|Home | About | Journals | Submit | Contact Us | Français|
Acute kidney injury (AKI) is a risk factor for long-term adverse outcomes including acute myocardial infarction (MI) and death. However, the relationship between severity of AKI and in-hospital outcomes in the setting of acute MI has not been well-documented.
The study population (n=59,970) was drawn from the ACTION Registry®-GWTG(tm), a nation-wide sample of MI patients admitted to 383 hospitals in the United States between July 2008 and September, 2009. AKI was defined using absolute changes in serum creatinine (SCr; peak SCr – admission SCr), and categorized as no AKI (SCr change<0.3 mg/dl), mild AKI (SCr change 0.3-<0.5 mg/dl), moderate AKI (SCr change 0.5-<1.0 mg/dl), and severe AKI (SCr change ≥1.0 mg/dl). Overall, 16.1% had AKI, including 6.5% with mild AKI, 5.6% with moderate AKI, and 4.0% with severe AKI. In-hospital mortality rates for those with mild, moderate and severe AKI were 6.6%, 14.2%, and 31.8% compared to 2.1% without AKI. The odds ratio for in-hospital death were 2.4 (95% CI 2.0-2.7), 4.5 (95% CI 3.9-5.1), and 12.6 (95% CI 11.1-14.3) for mild, moderate, and severe AKI compared to those without AKI. Although patients with AKI were less likely to undergo early invasive care or to receive antiplatelet therapies, rates of major bleeding ranged from 8.4% (no AKI) to 32.7% (severe AKI).
AKI is common and associated with mortality and bleeding, underscoring the importance of efforts to identify risk factors and prevent AKI in AMI care.
Acute kidney injury (AKI, previously known as acute renal failure) is a frequent complication that affects a substantial number of hospitalized patients annually. While exact numbers are not known, AKI occurs in an estimated 5% of all hospitalizations, and associated health care expenditures are more than 10 billion dollars annually.1;2 AKI during hospitalization also predicts future risk for end-stage renal disease3;4 and all-cause mortality.4;5
Chronic kidney disease is a strong predictor of in-hospital mortality in the acute myocardial infarction setting. We have recently shown that up to 40% of the AMI population has CKD, and this confers up to an 8-fold increased risk of mortality among affected patients.6 Similarly, in the post-myocardial infarction setting, AKI confers an independent and graded association with long-term mortality.2;5;7;8 Even small increases in serum creatinine, as low as 0.1 mg/dl, are associated with an increased risk of end-stage renal disease and all-cause mortality in this setting.9 AKI is common after cardiac surgery, and is associated with an increased risk of death.10
While the long-term complications of AKI have been examined, a comprehensive assessment of the early risk of AKI in the post-MI setting has not been carried out. Thus, the purpose of this analysis was to evaluate the prevalence of AKI among patients presenting with STEMI and NSTEMI, the relationship with treatments including catheterization and revascularization with AKI, and to examine the short-term outcomes associated with AKI.
Study patients were part of the National Cardiovascular Data Registry (NCDR) Acute Coronary Treatment and Intervention Outcomes Network (ACTION) Registry®–Get With The Guidelines (GWTG™) which is a nationally-representative, quality improvement registry of STEMI and NSTEMI patients. Enrollment began January 1, 2007. Peak creatinine data was added to the data collection form in August, 2008, but hospitals were allowed to retrospectively enter data. Thus, data for the present analysis includes patients from the July 1, 2008 to September 30, 2009 at 306 ACTION Hospitals. ACTION eligibility has been described previously;6 briefly, patients were eligible if they presented with an ischemic syndrome within 24 hours, and a primary diagnosis of myocardial infarction (STEMI or NSTEMI). For the present analysis, we excluded patients on dialysis or with missing dialysis status (n=1,729), or those with missing initial or peak serum creatinine (n=7,844). Of the 7,844 individuals with “missing” initial or peak creatinine values, 388 had missing values and 7456 patients did not have initial or peak creatinine values collected. Patients in whom serum creatinine was not collected (n=7456) were more likely to die within 24-hours of hospitalization compared to those with either missing or collected serum creatinine. Overall, the final sample size was 59,970. Each participating center abides by local ethical or institutional review.
Trained data collectors extracted data from existing medical records using an electronic case record form. All data elements can be found at www.ncdr.com. The NCDR Data Quality Program monitors data quality and completeness.
AKI was assessed based on the difference between the admission and peak serum creatinine measurements using the modified Acute Kidney Injury Network (AKIN) definition.11 We used serum creatinine in the model because AKI is a time of flux and eGFR is not reliable when a patient is not in steady-state. In addition, serum creatinine is also more analogous to the AKI categories as compared to the derived measurements, justifying why we used it in place of patients with and without CKD (where CKD is defined as eGFR<60 ml/min/1.73m2 using the MDRD equation). We defined no AKI as a change in creatinine <0.3 mg/dl; mild AKI as a change in creatinine of 0.3-<0.5 mg/dl, moderate AKI as change in serum creatinine 0.5-<1.0 mg/dl, and severe AKI as a change in serum creatinine ≥ 1.0 mg/dl. Furthermore, we evaluated whether more subtle creatinine changes within the “no AKI category” were associated with in-hospital mortality (<0.1, 0.1-<0.2, 0.2-<0.3 mg/dl). We also defined AKI based on percent changes in serum creatinine: <25%, 25-<50%, 50-<75%, and ≥75% as a sensitivity analysis.
The primary outcome of interest was in-hospital all-cause death. We also examined major bleeding, as defined in ACTION-GTWG as an absolute drop in hemoglobin of 4 g/dL, intra-cranial hemorrhage, documented or suspected retroperitoneal bleed, any red blood cell transfusion with a baseline hemoglobin of at least 9 g/dL, or any red cell transfusion with a hemoglobin <9 g/dL and a suspected bleed. Bleeding events were considered only if they occurred before coronary artery bypass grafting (CABG).
We stratified our primary analysis by CKD group calculated based upon the admission creatinine; groups were stratified based on CKD stage as previously defined.6 Using the MDRD equation, estimated glomerular filtration rate (eGFR) was calculated, and CKD was defined as eGFR<60 ml/min/1.73m2. We also stratified our analysis by in-hospital CABG, and in-hospital cardiogenic shock due to known high rates of adverse renal outcomes in patients with these characteristics. Cardiogenic shock was defined as a sustained (>30 minutes) episode of systolic blood pressure <90 mm Hg, and/or cardiac index <2.2 L/min/m2 determined to be secondary to cardiac dysfunction, and/or the requirement for parenteral inotropic or vasopressor agents or mechanical support (e.g., IABP, extracorporeal circulation, ventricular assist devices) to maintain blood pressure and cardiac index above those specified levels.
Patient baseline characteristics, treatments and interventions, and in-hospital outcomes were summarized by AKI category. Ordinal categorical and continuous variables were compared with the Kruskal Wallis test and categorical variables were compared with the chisquare test.
To evaluate the association between in-hospital mortality and AKI categories, we used logistic generalized estimating equations (GEE)12 with an exchangeable working correlation matrix to account for within-hospital clustering. Multivariable models included adjustment for age, prior peripheral arterial disease, systolic blood pressure (SBP) on presentation, heart rate on presentation, heart failure (HF) or shock on admission (HF only, shock only or HF with shock, vs. none), electrocardiographic findings (STEMI, ST-segment depression or transient ST-segment elevation vs. no ST-segment changes), initial troponin ratio, and initial serum creatinine. ST-segment changes included ST depression or transient ST elevations and no ST-segment changes included T-wave inversions and no electrocardiogram changes.13 Furthermore, we examined association between in-hospital morality and AKI categories across different subgroups. First, we stratified by patients with or without CKD calculated based upon the admission creatinine. Using the MDRD equation, estimated glomerular filtration rate (eGFR) was calculated, and CKD was defined as eGFR<60 ml/min/1.73m2. In addition, we stratified by in-hospital CABG, and in-hospital cardiogenic shock due to their known risks to the kidney.
To explore the relationship between in-hospital major bleeding and AKI categories, the logistic GEE method was used adjusting for covariates in the ACTION in-hospital major bleeding model, which included female sex, age, diabetes, prior peripheral arterial disease, body weight, home warfarin therapy, heart rate on presentation, SBP on presentation (SBP≤130 mmHg, SBP 130-160 mmHg and SBP≥160 mmHg), HF on presentation (HF only, shock only or HF with shock, vs. none), electrocardiographic findings (STEMI, ST-segment changes vs. no ST segment change), initial serum creatinine and initial Hb (Hb<12 g/dL vs. Hb≥12 g/dL). 14
Finally, we examined factors associated with AKI (modeled as mild or worse AKI compared to no AKI). In-hospital catheterization, CABG, and cardiogenic shock were included in the model if occurring before the documentation of peak serum creatinine levels. Continuous variables were tested for linearity and plotted against rates of mild or severe AKI to create dichotomous cutoff points where applicable. Cutoff points were considered where the relationship between the variable and mild or severe AKI became flat or nonlinear, and finalized once determined to be clinically appropriate. For heart rate, there was a J-shaped relationship with unadjusted mild or severe AKI, and therefore we analyzed it as a continuous variable by allowing for 2 slopes via fitting a linear spline with a knot at 80 beat/min. For interpretation, we provided per 10 beat/min increase or decrease instead of 1 beat/min increase or decrease. Similar logic was applied to other continuous variables (e.g., baseline hemoglobin, peak troponin, SBP, and weight).
We stratified our cohort by categories of the ACTION mortality model,13 Briefly, the ACTION mortality model is a mortality risk score developed in the ACTION registry to predict death in the setting of AMI. Risk factors include age, baseline serum creatinine, systolic blood pressure at presentation, baseline troponin elevation, heart failure and/or cardiogenic shock at presentation, ST-segment changes, heart rate at presentation, and prior peripheral arterial disease. We stratified our sample into 4 groups based on predicted baseline risk (<=30%, 31-40%, 41-50%, >50%) and then further stratified our data by AKI category. Data presented are crude rates of in-hospital mortality.
Data are presented as adjusted associations with odds ratios (OR) (95% confidence interval [CI]). We considered P value <0.05 as statistically significant. Analyses were performed with SAS software (version 9.2, SAS Institute, Cary, North Carolina).
The ACTION Registry®–Get With The Guidelines (GWTG™) is administered by the American College of Cardiology Foundation (ACCF) and is sponsored by Bristol-Myers Squibb/Sanofi Pharmaceuticals Partnership, Genentech, and Schering-Plough Corporation. The sponsors had no additional role in this project including the selection of topic, analysis of data, decision to publish, or approval of the manuscript prior to publication. The statistical analysis was performed by an independent statistician (AYC) from the Duke Clinical Research Institute.
Patients who developed AKI were older and more often male (Table 1). Admission eGFR was lower among participants who developed AKI compared to those who did not. In addition, CKD, hypertension, diabetes, and prior MI were more common among those who developed AKI. Figure 1 presents the prevalence of AKI by admission CKD status. Finally, patients who developed AKI had more in-hospital complications, with higher rates of CHF, shock, and higher levels of cardiac troponins (Table 2).
Overall, 16.1% (n=9,659) of patients developed AKI: 6.5% (n=3,919) with mild AKI, 5.6% (n=3,363) with moderate AKI, and 4.0% (n=2,377) with severe AKI. AKI prevalence by admission CKD status is presented in Figure 1. Factors associated with AKI in the acute MI setting are presented in Table 3; as expected, in-hospital CABG (OR 3.8), in-hospital shock (OR 4.0), in-hospital catheterization (OR 4.0), CHF on admission (OR 2.4), and presenting serum creatinine (OR 1.5 per 1 mg/dL increase) had strong associations with AKI. Most standard CVD risk factors were also risk factors for AKI, and several factors known to be associated with mortality in the AMI setting were also important AKI risk factors.15-17
Overall, there were 2482 deaths. Mortality increased with worsening AKI category (Table 2, Figure 2); those without AKI had a 2.1% mortality rate, mild AKI was associated with a 6.6% mortality rate, moderate AKI with a 14.2% mortality rate, and severe AKI with a 31.8% mortality rate.
In the overall sample, the adjusted odds ratio of death increased sharply with worsening AKI category, ranging from an OR of 2.4 (95% CI 2.0-2.7) amongst those with mild AKI, 4.5 (95% CI 3.9-5.1) amongst those with moderate AKI, and 12.6 (95% CI 11.1-14.3) amongst those with severe AKI, compared to the referent group of no AKI (Figure 2A). Results were slightly attenuated but still robust after excluding individuals with in-hospital shock. Similar trends were observed among those with and without CKD at presentation (Figure 2B), although the odds ratio for death was substantially higher amongst those without baseline CKD (OR 28.0, 95% CI 22.7-34.5), likely due to the lower rate of death in this sub-group among patients without AKI (p-value for interaction between CKD and AKI <0.0001). Finally, we observed similar increases in rates of death by AKI category in those patients who underwent in-hospital CABG as compared to those who did not (Figure 2C).
We examined the rates of in-hospital mortality by AKI category stratified by the ACTION mortality risk score13 (Figure 3); within each category of baseline mortality risk, in-hospital mortality rose sharply by AKI category.
In sensitivity analyses, we additionally adjusted our multivariable-adjusted model for in-hospital CABG, in-hospital catheterization, or in-hospital shock; results for those with mild, moderate, and severe AKI (relative to no AKI) were 2.11 (95% CI 1.82-2.44), 3.54 (95% CI 3.07-4.09), and 9.64 (95% CI 8.44-11.01). We observed a graded mortality risk by AKI category among those with serum creatinine changes as low as 0.1-<0.2 mg/dl (<0.1 mg/dl: 1.9%, 0.1-<0.2 mg/dl: 2.9%, 0.2-<0.3 mg/dl: 3.6%). When we secondarily defined AKI by percent change in serum creatinine, mortality rates were 2.5%, 7.0%, 13.6%, and 27.9% among those with <25%, 25-<50%, 50-<75%, and ≥75% change in serum creatinine, respectively. In multivariable analyses, ORs for death were 2.5 (25-<50%), 4.7 (50-<75%), and 11.9 (≥75%) compared to the referent group of no AKI. Both unadjusted and adjusted results were similar to the primary analyses.
Overall, 6184 patients experienced a major bleeding event. The characteristics of patients with and without an in-hospital major bleeding event are shown in Supplemental Table 1. Despite less frequent invasive therapy and less use of oral and intravenous antiplatelet agents (Table 1), patients with AKI were more likely to experience an in-hospital major bleeding event, ranging from 8.4% amongst those without AKI, to 32.7% amongst those with severe AKI (Table 2). After multivariable adjustment, the odds ratio for a major bleeding event by AKI category for mild AKI was 1.7 (95% CI 1.5-1.8), for moderate AKI: OR 2.1 (95% CI 1.9-2.3), and 3.1 (95% CI 2.8-3.4) for severe AKI compared to the referent group of no AKI. Sensitivity analyses stratified by in-hospital catheterization or by the number of anti-platelet medications treated with demonstrated consistency (Supplemental Table 2).
Overall, the prevalence of AKI in the setting of MI is 16%. Patients with AKI had a substantially higher risk of mortality than those without AKI. These observations persisted amongst those who did not experience in-hospital shock, in-hospital CABG, or have pre-existing CKD at baseline, all risk factors for AKI. Finally, patients with AKI are at a substantially higher risk of in-hospital major bleeding events, despite lower rates of invasive strategies.
The prevalence of AKI in our study was 16%, a rate that is comparable to what has been previously reported in diverse clinical settings.1;8;18;19 Following cardiac surgery, AKI may be as high as 24%,10 with up to 1.1% of patients requiring dialysis.20 When AMI is complicated by cardiogenic shock, AKI may affect more than half of all patients.21
Our study adds to the literature by extending what is known about the rates of in-hospital mortality in association with AKI in the setting of AMI. Scant literature exists that has systematically examined the association between AKI and mortality in the short-term setting. In a single-center study of 97 STEMI patients who required primary percutaneous coronary intervention and intra-aortic balloon pump, the AKI-associated mortality rate was 50%, with an in-hospital mortality relative risk of 12.3.21 Similarly, among 1038 patients who presented with STEMI at a medical center in Israel, worsening renal function (defined as an increase in serum creatinine of at least 0.5 mg/dl) was associated with an 11.4-fold increased risk of in-hospital mortality.8 Our results complement these findings in a substantially larger sample size using contemporary data, showing an increased mortality risk among patients admitted with MI with and without hemodynamic compromise and show a gradient of risk across increasing severity of AKI.
Outside of the AMI setting, substantial short-term increased risks associated with AKI have been observed. Among 27,068 patients who underwent coronary angiography over a 12-year period, small changes in serum creatinine (0.25-0.50 mg/dl) were associated with a 1.83 increased risk of mortality, whereas patients with a >1.0 mg/dl increase in serum creatinine had a 3.0-fold increased risk of mortality.22 These estimates are lower than those in the present work, and may in part represent the lower risk clinical setting, as well as selection bias in terms of patients selected to undergo coronary catheterization. Similarly, in a sample of 19,982 adults admitted to an urban medical center, patients with in-hospital increases in serum creatinine had a higher risk of mortality with odds ratios ranging from 4.1 to 16.4 for serum creatinine increases of 0.3 mg/dl to 2.0 mg/dl.1 Also, compared to patients without an increase in serum creatinine, adjusted hospitalization costs were more than $22,000 higher amongst those most severely affected.1 While physicians are generally aware of the risk of AKI associated with coronary angiography and guidelines have recommended measures to minimize this risk,23 our data highlight that AKI is observed in the MI setting regardless of treatment strategy and portends a risk of adverse outcomes.
Complications associated with AKI are particularly pronounced following cardiac surgery. A recent analysis of a multicenter cohort of 3500 adults who underwent cardiac surgery demonstrated that participants with as little as a 25% decrease in eGFR had a 4.0-fold increased risk of post-operative mortality, with rates as high as 9.5-fold increased when eGFR declined by at least 75%.10
The literature is far more robust in examining the association between AKI and longer-term mortality. Not surprisingly, as the length of time from the index event increases, relative increases in mortality rates appear lower. The long-term association of AKI with mortality was examined among 147,007 Medicare patients who presented with AMI from 1994 to 1996. At 10 years, the hazards for mortality ranged from 1.15 (for mild AKI) to 1.33 (for severe AKI).5 Similar findings were observed among 87,094 Medicare patients admitted to US hospitals between 1994 and 19959 and on the Survival and Ventricular Enlargement (SAVE) trial.7 A large meta-analysis of 48 studies comprised of 47,017 patients demonstrated a risk of death of 2.59 among studies with follow-up of at least 6 months.2 Finally, data from 920,985 patients in the Alberta Kidney Disease Network highlights the high longer-term risks associated with AKI, as well as the important contributions of baseline kidney function and proteinuria as an important predictor of disease.4 Taken together, these findings highlight the higher absolute and relative risk of death associated with AKI.
A striking finding from our work was the high rates of in-hospital bleeding events among those with worsening AKI, despite lower rates of usage of oral and intravenous antiplatelet agents. The reason for this is unclear, and may be related to the co-morbidities present in our patients with worsening AKI and the relationship between kidney and platelet dysfunction. Nonetheless, treatment algorithms should take worsening renal function in the AMI setting into account when making use of these agents such as updated dosing of renally-cleared medications including anticoagulants and antiplatelet agents.
There are several implications of this work. First, the prevalence of AKI in the AMI setting is high, and clinicians need to be aware of this important complication. Second, even small increases in serum creatinine are associated with increased risks of bleeding and in-hospital mortality. Particularly notable is the elevated risk of mortality among patients who did not experience in-hospital shock, undergo cardiac surgery, or have baseline CKD, underscoring the impact of AKI even without these important risk factors. Further, clinicians should recognize that AKI is a marker of risk in the setting of MI and is not solely associated with cardiac catheterization and exposure to intravenous contrast agents. Thus, efforts to better incorporate markers of AKI into risk prediction models in the AMI setting, as well as efforts to prevent AKI are critically important and may be a viable method for reducing mortality in the AMI setting. Given the importance of AKI in AMI, AMI registries should seek to collect multiple measures of serum creatinine, including discharge creatinine data to measure rates of recovery.
The strength of our work lies in the large, contemporary sample of MI patients. Further, we assessed outcomes in-hospital, which represents a high-risk setting for patients with AKI. Some limitations warrant mention. Our baseline serum creatinine measures were obtained on admission. Thus, we cannot exclude the possibility that the MI process itself may have already lead to an increase in serum creatinine; this would have lead us to underestimate the magnitude of AKI in our sample and would likely bias our findings towards the null. Second, we used a modified version of the AKIN11 criteria to define AKI, as information on urinary output was not available in our study sample. Universal definitions of AKI do not exist, although leading definitions put forward include AKIN11 and RIFLE.24 In sensitivity analyses, we demonstrated comparability of an AKI definition based on percent creatinine changes. We do not have information collected on the amount of contrast administered during catheterization, bypass times during CABG, nor detailed information on the length of hypotensive episodes during hospitalization. Our registry consists of in-hospital mortality, but not long-term follow-up. We have information on baseline and peak serum creatinine, but not intercurrent creatinine. In addition, we do not have information on acute dialysis. Patients in whom serum creatinine was not collected (and hence excluded from our analysis) were less healthy. We have dates but not times for the in-hospital adverse events in our analysis, limiting our ability to finely assess temporality between AKI and in-hospital adverse outcomes. We do not have information on stopping and restarting inpatient medications. Some of the data presented in Table 3 are unexpected, such as dyslipidemia in association with lower rates of AKI. Given the observational nature of our dataset, we are unable to fully dissect these relationships; further research in more controlled settings may be warranted. Finally, AKI is associated with a multitude of co-morbidities known to be associated with mortality in the MI setting. Thus, we cannot exclude the role of residual confounding in our associations. While we cannot determine a causal relationship between AKI and mortality or bleeding, it does not detract from the use of AKI as a marker of risk.
AKI occurs in 16% of patients hospitalized for acute MI. This common complication is strongly associated with mortality and bleeding. Recognition of these risks and employing strategies to avoid AKI may improve outcomes in MI.
Funding Sources: ACTION Registry®–GWTG™ is an initiative of the American College of Cardiology Foundation and the American Heart Association with partnering support from the Society of Chest Pain Centers, the American College of Emergency Physicians, and the Society of Hospital Medicine. The registry is sponsored in part by the Bristol-Myers Squibb/Sanofi Pharmaceuticals Partnership.
Conflict of Interest Disclosures: Dr. Roe: Research funding: Eli Lilly, Novartis, Merck-Schering Plough, Bristol-Myers Squibb, American College of Cardiology, American Heart Association. Consulting or honoraria: Glaxo Smith Kline, KAI Pharmaceuticals, Novartis, Eli Lilly, Bristol-Myers Squibb, Sanofi-Aventis, Astra Zeneca. Dr. Wiviott: Research funding: Daiichi Sankyo, Eli Lilly, and Schering-Plough. Consulting: AstraZeneca, Bristol-Myers Squibb, and Sanofi-Aventis, Arena. Honoraria: Bristol-Myers Squibb, Daiichi Sankyo, Eli Lilly, Schering-Plough, Merck, the Medicine’s Company, Bayer. ACTION-GTWG Registry is administered by the American College of Cardiology Foundation (ACCF) and is sponsored by Bristol-Myers Squibb/Sanofi Pharmaceuticals Partnership, Genentech, and Schering-Plough Corporation