reimbursement system, all the commercial dialyzers are classified into five categories in accordance with the clearances for β2-microglobulin (β2-MG, MW 11800) under *Q <sup>B</sup>*=200 mL/min, *Q <sup>D</sup>*=500 mL/min for dialyzers with surface area of 1.5 m2 (Table 3). Classes IV and V dialyzers, clearances for β2-MG greater or equal to 50 and 70 mL/min,

Figure 4-3. Time courses of the sieving coefficient for albumin under a fixed albumin concentration (3.64 mg/mL) in two new PEPA membrane dialyzers Curves are similar to the ones found with **Figure 10.** Time courses of the sieving coefficient for albumin under a fixed albumin concentration (3.64 mg/mL) in two new PEPA membrane dialyzers Curves are similar to the ones found with PSf membrane. *Q*Bi=200 mL/min, *Q*F=10 mL/min, Volume of test sol'n=2.0 L.

respectively, are the "super-high flux" models and more than 95 % of Japanese dialysis patients are treated with dialyzers of this kind [12]. These dialyzers had been used also for on-line hemodiafiltration (HDF) with considerable amount of albumin removal (> 3 g/treatment) until 2010 before on-line HDF has been officially announced to be included in the reimbursement system. Table 4-2. Classification of dialyzers in Japanese reimbursement system *QBi* = 200 mL/min, *QF* = 10 mL/min, Volume of test sol'n = 2.0 L


1) Flow conditions: *QB* = 200 mL/min, *QD* = 500 mL/min, *QF* = 10 mL/min/m2. *2) A0* = 1.5 m2 1. Flow conditions: *Q <sup>B</sup>*=200 mL/min, *Q <sup>D</sup>*=500 mL/min, *Q <sup>F</sup>*=10 mL/min/m2 .

PSf membrane

#### *3) If A0* is NOT1.5 m2, use of the closest model is recommended. Clearance for 2. *A <sup>0</sup>*=1.5 m2

2-MG under *A0* = 1.5 m2 may be estimated by using the performance evaluation equations with *KoA* as a constant. 3. *If A <sup>0</sup>* is NOT1.5 m2 , use of the closest model is recommended. Clearance for β2-MG under *A <sup>0</sup>*=1.5 m2 may be estimated by using the performance evaluation equations with *K <sup>o</sup> A* as a constant.

**Table 3.** Classification of dialyzers in Japanese reimbursement system

Although 99 uremic toxins are compiled by Vanholder [13], clinicians and researchers have different opinions on which solutes to be removed or up to how much albumin may be leaked out. Figure 11 shows the relationship between the reduction rate of β2-MG and albumin loss taken with various dialyzers in different modalities. Only a limited increase in β2-MG reduction was found with the increase of albumin removal. Therefore β2-MG removal may not be directly related to convection transport when super-high flux dialyzers are used. In other words, super high-flux dialyzers are the ones in which β2-MG removal does not correlate with the amount of albumin loss or the convection transport.

respectively, are the "super-high flux" models and more than 95 % of Japanese dialysis patients are treated with dialyzers of this kind [12]. These dialyzers had been used also for on-line hemodiafiltration (HDF) with considerable amount of albumin removal (> 3 g/treatment) until 2010 before on-line HDF has been officially announced to be included in the reimbursement

Figure 4-3. Time courses of the sieving coefficient for albumin under a fixed albumin concentration (3.64 mg/mL) in two new PEPA membrane dialyzers Curves are similar to the ones found with

**Figure 10.** Time courses of the sieving coefficient for albumin under a fixed albumin concentration (3.64 mg/mL) in two new PEPA membrane dialyzers Curves are similar to the ones found with PSf membrane. *Q*Bi=200 mL/min, *Q*F=10

**0 100 200 300 400 500 600 Time [min]**

*QBi* = 200 mL/min, *QF* = 10 mL/min, Volume of test sol'n = 2.0 L

**new FDX new FDY**

I < 10 low

V >= 70 high

1) Flow conditions: *QB* = 200 mL/min, *QD* = 500 mL/min, *QF* = 10 mL/min/m2.

*3) If A0* is NOT1.5 m2, use of the closest model is recommended. Clearance for 2-MG under *A0* = 1.5 m2 may be estimated by using the performance

[mL/min] Reimbursement

.

, use of the closest model is recommended. Clearance for β2-MG under *A <sup>0</sup>*=1.5 m2 may be estimated

system. Table 4-2. Classification of dialyzers in Japanese reimbursement system

PSf membrane

Class 2-MG clearance

II >=10~< 30

III >=30~< 50

IV >=50~< 70

evaluation equations with *KoA* as a constant.

by using the performance evaluation equations with *K <sup>o</sup> A* as a constant.

**Table 3.** Classification of dialyzers in Japanese reimbursement system

1. Flow conditions: *Q <sup>B</sup>*=200 mL/min, *Q <sup>D</sup>*=500 mL/min, *Q <sup>F</sup>*=10 mL/min/m2

*2) A0* = 1.5 m2

**0.00**

mL/min, Volume of test sol'n=2.0 L.

**0.02**

**0.04**

*s.c.4* **for albumin [-]**

176 Updates in Hemodialysis

**0.06**

**0.08**

**0.10**

2. *A <sup>0</sup>*=1.5 m2

3. *If A <sup>0</sup>* is NOT1.5 m2

2-microglobulin (MW: 11,800) and albumin loss **Figure 11.** Relationship between the reduction rate of β2-microglobulin (MW: 11,800) and albumin loss.

Figure 4-4. Relationship between the reduction rate of

Figure 12 shows the same relationship between the reduction rate of α1-microglobulin (α1-MG, MW 33,000) and albumin loss. Up to albumin loss of 3 g/session, almost linear relationship was observed, meaning that it may not be possible to remove α1-MG without removing albumin, although the molecular weight of albumin is twice as large as that of α1-MG. There is no such article that reports α1-MG is toxic; moreover, α1-MG is not even included in Vanholder's list [13]. We, however, experienced fairly good number of patients who have become better with albumin loss of 3 g or more for bone pain, shoulder pain, and improvement of fingertip power, and 5 g or more for finger numbness, restless legs syndrome. Therefore α1- MG may be a possible surrogate marker of HDF treatment for those who have symptoms with normal HD therapy. Relief of clinical symptoms with various treatment modalities is sum‐ marized in Figure 13.

Figure 4-5. Relationship between the reduction rate of **Figure 12.** Relationship between the reduction rate of α1-microglobulin (MW: 33,000) and albumin loss.

α1-microglobulin (MW: 33,000) and albumin loss

Figure 4-6. Relief of clinical symptoms by employing various

protein-losing treatment modes. **Figure 13.** Relief of clinical symptoms by employing various protein-losing treatment modes.

### **5.3. Consideration of on-line HDF**

Figure 4-5. Relationship between the reduction rate of

Fingertip pinch power Shoulder pain

**Figure 12.** Relationship between the reduction rate of α1-microglobulin (MW: 33,000) and albumin loss.

**FX FDX FDY**

Reduction reta

0

10

20

Reduction rate of α1-MG

30

(%)

40

50

 of

α1-MG (%)

178 Updates in Hemodialysis

α1-microglobulin (MW: 33,000) and albumin loss

Figure 4-6. Relief of clinical symptoms by employing various

Albumin loss [mg]

APS-21E(50L QB200)

Finger numbness Restless leg syndrome

TS-2.1UL(3.5h50L QB240) TS-2.1UL(50L QB240) FDY-210GW(3.5h50L QB240) FDY-210GW(50L QB240) FDY-210GW(HD QB200) FDY-210GW(50L QB200) FDY-250GW(HD QB200) FDY-250GW(HD QB240) FDY-250GW(40L QB200) FDY-250GW(50L QB200) FDY-250GW(50L QB240) FDY-250GW(60L QB200) APS-25SA(HD QB200) APS-25SA(50L QB200) APS-25SA(50L QB240) APS-21E(HD QB200) APS-25SA(post12L QB200) FDY-210GW(post10L QB200)

0 1000 2000 3000 4000 5000 6000

protein-losing treatment modes.

**Figure 13.** Relief of clinical symptoms by employing various protein-losing treatment modes.

Pruritus, Prevention of symptoms

Dialysis induced hypotension (Italian study)

0 1000 2000 3000 4000 5000 6000 7000

Albumin loss [mg]

On-line HDF is mostly performed in post-dilution mode with high *Q <sup>B</sup>* (400 mL/min) in European countries, whereas that is mostly performed in pre-dilution mode with limited *Q <sup>B</sup>* (250 mL/min) in Japan. Diafilters preferred in post-dilution and pre-dilution HDF must be designed under different concepts. Membrane for the post-dilution HDF requires a limited permeability for albumin, otherwise unexpected large amount of albumin may be leaked out. Therefore relatively large surface area is preferred for achieving large amount of fluid exchange (20 L/session). Since usually higher clearances are expected with post-dilution HDF, membrane for the pre-dilution HDF prefers higher solute permeability that may allow much albumin to penetrate across the membrane. Amount of albumin loss, however, may be relatively easily controlled by changing the amount of ultrafiltration that is usually around 60 L/session. In the recent market, since diafilters specifically designed for either post-dilution or pre-dilution are available, choice of diafilters must be paid much attention not only for effective treatment but also for safety. Moreover, a proposal of technical specifications for the future diafilters is also reported [14].

Many randomized control studies have been done in order to verify superiority or better outcomes of on-line HDF [15-[19]; however, we have not yet come into a conclusion that states on-line HDF is better than other treatment modalities. These studies showed that on-line HDF was at least better than low-flux HD; however, the difference between on-line HDF and highflux HD was ambiguous [18, 19], in terms of survival rate within a study period of three years or so. Post–hoc analyses and sub-analyses of those studies showed superiority of on-line HDF with large amount of fluid exchange (at least > 15 L) to other treatment modalities in terms of dialysis-induced hypotension, reaction to ESR medications, as well as survival rate. Among them, the ESHOL study [20] greatly encouraged patients on dialysis as well as medical staffs in which on-line HDF showed better clinical outcomes in all the end points than high-flux HD. Many debates, however, still continues also elsewhere including in Japan where the number of patients on on-line HDF is rapidly growing and exceeded 10 % of the total patients [21].
