Clinical review: Optimal dose of continuous renal replacement therapy in acute kidney injury

Continuous renal replacement therapy (CRRT) is the preferred treatment for acute kidney injury in intensive care units (ICUs) throughout much of the world. Despite the widespread use of CRRT, controversy and center-specific practice variation in the clinical application of CRRT continue. In particular, whereas two single-center studies have suggested survival benefit from delivery of higher-intensity CRRT to patients with acute kidney injury in the ICU, other studies have been inconsistent in their results. Now, however, two large multi-center randomized controlled trials - the Veterans Affairs/National Institutes of Health Acute Renal Failure Trial Network (ATN) study and the Randomized Evaluation of Normal versus Augmented Level (RENAL) Replacement Therapy Study - have provided level 1 evidence that effluent flow rates above 25 mL/kg per hour do not improve outcomes in patients in the ICU. In this review, we discuss the concept of dose of CRRT, its relationship with clinical outcomes, and what target optimal dose of CRRT should be pursued in light of the high-quality evidence now available.


Introduction
Extracorporeal renal replacement therapy (RRT) has been used for the supportive treatment of acute kidney injury (AKI) for over 60 years. Whereas treatment method ologies have become increasingly sophisticated, the basic principles guiding the use of RRT have changed little. Th ey aim at averting the immediately life-threatening consequences of severe renal dysfunction (acidosis, electrolyte imbalances, uremia, and fl uid overload), preserving life, and allowing time for organ recovery to occur. Following these principles, one could argue that, as long as a minimum level of RRT is provided, the quality, quantity, and timing of RRT should not greatly aff ect outcome. However, considerable evidence that the develop ment of AKI is independently associated with mortality now exists [1], suggesting that optimization of RRT could benefi t survival. Until now, however, little evidence or consensus regarding the optimum delivery of RRT has existed, resulting in wide variations in clinical practice [2]. Furthermore, irrespective of intensity, the superiority of continuous renal replacement therapy (CRRT) over intermittent hemodialysis (IHD) for RRT in the intensive care unit (ICU) remains controversial, although there is increasing international consensus that initial therapy with CRRT is preferable in hemodynamically unstable patients [3][4][5][6]. Given that most patients critically ill with AKI do have hemodynamic instability, CRRT has become, worldwide, the dominant form of RRT for AKI in the ICU [2]. Furthermore, since a large majority of RRT treatments for AKI are now given in the ICU [7], CRRT is likely to be the most common modality for the initial treatment of AKI by RRT in the developed world.
As a continuous therapy, CRRT has the benefi t of better mimicking normal renal function, while quantity of therapy delivered is more easily assessed and compared with endogenous renal function. Consequently, there has been continued interest in establishing the appropriate dose or intensity of this therapy. Two major multi-center randomized controlled trials (MC-RCTs) examining this question have now been performed: the Veterans Aff airs/ National Institutes of Health Acute Renal Failure Trial Network (ATN) study [4] and the Randomized Evaluation of Normal versus Augmented Level (RENAL) Replacement Th erapy Study [8]. Together, these trials provide, for the fi rst time, high-quality evidence to guide the prescription of CRRT in the ICU. Th is review examines the results from these and earlier studies to summarize our current understanding of the most appropriate dose of CRRT to deliver during AKI in critical illness.

Assessing dose of continuous renal replacement therapy
To compare diff ering intensities of RRT, some measure of quantity of blood cleansing is needed. Quantity or dose of Abstract Continuous renal replacement therapy (CRRT) is the preferred treatment for acute kidney injury in intensive care units (ICUs) throughout much of the world. Despite the widespread use of CRRT, controversy and centerspecifi c practice variation in the clinical application of CRRT continue. In particular, whereas two single-center studies have suggested survival benefi t from delivery of higher-intensity CRRT to patients with acute kidney injury in the ICU, other studies have been inconsistent in their results. Now, however, two large multi-center randomized controlled trials -the Veterans Aff airs/ National Institutes of Health Acute Renal Failure Trial Network (ATN) study and the Randomized Evaluation of Normal versus Augmented Level (RENAL) Replacement Therapy Study -have provided level 1 evidence that effl uent fl ow rates above 25 mL/kg per hour do not improve outcomes in patients in the ICU. In this review, we discuss the concept of dose of CRRT, its relationship with clinical outcomes, and what target optimal dose of CRRT should be pursued in light of the high-quality evidence now available.
RRT delivered is traditionally assessed by measuring clearance of urea (that is, the rate of plasma cleansing required to account for the rate of urea removal if the treatment separated all urea from the plasma processed). Such measurements are comparable to endogenous renal function assessed as glomerular fi ltration rate (GFR). Urea is chosen as an easily measured prototype for lowmolecular-weight products of metabolism. While tubular reabsorption prevents urea clearance from being an accurate measure of kidney GFR, such concerns do not apply to extracorporeal blood purifi cation. Measuring total urea in the dialysis or ultrafi ltration effl uent and con tinuous plasma urea concentration could allow calculation of clearance. However, this is cumbersome and approxi mations of dose are instead estimated from machine settings. Th is process is very easy for continuous therapies. For a small uncharged molecule such as urea, clearance during continuous hemofi ltration (HF) is essentially equal to the ultrafi ltration rate because the fi lter membrane negligibly impedes the passage of urea. Similarly, during continuous hemodialysis (HD), when the dialysis fl ow rate is much slower than the blood fl ow rate, urea concentration in the dialysate will equilibrate with that in the plasma and clearance can be approximated by the dialysate fl ow rate. Th us, provided that no pre-dilution is applied, in any HF, HD, or hemodiafi ltration continuous therapy, the total clearance is approxi mately the total fl ow of fl uid out of the dialyzer or fi lter (that is, the effl uent fl ow rate).
Th e above approximations have been shown to correlate well with a more formal set of measurements of urea clearance [9]. Th us, the dose of CRRT is reported as effl uent fl ow in milliliter per hour or milliliter per kilogram of body weight per hour. Infusion of HF replace ment solution as pre-dilution will reduce eff ective effl uent dose by the degree to which the plasma is diluted. Since this dilution eff ect is dependent on circuit blood fl ow, replacement fl uid rate, and hematocrit, these factors have to be measured to correctly assess dose when predilution is applied [9]. Most importantly, time spent off CRRT (circuit clotting, radiological investigations, and/ or surgical procedures) can be substantial and can impact azotemic control [10]. Th e impact of these factors must be accounted for when prescribing therapy as it is the delivered and not the prescribed dose that will aff ect outcome. Finally, unlike urea, metabolites of higher molecular weight will interact signifi cantly with the dialyzer membrane. Similarly, other substances that significantly bind to plasma proteins may be relatively retained in the plasma. In these cases, changes in effl uent fl ow rate may correlate poorly with changes in solute clearance. Th us, it should be remembered that effl uent measures of dose, while reproducible and convenient, do not measure all aspects of blood cleansing in CRRT.

Clinical variation in continuous renal replacement therapy dose
Prescribed and delivered inten sities of CRRT vary widely. Two multi-center observa tional studies have documented CRRT dose in the ICU; in both studies, interquartile ranges for dose were large (15.3 to 27.7 mL/kg per hour in the BEST Kidney [Beginning and Ending Supportive Th erapy for the Kidney] study [11] and 22.1 to 33.9 mL/ kg per hour in the DO-RE-MI [Dose Response Multicentre International collaborative initiative] study [12]). As units participating in these studies were recruited by expression of interest, one can assume far wider variation and lower delivered doses in units without an interest in research and the provision of CRRT. Much of this variation in dose may be accounted for by the use of standard doses of CRRT unadjusted for weight. For example, in the DO-RE-MI study, patients receiving the lowest CRRT doses were 33% heavier than those receiving the higher doses [13]. Th is greatly confounds the interpretation of observational data on CRRT dose since weight is likely to have a complex causal relationship with outcome. Furthermore, even when used, weight estimation in the ICU can be highly errorprone and it is far from clear whether estimates of ideal body weight, pre-morbid weight, 'dry' weight, or actual body weight should be used [13].

Dose-response relationship in continuous renal replacement therapy
Prior to the ATN and RENAL trials, four single-center studies had prospectively examined the relationship between CRRT dose and survival in critical illness ( Table 1). All of these trials were, however, relatively small and varied in geographical location, patient case mix, and mode of CRRT. In the fi rst of these, Ronco and colleagues [14] randomly assigned 425 critically ill patients with AKI treated using continuous veno-venous hemofi ltration (CVVHF) at a single center to ultra fi ltration rates of 20, 35, or 45 mL/kg per hour. Survival 15 days after dis continuation of CRRT was signifi cantly better in the highestand intermediate-dose arms in comparison with the lowest-dose arm (58% and 57% versus 41%, respec tively; P <0.001). Th ree smaller randomized control trials subsequently examined RRT dose-response relationships in the ICU. Saudan and colleagues [15] randomly assigned 206 patients with AKI in a single-center trial comparing CVVHF (mean ultrafi ltration rate of 25 ± 5 mL/kg per hour) with continuous veno-venous hemodiafi ltration (CVVHDF) with the addition of a mean dialysate fl ow rate of 15 ± 5 mL/kg per hour to a similar mean ultrafi ltration rate (24 ± 6 mL/kg per hour): survival rates after 90 days were 34% in the CVVHF group and 59% in the CVVHDF group (P = 0.0005), analyzed on an intentionto-treat basis. While higher-intensity therapy involved a diff erent modality, evidence suggests that solute clearances in continuous high-fl ux dialysis and CVVHF are quite similar [16], so this study is best regarded as investigating the eff ect of a dose increment.
On the other hand, two studies [17,18] that included 200 and 106 patients, respectively, failed to demonstrate any benefi cial eff ects of increased CRRT intensity on patient survival or renal recovery. Similarly, in the prospective observational DO-RE-MI study [12], which included 338 patients undergoing CRRT for AKI in the ICU, higher-dose CRRT (defi ned as greater than 35 mL/kg per hour) was not associated with increased survival even after statistical adjustment for baseline characteristics (adjusted odds ratio for death 1.21, 95% confi dence interval 0.66 to 2.21; P = 0.537). Disparity in these results, concerns over the inherent biases of single-center studies [19], and consequent uncertainty about the optimum dosing of CRRT in the ICU led to the initiation of the ATN and RENAL studies.
Th e ATN and RENAL studies were large, multi-center randomized controlled trials investigating the eff ects of RRT dose on survival and were conducted in ICUs through out the US and in Australia and New Zealand, respectively ( Table 2). Th e ATN study [4] randomly assigned critically ill adults with AKI that required RRT to high-intensity or low-intensity RRT. High-intensity therapy consisted of pre-dilution CVVHDF to provide a total effl uent fl ow rate of 35 mL/kg per hour or six sessions of IHD per week [4]. Low-intensity therapy consisted of CVVHDF to provide a total effl uent fl ow rate of 20 mL/kg per hour or thrice-weekly IHD. A very small number of patients received slow extended-duration dialysis six or three times weekly in the highintensity and low-intensity groups, respectively, in centers where CRRT was not available [4]. Within treatment groups, patients were allocated to CRRT or IHD when their cardiovascular Sequential Organ Failure Assessment (SOFA) score was 3 or 4 and received IHD if their cardiovascular SOFA score was not more than 2. However, patients receiving only CRRT switched to IHD if their cardiovascular SOFA score was 0 or 1 for more than 24 hours. CRRT was provided to 69.7% of patients as their initial therapy [20]. Switching between modalities occurred at similar rates in high-intensity and lowintensity groups [20]. Among patients who survived to day 60, 84.2% received IHD at some stage during their ICU stay [20]. Given the controversies in dose comparison between treatment modalities and the use of IHD in many patients, the ATN trial might best be described as a test of maximization of intensity of RRT within current US practice rather than a direct test of a dose-response relationship for CRRT. In comparison, 1,508 critically ill adults meeting predetermined criteria for the initiation of RRT in the RENAL study [8] were randomly assigned to post-dilution CVVHDF with an effl uent fl ow of 40 mL/kg per hour or 25 mL/kg per hour. All patients received CRRT as their fi rst mode of RRT; only a small proportion of patients (approximately 7%) received IHD later in their ICU stay, and this proportion was similar to that of patients who remained dependent on dialysis at day 90. A small number of patients requiring an extended period of post-ICU RRT accounted for the higher mean duration of RRT (11.5 ± 18.0 days and 13.0 ± 20.8 days in the two treatment groups) despite a duration of study treatment of only 5.9 ± 7.7 and 6.3 ± 8.37 days. Th e RENAL study thus constitutes a more direct measure of the relationship between intensity of CRRT and survival.
Both studies failed to detect any reduction in mortality associated with a more intensive RRT. In addition, no signifi cant diff erences in mortality rates were observed between high-intensity and low-intensity treatment in pre-specifi ed subgroups in either study. Th ese subgroups included patients with sepsis and those requiring vasopressors. Th ese results now provide defi nitive evidence that an increase of CRRT dose beyond conventional effl uent fl ow rates of 25 mL/kg per hour is not benefi cial for unselected ICU patients with AKI. Importantly, both the ATN study and the RENAL study demonstrated that the prescribed dose is typically 10% to 15% less than the delivered dose in these patients, presumably because of treatment downtime. Th is observation is important because downtime may be longer outside of the environment of a clinical trial and clinicians seeking to replicate these outcomes will need to account for this in their prescription. In contrast to the within-trial comparisons, important diff erences between overall survival and recovery of renal function in the ATN and RENAL studies can be highlighted (Table 2). Th ese diff erences do not detract from the consistent primary results within the trials and may be accounted for by diff erences in population and case mix. However, they also suggest that other aspects of RRT provision, such as timing of therapy [21] and prevalence of IHD, may impact outcomes and should be the subject of further study [3].
After the publication of the above pivotal trials, it was speculated that the relationship between CRRT dose and survival is at a plateau between 20 and 40 mL/kg per hour but that lower or much higher doses could be associated with worse outcomes (Figure 1) [22]. Certainly, there is now little incentive to use doses of greater than 25 mL/kg per hour in unselected ICU patients. It may be speculated that individual patients might benefi t from diff ering doses of CRRT; however, there was no suggestion that specifi c groups benefi ted in subgroup analysis of the ATN and RENAL trials, indicating that identifi cation of such patients may be diffi cult. Similarly, diff erent target doses of CRRT may be suffi cient at various stages of critical illness, but this question has not been examined in prospective random ized controlled trials.
Th e absence of improved survival with dose increment above 20 mL/kg per hour should not be taken as an indica tion that assessment of dose is irrelevant, and lower intensities still will suffi ce. It can be reasonably presumed that some relationship for increasing survival with intensity of therapy exists as one increases intensity from no therapy to a greater level of therapy, with a plateau response occurring at or before 20 to 25 mL/kg per hour. It seems unlikely that this plateau is occurring much before a dose of 20 mL/kg per hour. In the DO-RE-MI study [12], patients in the second tertile of dose (23.6 to 30.9 mL/kg per hour) had the lowest mortality, with an adjusted odds ratio for death of 0.67, compared with doses of less than 23.6 mL/kg per hour, but this trend was far from statistically signifi cant (P = 0.196).
Historically, introduction of CRRT into the ICU led to improved survival when compared with previous use of thrice-weekly conventional IHD or acute peritoneal dialysis (PD) [23]. In this setting, use of CRRT led to improved azotemic control in comparison with the historical cohort, suggesting that increased survival may have been partly associated with delivery of a higher standardized dose of therapy. Similarly, a randomized control trial of CVVHF versus acute PD in patients with falciparum malaria or sepsis [24] demonstrated that CVVHF (mean prescribed dose of 19 mL/kg per hour) was associated with improved survival over continuous PD (2-L exchanges with 30-minute dwell time) (mortality of 15% versus 47%; P = 0.005). While modality-related factors undoubtedly played a role in these results, the rate of decline in creatinine was twice as fast in those treated with CVVHF in comparison with those treated with PD, suggesting that the dose of small-molecular clear ance delivered may have, in part, determined outcome.
By analogy, in the setting of end-stage renal disease, while doses of thrice-weekly IHD above an equilibrated Kt/V (eKt/V) of greater than 1.16 have not been asso ciated with improved outcomes [25], observational data have shown that an eKt/V of less than 1.05 is linearly associated with decreased survival in patients with endstage renal disease [26,27]. Th rice-weekly IHD with an eKt/V of 1.05 provides azotemic control similar to that of approximately 11 mL/kg per hour of CRRT, according to the peak concentration method of dose comparison [28], and somewhat more if time-averaged urea concentration [29] is used as a comparator. It seems unlikely that the critically ill will have a lower requirement for RRT in comparison with stable HD patients. Th ese considerations, therefore, suggest that we should not tolerate doses of CRRT in critically ill patients near the infl exion point of the dose-response relationship seen in the chronic HD population. Collectively, all of the observations provide strong circumstantial evidence that CRRT doses of less than 20 mL/kg per hour are likely to be harmful and should be avoided. Th us, to guarantee outcomes similar to those seen in the ATN and RENAL studies and to have confi dence that an eff ective dose of CRRT has been delivered, clinicians should prescribe 20 to 30 mL/kg per hour on the basis of body weight at time of commencement of CRRT and ensure that excessive periods off of treatment are not compromising dose delivered. Notably, international surveys [11,12] have reported that, even now, a signifi cant minority of patients receive treatment intensities of less than 20 mL/kg per hour and may be suboptimally treated. If RRT doses are increased by adoption of these targets, clinicians should be aware that adjustments in drug (most importantly, antibiotic) dosing may be required to allow for increased clearances.

Special circumstances
A number of authors have hypothesized that subgroups of the critically ill, particularly those with sepsis or multiorgan failure, might benefi t from a more intensive CRRT. It is thought that removal of circulating pro-and antiinfl ammatory mediators might blunt systemic derange ment of immune responses in severe critical illness [30,31]. In the ATN and RENAL trials, there was no benefi t from higher dose in any subgroup, including sepsis and vasopressor requirement. However, animal models [32][33][34], which have suggested benefi cial eff ects of CRRT on the severity of systemic infl ammatory responses, employed very-high-dose therapy (greater than 50 mL/kg per hour) initiated very early in the course of illness (prior to overt renal dysfunction). Dose-response studies of conventional CRRT, such as the ATN and RENAL trials, were not intended to examine these pleiotropic eff ects of CRRT in critical illness. A detailed discussion of this topic is beyond the scope of this review; however, as such immunemodulating eff ects are often cited as a rationale for dose adjustment of CRRT in severe sepsis, it is pertinent to briefl y discuss the current evidence base for such use of high-intensity CRRT in carefully selected patients.
Very-high-volume CRRT has been examined in uncontrolled clinical trials and case series with sugges tion of benefi t [35][36][37][38][39]. As removal of middle-molecularweight mediators was desired, purely convective therapy (HF) was used in most of these studies. Very-highvolume HF has been defi ned as a dose of greater than 50 mL/kg per hour [40], although many studies have used higher doses. In practice, delivery of doses of CVVHF in excess of 50 mL/kg per hour is complex, presenting a number of clinical and technical challenges [40] that may account for the lack of high-quality clinical trial data to confi rm or refute a benefi t. An MC-RCT, the IVOIRE (High Volume in Intensive Care) study, examining the use of high-volume versus intermediate-dose CRRT (70 versus 35 mL/kg per hour for 96 hours in patients with septic shock and moderate kidney injury) has recently stopped recruiting as interim analysis revealed lowerthan-expected mortality in all patients, rendering the planned study size (480 patients) insuffi ciently powerful to prove an eff ect on survival [41]. Interest has thus focused on other methods of blood purifi cationincluding coupled plasma fi ltration with bioadsorption [42] or hemofi ltation using high-cutoff point (greater than 50 kDa) membranes [43,44] -that may be easier to clinically implement. As yet, the adjunctive uses of CRRT and related technologies for the treatment of severe sepsis and multi-organ failure remain experimental and are subject to ongoing trials. However, given the complexity of these treatments and the diverse patient population, direct proof of survival benefi t may be diffi cult to obtain and surrogate endpoints such as vasopressor requirements and organ failure scoring may form the basis of practice recommendations in the future.

Conclusions
Th e ATN and RENAL studies have now established an upper limit of intensity for CRRT. In addition, they found These studies indicate a plateau response at the dose ranges examined. To reproduce these results, clinicians will need to prescribe continuous renal replacement therapy doses above the lower target dose in the trial protocols (20 or 25 mL/kg per minute) as larger periods of fi lter downtime can be expected outside a clinical trial environment. Below this best-practice region, survival is likely to be dose-dependent; however, the exact nature of this relationship has not been formally determined. Doses above the best-practice region are unlikely to be benefi cial to unselected patients and could potentially be harmful. ATN, Veterans Aff airs/National Institutes of Health Acute Renal Failure Trial Network; RENAL, Randomized Evaluation of Normal versus Augmented Level. Adapted from [18].
no evidence to suggest that any specifi c subgroups would benefi t from higher doses of RRT, refuting previous smaller studies [14]. Th e fi ndings of these studies, however, do not imply that the estimation of dose is unimportant. A dose-response relationship is likely at lower treatment intensities (that is, less than 20 mL/kg per hour), and a signifi cant minority of patients still receive treatment intensities less than this [11,12]. Given the likelihood of a dose-response relationship somewhere below 20 mL/kg per hour, delivery of doses lower than this should be avoided [22]. To ensure outcomes similar to those seen in the ATN and RENAL trials, clinicians should also prescribe CRRT on the basis of patient body weight to the established effl uent fl ow rate target of 20 to 25 mL/kg of body weight per hour.