Rimmele and Kellum have a nice article from 2012 summarising the key points of this practice. It has been asked about in Question 5 from the first paper of 2003. The answer to that question is a succinct summary of the evidence. In contrast, what follows below is more of a rant.
How "high" is high volume?
"High volume hemofiltration" had indeed suffered from a lack of a firm definition in 2003. The term "high volume hemofiltration" has been now been applied to high intensity hemodialysis in general, i.e. the use of a dialysis dose greater than 50ml/kg/hr for a 24 hour day, or using insane 100-120ml/kg/hr doses for a shorter period.
Taken literally, it is essentially the practice of using ultrafiltration as the primary mechanism of solute removal. Thus, transmembrane pressure is used to drive vast deluges of plasma through the filter, thereby allowing solute drag to pull larger amounts of unpleasant substances from the patient. The practice originates from the understanding that convection is better capable of clearing middle molecules more effectively, and thus excessive reliance on convection may result in a high-volume removal of such molecules. This "middle molecules" category contains within it both undesireable elements (eg. inflammatory cytokines) and essential elements (eg. antiinflammatory cytokines, vitamins and amino acids).
Proposed mechanism: how could this possibly be helpful
Proinflammatory mediators contribute to the pathogenesis of multi-organ system failure in systemic inflammatory states. Thus, if one were able to extract these from the patient's bloodstream, one might be able to avert the worst of a SIRS response.
These molecules are predominatly small-to-medium water soluble molecules - generally speaking, polypeptide cytokines. One could expect these molecules to be readily removed by ultrafiltration. Ultrafiltration should be the ideal means of removing these molecules because their size (5kDa-60kDa) makes them sub-optimal candidates for removal by countercurrent diffusion in dialysis.
Could this be true? We certainly have thought so in the past. Bellomo et al report on a series of encouraging pig experiments, where the septic pigs were rescued from haemodynamic collapse by high volume haemofiltration. Most excitingly, the reverse occurred when the septic ultrafiltrate was infused back into other pigs: they became profoundly shocked and their hearts failed. Clearly, something proinflammatory was being filtered out of the pigs. More good arguments for why this treatment should work are offered in some enthusiastic papers from the late 1990s. The linked article is a good overview of all the convincing-sounding animal studies.
The advantage of using "high volume" rather than routine CVVHF is the rapid emptying of the intravascular compartment. As you clear the proinflammatory mediators from the blood, you increase the concentration gradient for them to come out of the tissues, increasing the overall clearance. Furthermore, using higher transmembrane pressures may may increase the exposure of more membrane sites, improving ultrafiltration rates. Lastly, the membranes themselves may play a role in removing the cytokines by adsorption (i.e direct deposition on to the membrane surface).
Everything would be fine if the only small-to-medium water-soluble molecules in the human body were nasty pro-inflammatory mediators. Plainly, this is not the case. In fact, there are a lot of very useful molecules which also get removed from the body at a rate proportional to ultrafiltration. These include water-soluble vitamins, amino acids, anti-inflammatory cytokines, and therapeutic agents (eg. antibiotics).
Furthermore, dragging blood at a rapid rate though a haemofilter is not a benign process; the increased rate of haemolysis and complement activation would potentiall offset any cytokine-depletion benefit.
Lastly, think of the circuit. They are not designed to run at such a massive dose rate. The membrane is more likely to rupture if the transmembrane pressure is ridiculously high. The increased rate of circuit loss may be prohibitively expensive. Or, the rate of blood loss.
The Prismaflex people apparently claim that 8 litre exchanges are possible, which would be a dialysis dose of about 120ml/kg/hr for a 70kg patient. However, this would mean that the bedside nurse would have to carry away 8 litres of effluent every hour, and return with 8 litres of replacement fluid. That would be 128 litres of fluid carried back and forth over the course of an eight-hour shift.
This would not be a popular move.
Evidence for high volume haemofiltration
- Early (1992) pig studies were encouraging. The anaesthetised pigs did well.
- Early (2000) small-scale human trials also demonstrated a haemodynamic improvement.
- Small early cohort studies had demonstrated an improvement in survival and hemodynamic performance with doses around 100ml/kg/hr, in patients with sepsis, pancreatitis, and SIRS due to post-cardiac arrest global ischaemia.
- Some trials developing between 2002 and 2009 (including the RENAL study, which is discussed elsewhere) built the impression that high dialysis doses do not achieve anything. Critics countered with the observation that not all the studied patients had sepsis.
- In 2012, people were still excited about this and described HVHF as "a promising, though still experimental, option".
- The IVOIRE study (2013) compared doses of 75ml/kg/hr with 35ml/kg/hr. They found no benefit in either primary or secondary outcomes, and were forced to conclude that this therapy "cannot be recommended for septic shock complicated by AKI".
- A recent (2014) meta-analysis of high volume hemofiltration in patients with sepsis-associated AKI identified four RCTs, finding no evidence of benefit in primary or secondary outcomes, and attributing a greater frequency of wild electrolyte derangement to the high-volume group.
This section probably links rather well with the evidence for the optimim dose of dialysis.
Indeed many of the studies quoted above can be used to support the cardinal statement: that anything over 25 ml/kg/hr is probably useless, and might be dangerous.