Besides, selleckchem that the two propensity models yielded similar results made the hypothesis of having omitted an important confounding variable unlikely.Second, we encountered the same problem as others [31,32]: the 6- and 12-h urine outputs were not recorded in our database. Therefore, patients were classified according to the GFR criteria only. Patients classified according the GFR criteria seem to be more severely ill and have slightly higher mortality rates than their counterparts classified according to the urine output criteria [33,34]. Having considered both criteria may have resulted in a different estimation of RRT efficacy. Yet, urine output does not differentiate functional (pre-renal) AKI from organic AKI and new serum or urine biomarkers are probably much more reliable for the early diagnosis of AKI.

Third, the MDRD equation used to estimate baseline creatinine values has not been validated in ICU patients. Nevertheless, the sensitivity analysis including only data from patients with a normal serum creatinine value on ICU admission yielded similar results as the full analysis, showing that the use of the MDRD equation did not bias the results.Fourth, the use of the MDRD equation to estimate baseline creatinine values refrains from precisely establishing AKI onset (that is, patients with an apparent early-onset AKI may in fact have developed AKI for several days before ICU admission). This could be problematic in that the prognosis of early AKI may differ from that of late AKI.

That results of the sensitivity analysis, including only data from patients with a normal serum creatinine value on ICU admission, yielded similar results as the full analysis runs counter to the hypothesis of differential prognosis and impact of RRT between early and late Cilengitide AKI. However, this issue needs further evaluation.Fifth, it might be argued that RRT initiation may have prevented R or I class patients from reaching a higher RIFLE class (thus leading to an underestimation of their degree of renal dysfunction, and subsequent comparison of RRT patients with non-RRT patients having a more severe renal dysfunction). Yet, this limit, which is inherent to the RIFLE classification, does not apply to F class patients. Since odds ratios of mortality associated with RRT in the whole population were similar as those in the F class patients, it is very unlikely that results were flawed by a potential misclassification bias induced by an underestimation of renal dysfunction in RRT patients.Sixth, the prognostic impact of the dose and initial modality of RRT was not assessed. It must be emphasized, however, that all randomized controlled trials conducted so far have showed equivalence between high and low doses, and continuous and intermittent RRT [16-21].