Solute uptake by sorbents long has been considered a possible treatment for uremia (1–4), but the extent to which sorbents can increase solute clearances above the levels that are obtained by hemodialysis remains uncertain. In most sorbent systems, blood is separated from the sorbent by a semipermeable membrane to avoid the adverse effects of direct blood–sorbent contact. Solutes therefore must diffuse through the membrane before binding to the sorbent. Our study sought to define better the capacity for solute clearance of this two-step process. We first modeled the effect of adding a sorbent to the dialysate when dialysis is carried out by countercurrent flow of plasma and dialysate on opposite sides of a semipermeable membrane. The predicted effect of a sorbent in this case is to lower the free solute concentration in the dialysate, and adding a sorbent to the dialysate can increase solute clearances to the same extent as increasing dialysate flow (Qd).
One potential application of sorbents is to increase the clearance of solutes that bind to plasma proteins. These solutes are cleared poorly by conventional hemodialysis (5–9). Moreover, although not conclusive, a variety of evidence suggests that protein-bound solutes contribute to uremic toxicity (5,6,10–12), and mathematical modeling predicts that the clearance of protein-bound solutes can be increased more than the clearance of unbound solutes by increasing Qd or adding a sorbent to the dialysate. During conventional hemodialysis, values for Qd and the dialyzer mass transfer area coefficient (KoA) are set to achieve high single-pass clearance of small unbound solutes such as urea. Lowering the free solute concentration in the dialysate by increasing Qd or adding a sorbent cannot raise the clearance of these solutes much further, but the case is different for solutes that bind avidly to plasma proteins. The clearance of such solutes during conventional dialysis is low because only the free solute fraction is available for diffusion across the membrane. In this setting, modeling predicts that as long as KoA is large enough, solute clearances can be increased greatly by lowering the free solute levels in the dialysate. Our study tested this prediction using an in vitro dialysis system. Results showed that adding activated charcoal to the dialysate and increasing Qd similarly increased the clearances of protein-bound test solutes without significantly altering the clearance of urea. These results suggest that adding sorbents to the dialysate could increase the clearance of protein-bound solutes above the levels that are obtained with conventional hemodialysis.
Materials and Methods
In Vitro Dialysis System
Clearances of the protein-bound solutes indican, p-cresol sulfate (PCS), and p-cresol and of urea were measured during dialysis in vitro. Fluid that represented a patient’s plasma was placed in a continuously stirred 1.0-L reservoir and dialyzed using a Prisma system (Gambro, Lakewood, CO) set in the continuous venovenous hemodialysis mode. The reservoir and dialysate fluids were prepared to have identical electrolyte concentrations that approximated 140 mEq/L Na, 4.0 mEq/L K, 2.5 mEq/L Mg, 3.1 mEq/L Ca, and 5.0 mg/dl PO4. Phosphate was used instead of a bicarbonate as a buffer so that the pH of the fluids could be adjusted to approximately 7.4 without having to control the Pco2. In addition, BSA (CalBiochem 12659, La Jolla, CA) was added to the reservoir fluid at a concentration of 4.0 g/dl and indican, PCS, p-cresol, and urea were added to provide concentrations of approximately 1.6, 1.9, 0.9, and 160 mg/dl, respectively, at the beginning of each dialysis run. Batches of reservoir fluid that contained albumin but not test molecules were predialyzed to remove impurities. Preliminary studies showed that this increased the binding of indican and PCS from approximately 80% to approximately 90%, presumably reflecting the removal of substances in the commercial albumin preparation that competed for protein-binding sites. PCS, which is not available commercially, was synthesized as described previously (9). Dialysis experiments were performed with plasma flow (Qp) set at 40 ml/min using a Prisma M100 Set that includes a 0.9-m2 kidney composed of AN69 hollow fibers with wall thickness 50 μm. Separate dialysate experiments (n = 4 each) were performed with Qd set at 42 ml/min without sorbent and with Qd set at 42 ml/min with 3 g/L activated charcoal (Norit E Supra USP; Norit Americas, Marshall, TX) added to the dialysate. When charcoal was added, the dialysate reservoir was stirred continuously to keep the charcoal in suspension. Additional experiments (n = 4) were performed using a Fresenius H machine (Fresenius, Gurnee, IL) so that the effect of adding a sorbent could be compared with the effects of increasing Qd without any sorbent. In these experiments, the same reservoir and dialysate fluid compositions were used except that the concentration of Na was lowered to approximately 130 mEq/L. The Prisma M100 cartridge again was used and Qp was set at 40 ml/min, whereas Qd was set at 500 ml/min and the ultrafiltration rate was set at zero.
Dialysis experiments were carried out during 80 min with plasma recirculated from the 1-L reservoir, while dialysate flow was single pass. Qp rates were measured volumetrically at the beginning and end of each experiment, and Qd rates also were measured volumetrically in experiments that were performed with the Fresenius machine. Qd was calculated from the weight of waste fluid in experiments that were performed with the Prisma system except that in two experiments the measurement was omitted and the set flow rate of 41.7 ml/min was assumed to be correct. A reservoir fluid sample was collected at the beginning and end of each experiment for measurement of the unbound fraction of indican, PCS, and p-cresol. Additional reservoir fluid samples were collected at 0, 10, 20, 30, 40, 60, and 80 min for measurement of solute concentrations that were used in clearance calculations.
Chemical Assays and Clearance Calculations
Urea was measured using a commercial kit (1770-500; Thermo Electron Corp., Melbourne, Australia). Indican, PCS, and p-cresol were assayed by HPLC as described previously (9). Microcon YM-10 tubes (Millipore, Billerica, MA) were used to obtain ultrafiltrate from aliquots of the reservoir fluid samples that were obtained at 0 and 80 min in each experiment, and values for the free, unbound fraction f of indican, PCS, and p-cresol then were calculated as the solute concentrations in the ultrafiltrate divided by the total solute concentrations in the reservoir fluid. Clearance values for urea were calculated from the best fit slope of log values for concentration at 0, 10, 20, 30, and 40 min and from measured reservoir volume. Values for the clearances of indican, PCS, and p-cresol were calculated in the same way using concentration values at 0, 10, 20, 30, 40, 60, and 80 min. Values for R2 for the slope determinations averaged 0.999 ± 0.003, 0.999 ± 0.001, 0.999 ± 0.000, and 0.994 ± 0.006 for indican, PCS, p-cresol, and urea, respectively.
Values for KoA for the M100 cartridge were determined in separate experiments (n = 3) in which dialysis was performed without albumin and the reservoir volume was increased to 4.0 L. In these experiments, the Fresenius H machine was used with the ultrafiltration rate set at zero; Qp was set at 300 ml/min, and Qd was set at 500 ml/min. Clearances for indican, PCS, p-cresol, and urea were measured as the average of the arterial-venous extraction multiplied by Qp measured three times during 20 min in each experiment. KoA values for each solute then were obtained from the measured values for clearance, Qp, and Qd using the equation described by Michaels (13). Separate experiments were performed in triplicate to confirm that the E Supra carbon used could absorb the test solutes. Carbon at a concentration of 4 g/L was added directly to continuously stirred aliquots of an aqueous solution that contained indican, PCS, p-cresol, and urea in the concentrations present in the artificial plasma at the beginning of the dialysis experiments. After 10 min, samples were passed through a 0.2-μm filter, and free solute concentrations were measured in the filtrate. Results showed that the carbon absorbed 100 ± 0% of indican, 98 ± 1% of PCS, 100 ± 0% of p-cresol, but only 5 ± 1% of urea.
Mathematical Modeling of the Effect of Sorbent on Dialytic Clearance
Solute clearances were predicted using a previously described mathematical model that extends the standard treatment of Michaels (13) to include solutes that bind to proteins (8). The model incorporates the usual assumptions of perfect mixing (no boundary layers) in plasma and dialysate streams that flow countercurrent past a membrane whose solute permeability is unaffected by the fluid flows. For solutes that bind to proteins, the model incorporates the additional assumptions that only the free portion of the solute contributes to the concentration gradient’s driving diffusion across the membrane, and that binding can be described by an association constant KA such that
;)
where Cbound is the concentration of solute bound to albumin, Cfree is the free (unbound) solute concentration, and Calb is the total albumin concentration. The model further assumes that association-dissociation of solutes from binding proteins is rapid compared with the transit time of plasma through the dialyzer. The assumptions above can be expressed as a series of differential equations. For unbound solutes (f = 1.0), the solution to these equations is the classic expression of Michaels (13), which predicts clearance from values for Qp, Qd, and KoA. For bound solutes, the solution to the differential equations is provided by the expression
where Cl is the solute clearance and φ and θ are defined as
and
and where the free fraction of the solute f can be expressed as
where Cp is the total solute concentration (bound plus unbound). Equation (2) thus provides the clearance of a protein-bound solute as a function of Cp, Calb, KA, Qp, Qd, and KoA. This equation is applicable only when there is no ultrafiltration, so Calb and Qp are constant along the length of the dialyzer and when f also can be assumed to be constant along the length of the dialyzer (8). The authors have developed a more general treatment of the clearance of protein-bound solutes that incorporates additional differential equations to describe transport by both convection and diffusion (14). An analytic solution has not been obtained for this larger set of differential equations, but clearance values can be obtained with mathematical software that solves the equations by iteration. For our study, predicted clearance values were obtained using a recently described computer program that solves the equations using Excel (Dr. Addis Clearance Calculator, downloadable at http://www.stanford.edu/∼twmeyer/) (15). The maximal effect of adding sorbent to the dialysate was calculated on the basis of the assumption that a completely effective sorbent would lower the free solute concentration throughout the dialysate compartment to zero. For obtaining maximal clearance values with a sorbent, the published model therefore was modified so that the dialysate solute concentration was set at zero in each of the 1000 consecutive slices into which the computer program divides the length of the dialyzer to calculate solute transport. This modified version of the model, called Sorbent Dialysate Max Clearance Calculator, is available online at the same address.
Results
Results of in vitro dialysis experiments are summarized in Table 1. We first measured clearances with Qd approximately equal to Qp at the relatively low value of 42 ± 3 ml/min. Measured urea clearance with these flows was 34 ± 1 ml/min. To model urea clearance, we used the mean KoAurea value of 286 ± 38 ml/min that was obtained in separate experiments that were performed without albumin. Modeling yielded a predicted urea clearance of 40 ml/min, which was slightly greater than the observed value. Essentially, with KoAurea high relative to the Qp and Qd, the model predicts that urea concentration comes close to equilibrium across the dialyzer membrane near the plasma inlet/dialysate outlet. Nonuniform perfusion of the kidney at low flows presumably causes the observed clearance to fall slightly below the near-perfect urea exchange predicted by the model.
As expected, with a low Qd, the clearances of the protein-bound solutes indican, PCS, and p-cresol were much lower than the clearance of urea. Indican, 13 ± 4% of the total concentration of which was unbound, was cleared at the rate of only 5 ± 1 ml/min and PCS, 10 ± 2% of which was unbound, was cleared at the rate of only 4 ± 1 ml/min. P-cresol exhibited less extensive binding, with 35 ± 4% unbound, and was cleared at the higher rate of 14 ± 1 ml/min. In each case, modeling accurately predicted the effect of protein binding on the clearance, with the observed clearance rate being slightly less than the modeled clearance rate. The KoA values for the M90 kidney that was used in modeling were the mean values of 200 ± 5 ml/min for indican, 207 ± 5 ml/min for PCS, and 259 ± 3 ml/min for p-cresol as determined in separate experiments that were performed without albumin. As was the case with urea, the KoA values were large relative to the Qd. The model again predicts near equilibration of solute levels across the membrane but with the difference that the solute level in the dialysate approaches the unbound solute level in the plasma. The predicted clearance value for a protein-bound solute at low Qd thus approaches the Qd rate multiplied by the fraction of the solute that is unbound.
The addition of sorbent charcoal to the dialysate while Qd was maintained at 42 ± 2 ml/min had no effect on the clearance of urea. The lack of a sorbent effect on urea clearance was in accordance with the prediction of the model. It should be noted that because activated charcoal does not bind urea, addition of charcoal to the dialysate could not have increased the urea clearance with any dialysis prescription, but even a sorbent that did bind urea would be predicted to have little effect in this case. By lowering the free urea concentration in the dialysate, a sorbent that bound urea would increase the gradient for urea diffusion. With an adequately large membrane, this would tend to increase urea clearance but only up to the limiting value imposed by the Qp. In our experiment, the predicted urea clearance already was close to the Qp without any sorbent and therefore could not be increased significantly.
Although it did not alter the urea clearance, the addition of charcoal to the dialysate greatly increased the clearances of the protein-bound solutes indican, PCS, and p-cresol. For each solute, the clearance with charcoal added to the dialysate was more than double the clearance without charcoal. Of note, the clearance of each of the protein-bound solutes was increased well above the value that was obtained by multiplying the unbound solute fraction by the Qp rate, indicating that the solutes dissociated from albumin as plasma flowed through the kidney. Clearances for these solutes remained lower, however, than the values that were predicted on the basis of the assumption that the charcoal was fully effective in lowering the free concentrations of the bound solutes in the dialysate.
As expected, increasing the Qd to reduce solute concentrations in the dialysate compartment also increased solute clearances. The clearance of urea, which was unaffected by charcoal, was raised to a value that was indistinguishable from the theoretical maximum that was imposed by the Qp when Qd was increased to an average of 502 ± 4 ml/min. For indican, PCS, and p-cresol, the effect of increasing Qd was similar to the effect of adding charcoal to the dialysate. For each solute, the clearance at Qd approximately 500 ml/min was more than double the clearance with Qd approximately 42 ml/min. The clearances of indican and PCS, however, remained lower than the values that were predicted on the basis of the assumption that the increase in Qd was distributed uniformly throughout the dialysate compartment.
Discussion
Blood perfusion over sorbents was considered as an alternative to hemodialysis early during the development of renal replacement therapy. Yatzidis and colleagues (16,17) first showed that activated charcoal removed creatinine, urate, and phenols from the blood of uremic patients. Contact of blood with charcoal, however, caused destruction of platelets, complement activation, and sorbent embolization.
The adverse effects of direct blood–sorbent contact prompted the encapsulation of sorbents within various membrane materials. During the 1970s, numerous studies assessed solute removal from uremic patients by hemoperfusion over encapsulated sorbents (1,18). In general, the results were disappointing. Activated charcoal, the most commonly used sorbent, does not take up urea, water, or inorganic ions. Sorbent hemoperfusion therefore could serve only as an adjunct to hemodialysis. Moreover, the highest clearances for solutes such as creatinine and urate that were obtained with hemoperfusion were only slightly greater than those that were obtained with then-current hemodialysis methods, and solute clearances declined during treatment, presumably reflecting limited sorbent capacity. As hemodialysis evolved to provide higher clearances, consideration of sorbent treatment for uremia largely was abandoned.
The sorbent studies that were performed in the 1970s largely were restricted to empiric testing of various sorbent devices. Little attempt was made to model solute removal or predict clearance values. In particular, it seems generally not to have been appreciated that when sorbents are encapsulated, solutes must diffuse across the encapsulating membrane in response to a concentration gradient that is created when sorbent binding lowers the free solute concentrations in the sorbent compartment. The effect of the encapsulating membrane on the clearances that were obtained with sorbent systems therefore was not analyzed.
Our study attempted to characterize more accurately the extent to which solutes are cleared when sorbents are separated from the plasma by semipermeable membranes. We were particularly interested in whether sorbents could increase the clearance of protein-bound solutes above the low levels that are provided by conventional hemodialysis (5–9). We therefore modeled the effect of adding a sorbent to the dialysate during countercurrent dialysis. In this case, the sorbent is separated from the plasma by the dialyzer membrane and reduces the free concentration of bound solutes in the dialysate. As illustrated in Figure 1, addition of a sorbent has the same effect as increasing the Qd, and the maximum effect that is achievable with a sorbent is equal to that of an unlimited Qd. Provided that Qp does not limit solute supply and that KoA is sufficient, reducing free solute concentration in the dialysate either by adding a sorbent or by increasing Qd will increase clearance, as further illustrated in Figure 1.
The results of our in vitro dialysis experiment conformed to the predictions of the mathematical model. Experiments first were performed without sorbent and with KoA much greater than Qd. When dialysis was performed using these parameters, the urea clearance was close to the limiting value imposed by Qp, whereas albumin binding greatly restricted the clearances of indican, PCS, and p-cresol. Increasing Qd to exceed KoA only slightly increased urea clearance but greatly increased the clearance of the protein-bound solutes. As predicted, adding charcoal to the dialysate did not increase urea clearance but had the same effect on the clearances of the bound solutes as increasing the Qd.
The clearances of protein-bound solutes that we achieved with charcoal, although much higher than those that were observed without sorbent in the dialysate, were less than the maximum theoretically predicted values. Presumably, the amount of sorbent used was not sufficient to reduce the free solute concentrations near to zero throughout the dialysate compartment. It should be noted that this maximal effect will be approached only when the dialysate is stirred well enough to ensure that solute that diffuses through the membrane is exposed immediately to sorbent particles and when the sorbent particles are able to absorb the solute very rapidly at low concentration. We did not analyze the extent to which failure to reach maximum predicted clearance values in our study reflected inadequate mixing of the sorbent in the dialysate as opposed to inadequate capacity of the sorbent particles for rapid solute uptake. An additional potential contributor to the failure to achieve the predicted possible clearances was the use of low Qp and Qd. Clearances that are obtained during sustained low efficiency dialysis have been found to be lower than those that are calculated on the basis of dialyzer KoA values measured at higher flows (19). Presumably, this discrepancy reflects nonuniform perfusion of the blood and dialysate compartments when dialyzers are used at flow rates for which they were not designed (20). In our study, clearances of indican and PCS also were below the predicted values when the Qd was increased to a high level. This again could reflect nonuniform perfusion of the dialysate compartment, which would be expected to depress the clearances of protein-bound solutes more than unbound solutes.
It should be noted that a sorbent, in theory, can increase clearance rates not only for solutes that bind to plasma proteins but also for any solutes that bind to the sorbent. In addition, many solutes that do not bind to proteins, such as creatinine and urate, do bind to charcoal and other sorbents (1,21). In contemporary hemodialysis, however, KoA and Qd have been increased to the point that small solutes that do not bind to proteins are cleared at a large fraction of the limiting rate determined by the blood flow through the dialyzer. As was the case with urea in our experiments, adding a sorbent would not increase greatly the clearance of such solutes even if it bound them avidly. In contrast, current KoA and Qd values are far below the levels that are required to provide maximal clearances for protein-bound solutes (8). This being the case, either increasing Qd or adding a sorbent to the dialysate can increase solute clearances. In effect, both maneuvers can increase the clearance of protein-bound solutes relative to the clearance of unbound solutes such as creatinine and urea.
The extent to which the clearance of a solute can be increased by adding a sorbent to the dialysate or by increasing Qd depends on the relation of Qd to KoA. Our initial experiments were performed with Qd set at 42 ml/min, whereas KoA values for the protein-bound solutes ranged from 200 to 259 ml/min. Use of KoA greater than Qd allowed us to achieve a large increase in the clearances of the protein-bound solutes by adding charcoal to the dialysate. To obtain similar increases in the clearances of protein-bound solutes during clinical hemodialysis with Qd set at 600 to 800 ml/min would require kidneys with KoA values much larger than those in current use. It also would require the use of a large amount of sorbent. If the sorbent were charcoal, then the cost would not be great but might still exceed the cost of simply increasing the Qd.
By adding a sorbent to the dialysate in a single-pass system, we avoided solute accumulation on the sorbent over time. Sorbents also have been fixed to the dialysate side of hemodialysis membranes and suspended in recirculating dialysate as well as encapsulated in granular form and packed in hemoperfusion cartridges (1,4,13,18,22,23). The sorbents used have included albumin and various lipids and resins as well as activated charcoal (24–28). For any system in which a sorbent is separated from the plasma by a membrane, however, the limitations on clearance described in this study should apply. That is, the maximum clearance that is obtainable with the sorbent system can be determined by calculating the rate of solute diffusion across the membrane with the solute concentration in the sorbent compartment assumed to be equal to zero. It has been suggested that the permeability of dialysis membranes to solutes that bind to albumin may be increased when albumin is used as a sorbent (29), but we presume that exposure to powdered charcoal does not alter significantly membrane permeability. This being the case, the maximal effect of sorbent charcoal separated from plasma by membranes is equivalent to that of dialysis across the same membranes using an unlimited Qd. As an adjunct to conventional hemodialysis, addition of sorbents to the dialysate could increase the clearance of protein-bound solutes without greatly altering the clearance of unbound solutes. The model of serial diffusion and sorbent uptake also could be used to describe solute clearances that are obtained when membrane-enclosed sorbents are used to treat hepatic encephalopathy or poisoning.
Disclosures
None.
Figure 1: The predicted clearance of a protein-bound solute with conventional hemodialysis and the increase in this clearance that can be achieved either by raising dialysate flow (Q d) or by adding a sorbent to the dialysate. In each panel, dialyzer length is on the horizontal axis with the plasma inlet and dialysate outlet at the left (x = 0) and the plasma outlet and dialysate inlet at the right (x = 1). Solute concentration is on the vertical axis. The total plasma solute concentration (red line) is set equal to 1.0 at the plasma inlet, and the dialysate solute concentration (blue line) is set equal to 0.0 at the dialysate inlet. The broken red line represents the free concentration of the solute in the plasma, here set at 10% of the total. In each panel, the shaded region represents the magnitude of the gradient driving solute diffusion across the membrane. (A) Hemodialysis using conventional flow rates: blood flow (Q b) is 350 ml/min so that plasma flow (Q p) is 234 ml/min with a hematocrit of 33% and Q d is 700 ml/min. Even using a very large dialyzer with mass transfer area coefficient (K o A) at 2000 ml/min the modeled clearance of the protein-bound solute is only 67 ml/min, which is less than one third of the Q p. This is because the small concentration gradient that drives diffusion, as indicated by the shaded region, is dissipated when the solute concentration in the dialysate rises to approach the free solute concentration in the plasma. (B) Predicted effect of a large increase in Q d while other elements of the dialysis prescription are held constant. The solute concentration in the dialysate is reduced practically to zero, and the gradient that drives diffusion, although limited by the plasma free solute concentration, is maintained along the length of the dialyzer. The modeled clearance increases to more than half of the Q p and would rise higher if K o A were larger. (C) Effect of adding a sorbent to the dialysate while maintaining Q d at 700 ml/min with other elements of the dialysis prescription again held constant. If the sorbent is effective, then the free concentration of the solute in the dialysate, now represented by the dashed blue line, is maintained near zero. The gradient that drives diffusion and thus the clearance are the same as with the large increase in Q d illustrated in B, but solute is removed in less dialysate, with the total solute concentration in the dialysate, represented by the solid blue line, rising to exceed the plasma free solute concentration.
Table 1: Summary of in vivo dialysis experimentsa
This study was supported by the Research Service of the Veterans Administration and by the National Institutes of Health (R33 DK071251). Activated charcoal was a kind gift from NORIT Americas.
We thank John Moran and Satellite Healthcare for the loan of a dialysis machine, Gerry Kubovcik for help using the machine, and Ron VanGronigen for help with chemical assays.
Published online ahead of print. Publication date available at www.jasn.org.
See the related editorial, “Uremic Retention Solutes: The Free and the Bound,” on pages 675–676.
References
1. Chang TM: Assessments of clinical trials of charcoal hemoperfusion in uremic patients. Clin Nephrol 11: 111–119, 1979
2. Gurland HJ: Hemoperfusion in the treatment of uremia. Kidney Int Suppl 17: S47–S49, 1985
3. Winchester JF: Sorbent hemoperfusion in end-stage renal disease: An in-depth review. Adv Ren Replace Ther 9: 19–25, 2002
4. Mikhalovsky SV: Emerging technologies in extracorporeal treatment: Focus on adsorption. Perfusion 18[Suppl 1]: 47–54, 2003
5. Vanholder R, De Smet R, Lameire N: Protein-bound uremic solutes: The forgotten toxins. Kidney Int Suppl 78: S266–S270, 2001
6. Brunet P, Dou L, Cerini C, Berland Y: Protein-bound uremic retention solutes. Adv Ren Replace Ther 10: 310–320, 2003
7. Bammens B, Evenepoel P, Verbeke K, Vanrenterghem Y: Removal of the protein-bound solute p-cresol by convective transport: A randomized crossover study. Am J Kidney Dis 44: 278–285, 2004
8. Meyer TW, Leeper EC, Bartlett DW, Depner TA, Lit YZ, Robertson CR, Hostetter TH: Increasing dialysate flow and dialyzer mass transfer area coefficient to increase the clearance of protein-bound solutes. J Am Soc Nephrol 15: 1927–1935, 2004
9. Martinez AW, Recht NS, Hostetter TH, Meyer TW: Removal of P-cresol sulfate by hemodialysis. J Am Soc Nephrol 16: 3430–3436, 2005
10. De Smet R, Van Kaer J, Van Vlem B, De Cubber A, Brunet P, Lameire N, Vanholder R: Toxicity of free p-cresol: A prospective and cross-sectional analysis. Clin Chem 49: 470–478, 2003
11. Dou L, Bertrand E, Cerini C, Faure V, Sampol J, Vanholder R, Berland Y, Brunet P: The uremic solutes p-cresol and indoxyl sulfate inhibit endothelial proliferation and wound repair. Kidney Int 65: 442–451, 2004
12. Bammens B, Evenepoel P, Keuleers H, Verbeke K, Vanrenterghem Y: Free serum concentrations of the protein-bound retention solute p-cresol predict mortality in hemodialysis patients. Kidney Int 69: 1081–1087, 2006
13. Michaels AS: Operating parameters and performance criteria for hemodialyzers and other membrane-separation devices. Trans Am Soc Artif Intern Organs 12: 387–392, 1966
14. Meyer TW, Walther JL, Pagtalunan ME, Martinez AW, Torkamani A, Fong PD, Recht NS, Robertson CR, Hostetter TH: The clearance of protein-bound solutes by hemofiltration and hemodiafiltration. Kidney Int 68: 867–877, 2005
15. Walther JL, Bartlett DW, Chew W, Robertson CR, Hostetter TH, Meyer TW: Downloadable computer models for renal replacement therapy. Kidney Int 69: 1056–1063, 2006
16. Yatzidis H: A convenient haemoperfusion micro-apparatus over charcoal for the treatment of endogenous and exogenous intoxication. Its use as an effective artificial kidney. Proc Eur Dial Transpl Assoc 1: 83–87, 1964
17. Yatzidis H, Yulis G, Digenis P: Hemocarboperfusion-hemodialysis treatment in terminal renal failure. Kidney Int Suppl S312–S314, 1976
18. Gelfand MC, Winchester JF: Hemoperfusion results in uremia. Clin Nephrol 11: 107–110, 1979
19. Marshall MR, Golper TA, Shaver MJ, Alam MG, Chatoth DK: Urea kinetics during sustained low-efficiency dialysis in critically ill patients requiring renal replacement therapy. Am J Kidney Dis 39: 556–570, 2002
20. Leypoldt JK, Cheung AK, Agodoa LY, Daugirdas JT, Greene T, Keshaviah PR: Hemodialyzer mass transfer-area coefficients for urea increase at high dialysate flow rates. The Hemodialysis (HEMO) Study. Kidney Int 51: 2013–2017, 1997
21. Stefoni S, Coli L, Feliciangeli G, Baldrati L, Bonomini V: Regular hemoperfusion in regular dialysis treatment. A long-term study. Int J Artif Organs 3: 348–353, 1980
22. Randerson DH, Gurland HJ, Schmidt B, Farrell PC, Hone PW, Stokoe C, Zuber A, Blogg A, Fateh-Moghadam A, Marschner I, Kopcke W: Sorbent membrane dialysis in uremia. Contrib Nephrol 29: 53–64, 1982
23. Ash SR, Carr DJ, Sullivan TA: Sorbent suspension reactor for extracorporeal detoxification in hepatic failure or drug overdose. ASAIO J 50: lviii–lxv, 2004
24. Patzer JF 2nd, Bane SE: Bound solute dialysis. ASAIO J 49: 271–281, 2003
25. Evenepoel P, Maes B, Wilmer A, Nevens F, Fevery J, Kuypers D, Bammens B, Vanrenterghem Y: Detoxifying capacity and kinetics of the molecular adsorbent recycling system. Contribution of the different inbuilt filters. Blood Purif 21: 244–252, 2003
26. Whang R, Orndorff M, Papper S: Lipid-electrolyte (“lipo-lyte”) dialysis: Experimental and preliminary clinical observations. Clin Toxicol 3: 551–560, 1970
27. Asbach HW, Holz F, Mohring K, Schuler HW: Lipid hemodialysis versus charcoal hemoperfusion in imipramine poisoning. Clin Toxicol 11: 211–219, 1977
28. Winchester JF: Dialysis and hemoperfusion in poisoning. Adv Ren Replace Ther 9: 26–30, 2002
29. Stange J, Mitzner S: A carrier-mediated transport of toxins in a hybrid membrane. Safety barrier between a patients blood and a bioartificial liver. Int J Artif Organs 19: 677–691, 1996
