Acute kidney injury: epidemiology and diagnostic criteria Eric A.J Hostea,b and John A Kellumb Purpose of review To review recent advances in the definitions and diagnostic criteria for acute renal failure and acute kidney injury To explore how these changes impact the epidemiology and clinical implications for patients in the intensive care unit Recent findings Recently published consensus criteria for the definition of acute renal failure/acute kidney injury have led to significant changes in how we think about this disorder Studies from around the world, both in and out of the intensive care unit, have shown a dramatic incidence of acute kidney injury and high associated mortality This review considers these new findings and their historical context, and attempts to shed new light on this old problem Summary Small changes in kidney function in hospitalized patients are important and impact on outcome RIFLE criteria provide a uniform definition of acute kidney injury and are increasingly used in literature Keywords acute kidney injury, acute renal failure, RIFLE criteria Curr Opin Crit Care 12:531–537 ß 2006 Lippincott Williams & Wilkins a Intensive Care Unit, Ghent University Hospital, Ghent, Belgium and bThe Clinical Research, Investigation and Systems Modeling of Acute Illness (CRISMA) Laboratory, Department of Critical Care Medicine, University of Pittsburgh, School of Medicine, Pittsburgh, Pennsylvania, USA Correspondence to John A Kellum, MD, Department of Critical Care Medicine, University of Pittsburgh, Room 608, Scaife Hall, 3550 Terrace Street, Pittsburgh, PA 15261, USA Tel: +1 412 647 6966; fax: +1 412 647 8060; e-mail: kellumja@ccm.upmc.edu Current Opinion in Critical Care 2006, 12:531–537 Abbreviations AKI ARF GFR ICU MDRD RRT acute kidney injury acute renal failure glomerular filtration rate intensive care unit modification of diet in renal disease renal replacement therapy ß 2006 Lippincott Williams & Wilkins 1070-5295 Introduction For most clinicians working in intensive care and nephrology, the notion of acute renal failure (ARF) is that of severe organ dysfunction For research and even reimbursement purposes, ARF is often defined by the need for artificial support, termed renal replacement therapy (RRT) This narrow concept of ARF may, however, be far too limiting, and mounting evidence suggests that acute dysfunction of kidney manifest by changes in urine output and blood chemistries portend serious clinical consequences [1] The term ARF is relatively new in the medical lexicon Eknoyan [2] reminds us that the first description of ARF, then termed ischuria renalis, was by William Heberden in 1802 At the beginning of the 20th century, ARF, then named acute Bright’s disease, was well described in William Osler’s early works (1909), as a consequence of toxic agents, pregnancy, burns, trauma or operations on the kidneys During World War I the syndrome was named ‘War Nephritis’ [3] and was reported in several publications The syndrome was forgotten until World War II, when Bywaters and Beall [4] published their classical paper on crush syndrome It was Homer W Smith [5] who is credited with the introduction of the term ‘ARF’, in a chapter on ‘Acute renal failure related to traumatic injuries’, in his textbook The Kidney – Structure and Function in Health and Disease (1951) The same year a whole issue of the Journal of Clinical Investigation was dedicated to ARF [6] In most reviews and textbooks [7,8], the concept of acute kidney dysfunction still emphasizes the most severe forms with severe azotemia and often with oliguria or anuria It is only in the past few years that moderate decreases of kidney function have been recognized as important, e.g by the Sepsis-related Organ Failure Assessment score [9] and in studies on radiocontrastinduced nephropathy [10] Moreover, until very recently there was no consensus on the diagnostic criteria or clinical definition of ARF, resulting in multiple different definitions A recent survey revealed the use of at least 35 definitions in the literature [11] Apart from differences in patient characteristics, this is probably one of the main reasons that there is such a wide variation in the reported incidence and outcome of ARF (incidence ranges between and 31% [12,13] and mortality between 28 and 82% [13,14]) Obviously, if one study defines ARF as a 25% or greater 531 Copyright © Lippincott Williams & Wilkins Unauthorized reproduction of this article is prohibited 532 Renal system Similarly, Lassnig et al [16] saw, in a cohort of patients who underwent cardiac surgery, that acute kidney dysfunction, defined as an increase of serum creatinine of 0.5 mg/dl or above or a decrease greater than 0.3 mg/dl, was associated with worse survival The reasons why small alterations in renal function lead to increases in hospital mortality are unclear Possible explanations include the untoward effects of acute kidney dysfunction such as volume overload, retention of uremic compounds, acidosis, electrolyte disorders, increased risk for infection and anemia [17] Although acute kidney dysfunction could simply be co-linear with unmeasured variables that lead to increased mortality, multiple attempts to control for known clinical variables has led to the consistent conclusion that renal dysfunction is independently Figure Relationship between the definition used of acute renal failure (ARF) and the corresponding mortality observed Acute kidney injury and the RIFLE criteria Recognizing that early and/or milder forms of renal dysfunction have clinical importance and that staging (mild to severe) is desirable in order to better describe the syndrome, the Acute Dialysis Quality Initiative, a group of experts in acute kidney dysfunction, consisting of nephrologists and intensivists, proposed the RIFLE criteria for acute kidney dysfunction (http:// www.ccm.upmc.edu/adqi/ADQI2/ADQI2g1.pdf) [18] The acronym RIFLE stands for the increasing severity classes Risk, Injury and Failure, and the two outcome classes Loss and End-Stage Kidney Disease The three severity grades are defined on the basis of the changes in serum creatinine or urine output (Fig 2) where the worst of each criterion is used The two outcome criteria, Loss and End-Stage Kidney Disease, are defined by the duration of loss of kidney function The RIFLE criteria were published as a workgroup document on the Acute Dialysis Quality Initiative website in June 2003, published online in May 2004 and in print in August 2004 [18] Since then a number of papers have been published that use the RIFLE criteria [19,20,21,22,23,24,25,26,27,28,29] (Table 1) Most of the studies were published in the past year Figure The RIFLE classification scheme for acute kidney injury GFR criteria Urine output criteria Increased creatinine ×1.5 or GFR decrease > 25% UO < 0.5 ml/kg/h ×6h Increased creatinine × or GFR decrease > 50% UO < 0.5 ml/kg/h × 12 h Increase creatinine ×3 or GFR decrease >75% Failure or creatinine ≥ mg/dl UO < 0.3 ml/kg/h × 24 h or Anuria × 12 h Risk Injury (Acute rise of ≥ 0.5 mg/dl) Loss ESRD Seventeen published definitions of ARF were classified into grades on the basis of the criteria used As an arbitrary reference, a definition requiring a doubling of serum creatinine or a rise in serum creatinine by mg/dl was given a grade of More strict criteria were given proportionally high grades and less strict criteria, lower grades The corresponding observed mortality in the control group (for clinical trials) or overall mortality (for observational studies) were then charted There was a significant correlation between the definition grade and observed mortality From Kellum et al [11]; used with permission High sensitivity ria Another element that has emerged in recent years is the observation that small decreases in kidney function are important For example, Levy et al [10] found that a 25% increase of serum creatinine after administration of radiocontrast was associated with a worse outcome compared with those who did not experience a 25% or greater increase Chertow et al [15] defined hospital acquired acute kidney dysfunction as an increase of serum creatinine of above 0.3 mg/dl and found that this was independently associated with mortality associated with outcome Furthermore, more severe renal dysfunction tends to be associated with even worse outcome compared with milder abnormalities Oligu rise in serum creatinine and another study defines ARF only as the need for RRT, the two studies will not describe the same cohort of patients There is even a linear correlation between the degree of kidney dysfunction and the outcome of acute kidney dysfunction The more strict the definition of ARF, the greater the mortality (Fig 1) [11] High specificity Persistent ARF = complete loss of renal function > weeks End-stage renal disease The classification system includes separate criteria for serum creatinine and urine output (UO) The criteria that lead to the worst possible classification should be used Note that RIFLE-F is present even if the increase in serum creatinine is below 3-fold so long as the new serum creatinine is 4.0 mg/dl (350 mmol/l) or above in the setting of an acute increase of at least 0.5 mg/dl (44 mmol/l) The shape of the figure denotes the fact that more patients (high sensitivity) will be included in the mild category, including some without actually having renal failure (less specificity) In contrast, at the bottom, the criteria are strict and therefore specific, but some patients will be missed GFR, glomerular filtration rate; ARF, acute renal failure From Bellomo et al [18]; used with permission Copyright © Lippincott Williams & Wilkins Unauthorized reproduction of this article is prohibited Acute kidney injury Hoste and Kellum 533 Table Papers in which the RIFLE criteria for AKI were used Herget-Rosenthal et al [19] Hoste et al [20] Bell et al [21]a Abosaif et al [22]a Kuitunen et al [23] Cohort Aim of the study 85 ICU patients, with initial normal GFR evaluation of cystatin C vs creatinine 704 AKI patients treated with RRT 207 continuous RRT patients 183 ICU patients with AKI on admission 813 cardiac surgery patients AKI defined on glomerular filtration rate (1) or urine output and glomerular filtration rate (2) Outcome criteria Occurrence of AKI no 44/85 (51.8%) impact of BSI long-term outcome outcome incidence and outcome of AKI NA 2 yes yes no no NA RIFLEmax R: 3/85 (3.5%) I: 13/85 (15.3%) F: 28/85 (32.9%) L: no BSI 9.2%/ BSI 43.5% NA E: no BSI 0.5%/ BSI 8.1% R: 17/207 (8.2%) NA I: 50/207 (24.2%) F: 121/207 (58.5%) L: 3/207 (1.4%) E: 16/207 (7.7%) R: 60/159 (37.7%) 156/813 (19.2%) I: 56/159 (35.2%) F: 43/159 (27.0%) R: 88/813 (10.8%) I: 28/813 (3.4%) F: 40/813 (4.9%) I: 39/94 (41.5%) Guitard et al [24] 94 liver transplant patients incidence and outcome of AKI no 60/94 (63.8%) Hoste et al [25] 5383 ICU patients incidence and outcome of AKI no 3617/5383 (67.2%) F: 21/94 (22.3%) R: 670/5383 (12.4%) 18% I: 1436/5383 (26.5%) F: 1511/5383 (28.1%) R: 9.1% Uchino et al [26] Lin et al [27] Heringlake et al [28] Lopes et al [29]b 20 126 patients admitted to the hospital 46 ECMO patients 29 623 cardiac surgery patients 126 burn patients incidence and outcome of AKI incidence and outcome of AKI incidence and outcome of AKI incidence and outcome of AKI 2 no no no no 36/46 (78.3%) I: 5.2% F: 3.7% R: 7/46 (15.2%) 15.4% (range 3.1–75%) I: 18/46 (39.1%) F: 11/46 (23.9%) R: 9% (2–40%) 35.7% I: 5% (0.8–30%) F: 2% (0.6–33%) R: 14.3% I: 8.7% F: 12.7% R, RIFLE Risk class; I, RIFLE Injury class; F, RIFLE Failure class; L, RIFLE Loss; E, RIFLE End-Stage Kidney Disease AKI, acute kidney injury; BSI, bloodstream infection; ICU, intensive care unit; NA, not available/applicable; RRT, renal replacement therapy a Patients were classified at inclusion in the study (on admission to the ICU or at start of continuous RRT) b Patients were classified on occurrence of maximum RIFLE class during the first 10 days of hospital admission In addition, the Acute Kidney Injury Network organized two conferences endorsed by the different critical care and nephrology societies The aim of these conferences was to come to a broader consensus on the definitions and terminology for ARF In particular, this group has proposed the term ‘AKI’ to define the entire spectrum of acute renal dysfunction from its earliest and mildest forms to the need for RRT We will therefore adopt this term, as we have previously [1,25] Copyright © Lippincott Williams & Wilkins Unauthorized reproduction of this article is prohibited 534 Renal system Use of the RIFLE criteria An overview of the papers that used the RIFLE criteria for AKI is presented in Table All studies used the severity grading criteria Risk, Injury and Failure, but only two studies [20,21] also used the outcome criteria Loss and End-Stage Kidney Disease Both studies that used the outcome criteria were in a cohort of AKI patients defined by the need for RRT Only Bell et al [21] classified both severity grades and outcome classes All patients included in this study were treated with continuous RRT for AKI They had therefore severe AKI and there is rationale to classify these as Failure, according to the adaptations made to the RIFLE criteria after the First Acute Kidney Injury Network Conference (personal communication) Severity grading was performed according to the RIFLE criteria on creatinine and urine output criteria in seven out of the 10 studies that reported on severity grading [21,22,23,24,25,27,29] Lin et al [27], however, used different urine output criteria cut-offs compared with those of RIFLE The three remaining studies defined severity of AKI on a change of serum creatinine level and not on urine output [19,26,28] The reasons for this were diverse Herget-Rosenthal et al [19] compared assessment of glomerular filtration rate (GFR) by serum creatinine and cystatin C levels Uchino et al [26] retrospectively evaluated hospital-wide cases, which prevented assessment of urine output In addition, the study by Heringlake et al [28] was a large prospective study on practice patterns in cardiac surgery in German cardiovascular centers Presumably, the study coordinators chose to keep the questionnaire as compact as possible in order to get a large enough response Interestingly, one group chose to use the Cockcroft–Gault equation for assessment of GFR, rather than use a change in serum creatinine levels as all other authors did [22] When baseline serum creatinine level is unknown in a patient without a history of chronic kidney insufficiency, the Acute Dialysis Quality Initiative proposed the use of a baseline creatinine based upon the modification of diet in renal disease (MDRD) equation assuming a GFR > 75 ml/min/1.73 m2 [18] This was done in only three studies [25,26,27] Kuitunen et al [23] also used the MDRD formula, although not for assessment of a baseline creatinine level, but for assessment of GFR Most studies used the RIFLE criteria to assess the occurrence rate of AKI in specific cohorts of patients Two studies, however, used the RIFLE criteria for other means than this Herget-Rosenthal et al [19] evaluated whether a serum level of cystatin C is a better marker for GFR than a serum creatinine level and Hoste et al [20] used the RIFLE outcome criteria as a secondary outcome parameter in a study on the impact of bloodstream infection in AKI patients treated with RRT Occurrence rate of acute kidney injury defined by the RIFLE criteria The occurrence rate of AKI defined by RIFLE criteria in the different cohorts ranged from 15.4 to 78.3% (Table 1) This is higher than generally accepted when the classic terminology of ARF is used The large study by Uchino et al [26] demonstrated that almost 18% of hospitalized patients in a large tertiary care hospital had an episode of AKI defined by RIFLE on GFR criteria This is much higher than the incidence of 7.2% reported in a hallmark study on data from 1996 [30] and 4.9% in the same hospital on data from 1979 [31] Although the definition of AKI used in that study differs from the RIFLE criteria, the sensitivity seems comparable Nash et al and Hou et al [30,31] defined AKI as a rise in serum creatinine above 0.5 mg/dl for patients with a baseline below 1.9 mg/dl, above 1.0 mg/dl for patients with a baseline of 2–4.9 mg/dl and above 1.5 mg/dl for patients with serum creatinine level above 5.0 mg/dl An explanation for this may be that RIFLE criteria are more sensitive, especially for patients with acute or chronic disease; alternatively, the three cohorts may also have different baseline characteristics and/or different comorbidities The trend of increasing incidence for the same definition in the same institute suggests that the latter explanation seems more plausible Increasingly, patients are now older, suffer from more comorbidity such as diabetes or cardiovascular disease, and more patients are exposed to diagnostic and therapeutic procedures with potential harm for kidney function The two large studies in cardiac surgery patients indicate that the incidence of AKI after cardiac surgery is about 15–20% [23,28] This is a considerably higher incidence compared to the incidence of ARF of below 8% as is generally accepted in this specific cohort of patients [12,32–35] In a single-center, tertiary care, general intensive care unit (ICU) setting, two patients out of three experienced an episode of AKI [25] Again, this is a considerably higher incidence of renal dysfunction than generally reported (generally around 5% [36] up to 31% in specific subgroups [13,37]) Finally, small studies [24,27,29] in specific groups of patients such as patients with cardiogenic shock on extracorporeal membrane oxygenation, liver transplantation or burns also demonstrated high ICU period prevalence rates for AKI of 78, 64 and 35.7%, respectively In summary, the RIFLE criteria for AKI are certainly more sensitive compared with more traditional definitions of ARF The incidence of AKI defined by the RIFLE criteria is much higher (2–10 times higher) than Copyright © Lippincott Williams & Wilkins Unauthorized reproduction of this article is prohibited Acute kidney injury Hoste and Kellum 535 the incidence of ARF, but the incidence for both appears to be increasing Table Outcome of AKI defined by RIFLE criteria and by individual RIFLE severity grades after correction for other comorbidities OR (for LR)/ HR (for CPH) Statistical RIFLE (95% confidence test used criteria interval) Outcome of acute kidney injury defined by RIFLE criteria An overview of mortality for the individual grades of AKI in various populations is presented in Fig Importantly, all studies, with the exception of Bell et al [21], report a stepwise increase of mortality for increasing RIFLE class The study by Bell et al is, however, an exception in this series, as it included only patients that were treated with continuous RRT, suggesting that these patients already had severe AKI In all other studies, increasing severity classes of AKI indeed had worse outcome In four studies, a multivariable analysis was performed to assess the impact of AKI defined by RIFLE after correction for other comorbidities (Table 2) AKI defined by RIFLE criteria was associated with worse outcome in all four studies Kuitunen et al [23] Hoste et al [25] LR AKI OR: 2.616