In other words, treatment selection bias and patients’ underlying severity are major confounders making the assessment of RRT efficacy selleck chemical EPZ-5676 challenging.Our study brings a new insight in the field. By using the propensity technique, we were able to compare hospital mortality rates in matched patients with and without RRT, having a close probability of receiving RRT (somewhat as though RRT had been ‘randomly assigned’). Moreover, since the SAPS II score was included in the propensity regressions, matched patients with and without RRT had also a similar predicted hospital mortality. Consequently, the risk of biased assessment of the association between RRT and hospital mortality was minimized.Like in the interesting study of Elseviers et al.

[22] that reported an increased risk of death for RRT compared to conservative treatment in ICU patients after extensive adjustment on disease severity, we failed to demonstrate any beneficial effect of RRT. While it cannot be totally run out that RRT per se is potentially harmful (hemodynamic instability, central venous catheter-related blood stream infections, inflammation and coagulation disorders, which are common complications of RRT, may well have outweighed its metabolic benefits), these results emphasize the need for a critical reappraisal of current RRT practices and definitions of AKI. Particularly, it must be kept in mind that timing of RRT initiation is undoubtedly a key issue. In this regard, a plausible explanation for our findings is that RRT was in fact initiated too late.

Actually, patients were classified according to the glomerular filtration rate (GFR) criteria of RIFLE whereas increases in serum creatinine often lag behind the true reduction in GFR. Thus, although RRT was in place within 24 h after reaching maximum RIFLE class in the vast majority of patients, it might well have been initiated at a more advanced stage of renal dysfunction than clinically appreciated. So, our results do not imply, as one may believe at first sight, that RRT should be abandoned. Rather, the key message could be: ‘initiate RRT as early as possible’. That patients who received RRT had more coexisting organ failures on reaching maximum RIFLE class than their matched controls lends support to this hypothesis of delayed AKI diagnosis and RRT.

Since initiation of RRT when multiple organ failures are present probably limits its ability to improve patients’ outcomes, the utilization of highly sensitive and early diagnostic biomarkers such as cystatin C or neutrophil gelatinase-associated lipocalin (instead of serum creatinine) Cilengitide as triggers for RRT is worth considering for future investigations in the ICU [23-30].Despite the use of an original statistical approach minimizing the risk of bias, our study has potential limitations that merit consideration.First, residual confounding cannot be totally excluded because of the observational design.