"The difficulty lies, not in new ideas, but in escaping old ones, which ramify for those brought up with them, as most of us have been, into every corner of our minds."

John Maynard Keynes (1933)

Since its first description [1] continuous hemofiltration, or `continuous renal replacement therapy' (CRRT) as it is now called, has undergone remarkable growth [2]. In the modern ICU, CRRT is now performed using pump technology [3] and double-lumen central venous access [4]. In many ICUs, especially in Australia and in Europe, CRRT has become the dominant if not exclusive form of artificial renal support [5]. Furthermore, there has been growing research into its role as adjuvant therapy in sepsis [6]. Modifications to standard CRRT circuits are also being explored in an effort to increase such anti-inflammatory potential [7]. In the present review, we describe the current technology and state of CRRT in the ICU, address some of the controversies that surround its application in critically ill patients, and attempt to give the reader a state-of-the art view of its uses and clinical future.

Continuous arteriovenous therapies were first used for CRRT because they are simple and do not require a peristaltic blood pump. As such, they may have a place under emergency circumstances or in developing countries. However, the morbidity associated with arterial cannulation is substantial [8], and the cost of a simple blood pump module is only £2500. Thus, if one can afford to have an ICU at all, one can afford to have venovenous CRRT, which is much safer for the patient (Fig. 1) [8,9,10]. All CRRT modalities should now be venovenous.

Once venovenous therapy is applied, blood flow rate must be controlled. A peristaltic pump module is necessary to achieve this goal. This module must have the appropriate air-trap and pressure monitors to ensure patient safety. In this setting, either continuous venovenous haemofiltration (CVVH) or continuous venovenous haemodialysis (CVVHD), or a combination of both [continuous venovenous haemodiafiltration (CVVHDF)], may be chosen. All techniques will deliver excellent uraemic control provided ultrafiltrate flow and/or dialysate flow is adequate. In fact, with sufficient blood flow (≥ 200 ml/min) and membrane surface (≥ 0.8 m²), CVVH without pump-driven ultrafiltrate control initially will spontaneously deliver high ultrafiltration rates (1.5-2 l/h), and thereby high solute clearances without the need for counter-current dialysate flow [11]. To facilitate nursing care, however, ultrafiltration or dialysate flow should be pump-controlled (Figs 1 and 2). All new machines for CRRT possess such technology. If only a simple blood module is available, ultrafiltration (and replacement fluid) or dialysate flow rate can be controlled by means of a standard volumetric pump. Such volumetric pumps are ubiquitous in the ICU [12].

If it is necessary to combine diffusive and convective clearance, as in CVVHDF, this can be achieved by using the pump to control dialysate inflow and outflow. If the clearance is purely diffusive (dialysate inflow rate=dialysate outflow rate) and the membrane is low-flux (cellulose-based), then therapy is more appropriately called CVVHD. A consensus nomenclature has been published [13] that facilitates uniformity of communication and more precise exchange of ideas and clinical experience. The authors of that nomenclature currently prefer to use CVVH with pump-driven ultrafiltrate control because of its greater ability to remove middle molecules (most soluble mediators of sepsis are middle molecules), its safety, and ease of operation by nursing staff. This CVVH-based approach to the treatment of acute renal failure (ARF) has now been extended to include high-volume haemofiltration [7,14] and more complex circuit modifications [15,16] that are aimed at increasing blood purification in septic shock or at removing excess plasma water during cardiac surgery [17].

CRRT offers extraordinary advantages over IHD and peritoneal dialysis. With CRRT, volume control is continuous and immediately adaptable to the changing clinical circumstances (the immediate need for blood or blood products in a patient who is at risk for adult respiratory distress syndrome or who is on extracorporeal membrane oxygenation) that are common in the care of critically ill patients. Because of this adaptability, CRRT can immediately treat volume overload or, even better, prevent it without inducing acute volume depletion. The avoidance of intravascular volume depletion and hypotension prevents treatment-associated ischaemic renal injury, which is seen with standard IHD [18].

Uraemic control with CRRT is vastly superior to that achieved with standard IHD [19]. Thus, patients treated with CRRT consistently maintain lower urea and creatinine levels [20]. Recent data [21,22] show that delivery of a greater `dialysis dose' is associated with better outcome in critically ill patients. The preliminary results of a randomized controlled trial that compared daily IHD with IHD every 2 days in patients with ARF [23] also show a statistically significant increase in survival in those patients treated with daily IHD. The better level of uraemic control provided by CRRT may offer a survival advantage.

CRRT offers more rapid improvement and control of metabolic acidosis, and more rapid and reliable control of serum phosphate levels [24]. However, hypophosphataemia will develop during CRRT unless clinicians monitor serum phosphate levels and administer replacement as appropriate as soon as the serum phosphate is within the normal range. CRRT also allows better nutritional support. With standard IHD, adequate control of uraemia is difficult, and protein restriction is often applied to prevent high levels of serum urea. Such restrictions induce protein starvation and a highly negative daily nitrogen balance [25]. An aggressive, protein-rich nutritional policy can be implemented if CRRT is used [26]. Such a policy maintains nitrogen balance close to neutral and prevents protein malnutrition [27]. Amino acid losses through the filter do occur. However, they represent approximately 10% of administered amino acids, and such losses are not appreciably greater than those seen during a session of IHD or during peritoneal dialysis [28].

CRRT is mandatory in all patients who are at risk of or who have increased intracranial pressure (neurosurgical patients, patients with encephalitis or meningoencephalitis or acute liver failure). In a series of elegant studies [29,30], Davenport and coworkers showed that CRRT prevents the surge in intracranial pressure that is associated with intermittent therapies.

ICU patients with significant cardiac disease are best treated with CRRT. In patients with diuretic-resistant congestive cardiac failure, CRRT restores dry body weight, and prolongs symptom-free and oedema-free time [31]. Clinical benefits have also been reported for cardiac surgery patients [32]. The possible mechanisms include decreased myocardial oedema, a decrease in left ventricular end-diastolic pressure, optimization of the Starling relationship, increased myocardial performance, and the removal of circulating myocardial depressant factors [33]. Finally, in an era of increasingly aggressive cardiac surgery and artificial mechanical heart support, a number of patients develop ARF in the setting of postcardiotomy cardiogenic shock. These patients require temporary mechanical heart support (left or right ventricular assist devices, extracorporeal membrane oxygenation, haemopump) and renal replacement therapy, they are dependent on vasopressor drugs, and they are haemodynamically fragile. The use of standard IHD can be lethal in such patients. CRRT, on the other hand, is easily tolerated and becomes a useful tool for control of intravascular and extravascular volume.

Patients with ARF and septic shock are particularly suited to CRRT. In these patients haemodynamic instability is very common, and oliguria and anuria are typical. If appropriate fluid resuscitation, nutrition, blood and blood products administration is to take place under optimal physiological circumstances, CRRT must be used. CRRT also appears to have beneficial effects on haemodynamics and inflammation in animal models of sepsis [34,35,36]. Accordingly, there is a strong biological rationale for using CRRT in septic shock and ARF. More recently, standard CRRT technology has been modified by using a more porous membrane [36]; by coupling continuous plasma filtration with continuous sorption [37]; and by increasing the plasma water exchange rate [38]. These modifications are aimed at moving CRRT from the simple treatment of ARF to the adjunctive treatment of sepsis.

When standard IHD is used, the initiation of renal replacement therapy is often delayed by concerns about haemodynamic tolerance. With CRRT, renal replacement therapy can and should be started promptly and aggressively. Accordingly, we have proposed a set of indications (Table 1) that can be used as triggers for initiating artificial renal support in the ICU. Early initiation of CRRT may increase survival [39].

Several controversies surround the use of CRRT, including the following: does CRRT increase survival?; does the anticoagulation that is needed in CRRT constitute an important disadvantage of CRRT?; do the costs of CRRT compare favourably with other modalities?; should ICU or haemodialysis nurses manage the CRRT circuit?; and should CRRT be used in patients without ARF who have severe sepsis or septic shock?

The development of CRRT and its increasing use by intensivists has put a great deal of pressure on nephrologists to adapt and compete. Accordingly, hybrid strategies are now emerging. These strategies seek to reach a middle ground between CRRT and standard dialysis. One such approach is called `slow extended daily dialysis' [54]. With this approach, dialysis is extended to 6-12 h with intermediate blood flows and dialysate flows. This approach represents an improvement in the type of IHD applied to ICU patients, which may make it possible for IHD to return to the ICU in a more competitive manner in the next few years.

CRRT is now firmly established as a form of artificial renal support in the ICU. In many units and in many countries, it has superseded IHD. Intensivists have begun to use CRRT independently, and are now exploring the opportunities that CRRT provides as an adjunctive treatment for severe sepsis. In particular, in patients such as those with heart failure, acute liver failure or cerebral oedema, the physiological advantages of CRRT over standard IHD are overwhelming. Once the appropriate training of nursing staff and medical staff has been achieved, CRRT is easy to conduct, is safe and flexible, and it will easily become the only form of artificial renal support in the ICU. The future may see CRRT move beyond its initial goal of providing renal support to the area of immune modulation in sepsis. However, many investigations and technological changes will be necessary to establish the efficacy of CRRT in sepsis.