Abstract
High volume hemofiltration (HVHF) (200 ml/kg/h) improves hemodynamics in experimental septic shock but is difficult to apply clinically. Accordingly, we studied whether less intensive HVHF (80 ml/kg/h) can still improve hemodynamics in experimental septic shock. We also investigated its effect on the serum concentrations of several inflammatory mediators, including endothelin (ET-1), endotoxin (LPS), tumor necrosis factor- α (TNF- α ), and 6-keto prostaglandin F1 α (6-kepto PGF1 α). Sixteen anesthetized dogs were connected to a continuous veno-venous hemofiltration (CVVH) (filtration: 80 ml/kg/h) or sham circuit and endotoxin (0.5 mg/kg) was infused intravenously over 5 min. Hemodynamic variables were measured at baseline and at 15, 45, 90, and 180 min. The major hemodynamic finding was that endotoxin-induced hypotension was significantly attenuated by intensive CVVH (p < 0.04). Changes in cardiac output and right ventricular ejection fraction were equal in both groups. ET-1 levels, but not LPS, TNF- α , or 6-keto PGF1 α, were lower during CVVH (p = 0.042). Endotoxin or TNF- α were not found in the ultrafiltrate. Median clearances of ET-1 and 6-keto PGF1 α during intensive CVVH were 8.8 and 25.9 ml/m, respectively. We conclude that intensive CVVH attenuates the early component of endotoxin-induced hypotension and reduces serum concentrations of endothelin-1. The effect of CVVH on blood pressure is not explained by convective clearance of the mediators in question.
Continuous hemofiltration is now a common technique of renal replacement therapy in critically ill patients (1-3). Its choice in preference to intermittent hemodialysis (IHD) is supported by the several clinical advantages associated with its use (4-6). Over the last few years, animal experiments and human studies have shown that hemofiltration (HF) may remove some of the soluble inflammatory mediators of sepsis (7-9). Furthermore, these experiments have shown that it can attenuate the severity of the response to sepsis (10, 11). These findings suggest that it may have a role as an adjuvant treatment for septic shock. They also suggest that one of the mechanisms for the beneficial effect of HF in sepsis may be the convective clearance of soluble inflammatory mediators (12). The advantage of using convective clearance (hemofiltration) rather than diffusive clearance (hemodialysis) for blood purification during sepsis has been previously emphasized (13). It lies in the fact that most of the soluble mediators of the inflammatory response are of middle molecular weight (> 500 daltons and < 60,000 daltons). The movement of such solutes through filtering membranes is significantly more efficient during convection than during diffusion. Thus, increasing convective clearance should increase blood purification effectiveness in sepsis. In fact, the strongest beneficial effects have been shown in porcine endotoxemia when convective fluid clearance was increased to 200 ml/kg/h (so-called high volume hemofiltration) (14).
The findings that high levels of convection are beneficial in experimental sepsis have implications for the treatment of human sepsis. Unfortunately, sustaining ultrafiltration rates greater than 100 ml/kg/h is logistically very difficult in humans because of the high blood flows and filtration fractions required. Such levels of plasma water exchange are also expensive because of the high cost of necessary sterile replacement fluid. Accordingly, it has become important to test whether an intermediate level of plasma water exchange (one logistically applicable in humans, but much greater than in current routine practice) could still achieve beneficial hemodynamic effects in sepsis. It has also become important to test how such an effect can be explained; the convective removal of one or of some of the various groups of inflammatory mediators being a possibility. It is also important to establish whether the release and blood levels of mediators are still affected (through yet undetermined mechanisms) even when convective solute removal appears insufficient to account for any such changes. To study these issues, we performed a controlled animal experiment. We tested whether intensive continuous veno-venous hemofiltration (CVVH), at an ultrafiltration rate of approximately 80 ml/kg/h, can attenuate the systemic hemodynamic response to endotoxemia and whether it can decrease the blood concentration of soluble inflammatory mediators of different molecular size. Our findings demonstrate that intensive CVVH attenuates endotoxin-induced hypotension and decreases the blood levels of endothelin-1.
The study was approved by the Animal Care and Use Committee of the University of Pittsburgh Medical Center. After a 24-h fast, 16 male mongrel dogs weighing 19 to 22 kg were anesthetized intravenously with pentobarbital sodium (30 mg/kg). Each animal was intubated with a 9-F cuffed endotracheal tube and ventilated (Siemens-Servo 900B; Siemens, Selan, Sweden) at a tidal volume of 12 ml/kg and a frequency sufficient to maintain an arterial PaCO2 between 36 and 44 mm Hg. The adequacy of ventilation was monitored by end-tidal CO2 measures (Hewlett-Packard, Palo Alto, CA). Arterial blood gases were periodically sampled, and arterial pH was maintained between 7.35 and 7.45 by adjustments in ventilation. A 7-F balloon-tipped right ventricular ejection fraction pulmonary artery thermodilution catheter with a 15-cm proximal port (Edwards Laboratories, Irvine, CA) was advanced into the pulmonary artery through the right external jugular vein. A 5-F catheter with multiple holes was inserted into the right femoral artery and advanced into the abdominal aorta for measurement of arterial pressure. A 7-F polyvinyl chloride catheter was inserted into the right femoral vein for continuous infusion of pentobarbital at 2 to 4 mg/kg/h and sampling of blood. An 8.5-F pulmonary artery catheter sheath (Edwards) was inserted into the right femoral vein to provide vascular access for CVVH. A second 8.5-F pulmonary artery catheter sheath was inserted into the left jugular vein and provided the return limb of the veno-venous circuit. A high fidelity Millar pressure catheter and a conductance catheter were also inserted into the right carotid artery and advanced into the left ventricle under fluoroscopic guidance for the purpose of an associated study. Throughout the study, body temperature was maintained by use of a thermal pad, and core temperature was monitored via the pulmonary artery catheter.
Mean arterial pressures (Pa) and mean pulmonary artery pressures (Ppa) were derived from the aortic and pulmonary arterial wave forms, respectively. Pulmonary artery occlusion pressure (Ppao) was taken as the end-expiratory balloon occlusion pulmonary arterial pressure. Cardiac output (CO) and right ventricular ejection fraction (EF) were calculated as the mean values of five thermal injection maneuvers using saline at 4° C after inspection of the thermal decay curves for artifacts. Stroke volume (SV) was calculated as the ratio of CO to heart rate. Right ventricular end-diastolic volume (RVEDV) was calculated as the ratio of SV to EF. Left and right ventricular stroke work (SWLV and SWRV) was calculated as the product of SV and either the difference between Pa and Ppao or Ppa and central venous pressure, respectively. Calculated oxygen delivery (Do 2) was determined as the product of the arterial oxygen content, hemoglobin, CO, and 1.36. All these values were either measured or calculated for each step of the protocol as described below.
Some understanding of the principles of hemofiltration is necessary to appreciate the design and findings of this study. All water-soluble substances can be theoretically removed by hemofiltration. Lipid-soluble substances cannot. Hemofiltration removes water-soluble substances by the simple process of plasma water exchange. Plasma water is filtered through the membrane by means of the pressure generated by blood flow and solutes move with it (solvent drag). This process is exactly the same as that seen in the glomerulus. However, the membranes used for hemofiltration are generally less “leaky” than the glomerulus and typically only allow very limited passage of molecules > 20 to 30,000 daltons and do not allow albumin filtration. Thus, as solute is removable if free (not bound to protein) and of sufficiently small size, for solutes that are protein bound, only the free fraction is available for filtration. For each solute there is a “sieving coefficient,” which is a measure of its ability to cross the membrane. If the concentration of a solute is the same in the ultrafiltrate as it is in plasma water the sieving coefficient (S) is said to be 1 (e.g., urea). A large protein like albumin would have a sieving coefficient of 0. For small peptides such as ET-1 the sieving coefficient will be determined by the molecular size, electrical charge (the membrane is negatively charged and will repel negatively charged molecules), degree and tightness of protein binding, degree of membrane “fouling” and, in part, intensity of transmembrane pressure. The clearance of a given solute during hemofiltration is easily calculated as would be the case for a normal kidney using plasma concentration but substituting ultrafiltrate (UF) generation for urine output and UF concentration for urine concentration. Technically, hemofiltration is very simple and only requires blood flow and the appropriate membrane. The plasma water lost as UF needs to be replaced with a solution containing appropriate electrolyte concentrations. Removal of solutes with unicompartmental distribution follows first-order kinetics. The final effect of hemofiltration on the circulating concentration of a given solute will depend on the intensity of plasma water exchange over time, the sieving coefficient of that solute, its volume of distribution, its ability to be adsorbed to the membrane, and the generation rate of the solute at tissue level.
The CVVH circuit consisted of standard plastic tubing used for human CVVH (Hospal, Lyon, France). Such tubing was directly connected to a pulmonary artery catheter sheath placed in the femoral vein (outflow). Blood was pumped from the outflow catheter by the roller pump of a blood flow/air trap module (AK 10; Gambro, Lund, Sweden). It was delivered to a polyacrylonitrile AN69 filter (Multiflow 60; Hospal) 0.45 m2 in membrane surface. After traversing the filter, blood was returned to another sheath used as the return (inflow) vascular limb of the veno-venous circuit. Blood flow was set at 200 ml/min. Ultrafiltration was maintained at 1,750 ml/h through the duration of the experiment. This rate was achieved by connecting the ultrafiltration port of the filter with standard tubing used for intravenous volumetric pumps and connecting such tubing to a double-channel intravenous volumetric pump (Flo-guard 6300; Baxter, Deerfield, IL) set at 1,750 ml/h. Replacement fluid for the planned isovolemic plasma water exchange was delivered at the same rate via a second double-channel volumetric pump. The replacement fluid was prepared to reflect the electrolyte and buffer composition of plasma water. It contained the following solutes: Na at 135 mEq/L, HCO3 at 20 mEq/L, CI at 117 mEq/L, K at 5 mEq/L, PO4 at 5 mEq/L, Mg at 2 mEq/L, SO4 at 2 mEq/L, and Ca at 2 mEq/L. The fluid was prepared immediately before use to avoid precipitation of calcium with bicarbonate. Replacement fluid was delivered through a side port of the circuit in the prefilter, preroller pump position. Prior to connection, the circuit was primed with heparinized saline (5,000 IU of heparin in 1,000 ml of normal saline recirculated through the circuit for approximately 1 h). During its period of operation the circuit was anticoagulated by means of the prefilter administration of heparin at 1,500 IU/h. The circuit was connected to the animal and CVVH initiated and maintained for at least 10 min prior to the infusion of endotoxin.
The blood path of the sham circuit was exactly the same as that used for CVVH. The only difference was that the polyacrylonitrile hemofilter was replaced by a piece of plastic tubing of the same length. The circuit was primed and blood was pumped and anticoagulated in the same fashion as for CVVH. Because no UF was generated, no replacement fluid was administered.
After line insertion, once the animals had been hemodynamically stable for a sufficient period of time (heart rate change, blood pressure change, and cardiac output change < 10% during a 30-min period of monitoring), they were connected to their operating circuit for at least 10 min. Once the circuit was operative and the animal remained stable, blood samples were obtained for baseline measurements of mediators. Hemodynamic measurements were also performed at this time. Escherichia coli endotoxin (L-2880 lipopolysaccharide; Sigma Chemical, St. Louis, MO) was administered as a slow bolus over 5 min via the right atrial port at a dose of 0.5 mg/kg, and the animal's hemodynamic response was observed. Hemodynamic measurements were obtained again at 15, 45, 90, and 180 min after endotoxin administration. Blood was also collected from all animals at 15, 45, 90, and 180 min after endotoxin for the measurement of mediators. In those animals receiving CVVH, ultrafiltrate was simultaneously obtained for the calculation of mediator clearance. According to protocol, animals were maintained on the same setting of mechanical ventilation and pentobarbital anesthesia during the study and did not receive a constant infusion of fluids. However, a bolus of 200 ml of normal saline was administered during any episode of profound hypotension (Pa < 40 mm Hg) to prevent cardiovascular collapse and death. The total amount of fluid given for this purpose was recorded for each animal (see Results).
Blood was collected in chilled tubes containing EDTA (1 mg/ml) and centrifuged immediately at 3,000 rpm for 10 min. The supernatant was separated and placed in Eppendorf tubes, which were frozen at −70° C for later analysis. Blood obtained for 6-keto PGF1α analysis was collected in tubes containing acetylsalicylic acid to impede further generation of the analyte. Samples collected for the analysis of endotoxin levels were collected and stored in pyrogen-free glass tubes. Ultrafiltrate was collected and immediately stored at −70° C without any further processing. ET-1 was extracted from the plasma, and its concentration was determined by radioimmunoassay as described previously (15, 16). According to the manufacturer (Peninsula Labs, Inc., Belmount, CA), cross-reactivity of the anti-ET-1 antibody towards ET-1, ET-2, and ET-3 is 100, 7, and 7%, respectively. Endotoxin was measured by chromogenic assay (Biowhittaker, Walkerville, MD). For 6-keto-PGF1α, an ELISA (Cayman Chemical, Ann Arbor, MI) was used, and TNF-α was measured by ELISA (Quantakine; R&D Systems, Minneapolis, MN).
Intragroup comparisons over time were performed using nonparametric analysis of variance (Friedman's test) followed by pairwise comparisons when indicated using Wilcoxon's signed rank test. Correlations between the concentration of soluble mediators and hemodynamic variables were assessed using Spearman's test. Overall intergroup comparisons of the hemodynamic response to endotoxin and of the blood concentration of various mediators were performed using the area under the curve (AUC) method (17). Intergroup comparisons of blood pressure at given time points were performed using the Mann-Whitney U test. Results are presented as medians with ranges when reflective of nonparametric data and as means with standard error of the mean when reflective of parametric data. A p < 0.05 was considered statistically significant.
The administration of endotoxin-induced hypotension in both groups of animals (p < 0.02). Hypotension, however, was significantly less severe in the animals treated with intensive CVVH (median overall change in mean arterial pressure: −22.5 mm Hg with CVVH versus −35 mm Hg with sham treatment; p < 0.04). The beneficial effect of CVVH on the hypotensive effect of endotoxemia, was particularly striking in the first 15 to 45 min after the administration of endotoxin (Figure 1). The need for fluid administration to avoid cardiovascular collapse was similar in both groups: 184 ± 233 ml/animal (sham) versus 125 ± 185 ml (CVVH); NS. Cardiac output decreased (p < 0.03) during endotoxemia (hypodynamic septic shock). The decrease in cardiac output, however, was similar in both groups. Measures of left and right ventricular performance were significantly and adversely affected in both groups without significant differences between the two groups. The hemodynamic changes induced by endotoxin in both groups of animals are summarized in Table 1.
Fig. 1. Box plot illustrating the changes in mean arterial pressure (MAP) during endotoxemia in animals treated with CVVH (shaded bars) or sham-therapy (open bars) over time (in minutes). Each bar outside the box describes the 90th and 10th percentile value. Circles refer to values outside such limits. The midline in each box represents the median value. As can be seen, the MAP is significantly higher during CVVH throughout the study (p < 0.05) and particularly so at 15 and 45 min (p < 0.05).
[More] [Minimize]| Time (min) | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 15 | 45 | 90 | 180 | ||||||
| MAP, mm Hg | ||||||||||
| CVVH* | 95 (65–125)† | 100 (58–130) | 60 (30–110) | 55 (30–90) | 65 (30–95) | |||||
| Sham | 100 (65–130)‡ | 65 (40–70) | 45 (30–70) | 35 (20–55) | 55 (20–65) | |||||
| MPAP, mm Hg | ||||||||||
| CVVH | 10 (5–12.5) | 10 (7.5–12) | 10 (7–12) | 7.5 (6–10) | 8 (6–12) | |||||
| Sham | 12 (7–15) | 11 (6–15) | 8 (5–13) | 8 (15–13) | 8 (5–13) | |||||
| SVR, dynes · cm5 | ||||||||||
| CVVH | 3,312 (1,722–5,514) | 3,920 (1,982–6,257) | 2,530 (1,225–6,142) | 2,795 (1,235–5,056) | 4,497 (2,088–9,673) | |||||
| Sham | 2,478 (1,155–5,387) | 1,931 (955–3,868) | 1,754 (1,046–3,605) | 1,288 (954–2,181) | 2,333 (1,290–3,876) | |||||
| RAP, mm Hg | ||||||||||
| CVVH | 2.5 (0–5) | 2 (0–2.5) | 2 (0–2.5) | 1 (0–1) | 1 (0–3) | |||||
| Sham | 2.5 (1–7) | 1.5 (0–7) | 2 (1–3) | 2 (0–7) | 2 (0–7) | |||||
| PAOP, mm Hg | ||||||||||
| CVVH | 2.5 (−1 to 5) | 2 (−4 to 3) | 2 (−4 to 4) | 1 (−2 to 5) | 2 (0–4) | |||||
| Sham | 3 (1–4) | 1 (−3 to 7) | 1 (0–3) | 2 (0–7) | 1.5 (0–7) | |||||
| HR, bpm | ||||||||||
| CVVH | 100 (60–160) | 130 (110–200) | 130 (90–200) | 130 (100–200) | 130 (100–170) | |||||
| Sham | 125 (60–140)§ | 120 (110–200) | 140 (120–200) | 125 (110–160) | 125 (90–160) | |||||
| CO, L/m | ||||||||||
| CVVH | 2.47 (1.56–3.6)‖ | 1.72 (1.41–2.26) | 1.86 (1.38–2.38) | 1.25 (1.02–2.72) | 1.04 (0.8–1.8) | |||||
| Sham | 3.2 (1.47–5.2) | 2.13 (1.71–2.7) | 1.95 (1.42–2.6) | 1.65 (1.09–2.43) | 1.3 (0.9–1.9) | |||||
| RVSW, g-m/min | ||||||||||
| CVVH | 2.4 (0.7–4.6)¶ | 1.7 (0.7–2.5) | 1.4 (0.5–2.4) | 1.0 (0.4–2.2) | 0.8 (0.25–1.2) | |||||
| Sham | 3 (1.3–5.5)** | 1.9 (1–2.5) | 1.2 (0.5–2.3) | 1.0 (0.5–2.1) | 0.8 (0.5–1.5) | |||||
| LVSW, g-m/min | ||||||||||
| CVVH | 32 (18–47)1-164 | 16 (12–20) | 10 (6–16) | 8 (4–13) | 7 (2–14) | |||||
| Sham | 35 (17–48)‡ | 14 (8–22) | 9 (5–16) | 6 (2–10) | 7 (3–10) | |||||
| EF, % | ||||||||||
| CVVH | 21 (14–33)§§ | 21 (17–34) | 25.7 (13.3–32) | 18 (9.7–27.3) | 14 (9.3–17.3) | |||||
| Sham | 33.3 (21.7–46) | 25.3 (17.3–33.3) | 24.3 (17.3–36.7) | 25 (17–33.3) | 20 (15–48) | |||||
| RVEDV, ml | ||||||||||
| CVVH | 122 (51–153)1-166 | 68 (35–127) | 62 (40–123) | 62 (7–139) | 48 (20–89) | |||||
| Sham | 97 (74–131)1-166 | 68 (32–106) | 57 (33–80) | 55 (34–75) | 58 (27–74) | |||||
| SV, ml | ||||||||||
| CVVH | 26.4 (11–45)¶¶ | 13.7 (9.7–17.4) | 13.2 (8–16.4) | 11.5 (6–22.7) | 9.1 (4.5–18.3) | |||||
| Sham | 29.8 (17.7–40.3)1-165 | 16.6 (10.6–24.7) | 13.9 (10.1–18.9) | 13.5 (9.9–18.7) | 11.6 (9.8–13.5) | |||||
| cDO2, ml/m | ||||||||||
| CVVH | 403 (254–533)1-170 | 368 (228–573) | 350 (240–497) | 319 (207–424) | 197 (167–305) | |||||
| Sham | 456 (279–702) | 305 (253–377) | 250 (241–409) | 277 (150–357) | 217 (97–284) | |||||
Endotoxin administration induced an increase in serum endotoxin levels (p < 0.005) and in serum ET-1 concentrations (p < 0.005). There were no changes in 6-keto PGF1α and TNF-α levels, which were already abnormally high before the infusion of endotoxin (Table 2). The increase in ET-1 levels was significantly less in CVVH-treated animals (p = 0.042) (Figure 2). No such differences were present for the other three marker molecules. TNF-α and endotoxin could not be detected in the ultrafiltrate. ET-1 and 6-keto PGF1α were detected in the ultrafiltrate (Figure 2). The median sieving coefficient for ET-1 was 0.30 (0.1 to 0.7) and that for 6-keto PGF1α was 0.88 (0.1 to 1.0). Such sieving coefficients resulted in a median ET-1 clearance of 8.8 ml/m (1.5 to 20.8) and in a median 6-keto PGF1α clearance of 25.9 ml/m (2.5 to 32.7). Across both groups, endotoxin levels were positively correlated with ET-1 levels (r = 0.355; p = 0.0027) and negatively correlated with SWLV (r = −0.246; p = 0.039) and SVLV (r = −0.285; p = 0.0172). ET-1 was negatively correlated with many indices of left-sided myocardial performance including SV, SWLV, and CO (p < 0.0007 for all indices) (Figure 3) as well as SWRV (r = −0.562; p < 0.0001) (Figure 4) and Pa (r = −0.4; p < 0.0005).
| Time (min) | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 15 | 45 | 90 | 180 | ||||||
| ET-1, pg/ml | ||||||||||
| CVVH* | 10.8 (9.5–12.9)† | 14.9 (10.7–7.8) | 24.1 (11.9–183) | 40 (15.5–77.5) | 34 (12.8–154) | |||||
| Sham | 11.2 (9.4–15.7)‡ | 14.1 (11–13) | 43.5 (15.8–110) | 58 (36.5–42) | 94 (40.6–187) | |||||
| LPS, U/ml | ||||||||||
| CVVH | 1 (0.2–2,300)§ | 1,532 (1,287–5,000) | 1,511 (69–5,000) | 772 (273–5,000) | 637 (15–5,000) | |||||
| Sham | 0.6 (0.4–1.6)‖ | 2,244 (508–5,000) | 1,430 (525–5,000) | 1,680 (362–1,945) | 510 (182–1,651) | |||||
| TNF, pg/ml | ||||||||||
| CVVH | 12,385 (852–40,000) | 13,962 (1,176–40,000) | 14,574 (3,637–23,566) | 14,570 (738–23,255) | 11,627 (781–18,030) | |||||
| Sham | 2,118 (344–40,000) | 2,736 (886–40,000) | 2,762 (994–40,000) | 2,835 (448–40,000) | 1,276 (542–40,000) | |||||
| 6-keto PGF, pg/ml | ||||||||||
| CVVH | 317 (0–900) | 357 (329–1,157) | 459 (332–1,153) | 369 (354–1,156) | 470 (345–1,132) | |||||
| Sham | 374 (322–1,056) | 399 (333–1,015) | 383 (339–1,015) | 381 (333–1,165) | 479 (318–1,122) | |||||
Fig. 2. Area plots for 6 keto-PGF1α (left panel ) and ET-1 (right panel ) levels during endotoxemia in animals treated with CVVH (middle curves) or sham therapy (rear curves) over time (in minutes). The front curves represent ultrafiltrate levels over the same time period. ET-1 levels are significantly higher (p < 0.05) during sham treatment, particularly after 3 h of CVVH despite the low levels of ET-1 appearing in the ultrafiltrate (sieving coefficient: 0.3). By contrast, 6 keto-PGF1α levels are not different between CVVH and sham despite large amounts appearing in the ultrafiltrate (sieving coefficient: 0.88).
[More] [Minimize]
Fig. 3. Bivariate plot showing the relationship between endothelin-1 levels (ET-1) and left ventricular stroke work (LVSW). (Spearman's r = −0.633; p < 0.0001). The higher the ET-1 levels the lower the left ventricular performance.
[More] [Minimize]
Fig. 4. Bivariate plot showing the relationship between endothelin-1 levels (ET-1) and right ventricular stroke work (RVSW). (Spearman's r = −0.562; p < 0.0001). The higher the ET-1 levels the lower the right ventricular performance.
[More] [Minimize]During the last decade, there has been much interest in the benefits of blood purification by HF in sepsis (18). The biologic rationale for such interest derives from knowledge that some of the systemic effects of sepsis may be mediated by endogenous, water-soluble, small (< 500 daltons) and middle (> 500 but < 60,000 daltons) molecules and that the removal of such molecules from the bloodstream may be desirable. This view is supported by animal experiments that have shown beneficial effects of HF in the setting of bacteremia or endotoxemia (10). CVVH is a technique of hemofiltration that is based on the principle of convective plasma water exchange. CVVH has been shown to clear some of the putative mediators of sepsis (19, 20) and to affect circulating levels of TNF-α (12). In several animal models of sepsis, it has achieved beneficial effects on cardiac performance and pulmonary function (21, 22). Such effects, however, although statistically significant, have been clinically modest when low ultrafiltration rates typical of renal replacement therapy (10 to 15 ml/kg/h) are used. It is logical to believe that, if modest levels of plasma water exchange show a benefit, increasing the intensity of such exchange could increase the benefits of CVVH in sepsis.
To test this hypothesis, Grootendorst and colleagues (14) recently studied the impact of high volume hemofiltration (ultrafiltration at 200 ml/kg/h) in a model of porcine endotoxemia. They showed a major attenuation of endotoxin-induced hypotension and an improvement in cardiac performance. Such results reinforce the notion that the intensity of convective plasma water exchange during hemofiltration is important in achieving clinically significant goals. However, implementing and sustaining ultrafiltration rates of close to 200 ml/kg/h in human beings is difficult. This difficulty is due to the need for high blood flows (up to 500 ml/m depending on the degree of predilution) and high filtration fractions (30 to 40% depending on the degree of predilution and blood flow rate). In addition, the cost of administering such large amounts of replacement fluid is high: at approximately $2.00 per liter and a plasma water exchange rate of 200 ml/kg/h, such cost would be $32.00 an hour in an 80-kg patient! It is, therefore, important to establish whether less aggressive plasma water exchange (approximately half way between current renal replacement therapy levels and HVHF) would still have beneficial effects during endotoxemia. If so, a stronger case could be made for more intensive hemofiltration in patients with acute renal failure and severe sepsis. Furthermore, such treatment would be logistically possible and financially acceptable. We sought to address these issues by means of a controlled animal study in which a septic state had been induced by the administration of endotoxin. The results of our study demonstrate that intensive CVVH retains statistically significant and clinically meaningful beneficial effects on blood pressure during experimental endotoxemia. These effects were particularly striking in the first 15 to 45 min after exposure to endotoxin and would be particularly relevant to patients who, while receiving treatment with CVVH, were exposed to endotoxin. We were unable to replicate the beneficial effects on right ventricular performance and cardiac output obtained with high volume hemofiltration in the studies by Grootendorst and colleagues (14). There are several possible explanations for this difference. High volume hemofiltration levels of plasma water exchange may be needed to adequately attenuate myocardial depression. The different membrane (polyamide) used by Grootendorst and colleagues may more effectively remove myocardial depressant substances or, finally, species-related differences may account for the different response.
We also measured the blood and ultrafiltrate concentration of four “marker” molecules to better understand the effect of intensive hemofiltration on inflammatory mediators and their relation to the hemodynamic response. We chose ET-1 as a molecule representing small peptides, TNF-α for larger middle molecules (MW: between 17,000 daltons for monomer and 54,000 daltons for trimer) and 6-keto PGF1α as a molecule representing the eicosanoid family (MW: 348). Finally, we measured endotoxin, a very large molecule (> 1 million daltons), to determine whether its adsorption to the membrane may explain the differences between the two groups. Our choice was aimed at spanning the molecular weight range of putative inflammatory mediators. Another rationale was to study two vasoactive peptides (ET-1 as a potent vasoconstrictor and 6-keto PGF1α, a metabolite of prostacyclin, as a marker of vasodilator activity). It was hoped that by examining the differential effects of CVVH on these vasoactive substances, the hemodynamic effects of CVVH could be better understood.
We found that serum ET-1 increased after endotoxin and that ET-1 was removed by intensive CVVH (mean sieving coefficient of 0.3; median ET-1 clearance of 8.8 ml/m). In association with such removal, serum ET-1 concentrations were significantly lower than in the sham-treated group. Because of the low clearances, however, the effect of CVVH on ET-1 serum levels is not fully explained by convective removal. Other mechanisms (adsorption to the membrane or an effect on other unmeasured mediators with consequent decrease in ET-1 generation rate) have to be considered. If membrane adsorption had taken place, we would have expected a progressive increase over time in the ultrafiltrate concentration of ET-1 and/or a maximal decrease in levels early in the course of CVVH. We did not find such suggestive evidence. Our findings are more consistent with an effect of CVVH on the inflammatory response proximal to ET-1 generation and release. Such proximal inflammatory down-modulation would then decrease ET-1 release from the endothelium and lower its serum concentration. In keeping with this hypothesis, recent work by Kellum and colleagues (12) showed that TNF-α levels can be decreased by CVVH in septic patients even in the absence of TNF-α filtration. ET-1 concentrations were negatively correlated with most indices of cardiac function and with mean arterial blood pressure and positively correlated with endotoxin blood levels. ET-1 is now considered a polyfunctional cytokine with a proinflammatory effect. ET-1 levels increase 3- to 4-fold in human and experimental septic shock (23, 24) and its synthesis is stimulated by endotoxin both in vitro and in vivo (25). Its hemodynamic effects, however, are complex as shown by recent experimental studies using receptor antagonists (26, 27) and may be species-specific. The most commonly reported findings are increased pulmonary artery pressures (not found in our sham-treated animals), increased blood pressure (the opposite of what was seen in our model), decreased coronary artery blood flow, and a variable effect on cardiac output. In addition, ET-1 release triggers prostaglandin and nitric oxide release with counterbalancing effects on hemodynamic variables, making the interpretation of the moderate decrease in ET-1 levels achieved even more complex.
Intensive CVVH cleared 6-keto PGF1α from the bloodstream more effectively than ET-1, as would be expected given the molecular weight of the two substances. 6-keto PGF1α is a smaller molecule of the eicosanoid group. It was chosen to represent the fate of such molecules as thromboxane, other eicosanoids, and leukotrienes during intensive CVVH. Despite a mean 6-keto PGF1α clearance of 22.3 ml/m, its serum levels were unchanged during CVVH, suggesting that such clearance is insufficient to lower blood levels of eicosanoids as recently confirmed by another group (28). No correlation was found between 6-keto PGF1α levels and other mediators or hemodynamic variables.
Serum TNF-α levels were high throughout the study, including baseline observations, and higher than those seen in humans with severe sepsis (12). Such levels suggest that instrumentation had induced a degree of inflammatory response. They may also represent a cross-reaction of the human ELISA-specific antibody with other epitopes in the dog yielding falsely elevated concentrations. No difference in serum TNF-α concentration was demonstrated between the two groups. TNF-α has been detected in the ultrafiltrate of patients with severe sepsis, but we were unable to demonstrate it in the ultrafiltrate of our animals. Recent data, however, indicate that TNF-α filtration may be a time-dependent process: first, TNF-α binds to the membrane; then after membrane saturation, it is found in the ultrafiltrate (29). Such movement into the ultrafiltrate may only occur after 4 to 6 h of treatment. In addition, the pore size for the AN-69 membrane is normally distributed over a relatively wide range, with a mean size of approximately 20,000 daltons when tested ex vivo. Accordingly, TNF-α would be expected to cross the membrane either after fragmentation of its trimeric form (almost all of TNF-α circulates as such) into monomers or via the small number of pores that might approach the size of the trimer.
Endotoxin levels were also extraordinarily high after its administration. There were no differences in endotoxin concentration between the two groups. The beneficial effect of CVVH cannot, therefore, be explained by endotoxin adsorption to the membrane or by other mechanisms directly affecting endotoxin levels.
The early impact on blood pressure by intensive CVVH and its effects on ET-1 suggest an impact on mediator pathways, which are active in early endotoxemia. Such pathways may include a degree of activation of constitutive nitric oxide synthase (30) or the activation of the kinin system, particularly bradykinin (31). The latter system, however, is an unlikely candidate for attenuation by the AN-69 membrane because this membrane is well known to actually induce bradykinin release to clinically significant levels, especially under circumstances of decreased metabolism (32). In addition, such systems may operate only in very early sepsis and not in established sepsis where intervention is logistically possible in humans. Whatever these systems may be, they are likely to have repercussions on other biologic systems (such as ET-1). A membrane adsorption-mediated effect, which affects the early components of the inflammatory cascade, seems probable because convection could not be expected to act within 15 min. Recent data in humans, published after the completion of our experiments, support the concept that complement system activation may be decreased in sepsis through factor D membrane adsorption (19). Furthermore, platelet-activating factor (PAF) binding to polysulfone membranes has also recently been shown (33). PAF is released early in the course of sepsis and its membrane binding may be a determinant of the effect of CVVH on blood pressure.
First, we studied only one membrane. Several high flux membranes are available and have been used in prior animal experiments. We chose the AN-69 polyacrylonitrile membrane because it is used at our institution in patients. This membrane is also highly biocompatible, has strong adsorptive abilities, and has been demonstrated to facilitate cytokine removal. Second, we used only one ultrafiltration rate (80 ml/kg/h). Such an ultrafiltration rate, however, translates into a plasma exchange rate of 6 L/h in a 75-kg man (six times the intensity of standard therapy). It can be technically achieved and sustained for as long as 8 h in critically ill humans with septic shock. Third, we could not formally test the degree to which adsorption was fully responsible for our findings. We did not employ a capped ultrafiltrate port although this method has been used to study the separate effects of adsorption and convection. We did not use this methodology specifically because it has been demonstrated by Ronco and colleagues (33), with elegant membrane extraction studies, that adsorption to the membrane is dramatically increased by the process of ultrafiltration. Accordingly, adsorption in the absence of convection (capped port) would grossly underestimate the true adsorption that would occur under normal operating conditions (convective clearance). Forth, we gave only bolus endotoxin. Although this dose of endotoxin produces a syndrome similar to septic shock (vasodilatation, capillary leak, metabolic alterations, and so on), it is sublethal in the majority of cases and shows several differences from human sepsis. In human sepsis inflammation develops more slowly and is more sustained than that seen with bolus endotoxin infusion. Furthermore, blood purification strategies are instituted after the onset of sepsis rather than before. However, human sepsis and multiorgan dysfunction may be sustained by ongoing and recurrent inflammatory stimuli (e.g., endotoxin translocation from the gut, bacterial reinfection, etc.). The capacity of CVVH to modulate this ongoing inflammation by affecting biosynthesis of distal mediators (e.g., ET-1) is notable and may have significant implications for human sepsis. Despite such considerations, the effect of blood purification by CVVH on the autocrine and paracrine mechanisms by which cytokines act is unknown and as such, it is unclear if these effects will be beneficial for patients. Some of the improvement in blood pressure noted with CVVH has been attributed to the cooling effects of hemofiltration and subsequent vasoconstriction. In our study, body temperature was supported with a thermal regulating pad. Although temperature was not formally recorded at each time interval there was no obvious difference between the two groups in the constant pulmonary artery catheter thermistor display. Furthermore, it should be noted that the major effect of intensive CVVH seen in this study was over a short period of time immediately after endotoxin rather than increasingly over time as if secondary to worsening hypothermia.
In summary, we have studied the effects of intensive CVVH in canine endotoxemia and have found that it attenuates early hypotension in a statistically and clinically significant way. Using marker molecules, we have also found that such beneficial effect was not explained by endotoxin or TNF-α adsorption or prostaglandin removal. In fact, the timing of the hemodynamic effects and the alterations in ET-1 levels suggest an early activation system-related, adsorption-mediated mechanism of action. Convective removal of mediators may not have accounted for any of the observed hemodynamic effects of high volume CVVH. We believe that, based on the current available data, the application of hemofiltration (high volume or standard) as adjuvant treatment of sepsis or even as salvage therapy in intractable sepsis is not yet justified. However, we believe that pilot investigations are now warranted to better define the biologic and clinical effect of this approach in human sepsis and whether similar mechanisms of blood purification (adsorption but little convection) operate under such circumstances.
Supported by a Seed Grant from the Department of Anesthesiology and Critical Care Medicine, University of Pittsburgh Medical Center, and by a Grant from the Laerdal Acute Medicine Foundation, Stavanger, Norway.
| 1. | Bellomo R., Ronco C.The changing pattern of severe acute renal failure. Nephrology21998149154 |
| 2. | Silvester, W., R. Bellomo, and C. Ronco. 1998. Continuous versus intermittent renal replacement therapy in the critically ill. In C. Ronco and R. Bellomo, editors. Critical Care Nephrology, Kluwer Academic Publishers, Dordrecht. 1225–1239. |
| 3. | van Bommel E., Bouvy N. D., So K. L., Zietse R., Vincent H. H., Bruning H. A., Weimar W.Acute dialytic support for the critically ill: intermittent hemodialysis versus continuous arteriovenous hemodiafiltration. Am. J. Nephrol.151995192200 |
| 4. | Bellomo R., Ronco C.Acute renal failure in the intensive care unit: adequacy of dialysis and the case for continuous therapies. Nephrol. Dial. Transplant.111996426428 |
| 5. | Clark W. R., Mueller B. A., Alaka K. J., Macias W. L.A comparison of metabolic control by continuous and intermittent therapies in acute renal failure. J. Am. Soc. Nephrol.4199414131420 |
| 6. | Davenport A., Will E. J., Davison A. M.Effect of renal replacement therapy on patients with combined acute renal failure and fulminant hepatic failure. Kidney Int.411993S245S251 |
| 7. | Sanchez-Izquierdo Riera, J. A., J. L. Perez Vela, M. J. Lozano Quintana, E. Alted Lopez, B. Ortuno de Solo, and A. Ambros ChecaCytokine clearance during venovenous hemofiltration in the trauma patient. Am. J. Kidney Dis.301997483488 |
| 8. | van Bommel E. F. H.Should continuous renal replacement therapy be used for “non-renal” indications in critically ill patients with shock? Resuscitation331997257270 |
| 9. | Bellomo R., Tipping P., Boyce N.Continuous veno-venous hemofiltration with dialysis removes cytokines from the circulation of septic patients. Crit. Care Med.211993522526 |
| 10. | Lee P. A., Matson J. R., Pryor R. W., Hinshaw L. B.Continuous arteriovenous hemofiltration therapy for Staphylococcus aureus-induced septicemia in immature swine. Crit. Care Med.211993914924 |
| 11. | Heering P., Morgera S., Schmitz F. J., Schmitz G., Willers R., Schultheis H. P., Strauer B. E., Grabensee B.Cytokine removal and cardiovascular hemodynamics in septic patients with continuous hemofiltration. Intensive Care Med.231997288296 |
| 12. | Kellum J. A., Johnson J. P., Kramer D., Palevsky P., Brady J. J., Pinsky M. R.Diffusive versus convective therapy: effects on mediators of inflammation in patients with severe systemic inflammatory response syndrome. Crit. Care Med.26199819952000 |
| 13. | Bellomo R.Choosing a therapeutic modality: hemofiltration vs. hemodialysis versus hemodiafiltration. Semin. Dialysis919968892 |
| 14. | Grootendorst A. F., van Bommel E. F. H., van der Hoven B., van Leengoed L. A. M. G., van Osta G. A. L. M.High-volume hemofiltration improves hemodynamics of endotoxin-induced shock in the pig. J. Crit. Care719926775 |
| 15. | Gandhi C. R., Sproat L. A., Subbotin V. M.Increased hepatic endothelin-1 levels and endothelin receptor density in cirrhotic rats. Life Sci.5819965562 |
| 16. | Gandhi C. R., Kang Y., De Wolf A., Madariaga J., Aggarwal S., Scott V., Fung J.Altered endothelin homeostasis in patients undergoing liver transplantation. Liver Transpl. Surg.21996362369 |
| 17. | Matthews J. N. S., Altman D. G., Campbell M. J., Royston P.Analysis of serial measurements in medical research. B.M.J.3001990230235 |
| 18. | Bellomo R.Continuous hemofiltration as blood purification in sepsis. New Horiz.31995732737 |
| 19. | Gasche Y., Pascual M., Suter P. M., Favre H., Chevolet J. C., Schifferli J. A.Complement depletion during hemofiltration with polyacrylonitrile membranes. Nephrol. Dial. Transplant111996117119 |
| 20. | Tonnessen E., Hansen M. B., Hohndorf K., Diamants M., Bendtzen K., Wanscher M., Toft P.Cytokines in plasma and ultrafiltrate during continuous arteriovenous hemofiltration. Anaesth. Intensive Care211993752758 |
| 21. | Stein B., Pfenninger E., Grunert A., Schmitz J. E., Hudde M.Influence of continuous hemofiltration on hemodynamics and central blood volume in experimental endotoxic shock. Intensive Care Med.161990494499 |
| 22. | Stein B., Pfenninger E., Grunert A., Schmitz J. E., Deller A., Kocher F.The consequences of continuous haemofiltration on lung mechanics and extravascular lung water in a porcine endotoxic shock model. Intensive Care Med.171991293298 |
| 23. | Mitaka C., Hirata Y., Makita K., Nagura T., Tsunoda Y., Amaha K.Endothelin-1 and atrial natriuretic peptide in septic shock. Am. Heart J.1261993466468 |
| 24. | Levin E. R.Endothelins. N. Engl. J. Med.3331995356363 |
| 25. | Sagiura M., Inagami T., Kan V.Endotoxin stimulates endothelin release in vivo and in vitro as determined by radioimmunoassay. B.B.R.C.161198912201227 |
| 26. | Ruetten H., Thiemermann C., Vane J. R.Effect of the endothelin receptor antagonist, SB 209670, on circulatory failure and organ injury in endotoxic shock in the anaesthetized rat. Br. J. Pharmacol.1181996198204 |
| 27. | Weitzberg E., Hemsen A., Rudehill A., Modin A., Wanecek M., Lundberg J. M.Bosentan-improved cardiopulmonary vascular performance and increased plasma levels of endothelin-1 in porcine endotoxic shock. Br. J. Pharmacol.1181996617625 |
| 28. | Bottoms G., Fessler J., Murphey E., Johnson M., Latshaw H., Mueller B., Clark W., Macias W.Efficacy of convective removal of plasma mediators of endotoxic shock by continuous veno-venous hemofiltration. Shock51996149154 |
| 29. | van Bommel E. F. H., Hesse C. J., Jutte N. H. P. M., Zietse R., Bruining H. A., Weimar W.Cytokine kinetics (TNF-alpha, IL-1 beta, IL-6) during continuous hemofiltration: a laboratory and clinical study. Contrib. Nephrol.11619946275 |
| 30. | Szabo C., Mitchell J. A., Thiemermann C., Vane J. R.Nitric oxide-mediated hyporeactivity to noradrenaline precedes the induction of nitric oxide synthase in endotoxin shock. Br. J. Pharmacol.1081993786792 |
| 31. | Paya D., Stoelet J. C.Involvement of bradykinin and nitric oxide in the early hemodynamic effects of lipopolysaccharide in rats. Shock31995376379 |
| 32. | Schaefer R. M., Fink E., Shaefer L., Barkhausen R., Kulzer P., Heidland A.Role of bradykinin in anaphylactoid reactions during hemodialysis with AN 69 dialyzers. Am. J. Nephrol.131993473477 |
| 33. | Ronco C., Tetta C., Lupi A., Galloni E., Bettini M. C., Marian F., De Martino A., Montrucchio G., Camussi G., La Greca G.Removal of platelet activating factor in experimental continuous arteriovenous hemofiltration. Crit. Care Med.23199599107 |
