High-volume Hemofiltration in the Intensive Care Unit: A Blood Purification Therapy

HVHF is not well defined in medical literature. Terms such as “high-volume,”“high intensity,” or “high flow” are used in publications for a large range of ultrafiltration flow rates, including some that by contemporary standards are not “high.” Additional confusion stems from the fact that the same given ultrafiltration flow rate can be evaluated both in the “high-volume” arm of one study and in the “low-volume” arm of another one. For example, in the Acute Renal Failure Trial Network study, 35 ml · kg⁻¹· h⁻¹is the ultrafiltration flow rate for the group receiving the intensive therapy strategy, and, in the Boussekey et al.  study, this ultrafiltration flow rate was used in the “low-volume” hemofiltration group.4,14 

In 2002, the Acute Dialysis Quality Initiative workgroup defined HVHF as more than 35 ml · kg⁻¹· h⁻¹.15In 2012, this threshold is not universally accepted anymore, especially because, from a practical standpoint, 35 ml · kg⁻¹· h⁻¹does not seem that high. Consequently, Honoré and other experts recently updated and clarified the definition of HVHF.16They agreed that HVHF includes continuous high-volume treatment of more than 50 ml · kg⁻¹· h⁻¹24 h a day and intermittent high-volume hemofiltration with brief, very high-volume treatment at 100–120 ml · kg⁻¹· h⁻¹for a short period of 4–8 h, followed by conventional CVVH.17This latter strategy is also called “pulse HVHF” and was initially developed by Ronco et al.  18This definition is consistent with the 2001 Critical Care Nephrology conference held in Melbourne, Australia, where HVHF was defined as ultrafiltration flow rates greater than 50 ml · kg⁻¹· h⁻¹and very HVHF for ultrafiltration flow rates greater than 100 ml · kg⁻¹· h⁻¹.19At that time, it was also decided to universally use the “ml per kilogram per hour” as the unit to be used to express the ultrafiltration flow rate.

The question of the adequate dose for renal support in AKI patients requiring RRT was first illustrated in a landmark study by Ronco et al.  in 2000.20In this single-center Italian study, an ultrafiltration flow rate of 35 ml · kg⁻¹· h⁻¹improved survival in AKI patients receiving postdilution CVVH, as compared with those who received 20 ml · kg⁻¹· h⁻¹. Nearly simultaneously, Schiffl et al.  reported that intensive hemodialysis (performed daily) reduced mortality in patients with acute renal failure, as compared with conventional (alternate-day) intermittent hemodialysis.21Consequently, at that time, a relationship between the dose of renal support prescribed and patient outcomes, including mortality, was highly suspected.

However, subsequent randomized controlled studies did not confirm these results. Bouman et al.  found that survival and recovery of renal function were not different in 106 critically ill patients randomized in three different strategies of RRT: early HVHF, early low-volume hemofiltration, and late low-volume hemofiltration.22Similarly, Tolwani et al.  recently reported no difference in patient survival or renal recovery in 200 critically ill patients with AKI undergoing 35 ml · kg⁻¹· h⁻¹versus  20 ml · kg⁻¹· h⁻¹of continuous venovenous hemodiafiltration with prefilter replacement fluid.23On the contrary, Saudan et al.  published in 2006 interesting data suggesting that increasing dialysis dose by adding a dialysis component to CVVH increased survival. Indeed, in this trial that randomized 206 patients, the 28-day survival was significantly increased from 39% in the CVVH group to 59% in the continuous hemodiafiltration group (P = 0.03), and the 90-day survivals were even more pronounced (34 vs.  59%, P = 0.0005).24 

So why did all these randomized controlled studies not find the same results? First, they are all relatively small, single-center studies (two centers for the Bouman study), and consequently, their results have limited external validity. It is well known that interventions tested in single clinical environments are not necessarily generalizable to a broader population, especially in critical care.25Second, the trials were conducted by committed experts of RRT with a lack of blinding, increasing the risk of a possible Hawthorne effect (some patients belonging to one group consciously or unconsciously treated differently than the other).26Third, AKI patients studied in these trials were drastically different. For example, the percentage of septic patients was below 15% in the Ronco study, whereas it was 60% in the Saudan study. Mortality rates in these studies were also very different, demonstrating the patients' heterogeneity in terms of severity and consequently pointing out the difficulty in analyzing them collectively. In the Bouman study, survival at day 28 was indeed unusual, reaching 70%. Lastly, management of RRT was very different between these trials. Substitution fluid replacement was sometimes administered in postdilution (Bouman and Ronco studies) and sometimes in predilution (Saudan and Tolwani studies), therefore modifying the actual delivered dosage from the prescribed dosage. Furthermore, the percentage of patients achieving a delivered dosage greater than 80% of the prescribed dosage was not always provided. Whereas all patients reached values of ultrafiltration of at least 85% of the prescribed dose in the Ronco study, only 77% of patients achieved greater than 80% of the prescribed dosage in the Tolwani study.

Recently, the two large multicenter prospective randomized trials were published, producing nearly identical results. The Acute Renal Failure Trial Network study, from the Veterans Affairs/National Institutes of Health Acute Renal Failure Trial Network, enrolled 1,124 AKI patients from 27 tertiary-care Veterans Administration and university hospitals in the United States.4Patients were randomized to either an intensive or a conventional management strategy. Patients assigned to the intensive treatment strategy received intermittent hemodialysis six times per week or continuous venovenous hemodiafiltration at 35 ml · kg⁻¹· h⁻¹, depending on their hemodynamic stability. For patients receiving the less-intensive treatment strategy, intermittent hemodialysis was provided three times per week or continuous venovenous hemodiafiltration was prescribed to provide an effluent flow rate of 20 ml · kg⁻¹· h⁻¹. There was no significant difference between the two groups in terms of mortality rates from any cause at day 60, duration of RRT, AKI recovery, and recovery from nonrenal organ failure. The RENAL study enrolled 1,500 patients from 30 intensive care units in Australia and New Zealand. AKI patients were randomized to receive an “augmented” continuous RRT regimen of 40 ml · kg⁻¹· h⁻¹or a normal regimen at a dose of 25 ml · kg⁻¹· h⁻¹.3The dose of 25 ml · kg⁻¹· h⁻¹represents the average practice in Australia and New Zealand, which is also true in Europe and the United States.6,12,27Results of the RENAL study were similar to those of the Acute Renal Failure Trial Network study, with an odds ratio of 1.0 for the 90-day mortality. Finally, two recent meta-analyses including all of the above trials concluded that no benefit could be demonstrated by increasing RRT intensity beyond the currently recommended 20–25 ml · kg⁻¹· h⁻¹.28,29However, in order to achieve this dose in clinical practice, prescriptions of 25–30 ml · kg⁻¹· h⁻¹are generally required, and close monitoring of dose delivery is advised.3,–,5,12,30 

In conclusion, if the physician's singular goal is renal support through a RRT prescription for a patient with AKI, the best evidence currently available suggests that hemofiltration should be prescribed at 25 to 30 ml · kg⁻¹· h⁻¹.

Systemic inflammatory states, such as severe sepsis and acute pancreatitis, are known to be major causes of AKI and other organ failures in critically ill patients.31But they are also responsible for an important immunologic disturbance with the release into the bloodstream of numerous inflammatory mediators.32Some aspects of this systemic inflammatory response may be beneficial to combat systemic infection, but when it is excessive, uncontrolled, or unbalanced, it has deleterious effects, including multiorgan failure and death. Indeed, inflammatory mediators such as cytokines have direct harmful effects on tissues (cytotoxic effects),33and the prolonged release of antiinflammatory mediators, interpreted as a protective mechanism to prevent an excessive effect of the pro-inflammatory phase, leads to impaired immunity;34this defect plays a major role in mortality.35,36The overall concept of blood purification is therefore to attenuate this overwhelming systemic overflow of pro- and antiinflammatory mediators released at the early phase of sepsis and to restore a broad-based humoral homeostasis in order to improve outcome. A variety of mediators are involved in this inflammatory response, such as cytokines, chemokines, complement components, platelet-activating factor, leukotrienes, thromboxanes, and kinins.37In the past, every attempt to modulate this inflammatory response by targeting one component has failed, at least at the clinical phase.38Thus, the concept of blood purification has evolved toward a nonspecific removal of a large spectrum of inflammatory mediators.

Different pathophysiological theories have been proposed to support the concept of blood purification. First, Ronco and Bellomo hypothesized that shaving the cytokine peak concentration by removing them from the blood compartment during the early phase of sepsis could stop the inflammatory cascade, limit organ damage, and consequently decrease the incidence of multiorgan failure syndrome. This hypothesis is called the “peak concentration hypothesis.”39,40More recently, Honoré proposed the “threshold immunomodulation hypothesis,” which takes a more dynamic view, postulating that the cytokine removal from the blood compartment leads to the removal of cytokines located at the tissue level because of an equilibration of their concentrations between these two compartments.41,42This theory is interesting because it affects cytokines at the tissue level, which is where cytokines are harmful, and it also explains why numerous studies assessing blood purification techniques found an improvement of outcomes with no modification of cytokine blood concentrations as cytokines from the tissues replace those removed from the blood.43,44Di Carlo and Alexander also proposed the “mediator delivery hypothesis,” in which HVHF is responsible for an increase of the lymphatic flow by 20- to 40-fold because of the high amounts of crystalloid fluids used for replacement with this technique.45,–,48This leads to a significant drag and displacement of inflammatory mediators to the blood compartment, making them available for removal. In this theory, HVHF is therefore not only able to remove cytokines from the blood but also to insure that lymphatic transport from tissue and interstitium to the blood compartment occurs. Finally, Kellum has suggested that blood purification therapies act at the inflammatory cell level in order to restore the immune function through the regulation of monocytes and neutrophils.49,–,51A novel component of this hypothesis is that by removing mediators from the plasma in the situation of systemic inflammation, one can restore the concentration gradient from plasma to infected tissues (fig. 2). This gradient has large effects on leukocyte trafficking and bacterial clearance.52Thus the “cytokinetic model” may be more important than the cytotoxic models traditionally considered to explain the association between high cytokine levels and mortality. It needs to be acknowledged that neither of these different theories is based on robust data, and once discussed, they usually lead to subsequent questions. For example, do other extracorporeal blood purification techniques such as the use of endotoxin sorbents interfere with the inflammatory response in the same way as HVHF? It is wise to imagine that several different mechanisms may be responsible for the effects observed with these techniques. Moreover, because of the activation of leukocytes during the passage of blood through the artificial extracorporeal circuit, could HVHF interfere with the apoptotic or the leukocytic infiltration pathways of sepsis-induced AKI pathophysiology? Or, for the same reasons, could it be responsible for an aggravation of the tubular injury lesions observed in patients with septic induced-AKI, as recently hypothesized by Lerolle et al. ?53 

Regarding pathophysiology, HVHF is an attractive therapy in order to successfully remove a significant amount of a wide range of inflammatory molecules involved in the host inflammatory response. First, these circulating molecules are mostly water-soluble, and convection carries both plasma water and solutes across the membrane. Second, most inflammatory mediators are so-called “middle molecular weight molecules,” with a wide range of mass from 5 kDa to 60 kDa. Convection is far more effective than diffusion in removing middle molecules,54,55at least with traditional membranes. Third, depending on their composition, most hemofiltration membranes also have some adsorptive properties. This associated principle corresponds to the saturable fixation of some molecules directly on the membrane along an affinity gradient depending on ionic, hydrophobic, and van der Waals interactions (fig. 3). The ultrafiltrate contains the molecules from the plasma that have a molecular weight below the membrane cutoff, and adsorption allows the removal of some molecules with a molecular weight higher than the membrane cutoff. Therefore, a very large panel of inflammatory mediators can be affected by hemofiltration. Finally, the use of HVHF is important because it has been shown that conventional hemofiltration, with low ultrafiltration flow rates, is not effective for blood purification.56,57Increasing the ultrafiltration flow rate may also increase the adsorption properties of the hemofilter because of its effect on transmembrane pressure (greater membrane site recruitment) and exposure of more of the internal matrix of the filter, thus increasing the available adsorptive surface area.58 

Many animal studies have been performed to assess HVHF, especially in the 1990s, when HVHF was still very experimental in humans. Grootendorst et al.  reported in 1992 an improvement in right ventricular function and the cardiac performance in 18 pigs with endotoxin-induced shock when zero balance HVHF (ultrafiltrate = 6,000 ml/h) was applied. He hypothesized that some vasoactive mediators, responsible for the myocardial depression, were removed with HVHF.59Two years later, he also found in a gut ischemia and reperfusion porcine model that HVHF not only improved short-term hemodynamics, but also reduced macroscopic bowel damage observed at autopsy and improved 24-h survival.60In an experimental pancreatitis model, HVHF improved hemodynamics, reversed sepsis–induced immunoparalysis, and increased 60-h survival.61In addition, in this pancreatitis study, Yekebas et al.  showed that survival rates were best with a very high ultrafiltration flow rate (100 ml · kg⁻¹· h⁻¹) and frequent filter changes (every 12 h).51In septic dogs, Bellomo et al.  found that 80 ml · kg⁻¹· h⁻¹HVHF improved hemodynamic parameters compared with a sham circuit with no hemofilter. This study also raised the likelihood of the beneficial effect of the adsorption properties of the polyacrylonitrile hemofilter used in the HVHF group.62Rogiers et al.  found similar results in dogs with endotoxin-induced shock, showing an improved cardiac performance when HVHF was performed with polyacrylonitrile membranes compared with polysulfone membranes. However, these effects that are supposed to be related to the adsorption properties of the membrane are temporary because of saturation phenomenon.63 

Some animal studies assessed HVHF by looking at ultrafiltrate obtained from either healthy donor or septic donor animals and infused into a healthy acceptor animal. The ability for HVHF to remove toxic mediators is suggested when ultrafiltrate obtained from septic animals leads to hemodynamic disturbances or even death in healthy animals.64,–,66In a prospective randomized controlled study including 65 septic pigs, Lee et al.  reported an increase survival time in filtered animals compared with matched nonfiltered animals. Increments in survival time even increased directly with filtration fraction. Moreover, ultrafiltrate concentrate obtained from septic pigs produced death in healthy animals, whereas the infusion of clean ultrafiltrate concentrate produced no response, supporting the hypothesis that HVHF was able to remove harmful mediators.65 

Although most of the well-conducted animal studies found benefits in terms of hemodynamics, respiratory function, and survival, they should be interpreted with caution. Septic animal models, especially endotoxemia, are indeed very difficult to extrapolate to clinical conditions. For example, most infection models do not include antibiotics, and the blood purification therapy is usually started at the very early phase after the bacterial insult, which is far different from clinical settings.

Like in animals, numerous human studies have shown beneficial hemodynamic effects of HVHF. Journois et al.  were among the first to study HVHF as a blood purification technique in humans. They studied children undergoing cardiac surgery with cardiopulmonary bypass.67,68In 20 children, zero-balance, HVHF (100 ml · kg⁻¹· h⁻¹) showed a reduction in postoperative blood loss, time to extubation, and cytokine plasma levels.67In 2001, Cole et al.  assessed the hemodynamic impact of HVHF in a small, randomized crossover clinical trial. In 11 patients with septic shock and multiorgan failure, an 8-h period of HVHF (6 l/h) was associated with a greater reduction in norepinephrine requirements than a similar period of CVVH (1 l/h).58Reduction of vasopressor requirements with HVHF was also found more recently in another pilot randomized study comparing CVVH at 65 ml · kg⁻¹· h⁻¹versus  35 ml · kg⁻¹· h⁻¹in 20 septic shock patients with AKI.14 

Using mortality as the primary outcome, large randomized controlled studies of HVHF in septic shock are extremely difficult to conduct because the sample size requires several hundred subjects. Therefore, such studies have not yet been performed; however, one such study is currently ongoing in Europe, known as the IVOIRE study (hIgh VOlume in Intensive caRE), which compares 70 ml · kg⁻¹· h⁻¹versus  35 ml · kg⁻¹· h⁻¹. Although results from this study will be formally released within the next 6 months, we already know from the investigators that the enrollment process was very slow and therefore the study may be underpowered. Thus, the only available studies regarding mortality compared observed-mortality versus  expected-mortality based on the patients' severity scores at admission. However, despite being uncontrolled and thus limited in their ability to draw strong conclusions, at least six studies have found significant (sometimes spectacular) reductions in mortality rates with HVHF compared with predicted mortality.69,–,74In 2000, Honore et al.  reported a reduction of the mortality rate from 79% (expected mortality based on APACHE II and SAPS II scores) to 55%.70A few years later, Joannes-Boyau et al.  obtained a similar result with a predicted 28-day mortality of 70% and an observed mortality of 46% in a study assessing the effect of 40–60 ml · kg⁻¹· h⁻¹maintained for 96 h in patients with septic shock and multiorgan dysfunction syndrome.71In addition, Ratanarat et al.  reported a reduction of the 28-day mortality rate from 70% (predicted) to 47% (observed) when septic patients underwent daily pulse HVHF.74 

In patients without sepsis but with systemic inflammation, the effect of HVHF on mortality was evaluated in two randomized controlled trials. The largest study assessing mortality with HVHF in randomized patients was performed on 61 resuscitated cardiac arrest patients.75This whole-body ischemia-reperfusion represents an interesting “sepsis-like” syndrome.76Although the primary analysis was negative, after adjusting for imbalances in baseline characteristics, very HVHF (200 ml · kg⁻¹· h⁻¹during 8 h) was associated with improved 6-month survival (logistic regression odds ratio, 4.4; 95% CI, 1.1–16.6) and a decreased risk of death from early intractable shock.75Similarly, in 37 patients with acute pancreatitis, hemodynamics and short-time survival rates were significantly better with HVHF (4 l/h) than in patients receiving low-volume hemofiltration (1 l/h).77The most important recent studies assessing mortality with HVHF as a blood purification therapy are summarized in table 1.

Unlike these promising results using HVHF, standard “renal dose” continuous RRT appears to be ineffective as an immune-modulating therapy. In 1998, Kellum et al.  found modest effects of CVVH (2 l/h) compared with continuous hemodialysis on tumor necrosis factor and interleukin-6 in a small (n = 13) randomized crossover study.55When patients served as their own controls, median differences in the tumor necrosis factor level were only 16% and highly variable between patients. In 2002, Cole et al.  studied the effects of early CVVH (2 l/h) on plasma concentrations of several inflammatory mediators and on organ dysfunctions in 24 septic patients without AKI. In this randomized controlled trial, CVVH was not associated with a reduction in any plasma cytokine concentrations and did not decrease organ dysfunctions, as compared with no hemofiltration.56Recently, Payen et al.  found similar results (and even a trend toward worse outcomes) in septic shock patients without AKI who underwent CVVH (25 ml · kg⁻¹· h⁻¹for a 96-h period) at the early phase of sepsis compared with those who were managed conventionally. This study confirmed that low-volume hemofiltration (standard continuous RRT) is not a suitable technique for blood purification in sepsis but, as the authors stated in their conclusion, it did not rule out a beneficial effect of HVHF on the course of sepsis.57 

Other extracorporeal blood purification therapies are currently available besides HVHF. Indeed, coupled plasma filtration adsorption, polymyxin-B hemoperfusion, and the use of high cutoff membranes have also been proposed as adjuvant treatments for sepsis.78,–,80However, at this time, it is not possible to state which technique is most effective, because they have not been compared with each other so far. Nevertheless and interestingly, some additional hybrid techniques can synergistically combine HVHF and hemoadsorption in a technique named “high adsorptive hemofiltration.” This therapy consists of optimizing the performances of the hemofilters regarding cytokine and/or endotoxin removal through hemoadsorption by manipulating their composition or structure. For example, positive hemodynamic effects of a polyacrylonitrile hemofiltration membrane having endotoxin adsorption properties were recently reported in septic pigs.81The membrane surface polarity was modified by the addition of a polyethyleneimine coating, a positively charged polymer, allowing the membrane to catch negatively charged endotoxins via  surface adsorption (fig. 4). In the same line, other models where HVHF and high-permeability hemofiltration work synergistically have shown promising results.82 

Although new machines dedicated for RRT or blood purification are more reliable with very sensitive and precise pressure control and volume balance functions, the use of HVHF still requires attention to several technical aspects. First, in order to keep filtration fraction below 25%, the blood flow rate has to be increased. This consequently requires the use of large dialysis catheters with diameters such as 13.5–15 French inserted in an adequate location, the right jugular site perhaps being the best.83Second, the optimal pre- and postdilution ratio for replacement fluid is suggested to be 1/3–2/3 by some experts.17Indeed, a good compromise between the loss of efficacy because of predilution and its benefit regarding blood rheology conditions is suitable, although data addressing this particular aspect are limited.84Third, to date, no recommendation about anticoagulation can be made because of the lack of evidence of any superiority of any one strategy. However, citrate is increasingly employed because of its advantages in terms of filter lifetime and coagulation. Fourth, highly biocompatible synthetic membranes with a high exchange surface of at least 2 m²are suitable.17Fifth, temperature monitoring must be performed very closely because of the greater heat loss observed with high convective exchanges and the subsequent risk of hypothermia. Sixth, HVHF may be responsible for an important loss of small beneficial molecules including nutrients, amino-acids, vitamins, trace elements, and antimicrobials in the ultrafiltrate.85,86Thus, nutritive supplementation (e.g. , glutamine) and adaptation of drug prescriptions must be taken into account.87Antibiotics levels in particular should be checked whenever possible. When using pulse HVHF, it is therefore advisable to dose around the procedure, much the same way drugs are managed in intermittent hemodialysis. For the same reason, strict monitoring of sodium, potassium, phosphorus, glucose, and acid-base balance is mandatory.17 

Regarding safety, it is also important to recall that HVHF is usually performed in patients with a multiorgan failure syndrome, and therefore the adverse effects of the therapy are exacerbated by the severity of these patients. HVHF has the same risks as other extracorporeal therapies such as bleeding related to anticoagulation, infection, gas embolism, and hemodynamic intolerance, but the use of high fluid exchanges multiplies the risk of metabolic consequences if there is an error in the replacement fluid composition. In other words, physicians have to be aware that a small mistake in the therapy prescription can rapidly have major deleterious consequences. Finally, the difference between the prescribed and delivered dose can rapidly become important with HVHF if the therapy is interrupted for some reason (nursing activities, coagulation issues, transfer to the operating room, or radiology).

Physicians initiating RRT should first identify their goals. If the strategy is to support renal function, then HVHF is not necessary and ultrafiltration flow rates ranging from 25 to 30 ml · kg⁻¹· h⁻¹will be adequate for the majority of patients. Conversely, if the strategy is to overtake renal support and modulate systemic inflammation, HVHF appears to be a promising, though still experimental, option (fig. 5). Although there are still no large multicenter randomized controlled trials showing beneficial effects on mortality with HVHF, preliminary studies in humans and preclinical animal data support continued research in this area.

In our opinion, during the next decade, considerable work needs to be done in order to find and optimize the best blood purification strategy for systemic inflammatory states. Convection, diffusion, and adsorption should probably not be seen as opposed or competitive mechanisms for blood purification but rather as complementary ones. Since the mechanism by which blood purification therapies work on modulating the cytotoxic and cytokinetic effects of inflammatory mediators is still unclear, we believe that basic science studies are still warranted. In addition, more clinical trials conducted in both mono- and multicenters are suitable in order to evaluate the ability of HVHF to improve clinical outcomes (i.e. , mortality or organ failure) rather than to study surrogate markers (i.e. , inflammatory mediator clearance or physiologic variables such as oxygenation and hemodynamics).