Kidney function decline after a non-dialysis-requiring acute kidney injury is associated with higher long-term mortality in critically ill survivors

This study was a retrospective observational cohort study based on the National Taiwan University Hospital Study Group on Acute Renal Failure (NSARF) database established in the surgical intensive care units (ICU) of a medical center and its three branch hospitals in different cities [17‐23]. This database contains data prospectively collected from patients with AKI during their stay in ICU and continuously recorded data for outcome analysis. The AKI was defined and classified according to the risk, injury, failure, loss, and end-stage kidney (RIFLE) criteria [24]. In the present study, we enrolled all patients who experienced AKI and survived to discharge but did not receive renal replacement therapy throughout the index hospitalization between January 1 2002 and June 30 2010. Patients were excluded if they had a history of ESRD or kidney transplantation. The follow-up period was continued until December 31 2010.

The study was approved by the Institutional Review Board of National Taiwan University Hospital (No.31MD03). An informed consent was waived because there was no breach of privacy, and the study did not interfere with the clinical decisions related to patient care.

Baseline demographic and clinical data were assessed at the time of hospital admission, as previously reported [19]. Pertinent medical history included diabetes mellitus (DM, defined as being treated with oral hypoglycemic agents or insulin), hypertension (defined as taking anti-hypertensive drugs or systolic and diastolic blood pressure > 145/95 mmHg at the time of hospitalization), coronary arterial disease (CAD, defined by the diagnosis of ischemic heart disease prior to admission, and positive electrocardiographic findings), cerebral vascular accidents (CVA, ischemic or hemorrhagic), severe congestive heart failure (CHF, defined as a New York Heart Association (NYHA) functional class of III or IV), severe chronic obstructive pulmonary disease (COPD, defined as requiring long-term bronchodilators or steroids), metastatic carcinoma, and hematological malignancies. Body mass index (BMI) was calculated upon admission to the hospital.

All of the clinical parameters and biochemical data were recorded at ICU admission and at the peak of the AKI, including the acute physiology and chronic health evaluation (APACHE) II [25] and sequential organ failure assessment (SOFA) [26] scores. Data were also recorded on the use of inotropic drugs, expressed as inotropic equivalent (mcg/kg/min) = dopamine + dobutamine + 100 × epinephrine + 100 × norepinephrine + 100 × isoprotenolol + 15 × milrinone [27]; mechanical ventilation; cardiopulmonary resuscitation (CPR); intra-aortic balloon pump (IABP); extracorporeal membrane oxygenation (ECMO); total parenteral nutrition (TPN), and any history of surgery (elective or emergent). Body weight change was calculated as the percentage change from ICU admission to the peak of the AKI.

Baseline SCr was determined as the last value measured at least one month, but no more than a year, prior to the index admission [28]. For patients without valid premorbid data, we used the lowest SCr value during the index admission [29]. The estimated glomerular filtration rate (eGFR) was calculated via the Modification of Diet in Renal Disease (MDRD) equation [30]. By using both SCr and urine output criteria, the maximum RIFLE stage throughout the index hospitalization was determined [24].

All of the longitudinal SCr measurements after the index hospitalization and during the follow-up period were collected for each enrolled patient. For every post-discharge SCr measurement, the eGFR was calculated via the MDRD equation [30]. All of the eGFR data were derived from SCr measurements, with the exception of the eGFR at 90 days after the onset of the AKI (post-90d-eGFR). It was suggested that AKI patients should be evaluated for the new onset or worsening of pre-existing CKD at least 3 months after the onset of AKI [31‐33]. Thus, we estimated the SCr at 90 days after the onset of AKI via linear interpolation from the two closest SCr measurements before and after the 90-day period from the onset of AKI. The post-90d-eGFR was then calculated from the estimated SCr at 90 days after the onset of AKI. In patients who did not have a SCr measurement at > 90 days after the onset of an AKI, the post-90d-eGFR was characterized as unknown.

The endpoints included stages 3 to 5 CKD entries, ESRD, and mortality during the follow-up period. Stage 3 to 5 CKD entries were defined as the first day when the eGFR decreased below 60, 30, and 15 mL/min/1.73 m², respectively. ESRD was the certification by a nephrologist of the need for long-term dialysis, as determined from the national Taiwan Renal Registry Data System, which receives data on all dialysis patients every 3 months. Patient survival after discharge was determined from the National Health Insurance Research Database (NHIRD) [34] in February 2011. The NHIRD contains health care data on > 99% of the entire population in Taiwan and covers all inpatient and outpatient medical benefit claims. The survival period was calculated from the onset of AKI until each outcome, whichever came first.

Statistical analyses were performed using Stata 10.0 (StataCorp, College Station, TX, USA) and R 2.11.1 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided P value ≤ 0.05 was considered statistically significant. The distributional properties of the continuous variables were presented by either means ± SD or medians (25th, 75th percentiles), whereas categorical variables were presented as frequencies and percentages. The differences between the groups were assessed using the chi square (χ²) test, Fisher's exact test, and the two-sample t-test, as indicated.

The impact of AKI on the prognoses of the initial CKD or non-CKD patients was carefully evaluated stepwise. First, allowing 3 months as a reasonable period of the time for patients to recover from the AKI [31‐33], we considered 90 days after the onset of AKI as the starting point to examine the endpoints of CKD entries, using a Cox proportional hazards model. Specifically, stages 3, 4, and 5 CKD were assessed only in patients with post-90d-eGFR ≥ 60, 30, and 15 mL/min/1.73 m², respectively. Therefore, the patients eligible for the assessment of stage 3 CKD were also eligible for the risk assessments of stage 4 and 5 CKD, and the patients eligible for the risk assessment of stage 4 CKD were also eligible for the risk assessment of stage 5 CKD. In addition, all of the enrolled patients were eligible for the risk assessments of ESRD and long-term mortality after the onset of AKI. The Kaplan-Meier method was applied to estimate survival curves for all the outcomes.

Next, to examine the harmful effects of a CKD entry and ESRD on a patient's time to death, the time-dependent variables of stage 3, 4, and 5 CKD (CKD3, 4, 5-TD), ESRD (ESRD-TD), and the related interaction terms were put into the Cox regression analysis model. Technically, the original wide-form data (n = 634) were reconstructed into a long-form structure (n = 53,833) using the so-called counting process style of input for the specified Cox regression model [35, 36]. Then at each ordered event time, we recorded the values of the CKD3-TD, CKD4-TD, CKD5-TD, ESRD-TD, and the related interaction terms for each of the patients at risk in the transformed long-form data set.

The basic model-fitting techniques for (1) stepwise variable selection, (2) goodness-of-fit assessment, and (3) regression diagnostics (for example, the verification of proportional hazards assumption, residual analysis, detection of influential cases, and determination of multicollinearity) were used in our regression analyses.

According to the calculated post-90d-eGFR, renal function status at 90 days after the onset of AKI is shown in Table 2A. At that time, only 187 patients (29.5%) had a preserved eGFR, while 244 (38.5%) had CKD. A post-90d-eGFR could not be obtained in the 160 patients (25.2%) who had no SCr measurement at > 90 days after the onset of AKI and in the 43 patients (6.8%) who died before 90 days after the onset of AKI.

The long-term outcomes pertaining to stage 3, 4, and 5 CKD were only assessed in hospital survivors who had a post-90d-eGFR ≥ 60 (n = 187), 30 (n = 374), and 15 mL/min/1.73 m² (n = 421), respectively (Table 2B). All 634 patients were assessed for the long-term risks of ESRD and mortality (Table 2B). The median time interval between the onset of the AKI and the development of stage 3 and 4 CKD was 740 and 1,984 days, respectively. Only 35 patients (5.5%) developed stage 5 CKD, and 13 patients (2.1%) developed ESRD during the follow-up period (Figure 1A).

The median follow-up duration was 700.5 (330.5, 1,296) days. The endpoint of all-cause mortality occurred in 207 patients (32.7%). The Kaplan-Meier curves showed that the median patient survival time was 2,178 days after the onset of AKI (Figure 1B).

The development of advanced CKD and mortality are competitive endpoints, as many AKI survivors do not live long enough to develop advanced CKD. We therefore assessed the composite outcomes. The median survival time of patients assessed for the composite endpoints "stage 3 CKD or death", "stage 4 CKD or death", "stage 5 CKD or death", and "ESRD or death", were 685, 1,319, 1,743, and 2,048 days following the onset of AKI, respectively (Table 3 and Figure 1C).

A Cox proportional hazards model, adjusted for all of the variables listed in Table 1 including baseline eGFR, was conducted to identify the independent factors associated with each composite outcome (Table 4). It was found that age was consistently a risk factor for all of the composite endpoints. A higher baseline eGFR was protective against the composite outcomes of "stage 3 CKD or death" and "stage 4 CKD or death". When assessing the composite outcomes of "stage 4 CKD or death" and "stage 5 CKD or death", SCr at the peak of the AKI appeared to be an important risk factor.

Additionally, another Cox proportional hazards model, adjusted for all of the variables listed in Table 1, including baseline eGFR, was performed to evaluate the independent factors associated with long-term mortality. Time-dependent CKD3, 4, 5-TD, ESRD-TD, and the related interaction terms were also included in the regression model. After adjusting for the effects of the other covariates in the final model, the interaction terms of baseline eGFR and CKD3, 4, and 5-TD, but not ESRD-TD, significantly predicted long-term mortality (Table 5). The main CKD3, 4, and 5-TD factors themselves were not significant in the final model. The hazard ratios (HRs) product suggested a 0.7%, 2.3% (1.007*1.016 = 1.023), and 4.1% (1.023*1.017 = 1.041) increase in mortality risk for every 1 mL/min/1.73 m² decrease from baseline eGFR in individuals who progressed to stage 3, 4, and 5 CKD during the long-term follow-up period, respectively. In other words, among patients with the same baseline kidney function, those who progressed to more advanced CKD had a higher risk of death. Furthermore, among the patients who progressed to the same stage of CKD during the follow-up period, those who had a higher baseline eGFR had a higher risk of death. This indicated that mortality risk increased significantly in a graded manner with kidney function decline from baseline eGFR to the advanced stages of CKD during the follow-up period. The other risk factors for long-term mortality included age (HR 1.04), receiving general surgical services (HR 1.77), and the administration of CPR during ICU admission (HR 6.55). Conversely, patients who received organ transplantation (HR 0.29) and who had a higher mean arterial pressure at the peak of the AKI (HR 0.98) appeared to be at less of a risk for mortality during the follow-up period. The interaction term of CPR and ECMO indicating a higher grade of life support during resuscitation, also predicted a lower risk of death after discharge (HR 0.17). This final model passed the test for proportional hazards assumption and fit the data well (adjusted R² = 0.289).