Postoperative acute kidney injury (AKI) is associated with poor outcomes in surgical patients. This study aims to evaluate whether the timing of renal replacement therapy (RRT) initiation affects the in-hospital mortality of patients with postoperative AKI.
This multicenter retrospective observational study, which was conducted in the intensive care units (ICUs) in a tertiary hospital (National Taiwan University Hospital) and its branch hospitals in Taiwan between January, 2002, and April, 2009, included adult patients with postoperative AKI who underwent RRT for predefined indications. The demographic data, comorbid diseases, types of surgery and RRT, and the indications for RRT were documented. Patients were categorized according to the period of time between the ICU admission and RRT initiation as the early (EG, ≦1 day), intermediate (IG, 2–3 days), and late (LG, ≧4 days) groups. The in-hospital mortality rate censored at 180 day was defined as the endpoint.
Six hundred forty-eight patients (418 men, mean age 63.0±15.9 years) were enrolled, and 379 patients (58.5%) died during the hospitalization. Both the estimated probability of death and the in-hospital mortality rates of the three groups represented U-curves. According to the Cox proportional hazard method, LG (hazard ratio, 1.527; 95% confidence interval, 1.152–2.024; P = 0.003, compared with IG group), age (1.014; 1.006–1.021), diabetes (1.279; 1.022–1.601; P = 0.031), cirrhosis (2.147; 1.421–3.242), extracorporeal membrane oxygenation support (1.811; 1.391–2.359), initial neurological dysfunction (1.448; 1.107–1.894; P = 0.007), pre-RRT mean arterial pressure (0.988; 0.981–0.995), inotropic equivalent (1.006; 1.001–1.012; P = 0.013), APACHE II scores (1.055; 1.037–1.073), and sepsis (1.939; 1.536–2.449) were independent predictors of the in-hospital mortality (All P<0.001 except otherwise stated).
Citation: Shiao C-C, Ko W-J, Wu V-C, Huang T-M, Lai C-F, Lin Y-F, et al. (2012) U-Curve Association between Timing of Renal Replacement Therapy Initiation and In-Hospital Mortality in Postoperative Acute Kidney Injury. PLoS ONE 7(8): e42952. https://doi.org/10.1371/journal.pone.0042952
Editor: Emmanuel A. Burdmann, University of Sao Paulo Medical School, Brazil
Received: March 3, 2012; Accepted: July 16, 2012; Published: August 28, 2012
Copyright: © Shiao et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This study was financially supported by The Ta-Tung Kidney Foundation, Taiwan National Science Council (grant NSC 96-2314-B-002-164, NSC 96-2314-B-002-033-MY3, NSC 97-2314-B-002-155-MY2, NSC 98-2314-B-002-155-MY4), and NTUH 098-001177, NTUH 100-001667. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Postoperative acute kidney injury (AKI), with rates of occurrence ranging from 0.8 to 30% due to the various definitions of the pathology and the different associated surgery types, is associated with not only the increased healthcare costs and duration of hospital stay, but also higher morbidity and mortality rates , , . Among the critically ill patients admitted to the intensive care unit (ICU), approximately 4–15% develop AKI that requires renal replacement therapy (RRT), and AKI is associated with a significant in-hospital mortality rate of 50–80% , . To date, RRT remains a critical supportive measure for AKI . The goal of RRT is to maintain solute clearance and fluid balance to prevent subsequent multi-organ damage as renal function recovers. However, there are limited data documenting the significant factors used to determine if and when it is appropriate to initiate RTT in this setting. Because of the great impact of RRT-requiring AKI on patients’ prognoses, many investigators have made every effort to define the optimal timing of RRT initiation.
Although some meta-analyses have revealed the probable beneficial effects of “earlier” initiation of RRT for AKI , , the non-standardized triggers for this therapy and the heterogeneous populations represented in the published studies preclude the possibility of forming a definitive recommendation. Besides the traditional markers, other parameters, including the duration of time between admission to the ICU and the initiation of RRT , were used to determine the timing of the RRT. A prospective, observational study by Bagshaw et al. found that RRT initiation within 2 days of the ICU admission was associated with better survival compared to the patients who received RRT after 5 days in the unit . The beneficial effect of earlier RRT was also seen in other studies, but their cutoff points relative to the ICU admission varied (e.g., “on admission” versus “at 24 hours” by Andrade et al. , “<2 days” versus “≧2 days” by Payen et al. , and “<3 days” versus “≧3 days” by Ostermann et al. ). Furthermore, the association between the temporal relationship and the outcomes was not supported by a prospective, randomized trial, in which no survival difference was observed between patients starting RRT at a mean time of 1.5 days and 3 days after the ICU admission . The heterogeneity of the surgical/medical patient population included in these investigations may also explain the discordant results generated, as different pathologies have varying influences on the prognoses for AKI , . Additionally, the different patient categories based on the time to the initiation of the RRT in these studies may have also played a role in the lack of a general consensus in the results. Regardless, simply categorizing all the participants into 2 or 3 groups accordingly may not truly reflect the complete clinical scenarios.
Thus, we designed this study to evaluate the association between the timing of RRT relative to the ICU admission and the patients’ in-hospital mortality rate. We also analyzed the factors that may predict the mortality in the setting of postoperative AKI.
Study Design and Participants
This multicenter, retrospective observational study was conducted using the NSARF (National Taiwan University Hospital Study Group on Acute Renal Failure) database in which the data was prospectively collected in the ICUs in a tertiary hospital (National Taiwan University Hospital) and its branch hospitals in Taiwan for quality and outcome assurance , , . The Institutional Review Board of National Taiwan University Hospital approved the study (No. 31 MD03) and determined that there was no need for informed consent as there was neither breach of privacy nor interference with clinical decisions. The data were analyzed anonymously as well.
Adult patients who underwent major surgeries with postoperative AKI requiring RRT in the ICUs between January 1, 2002, and April 30, 2009, were eligible for this study. The exclusion criteria included undergoing chronic dialysis, less than 18 years of age, receiving RRT before the surgery. The enrolled patients were categorized into three groups based on the duration of time between the ICU admission and the initiation of the RRT. The patients were designated as follows: the early group (EG, defined as 0–1 day), the intermediate group (IG, 2–3 days), and the late group (LG, ≧4 days). Under the same practice principles, those enrolled in this study were treated and followed until November 30, 2009.
The baseline demographic data, comorbid diseases, types of surgery and RRT, and indications for RRT were documented. Biochemical data, such as the complete blood cell count, the blood urea nitrogen (BUN), the serum creatinine (sCr), the estimated glomerular filtration rate (eGFR) calculated using the Chinese Modification of Diet in Renal Disease equation , the serum albumin, and the serum potassium, were recorded when the patient was admitted to the ICU and when RRT was initiated. The severity scores, including the Glasgow Coma Scale (GCS) scores, the Acute Physiology and Chronic Health Evaluation II (APACHE II) scores , and the Sequential Organ Failure Assessment (SOFA) scores , were also measured at the two time points. Mechanical ventilation (MV), extracorporeal membrane oxygenation (ECMO) support, inotropic equivalents [IE, = dopamine + dobutamine + (epinephrine + norepinephrine + isoproterenol) × 100+ milrinone ×15 (mcg/kg/min)] , and net fluid balance (% body weight [BW], = [pre-RRT BW − baseline BW]/baseline BW) at RRT initiation were recorded as well.
The baseline sCr was determined as the last value measured one month to one year prior to the index admission , or the lowest sCr value before operation during the index admission . The peak sCr was defined as the highest sCr before the initiation of RRT in the ICUs. The predefined indications for RRT initiation at our institutes were as follows: (1) azotemia (BUN>80 mg/dl and sCr>2 mg/dl) with uremic symptoms (encephalopathy, pericarditis, pleuritis); (2) oliguria (urine amount <100 mL/8 h) or anuria refractory to diuretics; (3) fluid overload refractory to diuretics use with a central venous pressure (CVP)>12 mmHg; (4) hyperkalemia (serum potassium>5.5 mmol/L) refractory to medical treatment; and (5) metabolic acidosis (a pH <7.2 in arterial blood gas) , . Other definitions were made as followings: diabetes mellistus (DM), previous usage of insulin or oral hypoglycemic agents; hypertension, usage of anti-hypertension agents or blood pressure>145/95 mmHg at the time of hospitalization; heart failure, New York Heart Association functional class IV; chronic kidney disease (CKD), baseline eGFR ≦45 ml/min/1.73 m2 for more than 3 months ; neurological dysfunction, GCS <9 points ; sepsis, persisted or progressive signs and symptoms of the systemic inflammatory response syndrome with clinical evidence of infection ; RRT wean-off, cessation from RRT for at least 30 days . Besides, major surgeries were defined if the length of hospital stay for patients in a given diagnosis-related group exceeded 2 days , , and which were divided into 5 categories, namely, neurosurgery, cardiovascular surgery (CVS), chest surgery (defined as surgery in the chest area excluding CVS and thoracic vascular surgery), abdominal surgery, and others.
The modalities of the RRT for individual patients were initially chosen and may have been subsequently changed depending on the patients’ hemodynamic stability. For those who needed an IE of more than 15 mcg/kg/min to maintain a systemic blood pressure up to 120 mmHg, continuous venovenous hemofiltration (CVVH) was performed using the high-flux filters (Hemofilter, PAN-10, Asahi Kasei, Japan) and the HF-400 (Informed, Geneva, Switzerland). The hemofiltration and blood flows were 35 ml/kg/hr and 200 ml/min, respectively. The replacement fluid was bicarbonate-buffered and was predilutionally administered at a dynamically adjusted rate to achieve the desired fluid therapy goals. The default fluid was composed of Na 142 mEq/L, bicarbonate 33 mEq/L, Ca 2.6 mEq/L, and Mg 1.4 mEq/L. For patients with a required IE of 5–15 mcg/kg/min, sustained low efficiency daily dialysis (SLEDD) or diafiltration (SLEDD-f) was used with a blood flow of 200 ml/min, a dialysate flow of 300 ml/min, and a hemofiltration flow of 35 ml/kg/hour. The duration of the filtration procedures was approximately 6–12 hrs, depending on the amount of ultrafiltration. Intermittent hemodialysis was performed for 4 hours except for the first and second sessions using the low-flux polysulfone hemofilters (KF-18C, Kawasumi Laboratories, Japan) with dialysate and blood flows of 500 ml/min and 200 ml/min, respectively. , , . Double lumen catheters were placed for vascular accesses.
The endpoint of this study was in-hospital mortality censored at 180 days. The censoring period was calculated from the RRT initiation to the mortality (in non-survivors) or the hospital discharge (in survivors).
The statistical analyses were performed using the R 2.12.1 (R Foundation for Statistical Computing, Vienna, Austria) software. The continuous data were expressed as mean ± standard deviation unless otherwise specified and compared using the Kruskal-Wallis Rank Sum Test or Wilcoxon Rank Sum Test with Bonferroni correction. The categorical variables were shown as numbers (percentages) and analyzed using Fisher’s Exact Test with Bonferroni correction. The Kaplan-Meier survival curves with the log-rank test were drawn to show the differences in patient survival among the three groups. The generalized additive models (GAM) were applied to measure the probability of death according to the individual calendar day from the ICU admission to the RRT initiation. We then analyzed the independent predictors of in-hospital mortality with the stepwise selection method of the Cox proportional hazards model, and identified predictors for the EG and LG groups by the logistic regression method. All the variables were selected for multivariate analysis if they had a P≦0.1 on univariate analysis or if they are considered to be clinically important. The basic model-fitting techniques for variable selection, goodness-of-fit assessment, and regression diagnostics were used in our regression analyses to ensure the quality of analysis results. Specifically, we used the stepwise variable selection procedure with both significance level for entry and significance level for stay set to 0.15 or greater to select the relevant covariates for the final Cox proportional hazards model. In all statistical analyses, a two-sided p≦0.05 was considered statistically significant.
Figure 1 shows the flowchart of patient selecting and categorizing. After screening of 1263 patients, 648 (418 men, mean age 63.0±15.9 years) were enrolled and categorized into EG (n = 256), IG (n = 180), and LG (n = 212). There were 20 patients (3.1%) undergoing neurosurgery, 58 (8.9%) with chest surgery, 347 (53.5%) with CVS, 183 (28.2%) with abdominal surgery, and 40 (6.2%) with other surgeries. The indications for RRT included 358 patients (55.2%) for azotemia with uremic symptoms, 408 (63.0%) for fluid overload, 543 (83.8%) for oliguria, and 80 (12.3%) for hyperkalemia or acidosis. Most of our patients had more than one indication for RRT initiation.
EG, IG, and LG were defined as ≦ 1 day, 2–3 days, and ≧ 4 days between ICU admission and RRT initiation, respectively. Abbreviations: AKI, acute kidney injury; EG, early group; ICU, intensive care unit; IG, intermediate group; LG, late group; RRT, renal replacement therapy.
A total of 379 patients (58.5%) died during hospitalization period within 180 days. By Kaplan-Meier method, we demonstrated the much lower survival proportion of both EG and LG groups as compared with IG (Log Rank, P = 0.005) (Figure 2). While both the estimated probability of death estimated using GAM (Figure 3) and in-hospital mortality rates of the 3 groups (59.0%, 47.8%, and 67.0%, consecutively. figure not shown) represented U-shaped curves.
The survival period was calculated from RRT initiation. EG (blue line, n = 243), IG (green line, n = 146), and LG (red line, n = 227) were defined as RRT initiation ≦ 1 days, 2–3 days, and ≧ 4 days after ICU admission, respectively. Abbreviations: EG, early group; ICU, intensive care unit; IG, intermediate group; LG, late group; RRT, renal replacement therapy.
The figure was drawn using generalized additive model. Adjusted by sex, age, diabetes mellitus, chronic kidney disease, cirrhosis, extracorporeal membrane oxygenation support, initial neurological dysfunction, as well as sepsis, mean arterial pressure, inotropic equivalent, and Acute Physiology and Chronic Health Evaluation II scores at RRT initiation. Abbreviations: EG, early group; ICU, intensive care unit; IG, intermediate group; LG, late group; RRT, renal replacement therapy.
The patients in the EG had higher proportion of CKD (P = 0.015), CVS, and ECMO support, but less chest and abdominal surgery (P = 0.001). They had the worst conditions at ICU admission among the 3 groups, including the severity scores (highest APACHE II and SOFA scores), neurological function (lowest GCS), hemodynamics [highest IE and CVP level, with lowest mean arterial pressure (MAP)], and renal function (highest BUN and sCr, and lowest eGFR). Subsequently, they received RRT soonest, with a mean period of only 0.6 days after ICU admission, for the indications of oliguria/anuria rather than azotemia.
On the contrary, the patients in the remaining two groups had lower severity of illness (lower APACHE II and SOFA), hemodynamics (lower CVP, IE, and higher MAP), and renal function (lower BUN, sCr, and higher eGFR) when they were admitted to ICU. However, their renal functions gradually got worse during the ICU admission, which made them start RRT after 2.6 days and 17.9 days in IG and LG, respectively, of ICU admission. At RRT initiation, although their hemodynamics were still better (lower IE and higher MAP), but renal function turned to be worse (higher BUN and sCr, lower eGFR, and higher proportion of RIFLE-I & F) than EG, while the severity scores and consciousness were not of statistical differences.
As to the patients in LG, they were older but had the best condition at ICU admission (lowest SOFA scores and IE, along with highest MAP and eGFR). They had higher percentage of receiving chest surgery and more sepsis at RRT which were initiated due to azotemia rather than oliguria/anuria, consequently accompanying a higher sepsis-associated mortality rate. (All P<0.001 except otherwise stated).
In the comparisons between survivors and non-survivors, the non-survivors were older (P = 0.014), along with higher proportion of initial neurological dysfunction, sepsis at RRT initiation, ECMO support, CVVH, MV (P = 0.032), but less CKD. They underwent more chest surgery (P = 0.026) but less neurosurgery (P = 0.038), and had higher proportion of fluid overload (P = 0.013) as indication for RRT than survivors.
At ICU admission, they had worse hemodynamics [higher IE (P = 0.013) with lower MAP (P = 0.006)] and neurological function (GCS scores), and higher severity of illness [higher APACHE-II and SOFA scores (P = 0.004)], but better renal function [lower sCr with higher eGFR]. The tendency persisted up to RRT initiation. (All P<0.001 except otherwise stated).
Independent Predictors for In-hospital Mortality (Table 3)
Multivariate analysis was performed to determine the predictors for in-hospital mortality. The variables put into analysis included age, DM, liver cirrhosis, heart failure, baseline CKD, ECMO, MV, initial neurological dysfunction, sepsis and some parameters (RIFLE-I & F, MAP, IE, APACHE II scores) at RRT initiation, EG and LG as compared with IG, indications for RRT (azotemia, fluid overload & oliguria, and hyperkalemia & acidosis), as well as surgery types (neuro-, chest, cardiovascular, abdominal, and other surgery).
When compared with IG, LG [hazard ratio (HR), 1.527; 95% confidence interval (CI), 1.152–2.024; P = 0.003] independently predicted in-hospital mortality, but EG (HR, 1.213; 95% CI, 0.914–1.612; P = 0.182) did not. Other independent predictors included age (HR, 1.014; 95% CI, 1.006–1.021), DM (HR, 1.279; 95% CI, 1.022–1.601; P = 0.031), cirrhosis (HR, 2.147; 95% CI, 1.421–3.242), ECMO support (HR, 1.811; 95% CI, 1.391–2.359), initial neurological dysfunction (HR, 1.448; 95% CI, 1.107–1.894; P = 0.007), sepsis at RRT initiation (HR, 1.939; 95% CI, 1.536–2.449), as well as MAP (HR, 0.988; 95% CI, 0.981–0.995), IE (HR, 1.006; 95% CI, 1.001–1.012; P = 0.013), and APACHE II scores (HR, 1.055; 95% CI, 1.037–1.073) at RRT initiation. (All P<0.001 except otherwise stated).
Predictors for Entering EG and LG
Basic demographic data, comorbidities, clinical parameters (MAP, IE, APACHE II) at ICU admission, ECMO, MV, initial neurological dysfunction, sepsis at RRT, indications for RRT, and surgery types were put into logistic regression method to determine the predictors for entering EG and LD. Baseline CKD [odds ratio (OR), 1.805; 95% CI, 1.188–2.740; P = 0.006], undergoing CVS (OR, 2.816; 95% CI, 1.901–4.173), with higher ICU APACHE II scores (OR, 1.113; 95% CI, 1.076–1.152) along with fluid overload (OR, 3.754; 95% CI, 1.319–10.688; P = 0.013) were predictors for entering EG. Whereas those who undergoing chest surgery (OR, 4.236; 95% CI, 2.072–8.659) with higher ICU MAP (OR, 1.009; 95% CI, 1.000–1.019; P = 0.050), receiving MV support (OR, 4.497; 95% CI, 1.809–11.180; P = 0.001), complicating with sepsis (OR, 1.914; 95% CI, 1.180–3.106; P = 0.009) and azotemia (OR, 2.039; 95% CI, 1.356–3.064; P = 0.001) had higher probability to receive RRT at a later day (≧4 days). (All P<0.001 except otherwise stated).
To our knowledge, this is the first investigation revealing the U-curve association between the timing of RRT initiation after the ICU admission and in-hospital mortality in postoperative AKI.
Independent Predictors of In-hospital Mortality
Compatible with other studies , , the in-hospital mortality rate in our population was 58.5%. The current analysis also determined several independent predictors of the in-hospital mortality (table 3). A large body of evidence has shown significant associations between increased mortality rate and older age, DM, liver dysfunction, neurological dysfunction, higher APACHE II scores, and the presence of sepsis , , , , . Also a BP that falls below a predetermined threshold is associated with a higher IE and poorer prognosis in the critically ill patients .
ECMO, a temporary cardiopulmonary support, may be associated with ischemia/reperfusion (I/R) kidney injury which is a major cause of postoperative AKI in ECMO patients . Due to the interaction of numerous factors, including immune dysregulation, platelet dysfunction, and RRT-related hemodynamic instability and infection, the I/R AKI is associated with a high mortality rate ranging from 40–80% in those with ECMO needing RRT , .
Additionally, late initiation of RRT could also independently predict in-hospital mortality. In our study, the LG had a higher proportion of chest surgery, which was thought to more likely lead to hospital-acquired infections and the “septic AKI” . The patients in the LG could have either late onset AKI or delayed RRT for AKI that had been present for a longer duration. Late development of AKI is usually related to sepsis, while both the hospital-acquired infection  and septic AKI  are associated with higher in-hospital mortality. The longer duration of AKI  and later RRT initiation ,  are also shown to have a significant adverse impact on the patient survival.
It’s worthy of mention that the existing CKD is recognized as a potent predictor of AKI and may affect the patient outcome , . In our study, the baseline CKD was associated with higher proportion of “hyperkalemia or acidosis” (P = 0.046) as the indication for RRT initiation. It was also found to be an independent predictor for entering EG, but however, not for the in-hospital mortality.
Contributors of the Two Peaks in the U-curve
The first peak of the U-curve represented the combined effect of poor clinical conditions (higher APACHE II scores and IE, but lower MAP) and the I/R injuries related to ECMO. Whereas older age, sepsis with subsequent complications, and later initiation of RRT were responsible for the second peak in the curve.
Comparisons with Previous Works- The Role of Temporal Timing of RRT Initiation
A prospective, multicenter, observational study by Bagshaw et al. , which included 1238 patients with equal proportions of surgical and medical admissions, evaluated the relationship between the timing of the RRT initiation in severe AKI and the prognoses. The timing of the RRT was defined by different means including various levels of uremic toxins and the period from the ICU admission to the start of RRT. All patients were categorized into the early (<2 days), delayed (2–5 days), and late (>5 days) RRT groups. Similar to our results, the study found that patients in the late RRT group were older and had a higher proportion of patients affected by sepsis. These patients were also prone to be more azotemic and nonoliguric at RRT initiation. However, unlike our investigation in which only approximately 40% patients were in the EG, the majority (64%) of patients in that study underwent RRT within the first day of the ICU admission. Because the interval from hospital admission to the ICU admission was only one day , the “2-days interval” between hospitalization and the RRT initiation may be suggestive of a relatively late hospital admission. The study also hinted at the different clinical practice patterns applied among the various areas of medicine and/or surgery and the difficulty of recognizing the exact time point of the renal insult in the medical patients.
Another prospective study that evaluated the association between the timing of RRT initiation and the outcomes was performed by Iyem et al. . This study included 185 patients with RRT-requiring postoperative AKI after open-heart surgery. They found no survival difference between early RRT (immediate RRT initiation for AKI, n = 95) and late RRT (RRT start after 48 hours of AKI development, n = 90). However, the study population (all CVS patients) and the initial conditions (no statistical significance between the two groups) in this study were different from our own (only half underwent CVS and higher initial severity of illness in the EG group). CVS is associated with different risks and patient outcomes than those associated with noncardiac surgeries . Additionally, the Iyem et al. study only categorized patients into 2 groups, in which the late RRT group in the temporal sense was equivalent to our IG and LG combined. However, mixing the two groups may generate different results.
More recently, a retrospective, multicenter study  that included 203 patients with post-CVS AKI demonstrated that early RRT (≦3days after CVS) improved the survival benefit when compared to the survival of the late RRT (>3 days) group. Although the patient population and the categorization of the patients differed from our own, the results were consistent in the two studies if we took the EG and IG together as the equivalent of the early RRT group in former study.
Several significant aspects of our study must be addressed. First, because the main hospital where the study took place is a leading teaching hospital of CVS and ECMO, the proportion of the patients with ECMO support was higher than in other studies. Second, the APACHE II scores in current study were significantly lower than those in other studies which are conducted in the ICU predominantly admitting medical patients according to the traditional ICU admission criteria such as sepsis, shock, or respiratory failure. In our study, all were surgical patients, and nearly half underwent elective surgeries. Many of them admitted into ICU for surgery-related complications monitor but not for the critical illness. Besides, the use of ECMO may underestimate the severity scores because some parameters such as blood pressure under ECMO support were pseudo-normalized.  Third, the timing of the RRT initiation should be based on the onset of AKI, which was assumed to coincide with the time point of the surgery. However, this is only true for those undergoing CVS on ECMO support. In the current study, approximately 50% of the patients underwent noncardiac surgeries, resulting in quite a heterogeneous mix of etiologies leading to AKI. Thus, we used the ICU admission date, not the surgery day, as the initial time point. Forth, the study design of categorizing patients by every 2 calendar days (0–1, 2–3, ≧4 days) might not be the best method as it does not take factor in the role of the pathophysiology. Due to the significant factor “septic AKI” plays in the LG, determining a time point to differentiate septic from nonseptic AKI would include the pathophysiology in the analysis. However, the various intervals between the ICU admissions and RRT initiations due to the development of septic AKI (median, 2 days; inter-quartile ranges, 0 to 6 days)  make deciphering the optimal time point difficult.
Several limitations of this study should be addressed. First, as an observational study, it is potentially prone to bias. Second, the study was performed with a limited number of surgical patients predominantly undergoing CVS. Thus the results may not serve as a representative sample of the AKI patients without surgery or with different proportions of surgery types throughout the world. Third, since the term “postoperative AKI” not yet has a universal definition, we use its broad meaning to represent the AKI occurring in the surgical patients. However, we should address that the AKI in our patients (especially in LG) may be resulted from complications occurred later during admission rather than the surgery itself. Forth, the study was designed to enroll only patients with postoperative AKI requiring the RRT. Thus, we cannot compare current data with those whose renal function recovered without the RRT or those who died before the RRT initiation. Fifth, one of the predefined indications for RRT initiation, CVP>12 mmHg maybe not an adequate proxy for fluid overload. To address the issue of fluid overload more clearly, we provided the “net fluid balance” at RRT initiation of the three groups. However, the “net fluid balance” didn’t show its significance in the comparison between survivors and non-survivors as well as in the Cox proportional hazards model for outcome analysis.
Furthermore, multicenter randomized clinical trials based on a dependable compound AKI severity scoring system, which contain some early predictive biomarkers such as urine neutrophils gelatinase-associated lipocalin or even genetic risk factors , are warranted to confirm our findings.
This study found the U-curve association between the timing of RRT initiation after the ICU admission and the outcomes in postoperative AKI. The first peak of mortality rate was related to the poor clinical conditions and the I/R injuries related to ECMO support, and the second peak was associated with older age, sepsis, and later initiation of RRT. Current findings remind physicians to be alert to patients with certain factors that may predict poor prognoses after the RRT initiation.
Comparisons of demographic data and clinical parameters among the three groups (complete data).
The authors thank Dr. Fu-Chang Hu, M.S., Sc.D. and Miss Min-Tzu Huang for their help in statistical analyses.
Conceived and designed the experiments: KDW WJK MJW. Performed the experiments: JWH YMC MSW PRT. Analyzed the data: TSC PCW GHY TWK. Contributed reagents/materials/analysis tools: CFL HBT YFL. Wrote the paper: CCS VCW TMH SLL CTC.
- 1. Kheterpal S, Tremper KK, Englesbe MJ, O’Reilly M, Shanks AM, et al. (2007) Predictors of postoperative acute renal failure after noncardiac surgery in patients with previously normal renal function. Anesthesiology 107: 892–902.
- 2. Lassnigg A, Schmidlin D, Mouhieddine M, Bachmann LM, Druml W, et al. (2004) Minimal changes of serum creatinine predict prognosis in patients after cardiothoracic surgery: a prospective cohort study. J Am Soc Nephrol 15: 1597–1605.
- 3. Abelha FJ, Botelho M, Fernandes V, Barros H (2009) Determinants of postoperative acute kidney injury. Crit Care 13: R79.
- 4. Clark E, Wald R, Walsh M, Bagshaw SM (2011) Timing of initiation of renal replacement therapy for acute kidney injury: a survey of nephrologists and intensivists in Canada. Nephrol Dial Transplant 27: 2761–2767.
- 5. Ng KP, Chanouzas D, Fallouh B, Baharani J (2012) Short and long-term outcome of patients with severe acute kidney injury requiring renal replacement therapy. QJM 105: 33–39.
- 6. Rondon-Berrios H, Palevsky PM (2007) Treatment of acute kidney injury: an update on the management of renal replacement therapy. Curr Opin Nephrol Hypertens 16: 64–70.
- 7. Karvellas CJ, Farhat MR, Sajjad I, Mogensen SS, Leung AA, et al. (2011) A comparison of early versus late initiation of renal replacement therapy in critically ill patients with acute kidney injury: a systemic review and meta-analysis. Critical Care 15: R72.
- 8. Seabra VF, Balk EM, Liangos O, Sosa MA, Cendoroglo M, et al. (2008) Timing of renal replacement therapy initiation in acute renal failure: a meta-analysis. Am J Kidney Dis 52: 272–284.
- 9. Bagshaw SM, Uchino S, Bellomo R, Morimatsu H, Morgera S, et al. (2009) Timing of renal replacement therapy and clinical outcomes in critically ill patients with severe acute kidney injury. J Crit Care 24: 129–140.
- 10. Andrade L, Cleto S, Seguro AC (2007) Door-to-dialysis time and daily hemodialysis in patients with leptospirosis: impact on mortality. Clin J Am Soc Nephrol 2: 739–744.
- 11. Payen D, de Pont AC, Sakr Y, Spies C, Reinhart K, et al. (2008) A positive fluid balance is associated with a worse outcome in patients with acute renal failure. Crit Care 12: R74.
- 12. Ostermann M, Chang RW (2009) Correlation between parameters at initiation of renal replacement therapy and outcome in patients with acute kidney injury. Crit Care 13: R175.
- 13. Bouman CS, Oudemans-Van Straaten HM, Tijssen JG, Zandstra DF, Kesecioglu J (2002) Effects of early high-volume continuous venovenous hemofiltration on survival and recovery of renal function in intensive care patients with acute renal failure: a prospective, randomized trial. Crit Care Med 30: 2205–2211.
- 14. Ostermann M, Chang R (2008) Correlation between the AKI classification and outcome. Crit Care 12: R144.
- 15. Kolhe NV, Stevens PE, Crowe AV, Lipkin GW, Harrison DA (2008) Case mix, outcome and activity for patients with severe acute kidney injury during the first 24 hours after admission to an adult, general critical care unit: application of predictive models from a secondary analysis of the ICNARC Case Mix Programme Database. Crit Care 12 Suppl 1S2.
- 16. Shiao CC, Wu VC, Li WY, Lin YF, Hu FC, et al. (2009) Late initiation of renal replacement therapy is associated with worse outcomes in acute kidney injury after major abdominal surgery. Crit Care 13: R171.
- 17. Lin YF, Ko WJ, Wu VC, Chen YS, Chen YM, et al. (2008) A modified sequential organ failure assessment score to predict hospital mortality of postoperative acute renal failure patients requiring renal replacement therapy. Blood Purif 26: 547–554.
- 18. Wu VC, Ko WJ, Chang HW, Chen YW, Lin YF, et al. (2008) Risk factors of early redialysis after weaning from postoperative acute renal replacement therapy. Intensive Care Med 34: 101–108.
- 19. Ma YC, Zuo L, Chen JH, Luo Q, Yu XQ, et al. (2006) Modified glomerular filtration rate estimating equation for Chinese patients with chronic kidney disease. J Am Soc Nephrol 17: 2937–2944.
- 20. Knaus WA, Draper EA, Wagner DP, Zimmerman JE (1985) APACHE II: a severity of disease classification system. Crit Care Med 13: 818–829.
- 21. Vincent JL, Moreno R, Takala J, Willatts S, De Mendonca A, et al. (1996) The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. Intensive Care Med 22: 707–710.
- 22. Chen YS, Ko WJ, Lin FY, Huang SC, Chou TF, et al. (2001) Preliminary result of an algorithm to select proper ventricular assist devices for high-risk patients with extracorporeal membrane oxygenation support. J Heart Lung Transplant 20: 850–857.
- 23. Hsu CY, Ordonez JD, Chertow GM, Fan D, McCulloch CE, et al. (2008) The risk of acute renal failure in patients with chronic kidney disease. Kidney Int 74: 101–107.
- 24. Lafrance JP, Miller DR (2010) Defining acute kidney injury in database studies: the effects of varying the baseline kidney function assessment period and considering CKD status. Am J Kidney Dis 56: 651–660.
- 25. Wu VC, Ko WJ, Chang HW, Chen YS, Chen YW, et al. (2007) Early renal replacement therapy in patients with postoperative acute liver failure associated with acute renal failure: effect on postoperative outcomes. J Am Coll Surg 205: 266–276.
- 26. Hsu CY, Chertow GM, McCulloch CE, Fan D, Ordonez JD, et al. (2009) Nonrecovery of kidney function and death after acute on chronic renal failure. Clin J Am Soc Nephrol 4: 891–898.
- 27. de Mendonca A, Vincent JL, Suter PM, Moreno R, Dearden NM, et al. (2000) Acute renal failure in the ICU: risk factors and outcome evaluated by the SOFA score. Intensive Care Med 26: 915–921.
- 28. Wu VC, Wang YT, Wang CY, Tsai IJ, Wu KD, et al. (2006) High frequency of linezolid-associated thrombocytopenia and anemia among patients with end-stage renal disease. Clin Infect Dis 42: 66–72.
- 29. Lindenauer PK, Pekow P, Wang K, Mamidi DK, Gutierrez B, et al. (2005) Perioperative beta-blocker therapy and mortality after major noncardiac surgery. N Engl J Med 353: 349–361.
- 30. Uchino S, Bellomo R, Goldsmith D, Bates S, Ronco C (2006) An assessment of the RIFLE criteria for acute renal failure in hospitalized patients. Crit Care Med 34: 1913–1917.
- 31. Eachempati SR, Wang JC, Hydo LJ, Shou J, Barie PS (2007) Acute renal failure in critically ill surgical patients: persistent lethality despite new modes of renal replacement therapy. J Trauma 63: 987–993.
- 32. Minne L, Abu-Hanna A, de Jonge E (2008) Evaluation of SOFA-based models for predicting mortality in the ICU: A systematic review. Crit Care 12: R161.
- 33. Balk RA (2000) Severe sepsis and septic shock. Definitions, epidemiology, and clinical manifestations. Crit Care Clin 16: 179–192.
- 34. Uchino S (2008) Outcome prediction for patients with acute kidney injury. Nephron Clin Pract 109: c217–223.
- 35. Gill N, Nally JV Jr, Fatica RA (2005) Renal failure secondary to acute tubular necrosis: epidemiology, diagnosis, and management. Chest 128: 2847–2863.
- 36. Rosner MH, Okusa MD (2006) Acute kidney injury associated with cardiac surgery. Clin J Am Soc Nephrol 1: 19–32.
- 37. Sadis C, Teske G, Stokman G, Kubjak C, Claessen N, et al. (2007) Nicotine protects kidney from renal ischemia/reperfusion injury through the cholinergic anti-inflammatory pathway. PLoS One 2: e469.
- 38. Bagshaw SM, Uchino S, Bellomo R, Morimatsu H, Morgera S, et al. (2007) Septic acute kidney injury in critically ill patients: clinical characteristics and outcomes. Clin J Am Soc Nephrol 2: 431–439.
- 39. Glance LG, Stone PW, Mukamel DB, Dick AW (2011) Increases in Mortality, Length of Stay, and Cost Associated With Hospital-Acquired Infections in Trauma Patients. Arch Surg.
- 40. Coca SG, King JT Jr, Rosenthal RA, Perkal MF, Parikh CR (2010) The duration of postoperative acute kidney injury is an additional parameter predicting long-term survival in diabetic veterans. Kidney Int 78: 926–933.
- 41. Piccinni P, Dan M, Barbacini S, Carraro R, Lieta E, et al. (2006) Early isovolaemic haemofiltration in oliguric patients with septic shock. Intensive Care Med 32: 80–86.
- 42. Waikar SS, Liu KD, Chertow GM (2008) Diagnosis, epidemiology and outcomes of acute kidney injury. Clin J Am Soc Nephrol 3: 844–861.
- 43. Wu VC, Huang TM, Lai CF, Shiao CC, Lin YF, et al. (2011) Acute-on-chronic kidney injury at hospital discharge is associated with long-term dialysis and mortality. Kidney Int 80: 1222–1230.
- 44. Uchino S, Bellomo R, Kellum JA, Morimatsu H, Morgera S, et al. (2007) Patient and kidney survival by dialysis modality in critically ill patients with acute kidney injury. Int J Artif Organs 30: 281–292.
- 45. Iyem H, Tavli M, Akcicek F, Buket S (2009) Importance of early dialysis for acute renal failure after an open-heart surgery. Hemodial Int 13: 55–61.
- 46. Brienza N, Giglio MT, Marucci M (2010) Preventing acute kidney injury after noncardiac surgery. Curr Opin Crit Care 16: 353–358.
- 47. Garcia-Fernandez N, Perez-Valdivieso JR, Bes-Rastrollo M, Vives M, Lavilla J, et al. (2011) Timing of Renal Replacement Therapy after Cardiac Surgery: A Retrospective Multicenter Spanish Cohort Study. Blood Purif 32: 104–111.
- 48. Tsai CW, Lin YF, Wu VC, Chu TS, Chen YM, et al. (2008) SAPS 3 at dialysis commencement is predictive of hospital mortality in patients supported by extracorporeal membrane oxygenation and acute dialysis. Eur J Cardiothorac Surg 34: 1158–1164.
- 49. Elahi M, Asopa S, Pflueger A, Hakim N, Matata B (2009) Acute kidney injury following cardiac surgery: impact of early versus late haemofiltration on morbidity and mortality. Eur J Cardiothorac Surg 35: 854–863.