[Skip to Content]
Access to paid content on this site is currently suspended due to excessive activity being detected from your IP address 34.236.190.216. Please contact the publisher to request reinstatement.
[Skip to Content Landing]
Figure.
Net Ultrafiltration (NUF) Rate and Survival From Gray Model
Net Ultrafiltration (NUF) Rate and Survival From Gray Model

Hazard ratios (blue solid lines) are shown with 95% CIs (blue dotted lines). The orange line indicates a hazard ratio of 1. A hazard ratio less than 1 suggests that the NUF rate is associated with lower mortality, and a hazard ratio greater than 1 suggests that the NUF is associated with higher mortality. A, The risk of death associated with an NUFrate greater than 1.75 mL/kg/h compared with an NUFrate slower than 1.01 mL/kg/h was 51% for day 7 to 12, 52% for day 13 to 26, and 66% for day 27 to 90. B, An NUFrate from 1.01 to 1.75 mL/kg/h was not associated with death. C, For an NUFrate greater than 1.75 mL/kg/h compared with an NUFrate from 1.01 to 1.75 mL/kg/h, the risk of death was 44% for day 7 to 12, 42% for day 13 to 26, and 77% for day 27 to 90. D, Every 0.50-mL/kg/h increase in NUFrate was associated with death: 5% for day 3 to 6, 8% for day 7 to 12, 11% for day 13 to 26, and 13% for day 27 to 90.

Table 1.  
Baseline Patient Characteristics by NUF Rate
Baseline Patient Characteristics by NUF Rate
Table 2.  
Processes of Care During NUF and Outcomes
Processes of Care During NUF and Outcomes
Table 3.  
Association of NUF With Survival From Gray Model
Association of NUF With Survival From Gray Model
Table 4.  
Summary of Complications by NUF Rate
Summary of Complications by NUF Rate
Supplement.

eAppendix 1. The RENAL Cohort

eAppendix 2. Statistical Method for Imputation of Missing Premorbid Creatinine Values and Estimated Glomerular Filtration Rate

eAppendix 3. Multivariable Gray Piecewise-Constant Time-Varying Coefficients Regression Model for Time to Mortality

eAppendix 4. Joint Model

eAppendix 5. Statistical Methods for Propensity Score Estimation and Matching

eAppendix 6. Multivariable Logistic Regression Model

eTable 1. Number of Deaths Within Each Time Interval From Primary Gray Model

eTable 2. Characteristics of Patients With Missing Treatment Duration During Continuous Venovenous Hemodiafiltration

eTable 3. Patient Characteristics Before and After Propensity Score Matching

eTable 4. Association of NUF Rate With Mortality From Primary Gray Model

eTable 5. Sensitivity Analyses of NUF Rate and Mortality

eTable 6. Association of NUF Rate With Risk-Adjusted Mortality From Logistic Regression

eTable 7. Association of NUF Rate With Risk-Adjusted Mortality Using NUF as a Continuous Variable From Logistic Regression

eTable 8. Subgroup Analyses of NUF Rate and Mortality

eFigure 1. Distributions of Imputed and Unimputed Premorbid Serum Creatinine Levels and Estimated Glomerular Filtration Rate

eFigure 2. Association of NUF Rate With Crude 90-Day Mortality

eFigure 3. Violation of Proportional Hazards Assumptions by NUF Rate Variable

eFigure 4. Distribution of Propensity Scores for Patients Who Received NUF Rate >1.75 mL/kg/hr Among Matched (N=405) and Unmatched Patients (N=532)

eFigure 5. Logistic Regression Model Calibration and Discrimination

eFigure 6. Crude 90-Day Mortality and Predicted Risk of Death by NUF Rate

eFigure 7. Association of NUF Rate With 90-Day Survival

eReferences

1.
Balakumar  V, Murugan  R, Sileanu  FE, Palevsky  P, Clermont  G, Kellum  JA.  Both positive and negative fluid balance may be associated with reduced long-term survival in the critically ill.  Crit Care Med. 2017;45(8):e749-e757. doi:10.1097/CCM.0000000000002372PubMedGoogle ScholarCrossref
2.
Vaara  ST, Korhonen  AM, Kaukonen  KM,  et al; FINNAKI Study Group.  Fluid overload is associated with an increased risk for 90-day mortality in critically ill patients with renal replacement therapy: data from the prospective FINNAKI study.  Crit Care. 2012;16(5):R197. doi:10.1186/cc11682PubMedGoogle ScholarCrossref
3.
Kidney Disease Improving Global Outcome (KDIGO) Group. KDIGO Clinical Practice Guideline for Acute Kidney Injury. https://kdigo.org/wp-content/uploads/2016/10/KDIGO-2012-AKI-Guideline-English.pdf. Accessed May 1, 2019.
4.
Palevsky  PM, Liu  KD, Brophy  PD,  et al.  KDOQI US commentary on the 2012 KDIGO Clinical Practice Guideline for Acute Kidney Injury.  Am J Kidney Dis. 2013;61(5):649-672. doi:10.1053/j.ajkd.2013.02.349PubMedGoogle ScholarCrossref
5.
Bouchard  J, Soroko  SB, Chertow  GM,  et al; Program to Improve Care in Acute Renal Disease (PICARD) Study Group.  Fluid accumulation, survival and recovery of kidney function in critically ill patients with acute kidney injury.  Kidney Int. 2009;76(4):422-427. doi:10.1038/ki.2009.159PubMedGoogle ScholarCrossref
6.
Bellomo  R, Cass  A, Cole  L,  et al; RENAL Replacement Therapy Study Investigators.  An observational study of fluid balance and patient outcomes in the Randomized Evaluation of Normal vs. Augmented Level of Replacement Therapy trial.  Crit Care Med. 2012;40(6):1753-1760. doi:10.1097/CCM.0b013e318246b9c6PubMedGoogle ScholarCrossref
7.
Burton  JO, Jefferies  HJ, Selby  NM, McIntyre  CW.  Hemodialysis-induced repetitive myocardial injury results in global and segmental reduction in systolic cardiac function.  Clin J Am Soc Nephrol. 2009;4(12):1925-1931. doi:10.2215/CJN.04470709PubMedGoogle ScholarCrossref
8.
Silversides  JA, Pinto  R, Kuint  R,  et al.  Fluid balance, intradialytic hypotension, and outcomes in critically ill patients undergoing renal replacement therapy: a cohort study.  Crit Care. 2014;18(6):624. doi:10.1186/s13054-014-0624-8PubMedGoogle ScholarCrossref
9.
Murugan  R, Balakumar  V, Kerti  SJ,  et al.  Net ultrafiltration intensity and mortality in critically ill patients with fluid overload.  Crit Care. 2018;22(1):223.PubMedGoogle ScholarCrossref
10.
Flythe  JE, Kimmel  SE, Brunelli  SM.  Rapid fluid removal during dialysis is associated with cardiovascular morbidity and mortality.  Kidney Int. 2011;79(2):250-257. doi:10.1038/ki.2010.383PubMedGoogle ScholarCrossref
11.
Kim  TW, Chang  TI, Kim  TH,  et al.  Association of ultrafiltration rate with mortality in incident hemodialysis patients.  Nephron. 2018;139(1):13-22. doi:10.1159/000486323PubMedGoogle ScholarCrossref
12.
Movilli  E, Gaggia  P, Zubani  R,  et al.  Association between high ultrafiltration rates and mortality in uraemic patients on regular haemodialysis: a 5-year prospective observational multicentre study.  Nephrol Dial Transplant. 2007;22(12):3547-3552. doi:10.1093/ndt/gfm466PubMedGoogle ScholarCrossref
13.
Saran  R, Bragg-Gresham  JL, Levin  NW,  et al.  Longer treatment time and slower ultrafiltration in hemodialysis: associations with reduced mortality in the DOPPS.  Kidney Int. 2006;69(7):1222-1228. doi:10.1038/sj.ki.5000186PubMedGoogle ScholarCrossref
14.
Bellomo  R, Cass  A, Cole  L,  et al; RENAL Replacement Therapy Study Investigators.  Intensity of continuous renal-replacement therapy in critically ill patients.  N Engl J Med. 2009;361(17):1627-1638. doi:10.1056/NEJMoa0902413PubMedGoogle ScholarCrossref
15.
von Elm  E, Altman  DG, Egger  M, Pocock  SJ, Gøtzsche  PC, Vandenbroucke  JP; STROBE Initiative.  The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.  Ann Intern Med. 2007;147(8):573-577. doi:10.7326/0003-4819-147-8-200710160-00010PubMedGoogle ScholarCrossref
16.
Psaty  BM, Koepsell  TD, Lin  D,  et al.  Assessment and control for confounding by indication in observational studies.  J Am Geriatr Soc. 1999;47(6):749-754. doi:10.1111/j.1532-5415.1999.tb01603.xPubMedGoogle ScholarCrossref
17.
Siew  ED, Peterson  JF, Eden  SK, Moons  KG, Ikizler  TA, Matheny  ME.  Use of multiple imputation method to improve estimation of missing baseline serum creatinine in acute kidney injury research.  Clin J Am Soc Nephrol. 2013;8(1):10-18. doi:10.2215/CJN.00200112PubMedGoogle ScholarCrossref
18.
Buuren  S, Groothius-Oudshoorn  K.  MICE: multivariate imputation by chained equations in R.  J Stat Softw. 2011;45(3):1-67. doi:10.18637/jss.v045.i03Google ScholarCrossref
19.
Závada  J, Hoste  E, Cartin-Ceba  R,  et al; AKI6 Investigators.  A comparison of three methods to estimate baseline creatinine for RIFLE classification.  Nephrol Dial Transplant. 2010;25(12):3911-3918. doi:10.1093/ndt/gfp766PubMedGoogle ScholarCrossref
20.
Gray  RJ.  Flexible methods for analyzing survival data using splines, with applications to breast cancer prognosis.  J Am Stat Assoc. 1992;87(420):942-951. doi:10.1080/01621459.1992.10476248Google ScholarCrossref
21.
Gray  RJ.  Spline-based tests in survival analysis.  Biometrics. 1994;50(3):640-652. doi:10.2307/2532779PubMedGoogle ScholarCrossref
22.
Kasal  J, Jovanovic  Z, Clermont  G,  et al.  Comparison of Cox and Gray’s survival models in severe sepsis.  Crit Care Med. 2004;32(3):700-707. doi:10.1097/01.CCM.0000114819.37569.4BPubMedGoogle ScholarCrossref
23.
Valenta  Z, Weissfeld  L.  Estimation of the survival function for Gray’s piecewise-constant time-varying coefficients model.  Stat Med. 2002;21(5):717-727. doi:10.1002/sim.1061PubMedGoogle ScholarCrossref
24.
Rizopoulos  D.  Dynamic predictions and prospective accuracy in joint models for longitudinal and time-to-event data.  Biometrics. 2011;67(3):819-829. doi:10.1111/j.1541-0420.2010.01546.xPubMedGoogle ScholarCrossref
25.
Wulfsohn  MS, Tsiatis  AA.  A joint model for survival and longitudinal data measured with error.  Biometrics. 1997;53(1):330-339. doi:10.2307/2533118PubMedGoogle ScholarCrossref
26.
Burton  JO, Jefferies  HJ, Selby  NM, McIntyre  CW.  Hemodialysis-induced cardiac injury: determinants and associated outcomes.  Clin J Am Soc Nephrol. 2009;4(5):914-920. doi:10.2215/CJN.03900808PubMedGoogle ScholarCrossref
27.
McIntyre  CW, Harrison  LE, Eldehni  MT,  et al.  Circulating endotoxemia: a novel factor in systemic inflammation and cardiovascular disease in chronic kidney disease.  Clin J Am Soc Nephrol. 2011;6(1):133-141. doi:10.2215/CJN.04610510PubMedGoogle ScholarCrossref
28.
Shivalkar  B, Flameng  W, Szilard  M, Pislaru  S, Borgers  M, Vanhaecke  J.  Repeated stunning precedes myocardial hibernation in progressive multiple coronary artery obstruction.  J Am Coll Cardiol. 1999;34(7):2126-2136. doi:10.1016/S0735-1097(99)00467-2PubMedGoogle ScholarCrossref
29.
McMullen  JR, Sherwood  MC, Tarnavski  O,  et al.  Inhibition of mTOR signaling with rapamycin regresses established cardiac hypertrophy induced by pressure overload.  Circulation. 2004;109(24):3050-3055. doi:10.1161/01.CIR.0000130641.08705.45PubMedGoogle ScholarCrossref
30.
Ritz  E, Wanner  C.  The challenge of sudden death in dialysis patients.  Clin J Am Soc Nephrol. 2008;3(3):920-929. doi:10.2215/CJN.04571007PubMedGoogle ScholarCrossref
31.
Palevsky  PM, Zhang  JH, O’Connor  TZ,  et al; VA/NIH Acute Renal Failure Trial Network.  Intensity of renal support in critically ill patients with acute kidney injury.  N Engl J Med. 2008;359(1):7-20.PubMedGoogle ScholarCrossref
32.
Demirjian  S, Teo  BW, Guzman  JA,  et al.  Hypophosphatemia during continuous hemodialysis is associated with prolonged respiratory failure in patients with acute kidney injury.  Nephrol Dial Transplant. 2011;26(11):3508-3514. doi:10.1093/ndt/gfr075PubMedGoogle ScholarCrossref
33.
Aubier  M, Murciano  D, Lecocguic  Y,  et al.  Effect of hypophosphatemia on diaphragmatic contractility in patients with acute respiratory failure.  N Engl J Med. 1985;313(7):420-424. doi:10.1056/NEJM198508153130705PubMedGoogle ScholarCrossref
34.
Lim  C, Tan  HK, Kaushik  M.  Hypophosphatemia in critically ill patients with acute kidney injury treated with hemodialysis is associated with adverse events.  Clin Kidney J. 2017;10(3):341-347.PubMedGoogle Scholar
35.
Yang  Y, Zhang  P, Cui  Y,  et al.  Hypophosphatemia during continuous veno-venous hemofiltration is associated with mortality in critically ill patients with acute kidney injury.  Crit Care. 2013;17(5):R205. doi:10.1186/cc12900PubMedGoogle ScholarCrossref
36.
Flythe  JE, Curhan  GC, Brunelli  SM.  Shorter length dialysis sessions are associated with increased mortality, independent of body weight.  Kidney Int. 2013;83(1):104-113. doi:10.1038/ki.2012.346PubMedGoogle ScholarCrossref
37.
Flythe  JE, Curhan  GC, Brunelli  SM.  Disentangling the ultrafiltration rate-mortality association: the respective roles of session length and weight gain.  Clin J Am Soc Nephrol. 2013;8(7):1151-1161. doi:10.2215/CJN.09460912PubMedGoogle ScholarCrossref
Limit 200 characters
Limit 25 characters
Conflicts of Interest Disclosure

Identify all potential conflicts of interest that might be relevant to your comment.

Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.

Err on the side of full disclosure.

If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.

Not all submitted comments are published. Please see our commenting policy for details.

Limit 140 characters
Limit 3600 characters or approximately 600 words
    Views 8,944
    Original Investigation
    Critical Care Medicine
    June 7, 2019

    Association of Net Ultrafiltration Rate With Mortality Among Critically Ill Adults With Acute Kidney Injury Receiving Continuous Venovenous Hemodiafiltration: A Secondary Analysis of the Randomized Evaluation of Normal vs Augmented Level (RENAL) of Renal Replacement Therapy Trial

    Author Affiliations
    • 1The Clinical Research, Investigation, and Systems Modeling of Acute Illness (CRISMA) Center, Department of Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
    • 2The Center for Critical Care Nephrology, Department of Critical Care Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
    • 3Department of Biostatistics, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania
    • 4Department of Medicine, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
    • 5The George Institute for Global Health and University of Sydney, Sydney, New South Wales, Australia
    • 6Renal Section, Veterans Affairs Pittsburgh Healthcare System, Pittsburgh, Pennsylvania
    • 7Department of Intensive Care Medicine, Austin Hospital, The University of Melbourne, Melbourne, Victoria, Australia
    JAMA Netw Open. 2019;2(6):e195418. doi:10.1001/jamanetworkopen.2019.5418
    Key Points español 中文 (chinese)

    Question  Is the net ultrafiltration (ie, fluid removal) rate associated with survival among critically ill patients with acute kidney injury?

    Findings  In this secondary analysis of a randomized clinical trial involving 1434 critically ill patients treated with continuous venovenous hemodiafiltration, a net ultrafiltration rate greater than 1.75 mL/kg/h compared with a net ultrafiltration rate less than 1.01 mL/kg/h was significantly associated with lower 90-day risk-adjusted survival.

    Meaning  Among critically ill patients with acute kidney injury being treated with continuous venovenous hemodiafiltration, net ultrafiltration rates greater than 1.75 mL/kg/h were associated with increased mortality.

    Abstract

    Importance  Net ultrafiltration (NUF) is frequently used to treat fluid overload among critically ill patients, but whether the rate of NUF affects outcomes is unclear.

    Objective  To examine the association of NUF with survival among critically ill patients with acute kidney injury being treated with continuous venovenous hemodiafiltration.

    Design, Setting, and Participants  The Randomized Evaluation of Normal vs Augmented Level (RENAL) of Renal Replacement Therapy trial was conducted between December 30, 2005, and November 28, 2008, at 35 intensive care units in Australia and New Zealand among critically ill adults with acute kidney injury who were being treated with continuous venovenous hemodiafiltration. This secondary analysis began in May 2018 and concluded in January 2019.

    Exposures  Net ultrafiltration rate, defined as the volume of fluid removed per hour adjusted for patient body weight.

    Main Outcomes and Measures  Risk-adjusted 90-day survival.

    Results  Of 1434 patients, the median (interquartile range) age was 67.3 (56.9-76.3) years; 924 participants (64.4%) were male; median (interquartile range) Acute Physiology and Chronic Health Evaluation III score was 100 (84-118); and 634 patients (44.2%) died. Using tertiles, 3 groups were defined: high, NUF rate greater than 1.75 mL/kg/h; middle, NUF rate from 1.01 to 1.75 mL/kg/h; and low, NUF rate less than 1.01 mL/kg/h. The high-tertile group compared with the low-tertile group was not associated with death from day 0 to 6. However, death occurred in 51 patients (14.7%) in the high-tertile group vs 30 patients (8.6%) in the low-tertile group from day 7 to 12 (adjusted hazard ratio [aHR], 1.51; 95% CI, 1.13-2.02); 45 patients (15.3%) in the high-tertile group vs 25 patients (7.9%) in the low-tertile group from day 13 to 26 (aHR, 1.52; 95% CI, 1.11-2.07); and 48 patients (19.2%) in the high-tertile group vs 29 patients (9.9%) in the low-tertile group from day 27 to 90 (aHR, 1.66; 95% CI, 1.16-2.39). Every 0.5-mL/kg/h increase in NUF rate was associated with increased mortality (3-6 days: aHR, 1.05; 95% CI, 1.00-1.11; 7-12 days: aHR, 1.08; 95% CI, 1.02-1.15; 13-26 days: aHR, 1.11; 95% CI, 1.04-1.18; 27-90 days: aHR, 1.13; 95% CI, 1.05-1.22). Using longitudinal analyses, increase in NUF rate was associated with lower survival (β = .056; P < .001). Hypophosphatemia was more frequent among patients in the high-tertile group compared with patients in the middle-tertile group and patients in the low-tertile group (high: 308 of 477 patients at risk [64.6%]; middle: 293 of 472 patients at risk [62.1%]; low: 247 of 466 patients at risk [53.0%]; P < .001). Cardiac arrhythmias requiring treatment occurred among all groups: high, 176 patients (36.8%); middle: 175 patients (36.5%); and low: 147 patients (30.8%) (P = .08).

    Conclusions and Relevance  Among critically ill patients, NUF rates greater than 1.75 mL/kg/h compared with NUF rates less than 1.01 mL/kg/h were associated with lower survival. Residual confounding may be present from unmeasured risk factors, and randomized clinical trials are required to confirm these findings.

    Trial Registration  ClinicalTrials.gov identifier: NCT00221013

    Introduction

    Fluid overload is a frequent complication present in more than two-thirds of critically ill patients with acute kidney injury and is independently associated with mortality.1,2 When fluid overload is resistant to treatment with diuretics, international practice guidelines recommend net ultrafiltration (NUF).3,4 These recommendations are supported by studies suggesting that NUF could reduce the number of deaths.5,6 However, uncertainty exists about the optimal rate of NUF in critically ill patients.

    A slower NUF rate is associated with prolonged exposure to tissue edema and organ dysfunction, whereas a faster rate is associated with hemodynamic stress.7,8 Both complications could decrease survival. A single-center observational study of critically ill patients receiving continuous venovenous hemodiafiltration (CVVHDF) and hemodialysis9 found that an NUF rate less than 20 mL/kg/d was associated with higher mortality compared with an NUF rate greater than 25 mL/kg/d. In contrast, emerging evidence from outpatients with end-stage renal disease receiving hemodialysis suggests that an NUF rate greater than 13 mL/kg/h per session compared with an NUF rate of 10 mL/kg/h or less is associated with mortality.10-13 However, the implications and generalizability of these study findings to patients undergoing continuous NUF are unclear.

    Thus, we performed a secondary analysis of the Randomized Evaluation of Normal vs Augmented Level (RENAL) of Renal Replacement Therapy clinical trial14 of critically ill patients treated with CVVHDF. In this study, we examine the association of NUF rate with risk-adjusted 90-day survival as well as adverse events during treatment.

    Methods

    We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.15 This retrospective cohort study was approved with a waiver of informed consent by the University of Pittsburgh’s Human Research Protection Office. Written informed consent was obtained from the patient or responsible surrogate by means of either a priori or delayed consent in the RENAL trial.14

    Population

    The RENAL study was a multicenter randomized clinical trial that compared the efficacy of 2 different intensities of solute control using CVVHDF in critically ill patients with acute kidney injury.14 The study was conducted in 35 intensive care units (ICUs) in Australia and New Zealand from December 30, 2005, to November 28, 2008. This secondary analysis was performed from May 31, 2018, to January 31, 2019. In brief, patients were eligible to participate in the study if they were critically ill adults with acute kidney injury, were deemed to require CVVHDF by a clinician, and fulfilled predefined criteria, including oliguria, severe organ edema, hyperkalemia, uremia, and/or severe metabolic acidosis (eAppendix 1 in the Supplement). The NUF rate was left to clinician judgment and was performed by decreasing the flow of the replacement fluid and the dialysate in equal proportion, so that effluent fluid volume exceeded replacement fluid and dialysate volumes.

    Variables

    The primary outcome was 90-day survival from study enrollment. The exposure variable was NUF rate, defined as the volume of net ultrafiltrate removed per hour, adjusted for patient body weight in kilograms. The hourly NUF volume was calculated after excluding the dialysate and replacement fluid volumes from the volume of ultrafiltrate (ie, NUF volume = ultrafiltrate volume − [replacement fluid + dialysate volume]). Subsequently, rate for duration of treatment was calculated using the following equation9: NUF rate (in milliliters per kilogram per hour) = cumulative NUF volume (in milliliters) / (weight at study enrollment [in kilograms] × treatment duration [in hours]). Daily patient fluid balance was first calculated as the difference between fluid administered (ie, intravenous fluids, blood products, enteral fluids, dialysate, and replacement fluid) and fluid lost (ie, dialysis effluent from CVVHDF, urine output, blood loss, enteral losses, and drain losses). We then excluded NUF volume from the output fluids since it was the exposure variable.9 Daily and cumulative fluid balance data were obtained from study enrollment until death, ICU discharge, or 28 days after enrollment. Complications and other adverse events were recorded during treatment.14

    We measured covariates for a risk-adjustment model to account for confounding by indication16 because patients who were older, sicker, and hemodynamically unstable would be expected to receive a lower rate and patients with organ edema and those requiring mechanical ventilation would be expected to receive a higher rate. These confounders included prespecified variables based on clinical experience and prior studies,5,6,9 including age category; female sex; premorbid estimated glomerular filtration rate (eGFR) based on most recent serum creatinine level, if known; duration from ICU admission to study enrollment; severity of illness assessed by Acute Physiology and Chronic Health Evaluation III (APACHE-III) score category (range, 0-299, with higher score indicating more severe illness) in the 24 hours prior to study enrollment; severity of organ dysfunction assessed by total Sequential Organ Failure Assessment (SOFA) score (range, 0-4 for each organ, with higher score indicating more severe organ dysfunction); presence of organ edema, sepsis, and use of mechanical ventilation; daily mean cardiovascular SOFA score during treatment; cumulative fluid balance from enrollment to ICU discharge; duration of CVVHDF in days; source of admission, including whether the patient was transferred from an emergency department, hospital ward, operating room after elective or emergency surgery, another hospital, or another ICU; hospital type; and hospital region. Race and ethnicity were not reported in the randomized clinical trial.

    For patients with unknown premorbid serum creatinine levels (637 [44.4%]), we used the multivariable imputation by chained equation method to impute creatinine values using age, sex, and weight as predictors (eAppendix 2 in the Supplement).17,18 We subsequently used the Modification of Diet in Renal Disease Study equation to determine eGFR using the imputed creatinine levels.19 There was no difference in distribution of imputed and unimputed creatinine and corresponding eGFR values (eFigure 1 in the Supplement).

    Statistical Analysis

    We examined several models, including linear, spline, median, tertiles, and quartiles, because of the nonlinear association of NUF rate with 90-day mortality (eFigure 2 in the Supplement). We selected tertiles owing to having the lowest Akaike information criterion. Thus, we stratified NUF rates into 3 groups: (1) low, less than 1.01 mL/kg/h; (2) middle, 1.01 to 1.75 mL/kg/h; and (3) high, greater than 1.75 mL/kg/h. Categorical variables are presented as numbers and percentages and compared using χ2 tests. Continuous variables are presented as medians and interquartile ranges (IQRs) and compared using the Wilcoxon rank sum test.

    Multivariable modeling of the association of NUF rate with survival was performed using Gray piecewise-constant time-varying coefficients regression (eAppendix 3 and eFigure 3 in the Supplement).20-23 We estimated risk-adjusted hazard ratios (aHRs) and their 95% CIs at 5 time intervals and 4 nodes (0-2 days, 3-6 days, 7-12 days, 13-26 days, and 27-90 days). The number of time intervals were selected based on prior work,9 and the duration of each time interval was selected by the model to ensure approximately equal distribution of deaths within each time interval (eTable 1 in the Supplement).21

    In these models, we used an NUF rate less than 1.01 mL/kg/h as the reference and adjusted for covariates with fixed effects for region and hospital type to account for nonindependence of NUF across hospitals. Models were fitted using 1341 patients (93.5%) after excluding patients with missing covariate data on source of ICU admission (91 [6.3%]), number of CVVHDF treatment days (1 [0.1%]), and mortality (1 [0.1%]). To predict the risk of death across a range of NUF rates, we restricted the cohort to NUF rate of 5 mL/kg/h or less (1428 patients [99.6%]). We also fitted similar models for patients with only available premorbid serum creatinine levels (797 [55.6%]). We performed longitudinal analyses using joint models to account for correlation between daily NUF rate and cardiovascular SOFA score over time and its association with survival (eAppendix 4 in the Supplement).24,25

    We assessed the robustness of findings in multiple sensitivity analyses. First, using propensity scores, we matched patients with NUF rates greater than 1.75 mL/kg/h on a 1:1 basis with patients with NUF rates of 1.75 mL/kg/h or less (eAppendix 5 and eFigure 4 in the Supplement). Second, we examined alternative thresholds by lowering and increasing the values by 0.05 mL/kg/h (ie, <0.96 mL/kg/h, 0.96-1.70 mL/kg/h, >1.70 mL/kg/h and <1.06 mL/kg/h, 1.06-1.80 mL/kg/h, >1.80 mL/kg/h). Third, we restricted rate to the first 72 hours of CVVHDF. Fourth, we classified patients using maximum NUF rate.

    Fifth, we excluded 92 patients with NUF rates less than 0.01 mL/kg/h. Sixth, we included 31 patients with missing treatment duration data by assigning 2 NUF rates (0 mL/kg/h and 1.43 mL/kg/h [the mean NUF rate]) before fitting the model. Seventh, we varied the time interval by moving the nodes in the Gray model 1 day higher (0-3 days, 4-7 days, 8-13 days, 14-27 days, and 28-90 days) and 1 day lower (0-1 day, 2-5 days, 6-11 days, 12-25 days, and 26-90 days). Eighth, we adjusted for individual baseline liver, coagulation, and respiratory SOFA scores instead of total SOFA scores. Ninth, we adjusted for use of red blood cells, fresh frozen plasma, platelets, cryoprecipitate, 20% albumin, and cumulative protein supplementation. Tenth, we adjusted for cumulative fluid balance that included NUF volume in output fluid calculation. Eleventh, we excluded cumulative fluid balance from the model.

    Twelfth, we stratified based on CVVHDF for less than 3 days and 3 or more days as well as less than 5 days and 5 or more days and among patients with negative daily fluid balance during ICU stay. Thirteenth, we fitted logistic regression with rate as a categorical variable after excluding collinearity using variance inflation factor (eAppendix 6 and eFigure 5 in the Supplement). To predict the risk of death across a range of rates, we restricted the analysis to NUF rates of 5 mL/kg/h or greater and used predictive margins adjusted for covariates to predict death for every 0.5-mL/kg/h increase in rate (eFigure 6 in the Supplement). Using subgroup analyses, we assessed for effect modification with a test for interaction in the Gray model between NUF rate and patient characteristics in prespecified subpopulations, including patients with and without organ edema; sepsis; premorbid eGFR less than 60 mL/min/1.73m2 and 60 mL/min/1.73m2 or greater; cardiovascular SOFA score less than 3 and 3 or higher; and high and low intensity CVVHDF.

    Statistical analyses were performed using SAS 9.4 (SAS Institute) and Stata version 15.1 for Windows (StataCorp). Gray and joint models were performed using R version 2.14.0 (The R Foundation) and multiple imputation using the MICE command in R version 3.4.2. All hypothesis tests were 2-tailed with a statistical significance level of P < .05.

    Results
    Patient Population and Characteristics

    Of 1508 patients enrolled in the RENAL trial,14 consent was withdrawn by 43 patients. Of the remaining 1465 patients, we excluded 31 patients for whom the treatment hours for CVVHDF were missing (eTable 2 in the Supplement). Of 1434 eligible patients, 477 patients (33.3%) received NUF less than 1.01 mL/kg/h, 479 patients (33.4%) received NUFfrom 1.01 to 1.75 mL/kg/h, and 478 patients (33.3%) received NUF greater than 1.75 mL/kg/h. Median (IQR) age was 67.3 (56.9-76.3) years; median (IQR) body weight was 80.0 (70.0-90.0) kg; and 924 patients (64.4%) were men (Table 1). Median (IQR) premorbid eGFR was 53.0 (32.6-73.9) mL/min/1.73 m2; median (IQR) APACHE-III score was 100 (84-118); and 634 patients (44.2%) died. Median (IQR) NUF rates within the 3 tertiles were as follows: 0.52 (0.06-0.79) mL/kg/h in the lowest tertile; 1.38 (1.19-1.55) mL/kg/h in the middle tertile; and 2.21 (1.96-2.68) mL/kg/h in the highest tertile (mean [SD] NUF rate, 1.43 [0.97] mL/kg/h).

    Net ultrafiltration rates greater than 1.75 mL/kg/h were associated with young age, female sex, lower body weight, higher eGFR, mechanical ventilation, and longer ICU stay than patients receiving NUF rates from 1.01 to 1.75 mL/kg/h or less than 1.01 mL/kg/h. Patients with NUF rates greater than 1.75 mL/kg/h had more severe organ dysfunction, as evidenced by higher median (IQR) total SOFA score (high: 8 [6-10]; middle: 8 [6-9]; low: 7 [5-9]; P = .001) and organ edema (high: 255 patients [53.3%]; middle: 207 patients [43.2%]; low: 172 patients [36.1%]; P < .001). There were also variations in source of admission to the ICU, type of hospital, country, and region (Table 1).

    Following initiation of CVVHDF, patients with NUF rates greater than 1.75 mL/kg/h were associated with having a similar median (IQR) cardiovascular SOFA score (high: 3 [1 to 4]; middle: 3 [1 to 4]; low: 3 [1 to 4]; P = .05; Table 2) and longer median (IQR) duration of treatment with CVVHDF (high: 7 [4 to 17] days; middle: 6 [3 to 11] days; low: 3 [2 to 6] days; P < .001). Patients receiving NUFgreater than 1.75 mL/kg/h had higher negative median (IQR) daily fluid balances (high: −658.0 [−1445.0 to 55.0] mL/d ; middle: −55.5 [−678.5 to 490.0] mL/d; low: 641.0 [−91.0 to 1793.0] mL/d; P < .001) and median (IQR) cumulative fluid balances (high: −3.6 [−7.7 to 0.4] L; middle: −0.4 [−3.5 to 2.4] L; low: 2.3 [−0.2 to 5.5] L; P < .001). Median (IQR) hourly NUF rates and cumulative NUF rates were also higher for patients receiving NUFgreater than 1.75 mL/kg/h for the duration of CVVHDF (hourly NUF rate: high: 167.4 [141.5 to 203.0] mL/h; middle: 106.0 [89.5 to 131.0] mL/h; low: 31.8 [0.7 to 62.5] mL/h; P < .001; cumulative NUF rate: high: 16.5 [8.5 to 28.4] L; middle: 8.5 [4.5 to 16.4] L; low: 1.7 [0.1 to 4.0] L; P < .001).

    Association of NUF Rate With Outcomes

    Patients with NUF rates greater than 1.75 mL/kg/h had a longer median (IQR) duration of mechanical ventilation (high: 7 [3-12] days; middle: 6 [2-12] days; low: 3 [1-7] days; P < .001), ICU length of stay (high: 9 [5-16] days; middle: 8 [4-15] days; low: 5 [2-10] days; P < .001), and hospital length of stay (high: 10 [6-16] days; middle: 9 [5-16] days; low: 6 [3-11] days; P < .001) compared with patients in the lowest and middle tertile (Table 2). A greater proportion of patients receiving NUF greater than 1.75 mL/kg/h were dependent on dialysis by day 28 (high: 64 of 292 surviving patients [21.9%]; middle: 37 of 317 surviving patients [11.7%]; low: 19 of 290 surviving patients [6.6%]; P < .001); however, there was no difference by day 90 (high: 17 of 246 surviving patients [6.9%]; middle: 17 of 291 surviving patients [5.8%]; low: 10 of 263 surviving patients [3.8%]; P = .28). This was primarily owing to higher mortality among patients receiving NUFgreater than 1.75 mL/kg/h (high: 232 patients [48.6%]; middle: 188 patients [39.2%]; low: 214 patients [44.9%]; P = .01) (eFigure 7 in the Supplement).

    Compared with NUF rates less than 1.01 mL/kg/h, NUF rates greater than 1.75 mL/kg/h were associated with lower survival, which was variable and yet persisted from day 7 to day 90. During this period, death occurred in 51 patients (14.7%) treated with NUF greater than 1.75 mL/kg/h compared with 30 patients (8.6%) treated with NUF less than 1.01 mL/kg/h from day 7 to 12 (aHR, 1.51; 95% CI, 1.13-2.02); 45 patients (15.3%) treated with NUF greater than 1.75 mL/kg/h compared with 25 patients (7.9%) treated with NUF less than 1.01 mL/kg/h from day 13 to day 26 (aHR, 1.52; 95% CI, 1.11-2.07); and 48 patients (19.2%) treated with NUF greater than 1.75 mL/kg/h compared with 29 patients (9.9%) treated with NUF less than 1.01 mL/kg/h from day 27 to 90 (aHR, 1.66; 95% CI, 1.16-2.39) (Table 3; Figure) (eTable 1 and eTable 4 in the Supplement). This association was not attenuated by organ edema strata, sepsis, eGFR less than 60 mL/in/1.73 m2, mean cardiovascular SOFA score of 3 or higher, or high-intensity CVVHDF (eTable 8 in the Supplement).

    A similar hazard was present for patients treated with NUF greater than 1.75 mL/kg/h compared with patients treated with an NUF rate from 1.01 to 1.75 mL/kg/h from day 7 to 90. During this period, 28 patients (7.1%) in the middle NUF rate group died from day 7 to 12 (aHR, 1.44; 95% CI, 1.10-1.90); 44 patients (12.1%) died from day 13 to 26 (aHR, 1.42; 95% CI, 1.07-1.89); and 29 patients (9.1%) died from day 27 to 90 (aHR, 1.77; 95% CI, 1.26-2.49) (Figure).

    Every 0.5-mL/kg/h increase in NUF rate was associated with increased mortality from day 3 to day 90 (days 3-6: aHR, 1.05; 95% CI, 1.00-1.11; days 7-12: aHR, 1.08; 95% CI, 1.02-1.15; days 13-26: aHR, 1.11; 95% CI, 1.04-1.18; days 27-90: aHR, 1.13; 95% CI, 1.05-1.22) (Figure). Of patients with available premorbid creatinine levels, NUF rates greater than 1.75 mL/kg/h were also associated with increased mortality (days 16-33; aHR, 1.75; 95% CI, 1.15-2.67; P = .02). Net ultrafiltration rate was significantly associated with risk of death in the presence of cumulative fluid balance (aHR, 1.00; 95% CI, 1.00-1.00; P for interaction = .001). Using a joint model, longitudinal increase in NUF rate was associated with risk of death (β = .056; P < .001).

    Sensitivity and Subgroup Analyses

    In the propensity score–matched cohort (405 matched pairs), an NUFrate greater than 1.75 mL/kg/h compared with 1.75 mL/kg/h or less was associated with mortality (192 patients [47.5%] vs 164 patients [40.5%]; P = .04; unadjusted odds ratio [OR], 1.33; 95% CI, 1.01-1.76; P = .04) (eTable 3 and eFigure 4 in the Supplement). An NUF rate greater than 1.75 mL/kg/h was associated with reduced survival using a lower threshold of 0.05 mL/kg/h (aHR, 1.62; 95% CI, 1.12-2.34) and a higher threshold of 0.05 mL/kg/h (aHR, 1.68; 95% CI, 1.16-2.43) (eTable 5 in the Supplement). Restricting to 72 hours of CVVHDF, NUF greater than 1.65 mL/kg/h compared with NUF less than 0.82 mL/kg/h was associated with increased mortality (aHR, 1.72; 95% CI, 1.20-2.47). Using maximum values, an NUF rate greater than 2.66 mL/kg/h compared with an NUF rate less than 1.57 mL/kg/h was associated with death (aHR, 1.95; 95% CI, 1.44-2.65).

    After excluding 92 patients with NUF rates less than 0.01 mL/kg/h, NUF rates greater than 1.75 mL/kg/h were associated with lower survival (aHR, 1.57; 95% CI, 1.08-2.28). A similar hazard persisted after including the 31 patients with missing treatment hours and assigning them an NUF rate of 0 mL/kg/h (aHR, 1.60; 95% CI, 1.12-2.29) as well as assigning them an NUF rate of 1.43 mL/kg/h (aHR, 1.65; 95% CI, 1.15-2.38). The association persisted when the nodes in the Gray model were increased (aHR, 1.68; 95% CI, 1.16-2.44) or decreased (aHR, 1.63; 95% CI, 1.14-2.33).

    An NUF rate greater than 1.75 mL/kg/h was associated with lower survival after adjusting for SOFA scores (aHR, 1.64; 95% CI, 1.14-2.38), use of blood products and protein supplementation (aHR, 1.60; 95% CI, 1.14-2.25), cumulative fluid balance including NUF volume (aHR, 1.74; 95% CI, 1.21-2.51), and after excluding cumulative fluid balance (aHR, 1.72; 95% CI, 1.20-2.47).

    Using stratified analysis, CVVDHF for 3 or more days and CVVDHF for 5 or more days were associated with death (≥3 days: aHR, 1.99; 95% CI, 1.22-3.23; ≥5 days: aHR, 1.93; 95% CI, 1.00-3.75). Of patients with negative fluid balance, the cutoff values were NUF rates less than 1.39 mL/kg/h, from 1.39 to 2.03 mL/kg/h, and greater than 2.03 mL/kg/h. Compared with NUF rates less than 1.39 mL/kg/h, NUF rates greater than 2.03 mL/kg/h were associated with death (aHR, 2.71; 95% CI, 1.56-4.69). Using logistic regression, an NUF rate greater than 1.75 mL/kg/h was not associated with death compared with an NUF rate less than 1.01 mL/kg/h (adjusted OR, 1.25; 95% CI, 0.91-1.73) (eTable 6 in the Supplement). However, every 0.5-mL/kg/h increase was associated with a 7% increase in odds of death (adjusted OR, 1.07; 95% CI, 1.00-1.15) (eTable 7 and eFigure 6 in the Supplement).

    Using subgroup analyses, an NUFrate greater than 1.75 mL/kg/h was associated with mortality among patients with and without organ edema (with organ edema: aHR, 1.61; 95% CI, 1.01-2.55; without organ edema: aHR, 1.75; 95% CI, 1.06-2.88); with and without sepsis (with sepsis: aHR, 2.19; 95% CI, 1.26-3.80; without sepsis: aHR, 1.72; 95% CI, 1.14-2.59); and with eGFR greater than 60 mL/min/1.73 m2 (aHR, 2.30; 95% CI, 1.21-4.38). Among 759 patients (52.9%) with eGFR less than 60 mL/min/1.73 m2, the cutoff values were NUF rates less than 0.95 mL/kg/h, from 0.95 to 1.71 mL/kg/h, and greater than 1.71 mL/kg/h. Compared with an NUFrate less than 0.95 mL/kg/h, an NUFrate greater than 1.71 mL/kg/h was associated with death (aHR, 1.53; 95% CI, 1.04-2.26). An NUFrate greater than 1.75 mL/kg/h was associated with mortality among patients with cardiovascular SOFA scores of 3 or greater (aHR, 1.89; 95% CI, 1.27-2.81) and high-intensity CVVHDF (aHR, 2.22; 95% CI, 1.35-3.66) (eTable 8 in the Supplement).

    Complications and Adverse Events

    A greater proportion of patients receiving NUF greater than 1.75 mL/kg/h developed hypophosphatemia compared with patients receiving NUF from 1.01 to 1.75 mL/kg/h and less than 1.01 mL/kg/h (high: 308 of 477 patients at risk [64.6%]; middle: 293 of 472 patients at risk [62.1%]; low: 247 of 466 patients at risk [53.0%]; P < .001) (Table 4). The frequency of hypophosphatemic episodes was also high (high: 1003 episodes; middle: 893 episodes; low: 627 episodes; P < .001). However, when adjusted for differences in effluent flow and duration of CVVHDF, an NUF rate greater than 1.75 mL/kg/h was not associated with risk of hypophosphatemia (adjusted OR, 1.02; 95% CI, 0.76-1.36; P = .89). More patients with NUF rates greater than 1.75 mL/kg/h developed cardiac arrhythmias requiring treatment and had an increased number of these episodes, but the associations were not significant (patients developing arrhythmia requiring treatment: high: 176 of 478 patients at risk [36.8%]; middle: 175 of 479 patients at risk [36.5%]; low: 147 of 477 patients at risk [30.8%]; P = .08; number of episodes: high: 286 episodes; middle: 264 episodes; low: 237 episodes; P = .08).

    Discussion

    Among critically ill patients receiving CVVHDF, we found that an NUF rate greater than 1.75 mL/kg/h compared with an NUF rate less than 1.01 mL/kg/h was associated with lower risk-adjusted 90-day survival between day 7 and day 90. These findings are aligned with several recent studies in outpatients with end-stage renal disease10-13 that found that higher NUF rates are associated with decreased survival.

    Our findings have several implications. First, the attributable risk associated with an NUF rate greater than 1.75 mL/kg/h was significantly higher (aHR, 1.66; 95% CI, 1.16-2.39) than risk associated with cumulative positive fluid balance (aHR, 1.00; 95% CI, 1.00-1.00) (eTable 4 in the Supplement). Moreover, there was an interaction between the NUF rate and cumulative fluid balance that considerably increased this risk, which may explain the high mortality among patients treated with CVVHDF. Notably, this risk was present only after day 7 and is easily modifiable by slowing the NUF rate to less than 1.75 mL/kg/h.

    There are many possible biological explanations for late mortality. Decreased circulating volume is associated with decreased coronary perfusion and myocardial ischemia.7,26 Repeated ischemia is associated with ventricular remodeling and heart failure.7 Gut hypoperfusion is associated with increased permeability, bacterial translocation, and endotoxemia, which is associated with chronic inflammation and cardiac stunning.27

    Hypotension associated with a high NUF rate may result in administration of fluid with subsequent fluid overload, which is associated with ventricular hypertrophy and fibrosis, predisposing the patient to heart failure and sudden death.28-30 The propensity toward higher frequency of cardiac arrhythmias in patients with a high NUF rate supported this finding. An NUF rate greater than 1.75 mL/kg/h was also associated with risk of hypophosphatemia, which has also been noted in the 2 different trials of intensity of solute control14,31 as well as with increased duration of CVVHDF.32 Hypophosphatemia may also predispose to cardiac arrhythmias and other undesirable biological effects.33-35

    Second, our study suggests that a more modest NUF rate less than 1.75 mL/kg/h is associated with lowest risk (eFigure 2 in the Supplement). This finding is consistent with other studies in patients with end-stage renal disease,13,36,37 in which lower rates and longer treatment duration were associated with survival. Nevertheless, randomized clinical trials are required to confirm our findings.

    Third, while a lower NUF rate might be associated with improved outcomes, it is likely to prolong treatment duration, and this has to be balanced against the need for fluid removal in a critically ill patient. For example, pulmonary edema in a patient with severe heart failure or refractory hypoxemia in a patient with acute respiratory distress syndrome may need a greater NUF rate for a short period of time to prevent sudden death.

    In a single-center study, Murugan et al9 found that among the subgroup of 487 patients who only received CVVHDF, an NUF rate less than 0.5 mL/kg/h compared with an NUF rate greater than 1 mL/kg/h was associated with higher mortality. Although the reason for the differences between the 2 studies is unclear, it is important to note that there are considerable differences in study design and patient population between them. Nevertheless, these differential findings emphasize the need for randomized clinical trials to examine the relationship of NUF rates with outcomes.

    Limitations

    Our study has several limitations. First, findings may be biased by measured and unmeasured confounding at the patient and hospital levels. Nevertheless, the joint model matched propensity score analysis, and the logistic regression provided alternative methods to handle measured confounders and support the primary analysis. Second, data on race/ethnicity, comorbid conditions, and episodes of hypotension during treatment were not measured. Third, there were 31 patients with missing treatment hours; however, including these patients in the analysis did not change the results. Fourth, fluid balance prior to initiation of CVVHDF was unavailable, a limitation that was addressed using the organ edema variable as a surrogate for fluid overload. Given these limitations, the risk associated with an NUF rate greater than 1.75 mL/kg/h is likely to be smaller than measured in this study.

    Conclusions

    In this study of critically ill patients receiving CVVHDF, an NUF rate greater than 1.75 mL/kg/h compared with an NUF rate less than 1.01 mL/kg/h was associated with lower survival. Although the study design does not exclude the possibility of residual confounding owing to unmeasured risk factors, a randomized clinical trial is required to validate these findings before they can be applied to clinical practice.

    Back to top
    Article Information

    Accepted for Publication: April 24, 2019.

    Published: June 7, 2019. doi:10.1001/jamanetworkopen.2019.5418

    Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2019 Murugan R et al. JAMA Network Open.

    Corresponding Author: Raghavan Murugan, MD, MS, FRCP, Center for Critical Care Nephrology, Department of Critical Care Medicine, University of Pittsburgh School of Medicine, 3347 Forbes Ave, Ste 220, Room 206, Pittsburgh, PA 15261 (muruganr@upmc.edu).

    Author Contributions: Dr Murugan and Ms Kerti had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

    Concept and design: Murugan, Gallagher, Kellum, Bellomo.

    Acquisition, analysis, or interpretation of data: Murugan, Kerti, Chang, Gallagher, Clermont, Palevsky, Bellomo.

    Drafting of the manuscript: Murugan, Bellomo.

    Critical revision of the manuscript for important intellectual content: All authors.

    Statistical analysis: Murugan, Kerti, Chang, Gallagher, Clermont.

    Administrative, technical, or material support: Murugan, Kerti, Bellomo.

    Supervision: Murugan, Clermont, Kellum, Bellomo.

    Conflict of Interest Disclosures: Dr Murugan reported receiving grants and personal fees from La Jolla Inc; grants from Bioporto, Inc, and the National Institute of Diabetes and Digestive and Kidney Diseases; and personal fees from Beckman Coulter and AM Pharma, Inc, outside the submitted work. Dr Chang reported receiving grants from the National Institutes of Health during the conduct of the study. Dr Gallagher reported receiving speaking fees from Amgen outside the submitted work. Dr Clermont reported receiving personal fees from UpToDate and grants from the National Institutes of Health and the National Science Foundation outside the submitted work. Dr Palevsky reported receiving personal fees from Novartis, GE Healthcare, HealthSpan Dx, and Baxter International and grants from Dascena outside the submitted work. Dr Kellum reported receiving personal fees from NxStage and grants and personal fees from Baxter International during the conduct of the study. Dr Bellomo reported receiving grants from Baxter International outside the submitted work. No other disclosures were reported.

    Disclaimer: This article was not prepared in collaboration with all the Randomized Evaluation of Normal vs Augmented Level (RENAL) study investigators and does not necessarily reflect the opinions or views of the RENAL study investigators, the George Institute, or the Australian and New Zealand Intensive Care Society.

    Additional Contributions: We thank the study participants of the Randomized Evaluation of Normal vs Augmented Level (RENAL) of Renal Replacement Therapy clinical trial, the RENAL study investigators, and the George Institute for Global Health for providing the study data. We also thank the Biostatistics and Data Management Core of the Clinical Research, Investigation, Systems Modelling of Acute Illness (CRISMA) Center for conducting data analysis. The RENAL trial was performed by the RENAL study investigators in collaboration with the Australian and New Zealand Intensive Care Society Clinical Trials Group and the George Institute for Global Health.

    References
    1.
    Balakumar  V, Murugan  R, Sileanu  FE, Palevsky  P, Clermont  G, Kellum  JA.  Both positive and negative fluid balance may be associated with reduced long-term survival in the critically ill.  Crit Care Med. 2017;45(8):e749-e757. doi:10.1097/CCM.0000000000002372PubMedGoogle ScholarCrossref
    2.
    Vaara  ST, Korhonen  AM, Kaukonen  KM,  et al; FINNAKI Study Group.  Fluid overload is associated with an increased risk for 90-day mortality in critically ill patients with renal replacement therapy: data from the prospective FINNAKI study.  Crit Care. 2012;16(5):R197. doi:10.1186/cc11682PubMedGoogle ScholarCrossref
    3.
    Kidney Disease Improving Global Outcome (KDIGO) Group. KDIGO Clinical Practice Guideline for Acute Kidney Injury. https://kdigo.org/wp-content/uploads/2016/10/KDIGO-2012-AKI-Guideline-English.pdf. Accessed May 1, 2019.
    4.
    Palevsky  PM, Liu  KD, Brophy  PD,  et al.  KDOQI US commentary on the 2012 KDIGO Clinical Practice Guideline for Acute Kidney Injury.  Am J Kidney Dis. 2013;61(5):649-672. doi:10.1053/j.ajkd.2013.02.349PubMedGoogle ScholarCrossref
    5.
    Bouchard  J, Soroko  SB, Chertow  GM,  et al; Program to Improve Care in Acute Renal Disease (PICARD) Study Group.  Fluid accumulation, survival and recovery of kidney function in critically ill patients with acute kidney injury.  Kidney Int. 2009;76(4):422-427. doi:10.1038/ki.2009.159PubMedGoogle ScholarCrossref
    6.
    Bellomo  R, Cass  A, Cole  L,  et al; RENAL Replacement Therapy Study Investigators.  An observational study of fluid balance and patient outcomes in the Randomized Evaluation of Normal vs. Augmented Level of Replacement Therapy trial.  Crit Care Med. 2012;40(6):1753-1760. doi:10.1097/CCM.0b013e318246b9c6PubMedGoogle ScholarCrossref
    7.
    Burton  JO, Jefferies  HJ, Selby  NM, McIntyre  CW.  Hemodialysis-induced repetitive myocardial injury results in global and segmental reduction in systolic cardiac function.  Clin J Am Soc Nephrol. 2009;4(12):1925-1931. doi:10.2215/CJN.04470709PubMedGoogle ScholarCrossref
    8.
    Silversides  JA, Pinto  R, Kuint  R,  et al.  Fluid balance, intradialytic hypotension, and outcomes in critically ill patients undergoing renal replacement therapy: a cohort study.  Crit Care. 2014;18(6):624. doi:10.1186/s13054-014-0624-8PubMedGoogle ScholarCrossref
    9.
    Murugan  R, Balakumar  V, Kerti  SJ,  et al.  Net ultrafiltration intensity and mortality in critically ill patients with fluid overload.  Crit Care. 2018;22(1):223.PubMedGoogle ScholarCrossref
    10.
    Flythe  JE, Kimmel  SE, Brunelli  SM.  Rapid fluid removal during dialysis is associated with cardiovascular morbidity and mortality.  Kidney Int. 2011;79(2):250-257. doi:10.1038/ki.2010.383PubMedGoogle ScholarCrossref
    11.
    Kim  TW, Chang  TI, Kim  TH,  et al.  Association of ultrafiltration rate with mortality in incident hemodialysis patients.  Nephron. 2018;139(1):13-22. doi:10.1159/000486323PubMedGoogle ScholarCrossref
    12.
    Movilli  E, Gaggia  P, Zubani  R,  et al.  Association between high ultrafiltration rates and mortality in uraemic patients on regular haemodialysis: a 5-year prospective observational multicentre study.  Nephrol Dial Transplant. 2007;22(12):3547-3552. doi:10.1093/ndt/gfm466PubMedGoogle ScholarCrossref
    13.
    Saran  R, Bragg-Gresham  JL, Levin  NW,  et al.  Longer treatment time and slower ultrafiltration in hemodialysis: associations with reduced mortality in the DOPPS.  Kidney Int. 2006;69(7):1222-1228. doi:10.1038/sj.ki.5000186PubMedGoogle ScholarCrossref
    14.
    Bellomo  R, Cass  A, Cole  L,  et al; RENAL Replacement Therapy Study Investigators.  Intensity of continuous renal-replacement therapy in critically ill patients.  N Engl J Med. 2009;361(17):1627-1638. doi:10.1056/NEJMoa0902413PubMedGoogle ScholarCrossref
    15.
    von Elm  E, Altman  DG, Egger  M, Pocock  SJ, Gøtzsche  PC, Vandenbroucke  JP; STROBE Initiative.  The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.  Ann Intern Med. 2007;147(8):573-577. doi:10.7326/0003-4819-147-8-200710160-00010PubMedGoogle ScholarCrossref
    16.
    Psaty  BM, Koepsell  TD, Lin  D,  et al.  Assessment and control for confounding by indication in observational studies.  J Am Geriatr Soc. 1999;47(6):749-754. doi:10.1111/j.1532-5415.1999.tb01603.xPubMedGoogle ScholarCrossref
    17.
    Siew  ED, Peterson  JF, Eden  SK, Moons  KG, Ikizler  TA, Matheny  ME.  Use of multiple imputation method to improve estimation of missing baseline serum creatinine in acute kidney injury research.  Clin J Am Soc Nephrol. 2013;8(1):10-18. doi:10.2215/CJN.00200112PubMedGoogle ScholarCrossref
    18.
    Buuren  S, Groothius-Oudshoorn  K.  MICE: multivariate imputation by chained equations in R.  J Stat Softw. 2011;45(3):1-67. doi:10.18637/jss.v045.i03Google ScholarCrossref
    19.
    Závada  J, Hoste  E, Cartin-Ceba  R,  et al; AKI6 Investigators.  A comparison of three methods to estimate baseline creatinine for RIFLE classification.  Nephrol Dial Transplant. 2010;25(12):3911-3918. doi:10.1093/ndt/gfp766PubMedGoogle ScholarCrossref
    20.
    Gray  RJ.  Flexible methods for analyzing survival data using splines, with applications to breast cancer prognosis.  J Am Stat Assoc. 1992;87(420):942-951. doi:10.1080/01621459.1992.10476248Google ScholarCrossref
    21.
    Gray  RJ.  Spline-based tests in survival analysis.  Biometrics. 1994;50(3):640-652. doi:10.2307/2532779PubMedGoogle ScholarCrossref
    22.
    Kasal  J, Jovanovic  Z, Clermont  G,  et al.  Comparison of Cox and Gray’s survival models in severe sepsis.  Crit Care Med. 2004;32(3):700-707. doi:10.1097/01.CCM.0000114819.37569.4BPubMedGoogle ScholarCrossref
    23.
    Valenta  Z, Weissfeld  L.  Estimation of the survival function for Gray’s piecewise-constant time-varying coefficients model.  Stat Med. 2002;21(5):717-727. doi:10.1002/sim.1061PubMedGoogle ScholarCrossref
    24.
    Rizopoulos  D.  Dynamic predictions and prospective accuracy in joint models for longitudinal and time-to-event data.  Biometrics. 2011;67(3):819-829. doi:10.1111/j.1541-0420.2010.01546.xPubMedGoogle ScholarCrossref
    25.
    Wulfsohn  MS, Tsiatis  AA.  A joint model for survival and longitudinal data measured with error.  Biometrics. 1997;53(1):330-339. doi:10.2307/2533118PubMedGoogle ScholarCrossref
    26.
    Burton  JO, Jefferies  HJ, Selby  NM, McIntyre  CW.  Hemodialysis-induced cardiac injury: determinants and associated outcomes.  Clin J Am Soc Nephrol. 2009;4(5):914-920. doi:10.2215/CJN.03900808PubMedGoogle ScholarCrossref
    27.
    McIntyre  CW, Harrison  LE, Eldehni  MT,  et al.  Circulating endotoxemia: a novel factor in systemic inflammation and cardiovascular disease in chronic kidney disease.  Clin J Am Soc Nephrol. 2011;6(1):133-141. doi:10.2215/CJN.04610510PubMedGoogle ScholarCrossref
    28.
    Shivalkar  B, Flameng  W, Szilard  M, Pislaru  S, Borgers  M, Vanhaecke  J.  Repeated stunning precedes myocardial hibernation in progressive multiple coronary artery obstruction.  J Am Coll Cardiol. 1999;34(7):2126-2136. doi:10.1016/S0735-1097(99)00467-2PubMedGoogle ScholarCrossref
    29.
    McMullen  JR, Sherwood  MC, Tarnavski  O,  et al.  Inhibition of mTOR signaling with rapamycin regresses established cardiac hypertrophy induced by pressure overload.  Circulation. 2004;109(24):3050-3055. doi:10.1161/01.CIR.0000130641.08705.45PubMedGoogle ScholarCrossref
    30.
    Ritz  E, Wanner  C.  The challenge of sudden death in dialysis patients.  Clin J Am Soc Nephrol. 2008;3(3):920-929. doi:10.2215/CJN.04571007PubMedGoogle ScholarCrossref
    31.
    Palevsky  PM, Zhang  JH, O’Connor  TZ,  et al; VA/NIH Acute Renal Failure Trial Network.  Intensity of renal support in critically ill patients with acute kidney injury.  N Engl J Med. 2008;359(1):7-20.PubMedGoogle ScholarCrossref
    32.
    Demirjian  S, Teo  BW, Guzman  JA,  et al.  Hypophosphatemia during continuous hemodialysis is associated with prolonged respiratory failure in patients with acute kidney injury.  Nephrol Dial Transplant. 2011;26(11):3508-3514. doi:10.1093/ndt/gfr075PubMedGoogle ScholarCrossref
    33.
    Aubier  M, Murciano  D, Lecocguic  Y,  et al.  Effect of hypophosphatemia on diaphragmatic contractility in patients with acute respiratory failure.  N Engl J Med. 1985;313(7):420-424. doi:10.1056/NEJM198508153130705PubMedGoogle ScholarCrossref
    34.
    Lim  C, Tan  HK, Kaushik  M.  Hypophosphatemia in critically ill patients with acute kidney injury treated with hemodialysis is associated with adverse events.  Clin Kidney J. 2017;10(3):341-347.PubMedGoogle Scholar
    35.
    Yang  Y, Zhang  P, Cui  Y,  et al.  Hypophosphatemia during continuous veno-venous hemofiltration is associated with mortality in critically ill patients with acute kidney injury.  Crit Care. 2013;17(5):R205. doi:10.1186/cc12900PubMedGoogle ScholarCrossref
    36.
    Flythe  JE, Curhan  GC, Brunelli  SM.  Shorter length dialysis sessions are associated with increased mortality, independent of body weight.  Kidney Int. 2013;83(1):104-113. doi:10.1038/ki.2012.346PubMedGoogle ScholarCrossref
    37.
    Flythe  JE, Curhan  GC, Brunelli  SM.  Disentangling the ultrafiltration rate-mortality association: the respective roles of session length and weight gain.  Clin J Am Soc Nephrol. 2013;8(7):1151-1161. doi:10.2215/CJN.09460912PubMedGoogle ScholarCrossref
    ×