Role of Acute Kidney Injury in High-Risk Surgical Patients. A Multicenter Prospective Cohort Study.

Background: Patients who develop postoperative acute kidney injury (AKI) have a poor prognosis, especially when undergoing high-risk surgery. Therefore, the objective of this study was to evaluate the outcome of patients with AKI acquired after noncardiac surgery and the possible risk factors for this complication. Methods: A multicenter, prospective cohort study with patients admitted to intensive care units (ICUs) after noncardiac surgery was conducted to assess whether they developed AKI. The outcomes of patients who developed AKI were then compared to those of patients who did not develop it. Results: A total of 29 ICUs participated, of which 904 involved high-risk surgical patients who were included in the study. The occurrence of AKI in the postoperative period was 15.8%, and the mortality rate of postoperative AKI patients at 28 days was 27.6%. AKI was strongly associated with 28-day mortality (OR = 2.91; 95% CI 1.51-5.62; P= 0.001), and higher length of ICU stay (P< 0.001), length of hospital stay (P<0.001) and length of mechanical ventilator stay (P<0.001). Independent factors for the risk of developing AKI were preoperative anemia (OR=7.0; P= 0.001), SAPS 3 ( OR= 1.04, P<0.001), postoperative vasopressor use (OR= 2.47; P<0.001), postoperative infection (OR= 8.8; P<0.001), need for reoperation (OR= 7.1; P<0.001) and elective surgery (OR= 0.4; P<0.001) was protective for AKI. In addition, patients with higher perioperative uid balance had more chance for AKI. Conclusions: AKI is associated with risk of death. Patients with anemia prior to surgery, who had a higher SAPS 3, needed a postoperative vasopressor or had postoperative infection or needed reoperation were more likely to develop AKI, as were those with a higher perioperative uid balance.


Background
Acute kidney injury (AKI) is characterized by a rapid and signi cant decrease in the glomerular ltration rate (GFR) and is usually of multifactorial origin [1]. Studies that have evaluated perioperative AKI were performed mostly after cardiac and vascular surgery, and there are still major de ciencies in the literature regarding the development of AKI in patients undergoing noncardiac surgery because most recent studies involve retrospective analyses [2,3].
The development of AKI is commonly associated with sepsis, low cardiac output and the postoperative period of major surgeries [4]. One out of every three cases of AKI occurs during the perioperative period [5]; such cases represent approximately 18 to 47% of the cases of hospital-acquired AKI [6].
The incidence of AKI in surgical patients varies according to the type and severity of surgery, with reported rates of 19% after cardiac surgery and approximately 12-13% after general and thoracic surgery [7]. In a large epidemiological study of patients undergoing major noncardiac surgery in intensive care units (ICUs) in Brazil, 30% of patients had postoperative complications, and AKI was the postoperative complication with the second-highest occurrence rate. Postoperative complications are common in highrisk patients after major surgery [8].
Therefore, our hypothesis is that perioperative AKI is common and is associated with a worse prognosis.
Thus, evaluating the incidence and characteristics of patients underwent major noncardiac surgery and developed postoperative AKI, in addition to determining whether perioperative volemic adjustments correlate with AKI, is relevant for adequately improving the therapeutic management of these patients.
The main objective of this study was to evaluate the incidence, impact on outcomes and main risk factors for developing AKI in patients underwent noncardiac surgery and developed AKI after admission to the ICU. Because of the importance of uid balance in surgical patients, we also evaluated the association between this variable and the development of AKI.

Methods
A prospective, multicenter cohort study was conducted between May 1 and November 1, 2018, with a 28day follow-up period. This study was approved by the Research Ethics Committee of the study coordinating center, Hospital Israelita Albert Einstein (CAAE: 55828016.1.1001.0071) and of all participating centers. A signed written informed consent form was obtained from all patients or their respective legal representatives. Two of the participating centers were exempted from providing the form due to the observational nature of the study. This study is a complementary analysis of a multicenter study recently published [8].
Patients aged 18 years or older who were undergoing noncardiac surgery requiring postoperative ICU care were included. Because the criteria for determining the need for postoperative intensive care were not standardized among the centers, all patients with this indication were high-risk.
Patients with terminal cancer, those receiving palliative care and those with severe liver failure (Child C) were excluded because they had lower, or no prospect of cure and their inclusion could lead to inaccurate results. Pregnant women were also excluded. Furthermore, we excluded patients with a length of hospital stay of less than 12 hours because it was not possible to determine whether they had received ICU care or because they were not considered high-risk. Patients who were readmitted to the ICU during the same hospitalization were excluded to avoid including them in the study more than once.
The following variables were evaluated: age; type of surgery; American Society of Anesthesiologists (ASA) classi cation; Simpli ed Acute Physiology Score III score (SAPS 3); Sequential Organ Failure Assessment score (SOFA); previous comorbidities; presence of infection, sepsis or septic shock during ICU stay; creatinine level (mg/dl) at admission and during ICU stay; daily uid balance; mean arterial blood pressure at ICU admission; mechanical ventilation; vasoactive drugs; length of hospital stay and complications. Day 1 was considered the day of surgery plus admission to the ICU before 11:59 pm. AKI was identi ed by the presence of at least one of the following parameters, Disease Improving Global Outcomes (KDIGO) criteria 2012 diagnostic score: increase in SCr by ≥ 0.3 mg/dl (≥ 26.5 µmol/l) within 48 hours; increase in SCr to ≥ 1.5 times baseline, which is known or presumed to have occurred within the prior 7 days; urine volume < 0.5 ml/kg/h for 6 hours or need for renal replacement therapy during ICU stay in patients with no history of chronic kidney failure [9,10]. Creatinine was measured at the entrance and reassessed in the rst 5 days of ICU stay. Oliguria was assessed for 24 hours to classify AKI severity.
The uid balance was calculated as the difference between the infused uids (crystalloids, colloids, uiddiluting drugs, blood derivatives and water via nasogastric tube) and eliminated uids (via diuresis, bleeding, dialysis, and drains).
The other postoperative complications were de ned as follows: cardiovascular, based on the need for vasopressors for more than one hour despite adequate volume resuscitation; respiratory, in the presence of a partial pressure of arterial oxygen to fraction of inspired oxygen ratio (PaO 2 /FiO 2 ) < 200 in patients without previous lung disease, the need for reintubation or the failure to wean from mechanical ventilation during the postoperative period; neurological, based on a sharply uctuating and nonzero Richmond Agitation-Sedation Scale (RASS) score [11], within 24 hours and agitation (determined by a RASS score greater than or equal to + 2); and gastrointestinal, in the presence of acute abdominal distension, uncontrolled nausea or vomiting, or moderate-to high-output stulas.
The main outcome was 28-day mortality after surgery, which was evaluated face-to-face or by telephone. A 28-day follow-up period was chosen to standardize the follow-up time speci cally related to surgery.

Statistical analysis
Considering data from the literature, we assumed a renal complication rate of 15% in high-risk surgical patients [4,7,12]. Therefore, we estimated that at least 1,000 patients would be required to conduct the study with the inclusion of fteen explanatory variables in a robust logistic regression model, with 28-day mortality as the dependent variable.
Categorical variables are presented as absolute and relative frequencies. Quantitative variables are expressed as the mean and standard deviation (SD) or as the median and interquartile range (IQR), when appropriate. We used the Kolmogorov-Smirnov test to evaluate the distribution pattern of continuous numerical variables.
Proportions were compared using the chi-square test or Fisher's exact test, as appropriate. Quantitative variables with multiple measurements were compared using general linear model (GLM) analysis, the consistency of the model was tested using Mauchly's sphericity test, and a post hoc Bonferroni correction was performed in these analyses. The Bonferroni correction set the signi cance cutoff P value; it was used to detect the time points at which the differences were signi cant in multiple comparisons.
The associations between explanatory and response variables were evaluated using xed logistic regression models. Variables that were statistically signi cant in the univariate analyses (p < 0.05) were selected for inclusion in the multiple logistic regression models. Collinearity was rst evaluated by examining the dispersion matrix and the Pearson correlation coe cient for continuous variables or the cross-tabulation for categorical variables. We also evaluated collinearity using the variance in ation factor (VIF). Variables with substantial collinearity (VIF ≥ 10) were excluded from the nal model. The results of the logistic regression analyses are expressed as the odds ratio (OR) and respective 95% con dence interval (95% CI).
All probabilities of signi cance (p-values) were two-tailed. P-values were considered statistically signi cant when p < 0.05. The Statistical Package for Social Sciences v. 26.0 (SPSS Inc.®; Chicago, IL, USA) and R v. 3.4.1 (R Foundation for Statistical Computing, Vienna, Austria) were used to perform the analyses.

Results
A total of 55 ICUs from 55 hospitals were selected for participation in the study (chosen for having accepted to participate in the screening). There were no signi cant differences in operational characteristics among the ICUs when the regions of the country were compared. Twelve of these ICUs (21.8%) were not eligible for participation for different reasons, i.e., 5 ICUs (9.1%) refused to participate because they did not treat the anticipated number of surgical patients, and 9 ICUs (16.4%) returned questionnaires with incomplete study data. Therefore, 29 ICUs participated in the study. During the study period, 25,500 patients underwent noncardiac surgery. Of these, 904 (3.5%; 95% CI 3.3% − 3.8%) were admitted to the ICUs and involved in analyze of the study.
The median (IQR) patient age was 62 (50-72) years, and 53.8% of the patients were male. The median (IQR) SAPS 3 was 42 (32-53) points. Approximately 80.4% of the patients had at least one comorbidity, and hypertension, cancer and smoking were the most frequently occurring.
The postoperative 28-day mortality rate for the entire cohort was 9.6%, and half of these patients (5.2%) presented with AKI. The total incidence of postoperative complications was 29.9%, with 15.8% of renal complications.
When comparing patients with and without AKI, it was identi ed that higher SAPS 3 scores, physical status other than ASA 2, diabetes and anemia, i.e., hemoglobin level prior surgery 10 g/dl, had a signi cant association with AKI (Table 2).
During the intraoperative period, the variables that were signi cantly associated with AKI were type of surgery, intraoperative hypotension without the need for reversal with vasopressors, need for vasopressors and blood transfusion and abdominal and oncological surgery ( Table 2).
In addition, during the postoperative period, AKI was more strongly correlated with a lower hemoglobin level, higher arterial lactate level and mechanical ventilation at ICU admission, in addition to the need for vasopressors during this period, the presence of postoperative infection and the need for unplanned reoperation ( Speci cally, regarding daily uid adjustment during the perioperative period, AKI exhibited an association in the univariate analysis with the median total volume, median uid balance and 0.9% saline volume received every 24 hours. However, when these variables were adjusted for the identi ed risk factors for developing AKI, only a high uid balance was strongly correlated with AKI (Table 3 and Fig. 2).
Regarding the evolution of the daily uid balance in patients with AKI, the GLM showed signi cant differences (P = 0.007) in uid balance from the day of surgery to the fth day between patients with and without AKI. The difference was most evident on the second (P = 0.05) and fth (P = 0.005) postoperative days (Fig. 3).

Discussion
The main ndings of this study were a high incidence of AKI (15.8%) with high mortality in noncardiac surgery patients admitted to the ICU and the association of AKI with infection, reoperation, circulatory shock, preoperative anemia, and positive uid balance.
Major surgeries account can reach 40% of cases of hospital-acquired AKI [13]. Patients with AKI admitted to the ICU during the postoperative period in cardiovascular surgery accounted for 7.7-40% of cases [14], and gastrointestinal tract surgery patients accounted for almost 22% [15].
The presence of AKI during the perioperative period in this subgroup of patients undergoing major surgery was associated with unfavorable outcomes, with a hospital mortality rate of 27.6%. In addition, the length of hospital stay of patients with AKI was higher than that of patients without AKI; patients with AKI spent approximately one extra week in the hospital. Similarly, a large multicenter observational study in European countries reported that mortality among patients with AKI was more than twice that observed in other patients [16]. In a recent study, the occurrence and severity of AKI were strongly associated with risk of death after surgery. However, the relationship between preoperative renal function, as assessed by serum creatinine-based estimated GFR, and the risk of death depended on patient age and whether AKI developed postoperatively [17]. In our study, we identi ed a high mean age (over 60 years), but we did not analyze small existing changes in renal function that could indicate a low physiological reserve and thus predispose the patient to a higher risk of AKI.
In our study, patients who developed AKI were more severe (higher SAPS 3), had a greater need for advanced support with catecholamines and frequently had infections. The systemic in ammatory response that occurs in patients with infection is correlated with the induction of AKI. Although the pathophysiology of sepsis-induced renal disorders is still poorly understood, experimental models suggest that, at least in the initial phase of the disease, functional changes associated with microvascular shunting and tubular cell injury are part of the picture [18]. In addition, the intensity of the in ammatory response to surgical trauma determines both tissue hypoperfusion and the ischemiareperfusion response, which intensi es the in ammatory response. Furthermore, mechanical ventilation, intra-abdominal hypertension, pain and surgical stress are factors that increase the release of antidiuretic hormones and contribute to the development of AKI. Prospective studies have reported a relationship between AKI and sepsis and the need for mechanical ventilation [6,19,20].
There is a close correlation between hypotension and the development of AKI. A recent meta-analysis showed that goal-directed hemodynamic adjustments reduce the likelihood of AKI in orthopedic and abdominal surgery patients [21]. We observed the most frequent use of vasopressors in this group of patients. The duration of intraoperative hypotension (particularly relative to the patient's normal arterial blood pressure), including brief episodes of mean arterial pressure lower than 55 mmHg, has been associated with kidney injury [5]. Careful maintenance of cardiovascular stability, including uid infusion, throughout this period is vital for protecting renal perfusion while avoiding volume overload. A study of high-risk surgical patients showed that the customized maintenance of blood pressure according to the patient's previous levels during surgery was able to reduce organ dysfunction [22]. Moreover, the association between the use of vasoactive drugs and mortality is consistent with the results of other studies [23][24][25].
Interestingly, a correlation was noted between preoperative anemia (such as hemoglobin concentration less than 10 g/dl) and impaired renal function. This pathophysiology is multifactorial, but mainly involves a decrease in the oxygen transport capacity and subsequent tissue hypoxia [26].
Besides, patients who developed AKI had a more positive mean daily uid balance. In fact, reduced ultra ltration is rarely the only cause of AKI; in ammation, direct vascular injury or tubular obstruction usually accompany kidney injury during the perioperative period. The subsequent rupture of intercellular junctions leads to the release of cells into the tubular lumen and to polarity loss of the transmembrane ion channel. It is expected that, in many cases, simply restoring the circulating volume does not improve the results and may be counterproductive [4,5]. Elevated intratubular pressure decreases glomerular ltration and activation of tubuloglomerular feedback, with consequent preglomerular vasoconstriction, which leads to an additional reduction in glomerular ltration. Studies have shown that excess uid is an independent factor for the development of AKI and that in patients with AKI, a more positive uid balance was correlated with higher mortality [27]. In general, a positive uid balance, if necessary, can be applied with caution within the rst 36 hours. After this, a more conservative approach to volume is the most appropriate and safe method. Regarding the type of solution used, epidemiologic data suggest that 0.9% saline solution, when compared with balanced salt solutions such as balanced solutions, may increase the risk of AKI, the need for renal replacement therapy and mortality in ICU patients. In addition, there is evidence of harm (increased rates of AKI) with the use of hetastarch solutions, which should generally be avoided [28]. In our study, we did not detect a negative in uence of any type of colloid on the development of AKI.
Major advances associated with a decrease in postoperative mortality have been observed worldwide in recent years, but the complication rates remain high [29]. AKI is associated with signi cant morbidity and mortality after major surgery. Patients should be strati ed according to their risk of developing AKI based on their exposure and susceptibility, including such factors as severity, the presence of sepsis, mechanical ventilation use and vasoactive drug use, among others [30, 31], and preemptive measures should be taken [5]. We must remember that there are well-established interventions in the KDIGO guidelines that, when used, can reduce the impact of AKI in high-risk surgical patients [32].

Study limitations
The main strengths of our study are associated with its multicenter nature, as it included ICUs located in several regions of Brazil. However, there was a reasonable rate of refusal to participate in the study that affected external validity to some degree. In addition, the adjusted regression model could be performed with a priori de ned AKI prediction variables, and no based on statistically important ndings from the unadjusted analysis.
There were failures in capturing some relevant data that could have been included in the analyses.
Additionally, the need to obtain informed consent in epidemiological studies such as this tends to skew the sample due to the nonconsent of more critical patients, whose families may be psychologically fragile. Another aspect to be considered is the lack of standardization among the centers regarding indications for postoperative intensive care. In addition, this study was not able to assess long-term complications and mortality, and some complications may have occurred after the study period.

Conclusion
AKI is a major complication in high-risk surgical patients and is associated with risk of death. Patients with anemia prior to surgery, those with a higher SAPS 3, those requiring vasopressors during the postoperative period and those with postoperative infection or the need for reoperation are the most likely to develop AKI, as are patients with a higher perioperative uid balance. Fluid type did not in uence AKI development.

Declarations
Ethics approval and consent to participate This study was approved by the Research Ethics Committee of the study's coordinating center, the Hospital Israelita Albert Einstein (CAAE: 55828016.1.1001.0071). And written informed consent was obtained from all patients.

Consent for publication
Written informed consent was obtained from all patients or their legal representatives.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
This research did not receive any speci c grant from funding agencies in the public, commercial, or notfor-pro t sectors.    Plot of AKI probability in relation to uid balance adjusted to SAPS 3 ( rst) and Youden diagram (second) showing more patients with AKI when SAPS 3 and uid balance were higher.