There have been great difficulties for decades in defining the syndrome which we now refer to as acute renal failure (ARF). The inconvenience of not having a single definition was first addressed by a review in which 28 studies published in the period between 1965 and 19891 were seen to have used different criteria for diagnosing the condition.

In the year 2001, Bellomo, Kellum and Ronco2 proposed unification of the definition of ARF, and considered that only two of the multiple functions of the kidneys–urine production and the excretion of waste products of nitrogen metabolism–are easily and routinely determined in all Intensive Care Units (ICUs).3 Accordingly, in 2004 the Second International Consensus Conference of the Acute Dialysis Quality Initiative (ADQI)4 proposed the diagnosis and stratification of ARF according to the RIFLE classification, contemplating three levels or stages of acute renal dysfunction (“risk”, “injury” and “failure”) according to the changes in plasma creatinine levels or urine output.

The validity of this approach and its capacity to predict mortality have been evaluated by a number of studies. Ricci et al.,5 in a systematic review of all the studies using the RIFLE criteria between 2004 and 2007, identified 24 articles (with over 71,000 patients) that were scantly comparable among each other and moreover had clear weaknesses: small sample sizes, heterogeneous populations (healthy subjects, severe burn victims, bone marrow transplant patients, heart surgery patients, liver transplant patients, etc.), or participation of a single institution (single-center studies). Not surprisingly, the review reported an extremely variable incidence of ARF (10–82%).

The results of a second consensus conference with the participation of representative specialists in Critical Care Medicine and Nephrology (the Acute Kidney Injury Network) were published in 2007,6 and proposed the AKIN classification with the purpose of improving the sensitivity and specificity of the RIFLE system.

Large studies have evaluated both systems. In 2008, Bagshaw et al.7 published the results of the analysis of a database containing over 120,000 patients, comparing the RIFLE and AKIN systems. The incidence of ARF was 36.1% and 37.1%, respectively, and despite the weaknesses, the authors concluded that AKIN does not offer advantages in terms of sensitivity, consistency or predictive capacity with respect to RIFLE.

In 2008, Osterman et al.8 applied the AKIN classification in 22,300 patients, recording an incidence of ARF of 35.4% and a good correlation between the severity of ARF and the prognosis–though only the most severe disease grade was identified as an independent mortality predictor. In 2009, Joannidis et al.9 published a comparison of the RIFLE versus the AKIN system in 14,300 patients, based on the data of the SAPS-3. The incidence of ARF was found to be 28.5% and 38.5%, respectively, and the RIFLE classification showed greater consistency and a high ARF detection rate. However, the lack of urine output information biased the results, since AKIN grades 1 and 2, and RIFLE grades I and F, could not be differentiated. Both of the mentioned studies demonstrated an increase in morbidity-mortality associated to the development of ARF.

The recommendations of an international consensus conference held in Canada in 200710 with the purpose of establishing guidelines for the prevention and management of ARF in critical patients were published in 2010. The mentioned consensus document considered cystatin C to be a promising marker that appears to detect ARF earlier than changes in plasma creatinine. However, both this and other biomarkers (NGAL, kidney injury molecule-1, interleukin-18) are still in the investigational stages. It therefore was concluded that plasma creatinine remains the main marker (despite the disadvantages related to the influence of body weight, catabolic status, the presence of rhabdomyolysis, dilutional effects and drugs or other substances that alter creatinine secretion), along with the measurement of diuresis. The document thus recommends the RIFLE and AKIN criteria for diagnosing ARF, and points out that although validation is still needed, the AKIN system is able to identify ARF earlier than the RIFLE classification.

The present study analyzes the incidence of ARF in different ICUs, using the RIFLE and AKIN criteria applied in strict compliance with the consensus definitions, and compares the usefulness of both systems in predicting mortality and duration of stay in the ICU.

According to our results, the incidence of ARF was greater than in the large studies published in the literature, on applying both the RIFLE system and the AKIN classification (Table 2). This increased incidence was observed for all grades of severity.

The RIFLE system classified a larger number of patients with ARF than the AKIN classification, in contrast to the observations of Brochard.10 The difference in the incidence of ARF according to RIFLE and AKIN was circumscribed to the intermediate and severe cases of ARF, since the proportion of patients with mild ARF was similar according to both RIFLE and AKIN. On the other hand, there were patients classified as having ARF according to the RIFLE system but not according to the AKIN classification, and vice versa. This suggested that the use of both systems would increase the capacity to detect ARF, as was effectively confirmed by the results obtained.

On comparing the RIFLE and AKIN classifications, the differences can only be attributed to the C criterion, since the U criterion has the same cutoff points in both systems. The weaknesses of using plasma creatinine for the diagnosis of ARF have been widely commented in the literature. In the case of the RIFLE system, the C criterion is based on comparison of the actual MPC of the patient with the BPC intended to represent “ideal creatinine”, and is estimated according to the MDRD formula. The latter has been validated only in patients who are stable (an uncommon circumstance in the critically ill), and ignores the nonlinear relationship between the creatinine levels and the glomerular filtration rate (GFR). Moreover, this formula includes information which is not usually reliable in the ICU: the patient weight is often not known or is altered as a consequence of losses or increments in body water, in the same way as the creatinine distribution volume. On the other hand, the formula also contemplates an arbitrary “normal” estimated GFR of 75ml/min/m2, when in fact this value may differ widely among normal individuals with different dietetic habits or with different renal functional reserves.3,12 The MDRD formula likewise does not take into account the changes in MPC due to non-renal causes, as in patients with diminished muscle mass or impaired liver function. All these factors question the usefulness of BPC in the RIFLE classification. In the case of the AKIN system, comparison is made of the actual MPC with the MPC value of the previous 48h. Here the previous creatinine value is not calculated but measured–suggesting that the results could be more reliable than in the case of the RIFLE system. However, a patient may have an already elevated previous MPC if ARF developed days before determination of the value, and consequently comparison of the two MPC values would be unable to detect ARF. Thus, analysis of the changes in creatinine is questioned in both methods, and discussion focuses on which of the two is more reliable. Since there is no gold standard for defining ARF, the specificity of the two methods cannot be compared. However, as suggested by Bellomo in 2004,3 the importance of the definition of ARF does not seem to center on whether it is “correct” or “reliable”, but on its usefulness. We thus consider that the most useful method should be that capable of detecting a larger number of cases of ARF with the purpose of adopting preventive or early management measures, despite the consequent incidence of false-positive results.

If we assume that the most useful method is that able to detect a larger number of patients with ARF, then the RIFLE system would be more useful than the AKIN classification. In turn, the use of RIFLE and AKIN simultaneously in one same patient would prove more useful than RIFLE alone.

In our study, the U criterion detected a larger number of patients with ARF than the C criterion, and was therefore more effective. This increased sensitivity of the U criterion could be explained by the fact that we stratified the patients in an hourly manner (according to a calculated hourly diuresis), in contrast to creatinine, which requires a new value after 24h in the case of the RIFLE system or after 48h in the AKIN classification.

In our experience there was great variation in the way of monitoring diuresis in the different participating centers, and the literature likewise describes a lack of uniformity in the way this parameter is determined. On the other hand, when the urinary tract is not catheterized it is not possible to record the urine produced every hour–theoretically making it impossible to apply the U criterion recommended by consensus in such patients who, although perhaps in a less critical condition, are not free from risk of developing ARF. For these reasons we decided to implement calculated hourly diuresis (CHD), making it possible to apply the RIFLE and AKIN systems with greater precision and obtain comparable results in all patients, i.e., with or without urinary catheterization.

Since diuresis was evaluated in ml/day, the sensitivity in diagnosing ARF was 66.7% and 62.8% in relation to RIFLE and AKIN, respectively; in other words, daily diuresis as an isolated value is of scant diagnostic utility. However, the decrease in hourly urine output showed a good correlation to the presence of ARF and to more severe grades of the disease. The diuresis/time ratio therefore could be regarded as having prognostic value.

The increase in the values of MPC among the patients with ARF was not correlated to progressing severity of ARF when using the AKIN system, though a correlation was observed in the case of the RIFLE classification. Here we must remember that the creatinine data were registered daily in the RIFLE system versus once every 48h in the AKIN system; as a result, we can assume that increases in MPC have prognostic value when creatinine is measured at least once every 24h. On the other hand, the isolated MPC values showed a sensitivity of 75.5% and 59.9% in relation to RIFLE and AKIN, respectively, i.e., their diagnostic value is poor.

In coincidence with the published data, our mortality rates were significantly higher among the patients with ARF than in the patients without ARF, on applying both the RIFLE and the AKIN classification–higher rates being associated to increased severity of the disease. The results were similar on stratifying the patients according to the U criterion or C criterion separately. The larger proportion of deaths recorded among non-surgical patients is also consistent with the findings in the literature.

In agreement with the observations of other authors, our data showed patients with ARF to be older, with a larger number of previous disease conditions, and with higher BMI values and APACHE II and SOFA scores than patients without ARF. There were no significant differences with respect to patient sex or days of stay in the ICU. Furthermore, we found the highest incidence of ARF to correspond to non-surgical patients.

Since there is no gold standard for knowing the true sensitivity and specificity of the RIFLE and AKIN systems, comparative assessment should aim to establish which of the two systems is more useful for improving the outcome in critically ill patients.

The literature continues to present data that are difficult to compare, despite the consensus-based definitions. Perhaps the incidences of ARF reported in the literature would have been greater if strict application of the U criterion had been observed.

Despite the recommendations, it remains unclear how diuresis should be measured, particularly in patients without catheterization of the urinary tract. In our experience, CHD allowed application of the urinary criterion in a standardized manner in all patients, independently of the diuresis monitoring intervals and the presence or absence of a bladder catheter. We therefore recommend its use in this context.

We feel the originality of our study to be fundamentally attributable to the fact of having applied the U criterion exactly as recommended by consensus, in contrast to most studies in which insufficient or absent diuresis information constituted an acknowledged weak point.

The authors declare that they have no conflicts of interest.