Clinical review: Anticoagulation for continuous renal replacement therapy - heparin or citrate?

Heparin is the most commonly prescribed anticoagulant for continuous renal replacement therapy. There is, however, increasing evidence questioning its safety, particularly in the critically ill. Heparin mainly confers its anticoagulant effect by binding to antithrombin. Heparin binds to numerous other proteins and cells as well, however, compromising its efficacy and safety. Owing to antithrombin consumption and degradation, and to the binding of heparin to acute phase proteins, and to apoptotic and necrotic cells, critical illness confers heparin resistance. The nonspecific binding of heparin further leads to an unpredictable interference with inflammation pathways, microcirculation and phagocytotic clearance of dead cells, with possible deleterious consequences for patients with sepsis and systemic inflammation. Regional anticoagulation with citrate does not increase the patient's risk of bleeding. The benefits of citrate further include a longer or similar circuit life, and possibly better patient and kidney survival. This needs to be confirmed in larger randomized controlled multicenter trials. The use of citrate might be associated with less inflammation and has useful bio-energetic implications. Citrate can, however, with inadequate use cause metabolic derangements. Full advantages of citrate can only be realized if its risks are well controlled. These observations suggest a greater role for citrate.


Introduction
Heparin is the most commonly prescribed anticoagulant for use during continuous renal replacement therapy (CRRT) [1]. By implication, it is considered the standard of care. Th ere is, however, increasing evidence question ing the safety of heparin particularly in the critically ill, and there are accumulating data on a potential better alternative, regional anticoagulation with citrate. In the present review, we discuss some of the complex interactions between heparin and the coagulation and infl ammation pathways, the associated consequences of such interactions in terms of effi cacy and safety in the critically ill, and the availability of the promising alternative citrate.

Heparin
In the present manuscript we often refer to heparin, not discriminating between unfractionated heparin and low molecular weight heparin (LMWH). Clinicians should, however, be aware of their diff erences [2]. LMWHs have a lower antithrombin (AT) dependency, cause less tissue plasminogen activator inhibitor depletion, exhibit less binding to proteins and cells, cause less activation of platelet factor 4 and cause less heparin-induced thrombocytopenia. Some of the eff ects mentioned below will therefore be less pronounced for LMWH. Many of these eff ects are, however, described for LMWH as well -in such cases, we have indicated this in the text or the tables.

Heparin resistance Low antithrombin due to consumption and inactivation
Th e anticoagulant eff ect of heparins depends on potentiating AT, the most important endogenous inhibitor of thrombin and other coagulation factors [2]. In critically ill patients, however, AT concentrations are often reduced as a result of consumption due to activation of coagulation in conditions such as sepsis or systemic infl ammation. AT is also reduced as a result of proteolytic degradation by granulocyte-derived elastase, and this eff ect is actually enhanced by heparin [3,4]. During infl am mation, elastase release is increased while its inactivation is decreased, because oxygen radicals decrease the activity of its natural inactivator α 1 -antitrypsin. Both high AT consumption and increased AT degradation thus contribute to heparin resistance during critical illness.
to proteins other than AT limits the amount of heparin available to act as a cofactor for AT and thus decreases its anticoagulant eff ect [5]. So-called heparin-binding proteins are released from endothelial stores [6]. Among these are acute-phase reactants such as platelet factor 4, histidine-rich glycoprotein, vitronectin, fi bro nec tin and lipopolysaccharide-binding protein, all of which increase in sepsis and other forms of infl ammation [2,7-9]. Furthermore, heparin avidly binds to apoptotic and necrotic cells to discrete domains released from the nucleus onto the membrane as the cell dies [10]. Affi nity of dead cells decreases if heparins with a lower molecular weight are used. Apoptosis is a key mechanism of cell injury in sepsis-related multiorgan failure, while necrotic cells are abundant in ischemia reperfusion injury. Importantly, the heparin-binding sites on apoptotic cells also signal phagocytotic clearance, and thus heparin may actually delay such clearance [10].
A recent study in critically ill patients with acute kidney injury receiving CRRT with LMWH anticoagulation (nadro parin) found that early fi lter clotting was asso cia ted with severe organ failure, consumptive coagulopathy and heparin resistance. LMWH resistance was inde pendent of low AT levels [11]. In addition to low AT levels, increased heparin binding is therefore a feature of critical illness, also with the use of LMWH, and contri butes to heparin resistance and possibly to other deleterious eff ects such as delayed phagocytotic removal of dead cells [10].

Heparin and bleeding
Th e main drawback of heparin is that it causes systemic anticoagulation in addition to circuit anticoagulation. Heparin thus increases the patient's risk of bleeding. Critically ill patients are at risk of bleeding due to recent surgery, trauma, mucosal lesions and coagulopathy. More or less severe bleeding events are reported in 10 to 50% of cases, depending on the population and the degree of anticoagulation [12,13]. Using no anticoagulation treatment is a safe option with regard to bleeding, but not to clotting, because the circuit life is generally reduced. Circuit anticoagu lation without systemic anticoagulation (for example, regional anticoagulation) represents an ideal solution. Antagonizing heparin by administering prota mine after the fi lter confers regional anticoagu lation. Th e circulating heparin-protamine complexes, however, may be harmful [14]. Regional anticoagulation with sodium citrate seems an attractive alternative. Th is topic is dis cussed below.

Heparins and infl ammation Proinfl ammatory eff ects: antithrombin mediated
Binding of heparin to AT not only potentiates its anticoagulant eff ects, but additionally inhibits the anti-infl ammatory actions of AT. Th e anti-infl ammatory eff ects of AT are exerted through binding to gluco s amino glycans on endothelial membranes, enhancing the formation of prostacyclin. Th e binding of AT to glucos amino glycans diminishes the adherence and migration of leukocytes, reduces platelet aggregation and decreases proinfl ammatory cytokine production [15,16]. Heparin binding to AT abolishes this eff ect [15]. Further more, during sepsis or ischemia reperfusion, elastase is increased. Recent data indicate that heparin, which normally poten tiates AT, inactivates AT in the presence of elastase. Th is process leads to proinfl am matory and procoagulant eff ects on the endothelium in sepsis, which may com pro mise the microcirculation and threaten tissue perfusion [4]. Heparin may thus have adverse eff ects on the microcirculation in sepsis [17,18].

Proinfl ammatory and anti-infl ammatory eff ects not related to antithrombin
Both unfractionated heparin and LMWH can mobilize infl ammatory mediators -such as myeloperoxidase, lactoferrin, elastase and platelet factor 4 -not only from circulating neutrophils and platelets and from cells adhered to and activated by the dialysis membrane, but also from glucosaminoglycans such as heparan sulfate on the endothelial surface [19][20][21]. Although this release may be interpreted as being protective for the endothelium, LMWH dialysis has been associated with higher serum concentrations of oxidized low-density lipoproteins com pared with citrate dialysis, possibly indicating systemic oxidizing eff ects by myeloperoxidase [20].
Mobilization of infl ammatory mediators from endothelial glucosaminoglycans may be especially relevant in sepsis and other forms of infl ammation. Heparin also releases the protective superoxide dismutase from the endothelium [22] and may have other anti-infl ammatory properties as well [23]. Th ese anti-infl ammatory properties may be partially mediated by inhibiting thrombin generation [24]. In addition, heparin -at much higher doses than used for anticoagulation -seems to attenuate endothelial dysfunction after ischemia-reperfusion injury by blocking the infl ammatory cascade at multiple levels, including complement activation [25], P-selectin and L-selectin-initiated cell adhesion and infl ux, and activation of the proinfl ammatory transcription factor NF-κB [26,27]. In contrast to the observations in plasma-free conditions, where heparin seems to reduce endotoxininduced cytokine synthesis dose dependently [9] and to bind TNF [23,28], both unfractionated heparin and LMWH cause a dose-dependent increase in IL-8 and IL-1β in plasma, which depends on the presence of lipopolysaccharide-binding protein [9]. Heparin binds to lipopolysaccharide-binding protein, and this binding facilitates the transfer of lipopolysaccharide to the CD14 receptor, thereby augmenting the endotoxin-induced activation of monocytes [9,18]. It should be noted that, for the prevention of clotting of the extracorporeal circuit, higher doses of heparins are often required, increasing the risk of proinfl ammatory eff ects [9,29].
Altogether, translation of these experimental results to critically ill patients remains speculative. Post hoc analysis of the major sepsis trials in humans suggests that lowdose heparin might have favorable eff ects on survival [23]. Th ere is, however, probably a bias -the patients receiving heparin were less critically ill compared with those judged unable to tolerate heparin. Altogether, due to its avid binding properties, both unfractionated heparin and LMWH interfere at multiple levels with the infl ammatory cascade (Figure 1). Th e eff ects depend on the dose, timing, the clinical setting and many unknown factors, leading to unpredictable eff ects.

Heparin-induced thrombocytopenia
A feared side eff ect of heparins is the development of heparin-induced thrombocytopenia, related to the bind ing of heparin to platelet factor 4 released from activated platelets. Some patients develop antibodies against this heparin-platelet factor 4 complex. Th e antibody-platelet factor 4-heparin complex subse quently binds to platelets, inducing platelet activation, aggregation and activation of the coagulation pathways. Th is sequence results in a loss of circulating platelets and a prothrombotic state [30]. Depending on the dose and type of heparin, on the population and on the criteria used for diagnosis, <1 to 5% of heparin-treated patients develop heparin-induced thrombocytopenia [31][32][33].

Citrate
Given the above concerns with heparin, why is regional citrate anticoagulation not the standard of care? For several reasons, physicians are reluctant to switch to citrate anticoagulation. Th ese reasons include concerns that the use of citrate is complex, carries a high risk of metabolic derangement, might even be dangerous and is expensive.

Citrate: anticoagulant and buff er
Citrate is both an anticoagulant and a buff er, and for this reason it can be diffi cult to use and understand. Sodium citrate, administered before the fi lter, chelates calcium. Th e associated regional hypocalcemia in the fi lter inhibits the generation of thrombin [34,35]. Citrate is partially removed by fi ltration or dialysis [36], and the remaining amount is rapidly metabolized in the citric acid (Krebs) cycle -especially in the liver, muscle and renal cortexwhile the chelated calcium is released and the lost calcium is replaced. Systemic coagulation is unaff ected. For anticoagulation, the citrate dose is adjusted to blood fl ow to attain an ionized calcium concentration <0.4 mmol/l in the fi lter; the lower the calcium concentration, the higher the degree of anticoagulation. Some protocols use a fi xed dose of citrate in relation to blood fl ow according to an algorithm, and target such doses at about 3 mmol citrate/l blood fl ow [37]. Other protocols measure postfi lter ionized calcium and adjust the citrate dose, which complicates the intervention but optimizes anti coagulation [13,38].
While the anticoagulant strength of the citrate solution depends on the citrate concentration, the buff er strength depends on the proportion of strong cations in the fl uid counterbalancing the citrate anion. One micromole of trisodium citrate provides the same buff er as 3 mmol sodium bicarbonate, assuming the citrate is completely metabolized. In some solutions, however, hydrogen is used instead of sodium for some of the cation. Hydrogen (citric acid) does not act as a buff er. For example, in the ACD-A solution (citrate dextose) that is used in many protocols, 30% of the cations consist of hydrogen. Th e buff ering capability of this solution is thus lower than when a pure trisodium citrate solution is used. In our view, the Stuart concept of acid-base provides an easier way to understand the buff ering eff ect of citrate for those familiar with the concepts. After metabolism of citrate, the remaining sodium increases the strong ion diff erence (SID) [39,40]: SID = (Na + + K + + Ca 2+ + Mg 2+ ) -(Cl -+ lactate -) An increased SID produces alkalosis, while the infusion of a zero-SID fl uid decreases SID and causes acidosis. Citrate is a triprotic acid with pK a values of 3.13, 4.76 and 6.40. A sodium citrate solution thus has a SID of zero until the citrate is metabolized. In conditions where citrate metabolism is grossly impaired, therefore, even trisodium citrate can fi rst cause some acidosis before it produces its alkalinizing eff ect.
Because sodium citrate acts as buff er, the replacement fl uids must be adjusted accordingly. A myriad of homemade systems are in use and, unfortunately, add to confusion among clinicians: for predilution or post dilution, for hemodialysis or hemofi ltration, or for their combination. All these modalities use diff erent fl uids and diff erent ways to attain metabolic stability. None of the systems has yet proven superiority. Th e pros and cons have been summarized in various reviews [38].

Metabolic consequences, tolerance and limitations
Reported derangements include metabolic alkalosis and acidosis, hypernatremia and hyponatremia, and hypocalcemia. Th ey are detected with proper monitoring and the local protocol should describe how to adjust fl ows under diff erent conditions to prevent such prob lems. Th e incidence of metabolic complications depends on the amount of citrate administered, the design, rules and fl exibility of the protocol, and its proper use. In the largest randomized controlled trial, metabolic control with citrate was better than with heparin [35].

Calcium
Citrate tolerance depends on the amount of citrate entering the body and the capacity to metabolize citrate. Th e amount of citrate needed for anticoagulation is lower when using continuous hemodialysis compared with hemo fi ltration, because dialysis can be performed at a lower blood fl ow and therefore requires a lower citrate dose. Citrate metabolism is diminished in case of liver failure and also of poor tissue perfusion [41,42], because the Krebs cycle only operates under aerobic conditions. Citrate accumulates if its metabolism is insuffi cient. Th e citrate molecule is not toxic in itself, but citrate accumulation causes ionized hypocalcemia in the systemic circulation, which may decrease cardiac contractility, cause hypo tension by decreasing vascular tone and contribute to systemic hypocoagulability. Monitoring of the patient's ionized calcium is therefore crucial. Ionized hypo cal cemia is the most sensitive indicator of citrate accumu lation, but it can also stem from other causes, including an insuffi cient replacement rate. Citrate accumulation, while decreasing ionized calcium, concomitantly increases the total calcium concentration, because the citrate-bound calcium increases and calcium is replaced. Th e total to ionized calcium ratio, therefore, is a useful test to detect citrate accumulation [43][44][45]. A rise in total to ionized calcium ratio above 2.25 should trigger con sideration of probable citrate accumulation [43]. Although citrate accumulation can also increase the anion gap, this variable is insensitive in critically ill patients. In the aforementioned randomized trial, citrate anticoagulation was better tolerated than heparin [37].

Metabolic alkalosis or acidosis
Notably, accumulation of citrate can additionally lead to acid-base derangements. Metabolic alkalosis occurs if a high sodium citrate load (for example, protocol violation or accidentally rapid infusion) is delivered and when the citrate is appropriately metabolized, providing bicar bonate equivalents or, according to the Stewart concept, leaving unopposed sodium (see above) [38]. Metabolic acidosis occurs if a large citrate dose is not metabolized; for example, in liver failure or poor peripheral perfusion. A high lactate level (for example, >5 mmol/l) in patients with liver failure before the start of citrate CRRT may act as a useful marker of greater risk for citrate accumulation. Septic patients with a high lactate level generally tolerate citrate remarkably well if circulation improves. Accumulation of citrate is easily detectable with standard mon itor ing: arterial blood gases, ionized and total calcium at 6-hour intervals. If citrate accumulation occurs together with metabolic acidosis and an increasing calcium ratio, the citrate dose should be decreased or discontinued, calcium replaced and the patient should receive bicarbonate replacement, preferably as part of the balanced replacement fl uids [38]. With proper monitoring, citrate accumulation causes no clinical symptoms.

Infl ammation
Apart from the potentially proinfl ammatory eff ects of heparin on the release of mediators from leukocytes and platelets (see above), it may be hypothesized that local hypocalcemia at a membrane level during anticoagulation with citrate might reduce infl ammatory mediator release from cells adhered to the membrane [46,47]. Activation of neutrophils and the release of mediators from intracellular granules are preceded by transient increases in cytosolic Ca 2+ caused by the mobilization of Ca 2+ from intracellular stores followed by the infl ux of Ca 2+ into the cell through the plasma membrane channels. Cytosolic Ca 2+ subsequently acts as an intracellular messenger [48][49][50]. It should be noted that the eff ects of citrate on complement activation, especially seen when using the old unsubstituted cuprophane membranes, are not uniform -some studies report no eff ect [51][52][53], others report a suppression of activation [46]. Degranulation seems to be mediated by a diff erent membrane receptor, not related to complement activation and neutropenia [54].

Energy and mitochondria
Citrate (C 6 H 5 O 7 ) is a source of energy, conferring 3 kcal/g (0.59 kcal/mmol) [55,56]. One gram of citrate corresponds to about 5 mmol. Part of the citrate administered before the fi lter enters the patient's circulation. Th e amount depends on the citrate infusion dose, its concentration in the fi lter and the amount removed by fi ltration or dialysis. Th e latter can be estimated, because the sieving coeffi cient of citrate is close to 1.0. Effl uent loss corresponds to effl uent fl ow × citrate concentration in the fi lter. Th e citrate concentration in the fi lter can be estimated by dividing the administered amount by blood fl ow. Postdilution hemofi ltration estimation is easiestthe amount lost by fi ltration equals the infusion rate × fi ltration fraction, and the amount entering the patient is equivalent to the infusion rate × (1 -fi ltration fraction). Assuming a blood fl ow of 200 ml/minute, a citrate infusion rate of 3 mmol/l blood fl ow (36 mmol/hour), a fi ltration fraction of 0.20, and a fi lter running 80% of the time, about 553 mmol citrate enter the patient's circulation per day, providing close to 333 kcal/day. Th e present calculation is an estimate, based on physiological principles. Actual measurements have been published in abstract form recently [57].
Note that citrate, although being an energy source, does not need insulin to enter the cell. Citrate is an easy fuel for the mitochondria, and may replenish the Krebs cycle if intermediates are scarce. Substrate availability is a crucial regulator of the Krebs cycle. Th e cycle is a central pathway providing energy and maintaining the mitochon drial redox state. In sepsis, inhibition of pyruvate dehydrogenase limits pyruvate conversion to acetylcoenzyme A, the main substrate of the cycle [58]. Further more, Weinberg and colleagues have shown that citric acid cycle intermediates may protect proximal tubules against injury and may promote recovery from sustained mitochondrial energetic defi cit. Th is defi cit occurs after hypoxia and reperfusion and acts by lowering the cellular burden of non-esterifi ed fatty acids that appear to account for much of the continuing mitochondrial dysfunction [59,60]. Th e serum concentrations of citrate reached during CRRT with citrate as an anticoagulant (0.3 to 0.5 mmol/l) [44] are in the range expected to modify tubular cell metabolism (personal communication).

Life-threatening adverse event
Th e main risk of citrate anticoagulation consists of the uncontrolled systemic infusion of a hypertonic citrate solution causing acute and severe hypocalcemia, hypotension and possibly cardiac arrest. Treatment consists of immediate calcium infusion and discontinuation of citrate infusion.

Costs
Although comparative studies are not available and expenses diff er between modalities, estimated costs of citrate anticoagulation are not substantially higher than those associated with heparin anticoagulation. Instead of measuring the activated partial thromboplastin time for the control of heparin anticoagulation, systemic ionized calcium and once-daily measurement of total calcium is suffi cient for the monitoring citrate accumulation, with more frequent measurement only in patients at risk of accumulation. Replacement fl uids including intravenous calcium are slightly more expensive when citrate is used. If the costs of less bleeding and longer circuit life are calculated, costs with citrate are probably less or may in fact represent cost savings.

Strategies to increase the safety of citrate
Th e use of citrate off ers many advantages to the patient, which can only be taken if the risks are well controlled. Th e incidence of metabolic complications largely depends on the design, rules and fl exibility of the protocol, and its proper use. Safe introduction of citrate should start with the choice of a well-designed and fl exible protocol with proven effi cacy and adjusted to the local preferences of modality and dose, and availability of fl uids and devices. Results of ionized calcium measurement should be available 24 hours a day. Implementation of the protocol includes the training of staff to create understanding and awareness of acid-base and calcium monitoring, and the correction of imbalance [38]. Safety is further improved by the incorporation of a citrate module in the CRRT device, blocking persistent citrate infusion if the blood pump stops. Pop-ups in the patient data management system and computerized algorithms may add to patient safety.

Effi cacy and side eff ects
Th e primary goal of anticoagulation for CRRT is to extend circuit life. Indeed, a shorter circuit life may result in underdosing of CRRT. Several studies have compared circuit life and bleeding complications under citrate treatment with nonrandomized controls on heparin treat ment (summarized in [13,38]). Groups were often not comparable, because the patients receiving citrate generally had a higher bleeding risk. Nevertheless, bleeding complica tions were lower in patients receiving citrate [61,62]. Circuit survival with citrate was usually longer, some times similar and, in some studies, shorter than with heparin (summarized in [38]). Five randomized con trolled studies comparing anticoagulation with citrate to anticoagulation with heparin have been published [63][64][65][66] (Table 1). Th ree of these studies report a signifi cantly longer circuit survival with citrate, four studies (a trend to) less bleed ing, and two studies less red blood cell transfusion using citrate.

Clinical outcomes
Most randomized studies evaluating heparin anticoagulation for CRRT are too small for a robust evaluation of clinical outcome [13]. A recent randomized study in 200 critically ill patients receiving CRRT for acute kidney injury, comparing anticoagulation with LMWH (nadroparin) to regional anticoagulation with citrate, unexpectedly showed a 15% absolute increase in 3-month survival using an intention-to-treat analysis [35] (Table 2). In addition to improved patient survival, kidney survival was also better with citrate. Th e diff erence was signifi cant for all per-protocol patients and tended to signifi cance for the sur viving patients. Among the higher proportion of surviving patients, therefore, more patients were free from chronic dialysis in the citrate group. Of note, the benefi t of citrate on survival could not be fully explained by less bleeding. Citrate appeared particularly benefi cial in surgical patients, younger patients, patients with sepsis (compared with no sepsis) and in patients with a high degree of organ failure. In none of the subgroups did citrate perform worse than heparin. Th ese fi ndings suggest the need for a better understanding of citrate anticoagulation and its challenges. Th ese fi ndings, however, were not confi rmed in a more recent multicenter randomized trial, including a slightly diff erent patient population, using a diff erent citrate protocol and reporting a shorter follow-up [66].

Conclusion
Th e use of heparin as an anticoagulant for CRRT is hampered by complex interactions between the drug and numerous acute phase proteins, and apoptotic and necrotic cells. In addition, heparin binding inhibits the anti-infl ammatory eff ects of AT, which may already be low, and triggers the release of infl ammatory mediators from blood and endothelial cells. Th is myriad of actions leads to an unpredictable dose-eff ect relationship and an uncontrollable and potentially deleterious interference with proinfl ammatory and anti-infl ammatory pathways. In randomized clinical trials, heparins are often less eff ective for preserving circuit life while they increase the patient's risk of bleeding. Th e benefi ts of citrate include less bleeding, a longer or similar circuit life, and possibly better patient and kidney survival. Th e last benefi t needs to be confi rmed in larger randomized controlled multicenter trials. Th e use of citrate might be associated with less infl ammation and has useful bio-energetic implications. Th ese observations call into question the primacy of heparins for circuit anticoagulation in the critically ill patient and suggest a greater role for citrate (Table 2). Th e full advantage of citrate anticoagulation can only be realized, however, if its risks are well controlled by means of a sound protocol, well-trained staff , and the incorporation of a citrate module in the CRRT device.