R-HuEPO treatment has renoprotective properties in experimental ischemic renal injury when given before, during or even after the injury
[5–7]. The effect of r-HuEPO on prevention and treatment of AKI in patients is still very controversial. In our study, we gave r-HuEPO shortly after cardiac surgery which was inefficient in preventing mild forms of AKI, independent of the dose used. Moreover, EPO did not modify NGAL urinary levels, nor did it reduce hospital stay and mortality. This contrasts with a recent study
 where erythropoietin given at the time of operation was efficient in preventing acute kidney injury, reducing ICU stays and decreasing mortality in cardiac surgery patients. However, timing of administration and type of r-HuEPO (epoetin-beta) differed from our study protocol. Another interventional study in high risk ICU patients, including patients with miscellaneous causes of kidney injuries, demonstrated that EPO was ineffective at preventing mild or severe AKI and did not modify mortality and need for dialysis
, similarly to our study. Notably in this study, patients were selected based on the rise in tubular biomarkers, and therefore were probably also treated after the onset of renal injury. In kidney transplant recipients, EPO administered prior to transplantation did not decrease the rate of delayed graft function
. Altogether, these studies show that erythropoietin may have some beneficial effect in human AKI when given before the ischemic injury. However, even an early post injury treatment, in opposition to what is observed in rodents, appears ineffective. This would limit the applicability of EPO treatment post-intervention in prevention of AKI in ICU patients or should lead to a more thorough identification of high-risk patients before intervention or before other procedures with a renal risk such as contrast media administration. Nevertheless, these conclusions are supported only by one positive monocentric study
 and more data are necessary. The ongoing EPO-TBI study (NCT 00987454) may provide an answer to this question. This study aims to include 606 patients having traumatic brain injury to evaluate EPO for brain protection but renal biomarkers will be also analyzed
. Finally, a study testing the effect of r-HuEPO using a scoring system identifying high-risk patients would be desirable.
Change in urinary NGAL levels was a primary outcome and pooled data showed no effect of r-HuEPO on uNGAL levels compared to placebo during the observation time. We observed that α-Epoetin 20’000ui had a tendency toward decreasing urinary NGAL in comparison to the higher dose, but not in comparison to placebo. The difference between high and low doses of r-HuEPO has not been described previously and we do not have a clear explanation for this result. This could be a statistical difference due to the small sample as clinically relevant parameters (death, length of stay, ICU admission, AKI incidence) were similar between the groups, although we cannot formally exclude a dose-dependent effect. The choice of NGAL levels as a primary outcome was made on the presumption of a better sensitivity of this test for AKI detection. However, establishing the sensitivity of NGAL is difficult given the lack of current gold standard for AKI diagnosis and we believe that NGAL levels over time may be of interest given its prognostic value for outcome
. In our patients we only measured NGAL, IL6, IL8 and cystatine C at ICU admission. Baseline values were not available. However, in patients suffering from AKI, both NGAL and cystatine C levels were significantly higher at admission than in non-AKI patients. Similar observations were made in patients that died during the observation period. Despite the low number of events, this tends to confirm that NGAL and cystatine C levels at admission (3–4 hours after surgery) are more valuable than creatinine levels in predicting AKI and mortality, as shown in other studies
[13, 14, 27].
In our study, the incidence of AKI (all stages) was 22.5%, but only three patients suffered from stage 2 AKI and no patient required dialysis. The incidence of severe AKI was low despite the high prevalence of diabetes and CKD and that most of the patients underwent valvular replacement. This may be related to the inclusion of elective patients only due to strict exclusion criteria. Furthermore no nephrotoxic drug was prescribed nor was a contrast agent administered during the observation period that could have increased the rate of renal events. Although 7 patients received contrast medium before surgery they did not have a higher rate of AKI. Recent observations also establish a lower incidence of severe AKI after cardiac surgery than previously reported, despite a remaining important repercussion of low grade AKI on mortality and costs
. This may be due to improvement in anesthetic care and surgical techniques. Given this limitation, we cannot demonstrate a protective effect of r-HuEPO on severe AKI after cardiac surgery. However, our study demonstrates that stage 1 AKI, a relatively frequent condition after heart surgery that is correlated to mortality, length of stay and costs, is not prevented by r-HuEPO treatment.
Erythropoietin has been shown to have immunomodulatory effects
[19, 20, 29]. In this study, IL6 and IL8 levels were elevated postoperatively and decreased thereafter. Administration of r-HuEPO shortly after surgery did not modify the plasma profile of these cytokines. As IL6 is emerging as a predictive factor for acute kidney injury and adverse post-operative outcomes
[17, 30], the absence of a beneficial effect of EPO on its levels parallels the unchanged incidence of acute kidney injury and mortality. This, together with the absence of a protective effect on renal function, does not support the use of r-HuEPO to decrease the anti-inflammatory response.
Our study has several limitations. First the number of events was quite small and for that reason the study may be underpowered even with the use of a sensitive primary outcome measure of change in urinary NGAL levels since the standard deviation of uNGAl was higher than expected. Secondly, we cannot exclude the presence of elevated NGAL levels before randomization, especially in patients with chronic kidney disease or diabetes, as blood was not taken from patients prior to surgery. Nevertheless diabetic patients did not show increased levels of uNGAL at randomization. As for CKD patients, data are difficult to analyze because of a small sample size. In each treatment group, the uNGAL level was higher at randomization in patients with CKD but the change in uNGAL level at 48 h and randomization or between treatment groups was not significantly different. Finally, we did not correct uNGAL for urinary creatinine levels, as this correction did not modify sensitivity of uNGAL in similar settings
. Nevertheless, this study is one of the few randomized studies studying EPO effects on renal function and is the only one testing the effect of EPO on inflammatory markers such as cytokines.