252 Clark et al. pore size has a relatively large impact on the ultrafil- tration capabilities of the membrane. B. Solute Removal by Diffusion Diffusion involves the mass transfer of a solute in re- sponse to a concentration gradient. For the extracor- poreal removal of a retained solute in an ARF patient, this concentration gradient exists across a semiperme- able membrane in a hemodialyzer or hemofilter. The inherent rate of diffusion of a solute is termed its dif- fusivity (3), whether this is in solution (such as dialy- sate and blood) or within an extracorporeal membrane. Diffusivity in solution is inversely proportional to sol- ute molecular weight and directly proportional to so- lution temperature. Solute diffusion within a membrane is influenced by both membrane thickness (diffusion path length) and membrane diffusivity (4), which is a function of both pore size and number (density). In conventional IHD, the overall mass transfer coefficient–area product (KoA) is used to quantify the diffusion characteristics of a particular solute-mem- brane combination (5). The overall mass transfer co- efficient is the inverse of the overall resistance to dif- fusive mass transfer, the latter being a more applicable quantitative parameter from an engineering perspec- tive: 1 Ko = Ro The overall mass transfer resistance can be viewed as the sum of resistances in series (3): Ro=Rbϩ Rm ϩ Rd where Rb, Rm, and Rd are the mass transfer resistances associated with the blood, membrane, and dialysate, respectively. In turn, each resistance component is a function of both diffusion path length (x) and diffusiv- ity (D): R = (x/D) ϩ (x/D) ϩ (x/D) OB M D The diffusive mass transfer resistance of both the blood and dialysate compartments for a hemodialyzer is primarily due to the unstirred (boundary) layer just adjacent to the membrane (6). Minimizing the thick- ness of these unstirred layers is primarily dependent on achieving relatively high shear rates, particularly in the blood compartment (7). For similar blood flow rates, higher blood compartment shear rates are achieved with a hollow fiber dialyzer than with a flat plate dia- lyzer. Indeed, based on the blood and dialysate flow rates (generally at least 250 and 500 mL/min, respec- tively) achieved in contemporary IHD with hollow fi- ber dialyzers, the controlling diffusive resistance is that due to the membrane itself. Another approach to quantifying diffusive mass transfer specifically through an extracorporeal mem- brane is use of Fick’s law of diffusion: N = D(dC/dx) where N is mass flux (mass removal rate normalized to membrane surface area), D is membrane diffusivity, an intrinsic membrane property for the particular solute being assessed, and dC/dx is the change in solute con- centration with respect to distance. This equation also can be expressed in a more applicable, integrated form: N=D(⌬C/⌬x) Thus, for a given concentration gradient across a mem- brane, the rate of diffusive solute removal is directly proportional to the membrane diffusivity and indirectly proportional to the effective thickness of the mem- brane. As described above, membrane diffusivity is deter- mined both by the pore size distribution and the num- ber of pores per unit membrane area (pore density). Diffusive mass transfer rates within a membrane de- crease as solute molecular weight increases due not only to effect of molecular size itself, but also to the resistance provided by the membrane pores (8). The difference in mean pore sizes between low-permeabil- ity dialysis membranes (e.g., regenerated cellulose) and high-permeability membranes (e.g., polysulfone, poly- acrylonitrile, cellulose triacetate) has a relatively small impact on small solute (urea, creatinine) diffusivities. This is related to the fact that even low-permeability membranes have pores sizes that are significantly larger than the molecular sizes of these solutes. However, as solute molecular weight increases, the tight pore struc- ture of the low-permeability membranes plays an in- creasingly constraining role such that diffusive removal of solutes larger than 1000 daltons is minimal by these membranes. On the other hand, the larger pore sizes that characterize high-flux membranes account for their higher diffusive permeabilities. In fact, based on the flow rates typically used in high-flux IHD, diffusion is the primary removal mechanism for solutes as large as inulin (5200 daltons) for all high-permeability mem- branes (9) and even ␤ 2 -microglobulin (11,000 daltons) for certain high-permeability membranes (10,11). Solute Removal in CRRT 253 C. Solute Removal by Convection Convective solute removal is primarily determined by the sieving properties of the membrane used and the ultrafiltration rate. The mechanism by which convec- tion occurs is termed solvent drag. If the molecular dimensions of a solute are such that sieving does not occur, the solute is swept (‘‘dragged’’) across the mem- brane in association with ultrafiltered plasma water. Thus, the rate of convective solute removal can be modified either by changes in the rate of solvent (plasma water) flow or in the mean effective pore size of the membrane. Both the water and solute permeability of an ultra- filtration membrane are influenced by the phenomena of secondary membrane formation (12) and concentra- tion polarization (13). The exposure of an artificial sur- face to plasma results in the nonspecific, instantaneous adsorption of a layer of proteins, the composition of which generally reflects that of the plasma itself. There- fore, plasma proteins such as albumin, fibrinogen, and immunoglobulins form the bulk of this secondary membrane. This layer of proteins, by serving as an ad- ditional resistance to mass transfer, effectively reduces both the water and solute permeability of an extracor- poreal membrane. Evidence of this is found in com- parisons of solute sieving coefficients determined be- fore and after exposure of a membrane to plasma or other protein-containing solution (8). In general, the ex- tent of secondary membrane development and its effect on membrane permeability is directly proportional to the membranes adsorptive tendencies (i.e., hydropho- bicity). Therefore, this process tends to be most evident for high-flux synthetic membranes, such as polyacryl- onitrile, polysulfone, and polymethylmethacrylate. Although concentration polarization (13) primarily pertains to plasma proteins, it is distinct from second- ary membrane formation. Concentration polarization specifically relates to ultrafiltration-based processes and applies to the kinetic behavior of an individual protein. Accumulation of a plasma protein that is predominantly or completely sieved (rejected) by a membrane used for ultrafiltration of plasma occurs at the blood com- partment membrane surface. This surface accumulation causes the protein concentration just adjacent to the membrane surface (i.e., the submembranous concentra- tion) to be higher than the bulk (plasma) concentration. In this manner, a submembranous (high) to bulk (low) concentration gradient is established, resulting in ‘‘back-diffusion’’ from the membrane surface out into the plasma. At steady state, the rate of convective trans- port to the membrane surface is equal to the rate of backdiffusion. The polarized layer of protein is the dis- tance defined by the gradient between the submem- branous and bulk concentrations. This distance (or thickness) of the polarized layer, which can be esti- mated by mass balance techniques, reflects the extent of the concentration polarization process. By definition, concentration polarization is applica- ble in clinical situations in which relatively high ultra- filtration rates are used. Therefore, in ARF, concentra- tion polarization may play a significant role in CVVH and CVVHDF, and the specific operating conditions used in these therapies influence the polarization pro- cess. Conditions that promote the process are high ul- trafiltration rate (high rate of convective transport), low blood flow rate (low shear rate), and the use of post- dilution (rather than predilution) replacement fluids (in- creased local protein concentrations) (14). The extent of the concentration polarization deter- mines its effect on actual solute (protein) removal. In general, the degree to which the removal of a protein is influenced is directly related to that protein’s extent of rejection by an individual membrane. In fact, con- centration polarization actually enhances the removal of a molecular weight class of proteins (30,000–70,000 daltons) that otherwise would have minimal convective removal. This is explained by the fact that the pertinent blood compartment concentration subjected to the ul- trafiltrate flux is the high submembranous concentra- tion primarily rather than the much lower bulk concen- tration. Therefore, the potentially desirable removal of certain proteins in this size range in ARF patients has to be weighed against the undesirable increase in con- vective albumin losses. This concern is particularly rel- evant in light of the growing interest in the use of high- volume hemofiltration (Ն6 L/h) for the treatment of septic conditions (with or without ARF) (15,16). On the other hand, the use of very high ultrafiltration rates in conjunction with other conditions favorable to protein polarization may significantly impair overall membrane performance. The relationship between ul- trafiltration rate and transmembrane pressure (TMP) is linear for relatively low ultrafiltration rates, and the positive slope of this line defines the ultrafiltration co- efficient of the membrane. However, as ultrafiltration rate further increases, this curve eventually plateaus (13). At this point, maintenance of a certain ultrafiltra- tion rate is only maintained by a concomitant increase in TMP. At sufficiently high TMP, fouling of the mem- brane with denatured proteins may occur and an irre- versible decline in solute and water permeability of the membrane ensues. Therefore, the ultrafiltration rate (and associated TMP) used for a convective therapy 254 Clark et al. with a specific membrane needs to fall on the initial (linear) portion of the UFR versus TMP relationship with avoidance of the plateau region. Convective solute removal can be quantified in the following manner (17): N=(1Ϫ ␴)Jv Cm where N is the convective flux (mass removal rate per unit membrane area), Jv is the ultrafiltrate flux (ultra- filtration rate normalized to membrane area), Cm is the mean intramembrane solute concentration, and ␴ is the reflection coefficient, a measure of solute rejection. As Werynski and Waniewski have explained (17), the pa- rameter (1 Ϫ ␴) can be viewed as the membrane re- sistance to convective solute flow. If ␴ equals 1, no convective transport occurs, while a value of 0 implies no resistance to convective flow. Of note, the appro- priate blood compartment concentration used to deter- mine Cm is the submembranous concentration rather than the bulk phase concentration. Therefore, this pa- rameter is significantly influenced by the effects of con- centration polarization. It is useful to individually assess the parameters on the right-hand side of the above equation and the man- ner in which changes in these parameters may affect the rate of convective solute transport. During a RRT, changes in the permeability properties of the hemofilter membrane or in the operating conditions may alter these parameters. However, a complex interplay exists between these parameters, and the net effect of changes in hemofilter membrane permeability or RRT operating conditions may be difficult to predict. To illustrate this point, the effect of a progressive decrease in membrane permeability as a membrane becomes fouled with pro- teins can be assessed. As a membrane becomes fouled with plasma proteins, the resistance to convective sol- ute flow (␴) increases such that the parameter (1 Ϫ ␴) decreases. In addition, fouling may result in a decrease in ultrafiltrate flux (Jv) despite attempted increases in TMP. This phenomenon is most relevant for CRRT sys- tems operated without a blood pump, such as CAVH and CAVHD. However, when the membranes become irreversibly fouled (i.e., gel formation occurs), even a hemofilter used in a venovenous system loses ultrafil- tration capabilities. Finally, polarization of solute at the membrane surface due to the fouling causes an increase in the submembranous blood compartment concentra- tion but a decrease in the filtrate concentration. The net effect on Cm, which essentially is a mean of the sub- membranous and filtrate concentrations, is difficult to predict and depends on the specific solute in question. In general, however, except for relatively large proteins capable of only minimal convective transport (e.g., al- bumin), fouling results in a decrease in Cm because the decrease in filtrate concentration is predicted to be greater than the increase in the submembranous con- centration. D. Interaction Between Diffusion and Convection In IHD and some continuous therapies, diffusive and convective solute removal occur simultaneously. How- ever, the effect of this combination on the total removal of a specific solute differs between intermittent and slow continuous therapies. In IHD, diffusion and con- vection interact in such a manner that total solute re- moval is significantly less than what would be expected if the individual components are simply added together. This phenomenon is explained in the following way. Diffusive removal results in a decrease in solute con- centration in the blood compartment along the axial length (i.e., from blood inlet to blood outlet) of the hemodialyzer/hemofilter. As convective solute removal is directly proportional to the blood compartment con- centration, convective solute removal decreases as a function of this axial concentration gradient. On the other hand, hemoconcentration resulting from ultrafil- tration of plasma water causes a progressive increase in plasma protein concentration and hematocrit along the axial length of the filter. This hemoconcentration and resultant hyperviscosity causes an increase in dif- fusive mass transfer resistance and a decrease in solute transport by this mechanism. The effect of this inter- action on overall solute removal in IHD has been an- alyzed rigorously by numerous investigators (17,18). The most useful quantification has been developed by Jaffrin (18): Kt=Kdϩ Qf ϫ Tr where Kt is total solute clearance, Kd is diffusive clear- ance under conditions of no ultrafiltration, and the final term is the convective component of clearance. The latter term is a function of the ultrafiltration rate (Qf) and an experimentally derived transmittance coefficient (Tr), such that: Tr = S(1 Ϫ Kd/Qb) where S is solute sieving coefficient. Thus, Tr for a particular solute is dependent on the efficiency of dif- fusive removal. At very low values of Kd/Qb, diffusion has a very small impact on blood compartment con- centrations and the convective component of clearance closely approximates the quantity S ϫ Qf. However, Solute Removal in CRRT 255 Table 1 Determinants of Solute Removal in IHD and CRRT IHD CRRT Small solutes (mw < 300) Diffusion: Q B Q D Membrane thickness Diffusion: Q D Convection: Q F Middle molecules (mw = 500–5,000) Diffusion Convection: Q F SC Convection: Q F Diffusion LMW proteins (mw = 5,000–50,000) Convection Diffusion Adsorption: site availability Convection Adsorption: site availability Large proteins (mw > 50,000) Convection Convection Q B , blood flow rate; Q D , dialysate flow rate; Q F , ultrafiltration rate; SC, sieving coefficient. Source: Ref. 62. with increasing efficiency of diffusive removal (i.e., in- creasing Kd/Qb), blood compartment concentrations are significantly influenced. The result is a decrease in Tr and, consequently, in the convective contribution to total clearance. Due to the markedly lower flow rates used in CRRT, the effect of simultaneous diffusion and convection on overall solute removal is quite different. Based on a comparison of clearances, the rate of diffusive removal of small solutes in CAVHD, CVVHD, or CVVHDF (17–34 mL/min) (19 –23) is only approximately 5– 15% of the rate achieved in IHD. Therefore, the small solute concentration gradient along the axial length of the filter (i.e., extraction) is minimal compared to that which is seen in an IHD setting, in which extraction ratios of 50% or more are the norm. This difference is demonstrated in the following comparison of CVVHD and IHD, both operated in the pure dialysis (diffusive) mode. For typical blood and dialysate flow rates of 300 and 500 mL/min, respectively, an expected diffusive urea clearance is approximately 200 mL/min for a high-efficiency dialyzer used in an ARF IHD applica- tion. Based on this clearance and an assumed arterial line BUN of 60 mg/dL, the resultant venous line BUN is 20 mg/dL. This significant decrease in the BUN oc- curring along the axial length of the dialyzer reduces potential convective solute removal, as explained above. On the other hand, typical blood and dialysate flow rates in a strictly diffusive CVVHD procedure are 200 and 17 mL/min, respectively (21,22), which result in a urea clearance of 17 mL/min due to saturation of the effluent dialysate stream (dialysis equilibrium) (20–22). Based on this clearance and the same arterial line BUN of 60 mg/dL, the resultant venous line BUN is 55 mg/dL. Thus, the minimal diffusion-related change in small solute concentrations along the filter allows any additional clearance related to convection to be simply additive to the diffusive component. In- deed, in a classic paper (20), Sigler and Teehan dem- onstrated this lack of interaction between diffusion and convection in a series of patients treated with CAVHD operating at a dialysate flow rate of 1 L/h and an ul- trafiltration rate range of 4 to 10 mL/min. III. SOLUTE REMOVAL MECHANISMS: INTERMITTENT HEMODIALYSIS VERSUS CRRT Application of the above principles allows a compari- son of solute-removal mechanisms for extracorporeal RRT used in ARF (Table 1). Solutes are divided into four categories: small solutes (<300 daltons), middle molecules (500 –5,000 daltons), low molecular weight (LMW) proteins (5,000–50,000 daltons), and large proteins (>50,000 daltons). Except for the LMW pro- tein category, the prototypical molecules (surrogates) in each category are similar for both ESRD and ARF. These common prototypical solutes are (a) urea, cre- atinine, phosphate, and amino acids (small solutes), (b) vitamin B 12 , vancomycin (24,25), and inulin (middle molecules), and (c) albumin (large molecules). For the LMW protein category, ␤ 2 -microglobulin is the focus in ESRD therapies (26), while inflammatory mediators, such as complement pathway products (MW 9–23 kDa) and cytokines (MW 15–50 kDa), are more of interest in the ARF setting (27–30). 256 Clark et al. As is the case in ESRD, optimized removal of sol- utes in the small solute, middle molecule, and LMW protein categories and minimal removal of albumin are therapy goals in ARF. However, as Table 1 indicates, the mechanisms by which solute removal within a par- ticular category occurs may differ significantly between the two types of therapies. For patients receiving IHD, small solute removal occurs almost exclusively by dif- fusion (24). As such, optimized small solute removal is achieved by employing dialysis conditions that min- imize diffusive mass transfer resistances, such as high flow rates and thin membranes (2). Likewise, for sol- utes in the middle molecule category, removal by high- flux IHD occurs predominantly by diffusion (31). Al- though LMW protein removal by high-flux dialyzers occurs primarily by convection or adsorption, diffusion can even play a significant role in the removal of sol- utes in the class (e.g., ␤ 2 -microglobulin) for some membranes (32). Only for a solute whose molecular weight is similar to or larger than that of albumin is convection essentially the sole removal mechanism during high-flux IHD. Recent ESRD data (33) dem- onstrate total protein losses during high-flux dialysis may be significant (up to 15 –20 g per treatment), at least for certain membrane-reuse combinations. Protein losses for IHD have not been quantified in ARF. The predominant mass transfer mechanism for each class of solutes may be significantly different for the slow continuous therapies. Small solute removal can occur exclusively by convection in CVVH (34,35), pre- dominantly by diffusion in CVVHD (21,22), or by ap- proximately equal contributions of both diffusion and convection in CVVHD (23). For a properly functioning filter, small solute sieving coefficients during CVVH are close to unity (36,37) such that clearances for these solutes are primarily determined by the ultrafiltration rate and the mode of replacement fluid administration (predilution vs. postdilution) (38). For the diffusion- based continuous therapies employing dialysate flow rates of 2 L/h or less, urea and creatinine clearances approximate the effluent dialysate flow rate because of the existence of dialysis equilibrium (20–22). For mid- dle molecule removal, Jeffrey et al. (25) have recently shown that convection is more important than diffusion for a surrogate solute (vancomycin: MW, 1448) when the same ultrafiltration rate (CVVH) and effluent dialy- sate flow rate (CVVHD) of 25 mL/min (1.5 L/h) is used. As the relative importance of convection in- creases with solute molecular weight, transmembrane removal of LMW proteins in ARF patients occurs al- most exclusively by this mechanism. However, adsorp- tive removal of inflammatory mediators in this class has also been demonstrated, and considerable contro- versy currently exists as to whether convection or ad- sorption optimizes mediator removal. Finally, in con- trast to the above IHD data for ESRD patients, Mokrzycki and Kaplan (39) have recently reported a relatively modest mean total protein loss of 1.6 g/day in seven CRRT patients. IV. REMOVAL OF SMALL NITROGENOUS WASTE PRODUCTS IN ARF A. Factors Influencing Small Solute Removal in ARF Factors that influence and impair small solute removal in ARF can be either patient related or therapy related. In the latter category, some of these factors are directly related to filter performance, while others relate to other aspects of the RRT. Protein hypercatabolism, total body water, and body size are all patient-related factors that significantly im- pact the degree to which small solute removal provides azotemia control in ARF. Acute renal failure in the ICU epitomizes a non–steady-state condition, as urea gen- eration rates and protein catabolic rates (PCRs) have been reported to vary on a daily basis (40,41). Protein hypercatabolism is nearly always present in this setting, with net normalized PCR (nPCR) values of 1.5 g/kg/ day or greater and net nitrogen deficits of 6 g/day or greater routinely reported (40,42– 44). The nPCR values in ARF are reflective of the met- abolic perturbations associated with ARF. The manner in which nPCR changes with time in critically ill pa- tients treated with a CRRT has been reported to be quite variable. Clark et al. (40) found a linear relation- ship between nPCR and time, ranging from 1.5 to 1.9 g/kg/day over the first several days of therapy in pa- tients treated with CVVH. On the other hand, Chima et al. (41) described an essentially random variation of nPCR with time in patients receiving CAVH. Body size and the extent of volume overload in ARF patients are also critical considerations in RRT pre- scription. For both nonuremics and patients with ESRD, numerous previous investigations have docu- mented that total body water closely approximates urea distribution volume, with values reported to be 0.55– 0.60 L/kg of lean body mass (45–48). However, the relationship between V and lean body mass in ARF patients is not nearly as well defined. Several factors in ARF make determination of this relationship quite Solute Removal in CRRT 257 difficult. These factors include severe volume overload and ongoing catabolism of lean body mass. Clark et al. (49) recently reported a mean V of 65% of body weight in a group of 11 hypercatabolic ARF patients whose mean IHD characteristics included 13 dialyses over a 24-day period. In concert with catabo- lism-induced loss of lean body mass, volume overload most likely accounts for the markedly higher fractional urea distribution volumes in ARF than those in ESRD patients and normal individuals. On the other hand, Clark et al. (50) found the mean value of V to be 0.55 L/kg of body weight in a group of 11 critically ill pa- tients receiving CVVH at steady state. As shown recently by Evanson et al. (51), failure to account for these volume disturbances may result in large discrepancies between prescribed and delivered HD doses. In a group of 45 patients who received a total of 136 HD treatments, these investigators used dialyzer KoA, prescribed blood flow rate and time, and a value of V equal to 0.60 ϫ pre-HD body weight to estimate prescribed Kt/V. Delivered Kt/V was esti- mated by an equation employing pre-HD and post-HD BUN values, a technique that may be problematic in ARF (see below). Nonetheless, a significant difference was observed between prescribed and delivered Kt/V per treatment (1.26 Ϯ 0.45 vs. 1.04 Ϯ 0.49, respec- tively; mean Ϯ SD). This difference appeared to be related primarily to the use of an estimated V, for pre- scription purposes, that was significantly less than the actual (kinetically derived) V. Our group has also high- lighted the detrimental effect on expected small solute removal if volume overload is neglected (52). In ad- dition, the large discrepancy between prescribed and delivered ARF dialysis doses observed in the Evanson et al. study has been corroborated by others (53). Severe volume overload may also adversely influ- ence small solute removal in relation to the large ultra- filtration requirements during IHD. During the rela- tively short duration of IHD (compared to CRRT), rapid osmolarity changes occur as large solute loads are removed from hypercatabolic patients. Especially in the early phase of a dialysis treatment, these osmolar changes may create a gradient for water movement from the intravascular space to the interstitial space. This water movement, in combination with the intra- vascular volume depletion occurring by extracorporeal ultrafiltration, may cause significant hypotension. A po- tential solution to this problem is the use of sequential ultrafiltration/dialysis (54), which involves an increase in overall treatment time if total urea removal is to be maintained. On the other hand, total urea removal is sacrificed if treatment time is kept constant. Dialytic sodium modeling is an alternative solution to this prob- lem, but formal reports describing its use in ARF are presently lacking. A number of RRT-related factors also influence small solute removal. Access recirculation may ad- versely affect the small solute clearances of any RRT. Although extensively investigated in ESRD patients with permanent (nonpercutaneous) vascular accesses (55,56), the determinants of percutaneous access recir- culation in ARF patients are not as well characterized. Percutaneous catheters designed for long-term use in chronic hemodialysis have been shown by Twardowski et al. (57) to have very low (Ϸ2%) degrees of recir- culation. At a blood flow rate of 250 mL/min, Kelber et al. (58) reported comparably low values for subcla- vian and internal jugular catheters used for IHD in ARF. However, mean recirculation was 10% for 24 cm femoral catheters while shorter (15 cm) femoral cath- eters exhibited an even greater value of 18%. At a blood flow rate of 400 mL/min, the value of this latter measurement increased to 38%. These data have re- cently been corroborated by Leblanc et al. (59). For small solutes, diffusive mass transfer resistances are an important consideration, and failure to apply the general principles discussed above may result in im- paired removal. A widespread misconception is that be- cause of the relatively open pore structure of highly permeable dialyzers (Kuf > 20 mL/h/mmHg), their urea removal capabilities are necessarily superior to those of low-permeability dialyzers (Kuf < 10 mL/h/mmHg). However, the thicknesses of highly permeable synthetic membranes (Ն25 ␮m) are substantially larger than those of low-flux cellulosic membranes, most of which have thicknesses of <10 ␮m (2). At blood flow rates (<300 mL/min) typically employed in ARF, the urea clearances for the two types of dialyzers are actually very similar. Thus, the enhanced diffusivity of urea in highly permeable synthetic membranes is negated by the large diffusive resistance associated with their rel- atively thick structures. To illustrate this point, in vitro urea clearances for a low-flux modified cellulosic dialyzer (Hemophan membrane: thickness Ϸ8 ␮m) and a high-flux poly- acrylonitrile dialyzer (AN69 membrane: thickness Ϸ25 ␮m) can be compared. At an in vitro blood flow rate of 200 mL/min, the urea clearance of a 0.9 m 2 Hem- ophan dialyzer (117 mL/min) is actually about 6% greater than that (166 mL/min) of a AN69 dialyzer with comparable surface area (1.0 m 2 ) (60). Although in- creasing blood flow rate would have a relatively greater impact on urea clearance for the high-flux dialyzer, this comparison still attests to the importance of membrane 258 Clark et al. Table 2 Continuous Renal Replacement Therapy Clearance Rate (mL/h)/Intermittent Hemodialysis Frequency (per week) Requirements for Varying Levels of Azotemia Control a Weight (kg) BUN = 60 mg/dL BUN = 80 mg/dL BUN = 100 mg/dL 50 886/4.4 668/3.2 535/<3.0 60 1097/5.2 823/3.8 649/3.0 70 1300/6.0 977/4.4 763/3.5 80 1500/6.9 1123/5.0 886/4.0 90 1686/NA 1279/5.6 1018/4.5 100 1911/NA 1432/6.2 1133/5.0 a BUN value is either continuous renal replacement therapy steady- state BUN or intermittent hemodialysis time-averaged BUN. NA, not achievable with daily dialysis. Source: Ref. 61. thickness in determining small solute clearances. In ad- dition, this comparison confirms the importance of us- ing fundamental mass transfer principles in choosing an extracorporeal device for ARF patients. Once an extracorporeal device and a specific RRT is chosen, adequate therapy prescription and delivery is imperative so that a selected target for metabolic control can be achieved. Two issues are pertinent in this regard. First, as previously discussed and as is the case in chronic hemodialysis, the amount of prescribed therapy is nearly always greater than the amount deliv- ered (51,53). Second, at present, exactly what should be the targets for metabolic control in both IHD and CRRT remain to be defined (see below). Nonetheless, the clinician needs to have a specific target in mind when a RRT is prescribed. B. Avoidance of Underdialysis: Use of Urea Kinetic Methods to Guide Therapy Prescription The recognition that both morbidity and mortality are inversely related to delivered HD dose in ESRD pa- tients has substantially changed clinical practices in the United States (60a,60b). A number of urea-based quan- tification methods that differ greatly in complexity and usefulness now are used in this setting. Investigators have recently begun to extrapolate some of these ESRD quantification techniques to the ARF setting. Examples of this are discussed below. We have recently developed a computer-based model designed to permit individualized RRT prescrip- tion for ARF patients (61). The critical input parameter is the desired level of metabolic control, which is the time-averaged BUN (BUN a ) or steady-state BUN (BUN s ) for IHD or CRRT, respectively. The basis for the model was a group of 20 patients who received uninterrupted CRRT for at least 5 days. In these pa- tients, the nPCR increased linearly (r = 0.97) from 1.55 Ϯ 0.14 g/kg/day (mean Ϯ SEM) on day 1 to 1.95 Ϯ 0.15 g/kg/day on day 6. The daily value of G, deter- mined from the above linear relationship, was em- ployed to produce BUN versus time curves by the di- rect quantification method for simulated patients of varying dry weights (50–100 kg) who received varia- ble CRRT clearances (500–2000 mL/h). Steady-state BUN versus time profiles for the same simulated pa- tient population treated with IHD regimens (K = 180 mL/min; t = 4 hr per treatment) of variable frequency were generated by use of the variable-volume single pool kinetic model. From these profiles, regression lines of required IHD frequency (per week) versus pa- tient weight for desired BUN a values of 60, 80, and 100 mg/dL were obtained. Regression lines of required CRRT urea clearance (mL/h) versus patient weight for desired BUN s values of 60, 80, and 100 mg/dL were also generated. The required amounts of IHD (treat- ment frequency) and CRRT (urea clearance) at these three levels of azotemic control were compared. The results of these analyses appear in Table 2 and Figs. 1 and 2. For the attainment of intensive metabolic control (BUN a = 60 mg/dL) at steady state, a required treatment frequency of 4.4 dialyses per week is pre- dicted for a 50-kg patient. However, the model predicts that the same degree of metabolic control cannot be achieved even with daily IHD therapy in hypercata- bolic ARF patients weighing more than 90 kg. Con- versely, for the attainment of intensive CRRT metabolic control (BUN s = 60 mg/dL), required urea clearances of approximately 900 and 1900 mL/h are predicted for 50-kg and 100-kg patients, respectively. Therefore, this model suggests that for many patients, rigorous azote- mia control equivalent to that readily attainable with most CRRTs can only be achieved with intensive IHD regimens. Therefore, these modeled data suggest that the complication of inadequate azotemic control is less likely to occur in hypercatabolic ARF patients if a CRRT is used. We also assessed the effect of variable IHD inter- mittence by plotting both IHD BUN a and CRRT BUN s versus the ratio nPCR/(Kt/V) d , where the denominator in the latter term represents the normalized daily ther- apy dose. As previously predicted and shown for pa- tients with nESRD (45) and ARF (40), a linear rela- tionship was observed when these regression analyses Solute Removal in CRRT 259 Fig. 1 Predicted CRRT urea clearance required for the attainment of varying desired levels of steady-state azotemia control (BUNs). The clearances shown are for patients ranging in size from 50 to 100 kg. The target BUNs values for curves A, B, and C are 100, 80, and 60 mg/dL, respectively. (From Ref. 61.) Fig. 2 Predicted IHD frequencies required for the attainment of varying desired levels of time-averaged azotemia control (BUNa). The frequencies are shown for patients ranging in size from 50 to 100 kg. The target BUNa values for curves A, B, and C are 100, 80, and 60 mg/dL, respectively. (From Ref. 61.) 260 Clark et al. Fig. 3 Steady-state RRT azotemia control versus the ratio nPCR/(Kt/V)d. The curves are shown for a patient of 70 kg dry weight. The CRRT line represents BUNs values, whereas the IHD line represents BUNa values. (From Ref. 61.) were performed (Fig. 3). The two regression lines shown are for simulated patient of dry weight 70 kg. Because nPCR was constant in these steady state sim- ulations (1.95 g/kg/day), variations in the abscissa were attributable entirely to changes in (Kt/V) d . In turn, changes in therapy dose were related to changes in K for CRRT and in treatment frequency for IHD. There- fore, the points determining the CRRT line represent K values ranging from 750 mL/h [highest nPCR/(Kt/V) d value] to 2000 mL/h [lowest nPCR/(Kt/V) d value]. Conversely, the points on the IHD line represent treat- ment frequencies ranging from three per week [highest nPCR/(Kt/V) d value] to seven per week [lowest nPCR/ (Kt/V) d value]. This figure shows that the degree of divergence between the CRRT BUN s and IHD BUN a lines decrease with increasing IHD frequency [i.e., de- creasing nPCR/(Kt/V) d ]. This convergence shows that the inherent inefficiency associated with an intermittent therapy, relative to that of a continuous therapy, de- creases with increasing frequency. Therefore, if IHD is the chosen therapy and the complication of inadequate metabolic control is to be avoided, high therapy fre- quency has specific benefits (62). In addition to the benefits specifically pertaining to the kinetics of solute removal, increased IHD frequency may result in de- creased ultrafiltration requirements per treatment. The avoidance of hypotensive episodes related to rapid ul- trafiltration rates may also indirectly improve solute re- moval by decreasing the risk of therapy interruptions. C. Small Solute Control in ARF: Effect of Amount of Delivered Therapy on Outcome Based on presently available data, precise targets for optimal metabolic control are not able to be provided for ARF patients treated with either IHD or CRRT. However, at least for IHD, rough guidelines exist. Kjellstrand has suggested that IHD should be initiated before the BUN reaches 100 mg/dL and that therapy should be delivered at a level to maintain the pre-di- alysis BUN below 100 mg/dL (63). Support for these recommendations is found in early comparative studies in which groups of patients received substantially dif- ferent levels of IHD therapy (64– 66). In these inves- tigations, survival was directly correlated with IHD in- tensity as measured by predialysis BUN, which ranged from approximately 90 to 150 mg/dL. In a more contemporary study, Gillum et al. (67) reported results from a multicenter, prospective study in which the effect of dialysis intensity on survival in patients with ARF was investigated. In this trial, a total of 34 patients with diverse ARF etiologies received ei- ther ‘‘intensive’’ or ‘‘nonintensive’’ dialysis. Daily di- Solute Removal in CRRT 261 alysis of 5–6 hours per treatment was generally pre- scribed in the intensive group, while the regimen in the nonintensive group consisted of 5-hour treatments ad- ministered daily to every third day. Mean predialysis azotemia control achieved in the two groups was very close to the target BUN and serum creatinine values of 60 and 5 mg/dL, respectively (intensive group), and 100 and 9 mg/dL, respectively (nonintensive group). However, prescribed blood and dialysate flow rates were not provided. In addition, data permitting an es- timation of the rate of interdialytic urea generation were not reported. Therefore, neither dialysis dose nor PCR could be estimated. Nevertheless, survival in the intensively treated group (41%) did not differ signifi- cantly from that in the nonintensive group (52%). Although serum creatinine was used as an efficacy parameter in the above study, recent data suggest that this parameter should not be used in this manner. Our group recently quantified steady-state creatinine kinetic parameters in a group of 11 critically ill ARF patients who received CVVH (68). In these patients, of whom four were women, the mean pretreatment serum cre- atinine was 5.6 Ϯ 2.6 mg/dL, while the value was 3.4 Ϯ 1.7 mg/dL at steady state, which occurred after a mean treatment period of approximately one week. A significant linear relationship was observed between steady-state serum creatinine and both creatinine gen- eration rate and lean body mass. Normalized to body weight, mean lean body mass was found to be 0.51 Ϯ 0.09 kg/kg, a value significantly lower than previously reported for both normal and ESRD patients. These data suggest that the steady-state serum creatinine is best viewed as a nutritional parameter rather than a therapy efficacy parameter in this patient population. Indirect support for this contention comes from recent data of Paganini et al. (53), who report that death is associated with a low rate of rise of serum creatinine in the ICU ARF population (see below). In a recent study of 58 consecutive ICU ARF pa- tients receiving IHD at the Cleveland Clinic, Tapolyai et al. (69) correlated patient outcome (survival vs. death) with a variety of patient-related and dialysis- related parameters. Patient demographics, hemody- namic status, and illness severity scores were similar in surviving and nonsurviving patients. Dialysis dose for each treatment was estimated by calculation of sin- gle-pool Kt/V. The prescribed Kt/V was not signifi- cantly different between the two groups. However, the mean delivered Kt/V per treatment was significantly higher among survivors (survivors: 1.09; nonsurvivors: 0.89). Although the actual Kt/V determination method was not specified in this preliminary report, these data suggest that survival in critically ill patients is corre- lated directly with delivered IHD dose. In a more recent publication (53), the Cleveland Clinic group has extended this analysis. These inves- tigators assessed outcome in 842 ICU ARF patients who received RRT between 1988 and 1994 at their in- stitution. The Cleveland Clinic Foundation (CCF) ARF scoring system (70) was employed to estimate illness severity. In this system, 23 different demographic, clin- ical, and laboratory parameters are used to produce a score ranging from 0 (low mortality) to 20 (high mor- tality). Eight factors were found to be associated strongly with poor patient outcome, including need for mechanical ventilation, leukopenia, thrombocytopenia, number of nonrenal organ system failure, and a low rate of increase in the serum creatinine. When patient outcome was adjusted for the CCF outcome score, sur- vival was correlated with delivered IHD (Kt/V > 1.0 per treatment). In a prospective study, Schiffl et al. (71) randomized 72 ICU ARF patients to either daily IHD or every other day IHD. Overall mortality in this study was 35% but was significantly lower in the daily IHD group (21%) than in the alternate day group (47%). Using weekly Kt/V as the discriminating parameter, these investiga- tors observed a significantly lower mortality in the high-dose group (Kt/V > 6.0 per week; 16%) than in the low-dose group (Kt/V < 3.0 per week; 57%). Both of the above studies have employed a single- pool quantification technique developed specifically for the ESRD population (72). The equation used in these studies contains constants accounting for the effects of intradialytic urea generation and ultrafiltration on de- livered dose. However, these constants were generated from ESRD patients. Therefore, extrapolation of this or any other equation developed specifically for ESRD pa- tients to ARF patients may be problematic, as recently demonstrated by Lo et al. (73). Recent studies also suggest that the intensity of CRRT influences outcome. Data from Storck et al. (74) suggest that greater intensity of CRRT produces is as- sociated with better patient outcomes. In this study, pa- tients were treated with either CAVH or CVVH such that a wide range of ultrafiltration rates was obtained. Survival was found to be significantly higher in the CVVH group than in the CAVH group, in which the mean ultrafiltration rates were 15.5 and 7.5 L/day, re- spectively. Whether the superior survival in the patients treated with CVVH rather than CAVH was related to the former’s greater convective removal of small sol- utes or larger substances could not be determined from the data provided. In addition, data from Paganini et [...]... Alaka KJ, Mueller BA, Macias WL A comparison of metabolic control by continuous and intermittent therapies in acute renal failure J Am Soc Nephrol 19 94; 4: 141 3 – 142 0 Chima CS, Meyer L, Hummell AC, Bosworth C, Heyka R, Paganini E, Werynski A Protein catabolic rate in patients with acute renal failure on continuous arterio- 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 venous hemofiltration and total... clearance of vancomycin during hemodialysis using polysulfone membranes Kidney Int 1989; 35: 140 9 – 141 2 268 87 88 89 90 Clark et al Quale J, O’Halloran J, DeVincenzo N, Barth R Removal of vancomycin by high-flux hemodialysis membranes Antimicrob Agents Chemother 1992; 36: 142 4 – 142 6 Bohler J, Reetze-Bonorden R, Keller E, Kramer A, Schollmeyer P Rebound of plasma vancomycin levels after hemodialysis... tachycardia were found in 27% of 92 patients with 2 4- hour Holter monitoring (73) The finding of high-grade ventricular arrhythmias in the presence of coronary artery disease was associated with increased risk of cardiac mortality and sudden death (72, 74) Whereas the dialysis method, membrane, and buffer used do not seem to have a direct effect on the incidence of arrhythmias (75), dialysis- associated hypotension... precipitating high-grade ventricular arrhythmias, irrespective of the type of dialysis (75,76) Use of digoxin in hemodialysis patients has raised concern regarding precipitation of arrhythmias, especially in the immediate postdialysis period, when both hypokalemia and relative hypercalemia may occur (71,72,77) Keller et al (78) studied 55 patients in a crossover study of ‘‘on-and-off’’ digoxin and... catheters for acute hemodialysis Clin Nephrol 1996; 45 :315 – 319 60 Van Stone J: Hemodialysis apparatus In: Daugirdas J, Ing T, eds Handbook of Dialysis 2d ed Boston: Little, Brown and Company, 19 94: 32 – 38 60a Clark WR, Rocco MV, Collins AJ Quantification of hemodialysis: analysis of methods and the relevance to patient outcome Blood Purif 1997; 15:92 – 111 60b NKF-DOQI Hemodialysis Adequacy Work Group... 108 Henderson LW Biophysics of ultrafiltration and hemofiltration In: Jacobs C, ed Replacement of Renal Function by Dialysis 4th ed Dordrecht: Kluwer Academic Publishers, 1995:1 14 – 118 Ofsthun NJ, Zydney AL Importance of convection in artificial kidney treatment Contrib Nephrol 19 94; 108: 53 – 70 Scott MK, Mueller BA, Clark WR Effect of membrane type on the performance of bleach-reprocessed highflux dialyzers... first 2 years of follow-up, after adjustments for age, primary renal diagnosis, comorbid conditions, and center size (28) Fig 3 Survival in a cohort of patients treated with hemo(Ⅲ) or peritoneal dialysis ( -) (A) The survival according to mode of dialysis therapy at 3 months (B) The survival in patients treated exclusively with one mode of dialysis therapy (From Ref 27.) Most of the above-mentioned data... of new dialysis patients, followed for a mean of 41 months, 133 patient had heart failure at baseline and 56% (N = 75) had recurrent heart failure during follow-up Two hundred and ninety-nine were free of heart failure on initiation of dialysis and 25% (N = 76) developed de novo heart failure (5) In patients treated only with peritoneal dialysis, 16.5% developed de novo heart failure versus 28.1% of. .. 38 39 40 41 Clark et al and kinetic characterization Kidney Int 19 94; 46 :1 140 – 1 146 Pascual M, Schifferli JA Adsorption of complement factor D by polyacrylonitrile dialysis membranes Kidney Int 1993; 43 :903 – 911 Goldfarb S, Golper TA Proinflammatory cytokines and hemofiltration membranes J Am Soc Nephrol 19 94; 5: 228 – 232 Bellomo R, Tipping P, Boyce N Tumor necrosis factor clearances during veno-venous... Co., 1981: 242 5 – 248 9 Colton C Analysis of membrane processes for blood purification Blood Purif 1987; 5:202 – 251 Sargent J, Gotch F Principles and biophysics of dialysis In: Maher J, ed Replacement of Renal Function by Dialysis 3rd ed Dordrecht: Kluwer Academic Publishers, 1989:89 – 91 Colton CK, Lysaght M Membranes for hemodialysis In: Jacobs C, ed Replacement of Renal Function by Dialysis 4th ed Dordrecht: . Mueller BA, Macias WL. A com- parison of metabolic control by continuous and inter- mittent therapies in acute renal failure. J Am Soc Ne- phrol 19 94; 4: 141 3 – 142 0. 41 . Chima CS, Meyer L, Hummell. of vancomycin mass transfer between well-per- fused compartments of the body (i.e., extracellular space) and poorly perfused compartments (i.e., intra- cellular space), relative to the rate of. FAILURE The identification of ␤ 2 -microglobulin as a precursor molecule in the development of dialysis- related amy- loidosis established low-molecular weight proteins as a new class of uremic toxins (92).