Vascular access is a major determinant of circuit survival. Both high arterial and venous pressures are detrimental. Access failure causes blood flow reductions, which are associated with early circuit clotting 5. In vitro studies have found that high venous pressures in the circuit reduce circuit life 10. Randomized studies in critically ill patients on CRRT which evaluate the effect of catheter site or design on circuit flow and survival are not available. Most information comes from observational and in vitro studies in chronic hemodialysis patients, who need their catheters intermittently and for a much longer time (reviewed in 11). Some general principles are summarized in Figure 2 and are discussed below.

According to Poisseuille's law, flow through a catheter is related to the fourth power of radius and inversely related to length, indicating that a thick (13 to 14 French) and short catheter is preferable. However, a more central position of the tip improves flow, dictating sufficient length. In chronic dialysis patients, best flows are obtained with the tip in the right atrium 1213. With the femoral route, tip position should be positioned in the inferior caval vein. Because the inner diameter counts, the material is crucial. In general, silicone catheters have thicker walls than polyurethane catheters. Another issue is the presence of side or end holes. Flow through end holes is laminar, which is optimal, whereas flow through side holes is turbulent and even locally stagnant, contributing to early clotting. Suctioning of side holes against the vessel wall may impair flow, which is minimized with side holes over the (near) total circumference and absent with end holes. Another important determinant of catheter flow is the patient's circulation. For example, catheter dysfunction was found to be associated with low central venous pressure 12.

Furthermore, kinking of the catheter may impair catheter flow. Subclavian access has an enhanced risk of kinking and of stenosis with longer catheter stay 141516. The right jugular route is the straightest route. Furthermore, high abdominal pressures or high or very negative thoracic pressures, occupancy by other catheters, patency or accessibility of veins, anatomy, posture, and mobility of the patient determine choice of the site. Ultrasound-guided catheter placement significantly reduces complications 17. An important issue is locking of the CRRT catheter when not in use by controlled saline infusion or by blocking with heparin or citrate solutions to prevent fibrin adhesion, which slowly reduces lumen diameter 1819.

Slow reaction to pump alarms contributes to stasis of flow and early filter clotting. Training includes the recognition and early correction of a kinked catheter and the adequate rinsing of the filter before use since blood-air contact activates coagulation 2021. Intermittent saline flushes have no proven efficacy 22. Filling of the air detection chamber to at least two thirds minimizes blood-air contact.

For several reasons, continuous venovenous hemofiltration (CVVH) appears to be associated with shorter circuit life than continuous venovenous hemodialysis (CVVHD) 23. First, for the same CRRT dose, hemofiltration requires higher blood flows. Higher blood flows give more flow limitation and more frequent stasis of blood flow. Second, hemofiltration is associated with hemoconcentration, occurring as a consequence of ultrafiltration. Within the filter, hematocrit (Ht), platelet count, and coagulation factors increase the likelihood of coagulation. Continuous venovenous hemodiafiltration (CVVHDF) combines the possible advantages of hemofiltration (higher middle molecular clearance) with less hemo-concentration. Higher solute clearances can be attained at relatively lower blood flows and may thus increase circuit survival. However, a prospective survey in children on 442 CRRT circuits (heparin and citrate) could not find a correlation between circuit survival and CRRT mode (CVVH, CVVHD, or CVVHDF) 24.

To minimize the procoagulant effects of hemoconcentration, it is recommended to keep the filtration fraction (the ratio of ultrafiltrate flow [QF] to blood flow [QB]) as low as possible; a value below 25% is generally recommended in postdilution mode. It may be more rational to adjust the filtration fraction to the patient's Ht because blood viscosity in the filter is the limiting factor. Although many factors contribute to blood viscosity, Ht is the main determinant and is available at bedside. A Ht in the filter (Ht_filter) of 0.40 may be acceptable. Ht_filter and the minimal QB required for the prescribed QF can be calculated at bedside.

Ht_filter = QB × Ht_patient/(QB - QF),

QB = QF × (Ht_filter/(Ht_filter - Ht_patient).

Another option for reducing the filtration fraction is to administer (part of) the replacement fluid before the filter.

In predilution CRRT, substitution fluids are administered before the filter, thus diluting the blood in the filter, decreasing hemoconcentration, and improving rheological conditions. One small randomized cross-over study (n = 15) and one study comparing 33 patients on predilution CVVH to 15 historical postdilution controls found longer circuit survival with predilution 2526 at the cost of a diminished clearance 26. However, compared to the historical controls, mean daily serum creatinine changes were not significantly different 25. Reduced filter downtime may compensate for the lower predilution clearance. Predilution particularly reduces middle molecular clearance 27, the clinical consequences of which are still unclear.

Clogging is due to the deposition of proteins and red cells on the membrane and leads to decreased membrane permeability. Clogging is detected by declining sieving coefficients of larger molecules and increasing transmembrane pressures. Clogging enhances the blockage of hollow fibers as well. The process is still incompletely understood, but interplay between the protein constitution of plasma, rheological characteristics of blood, capillary and transmembrane flow, membrane characteristics, and possibly the use of different resuscitation fluids influence this process 1027. It has been suggested that with predilution, membrane performance is better maintained by reducing protein adsorption. On the other hand, others have shown more protein adsorption with predilution 28. This may be explained by the higher ultrafiltration rate, opening more channels and thus increasing the actual surface and the amount of protein adsorbed. Future developments to reduce protein adsorption include hydrophilic modification of polyetersulfone 29.

Biocompatibility is significantly influenced by membrane characteristics. Main determinants are electronegativity of membrane surface and its ability to bind plasma proteins, as well as complement activation, adhesion of platelets, and sludging of erythrocytes 30 (Figure 1). Few studies have evaluated the influence of membrane material on filter run times. Membranes with high absorptive capacity generally have a higher tendency to clot. In a non-randomized controlled study, polyamide exhibited later clotting than acrylonitrile (AN69) 31. Modification of existing membranes to increase heparin binding (AN69ST) reduced clotting in intermittent hemodialysis 32. Newer membranes with various polyethersulfone coatings that reduce activation of coagulation are being developed 33. Up to now, large randomized controlled trials evaluating the influence of the type of membrane on circuit life during CRRT have been missing.

Filter size may play a role and larger surfaces may be of relevance for filter survival and solute clearance when CVVHD is applied. A comparison of two polysulphone hemofilters with different hollow fiber lengths showed transmembrane pressure and increased survival time being lower with the longer filter 34.

Unfractioned heparin (UFH) is the predominant anticoagulant. Its major advantages are the low costs, ease of administration, simple monitoring, and reversibility with protamine 945. The half-life of UFH is approximately 90 minutes, increasing to up to 3 hours in renal insufficiency due to accumulation of the smaller fragments. Monitoring with activated partial thromboplastin time (aPTT) is still the best option. Retrospective analyses indicate increased bleeding if systemic aPTT is longer than 45 seconds 31. At this low level of anticoagulation, activated clotting time is relatively insensitive for monitoring 46. However, aPTT appears to be an unreliable predictor of bleeding 947. Given these limitations, a possible scheme for UFH consists of a bolus of 30 IU/kg followed by an initial rate of 5 to 10 IU/kg per hour in patients with normal coagulation. However, the level of anticoagulation should be individualized. Apart from bleeding, major side effects of UFH include development of heparin-induced thrombocytopenia (HIT), hypoaldosteronism, effects on serum lipids, and AT dependency 47.

Low molecular weight heparins (LMWHs) exhibit several advantages, including lower incidence of HIT 48, lower AT affinity, less platelet and polymorphonuclear cell activation, less inactivation by platelet factor-4 (PF-4), higher and more constant bioavailability, and lack of metabolic side effects 474950. However, data on the use of LMWH in CRRT are limited 7515253. Dalteparin, nadroparin, and enoxaparin have been investigated. Their mean molecular weight is between 4.5 and 6 kDa, and their mean half-life ranges from 2.5 to 6 hours and is probably even longer in renal insufficiency. However, there are indications that LMWHs are eliminated by CRRT 54. Although some studies use LMWH in a fixed dose 752, continuous intravenous application of LMWH, aiming at systemic anti-FX levels of 0.25 to 0.35 U/ml, may be the safest option 53. However, anti-Xa may not be a reliable predictor of bleeding 55 and anti-Xa determinations are not generally available.

HIT is caused by a heparin-induced antibody that binds to the heparin-PF-4 complex on the platelet surface. This may or may not lead to platelet activation and consumption, thrombocytopenia, and both arterial and venous thrombosis. Depending on the dose and type of heparin, the population, and the criteria used, 1% to 5% of treated patients develop HIT 56. Platelet count typically rapidly decreases by more than 50% after approximately 1 week or earlier after previous use of heparin. Diagnosis depends on a combination of clinical and laboratory results 57. A reliable diagnosis is complicated by the fact that the incidence of a false-positive enzyme-linked immunosorbent assay test is high 58. Unfortunately, the more precise carbon 14-serotonin release assay is not routinely available. Awaiting final diagnosis, all kinds of heparins should be discontinued and an alternative anticoagulant started.

There are no randomized controlled trials showing which anticoagulant is best for HIT. The choice depends on local availability and monitoring experience. If citrate is used for anticoagulation of the circuit, separate thromboprophylaxis must be applied. Inhibition of thrombin generation can be obtained via direct inhibition of FIIa (r-hirudin, argatroban, or dermatan sulphate), FXa (danaparoid or fondaparinux), or both (nafamostat). Inhibition of platelet activation can be obtained by the use of prostaglandins (PGs) (summarized in 959). The use of r-hirudin is discouraged because of severe adverse events, extremely long half-life (170 to 360 hours), and the requirement of ecarin clotting time for monitoring 60. Given the long half-life of fondaparinux and danaparoid (more than 24 hours), monitoring of anti-Xa is mandatory. The clinical relevance of cross-reactivity of danaparoid with HIT antibodies is not known 61. Argatroban might be preferred because it is cleared by the liver and monitoring with aPTT seems feasible 62636465. The half-life is approximately 35 minutes in chronic dialysis, but longer in the critically ill. Up to now, clinical data in CRRT and availability of the drug have been limited.

Regional anticoagulation can be achieved by the prefilter infusion of citrate. Citrate chelates calcium, decreasing ionized calcium (iCa) in the extracorporeal circuit. For optimal anticoagulation, citrate flow is adjusted to blood flow, targeting at a concentration of 3 to 5 mmol/l in the filter 71. Postfilter iCa can be used for fine tuning of the level of anticoagulation, aiming at a concentration of iCa of less than 0.35 mmol/l (Table 1). However, others prefer a fixed citrate dose and do not monitor iCa in the circuit, thereby simplifying the procedure (summarized in 9). Citrate is partially removed by convection or diffusion and partially enters the systemic circulation, where iCa rises again due to the dilution of extracorporeal blood, the liberation of chelated calcium when citrate is metabolized, and the replacement of calcium. As a result, systemic effects on coagulation do not occur.

Apart from being an anticoagulant, citrate is a buffer substrate. The generation of buffer is related to the conversion of sodium citrate to citric acid:

Na₃ citrate + 3H₂CO₃ → citric acid (C₆H₈O₇) + 3NaHCO₃

Citric acid enters the mitochondria and is metabolized in the Krebs cycle, mainly in the liver but also in skeletal muscle and the renal cortex, leaving sodium bicarbonate.

Citrate removal by CRRT mainly depends on CRRT dose and not on modality. Citrate clearance approximates urea clearance. The sieving coefficient is between 0.87 and 1.0 and is not different between CVVH and CVVHD 7273. Citrate removal with CRRT also depends on citrate concentration in the filter and filtration fraction; high fractions are associated with relatively higher citrate clearance and a lower buffer supply to the patient.

The use of regional anticoagulation with citrate is limited by the patient's capacity to metabolize citrate, which is decreased if liver function or tissue perfusion fails 74. Due to the citrate load associated with transfusion, patients having received a massive transfusion are also at risk of citrate accumulation. If citrate accumulates, iCa decreases and metabolic acidosis ensues, since bicarbonate continues to be removed by filtration or dialysis, while citrate is not used as a buffer. In daily clinical practice, citrate measurement is hampered by the limited stability of the reagents. However, accumulation of citrate due to decreased metabolism can be detected accurately by the symptoms of metabolic acidosis, increasing anion gap, ionized hypocalcemia, and most specifically by an increased total/iCa concentration. A ratio of more than 2.1 predicted a citrate concentration of greater than 1 mmol/l with 89% sensitivity and 100% specificity 71. Others use a ratio of more than 2.5 for accumulation 75.

Accumulation of citrate can also be the result of an unintended citrate over-infusion or of decreased removal in case of a decline in membrane performance at constant citrate infusion. In these cases, ionized hypocalcemia occurs together with metabolic alkalosis. Both derangements are preventable by adherence to the protocol or are detectable early by strict monitoring.

Anticoagulation with citrate has complex metabolic consequences, which are related to the dual effects of citrate as an anticoagulant and a buffer. Manipulation of citrate or blood flow, ultrafiltrate, dialysate, or replacement rates, and their mutual relation changes the amount of buffer substrate entering the patient's circulation. For a constant buffer delivery, these flows are to be kept constant, while they can be adjusted to correct metabolic acidosis or alkalosis. Causes of metabolic derangements and possible adjustments are summarized in Table 2.

Citrate is either infused as a separate trisodium citrate solution or added to a calcium-free predilution replacement fluid. The strength of citrate solutions is generally expressed as a percentage (grams of trisodium citrate per 100 ml). Some of the solutions contain additional citric acid to reduce sodium load. Because anticoagulatory strength of the solution depends on the citrate concentration, it is best expressed as molar strength of citrate. Citrate solutions for postdilution CVVH(D) contain 133 to 1,000 mmol citrate per liter 737576777879808182. Citrate replacement solutions for predilution CVVH contain 11 to 15 mmol citrate per liter 838485868788 and for predilution CVVHDF, 13 to 23 mmol/l 4089909192. The buffer strength of the solution is related to the conversion of trisodium citrate to citric acid (see formula above) and therefore to the proportion of sodium as cation.

After the first report of Mehta and colleagues 76, a wide variety of homemade citrate systems for CRRT have been described. There are systems for CVVHD, predilutional or postdilutional CVVH, CVVHDF, and different doses of CRRT (1.5 to 4 liters per hour) (summarized in the electronic supplemental material in 9). None of the proposed systems can attain perfect acid-base control using one standard citrate, replacement, or dialysis solution. Each protocol has its own rules to correct metabolic acidosis or alkalosis or hypocalcemia or hypercalcemia.

Some of the published studies compare circuit life and bleeding complications with citrate to historical or contemporary non-randomized controls on heparin (summarized in 9) 939495. Because the citrate patients often had a higher risk of bleeding, groups are generally not comparable. Nevertheless, bleeding complications were generally reduced in the citrate groups. Circuit survival with citrate was usually improved (summarized in 9) 93, sometimes comparable 248495, and in some studies shorter than with heparin 8994. Differences in circuit life between studies can be explained in part by the wide variety of citrate dose (2 to 6 mmol/l blood flow), fixed citrate infusion or citrate dose titrated on postfilter iCa, the use of dialysis or filtration (predilution or postdilution), differences in CRRT dose and filtration fraction, or by a reduction in citrate flow used for control of metabolic alkalosis. Only two small randomized controlled studies comparing anticoagulation with citrate to UFH have appeared in a full paper. Both show a significantly longer circuit survival with citrate 4082, a trend toward less bleeding 40, and less transfusion with citrate 82.

It may be questioned whether the benefits of citrate (less bleeding, possibly a longer circuit survival, and less bio-incompatibility 969798) weigh against the greater risk of metabolic derangement and possible long-term side effects like increased bone resorption 99. Preliminary results from a large randomized controlled trial (of approximately 200 patients) comparing regional anticoagulation with citrate to nadroparin in postdilution CVVH show that citrate is safe and superior in terms of mortality to nadroparin (H.M. Oudemans-van Straaten, to be published).