Operational characteristics of continuous renal replacement modalities used for critically ill patients with acute kidney injury

Transcription

1 The International Journal of Artificial Organs / Vol. 31 / no. 6, 2008 / pp Review Operational characteristics of continuous renal replacement modalities used for critically ill patients with acute kidney injury Z. HUANG 1, J.J. LETTERI 2, W.R. CLARK 2,3, C. RONCO 4, D. GAO 5 1 School of Mechanical Engineering, Widener University, Philadelphia, Pennsylvania - USA 2 Gambro, Lakewood, Colorado - USA 3 Nephrology Division, Indiana University School of Medicine, Indianapolis, Indiana - USA 4 Department of Nephrology, Dialysis and Transplantation, St. Bortolo Hospital, International Renal Research Institute Vicenza (IRRIV), Vicenza - Italy 5 School of Biomedical Engineering, University of Washington, Seattle, Washington - USA ABSTRACT: Renal replacement therapy (RRT) is required in a significant percentage of patients developing acute kidney injury (AKI) in an intensive care unit (ICU) setting. One of the foremost objectives of continuous renal replacement therapy (CRRT) is the removal of excess fluid and blood solutes that are retained as a consequence of decreased or absent glomerular filtration. Because prescription of CRRT requires goals to be set with regard to the rate and extent of both solute and fluid removal, a thorough understanding of the mechanisms by which solute and fluid removal occurs during CRRT is necessary. The following provides an overview of solute and water transfer during CRRT and this information is placed in the appropriate clinical context with a discussion of recent clinical trials assessing the relationship between CRRT dose and patient survival. Moreover, the differences between solute removal in CRRT and other dialysis modalities, especially sustained lowefficiency dialysis (SLED) and extended daily dialysis (EDD), along with the potential clinical implications are discussed. (Int J Artif Organs 2008; 31: ) KEY WORDS: Clearance, Convection, Diffusion, Hemodialysis, Hemofiltration, Sieving coefficient INTRODUCTION Renal replacement therapy (RRT) is required in a significant percentage of patients developing acute kidney injury (AKI) in an intensive care unit (ICU) setting. One of the foremost objectives of continuous renal replacement therapy (CRRT) is the removal of excess fluid and blood solutes that are retained as a consequence of decreased or absent glomerular filtration. Because prescription of CRRT requires goals to be set with regard to the rate and extent of both solute and fluid removal, a thorough understanding of the mechanisms by which solute and fluid removal occurs during CRRT is necessary. The following provides an overview of solute and water transfer during CRRT, and this information is placed in the appropriate clinical context with a discussion of recent clinical trials assessing the relationship between CRRT dose and patient survival. Moreover, the differences between solute removal in CRRT and other dialysis modalities, especially sustained low-efficiency dialysis (SLED) and extended daily dialysis (EDD), along with the potential clinical implications are discussed. Characterization of filter performance in CRRT Clearance Quantification of dialytic solute removal is complicated by the confusion about the relationship between clearance and mass removal for different therapies. By definition (1), solute clearance (K) is the ratio of mass removal rate (N) to blood solute concentration (C B ): K = N / C B [1] From a kinetic perspective, Figure 1 depicts the relevant flows for determining CRRT clearances. In CRRT, Wichtig Editore, / $25.00/0

2 Renal replacement for acute kidney injury Fig. 1 - Relevant flow considerations for the determination of solute clearance in continuous renal replacement therapy. The modality represented is continuous venovenous hemodiafiltration (CVVHDF). C A, C V, C E concentration (arterial, venous, and effluent respectively). Q A : arterial blood flow rate (ml/min); Q V : venous blood flow rate (ml/min); Q D : dialysate flow rate (ml/hr); Q E : effluent flow rate (ml/hr); Q R : replacement fluid rate (ml/hr). the mass removal rate is estimated by measuring the actual amount of solute appearing in the effluent. The mass removal rate is the product of the effluent flow rate (Q E ) and the effluent concentration of the solute (C E ). In continuous venovenous hemodialysis (CVVHD) and continuous venovenous hemodiafiltration (CVVHDF), the effluent is dialysate and diafiltrate, respectively. For these therapies, the extent of solute extraction from the blood is estimated by the equilibration ratio (E), also known as the degree of effluent saturation. The benchmark for efficiency in these therapies is the volume of fluid (dialysate and/or replacement fluid) required to achieve a certain solute clearance target (see below). Clearance in postdilution continuous venovenous hemofiltration (CVVH) is the product of the sieving coefficient (see below) and the ultrafiltration rate (Q UF ) (2). For small solutes such as urea and creatinine, the sieving coefficient is essentially 1 (under normal filter operation). Therefore, small solute clearance in postdilution CVVH essentially is equal to the Q UF. On the other hand, estimation of clearance in predilution CVVH has to account for the fact that the blood solute concentrations are reduced by dilution of the blood before it enters the filter (3). Thus, the clearance has a dilution factor (see below). In essence, the dilution factor can be viewed as a measure of the extent to which predilution differs from postdilution for a specific combination of blood flow rate (Q B ) and Q UF. It should be noted here these clearance equations are ultrafiltrate-side measurements. Thus, they are quite satisfactory for small solutes but are not always appropriate for larger molecules, which may have an adsorptive clearance component. As an example, for the AN69 membrane, sieving coefficient interpretations for larger molecules may be difficult due to its adsorptive nature (4). Specifically, during the first 2 hours of use of an AN69 filter, adsorptive removal of large compounds, such as β2-microglobulin (B2M) and inflammatory mediators, may predominate over transmembrane removal. If a B2M sieving coefficient determination is made during this early period, the value may be very low (or zero), and the untrained observer may conclude that the membrane is not clearing any B2M. However, significant clearance is occurring by adsorption, but to appreciate this, a total clearance (i.e., blood-side) measurement needs to be made. After the first 2 hours of treatment, the membrane may reach adsorptive saturation when B2M breakthrough into the filtrate occurs with a measurable sieving coefficient then obtained. Thus, the timing of AN69 clearance determinations for larger molecules is important for correct interpretation. Sieving coefficient When a dialyzer is operated as a hemofilter (i.e., ultrafiltration with no dialysate flow), solute mass transfer occurs almost exclusively by convection. Convective solute removal is primarily determined by membrane pore size and treatment ultrafiltration rate (5). Mean pore size is the major determinant of a filter s ability to prevent or allow the transport of a specific solute. The sieving coefficient (S) represents the degree to which a particular membrane permits the passage of a specific solute: S = C UF / C P [2] In this equation, C UF and C P are the solute concentrations in the ultrafiltrate and the plasma (water), respectively. (As discussed below, a sieving coefficient measurement is influenced by the flow operating conditions under which the determination is made.) Irrespective of membrane type, all filters in the virgin state have small solute sieving coefficient values of one, and these values are typically not reported by filter manufacturers. Sieving coefficient values for solutes of larger molecular weight are more applicable, and manufacturers frequently provide data for one or more middle-mol- 526

3 Huang et al ecule surrogates, such as vitamin B 12, inulin, cytochrome C, and myoglobin. As is the case for solute clearance, the relationship between S and solute molecular weight is highly dependent on membrane mean pore size (6). Sieving coefficient data provided by manufacturers are usually derived from in vitro experimental systems in which (non-protein containing) aqueous solutions are used as the blood compartment fluid. In actual clinical practice, nonspecific adsorption of plasma proteins to a filter membrane effectively reduces the permeability of the membrane. Consequently, ex vivo sieving coefficient values are typically less than those derived from aqueous experiments, sometimes by a considerable amount. Fig. 2 - Transport mechanisms of diffusion and convection. Mechanisms of solute removal Diffusion Diffusion is the process of transport in which molecules that are present in a solvent and can freely cross across a semipermeable membrane tend to move from the region of higher concentration into the region of lower concentration (Fig. 2). In addition to the concentration gradient (dc), the diffusive flux (J X ) is influenced by membrane characteristics, namely surface area (A) and thickness (dx), solution temperature (T), and the diffusion coefficient of the solute (D). Fick s law of diffusion then gives the diffusive flux, defined as the solute mass removal rate due to diffusion normalized to membrane surface area (7): J X = DTA (dc/dx) [3] Diffusion is an efficient transport mechanism for the removal of relatively small solutes, but as solute molecular weight increases, diffusion becomes limited and the relative importance of convection increases. Convection Convection is the mass transfer mechanism associated with ultrafiltration of plasma water (8). If a solute is small enough to pass through the pore structure of the membrane, it is driven ( dragged ) across the membrane in association with the ultrafiltrated plasma water (Fig. 2). This movement of plasma water is a consequence of a transmembrane pressure (TMP) gradient. Quantitatively, the ultrafiltration flux (J F ), defined as the ultrafiltration rate normalized to membrane surface area, can be described by: J F = K F TMP [4] In this equation, K F is the membrane-specific hydraulic permeability (units: ml/hour/mmhg/m 2 ), and TMP is a function of both the hydrostatic and oncotic pressure gradients. (Note: K F differs from the ultrafiltration coefficient [K UF ] described. Adsorption For certain membranes, adsorption (binding) may be the dominant or sole mechanism by which some hydrophobic compounds (e.g., peptides and proteins) are removed (9). The adsorptive surface area of a membrane resides primarily in the pore structure rather than in the nominal surface area. As such, the adsorption of a lowmolecular-weight protein is highly dependent on access of the protein to a membrane s internal pore structure. Consequently, adsorption of peptides and low-molecularweight proteins, such as B2M and some inflammatory mediators, to low-flux membranes is not expected to be clinically significant, at least in comparison with that which occurs to high-flux membranes. As previously noted, the adsorption affinity of certain high-flux synthetic membranes for proteins and peptides is particularly high, attributable to the relative hydrophobicity of these membranes. 527

4 Renal replacement for acute kidney injury Fig. 3 - Relationship between ultrafiltration rate (Q UF ) and transmembrane pressure (TMP) during ultrafiltration. Q B : blood flow rate. Reprinted with permission, from (11). Water permeability (flux) of CRRT filters In clinical practice, membranes are incorporated into specific devices designed to optimize the performance of the membrane itself. These devices may either be designed as dialyzers, working prevalently in diffusion with a countercurrent flux of blood and dialysate, or designed as hemofilters, working prevalently in convection. Improvements in membrane design have allowed diffusive and convective mass transport to be combined, leading to therapies (high-flux dialysis and hemodiafiltration) in which the advantages of both mechanisms are significantly enhanced. Considerable confusion regarding the exact meaning of flux currently exists. As discussed above, the hydraulic flux of a membrane is the volumetric rate (normalized to surface area) at which ultrafiltration of water occurs. The clinical parameter used to characterize the water permeability of a specific dialyzer is the ultrafiltration coefficient (K UF : ml/hour/mm Hg) (10). The K UF of a filter is usually derived from in vitro experiments in which bovine blood is ultrafiltered at varying TMPs. The membrane characteristic having the largest impact on water permeability is pore size, such that ultrafiltrate flux is roughly proportional to the fourth power of the mean membrane pore radius (10). As such, small changes in pore size have a very large effect on water permeability. Extracorporeal membranes used for dialysis are classified according to their ultrafiltration coefficient as high flux or low flux. As shown in Figure 3, a defined relationship exists between ultrafiltration rate (Q UF ; y-axis) and TMP (x-axis). For each curve, at relatively low TMP values, there is a linear region the slope of the line in this region is essentially the K UF of the filter. As TMP increases, each curve eventually plateaus at a certain maximum Q UF (11). As mentioned previously, filter K UF is a value that is specific to a certain set of flow operating conditions, including Q B. In terms of clinical operation of a filter, the plateau portion of the curve is to be avoided because, in this region, an increase in TMP yields no additional increase in Q UF. Blood flow rate influences the nature of these curves in 2 ways. First, as Q B increases, the slope of the curve in the linear (low-tmp) region increases. Effectively, this means that to achieve a certain Q UF, a lower TMP is required. As an example, to achieve a Q UF of 60 ml/min, a TMP of only about 10 mm Hg is required at a Q B of 400 ml/min, while TMPs of about 50 and 100 mm Hg are required at Q B values of 300 and 200 ml/min, respectively. The second way in which Q B influences the nature of these curves is its effect on the maximum achievable (plateau) Q UF. This value is approximately 120 ml/min for a Q B of 400 ml/min, while it is only 100 ml/min for a Q B of 300 ml/min and only 65 ml/min for a Q B of 200 ml/min.the explanation for the behavior of these curves is related to the effect of higher Q B in preserving filter membrane function. Specifically, as Q B increases, a greater shear force is applied to the proteins comprising the secondary membrane. In this way, the secondary membrane is disrupted, and its negative impact on membrane function is blunted. Kinetic considerations for different CRRT techniques Several techniques are today available in the spectrum of CRRT. Techniques may differ in terms of vascular access and extracorporeal circuit design, frequency and intensity of treatment, predominant mechanism of transport utilized, and type of membrane. The following description is based primarily on the operational parameters normally employed and the target efficiency with respect to solute and fluid control. Continuous venovenous hemofiltration This technique utilizes high-flux membranes, and the prevalent mechanism of solute transport is convection. 528

5 Huang et al Ultrafiltration rates in excess of the amount required for volume control are prescribed, requiring partial or total replacement of ultrafiltrate losses with reinfusion (replacement) fluid. Replacement fluid can either be infused before the filter (predilution), after the filter (postdilution), or simultaneously at both locations (pre/post-dilution). Postdilution hemofiltration The location of reinfusion fluid delivery in the extracorporeal circuit during CVVH has a significant impact on solute removal and therapy requirements. For a given volume of replacement fluid over a wide molecular weight (MW) spectrum of uremic toxins, postdilution CVVH provides higher solute clearance than does predilution hemofiltration. As discussed below, the relative inefficiency of the latter mode is related to the dilutionrelated reduction in solute concentrations, which decreases the driving force for convective mass transfer. Despite its superior efficiency with respect to replacement fluid utilization, postdilution hemofiltration is limited inherently by the attainable Q B. More specifically, the ratio of the ultrafiltration rate to the plasma flow rate delivered to the filter, termed the filtration fraction, is the limiting factor. In general, a maximal filtration fraction of approximately 25% usually guides prescription in postdilution CVVH. At filtration fractions beyond these values, excessive hemoconcentration and secondary membrane effects become prominent and may impair hemofilter performance. The Q B limitations imposed by the use of temporary catheters accentuates the filtration fraction related constraints on maximally attainable ultrafiltration rate in postdilution CVVH. Therefore, the ultrafiltrate volumes shown by Ronco and colleagues to improve survival (12) can frequently be achieved only in the predilution mode. As discussed below, efficient utilization of replacement fluid in acute predilution CVVH is an important consideration. Predilution hemofiltration The use of predilution reinfusion has several advantages over postdilution with respect to solute removal (13), including the potential to increase filter patency. However, these mass transfer benefits must be weighed against the predictable dilution-induced reduction in plasma solute concentrations, one of the driving forces for convective solute removal. As suggested previously, the extent to which this reduction occurs is determined mainly by the ratio of the replacement fluid rate to the Q B. In fact, a frequently overlooked consideration is the important influence of Q B on solute clearance, particularly in CVVH. For small solutes, which are distributed in the blood water (BW) component within the blood passing through the hemofilter, the operative clearance equation in predilution CVVH is (13): K = Q UF S [Q BW / (Q BW + Q R )] [5] where Q BW is the blood water flow rate and Q R is the reinfusion (replacement) fluid rate. At a given Q UF value, predilution CVVH is always less efficient than postdilution CVVH with respect to fluid utilization, as discussed above. As Equation 5 indicates, the larger Q R is, relative to Q BW, the smaller is the entire fraction represented by the third term on the right-hand side. In turn, the smaller this term is, the greater is the loss of efficiency (relative to postdilution) due to dilution. Since employing a relatively low Q R is not an option in high-dose CVVH, due to the direct relationship that exists between Q UF and Q R, attention needs to be focused on achieving blood flow rates that are significantly higher than what have been used traditionally in CRRT (i.e., higher than 150 ml/min or less). In fact, widespread attainment of doses consistent with the intermediate- and high-dose arms in the study performed by Ronco and colleagues (35-45 ml/hour/kg) cannot occur unless blood flow rates of approximately 250 ml/min or more become routine in predilution CVVH. Evidence supporting the critical importance of Q B in predilution CVVH has been provided by Clark and colleagues (14). For patients of varying body weight, these investigators estimated the daily replacement fluid volumes required to provide a dose equivalent to 35 ml/hour/kg in postdilution. (A filter operation of 16 hours per day was assumed to account for differences in prescribed vs. delivered therapy time (15)). The replacement fluid requirements for attaining the above dose were determined as a function of Q B. For low blood flow rates ( 150 ml/min), the data suggested reinfusion fluid rates required to achieve this dose are impractically high (>100 l/day) in the majority of patients (>70 kg) due to a chasing the tail phenomenon described in the following way: To achieve the dose target, a high ultrafiltration rate is required. However, the concomitant requirement of a similarly high replacement fluid rate has a relatively substantial dilutive effect on solute concentrations at low Q B. On the other hand, for Q B values greater than 250 ml/min, the dilutive effect of the reinfu- 529

6 Renal replacement for acute kidney injury Clearance (ml/min) Ultrafiltration Rate (ml/min) Fig. 4 - Solute clearance as a function of ultrafiltration rate in predilution continuous venovenous hemofiltration. Reprinted with permission, from (3). sion fluid is attenuated significantly, and with the resultant improvement in fluid efficiency, the target dose can be delivered practically to a broad range of patients. Although the relationship between ultrafiltration rate and clearance is relatively clear in postdilution (see above), the same relationship may not be as predictable in predilution. Consequently, the claim that Q UF is a dose surrogate in predilution hemofiltration needs to be demonstrated. To this end, Huang and colleagues (3) have investigated the effect of Q UF on solute removal parameters in predilution CVVH. For a Q B of 200 ml/min, removal parameters at Q UF values of 20, 40, and 60 ml/min, corresponding to 17, 34 and 51 ml/hour/kg for a 70-kg patient, were measured for solutes of varying MW. The relationship between solute clearance and Q UF for urea, creatinine, vancomycin, and inulin is shown in Figure 4. Overall, these data are consistent with a convective therapy for two reasons. First, for each solute, the Q UF -to-clearance relationship is linear, confirming a direct relationship between these two parameters. Second, for a given Q UF over the solute MW range investigated, clearance is not strongly dependent on MW, at least in comparison with hemodialysis. Specifically, very little difference in clearance is observed between the two small solutes and between the two middle-mw surrogates as a function of Q UF. On the other hand, reflecting its diffusive basis, hemodialysis is associated with much larger differences in clearance over the same MW range. The authors concluded that, because an orderly relationship exists between Q UF and solute clearance, effluent rate is a reasonable dose surrogate in predilution CVVH, as has been suggested for postdilution CVVH (12) and for CVVHDF (16). Overall, these data seem to validate the use of effluent-based dosing, which has become the standard approach used for CRRT prescription in clinical practice. As noted above, however, the major drawback of predilution CVVH is its relatively low efficiency, resulting in relatively high replacement fluid requirements to achieve a given solute clearance. In a group of patients treated with a traditional blood flow rate for CRRT, Troyanov and colleagues have recently quantified the efficiency loss associated with predilution in the clinical setting (17). This study demonstrated the significant negative effect on efficiency when a relatively low Q B (less than 150 ml/min) is used with a relatively high Q UF and Q R in predilution CVVH. This specific combination of Q B = ml/min and Q UF = 4.5 l/hour (75 ml/min) is associated with a loss of efficiency of 30%-40% relative to postdilution for several different solutes. In other words, to achieve the same solute clearance, 30%-40% more replacement fluid is required in predilution under these conditions, relative to postdilution under the same conditions. However, it should be noted the likelihood of achieving such an ultrafiltration rate in postdilution is very remote at such a low Q B, as this would require a filtration fraction in excess of 50%. This condition is likely to lead to very short-term filter patency. Continuous venovenous hemodialysis Because of the nature of the membrane and the gradient provided by the dialysate, the prevalent mechanism of solute transport in this technique is diffusion. When CVVHD is performed with a relatively small surface area filter (<0.5 m 2 ), saturation of the dialysate is achieved at relatively low dialysate flow rates. For a 0.4-m 2 filter, Bonnardeaux et al (18) have shown that saturation of the effluent dialysate for urea and creatinine is preserved only up to a dialysate flow rate (Q D ) of approximately 16.7 ml/min (1 l/hour). For Q D values in the 2-3 l/hour range ( ml/min), an increase in Q D results in an increase in clearance. However, a divergence develops between the urea and creatinine clearance curves and the effluent 530

7 Huang et al dialysate curve, indicating a certain degree of nonsaturation of dialysate. Of course, the greater the degree of nonsaturation, the more inefficient is the procedure. For this relatively small filter, beyond a Q D value of approximately 3 l/hour, the urea and creatinine clearance curves were found to plateau. In this Q D range, further increases in Q D no longer resulted in an increase in clearance. A more contemporary study involving a larger surface area filter (0.9 m 2 ) demonstrated clearly the important effect of surface area on preserving dialysate saturation (19). For this larger filter, preservation of effluent dialysate saturation was achieved essentially over the entire Q D range for several relatively small molecules, namely urea, creatinine, phosphate, and uric acid. The molecule for which saturation was not achieved was B2M. The high MW of this compound (approximately 200 times that of urea) severely limits its diffusive capabilities and, therefore, its ability to saturate the dialysate. Continuous venovenous hemodiafiltration Continuous hemodiafiltration requires a high-flux hemodiafilter and operates combining the principles of hemodialysis and hemofiltration. As such, this therapy may allow for an optimal combination of diffusion and convection to provide clearances over a very broad range of solutes. Later generation CRRT machines allow a combination of predilution and postdilution with the aim of combining the advantages of both reinfusion techniques, as information from the chronic hemodiafiltration literature suggests a combination of predilution and postdilution may be optimal in terms of clearance and operational parameters (20). This may also be the case for CVVHDF in AKI, although this possibility has not been assessed carefully. The optimal balance is most likely dictated by the specific set of CVVHDF operating conditions, namely Q B, Q D, Q UF, and filter type. Troyanov et al (17) have performed a direct clinical comparison of CVVHDF and predilution CVVH with respect to urea and B2M clearance at a traditional blood flow rate of 125 ml/min. The study compared clearances at the same effluent rate over an effluent range of up to 4.5 l/hour. Urea clearance was significantly higher in CVVHDF than in predilution CVVH, and, in fact, the difference between the two therapies increased as effluent rate increased. These results are consistent with the penalizing effect of predilution, which is especially pronounced at low Q B. For B2M, the results were contrary to the conventional wisdom, which would suggest that a purely convective therapy such as CVVH should be inherently superior to a partly convective therapy such as CVVHDF for clearance of a molecule this size. However, once again, the penalty of predilution in CVVH was apparent, as the B2M clearances for the two modalities were equivalent except at very high effluent rates (greater than 3.5 l/hour). Comparison of diffusive and convective solute removal in different acute modalities It is important to emphasize that the relationship between solute clearance and flow rate in diffusive therapies differs significantly from that in convective therapies. Based on mass transfer considerations, the expected clearance of small solutes during CVVHD and postdilution CVVH is the same. However, as solute MW increases, the relevance of diffusion diminishes, and the benefits of convection become increasingly apparent. For acute dialysis modalities, these principles have been substantiated in both modeling and clinical studies. Liao and colleagues (21) performed a kinetic comparison of conventional intermittent hemodialysis (daily 4-hour treatments) (22), SLED (daily 12-hour treatments) (23, 24), and predilution CVVH (ultrafiltration rate, 35 ml/kg/hour) (12). Both SLED and EDD, which are essentially different acronyms for the same therapy, use equipment that is primarily designed for conventional hemodialysis. As such, the primary solute removal mechanism for these modalities is diffusion. The major differences between conventional hemodialysis and SLED/EDD are treatment duration (longer for SLED/EDD) and Q B and Q D (lower for SLED/EDD). Liao et al (21) employed the equivalent renal clearance concept (25) to compare effective solute removal for these modalities. Their analyses indicated similar effective urea clearances for CVVH and SLED of 33 and 31 ml/min, respectively, both of which were substantially higher than that delivered by daily intermittent hemodialysis (21 ml/min). On the other hand, the estimated B2M clearances for CVVH and SLED were 18 and 4 ml/min, respectively. Daily intermittent hemodialysis with a high-flux dialyzer was estimated to provide an intermediate B2M clearance of 7 ml/min. The predicted B2M concentration profiles appear in Figure 5. The profile specifically predicted for SLED is the result of ongoing solute generation 531

8 Renal replacement for acute kidney injury B2M (mg/dl) Fig. 5 - Predicted β2-microglobulin (B2M) concentration profiles for daily intermittent hemodialysis (IHD), sustained low-efficiency dialysis (SLED), and continuous venovenous hemofiltration (CVVH). Reprinted with permission, from (21). in the face of no removal by a low-flux filter. On the other hand, the combination of continuous operation and convection permit CVVH to achieve a significant reduction in B2M concentration over time. These modeled data are extended by a recent clinical study in which effluent collections were used to quantify solute removal in predilution CVVH (2.5 l/hour) and an EDD system utilizing a high-flux filter (26). (Consistent with the original description of the use of SLED in the United States, the above modeling study employed a low-flux dialyzer.) Indeed, Kielstein and colleagues corroborated Liao s findings by demonstrating that urea and creatinine removal during CVVH and SLED are similar. However, B2M removal was 2-fold greater in CVVH vs. SLED, even with the use of a highflux filter in the latter therapy. The kinetic capabilities of the different modalities used for AKI may have an important effect on patient outcome. As noted previously, randomized controlled trials initially established a direct relationship between the dose of convection-based CRRT and AKI patient survival. These studies demonstrated that for both CVVH (12) and CVVHDF (16), an effluent dose of at least ml/kg/hour (normalized to body weight) was associated with higher survival, than were lower therapy doses. More recent CVVHDF studies, employing lower effective CRRT doses and including the landmark ATN Trial (27, 28), have not corroborated this direct relationship between dose and survival. It should be emphasized that the use of both predilution and the incorporation of a significant component of diffusion in these two latter studies rendered their effective solute removal profiles quite different in comparison with the Ronco trial. Nevertheless, it is expected that effluentbased dosing will continue to be used in clinical practice and additional studies will refine our understanding of the relationship between CRRT dose and patient outcome. The relationship between treatment dose in SLED and patient outcome has also been reported recently on a preliminary basis. In a randomized controlled trial comparing relatively low-dose ( standard ) EDD and relatively high-dose ( intensified ) EDD (29), urea concentration was used to estimate the intensity of therapy, with a low target urea concentration associated with intensive therapy and a higher urea concentration associated with less intensive therapy. Therefore, what the investigators termed therapy intensity was really small-solute clearance, which in EDD (even with a high-flux dialyzer) is achieved almost entirely by diffusion. Since higher intensity (i.e., higher small-solute clearance) did not result in better patient survival, one plausible conclusion is that diffusionconstrained therapies have a limited impact on actual patient outcome due to the relatively narrow solute removal spectrum achieved. Although the hope was that the ATN Trial would provide additional dose/outcome data for SLED, the very low rate of SLED utilization in the study (less than 5% of hemodynamically unstable patients) precludes this possibility. A more recent study performed by Ricci and colleagues (30) reinforces the importance of convection in achieving solute clearance over a broad MW spectrum in CRRT. Based on a common prescription of 35 ml/kg/hour effluent flow rate, these investigators measured clearance of urea, creatinine, and B2M during CVVH and CVVHD in a cross-over study. The median urea (31.6 vs ml/min) and creatinine (38.1 vs ml/min) clearances in CVVH and CVVHD, respectively, were similar. However, median B2M clearance in CVVH was higher than that in CVVHD (16.3 vs. 6.3 ml/min, respectively; p=0.055). It must be emphasized that this borderline statistically significant difference was observed despite the fact that this trial was markedly underpowered. 532

9 Huang et al CONCLUSIONS Rational prescription of CRRT to critically ill patients with AKI is predicated upon an understanding of the basic principles of solute and water removal. In the preceding, the major ways in which filter function is characterized clinically have been reviewed. In addition, the fundamental mechanisms for solute and fluid transport have been discussed. Finally, these principles have been applied in a therapeutic context to the various CRRT modalities used by clinicians managing AKI patients and comparisons made to other acute dialysis modalities. Address for correspondence: William R. Clark, MD Gambro, Inc 4322 Wythe Lane Indianapolis, IN 46250, USA william.clark@us.gambro.com Conflict of interest statement W.R. Clark and J.J. Letteri are full-time employees of Gambro, Inc. Z. Huang has received research support from Gambro, Inc. REFERENCES 1. Clark WR, Henderson LW. Renal vs continuous vs intermittent therapies for removal of uremic toxins. Kidney Int 2001; 59 (Suppl 78): S Colton CK, Henderson LW, Ford C, Lysaght MJ. Kinetics of hemodiafiltration: Part I: in vitro transport characteristics of a hollow-fiber blood ultrafilter. J Lab Clin Med 1975; 85: Huang Z, Letteri JJ, Clark WR, Zhang W, Gao D, Ronco C. Ultrafiltration rate as dose surrogate in pre-dilution hemofiltration. Int J Artif Organs 2007; 30: Clark WR, Macias WL, Molitoris BA, Wang NH. Plasma protein adsorption to highly permeable hemodialysis membranes. Kidney Int 1995; 48: Clark WR, Ronco C. Continuous renal replacement techniques. Contrib Nephrol 2004; 144: Clark WR, Ronco C. Determinants of hemodialyzer performance and the effect on clinical outcome. Nephrol Dial Transplant 2001; 16 (Suppl 3): S Clark WR, Hamburger RJ, Lysaght MJ.. Effect of membrane composition and structure on performance and biocompatibility in hemodialysis. Kidney Int 1999; 56: Henderson LW. Biophysics of ultrafiltration and hemofiltration. In: Dortdrecht JC, ed. Replacement of renal function by dialysis (4th edn). Boston, MA: Kluwer Academic, 1996; Clark WR, Macias WL, Molitoris BA, Wang NH. ß2-Microglobulin membrane adsorption: equilibrium and kinetic characterization. Kidney Int 1994; 46: Clark WR. Quantitative characterization of hemodialyzer solute and water transport. Semin Dial 2001; 14: Kim S. Characteristics of protein removal in hemodiafiltration. Contrib Nephrol 1994; 108: Ronco C, Bellomo R, Hommel P, Brendolan A, Dan M, Piccinni P, LaGreca G. Effects of different doses in continuous veno-venous hemofiltration on outcomes in acute renal failure: a prospective, randomized trial. Lancet 2000; 355: Huang Z, Gao D, Clark WR. Convective renal replacement therapies for acute renal failure and end-stage renal disease. Hemodial Int 2004; 8: Clark WR, Turk JE, Kraus MA, Gao D. Dose determinants in continuous renal replacement therapy. Int J Artif Organs 2003; 27: Venkataraman R, Kellum JA, Palevsky P. Dosing patterns for continuous renal replacement therapy at a large academic medical center in the United States. J Crit Care 2002; 17: Saudan P, Niederberger M, De Seigneux S, Romand J, Pugin J, Perneger T, Martin PY. Adding a dialysis dose to continuous hemofiltration increases survival in patients with acute renal failure. Kidney Int 2006; 70: Troyanov S, Cardinal J, Geadah D, Parent D, Courteau S, Caron S, Leblanc M. Solute clearances during continuous venovenous haemofiltration at various ultrafiltration flow ra- 533

10 Renal replacement for acute kidney injury tes using Multiflow-100 and HF1000 filters. Nephrol Dial Transplant 2003; 18: Bonnardeaux A, Pichette V, Ouimet D, Geadeh D, Habel F, Cardinal J. Solute clearances with high dialysate flow rates and glucose absorption from the dialysate in continuous arteriovenous hemodialysis. Am J Kidney Dis 1992; 19: Brunet S, Leblanc M, Geadah D, Parent D, Courteau S, Cardinal J. Diffusive and convective solute clearances during continuous renal replacement therapy at various dialysate and ultrafiltration flow rates. Am J Kidney Dis 1999; 34: Salvatori G, Ricci Z, Bonello M, Ratanarat R, D Intini V, Brendolan A, Dan M, Piccinni P, Bellomo R, Ronco C. First clinical trial for a new CRRT machine: the Prismaflex. Int J Artif Organs 2004; 27: Liao Z, Zhang W, Poh CK, Huang Z, Hardy PA, Kraus MA, Clark WR, Gao D. Kinetic comparison of different acute dialysis therapies. Artif Organs 2003; 27: Schiffl H, Lang SM, Fischer R. Daily hemodialysis and the outcome of acute renal failure. N Engl J Med 2002; 346: Kumar VA, Craig M, Depner TA, Yeun JY. Extended daily dialysis: a new approach to renal replacement therapy in the intensive care unit. Am J Kidney Dis 2000; 36: Marshall MR, Golper TA, Shaver MJ, Alam MG, Chatoth DK. Sustained low-efficiency dialysis for critically ill patients requiring renal replacement therapy. Kidney Int 2001; 60: Clark WR, Leypoldt JK, Henderson LW, Mueller BA, Scott MK, Vonesh EF. Quantifying the effect of changes in the hemodialysis prescription on effective solute removal with a mathematical model. J Am Soc Nephrol 1999; 10: Kielstein JT, Kretschmer U, Ernst T, Hafer C, Bahr MJ, Haller H, Fliser D. Efficacy and cardiovascular tolerability of extended dialysis in critically ill patients: a randomized controlled study. Am J Kidney Dis 2004; 43: Tolwani AJ, Campbell RC, Stofan BS, Lai KR, Oster RA, Wille KM. Standard versus high-dose CVVHDF for ICU-related acute renal failure. J Am Soc Nephrol. Epub 2008 Mar The VA/NIH Acute Renal Failure Trial Network. Intensity of renal support in critically ill patients with acute kidney injury. N Engl J Med 2008; 359. Epub 2008 May Faulhaber-Walter R, Hafer C, Jahr N, Vahlbruch J, Haller H, Fliser D, Kielstein J. The Hannover-Dialysis-Outcome (HAND-OUT) Study: comparison of standard versus intensified extended daily dialysis in treatment of patients with septic acute renal failure on the intensive-care unit [abstract]. European Dialysis and Transplant Association Annual Congress, Barcelona, Spain Ricci Z, Ronco C, Bachetoni A, D Amico G, Rossi S, Alessandri E, Rocco M, Pietropaoli P. Solute removal during continuous renal replacement therapy in critically ill patients: convection versus diffusion. Crit Care 2006; 10: R