|Home | About | Journals | Submit | Contact Us | Français|
To determine whether acute renal failure (ARF) increases the long-term risk of progressive chronic kidney disease (CKD), we studied the outcome of patients whose initial kidney function was normal or near normal but who had an episode of dialysis-requiring ARF and did not develop end-stage renal disease within 30 days following hospital discharge. The study encompassed 556,090 adult members of Kaiser Permanente of Northern California hospitalized over an 8 year period, who had pre-admission estimated glomerular filtration rates (eGFR) equivalent to or greater than 45 ml/min/1.73 m2 and who survived hospitalization. After controlling for potential confounders such as baseline level of eGFR and diabetes status, dialysis-requiring ARF was independently associated with a 28-fold increase in the risk of developing stage 4 or 5 CKD and more than a twofold increased risk of death. Our study shows that in a large, community-based cohort of patients with pre-existing normal or near normal kidney function, an episode of dialysis-requiring ARF was a strong independent risk factor for a long-term risk of progressive CKD and mortality.
Acute renal failure (ARF) (also known as acute kidney injury) is one of the most important complications among hospitalized patients. ARF severe enough to require dialysis is associated with a high rate of in-hospital death.1,2 The public health importance of ARF is highlighted by recent reports that its incidence is increasing.3-5
Most studies on ARF have focused on short-term inhospital outcomes,6-10 and the literature is rather limited with regard to the long-term sequelae of ARF. A number of older publications indicated that kidney function often recovered to such an extent after ARF that long-term dialysis was not required,11-13 but these were conducted decades ago when the patient population suffering from ARF was very different. A recent meta-analysis of papers published from January 1985 to October 2007 concluded that the relative risk for chronic kidney disease (CKD) and end-stage renal disease (ESRD) after ARF could not be determined because of a lack of follow-up of appropriate non-ARF controls.14,15
We recently completed an analysis of hospitalized patients with a o preadmission-estimated glomerular filtration rate (GFR) <45 ml/min per 1.73 m2, and found that dialysis-requiring ARF superimposed on preexisting CKD was associated with a very high risk of precipitating ESRD.16 In this study, we focused on investigating long-term renal outcomes in the much larger population of patients with a normal or near normal kidney function. Specifically, we compared outcomes among hospitalized patients who had preadmission baseline-estimated GFR45 ml/min per 1.73 m2 who did or did not suffer dialysis-requiring ARF. We hypothesized that, although most patients who suffered dialysis-requiring ARF and survived hospitalization would recover sufficient kidney function to become dialysis independent, they are at an increased long-term risk of developing progressive CKD.
We identified 562,799 patients who had an index hospitalization during 1996–2003 with a preadmission-estimated GFR45 ml/min per 1.73 m2. There were 703 patients who suffered dialysis-requiring ARF, among whom 295 died (42%) and 65 survived but failed to regain sufficient renal function to become dialysis independent (9%). Thus, 343 patients survived and did not have ESRD within 30 days of hospitalization (49%) (Figure 1). Of the 562,076 hospitalized patients with baseline-estimated GFR45 ml/min per 1.73 m2 who did not experience dialysis-requiring ARF, 555,660 survived to discharge and did not develop ESRD within 30 days of hospitalization (99%) (Figure 1).
The median time lapse between the last outpatient serum creatinine measurement and index hospital admission was 60 days (interquartile range 9–235 days) for those who suffered ARF and 71 days (interquartile range 6–251) for those who did not. Table 1 shows the characteristics of the 703 patients who suffered ARF during hospitalization, the 343 non-ESRD survivors among them and the 3430 matched non-ARF patients.
Among the 703 ARF cases, a review of medical records among a random sample of 100 by a board-certified nephrologist confirmed that all patients indeed experienced ARF. The etiologies of ARF for these 100 random cases are shown in Table 2 and are comparable with that reported in earlier studies.8
During 10,344 person-years of follow-up, 322 patients developed progressive CKD, including 281 cases of observed estimated GFR <30 ml/min per 1.73 m2, and 41 cases of ESRD. The event rate for progressive CKD was 47.9 per 100 person-years for ARF patients and 1.7 per 100 person-years for non-ARF patients (Figure 2). It is noteworthy that all 41 ESRD cases occurred among patients who suffered ARF.
We found in a multivariable Cox regression analysis that, independent of baseline-estimated GFR, age, sex, race/ethnicity, diabetes mellitus, diagnosed hypertension, and documented proteinuria, an episode of dialysis-requiring ARF was associated with a 28-fold increase in the risk of future progressive CKD (adjusted hazard ratio 28.1; 95% CI 21.1–37.6).
All sensitivity analyses showed similar results. Adjusted relative risks (hazard ratios) were 28.6 (95% CI 20.2–40.4) when outcome o was changed to time to second observed estimated GFR <30 ml/min per 1.73 m2 or ESRD; 20.7 (95% CI 14.7–29.0) after excluding GFR values observed within 6 months after discharge; 24.7 (95% CI 16.9–36.2) after excluding GFR values observed within 12 months after discharge; 25.5 (95% CI 18.1–35.9) after adjusting for baseline comorbidities and reason for hospitalization.
Among patients with baseline-estimated GFR60 ml/min per 1.73 m2, adjusted relative risk for progressive CKD was 54.0 (95% CI 34.3–85.1); for patients with baseline-estimated GFR of 45–59 ml/min per 1.73 m2, it was 13.7 (95% CI 9.1–20.7).
In a secondary analysis with death as the primary outcome (total number of deaths was 508), an episode of dialysis-requiring ARF was also independently associated with long-term risk of mortality with an adjusted hazard ratio of 2.3 (95% CI 1.8–3.0).
In this large diverse contemporary cohort of patients receiving usual clinical care, we found that although most patients who suffered dialysis-requiring ARF and survived did recover sufficient kidney function to become dialysis independent in the short run, there was a greatly increased relative risk of developing progressive CKD (including ESRD) in the months to years after hospital discharge. Furthermore, ARF was independently associated with an increased risk of long-term mortality.
The underlying mechanism of the manner in which ARF leads to progressive CKD is not completely understood. But renal parenchymal injury sustained during episodes of ARF may lead to permanent tubulointerstitial fibrosis and a reduction in the number of functioning nephrons. Furthermore, an episode of ARF may—on the basis of animal models—lead to permanent damage to the renal micro-vasculature, and trigger inflammatory and fibrotic signaling pathways that predispose to future accelerated declines in GFR.17-23
As alluded to earlier, until last year, basically all previous studies of renal outcomes after ARF were limited by their case series design and lack of non-ARF comparison individuals.14,15 They also often had limited generalizability because of the nature of the patients studied (for example, only patients treated with continuous renal replacement therapies24 or only those who were in an intensive care unit).25,26 Recently, Ishani et al.27 showed among Medicare beneficiaries (aged 67 years) that ARF was independently associated with the future development of ESRD with an adjusted hazard ratio of 13 in the absence of documented CKD. However, both CKD and ARF were determined in that study using administrative diagnostic codes that are known to be insensitive and suffer from ‘code creep’ bias.28,29 Hence, it is not possible to assess the degree to which their observation was confounded by, for example, the level of pre-ARF renal dysfunction,30 as patients with lower GFR are more likely to develop both ARF31 and ESRD.32 Newsome et al.33 showed that among elderly Medicare patients suffering from acute myocardial infarction, in-hospital acute changes in serum creatinine (peak vs first inpatient serum creatinine) were independently associated with long-term risk of ESRD. However, residual confounding remains a concern, given their finding that seemingly minimal changes in serum creatinine concentration (0.1 mg per 100 ml) increased future risk of ESRD by as much as 45%, even after multivariable adjustment.
Our study extends this literature in several important ways. Our study population was more diverse in terms of age and underlying disease, and hence our results should be more generalizable. We used actual observed outpatient (calibrated) serum creatinine levels to define baseline kidney function. We were able to provide a more comprehensive view of the sequelae of ARF by obtaining data not only on ESRD27,33 but also on cases of progressive CKD (estimated GFR <30 ml/min per 1.73 m2). The magnitude of renal injury was clearly defined and chart validation was performed to validate the ascertainment of ARF (100% positive predictive value). We chose to study ARF episodes of sufficient severity (requiring dialysis) in which there would be a pathophysiological likelihood of a permanent residual renal parenchymal damage. We used matching, in addition to regression, models to adjust for potential confounding.30 Finally, we conducted a series of sensitivity analyses that enhance our confidence in the robustness of the reported effect sizes.
Our findings have important public health and population epidemiology implications. The incidence of ARF has increased substantially over the past few decades.3-5 We previously documented the very high risk of ESRD after acute chronic renal failure16 and we now show that, among the much larger number of patients with a preexisting normal or near normal kidney function, dialysis-requiring ARF greatly increases the risk of subsequent progressive CKD. These trends may help to explain the somewhat paradoxical observation that growth in the incidence of treated ESRD in the United States has outpaced growth in the prevalence of CKD,34 as well as in the prevalence of risk factors for kidney failure such as diabetes mellitus.35 Our findings show that the array of clinical consequences of ARF extends long after hospitalization.36 Moreover, if successful strategies of ARF prevention could be developed, they may reduce the incidence of ESRD.37
From a research point of view, future studies of interventions in ARF should examine not only inpatient outcomes38-40 but also extend follow-up beyond discharge toevaluate the impact of different treatment strategies on the subsequent level of kidney function (beyond the absence of dialysis dependency). Our study also contributes toward a novel insight into risk factors for ESRD. It is noteworthy that over an 8-year period of follow-up, we did not observe any cases of ESRD among patients who had baseline-estimated GFR45 ml/min per 1.73 m2, unless there was superimposed dialysis-requiring ARF. Studies on ESRD risk factors in the past have not considered the role of ARF,32,41-43 but the very high relative risk associated with ARF merits it being included among other more established ESRD risk factors, such as hypertension or diabetes.
Our findings also have important clinical implications. Given these data, ARF should not be viewed as a transient, self-limited problem requiring no additional medical attention after hospital discharge. Currently, only a minority of patients who suffered ARF are being followed up by nephrologists or are undergoing testing for evidence of subtle kidney disease or of risk factors for future loss of kidney function (such as albuminuria).44 Changing this and early institution of preventive measures, such as strict blood pressure control or aggressive renin–angiotensin blockade, may reduce the high risk of progressive CKD documented after ARF. Preventing the development of progressive CKD may also ameliorate the increased risk of death observed among survivors of ARF, as CKD is a strong and independent risk factor for cardiovascular disease and overall mortality.45
The limitations of this study need to be mentioned. We only studied patients who suffered dialysis-requiring ARF. Hence, we are unable to ascertain the impact of less-severe degrees of ARF/acute kidney injury. The precise etiology that led to ARF was not always clear; but it would be unlikely for dialysis to be needed for reversible pre-renal or post-renal causes, and the most common cause of intrinsic ARF among hospitalized patients is acute tubular necrosis.1,2 Our chart review findings (Table 2) are consistent with this. As our study was conducted among Kaiser health-plan enrollees, our results might not be entirely applicable to uninsured patients or to other health-care settings, although our cohort was diverse in terms of age, sex, race/ethnicity, and other factors. We did not obtain information on the graded severity of proteinuria, which was classified dichotomously. We did not analyze whether proteinuria or hypertension worsened after ARF. The observation that the adjusted relative risks for progressive CKD were greater for higher (60 ml/min per 1.73 m2) than for lower (46–59 ml/min per 1.73 m2) estimated GFR is due to the low absolute risk of progressive CKD among those with a more preserved kidney function.
In summary, our findings underscore the important role of dialysis-requiring ARF in determining the likelihood of progressive CKD among persons with a normal or near normal kidney function. Although it has been long recognized that dialysis-requiring ARF is associated with short-term inpatient mortality and morbidity, more attention now needs to be focused on the long-term sequelae of ARF months to years after hospital discharge.
The study population consists of adult members (20 years old) of the Kaiser Permanente of Northern California (Kaiser) who were hospitalized between 1 January 1996 and 31 December 2003 and who had a serum creatinine measured before hospitalization, giving an estimated GFR of 45 ml/min per 1.73 m2 by the Modification of Diet in Renal Disease equation.46,47
Kaiser is a large integrated health-care delivery system currently insuring more than 3.2 million individuals in the San Francisco Bay area. Other than at the extremes of socioeconomic spectrum and age, wherein there are slightly lower-than-population percentages, Kaiser membership is highly representative of local surrounding and statewide populations.48 Relevant coexisting conditions among Kaiser enrollees can be assessed using comprehensive health-plan demographic, hospitalization, ambulatory visit, laboratory, prescription medication, and diagnostic code data.45
All laboratory tests are financially covered as a part of basic Kaiser membership benefits. Essentially, all inpatient and outpatient laboratory test results are performed at or comprehensively obtained through a central Kaiser regional laboratory. Baseline kidney function before ARF was defined as estimated GFR from the last outpatient serum creatinine measurement before hospitalization. Subsequent kidney function evolution was defined using estimated GFR values from outpatient serum creatinine measurements after discharge. We chose to use outpatient serum creatinine measurements to define baseline kidney function, as these were more likely to reflect the usual steady state function in the absence of acute illness.16,30,31 The Kaiser regional health-plan laboratory had previously calibrated measurement of serum creatinine against the Modification of Diet in Renal Disease core laboratory to ensure a more accurate estimate of GFR (applicable to the study time period 1996–2003).49,50
Dialysis-requiring ARF was defined as having a peak inpatient serum creatinine level higher than the last outpatient serum creatinine (baseline) level by 50% and a receipt of renal replacement therapy during hospitalization. Renal replacement therapy including acute peritoneal dialysis, hemodialysis, or hemofiltration was identified by the International Classification of Disease, ninth Edition (ICD–9) procedure codes 54.98 and 39.95 and the Current Procedural Terminology (CPT) codes 90935, 90937, 90945, 90947, and 90999.
Non-ARF (unexposed) individuals were hospitalized patients with an outpatient-estimated GFR45 ml/min per 1.73 m2 documented from their latest serum creatinine measurement before admission but who did not suffer dialysis-requiring ARF.
For both ARF and non-ARF patients, we analyzed only the first hospital admission for each patient, which was preceded by a known outpatient serum creatinine level after 1 January 1996 (this admission is referred to as the index hospitalization). Patients who had ESRD (defined as receipt of kidney transplantation or already on maintenance dialysis) before index hospitalization were excluded from further consideration. ESRD status was ascertained by cross-linking the US Renal Data System (USRDS) and an internal Kaiser ESRD registry.16,31,51 At the time of cross-linkage, USRDS data were complete through to 31 December 2003.
We determined the proportion of patients who suffered dialysis-requiring ARF and became dialysis dependent within 30 days of hospital discharge using USRDS and the internal Kaiser registry. This classification of cases of ARF immediately precipitating ESRD was made only after a confirmatory chart review by a board-certified nephrologist. We chose a 30-day window, given the uncertainty inherent in the start date of ESRD from both USRDS and Kaiser ESRD registries. Furthermore, progression to ESRD at the time of hospital discharge or any time within this 30-day window was judged to be of similar importance clinically.
Patients who suffered dialysis-requiring ARF, who survived their index hospitalization, and who did not develop ESRD within 30 days of discharge were then compared with patients who did not suffer dialysis-requiring ARF, who also survived their index hospitalization, and remained ESRD-free for 30 days after hospital discharge. Our primary interest was to compare, between these two groups, the risk of developing progressive CKD in subsequent months to years. Progressive CKD was defined as the development of a subsequent estimated GFR30ml/min per 1.73 m2 or ESRD (that is, stage 4 or higher CKD by the National Kidney Foundation classification system47). Both were ascertained through to 31 December 2003.
Any observed differences in outcome among patients with and without dialysis-requiring ARF may be due to confounding by differences in baseline characteristics, such as level of kidney function, as reduced GFR is a strong predictor of ARF31 and of even lower GFR or ESRD (and other adverse outcomes).45 We used matching as one tool to control for confounding.30 Each dialysis-requiring ARF patient who survived hospitalization and who did not develop ESRD within 30 days of hospital discharge was matched to 10 hospitalized non-ARF patients on the basis of the following matching variables: preadmission-estimated GFR level (45–59 ml/min per 1.73 m2, 60–89 ml/min per 1.73 m2, and 90 ml/min per 1.73 m2), diabetes mellitus (present/absent), age (by 10-year categories), sex (male/female), and race/ethnicity (White, Black, Asian or Pacific Islander, Hispanic, Others, or Unknown). A 10:1 ratio was chosen to enhance statistical power, given the large number of non-ARF individuals available.
We performed survival analysis using Cox proportional hazards regression censoring for death, disenrollment from health plan, or end of follow-up on 31 December 2003. We confirmed that the proportional hazards assumption was not violated by comparing estimated log (−log(survivor function)) vs time (person-years) survivor curves. We determined the independent association between ARF and progressive CKD in a Cox model that took into account the matching design, and also adjusted for diagnosed hypertension and documented proteinuria. To additionally control for residual confounding within each age or GFR category, we also entered age and estimated GFR as linear terms into the regression model.
We then conducted a series of sensitivity analyses. To avoid potentially misclassifying low estimated GFR values observed during recovery after dialysis-requiring ARF as CKD, we repeated our analysis ignoring any serum creatinine values observed 6 or 12 months after index hospitalization. To be more certain of the chronicity of the low GFR to establish CKD, we examined time to the second (confirmed) estimated GFR30 ml/min per 1.73 m2. To minimize residual confounding, we additionally adjusted for the comorbid conditions listed in Table 1 and for major reasons for hospitalization using primary discharge diagnosis divided generally into ‘diseases of the circulatory system’ (390–459), ‘diseases of the respiratory system’ (460–519), ‘diseases of the gastrointestinal system’ (520–579), ‘infectious diseases’ (001–139), ‘neoplasms’ (140–239), and all others.16,52 We also conducted a stratified analysis by baseline-estimated GFR above or below 60 ml/min per 1.73 m2.
In secondary analysis, we examined death as an end point to determine whether dialysis-requiring ARF was independently associated with longer-term mortality. Deaths were identified from health-plan administrative databases and from member proxy reporting,45 Social Security Administration files,53 and the California Automated Mortality Linkage Information System.54 We used the same matched cohort design and Cox model that adjusted for age and estimated GFR as linear terms, as well as for diagnosed hypertension, documented proteinuria, other comorbid conditions listed in Table 1, and reason for hospitalization.
Analyses were performed using SAS (Cary, NC, USA). The institutional review boards of collaborating institutions approved the study. Waiver of informed consent was obtained because of the nature of the study.
We thank Niela Pomernacki for her expert technical assistance with the study. Some of the data reported here were supplied by the United States Renal Data System. The interpretation and reporting of these data are the responsibility of the authors and should in no way be seen as an official policy or interpretation of the United States government.
The National Institutes of Health funded our study (grants R01 DK67126 and U01 DK82223). The funding source had no role in the collection, analysis, or interpretation of data or in the decision to submit the paper for publication. The authors had full access to the data files for the study.
The authors declared no competing interests.