Acute kidney injury (AKI) is a prevalent and prognostically important complication of cardiac surgery, occurring in up to 30% of patients and substantially increasing short- and long-term morbidity and mortality.1 – 5 As there are currently no clinically proven therapies that can prevent or treat AKI in cardiac surgery,1 the best means for reducing the burden of AKI may be through risk factor modification.

Several risk factors for AKI have been identified.6 While most are not modifiable, anemia and red blood cell (RBC) transfusion have been identified as potentially modifiable risk factors.2 , 6 They may therefore be ideal targets for reducing the burden of AKI in cardiac surgery. Importantly, these risk factors are interrelated, and while several studies have explored their individual relationship with AKI, the effect of their interrelationship with AKI has not been clearly defined. Existing evidence suggests that this effect may be synergistic, such that anemic patients may be more susceptible to the deleterious effects of RBC transfusions on the kidney than non-anemic patients.7 – 9 This interrelationship, if proven valid, could have important clinical implications for risk management and patient optimization.10 , 11 The objective of this study was to explore the interrelationship of preoperative anemia, intraoperative anemia, and RBC transfusion on the day of surgery with AKI in cardiac surgery.

Methods

Study setting, patient sample, and data collection

Seventeen university-affiliated hospitals in Canada, United States, Australia, and New Zealand participated in this retrospective observational study. After obtaining institutional Research Ethics Board approval at the University Health Network (REB 11-0663-AE; initial approval Sept 2011) and at all participating centres, investigators at each hospital retrospectively collected data on 100 consecutive adult (> 18 yr) patients who underwent cardiac surgery with cardiopulmonary bypass (CPB) from January 1, 2010 to December 31, 2011. Cases involving heart transplantation, ventricular assist device placement, or repair of complex congenital abnormalities were excluded because these procedures were not performed at all participating hospitals. Patients were also excluded if they were dialysis-dependent before surgery, had no creatinine or hemoglobin measures before surgery, had no record of timing of perioperative transfusions, or had no creatinine measures after surgery. Based on previous experience,2 we anticipated that the number of patients obtained from the participating hospitals would include sufficient cases of AKI to allow us to identify the relationships of interest by multivariable analysis.

Using standardized case report forms, detailed perioperative data (including demographics, laboratory tests, medications used, nature of surgery, blood product transfusions on the day of or day after surgery, and major in-hospital postoperative complications) were collected from existing clinical databases and hospital charts. Data were entered into a computerized database with validity checks. All queries were resolved by referring to the patients’ original records.

Primary independent variables

The primary variables of interest were preoperative anemia, defined as a hemoglobin concentration < 130 g·L−1 in males and < 120 g·L−1 in females,12 intraoperative anemia, defined as hemoglobin < 80 g·L−1 during CPB,8 and RBC transfusion on the day of surgery.

Dependent variable

Patients with a > 50% increase in serum creatinine at any time during the first week after surgery were categorized as having had AKI. This creatinine threshold is consistent with current definitions of AKI.13

Statistical analyses

SAS™ version 9.3 (SAS Institute, Inc., Cary, NC, USA) was used for the statistical analyses. Categorical variables were summarized as frequency (percentage) and continuous variables as median [interquartile range] unless otherwise stated. Variability in the incidences of anemia and RBC transfusions across sites was assessed by comparing a logistic regression model, which included each site as a factor, with the null model (intercept only) using a likelihood ratio test.

The unadjusted relationship of preoperative anemia, intraoperative anemia, and RBC transfusion on the day of surgery with AKI was assessed by univariate Poisson regression with robust error variance.14 Multivariable Poisson regression was used for determining the independent relationships of these three variables as well as their two-way and three-way interaction terms with AKI, after accounting for overdispersion15 and controlling for the underlying risk of AKI using the Cleveland risk score (Table 1).16 , 17 Site was also included as a categorical variable to account for unmeasured case-mix differences. Interaction terms that were not statistically significant (P > 0.2) were removed and the model was reconstructed. Seventeen of the 1,444 patients in the study had missing variables and were excluded from these analyses, leaving 1,427 patients. The effect of clustering (within-site correlation) was evaluated by constructing the model with generalized estimating equations.18

Table 1 Cleveland risk score16 for acute kidney injury
Full size table

Bootstrap resampling was used for internal validation. One-thousand computer-generated samples, each including 1,427 patients, were derived from the study cohort by random selection with replacement, and the model was refitted for each sample. The mean and standard deviation bootstrap parameter estimates were used to assess the results of the multivariable modelling.

Results

Data from one site (n = 100) could not be used because the timing of the RBC transfusion was not recorded. An additional 139 patients met one or more of the exclusion criteria and were excluded from the analyses, leaving 1,444 patients in the study. Of these, 541 (37%) had preoperative anemia, 501 (35%) developed intraoperative anemia (13 had missing data), 619 (43%) received RBC transfusions on the day of surgery, and 238 (16%) developed AKI (37 of whom required dialysis). There was significant (P < 0.001) site variability in the incidences of all four variables (Figure). Sample characteristics are presented in Table 2. In univariable analysis, preoperative anemia, intraoperative anemia, and RBC transfusion on the day of surgery were all significantly associated with AKI, with their respective relative risks (95% confidence interval [CI]) being 1.9 (95% CI 1.5 to 2.3), 2.1 (95% CI 1.7 to 2.6), and 2.0 (95% CI 1.7 to 2.5).

Figure
figure 1

Rates of acute kidney injury, preoperative anemia, intraoperative anemia, and red blood cell (RBC) transfusions on the day of surgery at the participating sites

Full size image
Table 2 Patient demographics and perioperative variables in patients with and without acute kidney injury
Full size table

In multivariable analysis, of the interaction terms analyzed, only the interaction of preoperative anemia with RBC transfusion was associated with AKI and hence was retained in the final model (Table 3); the remaining interaction terms had P values > 0.2. With the inclusion of this interaction term, neither preoperative anemia nor RBC transfusion was independently associated with AKI (Tables 3 and 4). In patients with preoperative anemia, however, RBC transfusion was associated with a 1.9-fold (95% CI 1.4 to 2.5) increase in the relative risk of AKI (Table 4). Intraoperative anemia was independently associated with a 1.4-fold (95% CI 1.1 to 1.7) increase in the relative risk of AKI. An individual with the combination of preoperative anemia, intraoperative anemia, and RBC transfusion had a 2.6-fold (95% CI 2.0 to 3.3) increase in the relative risk of AKI over an individual with none of these risk factors (Table 4). The results of bootstrap resampling were consistent with the regression modelling (Table 3).

Table 3 The results of multivariable modelling for the three variables of interest and significant interaction terms
Full size table
Table 4 Risk-adjusted relative risks (RR) and 95% confidence intervals (CI) of various combinations of preoperative anemia, intraoperative anemia, and red blood cell (RBC) transfusion on the day of surgery for acute kidney injury (AKI)
Full size table

Discussion

In this retrospective cohort study that included 1,444 patients who underwent cardiac surgery at 16 hospitals, we explored the relationship between preoperative anemia, intraoperative anemia, and RBC transfusion on the day of surgery. We found that more than one-third of patients had one or more of these risk factors and that 16% of patients developed AKI. After accounting for the underlying risk of AKI and the influence of interactions amongst the risk factors, we found that an individual with the combination of preoperative anemia, intraoperative anemia, and RBC transfusion had a 2.6-fold (95% CI 2.0 to 3.3) increase in the relative risk of AKI over an individual with none of these risk factors (Table 4).

Anemia and RBC transfusion have been observed in multiple previous studies as important risk factors for AKI after cardiac surgery, but the interrelationship between these variables and AKI has not been as extensively reported.19 , 20 In a single-centre observational study comprised of 2,113 propensity-score matched pairs of anemic and non-anemic patients who underwent cardiac surgery with CPB and received up to three units of perioperative RBC transfusions, the risk of AKI was increased in direct proportion to the number of transfusions in both groups, but the increase was much more pronounced in anemic patients.7 In non-anemic patients, the AKI rate increased from 1.7% in non-transfused patients to 3.2% in those who received three units of RBCs (P = 0.1); however, in anemic patients, the rate increased from 1.8% to 6.6% (P < 0.0001).7 In another single-centre observational study by Loor et al., which included 9,144 patients who underwent cardiac surgery with CPB, intraoperative anemia, defined as a nadir hematocrit < 25%, was independently associated with AKI but an RBC transfusion alone was not. As in our study, their reported risk of AKI was highest when both anemia and transfusion were present.8 The influence of preoperative anemia, however, was not clearly described in their study. Similar findings were reported by Ranucci et al. in their study of 1,766 adult patients who underwent isolated coronary artery bypass graft surgery.9 Overall, therefore, our findings are consistent with the existing single-centre studies described above.

The mechanisms by which perioperative anemia and RBC transfusions may cause AKI in cardiac surgery have not been elucidated. Recent proteomic studies indicate that all patients undergoing cardiac surgery with CPB develop the early stages of ischemia-reperfusion kidney injury, but whether they go on to develop AKI depends on both the occurrence of other renal insults as well as the severity of the ensuing inflammatory response, renal hypoxia, and oxidative stress.21 Anemia and RBC transfusion could cause AKI either by harming the kidney directly or by increasing patients’ susceptibility to concomitant renal insults.

During storage, RBCs undergo several changes that may harm the kidney after transfusion. These changes include a decrease in 2,3-diphosphoglycerate, adenosine triphosphate, and S-nitrosohemoglobin as well as an increase in the concentrations of lactate, potassium, cytokines, iron, and free hemoglobin in the supernatant.22 – 26 Red blood cells also become progressively less deformable and more fragile during storage in a time-dependent manner. This results in the accumulation of hemoglobin-laden microvesicles in the supernatant as well as predisposing up to 25% of RBCs to early hemolysis within one hour after transfusion.27 – 29 Cumulatively, these changes may result in post-transfusion impairment of tissue oxygen delivery and exacerbation of inflammatory response and oxidative stress, thereby harming the kidney.19 In line with this hypothesis, some (but not all) retrospective studies have found an association between age of blood and adverse outcomes.30 , 31

The contributory effects of anemia to AKI are likely also multifactorial. First, anemic patients (both preoperative and intraoperative) have lower hemoglobin concentrations throughout the perioperative period than non-anemic patients,32 predisposing them to renal hypoxia.20 Second, many anemic patients have pre-existing subclinical kidney disease that may increase renal tubular oxygen consumption and oxidative stress,33 – 35 thus increasing their susceptibility to concurrent renal insults. Finally, anemic patients have abnormal iron metabolism36 – 38 that may affect the clearance of the large amount of iron released when RBCs are hemolyzed, either during storage or soon after transfusion, and this may potentially lead to the presence of free iron and hemoglobin in the circulation.39 , 40

Our study has several limitations. Since neither the cause nor the duration of preoperative anemia were known, it is possible that diseases associated with both anemia and AKI may account for some of the observed relationships. It is also possible that the observed relationships were unduly influenced by other unmeasured confounders such as colloid use.41 Another limitation is that the number of patients in various categories of anemia and transfusion are small (Table 4), resulting in relatively wide confidence intervals for some of the coefficients (Table 3). Finally, we could not elucidate the cause of anemia, transfusions, or AKI, and we did not have data on the age of the transfused RBCs.31 Furthermore, we did not use a pre-specified protocol to guide transfusion therapy across all participating sites, and intraoperative management of blood salvage, hemodilution, and the conduct of CPB was left to the individual centres. On the other hand, our study has a number of strengths. Consecutive patients underwent surgery at multiple hospitals; the data used in our study were collected by blinded data abstractors, and the interrelationships between the variables of interest were carefully explored. Nevertheless, further studies are required to confirm or refute our findings.

The natural clinical implication of our findings is that correcting preoperative anemia and avoiding intraoperative anemia and RBC transfusion on the day of surgery could potentially reduce the burden of AKI after cardiac surgery. One relatively simple strategy is to reduce perioperative hemodilution by minimizing fluid administration and using retrograde autologous priming of the cardiopulmonary circuit.42 Transfusion practice bundles that incorporate point-of-care coagulation testing may also achieve these objectives by reducing blood loss and transfusions through better management of coagulopathy.43 , 44 Erythropoietin stimulating agents may also be used to correct preoperative anemia, thereby avoiding intraoperative anemia and RBC transfusion,42 but the risk-benefit of this intervention in cardiac surgery has not been elucidated.45

Other potential options that are currently undergoing evaluation include optimizing oxygen delivery by modifying pump flow;46 washing of blood to remove the pro-inflammatory molecules, free hemoglobin, and iron that accumulate in the supernatant during storage;47 haptoglobin therapy to scavenge the free hemoglobin that can be present after CPB and blood transfusion;48 and prophylactic RBC transfusion one to two days before surgery in patients with preoperative anemia.49 This latter approach has been postulated to “reduce the risk of AKI by reducing the severity of anemia, reducing the need for RBC transfusions, allowing time for the transfused blood to recover from the deleterious changes that they undergo during storage, and allowing time for the kidneys to recuperate from the harmful effects of transfused blood before they are exposed to other renal insults” during surgery.49 , 50 The risk-benefit profiles of these investigational interventions are yet to be determined.

In conclusion, the results of this multicentre retrospective study showed that preoperative anemia, intraoperative anemia, and RBC transfusion on the day of surgery are interrelated risk factors for AKI after cardiac surgery. Our findings suggest that preventing or correcting these risk factors may therefore reduce the burden of AKI in this setting.