Statistical Analysis

Profile Analysis with Repeated Measurements (MANOVA or multivariate ANOVA) (24, 25) was used to study the hemostatic changes in both study groups over the course of 48 hours of extracorporeal treatment. This statistical method allows us to test three different hypothesis. If the profiles are horizontal (test of horizontality), there are no condition effects, whereas if the profiles are equal (test of identity) there are no group effects. Finally, if the profiles are neither horizontal nor equal they may still be parallel (test of parallelism), which is an indication that there is no interaction between the group effects and the condition effects; in other words, there were no different effects between CVVH and CVVHDF. p-values <0.05 were regarded as significant. The statistical analyses were carried out with the SPSS package for Windows v7.5.

Results

As shown in Table I, there was no significant difference in the changes of acid-base status associated with the type of CRRT (CVVH and CVVHDF). After 24 h of CRRT the acidosis had been corrected in both groups and the mean values of serum bicarbonate, pH and blood lactate were similar at the different points. At 24h the significant increase of serum bicarbonate in Group I could be explained by a minor baseline mean value with respect to Group II. Although basal blood lactate was not available at 24 h of CRRT, Group II, which had been treated with lactate containing CVVHDF, had a mean level lower than group I. On the other hand, the level of group I also decreased after 24 h of CVVHDF (at 48 h). Although theoretical concerns have been raised that lactate absorption from the dialysis solution may produce hyperlactemia, clinically significant changes in the serum lactate concentration were not observed. (26-28). In our case both CVVH and CVVHDF corrected the acidosis without significant increases of blood lactate levels as have been described previously (29).
Baseline levels of endothelial injury and hemostatic factors for patients were higher than the normal range for healthy adults (Tab. II).

TABLE II - ENDOTHELIAL DAMAGE AND HEMOSTASIS FACTORS IN SIRS PATIENTS AT BASELINE POINT NORMAL VALUES

The results from MANOVA are shown in Table III.

TABLE III - HAEMOSTATIC CHANGES DURING THE 48h OF CONTINUOUS RENAL REPLACEMENT THERAPY

The values for each factor taken at baseline time, at 24 hours and at 48 hours are shown with their respective probabilities of parallelism, horizontality and identity. The order of administering hemofiltration and hemodiafiltration did not affect the vWF, thrombomodulin and t-PA antigen, which are related to endothelial damage (parallelism test (p): 0.101, 0.505 and 0.941, respectively). Although the vWF and thrombomodulin showed a tendency to increase over the 48 hours and the t-PA antigen over the first 24 hours, these changes were not significant (horizontality test (p): 0.16, 0.071 and 0.525, respectively). The significantly higher values of thrombomodulin in group I than in group II (identity test (p): 0.014) and t-PA value in group II than in group I are attributable to higher baseline values obtained by chance, and not because of the treatment (parallelism test (p) >= 0.05). PAI-1 antigen and PAI-1 activity showed a tendency to decrease over 48 hours, and its functional change was statistically significant (horizontality test (p) = 0.012) (Fig. 1).

parallelism test (p) = 0.167
horizontality test (p) = 0.012
identity test (p) = 0.044

The significant differences in PAI-1 between groups I and II (identity test (p) = 0.044) were also at the baseline, so they would be due to chance in the allocation of patients but not to the treatment (parallelism test (p) >= 0.05).
The parameters of activated coagulation, F1+2 and TAT complexes, were also not affected significantly by the type of treatment (parallelism test (p): 0.083 and 0.978, respectively). No significant changes were detected in either of the two factors during the 48-hour treatment period (horizontality test (p) >= 0.05). There were also no significant hemostatic differences between the two groups with regard to these two factors (identity test (p): 0.161 and 0.266, respectively).

Discussion

In this study, the significant decrease in PAI-1 activity over 48 hours of CRRT (horizontality test (p) = 0.012) and also the tendency to decrease PAI-1 antigen, independently of the type of CRRT, are the most remarkable results. Furthermore, this decrease is more remarkable taking into account the short time of treatment with CRRT (48 h). Perhaps a longer CRRT time would have diminished the PAI-1 levels even more and consequently, the final values of PAI-1 would not have been very far from normal values. High levels of this factor have been associated with a higher mortality in patients with sepsis or MODS (30). In our case, to obtain a control group with similar characteristics (severe SIRS) but not requiring CRRT, is very difficult, and consequently, we are unable to know if the PAI-1 decrease was due to the CRRT and/or the SIRS spontaneous evolution. However, if we consider that 87.5% of our patients developed MODS in the following days (range:1-3 days) and PAI-1 increases in this syndrome have been described previously (30), a possible effect of the CRRT in the PAI-1 levels cannot be discarded. During CRRT the predominant mass transfer mechanism is different for each type of solute. Small solute (<300 Da) removal can occur by convection in CVVH or by approximately equal contributions of both diffusion and convection in CVVHDF (31). For middle-sized molecules (300-5000 Da) convection is more important than diffusion. Regarding low-molecular weight proteins (5000-50000 Da), such as many cytokines and complement pathway products which are increased in SIRS, convection is considered the most important removal mechanism. However, absorptive removal of these molecules has also been described (32). Finally, large molecules (>50000 Da) can only be filtrated using high-flux membranes or under certain temperature and pH conditions, which change membrane permeability (13, 33). Taking into account all these considerations several mechanisms could be proposed to explain the possible relationship we have found between CRRT and the PAI-1 decrease. One could be the decrease of PAI-1 in relation to the removal of certains cytokines, such as TNF-* and IL-1 (34). The removal of these two cytokines in CRRT has been described, although significant changes in their plasma levels have not been demonstrated in all cases (32, 35). On the other hand, a clearance of PAI-1 (molecular weight of approximately 50000 Da) (36) by convection and/or by adsorption could be proposed. Finally, as PAI-1 circulates in three main forms (60% inactive, 20% active and 20% bound to t-PA) it could be proposed that the relative increase in t-PA which was observed during the treatment could have displaced the active form of PAI-1 into a t-PA-PAI-1 complex. In this study, we did not find any significant difference in the changes of PAI levels between CVVH and CVVHDF. Although it may be due to the short time (24 h) of each CRRT technique, some of the previous studies have found no different effects on the removal of some inflammatory mediators in relation to CVVH or CVVHD (37). In addition, in our case we have compared CVVH and CVVHDF, which also include diffusion and convection. The lack of difference in the PAI-1 levels between these two techniques would allow us to choose either of these two CRRT modalities based only on the patient's characteristics. (38).
The other factors related to endothelial activation (vWF, thrombomodulin and t-PA) and to fibrinoformation (F1+2 and TAT complexes) were not affected significantly either by treatment order (parallelism test (p) >= 0.05) or the course of time (horizontality test (p) >= 0.05). To interpret the small changes observed during the 48 hours of CRRT, two possible explanations may be considered. The effect of blood contact with the extracorporeal circuit and the possible removal of these factors or their regulatory mediators (39). It seems that blood contact with the extracorporeal circuit may activate the thrombus formation through the intrinsic coagulation pathway and platelet activation. However these two mechanisms are not quantitatively important either in intermittent hemodialysis or in continuous therapy (10). On the contrary, there is no doubt that such a contact modifies endothelial function, thereby stimulating the release of factors that not only favour thrombotic activity but also fibronolysis, the third system involved in hemostasis (7, 11, 40).
The non-significant increase in vWF, thrombomodulin and t-PA antigen during the CRRT treatment, expresses persistence or even an increase in endothelial injury (41-44). This could be due to its own systemic inflammatory activity (45) as well as the effect of blood contact with the different surfaces of the circuit (40). In fact, increases in these factors have been described either in hemodialysis when cellulose membranes are used or in CRRT with more biocompatible membranes (35, 40). Regarding the non-significant increase in t-PA antigen during the first 24 hours of CRRT, it may be an effect of the release of PAF and proteases. Both of them are released by stimulated platelets and leukocytes upon passing through the extracorporeal circuit (46, 47). Other possibilities are the complement activation, the hypoxia described at the start of hemodialysis in some patients or the effect of heparin (40, 48).
The big differences found in baseline values for thrombomodulin, PAI-1 and t-PA between groups were not related to the type of ARF. ATN was present in 10 patients of group I and 13 of group II and in both groups 14 patients were oliguric. Important increases in thrombomodulin related to the number of organic failures or multiple vascular pathology have been described (49, 50). In our case, the mean value of the Multiple Organ System Failure Score of the Tran group (16) was similar for both groups (6.3 vs 6.4). However, one patient in group I had been treated surgically for a multiple vascular pathology and the thrombomodulin basal value was significantly higher (265 ng/mL) than the mean value of group I (98 ng/mL). This case could explain the baseline differences of mean thrombomodulin between group I and II. In relation to the t-PA, the significant basal differences could be attributed to the number of patients with hepatic dysfunction and, consequently, with the decrease of their hepatic metabolism. In group I, there was one patient whose t-PA basal value was 80 ng/mL . However, in group II there were four patients with very high t-PA basal values. This would explain the higher mean value of t-PA in group II than in group I.
Regarding the parameters of activated coagulation, neither F1+2 nor TAT complexes were altered significantly during the 48 hours of CRRT. Previous results in patients with DIC-associated ARF have shown a non-significant increase in TAT complexes at the beginning of the extracorporeal purification technique related to low anticoagulation of the circuit (51). In our study the small increase observed after 24 hours (p >= 0.05) of CRRT could be related to the impact of non-heparinization of the system in 13 patients who had a high risk of bleeding and/or an activated coagulation following initial blood-circuit contact. However the TAT complex values at 24 and 48 h of treatment in the heparinized and non-heparinized patients were not different.
In summary, there is a significant decline in PAI-1 activity in patients with ARF associated to severe SIRS and treated with CRRT for 48 h. The decrease is observed both during the CVVH and CVVHDF. Considering the important prognostic value of this parameter (30) and the high survival rate (52.5%) in spite of 87.5% of our patients with MODS, further studies should confirm these results and investigate the implicated mechanisms. Perhaps, these data could help resolve the controversy about the type of renal replacement therapy in patients with ARF in the intensive care units (52). The increments of von Willebrand factor, thrombomodulin and t-PA antigen observed as a result of severe SIRS evolution may be an expression of endothelial injury.