In critically ill patients, FeUrea was not helpful in differentiating transient AKI from persistent AKI. Both in the overall population and in the subgroup of patients receiving diuretics, FeUrea performed less well than FeNa or the U/P urea ratio.
There is little scientific evidence to support the use of FeUrea. Only three studies have evaluated the accuracy of FeUrea in distinguishing transient from persistent AKI [11
]. Their results are conflicting. In one study, FeUrea was 90% sensitive and 96% specific in differentiating transient from persistent AKI when a cutoff of 35% was used [11
]. Conversely, another study found very poor diagnostic accuracy of FeUrea [12
]. Several factors may explain these discordant results. First, these studies were single-center cohort studies and included only patients who were referred to nephrologists [11
]. In addition, the study populations were poorly described but include both critically ill patients and patients in wards. Therefore, selection bias and differences between the institutions and study populations may explain the discrepancies [11
]. Furthermore, FeUrea reflects the ratio of urea clearance over creatinine clearance ratio. Variations in creatinine clearance may therefore modify FeUrea. In the study that found good performance of FeUrea [11
], wide differences in creatinine clearance can be suspected between patients with transient AKI and those with persistent AKI: serum creatinine levels were 140 ± 22 μmol/L and 520 ± 22 μmol/L (means +/- SD) in these two groups, respectively.
Interestingly, the performance of urinary indices in our study was poor. Several factors may explain this finding. First, although many publications have advocated the use of urinary biochemistry indices to differentiate transient from persistent AKI, these indices have not been extensively studied in critically ill patients or in patients with sepsis [4
]. The few published studies have had several limitations: most of them were single-center case series or retrospective studies, the definition of AKI varied across the studies and the definition of transient AKI also varied, being subjective in most instances [24
]. Several of these studies included patients who did not have critical illnesses [11
]. In addition, we chose a definition allowing for a distinction between transient and persistent AKI. Our study was therefore not designed to evaluate the interest of these indices in distinguishing prerenal and intrinsic AKI. This point may partly explain the poor performance of the urinary indices in our study. Last, most of the studied patients had sepsis or shock at ICU admission. This condition is frequently associated with renal handling of sodium or water independently of an underlying AKI. This may also explain the poor performance of these indices in the studied population. Nevertheless, taking these factors into account, and although the usual urinary indices were able to differentiate transient from persistent AKI, the accuracy of the indices was poor and none of them were independently associated with the diagnosis of persistent AKI, indicating a need to identify other biomarkers.
Our study has several limitations. First, our definition of transient AKI was mainly based on renal function recovery. Indeed, an accurate definition of prerenal AKI would have required a highly subjective definition based on clinical histories, physical examinations and physicians' judgments [11
]. In addition, AKI is mainly due to sepsis in critically ill patients, and in this setting there is frequently a continuum between volume depletion and persistent kidney injuries rather than two distinct entities, with the two mechanisms being frequently associated. Therefore, we chose a definition that relies only on an objective criterion. This point needs to be taken into account to interpret our findings. In the same way, the course of kidney function may have been modified by factors following study inclusion. However, in the ICU setting, predicting which patients will have persistent AKI may help to optimize treatment, such as by promptly restoring renal perfusion, limiting fluids or starting RRT. Our definition was highly sensitive for detecting patients with transient AKI (none of the patients in this group required RRT) but lacked specificity, since only 50% of patients in the persistent AKI group required RRT. Additional studies may help to determine the definition that best matches the need for RRT.
Second, although renal function was assessed within a few hours after ICU admission, the time course of the urinary indices was not evaluated. FeNa is known to vary during the first 12 to 24 hours in critically ill patients [13
]. Few data are available on the time course of other urinary indices [13
]. Any variations might explain the poor performance of FeNa or the other urinary indices. Therefore, these variations need to be investigated to determine the optimal time for renal assessment in critically ill patients.
Last, few of our patients received diuretics. The poor performance of urinary indices in this subgroup may therefore be related to low statistical power.