Septic shock is usually accompanied by acute renal injury, heralded by a drop in diuresis. However, when standard intermittent haemodialysis (IHD) is used to treat renal failure, the initiation of renal replacement therapy is often delayed by concerns about haemodynamic tolerance. With the availability of continuous veno-venous haemofiltration, a safe procedure in haemodynamically unstable patients 1, there is no reason to delay renal replacement therapy 2. Moreover, haemofiltration has been reported to improve cardiopulmonary function in septic patients, even if they are not oliguric 34.

Based on these findings, in January 1995 we began to treat sepsis-related multiple organ failure with early continuous veno-venous haemofiltration combined with continuous veno-venous haemodiafiltration (CVVHDF). This strategy was our standard clinical practice for 10 years (from January 1995 to December 2004) and was accepted as a routine procedure by the Ethics Committee of the Société de Réanimation de Langue Française (Paris, France). Our clinical results are presented in this report, which focuses on the relation between rapid correction of metabolic acidosis with early renal replacement therapy and mortality.

The study group included 38 men and 22 women, whose mean age was 57 ± 16 years. The average measured body weight was 72 ± 13 kg. Circulatory failure was present at admission in 44 patients (primary shock) or occurred after several days of hospitalization in our unit in 16 patients (secondary shock). The number of patients included per year during this 10-year period is presented in Figure 1. These patients were predominantly medical (45 medical patients versus 15 surgical patients). The causative bacterial agent was subsequently identified from a positive blood culture or from a localized site of infection in 42 cases; Gram-positive and Gram-negative agents were responsible in 22 and 20 patients, respectively. The average SAPS II score for the whole group was 67 ± 18. During the same period, 35 patients who met the inclusion criteria but with a rapidly fatal underlying disease (McCabe score 2) were excluded from the study.

Individual changes in base excess during the observational period, defining refractory circulatory failure, are shown in Figure 2, and the trend in base deficit after 12 hours of CVVHDF was used to separate patients into two groups. In 40 responders (group 1), metabolic acidosis was reduced by the first 12 hours of CVVHVD. In 20 nonresponders (group 2), metabolic acidosis was unchanged or even deteriorated after 12 hours of CVVHDF (Figure 2). Table 1 summarizes the main physiological data for both groups. No between-group difference was observed apart from a significantly higher Logistic Organ Dysfunction Score in group 2. Mortality rate was 30% in group 1 and 100% in group 2 (P < 0.0000; Table 1 and Figure 3).

Twenty-eight patients recovered and 32 ultimately died, leading to a crude mortality rate of 53% for the whole group. This mortality rate was significantly lower than that predicted by SAPS II (79%; P = 0.01) leading to an standardized mortality ratio of 0.67.

Septic shock is a condition in which renal perfusion is markedly impaired 13, producing acute renal injury 6. Use of noradrenaline to maintain arterial pressure classically worsens renal hypoperfusion in the low output state 6, but it may improve this perfusion in the hyperdynamic state, a frequent pattern in septic shock 14. However, when shock becomes refractory, an acute renal failure syndrome may also develop 6. Even if it is essentially of prerenal origin, renal function is stopped and renal replacement therapy appears logical. Unfortunately, conventional IHD frequently worsens circulatory status in haemodynamically unstable patients, and renal replacement therapy by this method is usually delayed, pending circulatory improvement 4. Moreover, conventional IHD is usually considered only when metabolic disorders associated with organic renal failure are marked 15, which is not usually the case after an interruption to diuresis of only 6–12 hours. Both of these conditions impose an obligatory delay, during which most patients with refractory shock die.

An important finding of the present study was that a rapid metabolic improvement occurred during early CVVHDF in 67% of patients with refractory septic shock (group 1). Of course, the metabolic acidosis observed in our patients, which was associated with an increased lactate level, was interpreted as mainly resulting from circulatory failure. It was thus considered a consequence, not a cause, of circulatory failure. However, as indicated by abnormally high values of plasma phosphate, prerenal failure contributed in part to the metabolic acidosis. Thus, despite a short duration of renal impairment, renal acidosis was actually present. However, the metabolic improvement observed in group 1 did not appear to result mainly from the buffering action of CVVHDF, because it was associated with a circulatory improvement, permitting a rapid reduction in vasoactive support. This finding is at variance with clinical observations made during IHD. In a recent clinical report, despite adherence to practice guidelines, conventional IHD was associated with a drop in arterial pressure and with increased need for catecholamines in the majority of patients 16.

Concerning the evolution of renal function, the haemodynamic improvement observed during the first CVVHDF session in group 1 patients was associated with rapid renal recovery in half of the patients, as expected with a renal failure of prerenal origin. In the remaining patients from this group, however, this recovery was more progressive, suggesting that some tubular necrosis might have occurred. In these latter cases, additional CVVHDF sessions were required because of delayed recovery of renal function. Moreover, these additional sessions permitted rapid removal of the large volume of fluid given for resuscitation once circulatory failure had been corrected. This beneficial result is impossible with IHD, as was recently illustrated by the report cited above 16, in which only minimal reduction in body weight could be achieved at the end of renal replacement therapy. In three cases, in whom renal recovery was markedly delayed, renal replacement therapy was completed by IHD, which is a perfectly safe procedure when temporally distant from the initial haemodynamic problems.

Another important finding was that lack of metabolic improvement after 12 hours of CVVHDF (for example, unchanged base deficit despite the buffering action of CVVHDF) was associated with a 100% mortality rate. This finding of failure to improve metabolically after 12 hours might be important because it suggests that application of CVVHDF might be futile in such cases. Conversely, the lack of metabolic improvement after 12 hours of CVVHDF might prompt use of drotrecogin alfa, an expensive but efficient treatment for septic shock 17. The lack of improvement in base deficit after 24 hours of active treatment was recently underscored as a strong predictor of mortality in critically ill patients 18, suggesting the importance of this dosage in the monitoring of septic shock. In that report, a base deficit above 2.5 mmol/l or a blood lactate level greater than 1 mmol/l after 24 hours of adequate support were respectively associated with mortality rates of 71% and 82%.

The particularly poor prognosis of refractory septic shock has led some authors to consider whether strong vasoactive support may be futile in this setting 19. In this latter report, septic shock initially treated by dopamine and requiring subsequent noradrenaline infusion was associated with a 85% mortality rate. In another report 20 septic patients exhibiting at least three organ failures, as did our patients, had a mortality rate of 92%. The duration of lactic acidosis was found to be a good predictor of multiple organ failure 21, and early lactate clearance was associated with an improved outcome from septic shock in another study 22. In this context, our strategy of using early CVVHDF appeared to represent a rescue procedure; the final mortality rate of 53% for the whole group was significantly and markedly lower than that predicted by SAPS II (79%). However, this finding needs confirmation in prospective comparative studies. Despite our somewhat different procedure, including only middle volume haemofiltration, continuous dialysis and use of the same filter throughout the session, our results are similar to those recently reported by Honore and coworkers 23. In a group of 20 patients with refractory septic shock, whose SAPS II score was in the same range as that for our patients, those authors observed a mortality rate of 55%, in contrast to the predicted mortality of 79%.

Our present data suggest that early renal replacement therapy by CVVHDF might have a beneficial effect as an adjunctive treatment for refractory septic shock. This therefore raises a major question about the precise mechanism by which CVVHDF may improve this condition. Many humoral mediators that are potentially involved in the inflammatory response associated with sepsis are 'middle molecules', which are cleared by the technique. For this purpose, convection has been proved to be more efficient than diffusion 24. However, recent studies have yielded conflicting results. De Vriese and coworkers 25 demonstrated effective cytokine removal in patients with septic shock, but in a recent phase II randomized study Cole and coworkers 26 were unable to confirm that early use of haemofiltration at a filtration rate of 2 l/hour reduced the circulating concentration of cytokines associated with septic shock. Increasing the rate of fluid exchange across the membrane will increase convective transport, and this led the same group of investigators to suggest that high-volume haemofiltration might improve clearance of mediators in refractory septic shock 27. In the present study we did not use high-volume haemofiltration, as defined by Ronco and coworkers 28 as a threshold of 35 ml/kg per hour. Use of 2,000 ml/hour haemofiltration in our group of patients, whose average body weight was 72 kg, would have resulted in an average convection exchange of 28 ml/kg per hour. However, in a recent study Cole and coworkers 27 compared high-volume versus standard-volume haemofiltration in a group of 11 septic patients, and found that both techniques lowered the plasma concentration of mediators. The kinetics of this decrease suggested that it mainly resulted from membrane adsorption, which was achieved with both techniques, and in a recent pro/con debate the benefit of high-volume haemofiltration was challenged 29. Thus, membrane adsorption might be responsible, at least to some extent, for the haemodynamic improvement observed in the present study.

However, early and adequate renal replacement therapy might also have contributed to the improved prognosis. Acute renal failure per se exerts an independent and specific effect on the morbidity of critically ill patients 3031. The mortality rate in acute renal failure occurring in a septic context was 75% in a large prospective study, in which IHD was the quasi-exclusive therapeutic procedure 32. Critically ill patients probably benefit from a combination of diffusion and convection that provides sufficient elimination of small and larger toxins and, in contrast to the majority of other reports, we combined dialysis with our procedure, anticipating better treatment of renal failure syndrome 33. Also, this technique probably permitted a more rapid correction of metabolic acidosis by acting on the renal component of this disorder. However, if it should be considered, any impact of this choice on the clinical outcome remains purely hypothetical.

Without providing any idea regarding the precise mechanism, which would require additional comparative studies, our clinical report suggests that early renal replacement therapy by CVVHDF may improve the prognosis of the most severe forms of septic shock, and should be considered as an adjunctive therapy in sepsis-related multiple organ failure. Moreover, after 12 hours, this procedure distinguishes a subgroup of patients with a 100% probability of death and so it could perhaps help in deciding whether to institute a more expensive treatment, such as drotrecogin alfa.

• Early CVVHDF may improve the prognosis of sepsis-related multiple organ failure.

• Failure to correct metabolic acidosis rapidly during CVVHDF (for example nonresponders) is a strong predictor of mortality.

• By screening nonresponders, early CVVHDF could help in deciding whether to institute a more expensive treatment, such as drotrecogin alfa.

CVVHDF = continuous veno-venous haemodiafiltration; IHD = intermittent haemodialysis; SAPS = Simplified Acute Physiology Score.

The authors declare that they have no competing interests.

FJ and AV-B conceived and designed the study, and drafted the manuscript. FJ supervised and was responsible overall for all aspects of the study. BP acquired a substantial proportion of the data. KC, OP and AR performed data collection. AB and PA supplied statistical expertise.