Defining acute kidney injury: playing hide-and-seek with the unknown man?

Abstract

Acute kidney injury (AKI) occurs frequently in hospitalized patients, especially at the intensive care unit (ICU), and is associated with a high morbidity, mortality and cost [1,2]. The disappointing quest for a golden bullet to treat or prevent AKI during the 1990s has been replaced by a crusade to develop (early) biomarkers of AKI [3]. The underlying reasoning for this change in research perspective is that the single-shot therapies that worked so well in the rat models do not work in humans because we lag behind with accurate, early diagnosis. As a result, several biomarkers have been identified and tested over the last decade in an attempt to diagnose AKI earlier. In a recent issue of Kidney International, Endre et al. [4] report the negative results of a trial using biomarkers to guide administration of rHuEPO to prevent or attenuate AKI. Most of the interest in this study is that the biomarkers used were not at all able to establish an accurate diagnosis of AKI, although they have been claimed previously to be accurate [5]. History seems to repeat itself: whereas some of these biomarkers seem to work properly in well-controlled conditions and restricted populations, such as in children postcardiac surgery [6], their performance decreases in mixed or more complex populations [3], and confidence intervals for sensitivity and specificity become wide [7]. Although Endre et al. [4] have to be congratulated for their courage in setting up this study, it is exemplary for the difficulties faced by all initiatives in the field of AKI: as long as a coherent and meaningful definition of AKI is lacking, all attempts for early diagnosis, prevention or treatment are bound to fail. The several initiatives undertaken over the last years to harmonize the definition of AKI [8–10] must be greatly credited for having brought the problem of AKI to the attention of non-nephrologists. However, although the principles of RIFLE and AKIN are well accepted, most papers use a ‘convenience’ format of RIFLE or AKIN in real practice. Also, the boundaries of the diagnostic criteria, especially in relation to the baseline value of creatinine, seem to change continuously [8,11,12]. Most attempts to come up with one definition neglect the simple truth that ‘the’ AKI as a single entity does not exist. For the clinician, AKI is the reduction in glomerular filtration rate, when most likely the kidney has already been damaged substantially, while for the experimental physiologist, harm to even a few tubular cells as evidenced by biomarkers is enough to establish the diagnosis. AKI is a multifaceted condition, in terms of severity, underlying disease and circumstances, and sequence of events. By creating stages of AKI, both RIFLE [10] and AKIN [9] tried to tackle this problem. It has been demonstrated that these stages correlate with outcome [13,14]. This has been accepted as validation of these definitions. But here loom a number of conceptual problems. First, in order to define a change in creatinine, a baseline value is needed, changing the problem of definition of AKI into a problem of defining a baseline value of creatinine, which might lead to different classifications [15]. In the study of Endre et al. [4], although all patients complied with one definition of AKI based on baseline creatinine, this ‘baseline’ creatinine was defined according to a pre-specified algorithm, consisting of five different definitions, de facto resulting in five different types of AKI. It has been demonstrated that the selection of baseline creatinine influences the categorization of AKI [15]. Second, besides the difference between a baseline and a second value, also the further kinetics of the evolution of serum creatinine should be taken into account. The current definitions of RIFLE and AKIN take into account neither the kinetics of creatinine nor the duration of an increase. When a patient with an established normal kidney function comes in with acute dehydration, his creatinine on admission might be several folds his baseline value, but will decrease rapidly by rehydration. This is cardinal for the concept of ‘fluid responsiveness’ [16], and the best marker to be followed here is the rise in urinary output, a neglected parameter in many studies. In a septic patient, declining urinary output despite fluid loading is often the first sign of impending AKI, but creatinine will only rise slowly in the beginning. The distinction between renal and pre-renal, from the therapeutic viewpoint the most crucial distinction to be made, can only be made by evaluating fluid responsiveness or the effect of increasing renal perfusion, not by a definition based on single changes from baseline in serum creatinine or a biomarker [17]. Third, relative risk rather than predictive value is used to indicate the increase