The study was approved by the institutional review board of the French Society for Intensive Care Medicine (SRLF-CE-07-212), which waived the need for signed informed consent. Patients and their next of kin were informed, however, and none refused to participate. Three ICUs in university hospitals participated in the study between April and September 2008. Patients admitted to the participating ICUs were included, except those younger than 18 years of age, pregnant women, patients receiving dialysis for an underlying chronic kidney disease and patients with evidence of obstructive renal failure. Patients from whom urine could not be collected during the first six hours were excluded from this study.

Each patient was assessed during the first 12 hours following ICU admission. Plasma sodium, urea and creatinine levels were measured at ICU admission, and urine was collected over the next six hours.

AKI was defined according to the Acute Kidney Injury Network classification scheme 16 as a serum creatinine level increase of 26.4 μmol/L or more, a serum creatinine increase ≥ 150% from baseline or urine output < 0.5 mL/kg/hour for six hours or more. For patients whose baseline serum creatinine level was unknown, this variable was estimated using the Modification of Diet in Renal Disease (MDRD) formula 16 17 .

Transient AKI was defined as AKI (of any stage) with a cause of renal hypoperfusion (that is, shock; dehydration; a medication interfering with renal perfusion, such as angiotensine-converting enzyme inhibitor; and so on) and recovery within three days. Recovery was defined as reversal of oliguria (in the absence of diuretics), and/or a 50% or greater decrease in serum creatinine 18 , and/or return of serum creatinine to the baseline value (whether measured or estimated using the MDRD formula 16 17 ). Persistent AKI was defined as renal dysfunction without recovery within three days. Oliguria was defined as urine output < 0.5 mL/kg/hour for six hours or more.

The FeNa percentage was calculated as ([urinary sodium/serum sodium]/[urinary creatinine/serum creatinine]) ×100. The FeUrea percentage was calculated as ([urinary urea/serum urea]/[urinary creatinine/serum creatinine]) ×100.

The Logistic Organ Dysfunction (LOD) score and the Simplified Acute Physiology Score version II (SAPS II) score were calculated at study inclusion 19 20 , and the Knaus scale score was determined to evaluate chronic health status at ICU admission (A: no limitation of activity, B: moderate limitation, C: severe limitation, and D: bedridden or institutionalized) 21 . Sepsis was diagnosed using the criteria developed by the American College of Chest Physicians/Society of Critical Care Medicine consensus conference 22 . Individual organ failure was defined as a LOD score greater than 1 point for each system except the kidney 19 .

Patients remaining in the ICU for < 72 hours were secondarily excluded from the analysis, since they could not be classified as having transient or persistent AKI according to our definition. The results are reported as medians and interquartile ranges (IQRs), numbers and percentages or as means ± standard deviations (SD) to express the percentage changes. Categorical variables were compared using Fisher's exact test, and continuous variables were compared using the nonparametric Wilcoxon signed-rank test or the Mann-Whitney U test for pairwise comparisons. The Friedman test was used to compare continuous variables across the three groups.

To determine how well FeUrea distinguished transient from persistent AKI (our primary objective), we plotted the receiver-operating characteristic (ROC) curves of the proportion of true positives against the proportion of false positives, depending on the prediction rule used to classify patients as having persistent AKI. A 2 × 2 table was established to determine the sensitivity and specificity of FeUrea in diagnosing persistent AKI. Cutoff values, defined as threshold values that maximized the sum of sensitivity and specificity, were determined on the ROC curves. The positive and negative likelihood (LH) ratios were computed. The same strategy was used to assess our secondary objectives, namely, the performance of the usual urinary indices in these patients and the performance of the usual urinary indices and of FeUrea in the subgroup of patients receiving diuretics.

Last, to confirm the input of urinary indices to detect persistent AKI, we performed logistic regression analyses to identify variables significantly associated with persistent AKI measured by the estimated odds ratio (OR) with the 95% confidence interval (95% CI). Variables yielding P values < 0.20 in the bivariate analyses were entered into a backward stepwise logistic regression model in which persistent AKI was the variable of interest. The covariates were entered into the model with critical entry and removal P values of 0.2 and 0.1, respectively. Last, since the performance of FeUrea was the primary objective of this study, this variable was forced into the final model. Colinearity and interactions were tested. The Hosmer-Lemeshow test was used to check the goodness of fit of the logistic regression.

All tests were two-sided, and P values < 0.05 were considered statistically significant. Statistical tests were performed using the SAS version 6.12 software package (SAS Institute, Cary, NC, USA).

During the study period, 203 patients with a median age of 61 years (46 to 73) were included. Their main characteristics are reported in Table 1. According to our definitions, 67 patients (33%) had no AKI, 54 patients (26.6%) had transient AKI and 82 patients (40.4%) had persistent AKI.

At ICU admission, the median SAPS II score was 46 (34 to 60) and the median LOD score was 6 (4 to 9). Most patients were admitted for medical conditions (91.1%). The main risk factors for AKI were sepsis (67.5%), aminoglycoside therapy (20.7%), chronic heart failure (19.8%), chronic kidney disease (16.3%) and exposure to iodinated contrast agents (8.9%).

At the time of the study, no patient was being treated with renal replacement therapy (RRT). Forty-five patients required RRT during their ICU stay, usually during the first three days in the ICU (41 of 45 patients). Each of the patients requiring RRT during the first three days in the ICU had persistent AKI, whereas the remaining four patients had no AKI at ICU admission and required RRT later during their ICU stay.

Median FeUrea was 37% (26 to 49) overall, 39% (28 to 40) in patients without AKI, 41% (29 to 54) in patients with transient AKI and 32% (22 to 51) in patients with persistent AKI (P = 0.12). Figures 1a, b and 1c show the distributions of FeNa, FeUrea and urine/plasma (U/P) urea ratios, respectively, in each group.

The area under the ROC curve was 0.59 (95% CI 0.49 to 0.70; P = 0.06) (Figure 2). At the usual cutoff (35%), FeUrea predicted persistent AKI with 63% sensitivity and 54% specificity (Table 2), yielding a positive LH of 1.37 and a negative LH of 0.68. In the study population, the optimal cutoff was 37%. However, the performance of FeUrea at this cutoff was poor (66% sensitivity and 53% specificity) (Table 2).

The performance characteristics of classical urinary indices for detecting persistent AKI are reported in Table 2, with the usual and optimal cutoffs in the study population. Performance was best for the U/P urea ratio (ROC curve area under the curve (AUC) 0.71 (0.62 to 0.80)) (Figure 2). A U/P urea ratio < 12 had 66% sensitivity and 66% specificity for persistent AKI (positive LH, 1.94; negative LH, 0.52). When entered into a regression logistic model, none of these urinary indices were independently associated with persistent AKI. Three variables were found to be associated with persistent AKI: chronic kidney disease (OR 11.89, 95% CI 2.52 to 56.24; P = 0.02), need for vasopressors at ICU admission (OR 2.60, 95% CI 1.15 to 5.91) and oliguria at ICU admission (OR 2.50, 95% CI 1.11 to 5.63). The model had good calibration (goodness of fit P = 0.88). FeUrea was then forced into the final model and was not selected.

Overall, 67 patients (33%) received diuretics before or at ICU admission. Among them, 17 had no AKI (25.4% of patients without AKI), 18 had transient AKI (33.3% of patients with transient AKI) and 32 had persistent AKI (39% of patients with persistent AKI). The performance characteristics of urinary indices in patients undergoing diuretic therapy are reported in Table 2. As with the overall population, the performance of FeUrea in this patient subgroup was poor (ROC curve AUC 0.58 (0.41 to 0.75)). The U/P urea ratio performed satisfactorily in differentiating transient from persistent AKI (ROC curve AUC 0.82 (0.70 to 0.94)). With a U/P urea ratio cutoff of 10, sensitivity was 72%, specificity was 69%, positive LH was 2.32 and negative LH was 0.41.

Overall, 137 patients (67%) had sepsis at ICU admission. Among them, 43 had no AKI (64.2% of patients without AKI), 33 had transient AKI (61.1% of patients with transient AKI) and 61 had persistent AKI (74.4% of patients with persistent AKI). The performance characteristics of urinary indices in patients with sepsis are reported in Table 2. As with the overall population, the performance of FeUrea in this patient subgroup was poor (ROC curve AUC 0.56 (0.43 to 0.68)). The performance of other urinary indices was similar to that in the overall patient population.