Acute Renal Failure Definitions and Classification

Abstract

Epidemiology of Acute Renal Failure The development of acute renal failure (ARF) in the hospital setting continues to be associated with poor outcomes (1–7). Over the last three decades, several experimental models have identified pathophysiologic mechanisms associated with ARF and have enhanced our understanding of the disease (8–10). It is evident that ARF can result from alterations in renal perfusion, changes in glomerular filtration, and tubular dysfunction, and that correction of these factors can ameliorate the effects of ARF (11,12). On the basis of the identification of the underlying mechanisms, several new potential interventions have been developed that have been shown to alter the course of incipient and established ARF in experimental models (13–15). Application of these findings has resulted in improvements in the prevention of ARF due to radiocontrast agents, aminoglycoside antibiotics, and rhabdomyolysis (16,17). Several other agents are now in advanced stages of development or initial phases of clinical trials (18,19). In concert, advances in dialysis have occurred with the availability of continuous renal replacement therapies in addition to intermittent hemodialysis and acute peritoneal dialysis (20–22). It is well recognized that uncomplicated ARF can usually be managed outside the intensive care unit (ICU) setting and carries a good prognosis, with mortality rates less than 5% to 10% (23,24). In contrast, ARF complicating nonrenal organ system failure in the ICU setting is associated with mortality rates of 50% to 70%, which has not changed for several decades (6,25–30). These figures are in sharp contrast to the experience with acute myocardial infarction (AMI), where in-hospital mortality rates have declined from the range of 50% to approximately 6% over the past 25 to 30 yr. Much of the credit for improved AMI outcomes has been attributed to the use of coronary care units, cardiac catheterization, aspirin, β-adrenergic antagonists, thrombolytic therapy, and, more recently, specialized percutaneous coronary interventions and glycoprotein IIb–IIIa inhibitors (31–34). In spite of improved dialytic technology, including the development and refinement of continuous renal replacement therapies for the most critically ill patients, we have seen no material change in the high mortality rates associated with ARF. Indeed, we have not demonstrated any pharmacologic or other intervention effective in the early management of ARF. Interventions that have been deemed ineffective (or potentially harmful) include the following: loop and osmotic diuretic agents (35,36), “renal dose” dopamine (37,38), atrial natriuretic peptide (39,40), insulin-like growth factor-1 (41), and endothelin receptor antagonists (42). Although ARF may develop in 5% or more of hospitalized patients, the heterogeneity of ARF and associated comorbidity make its study more difficult than AMI and other, more discrete conditions. Among the impediments to progress in ARF research is the lack of a uniform definition of ARF that might be used to compare and contrast observational studies, and to allow for rational design of clinical trials. Relatively few studies have examined the incidence of hospital-acquired ARF. The oft-cited study by Hou et al. (23) reported an ARF incidence estimate of 4.9%. Shusterman et al. (24) conducted a similar study identifying ARF in 1.9% of hospitalized patients. A follow-up study recently published by Hou and colleagues (43) showed an increase in incidence (7%), but a similar spectrum of risk factors. ARF in the ICU setting has also been characterized in the last two decades. Lian[Combining Tilde]o et al. (6) found ICU patients with ARF to have associated organ failure, sepsis, and other complications. It is well recognized that the development of ARF is associated with an increase in mortality (22,25,26,44,45). It is also known that patients with ARF as part of multiorgan failure have the highest mortality rates. In several studies, sepsis-related ARF had a significantly worse prognosis than ARF in the absence of sepsis (46,47). It is also recognized that untreated ARF may contribute to a higher incidence of new-onset sepsis (48). Definition of ARF The spectrum of definitions in published studies of ARF is striking, ranging from severe (e.g., ARF requiring dialysis) to relatively modest observable increases in serum creatinine concentration (e.g., increase in serum creatinine of 0.3 to 0.5 mg/dl above baseline). Solomon et al. (36) used the definition of an increase in serum creatinine of 0.5 mg/dl within 48 h of radiocontrast exposure in a widely cited study that showed a borderline significant difference in ARF among individuals given saline infusion versus furosemide or mannitol before radiocontrast exposure. However, neither the Solomon et al. study nor others that use this ARF definition (including the Tepel et al. (49) and Kay et al. (50) publications on N-acetyl cysteine) have shown an association between a transient change in serum creatinine and morbidity, or the likelihood of long-term recovery of renal function. Many other definitions of ARF have been applied; some are outlined in Table 1. The most liberal of definitions have been used in intervention studies aimed at ARF prevention, usually in the context of radiocontrast exposure, one of the few instances in which ARF can be anticipated.Table 1: Alternative ARF definitions from several published studiesSeveral of the definitions are extremely complex (see Lian[Combining Tilde]o et al. (6) and others) and could allow excessive subjectivity in ARF determination. These would likely be impractical for prospective, multicenter investigations. Moreover, none of the definitions used to date take into account the modifying effects of age, gender, and race on creatinine generation (and thereby serum creatinine concentration in ARF). It is noteworthy that creatinine generation is typically higher among individuals who are younger, male, and African American (51). Therefore, at the same decrement in GFR, persons with different demographic characteristics may be more likely to “qualify” with ARF diagnoses, particularly those definitions that require a minimum peak creatinine (e.g., 50% increase, to at least 2.0 mg/dl). By use of this definition, we found a twofold increase in the incidence of amphotericin B–associated ARF among men (52). Whether male gender is a true risk for ARF or simply a risk for being diagnosed with ARF is unclear. Regardless, the association of ARF with male gender highlights one of the limitations of the use of a definition of ARF that is creatinine based and not age, gender, and race adjusted. Changes in serum creatinine are not specific and do not discriminate the nature and type of renal insult (e.g., ischemic, nephrotoxic) or the site and extent of glomerular or tubular injury, and levels are relatively insensitive to small changes in GFR (53). Moreover, changes in serum creatinine may lag behind changes (decline or recovery) in GFR by several days. Finally, because serum creatinine is influenced by one of the potential interventions for ARF (e.g., creatinine is removed by dialysis), its specificity for renal recovery is even more problematic. Empiric Evidence of the Focus on Creatinine and Urine Output We recently completed an analysis focusing on correlates of timing of nephrology consultation for ARF in the ICU (54). To avoid the complexities of comparing individuals whose ARF developed during a complicated ICU stay, we restricted our analysis to those patients with evidence of ARF at ICU admission and excluded individuals designated as “do not resuscitate.” We considered a wide array of demographic, clinical, laboratory, and physiologic variables (including pulmonary artery catheter data in some patients). The serum creatinine concentration and urine output (either as a continuous variable or the dichotomous “oliguria”) (<400 ml/d) were associated with the timing of consultation. There was no relation between the timing of consultation and hospital service, medical history, physiologic parameters, other laboratory studies, or the presence or absence of organ system failure, despite the fact that many of these factors have been shown to predict mortality in ARF in other studies. In other words, empiric evidence demonstrates that the definitions used in published reports are operative in practice, with little attention to risk profiles or associated nonrenal organ system failure. Analogous Definitions in Other Conditions Conceptual Framework for Disease Definitions. Disease definitions may be used to ascertain the presence of a disease in an individual or a population, guide the nature and timing of diagnostic and therapeutic interventions, and, in individual patients, help determine prognosis. The presence of any disease is inferred from a combination of clinical symptoms and signs, and alterations in biologic markers that can be reproducibly measured. Measures defining a disease should be responsive to change, track the natural history of the disease, and provide an assessment of the severity of injury. Consequences of the untreated disease and its response to specific interventions are additional criteria that might be considered when evaluating the choice of variables to define and classify a disease. Most disease definitions rely on the presence of specific markers that are measurably altered in response to an injury, and the sensitivity and specificity of any definition depends on the criteria used. These “response variables” may appear at varying time points in the disease and help define the course of the disease. Ideally, the magnitude and pattern of change of the response variable correlate with disease outcomes. For instance, AMI can be diagnosed with the combination of chest pain and elevated cardiac troponins or creatine phosphokinase. Gradations in the severity of signs (including electrocardiography) and symptoms and the levels of troponin and creatine phosphokinase profile allow further classification of the disease spectrum (e.g., angina, unstable angina, demand ischemia, silent ischemia, myocardial infarction). A key feature for AMI is that the clinical presentation is directly related to an underlying event (i.e., coronary thrombosis). Moreover, the markers are sensitive, specific, and correlate with the severity of injury, even in the absence of typical clinical features. In contrast, sepsis is heterogeneous in its presentation and affects multiple organs, so no single marker can be used to define the presence or absence of disease. Recognizing this limitation, a functional definition for sepsis has been based on events in the natural history of the sepsis syndrome: systemic inflammatory response syndrome, sepsis, severe sepsis, and septic shock (55). In the absence of specific markers, effective definitions for a disease rely on multiple parameters, some of which represent the specific response to the disease, whereas others reflect nonspecific consequences of the disease. Disease severity is graded on the basis of the presence of specific parameters. For instance, the transition from sepsis to severe sepsis requires the presence of sepsis-related organ dysfunction. These graded definitions are more readily applied to classify populations, although they have also been used to guide interventions in individual patients and research subjects (56,57). The multidimensional definition and classification construct has been applied to several diseases where a single specific diagnostic criterion is not available or is otherwise unsuitable. For example, the Ranson criteria enable early classification of severe acute pancreatitis (58). These criteria rely on the presence of three or more of the 11 criteria evident within 48 h of admission. Criteria include age and a series of laboratory parameters that reflect consequences of disordered pancreatic function (i.e., hyperglycemia in absence of diabetes, hypocalcemia, azotemia, anemia, hypoalbuminemia, leukocytosis, elevations of the hepatic enzymes lactate dehydrogenase and aspartate aminotransferase) and physiologic variables (i.e., metabolic acidosis, hypoxia) to grade the response. Interestingly, serum amylase and lipase, enzymes directly related to pancreatic injury (and analogous to creatinine in ARF), are not included in the scoring system. The number of positive criteria are associated with mortality ranging from <5% for zero to two criteria to 100% for seven to eight criteria (58). Similarly, the Child-Pugh classification is a means of assessing the severity of hepatic cirrhosis (59). It assigns scores for each of three laboratory parameters (bilirubin, albumin, and prothrombin time) and two clinical criteria representing the consequences of liver failure (encephalopathy and ascites). The individual scores are summed and then grouped as <7 (A), 7 to 9 (B), and >9 (C). A Child-Pugh “C” classification forecasts survival of less than 12 mo. Cancer staging similarly uses the tumor node metastasis grading system to classify malignant disease. Common to all of these classification systems is the lack of any single criterion to diagnose the disease and reliance on the consequences of the disease and on other factors the course of the disease for We that ARF be in a similar in of of and organ ARF has more in with sepsis or pancreatitis than with of in For any the to the disease is and where and when the in the natural history of the disease. The whereas the could time points for variables should allow of the disease and provide a to determine the time course of the disease. Renal functional alterations many (e.g., and markers specific for the site and pattern of injury are Several new are to study ARF and are to be for studies et al. and et al. demonstrated that may be an early marker for renal injury specific for the renal et al. and et al. have new markers for disease, including a new that is found in urine injury. Other studies have a spectrum of markers for renal injury but none have been in a number of patients advances in likely provide new for assessment Several for changes in GFR each has markers may be the most and for assessment of changes in GFR and these markers are being to be a marker for changes in GFR but requires in the ICU and other that use contrast agents are being in experimental models of ARF and could provide on the and extent of and the presence of renal It is evident that of the site and extent of injury would a of the underlying severity of renal dysfunction, could and could help guide specific therapeutic of for of from Other to ARF It is evident that definitions of ARF that rely on serum creatinine and urine output do not the spectrum of ARF in clinical in they do not the clinical to for therapeutic Indeed, in the timing of to renal injury has to heterogeneity in intervention studies in ARF and may be for the lack of It is also recognized that of care (including the timing of of of and the use of likely the course and outcomes ARF. diagnostic and classification are to allow for advances in ARF research and clinical We on two but related we to the of and specific markers to the site and severity of renal injury and to track functional change over we to more define the course of ARF in a of and factors that the renal response to injury and outcomes from ARF. this increase diagnostic specificity and allow for more of For instance, the of radiocontrast and underlying renal function determine the of creatinine and incidence of In most elevations in serum creatinine are seen within 48 h radiocontrast exposure, so that the of a or early therapeutic intervention can be In contrast, when the nature and timing of insult are less well characterized (e.g., of in a with in serum creatinine and urine output are less of the pattern of with the response could help to guide the timing of specific interventions, and compare two or more interventions in clinical trials. We the heterogeneity of ARF and the absence of specific markers for renal injury, an effective diagnostic and classification should parameters other than the renal response to injury. organ system also affects the ARF For instance, serum creatinine levels may the severity of renal in hepatic failure with elevated levels and the for may further GFR In other and response factors have been in the definition and classification of disease (see and the Failure classification for sepsis Criteria To the to we should into several factors that to injury, the nature and timing of the the response of the to the and the consequences of the ARF We that a new definition for ARF should from each of these would to a more definition for the and would the assessment of interventions in this Table a definition for ARF that variables from each of the is grade an risk of classification of acute renal a for staging ARF that is based on the criteria within each We based grade on the risk of ARF from studies. We grade the nature and timing of the insult on the basis of of the specific insult and the time from the insult to the of We grade the response variables the criteria by the Acute We define nonrenal organ by organ failure scores Gradations in the and reflect levels of risk for an whereas those in the and reflect an severity of We that at each of individual patients would be graded within each On the basis of the in the patients would be into the of ARF and would stages during the course of the disease and in ARF et al. demonstrated three of ARF On the basis of these is widely that clinical ARF has three and The of each on the basis of the presence of disease and the nature and type of et al. and have recently a classification for ARF that phases for this and recovery in response has also been established in experimental ARF the response can be graded on the basis of for the and this is being as a therapeutic (e.g., may the response to the in renal function in ARF. For any given the pattern of ARF on or to injury and the nature and severity or of injury. The renal response is likely to be on these two factors and may in determine nonrenal organ dysfunction. the course is nonrenal organ further the renal of acute renal failure aimed at the and established ARF. from with individuals on the basis of may be to more define the time course of disease and time points for intervention is similar in to sepsis syndrome, where of have been for therapeutic The multicenter of the to be within h of the organ system failure the nature and timing of the insult in ARF were one could a specific therapeutic to ameliorate the injury and the within a known therapeutic the renal response can be can be and on the presence or absence of the underlying nonrenal organ dysfunction, specific interventions, as dialysis could be the These The definitions could be used to the presence of disease by the response variable graded for the for ARF and of the increase in serum creatinine of mg/dl with a grade of and an insult grade of is more likely to represent significant injury (and to be by more than the and insult were and an grade of a similar increase in serum creatinine might require dialytic A of versus course could be in diagnostic and classification in aimed at and The multidimensional can be with A with creatinine and no grade seen h a contrast grade with a urine output of and a follow-up serum creatinine of mg/dl grade with no nonrenal grade would be A 48 h a change in serum creatinine to mg/dl and a in urine output to evidence of On the basis of these two time is evident that this has a in renal representing an from the injury or injury. the for change in renal function based on the timing of the contrast would have been a to creatinine within the from the response would an of the initial injury A therapeutic intervention with a that could ameliorate injury or could then be for this In a example, a who coronary artery and replacement is The is with a creatinine of 2.0 mg/dl and a history of serum creatinine from to mg/dl with urine output ml/d) on an There was no in during or and has not any There was no evidence for or on is with on and requires two to a of is 7 mg/dl and serum On the basis of of is to have of to and diagnostic would be considered to be acute on renal failure. Many and care (and would not with dialysis or in this because the is urine and has no By use of the classification this would be the presence of and in renal a could be to dialytic although is no evidence on which to make this data on outcomes associated with patients in clinical would be more to than data on all patients as acute on renal a Framework for and Although a more diagnostic and classification system would be to could also be used to of For example, in a comparing dialysis or one could that dialysis should for subjects with of or with of or A more and classification could new and markers from and investigations. For instance, were that predict the renal these could be included as new criteria in the Similarly, new injury markers could be included in the insult and provide more specific on the nature and severity of disease. Other markers of renal function (e.g., or of organ (e.g., could be included to other and The definitions are based on in other as acute liver disease, and The criteria are and to The could be applied to a or for individual The could also and However, are several limitations to (and our we that the are but the and of each to the definition and classification has not been We do not these are the or are others that to be We do not have markers for defining the consequences of renal dysfunction. We to define the nature and severity of underlying disease with the markers at We that the established for response variables are related to outcomes and are We the of the variables to determine additional definitions would be It is evident that studies to the of the definition in several However, we a new diagnostic and classification to help management and into this ARF continues to be a a significant of time for and to the that ARF to outcomes in the critically little progress has been in this of the key impediments to progress is the lack of a uniform definition for this disease. definitions that rely on changes in serum creatinine and urine output are neither nor We a new classification for ARF that from other in an to data and, we studies to the of construct and to the and of within each we (or a is not a On the we more in ARF effective interventions, and, mortality rates in the single study was by a from the for and by The to in Acute Renal Disease study the of of and