Cytokine Removal during Continuous Hemofiltration in Septic Patients

Sepsis is induced by local or systemic release of microorganism components, leading to the production of proinflammatory cytokines, such as tumor necrosis factor-α (TNF-α), interleukin-1β (IL-1β), and IL-6. They initiate an inflammatory cascade mediated by several cell lines and involving the complement, coagulation, and fibrinolytic systems (1,2). These phenomena are counterbalanced by a complex system of naturally occurring inhibitors of inflammation: Anti-inflammatory cytokines such as IL-10, soluble cytokine receptors such as soluble TNF receptors (sTNFR-I and sTNFR-II), and cytokine receptor antagonists such as IL-1 receptor antagonist (IL-1ra). Sepsis is characterized by release of excessive amounts of proinflammatory cytokines in the circulation, leading to a generalized and uncontrolled host response, overwhelming the natural inhibitors of inflammation (3).

It has been hypothesized that this excessive inflammatory response could be downregulated with the continuous renal replacement therapies (CRRT), by nonspecific extracorporeal removal of cytokines and other mediators (4). In vitro studies indicate that cytokine removal with current synthetic membranes is more convective than diffusive (5,6,7,8). Moreover, these membranes have a high adsorptive capacity for cytokines (5,6,7,8,9,10). The extracorporeal removal of inflammatory mediators has been confirmed in clinical studies (11,12,13,14,15,16,17,18,19,20), but few reported important effects on their plasma concentration (19,20). Knowledge of the relative contribution of membrane adsorption and convective elimination could lead to more efficient removal strategies, but so far a detailed quantitative assessment of adsorption and convection has not been performed in vivo. Furthermore, little is known about the concomitant removal of inhibitors of inflammation. If these are removed to the same extent as the proinflammatory cytokines, the pro/anti-inflammatory balance might remain unaffected. Finally, cytokines exert their effects at the tissue level and the significance of their presence in the circulation is undefined. Whether removal of cytokines from the systemic circulation has a beneficial effect on the short-term hemodynamic status and respiratory function of the septic patient is at present unclear.

To address these questions, a prospective trial was conducted in critically ill patients with sepsis and acute renal failure, treated with continuous venovenous hemofiltration (CVVH). The removal of selected inflammatory cytokines and inhibitors of inflammation was studied. To determine the relative importance of convective removal and membrane adsorption, mass balance studies were performed, with comparison of different flow rates and analysis of the effect of replacement of the AN69 hemofilter. In addition, the influence of CVVH on hemodynamic status and respiratory function was examined.

Materials and Methods

Results

Studies of Hemodynamic and Respiratory Function

At baseline, all patients had a hemodynamic profile, characteristic for hyperdynamic shock, with a low MABP, a high CO and a low SVR. Most patients required vasopressor (13 of 15) and/or inotropic (9 of 15) drugs. The start of CVVH was associated with a significant and sustained fall in CO and a rise in SVR (Figure 1). A trend toward a lower PAOP was observed (Figure 1). The other hemodynamic parameters did not change significantly. Central body temperature, the dopamine, dobutamine, epinephrine, and norepinephrine requirements, and the administered colloid volume remained constant throughout the study period. PaO₂/FiO₂ and positive end expiratory pressure did not change during the observation period.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Response of cardiac output (CO), systemic vascular resistance (SVR), and pulmonary artery occlusion pressure (PAOP) to the treatment with continuous venovenous hemofiltration. CO was significantly higher and SVR was significantly lower during hemofiltration than at baseline, as defined by both t = 0 and the mean of t = -4, t = -2 and t = 0. ^*P < 0.05. A trend toward a lower PAOP was observed. P = NS.

Cytokine Plasma Concentrations

TNF-α, IL-6, IL-10, IL-1ra, sTNFR-I, and sTNFR-II levels were above the detection limit in all patients. IL-1β was not detectable in 10 patients throughout the study period. The plasma concentrations (pg/ml) before the start of CVVH (t = 0) were: 68.4 ± 17.2 for TNF-α; 16.4 ± 2.8 for IL-1β; 3997 ± 2594 for IL-6; 495.5 ± 121.3 for IL-10; 13581 ± 4102 for IL-1ra; 19009 ± 2442 for sTNFR-I; and 31165 ± 2270 for sTNFR-II. The inlet plasma concentrations of the inflammatory cytokines as well as of the inhibitors of inflammation decreased significantly in the first hour after installation of a new hemofilter (t = 1 versus t = 0 and t = 13 versus t = 12) (Figure 2, A and B). No decreases of a similar consistency were observed at other time points. No changes were observed in the ratios between the serum concentrations of the soluble TNF receptors and of TNF-α and between the serum concentrations of IL-1ra and IL-1β (Figure 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Cytokine plasma concentrations expressed as a percentage of the concentration at t = 0. A significant fall in the plasma concentrations of all cytokines was observed at t = 1 and at t = 13. ^*P < 0.05. (A) Inflammatory cytokines. No statistical analysis was performed for interleukin-1β (IL-1β), since it was detected in only five patients. (B) Anti-inflammatory cytokines.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Ratios between the serum concentrations of the soluble tumor necrosis factor (TNF) receptors and of TNF-α and between the serum concentrations of IL-1 receptor antagonist (IL-1ra) and IL-1β. No significant changes in soluble TNF receptor I (sTNFR-I)/TNF-α and sTNFR-II/TNF-α were observed. No statistical analysis was performed for IL-1ra/IL-1β, since IL-1β was detected in only five patients.

Cytokine Removal

The mean ultrafiltration volume was 18.3 ± 1.6 L/12 h (25.4 ± 0.7 ml/min) with Q_B = 100 ml/min, and 31.9 ± 3.0 L/12 h (44.3 ± 1.5 ml/min) with Q_B = 200 ml/min. TNF-α, IL-6, IL-1ra, sTNFR-I, and sTNFR-II were detectable in the ultrafiltrate of all patients. IL-1β was found only in the ultrafiltrate of the patients with detectable plasma levels. IL-10 could not be detected in any of the ultrafiltrates despite high plasma levels. The overall mean SC were: 0.16 ± 0.02 for TNF-α; 0.22 ± 0.02 for IL-1β; 0.18 ± 0.01 for IL-6; 0 ± 0 for IL-10; 0.28 ± 0.03 for IL-1ra; 0.006 ± 0.0007 for sTNFR-I; and 0.003 ± 0.0001 for sTNFR-II.

The M_TR, expressed as a percentage of the M_I, was highest 1 h after the use of a new hemofilter (at t = 1 and t = 13), ranging from 43.0 ± 3.3% for IL-1ra at t = 13 to 25.3 ± 1.7% for sTNFR-II at t = 1 (Figure 4, A and B). Five hours later (at t = 6 and t = 18), total cytokine removal was already significantly lower, varying from 28.7 ± 2.7% for IL-1ra at t = 6 to 18.1 ± 1.8% for sTNFR-II at t = 6. After 12 h of membrane use (at t = 12 and t = 24), total removal ranged from 24.0 ± 2.5% for IL-1ra at t = 24 to 8.4 ± 2.8% for sTNFR II at t = 24. The contributions of convective elimination (M_UF) and membrane adsorption (M_AD) were calculated in absolute values (data not shown) and as a percentage of M_TR (Figure 4, A and B). Removal was due mainly to membrane adsorption for all inflammatory cytokines and especially for their inhibitors. Absolute M_UF remained stable over the observation period, whereas absolute M_AD was highest after installation of a new hemofilter and decreased steadily thereafter. M_AD, expressed as a percentage of M_TR, was 100% for IL-10 and nearly 100% for both soluble TNF receptors, at all time points. For the other cytokines, relative M_AD ranged from 92.6 ± 7.3% (IL-1β) to 61.0 ± 11.5% (IL-6) at t = 1 and 13, from 84.4 ± 9.5% (IL-1β) to 44.3 ± 4.1% (IL-6) at t = 6 and 18, and from 75.1 ± 9.5% (IL-1β) to 37.0 ± 9.5% (IL-6) at t = 12 and 24. At the latter two time points, each after 12 h of blood-membrane contact, a negative adsorptive removal rate was found for TNF-α, IL-1β, IL-6, sTNFR-I, and sTNFR-II in a few patients, indicating release from the membrane of previously bound cytokine.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

The total amount of cytokine removed (M_TR), expressed as a percentage of the amount present in the prefilter plasma (M_I). The removed amount at t = 6 and 18 and at t = 12 and 24 is significantly lower than at t = 1 and t = 13. ^*P < 0.05. For each value, the relative contribution of adsorption (% AD, hatched bars) and ultrafiltration (% UF, open bars) is indicated. (A) Inflammatory cytokines. No statistical analysis was performed for IL-1β, since it was detected in only five patients. (B) Anti-inflammatory cytokines.

The M_UF at Q_B = 200 ml/min as compared to Q_B = 100 ml/min was significantly higher for all cytokines, except for IL-10, for sTNFR-I at t = 12/24, and for IL 1-ra at t = 12/24 (Table 2). The M_AD was also significantly higher with the higher Q_B, with TNF-α at t = 12/24 and IL-10 at t = 12/24 as exceptions (Table 2).

View this table:

• View inline
• View popup

Table 2.

Comparison of the mass removal rate by ultrafiltration (M_UF) and by adsorption (M_AD) for the two blood flow rates (Q_B)^a

Discussion

The inflammatory cytokines, such as TNF-α, IL-1β, and IL-6, are essential in the local host response to a microbial challenge. However, their unbalanced production and systemic release may lead to diffuse tissue injury and the development of multiple organ failure (1,2,3). Several studies have found correlations between circulating levels of inflammatory cytokines and outcome of patients with sepsis (23). These mediators have therefore received attention as potential targets in the treatment of sepsis. Clinical trials with monoclonal antibodies and other anticytokine strategies have so far failed to show a clear survival benefit (24). Blockade of only one mediator, however, may be insufficient to downregulate the inflammatory response.

The application of the CRRT to remove nonspecifically inflammatory mediators from the circulation therefore offers an attractive alternative. It has been suggested that the high endogenous production and clearance rate of the cytokines and their short half-life in the circulation preclude important extra-corporeal removal (25). Several studies, examining removal of inflammatory mediators with CRRT in septic patients, reported the presence of mediators in the ultrafiltrate or a positive plasma clearance (11,12,13,14,15,16,17,18,19,20), but to our knowledge only two studies showed a significant lowering of their plasma concentrations (19,20). In the present study, a strikingly consistent fall in the plasma concentrations of all cytokines was found, 1 h after the start of CVVH as well as 1 h after the replacement of the AN69 hemofilter. At these time points, about one-third of the amount of cytokine present in the prefilter plasma was removed by passage through the hemofilter. For most cytokines, the fall in the plasma concentration was not sustained, corresponding with a progressive decrease in the removal rate with longer blood-membrane contact.

One of the objectives of the present study was to quantify convective transport and membrane adsorption and to determine their relative contribution to total cytokine removal, to delineate more efficient removal strategies. If removal is predominantly convective, the efficiency could be increased by inducing a higher ultrafiltration rate. If elimination is mainly due to adsorption, more benefit could be expected from frequent membrane changes or the use of alternating parallel membrane systems.

Convective removal of a solute depends on transmembrane pressure, on molecular weight (MW) and structure of the solute, as well as on the cutoff point of the membrane, which averages 35 to 40 kD for the AN69 membrane. TNF-α is present in the circulation as a biologically active trimer with a MW of 54 kD, together with the nonactive monomer of 17 kD (23). The TNF-α detected in the ultrafiltrate by the immunoassay is probably to a large extent the monomer, since only this form can pass the average pore size of the AN69 membrane. The MW values of IL-1β (17 kD), IL-6 (26 kD), and IL-1ra (17 to 22 kD) allow passage across the AN69 hemofilter, and these cytokines were indeed detected in the ultrafiltrate. None of the ultrafiltrate samples contained IL-10, which has a MW of 35 to 40 kD. sTNFR-I and sTNFR-II are the extracellular domains of the respective TNF receptors and have a MW of 30 and 33 kD. Their SC values were very low. The rate of convective cytokine removal remained stable during the 12-h hemofiltration at a particular Q_B. Convective transport, however, represented only a relatively small fraction of total cytokine removal.

In the present study, membrane adsorption emerged as the main clearance mechanism for cytokines. In vitro studies have demonstrated that the synthetic membranes, and especially the AN69 membrane, avidly adsorb cytokines and complement components (5,6,7,8, 26, 27). The AN69 membrane consists of a copolymer of acrylonitrile and sodium methallylsulfonate with a homogeneous, dense, and symmetric structure and a strong hydrophilicity. Due to these physicochemical properties, it has the consistency of a hydrogel. In contrast to the microporous asymmetric membranes, all polymeric chains over the entire breadth of the AN69 membrane are in contact with the blood compartment and thus accessible for adsorption. Adsorption is further favored by the strong negative charges provided by the methallylsulfonate groups. Clinical studies have confirmed that important adsorption also occurs in vivo (17, 18, 28). We found adsorption to be most pronounced immediately after the installation of a new hemofilter (at t = 1 and t = 13) with a steady decrease thereafter. After 12 h of blood-membrane contact (at t = 12 and t = 24), negative adsorptive removal was found for some cytokines in a few patients, indicating saturation of the membrane and release of previously bound cytokine. For the whole group, net adsorption remained positive.

Q_B was randomly set at 100 or 200 ml/min at the start of CVVH and switched to 200 or 100 ml/min, respectively, after 12 h of hemofiltration. The ultrafiltration rate was 75% higher with the Q_B of 200 ml/min as compared to 100 ml/min. The M_UF of the cytokines increased more or less in parallel with the ultrafiltration rate. However, the removal rate by adsorption was also significantly higher when the Q_B was set at 200 ml/min. The higher convective driving force may increase the surface area accessible for adsorption, by pushing the molecules deeper into the hydrogel. In this respect, it is known that only minimal adsorption occurs when the ultrafiltrate line is clamped (29). Additionally, a higher load of cytokines is presented at the membrane per unit of time at higher blood flows and may thus also favor the adsorption process. Since we did not control ultrafiltration rate independently of Q_B, we cannot distinguish between their respective importance.

The present results indicate that optimal cytokine removal can be achieved with a combination of a high Q_B/ultrafiltration rate and frequent membrane changes. Frequent membrane changes, however, are impractical, labor-intensive, and expensive. Alternative solutions might be examined such as parallel alternating membrane systems or cartridges containing adsorptive microparticles. Whether a stable lowering of the cytokine plasma concentrations could be achieved with these measures remains to be determined.

CVVH indiscriminately removes all solutes that can pass across the membrane or adsorb to the membrane. In a few studies, the removal of inhibitors of inflammation has been investigated. Van Bommel et al. found an increase in the ratio between the soluble TNF receptors and TNF-α upon the start of continuous hemofiltration (13). In another study, no reduction in the plasma levels of various pro- and anti-inflammatory cytokines was achieved with CVVH (11). In the present study, the removal rates of the inhibitors of inflammation paralleled those of the inflammatory cytokines. Furthermore, no significant changes in the ratio between the serum concentrations of TNF-α and its soluble receptors and between the serum concentrations of IL-1β and IL-1ra were found. The complex interplay between pro- and anti-inflammatory mediators, however, might not be entirely represented by a simple ratio between selected cytokines and their inhibitors. The present results therefore do not allow solid conclusions about the effect of CVVH on the pro/anti-inflammatory balance in sepsis.

Another objective of the study was the assessment of the impact of CVVH on respiratory and hemodynamic parameters during the first 24 h of treatment. No changes in oxygenation were observed. The hemodynamic response to CVVH was characterized by a stable MABP and heart rate. CO fell and SVR increased immediately after the start and remained stable thereafter. PAOP decreased, but this trend was not statistically significant. The fall in CO and the rise in SVR coincided with a decrease of the plasma concentration of all cytokines. The hemodynamic parameters did not change after the replacement of the hemofilter at t = 12, which was followed by a comparable fall of the cytokine plasma concentrations. Therefore, it appears improbable that the observed hemodynamic changes are caused by or related to an alteration in the inflammatory status. Although in our patients a zero fluid balance was aimed for, mild hypovolemia may have been induced by the extracorporeal circulation of blood and the time delay between ultrafiltration and infusion of the substitution fluid, and thus may have been responsible for the observed rise in SVR and fall in CO. Only a trend toward a lower PAOP was noted, but this parameter is a rough measure of left ventricular enddiastolic pressure and may not reflect left ventricular preload in sepsis and adult respiratory distress syndrome.

In conclusion, the present study demonstrates that CVVH with an AN69 membrane in septic patients removes cytokines from the circulation, mainly by membrane adsorption. The current data should not be generalized to other membranes, since their physicochemical characteristics and adsorptive capacities might be very different. Inhibitors of inflammation were removed to the same extent as the inflammatory cytokines, which may explain the lack of influence on short-term hemodynamics and respiratory function. Alternatively, the fall in the cytokine plasma concentrations obtained in our study might have been too small or too short-lived to produce a measurable effect on hemodynamic and respiratory parameters.

CVVH is a valuable tool in the treatment of acute renal failure in the intensive care unit patient. More work is required to determine whether optimization of the cytokine removal strategy leads to a sustained decrease of their plasma concentrations and improves the outcome of the patients.