Comparison of standard and accelerated initiation of renal replacement therapy in acute kidney injury

Transcription

1 International Society of Nephrology clinical trial see commentary on page 670 Comparison of standard and accelerated initiation of renal replacement therapy in acute kidney injury Ron Wald 1,2, Neill K.J. Adhikari 3, Orla M. Smith 2,4, Matthew A. Weir 5, Karen Pope 6, Ashley Cohen 6, Kevin Thorpe 6,7, Lauralyn McIntyre 8, Francois Lamontagne 9, Mark Soth 10, Margaret Herridge 11, Stephen Lapinsky 12, Edward Clark 13, Amit X. Garg 5, Swapnil Hiremath 13, David Klein 2,6,14, C. David Mazer 2,15, Robert M.A. Richardson 16, M. Elizabeth Wilcox 10, Jan O. Friedrich 2,14, Karen E.A. Burns 2,14, Sean M. Bagshaw 17 on behalf of the Canadian Critical Care Trials Group 1 Division of Nephrology, St Michael s Hospital and University of Toronto, Toronto, Ontario, Canada; 2 Li Ka Shing Knowledge Institute of St Michael s Hospital, Toronto, Ontario, Canada; 3 Department of Critical Care Medicine, Sunnybrook Health Sciences Centre and Interdepartmental Division of Critical Care, University of Toronto, Toronto, Ontario, Canada; 4 Department of Critical Care Medicine, St Michael s Hospital, Toronto, Ontario, Canada; 5 Division of Nephrology, London Health Sciences Centre and Western University, London, Ontario, Canada; 6 Applied Health Research Centre, St Michael s Hospital, Toronto, Ontario, Canada; 7 Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada; 8 Division of Critical Care Medicine, The Ottawa Hospital and the University of Ottawa, Ottawa, Ontario, Canada; 9 Division of Critical Care Medicine, Centre Hospitalier Universite de Sherbrooke, Sherbrooke, Quebec, Canada; 10 Division of Critical Care Medicine, St Joseph s Healthcare, Hamilton, Ontario, Canada; 11 Division of Critical Care, University Health Network, Toronto, Ontario, Canada; 12 Division of Critical Care, Mt Sinai Hospital, Toronto, Ontario, Canada; 13 Division of Nephrology, The Ottawa Hospital and the University of Ottawa, Ottawa, Ontario, Canada; 14 Critical Care Department, St Michael s Hospital, Toronto, Ontario, Canada; 15 Department of Anesthesiology, St Michael s Hospital, Toronto, Ontario, Canada; 16 Division of Nephrology, University Health Network, Toronto, Ontario, Canada and 17 Division of Critical Care Medicine, Faculty of Medicine and Dentistry, University of Alberta, Edmonton, Alberta, Canada In patients with severe acute kidney injury (AKI) but no urgent indication for renal replacement therapy (RRT), the optimal time to initiate RRT remains controversial. While starting RRT preemptively may have benefits, this may expose patients to unnecessary RRT. To study this, we conducted a 12-center open-label pilot trial of critically ill adults with volume replete severe AKI. Patients were randomized to accelerated (12 h or less from eligibility) or standard RRT initiation. Outcomes were adherence to protocol-defined time windows for RRT initiation (primary), proportion of eligible patients enrolled, follow-up to 90 days, and safety in 101 fully eligible patients (57 with sepsis) with a mean age of 63 years. Median serum creatinine and urine output at enrollment were 268 micromoles/l and 356 ml per 24 h, respectively. In the accelerated arm, all patients commenced RRT and 45/48 did so within 12 h from eligibility (median 7.4 h). In the standard arm, 33 patients started RRT at a median of 31.6 h from eligibility, of which 19 did not receive RRT (6 died and 13 recovered kidney function). Clinical outcomes were available for all patients at 90 days following enrollment, with mortality 38% in the accelerated and 37% in the standard arm. Two surviving patients, both randomized to standard RRT initiation, were still RRT dependent at day 90. Correspondence: Ron Wald, Division of Nephrology, St Michael s Hospital and University of Toronto, 61 Queen Street East, 9-140, Toronto, Ontario M5C 2T2, Canada. waldr@smh.ca Received 30 January 2015; revised 22 April 2015; accepted 30 April 2015; published online 8 July 2015 No safety signal was evident in either arm. Our findings can inform the design of a large-scale effectiveness randomized control trial. Kidney International (2015) 88, ; doi: /ki ; published online 8 July 2015 KEYWORDS: acute kidney injury; randomized controlled trial; renal replacement therapy There is considerable uncertainty regarding the optimal timing of renal replacement therapy (RRT) initiation in critically ill patients with acute kidney injury (AKI). 1 Although the need to initiate RRT is unequivocal in patients with life-threatening AKI and complications that are refractory to medical measures, the advantages of commencing RRT in the absence of such complications are debatable. Earlier initiation of RRT may confer benefit through accelerated achievement of euvolemia, removal of toxic solutes, achievement of acid-base homeostasis, and prevention of overt complications attributable to AKI. On the other hand, spontaneous recovery of kidney function may occur in selected patients with severe AKI. Earlier initiation in these patients unnecessarily exposes them to the potential harms of vascular access (for example, hemorrhage, thrombosis, bacteremia) and the complications of RRT (for example, intradialytic hypotension, hypersensitivity to the extracorporeal circuit, clearance of trace elements, and antibiotics) along with added resource utilization. Kidney International (2015) 88,

2 clinical trial R Wald et al.: Standard versus accelerated initiation of RRT in AKI The preponderance of data, mostly derived from observational studies with variable definitions of early RRT initiation, suggests that earlier initiation of RRT is associated with lower all-cause mortality. 2 The few randomized controlled trials (RCTs) to explore this question have been inconclusive, owing to small sample size and limited generalizability to the wide population of critically ill patients. 3,4 We have developed a research program to establish whether earlier initiation of RRT in the setting of severe AKI, but in the absence of pressing indications, improves survival compared with a more conservative approach of watchful waiting and supportive care in which RRT is initiated only when classic indications arise. 5 7 We report the results of a pilot RCT that was designed to evaluate the feasibility and safety of our protocol. RESULTS We identified 1811 patients who met preliminary screening criteria, of whom 262 were provisionally eligible. Attending clinicians determined that 46 patients required urgent initiation of RRT, and that 84 patients required deferral of RRT. A description of the patients excluded by attending clinicians is found in Supplementary Appendix 1 online. Of the remaining 132 patients, 101 (77%) were enrolled in the trial. Twenty patients were enrolled via deferred consent, and none of these patients subsequently withdrew participation (Figure 1). One patient died between the acquisition of consent and randomization. Patient characteristics Patient characteristics at the time of randomization are summarized in Table 1 and were similar in both treatment arms. RRT initiation Accelerated RRT initiation arm. All 48 patients assigned to accelerated RRT initiation commenced RRT. The median time from meeting full eligibility criteria to RRT initiation was 7.4 h (interquartile range ; mean 9.7 ± 12.0 h; Table 2 and Figure 2). Of these, 45 (94%) patients commenced RRT within the prescribed 12 h window. Delays in RRT initiation beyond the 12 h window occurred in 3 patients for the following reasons: hemodynamic instability (resulting in a 1 h and 25 min delay), vascular access insertion difficulties (resulting in an 8 h and 15 min delay), and a postrandomization decision by the attending nephrologist to defer RRT and prescribe furosemide, although RRT was ultimately commenced (resulting in an 86 h delay). Standard RRT initiation arm. Among the 52 participants randomized to standard RRT initiation, no patient started RRT within 12 h of fulfilling eligibility. In the 14 days following randomization, 33 (63%) participants commenced RRT with a median time of 31.6 h (interquartile range, ; mean 51.6 ± 52.0 h) from the time of eligibility (Table 2 and Figure 2). Six (12%) patients died before RRT commenced, whereas 13 (25%) recovered from AKI or maintained adequate kidney function with no need to start Accelerated (n = 48) 0 Participant crossover 48 (100%) Followed up at day 90 Patients screened Provisionally eligible 262 Fully eligible 132 Consented 101 Standard (n = 52) 0 Participant crossover 52 (100%) Followed up at day Excluded as RRT initiation believed to be mandated 84 Excluded as RRT deferral believed to be mandated 23 Patient/SDM declined or 12 h window elapsed 8 Not enrolled for other reasons* 1 Patient died before randomization Figure 1 Patient flow through the trial. *Reasons for nonenrollment included participation in other trials (n=2), discharge from ICU (n=2), lack of NGAL supplies (n=2), nephrology team not available (n=1), lack of health insurance (n=1). ICU, intensive care unit; NGAL, neutrophil gelatinase-associated lipocalin; RRT, renal replacement therapy; SDM, substitute decision maker. RRT. The baseline characteristics of the participants enrolled in the standard arm stratified by receipt of RRT, death, and recovery are displayed in Supplementary Appendix 2 online. Individuals who remained alive and recovered kidney function without RRT were younger, had a lower SOFA (Sequential Organ Failure Assessment) score, were less likely to be septic, and had lower serum creatinine and urea concentrations at the time of enrollment as compared with individuals who commenced RRT. Baseline neutrophil gelatinase-associated lipocalin (NGAL) values were similar in patients who commenced RRT and those who recovered without RRT. The initiation of RRT was triggered by a protocol-driven criterion in a minority (11/33) of patients who commenced RRT in the standard arm (PaO2/FiO2 o200 in 7 patients, serum potassium concentration 46 mmol/l in 1 patient, and persistent AKI for 472 h in 3 patients). RRT initiation was based on clinician discretion in the remaining 22 patients; the most common reason cited for starting RRT was a perception of volume overload and/or oligoanuria. 898 Kidney International (2015) 88,

3 R Wald et al.: Standard versus accelerated initiation of RRT in AKI clinical trial Table 1 Baseline characteristics Accelerated (n = 48) Standard (n = 52) Age, years 62.2 ± ± 13.6 Women 13(27) 15(29) Baseline egfr, ml/min per 74.8 ± ± m 2 Hypertension 23 (48) 30 (58) Diabetes mellitus 19 (40) 20 (38) Heart failure 4 (8) 7 (13) Cardiopulmonary bypass in 2(4) 4(8) past 7 days Sepsis 26 (54) 30 (58) Receipt of IV contrast in past 11 (23) 12 (23) 7days SOFA score 13.3 ± ± 3.0 The 24-h urine output, ml a 355 ( ) 434 ( ) Cumulative fluid balance, ml b 5144 ( ) 4821 ( ) Mechanically ventilated 45 (94) 48 (92) Receipt of vasopressor(s) 42 (88) 43 (83) Serum creatinine, μmol/l 286 ( ) 250 ( ) Whole-blood NGAL, ng/ml ( ) ( ) Urea b, mmol/l 17.7 ( ) 16.9 ( ) Potassium a, mmol/l 4.3 ± ± 0.6 Bicarbonate b, mmol/l 20.1 ± ± 4.2 Hemoglobin, g/l 92 ± ± 19 White blood cells, 10 9 /l 16.1 ± ± 12.0 Abbreviations: egfr, estimated glomerular filtration rate; IV, intravenous; NGAL, neutrophil gelatinase-associated lipocalin; SOFA score, sequential organ failure assessment score. Conversion for serum creatinine: 1 mg/dl = 88.4 μmol/l. Conversion for serum urea: 1 mg/dl = mmol/l. Continuous data are presented as means ± s.d. or medians (interquartile range), as appropriate. Categorical data presented as number (%). a Data missing on one patient. b Data missing on three patients. At the time of RRT initiation, serum creatinine and urea concentrations were both higher in the standard RRT initiation arm as compared with the accelerated arm (Table 2). All other clinical and biochemical parameters were similar at the time of RRT initiation. Once RRT commenced, the features of the delivered therapies were similar in both treatment arms (Table 3). Safety We observed 11 serious adverse events among 11 patients and 44 adverse events among 19 patients in the accelerated RRT arm. In the standard RRT initiation arm, there were 18 serious adverse events in 15 patients and 56 adverse events in 23 patients (Table 4). RRT-related hemodynamic instability and other RRT-related complications were similar in both treatment arms (Table 4). Clinical outcomes We ascertained the vital status for all enrolled patients at day 90. A total of 18 (38%) patients randomized to accelerated RRT died compared with 19 (37%) in the standard RRT arm (Table 5). None of the surviving patients in the accelerated arm were RRT dependent at day 90 as compared with 2/52 (4%) patients randomized to standard RRT initiation. Additional clinical outcomes are displayed in Table 5. Table 2 Patient characteristics at the time of renal replacement therapy initiation Accelerated (n = 48) Standard (n = 33) Mean time from eligibility 9.73 ± ± to RRT, h Median time from eligibility 7.38 ( ) ( ) to RRT, h Cumulative fluid balance, ml a 5085 ( ) 5146 ( ) Urine output, ml 400 ( ) 265 (80 755) SOFA score 12.0 ± ± 2.0 Serum creatinine, μmol/l a 326 ± ± 193 Urea, mmol/l b 19.3 ± ± 11.3 Potassium, mmol/l 4.2 ± ± 0.7 Bicarbonate, mmol/l 20.7 ± ± 4.4 Hemoglobin, g/l 90 ± ± 14 White blood cells, 10 9 /l a 14.6 ± ± 7.4 Abbreviations: BUN, blood urea nitrogen; RRT, renal replacement therapy; SOFA score, sequential organ failure assessment score. Conversion for serum creatinine: 1 mg/dl = 88.4 μmol/l. Conversion for serum urea: 1 mg/dl = mmol/l. Continuous data are presented as means ± s.d. or medians (interquartile range), as appropriate. Categorical data presented as number (%). a Data missing on one patient. b Data missing on 14 patients. DISCUSSION In this pilot RCT, we demonstrated the feasibility of conducting a large definitive trial comparing two strategies for RRT initiation among critically ill patients with AKI. Despite the time constraints associated with protocol implementation, we enrolled 77% of eligible patients with 97% adherence to the window for RRT initiation and achieved 100% follow-up to 90 days. Though adverse events were common and expected in this population of critically ill patients, no salient differences were observed between the groups. This study complements other trials that have been completed or that are ongoing. Bouman et al. 3 randomized 106 patients with AKI to early (soon after meeting criteria for AKI) or delayed (following development of hyperkalemia, pulmonary edema, or plasma urea 440 mmol/l) RRT initiation. There was no difference in mortality; however, many patients had cardiac surgery-associated AKI, thus limiting generalizability. A recent trial enrolled 208 patients (mean age 42 years) with community-acquired AKI (450% associated with tropical infections or obstetric complications) at a single center in India. 4 In the early arm, RRT was started after serum urea exceeded 25 mmol/l or serum creatinine exceeded 619 μmol/l regardless of other AKI complications. In the usual care arm, RRT was only initiated in the setting of medically refractory hyperkalemia, acidosis, or volume overload or if uremic symptoms developed. No differences in mortality or kidney recovery were observed. The ongoing IDEAL-ICU French multicenter RCT is randomizing critically ill patients (target n = 824) with septic shock and severe AKI (defined as a threefold rise in serum creatinine and urine output o0.3 ml/kg per h for 12 h) to RRT initiation within 12 h of eligibility (early arm) versus RRT initiation after h (later arm). 8 In addition, the AKIKI (French Artificial Kidney International (2015) 88,

4 clinical trial R Wald et al.: Standard versus accelerated initiation of RRT in AKI Kidney Initiation in Kidney Injury) trial is randomizing 620 patients with severe AKI to immediate RRT initiation versus initiation of RRT only when an urgent indication arises ( These studies and ours in the AKI setting parallel work that has been done to elucidate the optimal time for dialysis initiation in the chronic kidney disease population. The IDEAL trial demonstrated that a strategy of initiating dialysis in patients with higher endogenous kidney function was not superior to a strategy of waiting for kidney function to decline to lower levels. 9 There are theoretical benefits for the preemptive initiation of RRT in critically ill patients with AKI. This strategy avoids life-threatening complications of uremia, acidosis, hyperkalemia, and volume overload, and a meta-analysis of mostly observational studies suggested a reduction in mortality (odds ratio 0.45, 95% confidence interval ). 2 Though compelling, observational data are limited by heterogeneous indicators of timing (for example, serum urea, AKI stage, 13 and duration of stay in the intensive care unit 5,14 ), nonrandom allocation of exposures, and failure to include patients with severe AKI who did not receive RRT (that is, treatment selection bias). 15 Enthusiasm for adoption of a preemptive/early strategy of RRT initiation is tempered by the fact that RRT is associated Proportion without RRT Time to RRT initiation Standard RRT initiation Accelerated RRT initiation Time to RRT (h) Figure 2 Time from eligibility to renal replacement therapy (RRT) initiation stratified by treatment strategy (Po0.0001). with risks and is resource intensive. Systematic preemptive RRT may be unnecessary for those patients with severe AKI who are destined to recover, as was observed in approximately one-quarter of the patients in this trial allocated to standard RRT initiation. Notwithstanding these issues, it appears that earlier or preemptive RRT initiation may have already permeated current practice as classic indications, at least those designated as triggers for RRT initiation in the standard arm of our protocol, are infrequently present at RRT initiation in contemporary practice. 16,17 The controversy surrounding this topic and the variability in clinical practice because of the lack of reliable evidence highlight the pressing need to determine the optimal timing of RRT initiation in an adequately powered RCT. This research priority has been articulated by the Acute Kidney Injury Network, 18 the Kidney Disease improving Global Table 4 Safety events occurring during the first 14 days Accelerated Standard Serious adverse events (total) Serious adverse events/patient Characterization of serious adverse events Death 6 7 Arrhythmia 0 2 Hypotension 3 3 Hemorrhage 1 3 Hypocalcemia 0 1 Ischemic bowel 0 1 Sepsis 1 1 Adverse events (total) Adverse events/patient RRT-related complications Episodes of hemodynamic instability with RRT RRT-associated arrhythmia 1 5 RRT-associated seizure 2 0 Days with serum potassium o3 mmol/l 3 7 Days with serum phosphate o0.5 mmol/l 2 4 Days with total calcium o2.00 mmol/l or ionized calcium o1.00 mmol/l Hemorrhage with CVC insertion 0 0 Pneumothorax with CVC insertion 1 1 CVC-associated bacteremia 0 1 CVC-associated thrombus 0 0 Inadvertent arterial puncture during CVC 1 0 insertion Other CVC-related events 1 0 Abbreviations: CVC, central venous catheter; RRT, renal replacement therapy. Table 3 Characteristics of renal replacement therapy delivered during the 14 days from randomization Accelerated (n = 48, 256 sessions) Standard (n =33, 216 sessions) First modality IHD 15 (31) 11 (34) First modality CRRT or SLED 33 (69) 22 (66) Mean dose prescribed CRRT, ml/kg per h 28.6 ± ± 28.6 Ultrafiltration per session a, ml 2303 ± ± 1032 Last RRT day 6 (5 9) 7 (6 10) Fluid balance at day 14 since enrollment, ml 1336 ( 5834 to 1627) 57 ( 4355 to 5366) Abbreviations: CRRT, continuous renal replacement therapy; IHD, intermittent hemodialysis; RRT, renal replacement therapy; SLED, sustained low efficiency dialysis. Continuous data are presented as means± s.d. or medians (interquartile range), as appropriate. Categorical data presented as number (%). a RRT session defined as each 24 h period during which CRRT was administered or a single session of SLED or IHD. 900 Kidney International (2015) 88,

5 R Wald et al.: Standard versus accelerated initiation of RRT in AKI clinical trial Table 5 Clinical outcomes Accelerated RRT (n = 48) Standard RRT (n =52) P-value Death in ICU 13 (27) 16 (31) 0.69 Death in hospital 16 (33) 19 (37) 0.74 Death by day (38) 19 (37) 0.92 Alive and dialysis dependent at day (6) 0.16 Duration of ICU stay among survivors, days 11 (8 29.5) 13.5 (8 32) 0.93 Duration of hospitalization among survivors, days 29 (20 49) 31 (20 51) 0.93 Abbreviations: ICU, intensive care unit; RRT, renal replacement therapy. Continuous data are presented as mean ± s.d. or median (interquartile range) as appropriate. Categorical data presented as number (%). Outcomes Clinical Practice Guidelines, 1 and the National Institute for Health and Care Excellence in the United Kingdom. 19 Nephrologists and intensivists have also expressed broad support in surveys. 6,20 This pilot trial has several strengths. We readily enrolled the target population and achieved all predefined feasibility objectives, including commencement of RRT within 12 h for almost all patients randomized to accelerated RRT initiation, despite the challenges of timely consent, central venous catheter insertion, and RRT machine set-up. In the standard arm, RRT was appropriately delayed resulting in a clinically important median time difference of 24 h (mean 42 h) between the groups in the initiation of RRT. We enrolled patients with AKI of varying etiologies and targeted patients to whom the question of RRT timing is relevant by including only those with no need for imminent RRT. Given the limited evidence surrounding the optimal dose, modality, and mode of RRT, we allowed clinicians to administer RRT as per usual local practice. This pragmatic approach acknowledged the wide range of clinically acceptable approaches to RRT administration, facilitated the participation of multiple centers without disrupting local practices, and will enhance the likelihood of broad international acceptance that will be required for a future large definitive trial. The clinical course of patients allocated to standard RRT initiation provided important insights into the divergent outcomes when applying a conservative RRT strategy in patients with AKI. Approximately one-quarter of patients randomized to standard RRT initiation recovered kidney function spontaneously, suggesting that a strategy of accelerated RRT might expose some patients to unnecessary RRT. Using biochemical criteria and input from attending clinicians, we tried to enrich our cohort with patients who were most likely to eventually receive RRT, recognizing that it is challenging to reliably predict worsening AKI or kidney recovery. 21 Nonetheless, clinicians were relatively successful at identifying patients whose AKI was destined to recover and were asked to exclude such patients from the trial. Such patients had less severe AKI and illness acuity at the time of screening and, indeed, over 80% of these patients survived to hospital discharge without the initiation of RRT. There are several limitations to consider. Although we incorporated a novel kidney damage biomarker (NGAL) to define severe AKI that might anticipate the need for RRT, its contribution proved to be limited. Prior data showed that NGAL was a biomarker of early AKI 22,23 that was strongly associated with progression of AKI 24 and the subsequent receipt of RRT. 25,26 As such, we believed that an NGAL of 400 ng/ml would enrich our enrolled population with individuals having a higher likelihood of requiring RRT. Moreover, NGAL concentration can be measured at the point of care within min, making this biomarker suitable for a trial testing a time-sensitive intervention. We found NGAL values to be universally elevated, highlighting the successful recruitment of patients with severe AKI; however, it did not discriminate well between individuals in the standard RRT arm who did and did not go on to commence RRT, perhaps because of the high frequency of sepsis in both treatment arms that may independently raise serum NGAL concentration. 27 The requirement of a central venous pressure of 8 mm Hg was intended to enroll patients whose AKI was not volume sensitive and would be unlikely to improve with fluid resuscitation alone. However, accumulating data have cast doubt on the practice of determining fluid responsiveness based on central venous pressure measurements. 28 As a result, our exclusion of individuals with nonelevated central venous pressure (or who lacked a central line making the determination of the central venous pressure impossible) might have compromised generalizability because of the exclusion of some patients for whom the research question posed by the trial was relevant. Finally, although we aimed to use objective criteria for screening and trial enrollment, we mandated that patients who were provisionally eligible be confirmed for participation by the attending physicians. Although this subjective assessment may potentially have led to the systematic exclusion of patients with a possible center or physician effect, allocation concealment would have prevented the clinicians ability to anticipate the treatment arm of the next enrolled patient. Ultimately, we believed that the opportunity for a clinician veto was necessary for physician acceptance of the trial; this phenomenon has recently been characterized in another trial in the critical care setting. 29 Conclusions This pilot RCT of critically ill patients with severe AKI demonstrates the feasibility of implementing a protocol comparing accelerated RRT initiation to a conservative Kidney International (2015) 88,

6 clinical trial R Wald et al.: Standard versus accelerated initiation of RRT in AKI strategy triggered by clinical indications and physician judgment. The experiences gleaned from the pilot will guide the design and implementation of a large-scale RCT that will determine whether accelerated RRT initiation confers improved survival in AKI. MATERIALS AND METHODS Study design We conducted an open-label parallel arm feasibility RCT in intensive care units at 12 sites in Canada between May 2012 and November 2013 (see Supplementary Appendix 3 online for participating sites). The trial was registered at clinicaltrials.gov (NCT ) and full details of the protocol have been published. 7 We established an independent Data Safety Monitoring Board to monitor the conduct of the trial. Ethics The research ethics boards at each participating site approved the trial. We obtained written informed consent from the patient or substitute decision maker. At four sites a deferred consent process was permitted such that participants could be randomized if they lacked capacity and did not have an available substitute decision maker. Participants enrolled under this condition were asked to provide formal consent once they regained capacity, at which time they had the opportunity to withdraw. Patient recruitment Trained coordinators screened the intensive care unit population 1 2 times per day on weekdays to identify patients who met the eligibility criteria. In preliminary screening, we identified adults ( 18 years of age) admitted to an intensive care unit with evidence of kidney dysfunction (defined as a serum creatinine 100 μmol/l for women or 130 μmol/l for men). We then assessed additional inclusion criteria (all of which needed to be fulfilled) including: (1) presence of severe AKI (defined by the presence of two of the following three criteria: (i) a twofold increase in serum creatinine from baseline, (ii) urine output o6 ml/kg in the preceding 12 h, or (iii) whole-blood NGAL 400 ng/ ml); (2) the absence of urgent indications for RRT initiation (defined as serum potassium 5.5 mmol/l and serum bicarbonate 15 mmol/ l); and (3) low likelihood of volume-responsive AKI (defined as central venous pressure 8 mm Hg). We excluded patients with any of the following characteristics: lack of commitment to ongoing life support, including RRT; presence of an intoxication requiring extracorporeal removal; RRT within the previous 2 months; clinical suspicion of renal obstruction, rapidly progressive glomerulonephritis, or interstitial nephritis; prehospitalization estimated glomerular filtration rate o30 ml/min per 1.73 m 2 ; and the passage of 448 h since doubling of baseline serum creatinine. If all of the inclusion criteria were met and no exclusion criteria were present, we considered the participant to be provisionally eligible. To confirm that equipoise existed regarding the trial intervention, we then asked the attending intensivist and nephrologist if they believed that either immediate RRT initiation or RRT deferral was mandated. If either physician answered affirmatively to either question, we excluded the patient. If both physicians answered no to both questions, we considered the participant fully eligible and efforts to obtain consent commenced. We measured whole-blood NGAL concentration for all participants, even if this component was not utilized as an inclusion criterion. A ml sample of blood was drawn from an indwelling arterial or central venous catheter into an EDTA-filled test tube. Blood was then transferred to a test strip (Alere, San Diego, CA) and measured by fluorescence immunoassay on a point-of-care platform (Triage MeterPro, Alere). Randomization and interventions We randomized patients 1:1 to accelerated or standard RRT initiation using randomly permuted blocks of size 2 and 4, stratified by center. For participants randomized to accelerated RRT initiation, we took measures to ensure prompt establishment of vascular access and initiation of RRT within 12 h of the patient fulfilling study eligibility. This 12 h window included the time required to obtain consent. In the standard RRT initiation arm, we discouraged initiation of RRT unless one of the following conditions developed: serum potassium 46.0 mmol/l, serum bicarbonate o10 mmol/l, or PaO 2 / FiO 2 o200 with infiltrates on chest radiograph compatible with pulmonary edema. Clinicians were still free to commence RRT (as long as the initial 12 h window from eligibility elapsed) but we asked that they declare their reason for starting RRT if none of the aforementioned conditions was present. If AKI persisted for 72 h after eligibility was met, clinicians were free to initiate RRT at their discretion without declaring a reason. RRT delivery Once RRT commenced, it was delivered using identical principles in both treatment arms. We made nonbinding recommendations regarding the modality choice, RRT intensity, and other aspects of the RRT prescription based on current best practice standards (Supplementary Appendix 4 online). 1 RRT was to continue until patient death, change in goals of care, or recovery of kidney function (defined in Supplementary Appendix 5 online). Follow-up Following randomization, we tracked all patients daily for 14 days to collect RRT prescription and administration details (if applicable), severity of illness as measured by the modified SOFA score 30 (Supplementary Appendix 6 online), laboratory data, and physiologic data. We followed all patients for 90 days to ascertain vital status and RRT requirement. Outcomes The primary outcome was the proportion of patients in each arm who commenced RRT within the protocol-specified window ( 12 and 412 h from eligibility in the accelerated and standard groups, respectively). Our a priori feasibility target called for 90% of patients who would start RRT to do so within these prespecified windows. We also evaluated the time difference from eligibility to RRT initiation between the groups. We considered the following secondary outcomes: (1) the proportion of patients successfully consented among those fully eligible (feasibility target 50%), (2) the proportion of patients followed to day 90 (feasibility target 95%), and (3) serious adverse events and adverse events attributable to study participation including RRT-associated complications and vascular access-associated complications (defined in Supplementary Appendix 7 online). Sample size determination We targeted recruitment of 100 patients (50/arm) that would enable the evaluation of 90% protocol adherence in each group with a 95% confidence interval of 79 96%; we considered the lower limit of this 902 Kidney International (2015) 88,

7 R Wald et al.: Standard versus accelerated initiation of RRT in AKI clinical trial confidence interval to be an acceptable level of adherence in the context of a large trial. Statistical analyses All analyses adhered to an intention-to-treat principle. As this was a feasibility study, most data included were descriptive and minimal inferential statistics were planned. The proportion of patients who met the various feasibility targets were described as a percentage. Wilson confidence intervals were calculated. Continuous variables were described as means (with s.d.) or medians (with interquartile range). Continuous data such as physiological and laboratory variables and duration of stay of patients were compared between treatment groups using the Wilcoxon test. Categorical variables were described as a number (%). Proportional outcomes were compared between treatment groups using Pearson s χ 2 test. Kaplan Meier estimates were used to estimate the survivor function of time to RRT. Patients who died or survived to 90 days without RRT were considered censored. A log-rank test was used to compare the time to RRT between treatment groups. All analyses were performed with R (R Core Team, Vienna, Austria, 2013). DISCLOSURE RW and SMB have served as paid consultants to Baxter and have received speaker fees from Baxter. They have also received unrestricted grant support from Baxter, in partnership with the Canadian Institutes of Health Research, to fund a multicenter international randomized controlled trial to evaluate whether the timing of renal replacement therapy initiation in RRT modifies patient survival. RW has received speaker fees from Alere. All the other authors declared no competing interests. ACKNOWLEDGMENTS We appreciate the efforts of the data safety monitoring board members: Stuart Goldstein (Chair), Eric Hoste, and Patrick Murray. We are grateful for the thoughtful review and input provided by the Grants and Manuscripts committee of the Canadian Critical Care Trials Group. Some of the data in this paper were presented in abstract form at the American Society of Nephrology Kidney Week Late- Breaking Clinical Trials Session in November 2014 (Philadelphia, PA). This study was funded by grants from the Canadian Institutes of Health Research and the University of Alberta Hospital Foundation. Alere provided the Triage MeterPro that was used to measure wholeblood NGAL. SMB holds a Canada Research Chair in Critical Care Nephrology and is a Clinical Investigator supported by Alberta Innovates Health Solutions (AI-HS). SUPPLEMENTARY MATERIAL Appendix 1. Characteristics and outcomes of patients who were provisionally eligible for STARRT-AKI but excluded based on clinician judgment. Appendix 2. Baseline characteristics of patients in the standard arm stratified by receipt of renal replacement therapy within 14 days of randomization. Appendix 3. Research team by study site. Appendix 4. Guidelines for administration of RRT in the STARRT-AKI Trial. Appendix 5. Cessation of RRT. Appendix 6. Calculation of the modified SOFA score in the STARRT-AKI trial. Appendix 7. Definitions of safety outcomes related to administration of RRT or vascular access for RRT. Supplementary material is linked to the online version of the paper at REFERENCES 1. KDIGO Clinical Practice Guideline for Acute Kidney Injury. Kidney Int 2012; 2: Karvellas CJ, Farhat MR, Sajjad I et al. A comparison of early versus late initiation of renal replacement therapy in critically ill patients with acute kidney injury: a systematic review and meta-analysis. Crit Care 2011; 15: R Bouman CS, Oudemans-Van Straaten HM, Tijssen JG et al. Effects of early high-volume continuous venovenous hemofiltration on survival and recovery of renal function in intensive care patients with acute renal failure: a prospective, randomized trial. Crit Care Med 2002; 30: Jamale TE, Hase NK, Kulkarni M et al. Earlier-start versus usual-start dialysis in patients with community-acquired acute kidney injury: a randomized controlled trial. Am J Kidney Dis 2013; 62: Bagshaw SM, Uchino S, Bellomo R et al. Timing of renal replacement therapy and clinical outcomes in critically ill patients with severe acute kidney injury. J Crit Care 2009; 24: Clark E, Wald R, Walsh M et al. Timing of initiation of renal replacement therapy for acute kidney injury: a survey of nephrologists and intensivists in Canada. Nephrol Dial Transplant 2011; 27: Smith OM, Wald R, Adhikari NK et al. Standard versus accelerated initiation of renal replacement therapy in acute kidney injury (STARRT-AKI): study protocol for a randomized controlled trial. Trials 2013; 14: Barbar SD, Binquet C, Monchi M et al. Impact on mortality of the timing of renal replacement therapy in patients with severe acute kidney injury in septic shock: the IDEAL-ICU study (initiation of dialysis early versus delayed in the intensive care unit): study protocol for a randomized controlled trial. Trials 2014; 15: Cooper BA, Branley P, Bulfone L et al. A randomized, controlled trial of early versus late initiation of dialysis. N Engl J Med 2010; 363: Gettings LG, Reynolds HN, Scalea T. Outcome in post-traumatic acute renal failure when continuous renal replacement therapy is applied early vs. late. Intensive Care Med 1999; 25: Liu KD, Himmelfarb J, Paganini E et al. Timing of initiation of dialysis in critically ill patients with acute kidney injury. Clin J Am Soc Nephrol 2006; 1: Carl DE, Grossman C, Behnke M et al. Effect of timing of dialysis on mortality in critically ill, septic patients with acute renal failure. Hemodial Int 2010; 14: Shiao CC, Wu VC, Li WY et al. Late initiation of renal replacement therapy is associated with worse outcomes in acute kidney injury after major abdominal surgery. Crit Care 2009; 13: R Modem V, Thompson M, Gollhofer D et al. Timing of continuous renal replacement therapy and mortality in critically ill children*. Crit Care Med 2014; 42: Ostermann M, Dickie H, Barrett NA. Renal replacement therapy in critically ill patients with acute kidney injury when to start. Nephrol Dial Transplant 2012; 27: Bagshaw SM, Wald R, Barton J et al. Clinical factors associated with initiation of renal replacement therapy in critically ill patients with acute kidney injury-a prospective multicenter observational study. J Crit Care 2012; 27: Vaara Suvi T, Matti Reinikainen, Ron Wald et al. Timing of RRT Based on the Presence of Conventional Indications. Clin J Am Soc Nephrol 2014; 9: Kellum JA, Mehta RL, Levin A et al. Development of a clinical research agenda for acute kidney injury using an international, interdisciplinary, three-step modified Delphi process. Clin J Am Soc Nephrol 2008; 3: National Institute for Health and Care Excellence Acute kidney injury: prevention, detection and management of acute kidney injury up to the point of renal replacement therapy. National Institute for Health and Care Excellence Thakar CV, Rousseau J, Leonard AC. Timing of dialysis initiation in AKI in ICU: international survey. Crit Care 2012; 16: R Wilson FP. A policy of preemption: the timing of renal replacement therapy in AKI. Clin J Am Soc Nephrol 2014; 9: Mishra J, Dent C, Tarabishi R et al. Neutrophil gelatinase-associated lipocalin (NGAL) as a biomarker for acute renal injury after cardiac surgery. Lancet 2005; 365: Shapiro NI, Trzeciak S, Hollander JE et al. The diagnostic accuracy of plasma neutrophil gelatinase-associated lipocalin in the prediction of Kidney International (2015) 88,

8 clinical trial R Wald et al.: Standard versus accelerated initiation of RRT in AKI acute kidney injury in emergency department patients with suspected sepsis. Ann Emerg Med 2010; 56: 52 9 e Koyner JL, Garg AX, Coca SG et al. Biomarkers predict progression of acute kidney injury after cardiac surgery. J Am Soc Nephrol 2012; 23: Haase M, Bellomo R, Devarajan P et al. Accuracy of neutrophil gelatinaseassociated lipocalin (NGAL) in diagnosis and prognosis in acute kidney injury: a systematic review and meta-analysis. Am J Kidney Dis 2009; 54: de Geus HR, Bakker J, Lesaffre EM et al. Neutrophil gelatinase-associated lipocalin at ICU admission predicts for acute kidney injury in adult patients. Am J Respir Crit Care Med 2010; 183: Kumpers P, Hafer C, Lukasz A et al. Serum neutrophil gelatinaseassociated lipocalin at inception of renal replacement therapy predicts survival in critically ill patients with acute kidney injury. Crit Care 2010; 14: R Marik PE, Cavallazzi R. Does the central venous pressure predict fluid responsiveness? An updated meta-analysis and a plea for some common sense. Crit Care Med 2013; 41: Cook D, Arabi Y, Ferguson N et al. Physicians declining patient enrollment in a critical care trial: a case study in thromboprophylaxis. Intensive Care Med 2013; 39: Ferreira FL, Bota DP, Bross A et al. Serial evaluation of the SOFA score to predict outcome in critically ill patients. JAMA 2001; 286: Kidney International (2015) 88,