|Home | About | Journals | Submit | Contact Us | Français|
Acute kidney injury (AKI) is common in hospitalized patients. The impact of AKI on long-term outcomes is controversial.
Systematic review and meta-analysis.
Persons with AKI.
MEDLINE and EMBASE databases were searched from 1985 through October 2007. Original studies describing outcomes of AKI for patients who survived hospital discharge were included. Studies were excluded from review when participants were followed for < 6 months.
AKI, as defined by acute changes in serum creatinine or acute need for renal replacement therapy.
CKD, cardiovascular disease, mortality.
Forty nine studies that contained a total of 47,017 participants were reviewed. Fifteen of these studies reported long-term data for patients without AKI. The incidence rate of mortality was 8.9 per 100 person-years in survivors of AKI and was 4.3 per 100 patient-years in survivors without AKI (RR 2.59, 95% CI 1.97-3.42). AKI was associated independently with mortality risk in 6 of 6 studies that performed multivariate adjustment (adjusted RR 1.6-3.9) and it was associated with myocardial infarction in 2 of 2 studies (RR 2.05, 95% CI 1.61-2.61). The incidence rate of CKD after an episode of AKI was 7.8 per 100 patient years and the rate of ESRD was 4.9 per 100 patient-years.
The relative risk for CKD and ESRD after AKI was unattainable due to lack of follow-up of appropriate non-AKI controls.
The development of AKI, as defined by acute changes in serum creatinine characterizes hospitalized patients at elevated risk of long-term adverse outcomes.
Acute kidney injury (AKI) is a common complication in hospitalized patients. As a conservative estimate, in the U.S. about 17 million admissions a year are complicated by AKI, resulting in additional costs to the health care system of $10 billion.1 Furthermore, perhaps partly due to increased recognition, the incidence of AKI has risen from approximately 60 to 500 per 100,000 population in the last decade.2, 3 Concurrently, the incidences of chronic kidney disease (CKD) and end-stage renal disease (ESRD) have increased over the past several years. The association between the development of AKI and higher in in-hospital mortality has been well-known for decades and was reported in several studies.4-10 However, the association of AKI with long term mortality has received less attention, probably due to the apparent reversibility of the clinical episode as observed by subsequent improvements in serum creatinine.
Studies in experimental animals however demonstrate that AKI causes permanent damage to the microvasculature and subsequent abnormalities in kidney structure and function.11-13 Through inflammatory and fibrotic signaling pathways, the residual kidney damage can lead to progressive structural kidney damage, which may then predispose to worsening hypertension, proteinuria, and more rapid declines in glomerular filtration rate (GFR), 14-16 all of which are well-known risk factors for cardiovascular disease.17-19
Currently, little attention is given to AKI after it resolves in clinical practice throughout the healthcare system.20 Understanding the impact of AKI on long term outcomes may identify a segment of patients during hospitalization who are at greater risk of longer term sequelae, guiding follow-up care after discharge. A better understanding of the association may also highlight the importance of preventing AKI. Additionally, if AKI episodes confer increased risk for CKD, cardiovascular disease, and mortality, then therapies known to prevent these outcomes (e.g., angiotensin-converting enzyme [ACE] inhibitors, statins, tighter blood pressure control) could be applied earlier and more aggressively. We conducted this systematic review and meta-analysis to characterize fully the relations between AKI and the long-term outcomes of CKD, cardiovascular disease, and death.
This study was performed in accordance with published guidelines for systematic review, analysis, and reporting for meta-analyses of observational studies.21
A study published from January 1985 onward was eligible for inclusion if it conducted at least 6 months of follow-up of patients after a defined episode of AKI and reported the incidence of at least one of the following outcomes: A) survival; B) cardiovascular events (myocardial infarction, stroke, or congestive heart failure); or C) CKD. We excluded studies that contained fewer than 50 participants in the initial cohort and studies that exclusively focused on “AKI syndromes” (e.g., rapidly progressive glomerulonephritis, hemolytic-uremic syndrome, hepatorenal syndrome, or after bone marrow transplantation). Studies were not required to include a control group without AKI.
We searched the MEDLINE and EMBASE (January 1985 - October 2007) databases using the following terms: Renal Insufficiency, acute; acute kidney, acute kidney failure, acute kidney tubule necrosis, kidney tubule necrosis; survival (explode), prognosis (explode), treatment outcome, recovery of function, follow-up studies. An additional search was performed via the Science Citation Index Expanded on the Web of Science, and the references of all selected articles were reviewed to identify any eligible studies. There was no language or age restriction. Each article chosen by the primary reviewer was reviewed by a second reviewer to confirm eligibility.
The primary outcome measures were rates of long-term mortality, cardiovascular events (myocardial infarction [MI], stroke, and congestive heart failure), and incident CKD in survivors of AKI. We assessed rates of each of these endpoints as defined by the number of events in survivors beyond hospitalization or other relevant short-term time period (28 or 30 days) divided by the patient-years of follow-up. Regarding the numerator, many of the studies reported long-term outcomes only in survivors of hospitalization, however, some studies (8 total) only listed cumulative survival for all patients, including those who died in-hospital or within 30 days of admission. In these eight studies, 22-29 we subtracted the number of acute deaths from the total number of deaths in the cohort during the entire period of follow-up to obtain the numerator for calculation of the mortality rate. The denominator for the event rate calculation (patient-years) was estimated by addition of the following two values: A) the product of the duration of follow-up and the number of patients who survived until the last follow-up period of the study; B) the product of the duration of follow-up and the number of patients who had died by the time of the last follow-up period divided by 2 (each person who died contributed half of the number of patient-years).
Two reviewers (BY, SC) extracted data using a standardized data extraction form. The reviewers extracted data on the characteristics of participants (number, age, sex), clinical setting, type of study (prospective vs. retrospective), dates of enrollment, definition of AKI, definition of CKD, and incidence of the outcomes (mortality, cardiovascular events, and CKD) among participants with and without AKI. Due to significant heterogeneity of definitions of AKI, we created the following 3 categories of AKI severity: Mild AKI, as defined by a threshold change in creatinine > 0% but < 50%; Moderate AKI, as defined by a change in serum creatinine ≥ 50%; and Severe AKI, defined as dialysis-requiring AKI. These groups were not mutually exclusive, as patients who required dialysis in studies that defined AKI as only a change in creatinine above a certain threshold were included with the rest of the patients in the cohort with AKI. Only one study clearly reported long-term survival for those who required dialysis separately from those without while including patients with less severe forms of AKI in the cohort.30 We were not able to separate these participants out in the other cohorts because we did not have individual patient data. Methodological quality was assessed using criteria adapted from Hayden et al.31 Any discrepancies on data extraction or quality score were resolved by a third reviewer.
Overall results for each outcome were mathematically pooled, and rate ratios were obtained using techniques that accounted for within and between study heterogeneity (random effects method of DerSimonian and Laird).32 We formally assessed heterogeneity of effects between studies with the Cochran Q and the I2 statistics.33 To examine the association between study-level characteristics and outcomes, we fitted random-effects meta-regression models to the natural logarithm of the relative risks by using the meta-regression procedure in Comprehensive Meta Analysis Version 2.0 (Englewood, NJ). At the study level, we evaluated the association of the following patient and study characteristics with mortality outcomes after AKI: clinical setting, definition of AKI, and duration of AKI (transient vs. persistent). In the 3 studies that separated AKI into 2 categories of duration, transient vs. persistent was defined as the following: Welton et al defined transient AKI as return of serum creatinine to within 10% of baseline value at day 3; Liano and Loef et al defined transient AKI as serum creatinine value at the time of discharge returning to preoperative level or below. All meta-analyses were performed using Comprehensive Meta Analysis Software Version 2.0 (Englewood, NJ).
We identified 3,808 citations meeting our search criteria. After excluding 689 duplicate citations, 3119 abstracts were evaluated, and 419 articles were selected for further review (Figure 1). Reasons for exclusion are outlined in Figure 1. Forty eight articles were deemed eligible for this systematic review.22-30, 34-73
Study and patient characteristics from the selected articles are summarized in Table 1. The most common clinical setting for the studies of AKI were critical illness (n=20, 42%), general hospitalized patients (n=10, 21%), and cardiac surgery (n=9, 19%). Most studies were single center (n=41, 85%) and retrospective (n=41, 85%). Five (10%) studies included pediatric patients.
The definitions of AKI varied substantially (Table 2). Maximum length of follow-up ranged from 6 months to 17 years. In-hospital mortality ranged from 6-80%.
Fifteen of the 48 studies (31%) provided mortality data on non-AKI controls. The mortality rate in these studies was 8.9 per 100 person-years for patients who survived hospitalization with AKI and 4.3 per 100 person-years for patients who survived hospitalization without AKI (Rate Ratio [RR] 2.59, 95% CI 1.99-3.42) (Figure 2).
Only two studies examined cardiovascular endpoints after AKI. At 1 year after AKI, 15.4% (56/377) of survivors of AKI and 7.0% (817/11755) of survivors without AKI suffered a MI (relative risk 2.05, 95% CI 1.61-2.61, I2 = 0%).25, 65 One of the two studies examined the risk of MI at 3 points of follow-up (0.5 years, 1 year, 5 years) and the increased risk of MI persisted over time (relative risk 1.6, 1.85, 1.75, respectively, at each time point).
Twenty seven (56%) studies provided data on the incidence of CKD or ESRD in patients who survived hospitalization with AKI. Eleven of these studies reported some form of CKD as an outcome (definitions varied- see footnotes in Table 3), and 23 reported ESRD as an outcome (7 studies reported the incidence of both CKD and ESRD). The rate of CKD after an episode of AKI was 7.8 per 100 patient years and the rate of ESRD was 4.9 per 100 patient-years. In studies that definitively excluded patients with pre-existing CKD, the rate of CKD after AKI was 6.2 per 100 person-years 28, 37, 44, 58, and the rate of ESRD was 4.2 per 100 person-years.39, 43, 58, 59, 66 None of the studies compared the rate of CKD or ESRD after AKI to patients within the same cohort without AKI, thus we could not determine the rate ratio for CKD or ESRD after an episode of AKI. None of the studies examined the incidence or rate of CKD after AKI that resolved compared to AKI without recovery.
Meta-regression analyses indicated that two study level factors, severity of AKI (β=0.12, 95% CI 0.004-0.22) and length of follow-up (β= -0.05, 95% CI -0.09- -0.02), were associated with the magnitude of long-term risk of premature death. In addition to these two study level factors, we also planned a priori to examine the relationship between AKI and long-term death among clinical settings and duration of AKI (transient vs. persistent) (Table 3).
We examined the association between AKI and death stratified by severity of AKI (Figure 3). The least severe definition of AKI among the studies was a rise in serum creatinine of ≥ 25% or a decrease in creatinine clearance ≥ 10% (3 studies). Mild AKI was associated with approximately a 70% increase in mortality risk, as the rate of death was 6.3 per 100 person-years in survivors of mild AKI and was 3.4 in survivors without AKI (RR 1.67, 95% CI 1.41 - 1.98). The risk of long-term death was nearly three-fold higher in patients with moderate and severe AKI compared to patients without AKI (Figure 3).
The long-term mortality rate and rate ratios for death after AKI varied according to the clinical setting, and ranged from 4.5 deaths per 100 person-years in patients who underwent aortic surgery (RR 1.45, 95% CI 1.17-1.79) to 13.1 deaths per 100 person-years in patients who underwent percutaneous coronary intervention (RR 2.89, 95% CI 2.32-3.61) to (Table 3). Compared to those without AKI, the risk for long-term death was significantly greater in patients who survived AKI across every clinical setting. In general, studies were statistically heterogeneous.
Three studies examined the relationship between AKI and long-term death stratified by duration of AKI (transient vs. persistent).58, 60, 72 Both transient and persistent AKI were associated with an increased risk of death compared to patients without AKI (RR 2.54, 95% CI 2.10-3.06 and RR 2.46, 95% CI 1.68-3.60, respectively) (Table 3). When compared directly, the risk of death was not different in patients with transient or persistent AKI (persistent vs. transient RR 1.15, 95% CI 0.84-1.57).58, 60, 72
We were able to significantly diminish statistical heterogeneity when we grouped studies by both severity and clinical setting. For example, the I2 statistic dropped to 0% and 42%, respectively for the following subgroups: A) Moderate AKI in patients undergoing PCI,25, 52, 65 B) AKI requiring renal replacement therapy in patients undergoing cardiac surgery.53, 57 Although the sample size was lower, the point estimate for the risk of mortality with AKI remained elevated and statistically significant for these sub-groups. The I2 was 0% in the 3 studies of mild AKI and the 2 studies of AKI after left ventricular assist device placement.
We assessed the methodological quality of the studies using the criteria by Hayden et al.31 Only 33 studies (69%) reported whether patients were lost to follow-up, and 25 of those 33 studies had attrition rates of < 10%. Only 6 studies (13%) adequately adjusted for relevant confounding factors. Three of the 48 (6%) studies met all 6 criteria for quality, 21 (44%) studies met 5 criteria, and 24 (50%) studies met 4 or less of the criteria.
Six studies examined the risk of long-term mortality in patients after an episode of AKI, adjusted for factors such as older age, baseline kidney function, history of MI, peripheral vascular disease, diabetes, and hypertension, among others. The risk for long-term death in patients with AKI was consistently greater than patients without AKI, with all adjusted point estimates exceeding 1.6 after multivariate adjustment.25, 26, 51, 52, 60, 72 Three studies 25, 51, 52 used logistic regression models and reported adjusted odds ratios and 3 studies26, 60, 72 used Cox Proportional Hazards models and reported adjusted hazards ratios. In these six studies, the independent strength of association between AKI and long-term death was similar to the strength of association between other established risk factors and long-term death.
This study demonstrates that patients who survive AKI have a higher rate of long-term mortality, and other adverse outcomes than patients who survive hospitalization without AKI. The associations were consistent in every clinical setting. With increasing severity of AKI, the association between AKI and death was even stronger, and the association with long-term mortality seemed to be present even in patients with more rapidly reversible AKI. These findings expand upon the results from our previous systematic review that demonstrated the increased risk of short-term death associated with very small changes in serum creatinine.74 Thus, all severities of AKI, even the most mild forms, are associated with both short and long-term survival.
Although we demonstrated a consistent, reproducible association between AKI and long-term adverse outcomes, this does not imply causality. Patients with AKI are often sicker than those without AKI, and AKI is often the complication of multiple organ dysfunction. However, in studies that performed multivariate analyses, the strength of association between AKI and mortality was similar, or greater than that associated with other established risk factors including diabetes, peripheral vascular disease, and chronic obstructive lung disease. Thus, at the very least, AKI can serve as an easily assessable prognostic marker of increased long-term mortality risk.
It remains possible that some of the association between AKI and poor long-term outcomes is causal in nature. As already mentioned, the relationship remained present after controlling for important co-morbidities. In addition, there is biologic support for this relationship as data from experimental animals does demonstrate that AKI can induce tissue injury in other organ systems such as the heart (apoptosis),75 lungs (loss of pulmonary vascular integrity),76 and kidney (fibrosis).11, 12 Despite the reversible nature of clinical AKI where serum creatinine returns to baseline after the episode, AKI may result in permanent injury to other vital organs and thus affect future survival.
If AKI is causally associated with long-term mortality, and not simply a marker for unmeasured risk factors, the most common pathway would be through progression to CKD after an episode of AKI. Although 27 studies examined the cumulative incidence of CKD or ESRD in survivors of AKI, we could not calculate risk ratios for the development of CKD because none of the studies included comparable controls without AKI.
The results of any systematic review are limited by the quality of the primary studies. We assessed the methodologic quality of these studies using accepted criteria. In 41 of the 48 studies, data were collected retrospectively using existing health records. The majority of the studies were single-center studies, thus limiting total achievable sample size and generalizability. Fifteen of the 48 studies did not report whether patients were lost to follow-up, and only 6 of the 15 studies (40%) that followed non-AKI patients adequately adjusted for relevant confounding factors. Nonetheless, the lower bound confidence interval of the risk ratio remained above 1 for all studies that presented adjusted long-term mortality, supporting a consistent relationship. Primary articles used several different definitions for both AKI and CKD, complicating the clinical application of these results. However, the effect was observed across all thresholds of AKI definition, and all-cause mortality is an outcome less prone to bias. It is unfortunate that no studies were able to provide a point estimate for the hazard ratio of CKD or ESRD because of a lack of non-AKI controls. Few studies excluded patients with prior CKD from the reported cumulative incidence of CKD. Also, no studies examined the pattern or slope of GFR decline after AKI over time. This type of analysis would help enlighten the medical community as to whether AKI evolves into CKD abruptly and then plateaus, whether GFR declines in a “step-function”, or whether GFR decreases at a constant rate after an episode of AKI. Many studies also did not have access to a previous “baseline” measure of renal function, prior to the acute illness that would have increased the sensitivity of AKI diagnosis. Most patients were classified as AKI if kidney function deteriorated further after the start of hospitalization, thus representing a more severe phenotype of AKI. Finally, our point estimates should be interpreted with caution, as most pooled estimates invoked a great deal of statistical heterogeneity. We attributed the cause of heterogeneity to the interaction between the definitions of AKI (severity) and clinical setting, as we were able to substantially ameliorate statistical heterogeneity when we accounted for these factors.
While the precise risk for adverse outcomes and death may not be known exactly from our meta-analysis, the consistency of our results suggest that patients who suffer AKI should be monitored closely even after recovery of kidney function for the development of CKD, myocardial infarction, and other potential mediators of poor long-term survival. Future longitudinal studies of AKI should examine larger cohorts of patients (via multi-center prospective studies), with appropriate non-AKI controls, repeated follow-up, and utilize both continuous outcomes (measures of kidney function) and hard-outcomes such as CKD, ESRD, cardiovascular events, and death to truly elucidate the consequences of AKI. Future studies should also rigorously adjust for the level of baseline kidney function when determining the impact of AKI towards CKD. If the suggestion of harm from AKI from this meta-analysis is validated, then determining effective treatment strategies to prevent AKI and adverse outcomes following AKI will be the next challenge that lies ahead for nephrologists and intensivists.
Financial Disclosure: None.