Citrate anticoagulation for continuous venovenous hemofiltration*

Heleen M. Oudemans-van Straaten, MD, PhD; Rob J. Bosman, MD; Matty Koopmans, RN;Peter H. J. van der Voort, MD, PhD, MSc; Jos P. J. Wester, MD, PhD; Johan I. van der Spoel, MD;Lea M. Dijksman, MSc; Durk F. Zandstra, MD, PhD

Acute renal failure (ARF) in crit-ically ill patients represents astrong and an independent riskfor mortality (1). Prognosis is es-

pecially poor if renal replacement therapy(RRT) is required. Strategies of RRT mayinfluence outcome. Among these is RRTdose (2, 3).

Anticoagulation is required to preventclotting in extracorporeal circuits. Hepa-rins are the classic choice. Their maindrawback is bleeding because of systemicanticoagulation (4). Citrate is an attractivealternative (5). Citrate chelates calcium de-creasing ionized calcium in the circuit. Be-cause calcium is a cofactor in the coagula-

tion cascade, thrombin generation isinhibited. Citrate and calcium are partiallyremoved by filtration or dialysis (6, 7). Theremaining citrate is rapidly metabolized ifliver function andmuscle perfusion are suf-ficient (8). Calcium is replaced. Systemiceffects on coagulation are thus avoided.However, because citrate is substrate forbuffer as well, its use may cause metabolicderangements (9). The three small ran-domized controlled studies comparing ci-trate with heparin in critically ill patientsfound longer or similar circuit life and lessbleeding or transfusion with citrate (10–12). Large randomized studies are notavailable yet. The aim of this study was tocompare the safety and efficacy of regionalanticoagulation for continuous venovenoushemofiltration (CVVH) with citrate to ourstandard systemic anticoagulation with thelow-molecular weight heparin nadroparin(13, 14). Low-molecular weight heparins areeffectively used by others as well (15, 16).

MATERIALS AND METHODS

Study Design and Setting. This nonblindedsingle-center randomized controlled trial

comparing the safety and efficacy of two anti-coagulant strategies for CVVH was conductedin an 18-bed closed format general intensivecare unit of a teaching hospital. CVVH is theonly modality of RRT in the unit and is per-formed under responsibility of the intensiv-ists. The Institutional Review Board approvedthe study according to European and Dutchlegislation. At that time, the need for informedconsent was waived because the two modali-ties were standard practice in the unit, andpatients with an increased risk of adverseevents to either intervention were not in-cluded (13, 14, 17). All patients or their legalrepresentatives received written informationexplaining that data collected in the patientdata management system were used to evalu-ate treatment.

Patients and Randomization. All adultcritically ill patients with ARF requiring RRTin the unit were eligible for inclusion. CVVHwas initiated when after resuscitation of thecirculation oliguria persisted and was accom-panied by a steep rise in serum creatinine, orat a nondeclining rise in creatinine in nono-liguric patients. Exclusion criteria were livercirrhosis Child-Pugh C, (suspicion) of bleed-ing necessitating transfusion or fall in hemo-globin �0.5 mmol/L within 24 hours, surgerywithin 24 hours before CVVH, need of thera-peutic anticoagulation, (suspected) heparin-

*See also p. 764.From the Department of Intensive Care Medicine

(HMO-VS, RJB, MK, PHJVDV, JPJW, JIVDS, DFZ); andTeaching Hospital (LMD), Onze Lieve Vrouwe Gasthuis,Amsterdam, The Netherlands.The authors have not disclosed any potential con-

flicts of interest.For information regarding this article, E-mail:

h.m.oudemans-vanstraaten@olvg.nlCopyright © 2009 by the Society of Critical Care

Medicine and Lippincott Williams & Wilkins

DOI: 10.1097/CCM.0b013e3181953c5e

Objective: Continuous venovenous hemofiltration (CVVH) is ap-plied in critically ill patients with acute renal failure for renal re-placement. Heparins used to prevent circuit clotting may causebleeding. Regional anticoagulation with citrate reduces bleeding, buthas metabolic risks. The aim was to compare the safety and efficacyof the two.

Design: Randomized, nonblinded, controlled single-center trial.Setting: General intensive care unit of a teaching hospital.Patients: Adult critically ill patients needing CVVH for acute

renal failure and without an increased bleeding risk.Interventions: Regional anticoagulation with citrate or systemic

anticoagulation with the low-molecular weight heparin nadroparin.Measurements and Main Results: End points were adverse

events necessitating discontinuation of study anticoagulant, trans-fusion, metabolic and clinical outcomes, and circuit survival. Of the215 randomized patients, 200 received CVVH per protocol (97 citrateand 103 nadroparin). Adverse events required discontinuation ofcitrate in two patients (accumulation and clotting) of nadroparin in 20(bleeding and thrombocytopenia) (p < 0.001). Bleeding occurred in 6

vs. 16 patients (p � 0.08). The median number of red blood cell unitstransfused per CVVH day was 0.27 (interquartile range, 0.0–0.63) forcitrate, 0.36 (interquartile range, 0–0.83) for nadroparin (p � 0.31).Citrate conferred less metabolic alkalosis (p � 0.001) and lowerplasma calcium (p < 0.001). Circuit survival was similar. Three-month mortality on intention-to-treat was 48% (citrate) and 63%(nadroparin) (p � 0.03), per protocol 45% and 62% (p � 0.02). Citratereduced mortality in surgical patients (p � 0.007), sepsis (p � 0.01),higher Sepsis-Related Organ Failure Assessment score (p � 0.006),and lower age (p � 0.009).

Conclusions: The efficacy of citrate and nadroparin anticoag-ulation for CVVH was similar, however, citrate was safer. Unex-pectedly, citrate reduced mortality. Less bleeding could onlypartly explain this benefit, less clotting could not. Post hoc citrateappeared particularly beneficial after surgery, in sepsis and se-vere multiple organ failure, suggesting interference with inflam-mation. (Crit Care Med 2009; 37:545–552)

KEY WORDS: citrate; hemofiltration; acute renal failure; heparin;nadroparin; anticoagulation; sepsis

545Crit Care Med 2009 Vol. 37, No. 2

induced thrombocytopenia, chronic dialysis,and do-not-resuscitate orders. Shock-relatedhepatitis and low-dose heparin for thrombo-prophylaxis were no exclusion criteria. Ran-domization was computer based in two blocksof 100 patients and an additional block of 20patients to compensate for dropouts. Wheninclusion and exclusion criteria were checkedin the patient data management system, thesystem automatically randomized the patient.

Study Protocol. In the nadroparin group,2850 IU of nadroparin (Sanofi-Synthelabo,Maassluis, The Netherlands) were added to the1 L priming solution. Patients received anintravenous bolus of 2850 IU at initiation ofCVVH, or 3800 IU when body weight exceeded100 kg, followed by a continuous infusion inthe extracorporeal circuit of 380 or 456 IU/hr,respectively, without anti-Xa monitoring (14,15). The hospital pharmacy prepared the ci-trate solution in 500 mL bags, containing 500mmol/L citrate, 1352 mmol/L sodium, and148 mmol/L hydrogen. Citrate dose was 3mmol/L blood flow (17). Citrate patients re-ceived standard thromboprophylaxis (unfrac-tionated heparin up to 10,000 IU/day or na-droparin up to 3800 IU/day). If adverse adventsneeded discontinuation of nadroparin or ci-trate, CVVH was continued using citrate re-spectively, nadroparin or no anticoagulationaccording to clinical judgment.

CVVH Protocol. We performed postdilutionCVVH targeting a blood flow of 220 mL/min,filtrate flow of 4000 mL/hr for each new cir-cuit and a filtration fraction below 32%. Aslong as the patient needed vasopressors, a newcircuit was connected if the filtrate flow de-creased below 4000 mL/hr to maintain trans-membranous pressure below 300 mm Hg. Innonvasodepressor-dependent patients, filtrateflow could be decreased after 24 hours to 2000mL/hr minimally. Circuits were disconnectedat high prefilter or transmembrane pressure(both more than 300 mm Hg), if vascularaccess failed, routinely after 72 hours or forclinical reasons (renal recovery, transport). Af-ter disconnection, a new circuit was initiatedimmediately if the patient remained vasopres-sor dependent. If not, CVVH was postponed for24 hours awaiting renal recovery (13). Weused a 1.9 m2 cellulose triacetate hollow fibermembrane (UF 205, Nipro, Osaka, Japan) andin the first half of this study the Diapact con-tinuous RRT (CRRT) device (B-Braun, Mel-sungen, Germany), thereafter we used theAquarius device (Edwards LifeSciences SA,Switzerland).

Replacement fluids were heated to 39°C. Inthe nadroparin group, we used commercialfluids, bicarbonate buffered (35 mmol/L) incase of metabolic acidosis and/or hyperlac-tatemia, and lactate buffered (42 mmol/L), ifacidosis was corrected and plasma lactate re-mained below 5 mmol/L (SH 53 HEP and BH504, Dirinco, Rosmalen, The Netherlands). Inthe citrate patients, a combination of bufferfree (SH 44 HEP part I, Dirinco, Rosmalen,

The Netherlands), containing 109.5 mmol/LNa�, 2.0 mmol/L K�, 1.81 mmol/L Ca��, 0.52mmol/L Mg��, 116.2 mmol/L Cl�, 3 mmol/Llactate, 1 g/L glucose, and bicarbonate-buffered fluids were used, guided by plasmabicarbonate and pH according to a computer-driven algorithm (www.nvic.nl). To maintainplasma-ionized calcium between 0.9 and 1.0mmol/L, calcium-magnesium-chloride (0–0.4mmol/hr of calcium and 0–0.24 mmol/hrmagnesium) was administered via a separateline or if not available into the venous cham-ber. Vascular access was obtained via a 12–14Fdouble-lumen catheter.

Study End Points. Primary outcomes weresafety and efficacy. We defined safety as theabsence of adverse events necessitating dis-continuation of study anticoagulant: e.g., ac-cumulation of citrate in the citrate group (de-fined as calciumtotal/calciumionized �2.25)(17), and bleeding (requiring �2 red bloodcells [RBCs] units or causing a �0.5 mmol/Lfall in hemoglobin within 24 hours) or sus-pected heparin-induced thrombocytopenia inthe nadroparin group (18). Bleeding was

quantified by transfusion rate. Trigger hemo-globins for transfusion were 4.0 or 4.5 mmol/Lin patients �40 years or between 40 and 60years without cardiopulmonary limitations,and 5.0 or 5.5 mmol/L in patients �60 yearswithout cardiopulmonary limitation or bleed-ing. Clinicians adhered to these triggers withsupport of the patient data management sys-tem. Efficacy was defined as circuit survival.

Secondary outcomes were hospital and3-month mortality. Of the patients dischargedalive, we queried outcome in the other hospi-tal or at home. Follow-up time was 3 monthsfor all patients.

Concomitant Treatment. Concomitant va-soactive treatment consisted of fluids (crystal-loids and gelatins; starches were not used),inotropics, vasopressors, and/or vasodilatorsto optimize circulation targeting mean arte-rial pressure above 60 mm Hg, cardiac indexabove 2.5 L/min/m2, and a central minus pe-ripheral temperature �4°C. Patients in shockreceived dexamethasone 1 mg/kg once (19),followed, after 24 hours, by a 4-day taperingscheme with prednisolon if the circulation re-

Figure 1. Enrollment, exclusion, and randomization of study patients. ICU, intensive care unit; CVVH,continuous venovenous hemofiltration; HIT, heparin-induced thrombocytopenia.

546 Crit Care Med 2009 Vol. 37, No. 2

mained vasopressor dependent. All patients re-ceived enteral nutrition and selective decon-tamination of the digestive tract (20).

Data Collection. Severity of illness and or-gan failure were scored using the Acute Phys-iology and Chronic Health Evaluation II andIII systems, the Simplified Acute PhysiologyScore II system, and the Sepsis-Related OrganFailure Assessment (SOFA) score as defined bythe Dutch National Intensive Care Evaluation

(www.stichting-nice.nl) (21–24). Renal func-tion was classified according to the RIFLE(Risk, Injury, Failure, Loss, End stage kidneydisease) System (25). All data were prospec-tively collected in the patient data manage-ment system of the unit. A research nurse(M. K.), who did not participate in patientcare, was responsible for the database.

Definitions. (Cardiac) surgical was de-fined as (cardiac) surgery within the week

before intensive care unit admission, medi-cal as no surgery in this period, sepsis ac-cording to the International Sepsis Defini-tion (26), duration of CVVH as the time frominitiation of the first session to disconnec-tion of the last, duration of ARF as the timefrom initiation to the end of RRT (includingtemporary intermittent hemodialysis afterdischarge), renal recovery as independencefrom RRT.

Table 1. Baseline characteristics of study patients and delivered continuous venovenous hemofiltration

Citratea

(n � 97)Nadroparina

(n � 103)

DemographicsAge (yrs) (IQR) 73 (64–79) 73 (67–79)Weight (kg) (IQR) 84 (80–87) 85 (82–88)Male sex, n (%) 66 (68) 70 (68)

Type of admissionMedical, n (%) 46 (47) 50 (49)Cardiac surgical, n (%) 27 (28) 30 (29)Other surgical, n (%) 24 (25) 23 (22)

Severity of illness at ICU admissionAcute Physiology and Chronic Health Evaluation II

Score, points (95% CI) 28 (27–30) 28 (27–29)Predicted mortality (%) (95% CI) 57 (51–62) 57 (53–62)

Simplified Acute Physiology Score IIScore, points (95% CI) 59 (55–62) 61 (58–64)Predicted mortality (%) (95% CI) 61 (56–66) 63 (59–68)

Acute Physiology and Chronic Health Evaluation III score,points (95% CI)

105 (98–111) 105 (100–111)

Plasma creatinine (mg/dL) (IQR) 2.3 (1.0–3.4)b 2.3 (1.5–3.6)c

Severity of illness at start CVVHSequential Organ Failure Assessment score, points (IQR) 11 (10–13) 11 (10–14)C-reactive protein (mg/L) (IQR) 111 (60–229) 113 (45–181)Plasma lactate (mmol/L) (IQR) 1.8 (1.2–2.3) 1.7 (1.3–3.6)Mean arterial pressure (mm Hg) (IQR) 60 (53–70) 60 (53–68)Cumulative fluid balance from ICU admission (L) (IQR) 6.2 (2.2–10.3) 6.5 (2.3–12.1)Dopamine dose (�g/kg/min) (IQR) 6.1 (3.5–10.3) 6.5 (3.0–9.7)Noradrenalin dose (�g/kg/min) (IQR) 0.000 (0.000–0.019) 0.000 (0.000–0.032)PaO2/FIO2 ratio (mm Hg/fraction) (IQR) 195 (153–246) 184 (149–245)Mechanical ventilation, n (%) 86 (89) 94 (91)Diuresis 6 hrs before (mL) (IQR) 130 (34–370) 150 (30–350)Plasma creatinine (mg/dL) (IQR) 3.8 (2.7–4.8)d 3.7 (2.8–5.1)e

Plasma urea (mg/dL) (IQR) 78 (56–106)f 78 (56–112)g

RIFLE score, points (CI)h 3 (2–3) 3 (2–3)Hemoglobin (mmol/L) (IQR) 5.8 (5.2–6.2)i 5.6 (5.2–6.2)j

Platelet count (109/L) (IQR) 144 (97–181) 125 (78–193)Prothrombin time (sec) (IQR) 12 (11–15) 13 (11–16)Activated partial thromboplastin time (sec) (IQR) 34 (28–42) 34 (30–42)

Causes of acute renal failurek

Sepsis, n (%) 42 (43) 50 (49)Cardiac failure, n (%) 57 (59) 49 (48)Absolute hypovolemia, n (%) 20 (21) 13 (13)Major surgery, n (%) 43 (44) 43 (42)Toxic, n (%) 5 (5) 5 (5)

Delivered CVVHTime ICU admission to start CVVH (days) (IQR) 1.9 (0.24–3.2) 1.8 (0.28–3.4)Duration of CVVH period (days) (IQR) 2.7 (1.6–6.8) 3.2 (1.7–7.6)Hours on CVVH device (IQR) 58 (36–105) 63 (31–109)Duration of acute renal failure (days) (IQR)l 2.8 (1.6–7.3) 3.2 (1.4–7.9)Ultrafiltrate flow CVVH period (mL/kg/hr) (IQR)m 36 (32–39) 33 (30–35)Ultrafiltrate flow on CVVH (mL/kg/hr) (IQR)n 41 (40–43) 40 (38–42)

IQR, interquartile range; ICU, intensive care unit; CI, confidence interval; CVVH, continuous venovenous hemofiltration; RIFLE, acute renal failureclassification score, Risk � 1, Injury � 2, Failure � 3.

aDifferences between randomized groups were not statistically significant; b203 (117–301) �mol/L; c198 (132–318) �mol/L; d332 (240–420) �mol/L;e326 (247–448) �mol/L; f28 (20–38) mmol/L; g28 (20–40) mmol/L; hWe scored RIFLE-risk as 1, injury as 2, and failure as 3; i9.2 (8.4–10) g/dL; j9.0 (8.4–10)g/dL; kMost patients had more than one cause, “toxic” was never the sole cause of acute renal failure; lIncluding intermittent hemodialysis after ICUdischarge; mIncluding filter down-time; nConnected to CVVH device.

547Crit Care Med 2009 Vol. 37, No. 2

Statistical Analysis. We based the power ofthe study on adverse events necessitating dis-continuation of study medication (safety). Onaccount of a 12.3% incidence of bleeding dur-ing heparin-based CVVH (27) and a 1.5% in-cidence of heparin-induced thrombocytopenia(18, 28), we hypothesized that, given the ex-clusion criteria, discontinuation of nadroparin

would be necessary in 12.5% of the patientswhereas citrate accumulation would developin 2% (9). A sample size of 200 patients wouldgive the study 80% power with a two-sided �level �0.05 to detect this difference. Althoughwe did not anticipate a mortality difference,the study would be large enough to detect atrend to whatever side. Analysis was per-

formed using per protocol approach, exclud-ing the patients who did not receive CVVH orwere erroneously randomized, because theprimary end points could not be (reliably)evaluated in these patients. We additionallyperformed an intention-to-treat analysis forhospital and 3-month mortality, avoiding anybias. Patients needing discontinuation of

Table 2. Prevalence of metabolic derangements before start of hemofiltration and at the end of individual hemofiltration sessions

Before Hemofiltrationa End of Hemofiltration Sessions

pbCitrate (%) Nadroparin (%) Citrate (%) Nadroparin (%)

Metabolic eventPlasma sodium �130 (mmol/L) 1 5 1 8 �0.001Plasma sodium �150 (mmol/L) 2 0 0.4 0 0.50Plasma bicarbonate �20 (mmol/L) 45 49 3 5 0.50Plasma bicarbonate �30 (mmol/L) 3 7 9 26 �0.001Plasma pH �7.30 31 30 5 5 0.84Plasma pH �7.50 2 4 9 19 0.001Plasma lactate �2 (mmol/L) 48 47 35 64 �0.001

Plasma ionized calcium (mmol/L)�0.9 13 9 6 2 �0.0010.9–1.19 67 75 80 541.19–1.31 18 16 11 33�1.31 3 0 2 11

aDifferences between randomized groups before start of hemofiltration were not significant. Values are presented as % of the sessions; bp valuesrepresent the difference between randomized groups at the end of the sessions.

Table 3. Safety, efficacy, and clinical outcomes

Citrate (n � 97) Nadroparin (n � 103) p

SafetyAdverse events needing discontinuation of study anticoagulant, n (%) 2 (2) 20 (19) �0.001

Bleeding, n (%) 6 (6) 16 (16) 0.08Heparin-induced thrombocytopenia, n (%) 3 (3) 4 (3) 0.90Transfusion

Red blood cellsDuring CVVH period, number of patients (%) 56 (58) 62 (60) 0.89Per CVVH day, number of units (IQR) 0.27 (0.0–0.63) 0.36 (0.0–0.83) 0.31

Quarantaine plasma, number of patients (%) 8 (8) 11 (11) 0.63Platelets, number of patients (%) 6 (6) 9 (9) 0.59

Hemoglobin start-end CVVH (mmol/L) (IQR) �0.10 (�0.06 to 0.05)a 0.20 (�0.07 to 0.07)b 0.36Platelet count start-end CVVH (109/L) (IQR) �6 (�32 to �13) �8 (�32 to �9) 0.46

EfficacyCircuit survival time (hr) (IQR) 27 (13–47) 26 (15–43) 0.68Total filtrate volume per session (L) (IQR) 90 (48–158) 86 (52–141) 0.55Total filtrate volume per hour (mL/kg/hr) (IQR) 41 (34–50) 39 (34–46) 0.27Reasons of circuit disconnection (%)Circuit failure/catheter failure/otherc 46/8/46 49/7/44 0.77

Renal recoveryRenal recovery (all patients), n (%) 67 (69) 54 (52) 0.02Renal recovery (surviving patients), n (%) 55 (97) 38 (86) 0.08

Recovery in the intensive care unit/after intensive care unitdischarge (n)

50/5 31/7

Nonrecovery (surviving patients), n (%) 2 (2) 6 (6)Creatinine hospital discharge (mg/dL) (IQR) 1.4 (1.0–2.3)d 1.4 (1.1–2.5)e 0.70

MortalityAll randomized patients

Mortality hospital (%) (95% CI) 42 (31–51) 57 (48–67) 0.02Mortality 3-month (%) (95% CI) 48 (38–58) 63 (53–72) 0.03

Per protocol patientsMortality hospital (%) (95% CI) 41 (31–51) 57 (48–67) 0.03Mortality 3-month (%) (95% CI) 45 (35–55) 62 (53–72) 0.02

CVVH, continuous venovenous hemofiltration; IQR, interquartile range; CI, confidence interval.a�0.16 (�0.097 to 0.081) g/dL; b0.32 (�0.11 to 0.11) g/dL; cOther: transport, death, 72 hr, or clinical decision for interruption anticipating recovery

of renal function; d127 (88–200) �mol/L; e124 (100–221) �mol/L.

548 Crit Care Med 2009 Vol. 37, No. 2

study anticoagulant were analyzed in the ini-tially randomized group.

The data are presented as means (95% con-fidence interval) or medians (interquartileranges [IQRs]). We compared variables usingthe Student’s t test, chi-square (two-sided), orMann-Whitney U test, as appropriate. The ef-fect of study anticoagulant on circuit survivaltime and (time to) death at 3 months wasassessed using Kaplan-Meier estimates withlog-rank testing. For 3-month mortality, weadditionally performed a Cox’s proportionalhazard regression analysis, using stepwise for-ward inclusion of all variables significantlyrelated to 3-month mortality in univariateanalysis (p � 0.05). Variables were removedwhen not significantly improving the model.To explain the unexpected difference in mor-tality, we stratified post hoc for complemen-tary subgroups: medical vs. surgical, sepsis vs.nonsepsis, SOFA score (higher vs. lower thanmedian), and age (higher vs. lower than me-dian). A p value �0.05 was considered statis-tically significant. Because of the explorativenature of the study, we did not correct formultiple testing. We used SPSS 13.0 (SPSS,Chicago, IL) for analysis.

RESULTS

From March 2003 until November2006, 5712 patients were admitted to theintensive care unit, 365 had an indicationfor RRT. Of these, 170 met the exclusioncriteria leaving 215 patients for randomiza-tion. Fifteen patients were subsequentlycensored from per protocol analysis be-cause they fulfilled the exclusion criteria ordid not receive CVVH (Fig. 1). Baselinecharacteristics, timing of initiation ofCVVH, delivered CVVH dose (filtrate flow),and duration of the RRT were similar be-tween groups (Table 1).

Metabolic Control. Nadroparin pa-tients more frequently developed meta-bolic alkalosis and hyperlactatemia,whereas initial hypocalcemia was less of-ten corrected in the citrate patients (Ta-ble 2). Mean, lowest and highest glucoseconcentrations, and insulin dose were notsignificantly different between groups, nei-ther was C-reactive protein (data notshown).

Safety and Efficacy. Citrate was prema-turely discontinued in two patients be-cause of accumulation and early clotting(protocol violation), and nadroparin in 20on account of bleeding (16 patients), se-vere thrombocytopenia (6 patients) orboth (Table 3). Six patients had bleedingin the citrate group. Sites of bleedingwere not different. The difference intransfusion rate was nonsignificant. Cir-cuit survival and the total amount of fil-

trate per circuit were similar (Table 3,Fig. 2), also when adjusted for differentreasons of circuit disconnection or fil-trate flow at disconnection.

Of the 20 patients requiring prema-ture discontinuation of nadroparin, 12continued CVVH with citrate, 5 withoutanticoagulation, and 3 with fondaparinuxor lepirudin. In these 20 patients, themedian number of RBC per CVVH daywas significantly higher than that in theother nadroparin patients (0.65 [IQR0.22–1.38] vs. 0.27 [IQR 0–0.65] [p �0.01]) or in the citrate patients. They hadhigher SOFA scores (13.0, IQR 11.7–15.0)than all other patients (p � 0.005) and ahospital mortality of 80%. Acute Physiol-ogy and Chronic Health Evaluation andSimplified Acute Physiology scores, bodyweight, and urinary output were not dif-ferent.

Secondary Outcomes. Citrate patientshad a higher rate of renal recovery (Table3). Length of mechanical ventilation, in-tensive care unit, and hospital stay werenot different between groups. Among the215 randomized patients, hospital and3-month mortality rates were 15% lowerin the citrate than in the nadroparingroup (Table 3); among the 200 per pro-tocol patients, hospital and 3-monthmortality rates were 16% and 17% lower(Table 3, Fig. 3). To explore whether thesurvival benefit had been consistent intime, we retrospectively calculated mor-tality for each sequential cohort of 50patients. Over the entire study period theabsolute survival benefit for citrate variedbetween 10% and 23%.

Post hoc analysis showed that mortal-ity was lower with citrate in all subgroups(Fig. 3). The difference was significant for

patients after surgery, with sepsis, higherthan median SOFA score (11 points), orlower than median age (73 years).

In none of these subgroups, transfu-sion rates were different between ran-domized groups. Septic patients hadhigher RBC transfusion rates, but rateswere not different between other comple-mentary subgroups. Septic patients hadhigher SOFA scores than nonseptic pa-tients: 11 (IQR 9–14) vs. 10 (IQR 9–12)(p � 0.01). Medical patients had moreoften sepsis than surgical patients: 55%vs. 38% (p � 0.02).

Higher age, severity of illness scores,SOFA score and dopamine dose, morepositive fluid balance, lower creatinineand diuresis, later initiation, and higherRBC transfusion rate were related tomortality at univariate analysis (Table 4),whereas CVVH dose, metabolic alkalosisduring CVVH, type of admission, and thepresence of sepsis, or cardiac failure werenot. The significant variables were in-cluded in a Cox’s proportional regressionanalysis for 3-month mortality. The bestmodel showed that nadroparin anticoag-ulation, age, SOFAstart CVVH, and RBCtransfusion rate were independent pre-dictors of mortality (Table 4).

DISCUSSION

The present single-center randomizedcontrolled trial in critically ill patients withARF without an increased bleeding riskfound that regional anticoagulation withcitrate for CVVH is safer, better toleratedthan, and as effective as anticoagulationwith nadroparin. Unexpectedly, the studyshows a survival benefit for citrate. An ef-fect of anticoagulation for CRRT on sur-

Figure 2. Graphs represent circuit survival. Dotted lines represent the citrate patients; continuouslines represent the nadroparin patients. In the right graph, circuits disconnected for catheter failure,recovery of renal function, (impending) death or logistic reasons are censored. p values (log rank) forboth populations are 0.92.

549Crit Care Med 2009 Vol. 37, No. 2

vival is novel and may confer new in-sight in the pathophysiology of citratefor CRRT.

Although our primary concern waswhether citrate would be safe, the studyindicates that nadroparin was less safethan citrate even though we excludedfour times more patients from study en-rollment for nadroparin-related than forcitrate-related anticipated adverse events.Nevertheless, nadroparin patients tendedto bleed more and nadroparin had to bediscontinued more frequently for adverseevents. Only one patient had signs of ci-trate accumulation. The difference intransfusion between randomized groupswas, however, not significant. This maybe because nadroparin was discontinued

if bleeding developed. Transfusion addi-tionally reflects blood loss from samplingand circuit clotting, which was not differ-ent between groups. There is some con-cern in the literature about the use oflow-molecular weight heparins in CRRT(29). However, RBC transfusion in ournadroparin group was lower than in hep-arin groups of other controlled studies(10, 30), possibly because we did not en-roll patients with a bleeding risk anddiscontinued nadroparin if bleeding de-veloped. Although low-molecularweight heparins accumulate in renalfailure, they are removed with CRRT(31). Whether anti-Xa monitoringwould decrease the risk of bleeding isnot known.

Contrary to expectation, citrate patientshad less metabolic alkalosis than nadropa-rin patients. This may be because citratewas administered in a fixed relation toblood flow and two replacement fluids wereused targeting normal pH. Citrate patientshad lower-ionized calcium, which was adeliberate choice of the protocol.

In this study, efficacy of citrate interms of circuit survival was similar tonadroparin. Although reasons for circuitdisconnections were different betweenpatients, criteria were preset and equalbetween groups. In several (10, 11) butnot all (12, 32) previous studies, citrateconferred a longer circuit survival thanheparin (summarized in Ref. 33). Factorsnegatively influencing circuit life in ourstudy were the use of postdilution hemo-filtration with relatively high filtrationfractions and routine disconnection after72 hours. Also, citrate dose was relativelylow and not titrated to postfilter calcium(33). Finally, calcium replacement in thevenous chamber counteracts citrate antico-agulation facilitating clotting at this site.

An unexpected finding is the survivalbenefit and higher rate of renal recovery inpatients randomized to citrate, presentthroughout the entire study period. Thesurvival benefit was independent from fac-tors known to affect survival such as age,severity of illness, and even transfusionrate. CVVH dose and timing, and fluid bal-ance were not different between studygroups. Citrate appeared particularly bene-ficial in surgical patients, patients with sep-sis, severe organ failure, and in relativelyyounger patients. There are several hypo-thetical explanations. Citrate may be favor-able, nadroparin unfavorable, or both. Ex-planations for the survival benefit may berelated to bleeding, biocompatibility, anti-inflammatory effects, and mitochondrialfuelling.

Less bleeding may be a factor, but can atmost partially explain the benefit. The differ-ence in bleeding between randomized groupswas not significant, transfusion rates werenot different, neither in the entire populationnor in any subgroup, and transfusion ratewasan independent predictor of mortality in ad-dition to study anticoagulant. Patients withbleeding during nadroparin had more severeorgan failure, higher transfusion rates, and anextremely high mortality.

Biocompatibility may be an additionalfactor. In intermittent hemodialysis, ci-trate anticoagulation (compared withheparin or dalteparin) almost completelyabolishes both polymorphonuclear andplatelet degranulation and lowers plasma

Figure 3. Three-month survival of the study patients. Graphs represent survival curves of the perprotocol study patients, and of the post hoc complementary subgroups surgical vs. medical, Sepsis-related Organ Failure Assessment (SOFA) score lower or equal to vs. higher than median (11 points),no sepsis vs. sepsis, age lower or equal to vs. higher than median (73 years). Dotted lines represent thecitrate patients; continuous lines represent the nadroparin patients. Log-rank testing was used tocalculate p values for comparison of survival between groups.

550 Crit Care Med 2009 Vol. 37, No. 2

oxidized-low density lipoprotein indicat-ing less cellular activation and lipid per-oxidation (34). It may be hypothesizedthat hypocalcemia in the filter down-regulates inflammation (35, 36). In oursetting, about 13 L of blood per hoursustain hypocalcemia while passing thefilter. Therefore, hypocalcemia in the fil-ter may not only mitigate membrane in-duced, but also systemic inflammation.An alternative explanation may be thatcitrate is readily available mitochondrialfuel, about 20 g/day. Substrate availabilityis a crucial regulator of the citric acidcycle, a central pathway recovering en-ergy and maintaining redox state (37). In

sepsis, inhibition of pyruvate dehydroge-nase limits pyruvate conversion to acetyl-coenzyme-A, the main substrate of the cy-cle. Finally, heparins may be harmful byreducing the anti-inflammatory effects ofantithrombin (38, 39) and mobilizing my-eloperoxidase from circulating and adheringinflammation-primed leukocytes (40). Theseexplanations need further investigation.

This study has the limitation that it issingle centered. A typical patient popula-tion may bias single-center studies. Theage of our patients is remarkably highand this may explain that mortality inthis study is higher than in previous stud-ies (2, 3, 27). Single-center studies may

on the other hand be a better test situa-tion because cointerventions are stan-dardized, avoiding the confounding effectof nonuniform cointerventions. Althoughhemofiltration prohibits blinding, ran-domization was free from bias, transfu-sion targets were strictly followed, andthe end point mortality is objective. How-ever, the primary end points could havebeen biased despite using a protocol. Itshould further be noted that 12 of the 20patients needing discontinuation of na-droparin continued CVVH with citrate.These nadroparin patients may have bene-fited from citrate but not after having ex-perienced a severe adverse event possiblyrelated to nadroparin. However, all citratepatients received low-dose nadroparin forthrombosis prophylaxis.

This study has the strength that wedid not change daily practice for studypurpose and that more than half of thepatients needing RRT were randomized.Improving survival in this populationwith high mortality is clinically relevant.

Furthermore, this is the first largestudy evaluating the safety of citrate andalthough the mortality benefit was notanticipated, the sample is large enough toshow the present difference. The benefitof citrate existed throughout the entirestudy period, which strengthens the re-sults. Safe performance of citrate anti-coagulation requires the implementa-tion of a local protocol and training.Safety will further improve as a citratemodule is incorporated in the CRRTdevice.

In conclusion, this randomized con-trolled trial in critically ill patients withARF without an increased bleeding riskFigure 3. Continued.

Table 4. Variables significant at univariate analysis and at multivariate Cox’s proportional-hazard analysis for 3-mo mortality

Variable

Univariate Multivariate

R p Hazard Ratio 95% Confidence Interval p

Anticoagulant strategya 0.17 0.02 0.7 (0.45–0.98) 0.04Age (yrs)b 0.28 �0.001 1.06 (1.04–1.08) �0.001Sepsis-Related (later Sequential) Organ

Failure Assessment start CVVHc0.31 �0.001 1.3 (1.17–1.34) �0.001

Number of red blood cell units per CVVH dayb 0.23 0.001 1.5 (1.19–1.96) 0.001Acute Physiology and Chronic Health

Evaluation II scoreb0.17 �0.001

Simplified Acute Physiology II scoreb 0.27 �0.001Dopamine dose start CVVHb 0.21 0.003Fluid balanced 0.18 0.01Creatininestart CVVH

c 0.17 0.02Diuresisstart CVVH

e 0.15 0.03Timing of CVVHf 0.17 0.02

CVVH, continuous venovenous hemofiltration.aThe hazard is lower with citrate; bLower in survivors; cLower in survivors; dLess positive in survivors; eHigher in survivors; fEarlier in survivors.

551Crit Care Med 2009 Vol. 37, No. 2

shows that regional anticoagulation withcitrate for CVVH is safer than anticoagula-tion with nadroparin. Citrate was bettertolerated and provided better acid-basecontrol. Citrate anticoagulation reducedmortality, which can at most partially be at-tributed to less bleeding, and not to less clot-ting. Post hoc analysis suggests that citratemay be particularly beneficial in patients aftersurgery, with sepsis, a high degree of organfailure, or younger age, suggesting a role ofcitrate in the metabolism of inflammation.

REFERENCES1. Metnitz PG, Krenn CG, Steltzer H, et al: Effect of

acute renal failure requiring renal replacementtherapy on outcome in critically ill patients. CritCare Med 2002; 30:2051–2058

2. Ronco C, Bellomo R, Homel P, et al: Effectsof different doses in continuous veno-venoushaemofiltration on outcomes of acute renalfailure: A prospective randomised trial. Lan-cet 2000; 356:26–30

3. Saudan P, Niederberger M, De Seigneux S, et al:Adding a dialysis dose to continuous hemofiltra-tion increases survival in patients with acute re-nal failure. Kidney Int 2006; 70:1312–1317

4. van de Wetering J, Westendorp RG, van derHoeven JG, et al: Heparin use in continuousrenal replacement procedures: The strugglebetween filter coagulation and patient hem-orrhage. J Am Soc Nephrol 1996; 7:145–150

5. Mehta RL, McDonald BR, Aguilar MM, et al:Regional citrate anticoagulation for continu-ous arteriovenous hemodialysis in criticallyill patients. Kidney Int 1990; 38:976–981

6. Chadha V, Garg U, Warady BA, et al: Citrateclearance in children receiving continuousvenovenous renal replacement therapy. Pedi-atr Nephrol 2002; 17:819–824

7. Swartz R, Pasko D, O’Toole J, et al: Improv-ing the delivery of continuous renal replace-ment therapy using regional citrate antico-agulation. Clin Nephrol 2004; 61:134–143

8. Apsner R, Schwarzenhofer M, Derfler K, et al:Impairment of citrate metabolism in acute he-patic failure. Wien Klin Wochenschr 1997; 109:123–127

9. Oudemans-van Straaten HM, Wester JP, dePont AC, et al: Anticoagulation strategies incontinuous renal replacement therapy: Canthe choice be evidence based?. Intensive CareMed 2006; 32:188–202

10. Monchi M, Berghmans D, Ledoux D, et al:Citrate vs. heparin for anticoagulation incontinuous venovenous hemofiltration: Aprospective randomized study. IntensiveCare Med 2004; 30:260–265

11. Kutsogiannis DJ, Gibney RT, Stollery D, et al:Regional citrate versus systemic heparin an-ticoagulation for continuous renal replace-ment in critically ill patients. Kidney Int2005; 67:2361–2367

12. Betjes MG, van Oosterom D, van Agteren M,et al: Regional citrate versus heparin antico-agulation during venovenous hemofiltration

in patients at low risk for bleeding: Similarhemofilter survival but significantly lessbleeding. J Nephrol 2007; 20:602–608

13. Oudemans-van StraatenHM,BosmanRJ, van derSpoel JI, et al: Outcome of critically ill patientstreated with intermittent high-volume haemofil-tration: A prospective cohort analysis. IntensiveCare Med 1999; 25:814–821

14. de Pont AC, Oudemans-van Straaten HM,Roozendaal KJ, et al: Nadroparin versusdalteparin anticoagulation in high-volume,continuous venovenous hemofiltration: Adouble-blind, randomized, crossover study.Crit Care Med 2000; 28:421–425

15. Reeves JH, Cumming AR, Gallagher L, et al:A controlled trial of low-molecular-weightheparin (dalteparin) versus unfractionatedheparin as anticoagulant during continuousvenovenous hemodialysis with filtration. CritCare Med 1999; 27:2224–2228

16. Joannidis M, Kountchev J, RauchenzaunerM, et al: Enoxaparin vs. unfractionated hep-arin for anticoagulation during continuousveno-venous hemofiltration: A randomizedcontrolled crossover study. Intensive CareMed 2007; 33:1571–1579

17. Meier-Kriesche HU, Gitomer J, Finkel K, etal: Increased total to ionized calcium ratioduring continuous venovenous hemodialysiswith regional citrate anticoagulation. CritCare Med 2001; 29:748–752

18. Warkentin TE: Platelet count monitoringand laboratory testing for heparin-inducedthrombocytopenia. Arch Pathol Lab Med2002; 126:1415–1423

19. JansenNJ, vanOeverenW, van denBroek L, et al:Inhibition by dexamethasone of the reperfusionphenomena in cardiopulmonary bypass. J ThoracCardiovasc Surg 1991; 102:515–525

20. Stoutenbeek CP, van Saene HK, Miranda DR,et al: A new technique of infection preven-tion in the intensive care unit by selectivedecontamination of the digestive tract. ActaAnaesthesiol Belg 1983; 34:209–221

21. Knaus WA, Draper EA, Wagner DP, et al:APACHE II: A severity of disease classificationsystem. Crit Care Med 1985; 13:818–829

22. Knaus WA, Wagner DP, Draper EA, et al: TheAPACHE III prognostic system. Risk prediction ofhospital mortality for critically ill hospitalizedadults. Chest 1991; 100:1619–1636

23. Le Gall JR, Lemeshow S, Saulnier F: A newSimplified Acute Physiology Score (SAPS II)based on a European/North American multi-center study. JAMA 1993; 270:2957–2963

24. Vincent JL, Moreno R, Takala J, et al: TheSOFA (Sepsis-related Organ Failure Assess-ment) score to describe organ dysfunction/failure. On behalf of the Working Group onSepsis-Related Problems of the European So-ciety of Intensive Care Medicine. IntensiveCare Med 1996; 22:707–710

25. Bellomo R, Ronco C, Kellum JA, et al: Acuterenal failure - definition, outcome measures,animal models, fluid therapy and informa-tion technology needs: The Second Interna-tional Consensus Conference of the Acute

Dialysis Quality Initiative (ADQI) Group. CritCare 2004; 8:R204–R212

26. Levy MM, Fink MP, Marshall JC, et al: 2001SCCM/ESICM/ACCP/ATS/SIS InternationalSepsis Definitions Conference. Crit Care Med2003; 31:1250–1256

27. Bouman CS, Oudemans-van Straaten HM,Tijssen JG, et al: Effects of early high-volumecontinuous venovenous hemofiltration onsurvival and recovery of renal function inintensive care patients with acute renal fail-ure: A prospective, randomized trial. CritCare Med 2002; 30:2205–2211

28. Verma AK, Levine M, Shalansky SJ, et al:Frequency of heparin-induced thrombocyto-penia in critical care patients. Pharmaco-therapy 2003; 23:745–753

29. Uchino S, Bellomo R, Morimatsu H, et al:Continuous renal replacement therapy: Aworldwide practice survey: The Beginningand Ending Supportive Therapy for the Kid-ney (B.E.S.T. Kidney) Investigators. Inten-sive Care Med 2007; 33:1563–1570

30. Gabutti L, Marone C, Colucci G, et al: Citrateanticoagulation in continuous venovenous he-modiafiltration: A metabolic challenge. IntensiveCare Med 2002; 28:1419–1425

31. Isla A, Gascon AR, Maynar J, et al: In vitroand in vivo evaluation of enoxaparin removalby continuous renal replacement therapieswith acrylonitrile and polysulfone mem-branes. Clin Ther 2005; 27:1444–1451

32. Brophy PD, Somers MJ, Baum MA, et al:Multi-centre evaluation of anticoagulation inpatients receiving continuous renal replace-ment therapy (CRRT). Nephrol Dial Trans-plant 2005; 20:1416–1421

33. Joannidis M, Oudemans-van Straaten HM: Clini-cal review: Patency of the circuit in continuousrenal replacement therapy. Crit Care 2007; 11:218

34. Gritters M, Grooteman MP, Schoorl M, et al:Citrate anticoagulation abolishes degranulationof polymorphonuclear cells and platelets and re-duces oxidative stress during haemodialysis.Nephrol Dial Transplant 2006; 21:153–159

35. Berridge MJ, Bootman MD, Roderick HL: Cal-cium signalling: Dynamics, homeostasis and re-modelling. Nat Rev Mol Cell Biol 2003;4:517–529

36. Hofer AM, Brown EM: Extracellular calciumsensing and signalling. Nat Rev Mol Cell Biol2003; 4:530–538

37. Owen OE, Kalhan SC, Hanson RW: The keyrole of anaplerosis and cataplerosis for citricacid cycle function. J Biol Chem 2002; 277:30409–30412

38. Warren BL, Eid A, Singer P, et al: High-doseantithrombin III in severe sepsis: A randomizedcontrolled trial. JAMA 2001; 286:1869–1878

39. Schouten M, Wiersinga WJ, Levi M, et al: Inflam-mation, endothelium, and coagulation in sepsis.J Leukoc Biol 2008; 83:536–545.

40. Borawski J: Myeloperoxidase as a marker of he-modialysis biocompatibility and oxidative stress:The underestimated modifying effects of heparin.Am J Kidney Dis 2006; 47:37–41

552 Crit Care Med 2009 Vol. 37, No. 2