Drug removal in continuous kidney replacement therapy

INTRODUCTION — Continuous kidney replacement therapy (CKRT) is occasionally required for critically ill patients with acute kidney injury (AKI) [1]. Some, but not all, drugs are removed by CKRT. For patients treated with CKRT, it is important to understand the factors that determine drug removal to permit optimal drug dosing [2]. This is particularly true for antibiotics since subtherapeutic levels are commonly observed in patients on CKRT [3-6] and associated with antibiotic failure [3,7].

This topic reviews drug removal and an approach to drug dosing in patients on CKRT. The general mechanisms underlying solute removal by kidney replacement therapies (KRTs) are discussed elsewhere. (See "Mechanisms of solute clearance and ultrafiltration in peritoneal dialysis".)

Other issues related to CKRT, including optimal prescription, and anticoagulation are discussed elsewhere. (See "Continuous kidney replacement therapy in acute kidney injury" and "Anticoagulation for continuous kidney replacement therapy".)

PHARMACOKINETICS OVERVIEW — The effectiveness and potential toxicity of drugs are related to their concentration at the site of action, which is practically assumed to be the blood concentration. Drug concentration is determined by absorption, volume of distribution, and clearance, all which may be altered among critically ill patients [8].

In patients on CKRT, drug clearance is the sum of metabolisms and excretion by the liver and gastrointestinal tract, removal by residual kidney function that may be present, and removal by KRT. Drug clearance is difficult to estimate and may change from day to day among critically ill patients, particularly as kidney function deteriorates and CKRT is initiated, and as kidney function begins to improve and CKRT is discontinued or the patient is transitioned to other KRT modalities [9-13].

Hepatic drug metabolism may also change in patients who develop acute kidney injury (AKI). End-stage kidney disease (ESKD) and AKI both reduce hepatic drug metabolism compared with healthy individuals [14]. However, patients with AKI have higher hepatic clearance than patients with ESKD [14]. This is potentially important since drug dosing guidelines have been derived from pharmacokinetic studies that were performed in patients with ESKD rather than those with AKI. Adherence to such guidelines may result in underdosing of drugs among AKI patients [14-16]. As a result, appropriate dosing in AKI may be higher than what is used in patients with ESKD to achieve therapeutic serum concentrations. This has been observed for imipenem [15], meropenem [14], and vancomycin [16].

Because of the potential changes in total clearance, drug doses require ongoing consideration, particularly with respect to hepatic and kidney function.

Not all drugs require dose changes to account for CKRT drug removal. Although the CKRT-mediated elimination of drugs may be the most visible drug removal mechanism, it may not be the dominant clearance mechanism. Drugs that are primarily cleared by hepatic or gastrointestinal metabolism and do not have significant pharmacologically active metabolites that are excreted by the kidney do not have to be adjusted for reduced kidney function or CKRT clearance.

FACTORS THAT AFFECT CLEARANCE BY CKRT — The degree to which drugs are removed by CKRT is determined by drug characteristics and by CKRT features.

Drug characteristics — Drug characteristics that affect clearance by all CKRT modalities include the degree of protein binding, volume of distribution, interactions of the drug with the dialyzer/hemofilter membrane, and, potentially, molecular weight.

In practice, most drugs are small enough to be removed equally well by all CKRT modalities. Larger (>5000 Daltons) drugs may be less well cleared by continuous venovenous hemodialysis (CVVHD) and continuous venovenous hemodiafiltration (CVVHDF) compared with continuous venovenous hemofiltration (CVVH). However, few drugs fall into this size category; insulin is one such drug at 5805 Daltons but is dosed to a target effect.

Protein binding — The degree of protein binding is a major determinant of the extent to which the drug is removed by CKRT.

Only unbound drug can be removed by CKRT. The degree of protein binding is determined by pH, molar concentration of both the drug and protein, and the presence of displacing drugs as well as by circulating bilirubin, and free fatty acids. The importance of these factors varies from drug to drug. The percent protein binding can be found for all drugs in Lexicomp under Pharmacodynamics/Kinetics or in the drug package insert.

However, protein binding is lower in patients with acute kidney injury (AKI) compared with healthy subjects [4,17,18]. Reasons for this are not entirely clear but may be related to kidney failure or hypoalbuminemia. As a result of the lower protein binding, dosing strategies that use the unbound fraction of the drug estimates derived from healthy subjects to estimate drug removal by CKRT may underestimate the degree of drug removal. This underestimation may be by a factor of two or more, although the degree to which this is clinically significant is not known [19,20].

Volume of distribution — The apparent volume of distribution is the theoretical volume of water the drug would occupy if the body were a single homogenous reservoir whose concentration is equal to the plasma concentration. Drugs that are lipid soluble or highly tissue bound have a large volume of distribution. Drugs with a large distribution volume are removed less efficiently by CKRT. Drugs that are limited to intravascular compartment have a small volume of distribution and are efficiently removed by CKRT.

Among critically ill patients, the volume of distribution is often increased because of severe volume overload [8]. The volume of distribution may also be increased by other factors including mechanical ventilation, hypoalbuminemia, and, in neonates, extracorporeal circuit volumes [5,21,22].

Molecular weight — The efficiency with which solutes, including drugs, are removed by CKRT is affected by molecular weight, and, for higher molecular weight solutes, the CKRT modality.

Drugs ≤2000 Daltons readily cross the membrane and are small enough to be removed equally by all CKRT modalities.

For drugs between 2000 and 15,000 Daltons, the membrane's pore size becomes a limiting factor in CKRT clearance, with increasing hindrance as molecular weight goes up. For such agents, the degree to which molecular weight affects clearance depends upon the type of CKRT. (See "Continuous kidney replacement therapy in acute kidney injury".)

The drug molecular weight affects clearance with CVVHD but less so than in CVVH, unless the drug is too big to fit through the pores of the CVVH membrane. With CVVHD, higher-molecular-weight drugs are removed more slowly compared with lower-molecular-weight drugs. This is because CVVHD relies on diffusion, and higher-molecular-weight molecules diffuse more slowly than smaller molecules. CVVH and, in part, CVVHDF rely on convection. With convection, most drugs are removed at the same rate, as long as the unbound drug molecule is small enough to cross the hemofilter/hemodialyzer membrane.

The influence of molecular weight on CVVHDF clearance falls in between CVVHD and CVVH depending on the ratio of dialysate inflow to total effluent flow. (See 'CKRT features' below.)

Drugs >15,000 Daltons are not appreciably removed by any CKRT modality, as they do not cross the membrane.

CKRT features — Important CKRT features that affect drug clearance include characteristics of dialyzer/hemofilter membrane and operating conditions (ie, flow rate settings).

Characteristics of membrane — Characteristics of the dialysis membrane/hemofilter that affect clearance include permeability (molecular weight cut-off) and factors such as charge that may influence binding to the drug [23].

Membrane permeability is an important determinant of drug removal. All membranes used for CKRT today are high flux. High-flux membranes have greater permeability for larger molecules compared with conventional membranes that may be used for intermittent hemodialysis. Whereas a drug like vancomycin at 1450 Daltons is poorly dialyzed by intermittent hemodialysis using conventional, less permeable membranes, it is appreciably removed when high-flux membranes are used [24].

Drug clearance may be increased by binding of the drug to the membrane, although this is rarely of sufficient magnitude to affect dosing of the drug [25-27]. Colistin is the only drug that has demonstrated sufficient binding to CKRT membranes to necessitate dosing adjustments. In one study, 40 to 60 percent of total colistin clearance occurred via CKRT [28]. Only a small fraction of the cleared drug was recovered in the effluent, suggesting that adsorption to the membrane contributed to extracorporeal clearance.

The mechanisms underlying the drug-membrane interaction are poorly characterized, although drug and membrane charge are commonly proposed to contribute. Charged membranes (which would be more likely to interact with charged solutes), such as the negatively charged AN69 membrane filter, are less commonly used for CKRT [27]. Clinically significant CKRT drug clearance changes due to membrane charge have not been demonstrated with AN69 versus uncharged membranes. The more commonly used polysulfone membranes are not charged.

Modality and operating conditions (flow rate settings) — The specific CKRT modality and flow rate setting may affect clearance.

As noted above, convective therapies (CVVH) may clear large solutes better than diffusive therapies (CVVHD and CVVHDF). However, for most drugs that are used in the intensive care unit (ICU), the differences in clearance provided by different modalities are negligible [29]. (See 'Molecular weight' above.)

Depending on the specific CKRT modality, drug clearance increases with the ultrafiltration, dialysate, or effluent flow rate.

Variables that affect clearance and estimates of clearance that are specific to each modality are discussed here.

Continuous venovenous hemofiltration (CVVH) — In CVVH, plasma is filtered through a dialyzer but not exposed to dialysate flowing on the other side of the membrane. As a result, solute removal occurs entirely by convection (or solvent drag), with no solute loss by diffusion. In convection, the frictional forces between water and solutes result in the movement of small- and middle-molecular-weight solutes (<5000 Daltons) in the same direction as plasma water.

In addition to factors defined above (ie, volume of distribution, protein binding, and drug-membrane interactions), drug removal by CVVH is determined by the ultrafiltration rate and the method of administering replacement fluid.

●Molecular weight – As noted above, for most drugs, the molecular weight is not an important determinant of drug removal by CVVH, since most drugs are smaller than the maximum size that is freely removed by convection. (See 'Molecular weight' above.)

●Method of fluid replacement – Effluent volumes of 20 to 25 mL/kg/hour are commonly used regardless of the type of CKRT [30]. When CVVH is used, a similarly large volume of replacement fluid is required to prevent fluid depletion.

Replacement fluid may be administered before (prefilter replacement (figure 1)) or after the hemofilter (postfilter replacement (figure 2)). (See "Alternative kidney replacement therapies in end-stage kidney disease", section on 'Substitution fluid infusion methods'.)

The method of fluid replacement (ie, pre- or postfilter) affects both the efficiency of solute clearance and estimates of solute clearance. CVVH drug clearance with prefilter fluid replacement is slightly lower than postfilter systems for any given ultrafiltration rate.

Prefilter fluid replacement reduces the CVVH drug clearance by decreasing the concentration of the drug just prior to convection across the hemofilter membrane; some of the formed ultrafiltrate is simply prefilter fluid replacement solution that has no drug in it. The degree of reduction in clearance is a function of the ratio between the rate of prefilter replacement fluid administration and blood flow rate.

●Ultrafiltration rate – The ultrafiltration rate is set by the clinician (see "Prescription of continuous kidney replacement therapy in acute kidney injury in adults", section on 'CKRT dose'). The higher the ultrafiltration rate, the greater the drug clearance. The degree to which the ultrafiltration increases drug clearance is demonstrated in the section below. Predilution fluid substitution dilutes blood proteins and cell concentrations and allows for greater ultrafiltration rates than those seen in postdilution fluid substitution.

●Calculation of clearance ─ The clearance of specific drugs by CVVH may be calculated from the ultrafiltration rate and the solute-specific sieving coefficient (SC), providing the SC is known or the drug concentration can be measured in plasma and ultrafiltrate [31].

The SC is the mathematical expression of the ability of a specific solute to convectively cross a specific membrane. The SC is determined from the ratio of the solute concentration in the ultrafiltrate to the solute concentration in the plasma. The plasma solute concentration is estimated from the average of the arterial (A) and venous (V) drug concentrations.

Thus:

UF is the solute concentration in the ultrafiltrate [32]. An SC of 1 means that the solute freely crosses the membrane and is removed in the same concentration as in the plasma water. An SC of 0 means that there is no solute in the ultrafiltrate and thus no drug removal by CKRT. This is generally due either to large molecular size or to extensive protein binding of the drug.

The arterial concentration is the plasma drug concentration before the dialyzer/hemofilter. The venous concentration is the plasma drug concentration after the dialyzer/hemofilter. The venous concentration is equal to arterial if the SC is 1 and slightly greater than arterial if the SC is less than 1 because more water than solute is removed. This difference is not clinically significant for most drugs, and it can be assumed that the arterial and venous concentrations are the same. There is therefore no need for a venous sample, and the above equation can be simplified to [33]:

Although the physiologically important concentration is plasma water rather than measured plasma, the difference between the measured plasma and plasma water concentrations can be ignored since plasma water makes up approximately 93 percent of the plasma (which is deemed sufficiently close to unity).

The SC is approximately constant during CVVH. As a result:

The SC for several drugs varies depending upon the membrane (ie, polyacrylonitrile or polyamide membranes versus polysulfone membranes) [34]. (See 'Characteristics of membrane' above.)

SC data from our experience and the literature are displayed in the table (table 1) [35]. It is not known with certainty, however, whether SC data derived from healthy individuals apply to septic, acutely ill individuals. Also shown for comparison is the unbound fraction, which, for most drugs, correlates closely with the SC.

The method of fluid replacement (ie, pre- or postfilter) affects estimates of solute clearance. As noted above, CVVH drug clearance with prefilter fluid replacement is slightly less efficient than postfilter systems for any given ultrafiltration rate, but the prescribed ultrafiltration rate often is greater in predilution mode than in the postdilution mode.

The calculation given in the previous section for the rate of solute clearance assumes that replacement fluid is administered after the hemofilter. Calculation of CVVH drug clearance must correct for the rate of prefilter replacement relative to blood flow.

For drugs that partition into red blood cells (RBCs), CVVH drug clearance using prefilter fluid replacement = UF × SC x [Qb / (Qb + Qrep)]

If the drug does not partition into RBCs and is restricted to the extracellular compartment, blood flow needs to be corrected for hematocrit:

where Qb is the blood flow rate and Qrep is the rate of prefilter fluid replacement. However, most of the time, it is not known whether or not the drug partitions into RBCs, so we tend to use the simpler equation.

Continuous venovenous hemodialysis (CVVHD) — In CVVHD, plasma is exposed to drug-free dialysate, which runs countercurrent to plasma flow through the dialyzer. Drug removal during CVVHD occurs primarily by diffusion into the drug-free dialysate, as per the second law of thermodynamics. The removal of a solute by dialysis is dependent on the degree to which the solute diffuses across a particular membrane. Convection is less important (though can occur) since the ultrafiltration rate is held at a much lower level than with hemofiltration alone.

In addition to the volume of distribution, protein binding, and drug-membrane interactions, drug removal by CVVHD is influenced by drug molecular weight and the dialysate flow rate [36].

Diffusion rates for any solute are indirectly related to solute molecular mass. Thus, the rate of solute transfer by diffusion decreases as the molecular size increases. This is by contrast to solute transfer by convection, which, for most solutes, is not affected by solute molecular size. As a result, CVVH may theoretically be more effective than CVVHD in removing cytokines and solutes with relatively higher molecular weights.

In practice, however, CVVH does not appear to provide significantly better middle-molecule clearance compared with CVVHD [37]. It is possible that the passage of higher-molecular-weight solutes across membranes is lower than expected in CVVH because of the accumulation of plasma protein at the membrane that creates an additional barrier for solute to cross (a process referred to as protein concentration polarization) [38,39].

●Calculation of clearance – The clearance by CVVHD is the product of the ratio of the dialysate-specific solute concentration to the plasma-specific solute concentration (D/P) times the dialysate effluent flow rate.

This assumes no clearance from the plasma by adsorption to the dialysis membrane. The D/P ratio is sometimes called the saturation ratio or saturation coefficient because it is a measure of the degree of saturation of the dialysate with the solute in question. A D/P of unity means the dialysate is fully saturated with the solute, and the clearance is equal to the effluent flow rate.

When dialysate flow rate is slow (as it is with CVVHD compared with intermittent hemodialysis), there is usually enough time for drugs to cross the membrane and fully saturate the dialysate. At these slower flow rates, D/P will be similar to the SC seen in CVVH. The faster that dialysate flows through a circuit, the less time there is for the drug to come into equilibrium; consequently, D/P is dependent on the ratio between dialysate and blood flow, and a tripling of dialysate rate will not triple drug clearance rate. The larger the drug, the more profound this effect becomes [40]. (See "Short daily home hemodialysis: The low dialysate volume approach".)

Continuous venovenous hemodiafiltration (CVVHDF) — Continuous hemodiafiltration utilizes both convective (CVVH) and diffusive (CVVHD) solute removal mechanisms. A dialysate is run through the system, but a pump is set such that much more fluid comes out of the dialysate efflux port than entered the dialysate influx port. The fluid coming out of the efflux port is a combination of spent dialysate and formed ultrafiltrate (ie, effluent).

●Calculation of clearance – Because CVVHDF effluent contains both dialysate and ultrafiltrate, drug removal can be calculated using CVVHD equations ensuring to account for all effluent.

If prefilter fluid replacement is used in CVVHDF, the dilutional effect must be accounted for.

Solute losses during combined convection and diffusion are less than the sum of each transport process individually since the presence of convectively derived solute in the dialysate decreases the concentration gradient that drives diffusion [41-43]. This effect can be overcome by increasing the dialysate flow rate and is less important at lower ultrafiltrate rates [44].

OUR APPROACH TO DETERMINING THE OPTIMAL DRUG DOSE — Lexicomp has updated CKRT dosing guidelines for most drugs. This section describes the principles that should be used when determining any drug dose for CKRT at any set of flow characteristics.

Estimate of loading dose — It is important to give an adequate loading dose, particularly with water-soluble drugs that have small volumes of distribution (<0.7 L/kg). Such drugs are likely to be affected by fluid overload. Underdosing may result if volume of distribution is underestimated. (See "Evaluation and management of suspected sepsis and septic shock in adults", section on 'Dosing'.)

We calculate the loading doses based on the ideal body weight plus added weight from volume overload (which may be estimated from the change in weight from the baseline weight determined on admission). Doses estimated from the actual weight (rather than ideal body weight) are more likely to be large enough to "fill" the volume with sufficient drug to achieve therapeutic drug levels. The volume of distribution (in L/kg) may be found in Lexicomp, in the section on Pharmacodynamics/Kinetics, and in many other electronic references, package inserts, and published books.

Estimate of maintenance dose — Many drugs that are not titrated to clinical response do not have to be dose adjusted for reduced kidney function or CKRT. Drugs that are primarily cleared by hepatic or gastrointestinal metabolism and do not have significant pharmacologically active metabolites that are excreted by the kidney do not have to be adjusted for reduced kidney function or CKRT clearance.

Our approach to maintenance drug dosing for drugs that require dose adjustment for CKRT depends on the specific drug (algorithm 1).

For many medications (pain medications, sedatives, vasopressors), we titrate the maintenance dose based on the clinical response. The dose of such drugs is usually not altered by initiation of CKRT.

Ideally, for medications for which there is not an observable clinical response to follow and which require dose adjustment for reduced kidney function, we titrate the dose based on plasma concentrations. Therapeutic drug monitoring is the best method to provide accurate dosing among critically ill patients who have alterations in changes in the volume of distribution, protein binding, and total clearance.

For many drugs (eg, aminoglycosides, vancomycin, calcineurin inhibitors, and antiseizure medications), the clinical laboratory can provide drug plasma concentration within a few hours of phlebotomy. Plasma drug concentrations can be used in standard, first-order equations in patients receiving CKRT because the CKRT runs continually and at a constant rate, providing there are no unanticipated interruptions of treatment. Interruptions in CKRT decrease clearance. The appropriate timing to obtain the plasma measurement may vary according to individual pharmacy and hospital guidelines but is generally 30 to 60 minutes after the drug infusion has ended; this allows for optimal distribution of the drug.

However, monitoring plasma concentrations may not be possible. Many drugs can be measured only in specialized laboratories, requiring a "send-out" test that may take days or weeks to be reported back.

For medications that cannot be titrated to clinical response nor easily measured, we estimate the dose based on CKRT clearance and residual kidney function. CKRT clearance must be modified based on the degree of protein binding for the specific drug. The degree of protein binding is described in Lexicomp under Pharmacodynamics/Kinetics but is also described in multiple other references including drug package inserts.

For patients with significant urine output (ie, >20 mL per hour), we add clearance provided by kidney function. Thus we obtain a 24-hour urine collection for determination of creatinine and urea clearance in all CKRT patients who have a urine output ≥20 mL/hour. Significant residual kidney function is unlikely in patients with urine output <20 mL/hour. We use the mean of creatinine and urea clearances to determine residual kidney function since use of creatinine alone significantly overestimates true glomerular filtration rate (GFR) among patients with markedly reduced kidney function [45]. Cystatin C-derived GFR estimates may be a preferable option to creatinine or urea-based estimates [46]. If the patient has been on CKRT for a few days, the serum creatinine and urea are at steady state (ie, stable over several days) and can be used in the calculation of clearance. If CKRT is recently initiated and the serum creatinine and blood urea nitrogen (BUN) are still decreasing, the means of the serum creatinine and BUN during the timing interval may be used to calculate clearance, or a single midpoint value may be used.

We calculate clearance as follows:

●Drugs that are not protein bound – For drugs that have no protein binding, drug clearance is equal to the CKRT effluent rate since CKRT waste fluid is generally well equilibrated with small molecules with low protein binding [19,29,47].

For patients with urine output ≥20 mL/hour, we add residual kidney function:

GFR (mL/day) = (urine creatinine [mg/dL] / plasma creatinine [mg/dL] + urine urea nitrogen [mg/dL] / plasma urea [mg/dL]) x urine volume (mL/day) / 2

The GFR is converted to mL/min by multiplying by 1000 (to convert to mL) and then dividing by 1440 (to convert to mL/min) (or by multiplying by 1000/1440, which is equal to 0.694). This number is added to the CKRT clearance for total clearance in mL/min.

The GFR is divided by 1440 (to convert to mL/min). This number is added to the CKRT clearance for total clearance in mL/min.

As noted above, once the CKRT clearance (plus residual kidney function, if need be) is estimated in mL/min, we administer the maintenance dose that is recommended for patients with an equivalent level of kidney function. This dose can be found in Lexicomp under the section on "Dosing: Renal impairment" and in several other references including package inserts.

●Drugs that are protein bound – Even a small percentage of protein binding influences clearance. The clearance estimate is adjusted for protein binding. To account for protein binding, the CKRT effluent rate is multiplied by the unbound fraction:

For patients with residual kidney function, residual kidney function is added as described above. Once the CKRT clearance (plus residual kidney function, if need be) is estimated in mL/min, we administer the maintenance dose that is recommended for patients with an equivalent level of kidney function (ie, creatinine clearance). This dose can be found in Lexicomp under the section on "Dosing: Renal impairment" and in several other references including package inserts.

This approach can only be used for drugs ≤2000 Daltons. For larger drugs, the membrane's pore size becomes a limiting factor in CKRT clearance, with increasing hindrance as molecular weight goes up. A precise estimate of clearance is difficult for drugs between 2000 and 15,000 Daltons. However, very few drugs fall into this size range. As noted above, insulin is one such drug at 5808 Daltons, but insulin is dosed to target effects. (See 'Pharmacokinetics overview' above.)

There is very little, if any, CKRT clearance of drugs larger than 15,000 Daltons [48]. Most drugs this size are biological agents (such as monoclonal antibodies and soluble receptor antagonists) that are titrated to a specific clinical response or given as a single dose.

Lexicomp CKRT dosing guidelines have been updated and incorporate modern CKRT technology, the most commonly used CKRT flow rates, and apply the latest pharmacodynamic and pharmacodynamic principles. We use these dosing recommendations rather than older published references for dosing. Several studies of beta lactam antibiotics have shown that up to a quarter of patients receiving doses guided by older standards fail to attain therapeutic targets [21].

SUMMARY AND RECOMMENDATIONS

●General principles – The effectiveness and toxicity of drugs are related to their concentration. Factors that affect drug concentration are altered among critically ill patients, particularly patients with acute kidney injury (AKI) on continuous kidney replacement therapy (CKRT). (See 'Pharmacokinetics overview' above.)

●Factors that affect drug clearance by CKRT

•Drug characteristics that affect CKRT clearance include the volume of distribution and degree of protein binding. Size is theoretically important in that larger drugs may be hindered by the dialysis/hemofilter membrane and be less well cleared by diffusive therapies compared with convective therapies. In practice, most drugs are small enough to pass through the membrane and to be cleared equally well by all modalities. (See 'Drug characteristics' above.)

•CKRT features that affect drug clearance include characteristics of dialyzer/hemofilter membrane (largely permeability) and operating conditions (ie, flow rate settings). (See 'CKRT features' above.)

●Approach to determining optimal drug dose

•Estimation of loading dose – It is important to give an adequate loading dose, particularly with water-soluble drugs that have small volumes of distribution (<0.7 L/kg). We calculate the loading doses based on the ideal body weight plus added weight from volume overload (which may be estimated from the change in weight from baseline on admission). (See 'Estimate of loading dose' above.)

•Estimation of maintenance dose – Our approach to maintenance drug dosing is as follows:

-For many medications (pain medications, sedatives, vasopressors), we titrate the dose based on the clinical response.

-For medications for which there is not an observable clinical response to follow, we titrate the dose based on plasma concentrations.

-For medications that cannot be titrated to clinical response nor easily measured, we estimate the dose based on CKRT clearance and residual kidney function. The estimated CKRT clearance must be modified based on the degree of protein binding (algorithm 1). (See 'Our approach to determining the optimal drug dose' above.)