The Assessment, Serial Evaluation, and Subsequent Sequelae of Acute Kidney Injury (ASSESS-AKI) Study: Design and Methods

Alan S Go; Chirag R Parikh; T Alp Ikizler; Steven Coca; Edward D Siew; Vernon M Chinchilli; Chi-yuan Hsu; Amit X Garg; Michael Zappitelli; Kathleen D Liu; W Brian Reeves; Nasrollah Ghahramani; Prasad Devarajan; Georgia Brown Faulkner; Thida C Tan; Paul L Kimmel; Paul Eggers; John B Stokes


BMC Nephrology 

In This Article


Study Organization

The ASSESS-AKI Study consists of a Data Coordinating Center (Pennsylvania State University), three Clinical Research Center networks through Kaiser Permanente Northern California (Oakland, CA; San Francisco, CA; Walnut Creek, CA), Vanderbilt University (Nashville, TN), and the Translational Research Investigating Biomarker Endpoints in Acute Kidney Injury (TRIBE-AKI) network (New Haven, CT; Cincinnati, OH; London, Ontario; Montreal, Quebec). A central laboratory for analysis of core biochemistries is located at the University of Minnesota and an electrocardiography reading center is based at Wake Forest University (Figure 1). In addition, ASSESS-AKI includes an External Advisory Committee and NIDDK project scientists.

Figure 1.

Organizational structure for the ASSESS-AKI Study. NIDDK = National Institute of Diabetes, Digestive and Kidney Diseases; DCC = Data Coordinating Center; CRC = Clinical Research Center network; and ECG = electrocardiography

The Institutional Review Boards of the Data Coordinating Center and participating Clinical Research Centers' institutions approved the study.

Study Design

ASSESS-AKI will employ a parallel, matched, prospective cohort design of adult participants with and without AKI. In addition, ASSESS-AKI will attempt to enroll and prospectively follow all eligible children undergoing cardiac surgery requiring cardiopulmonary bypass who are participants in the TRIBE-AKI Consortium study. The study will enroll 1200 participants (1100 adults, 100 children) with approximately 50% of adult participants having AKI and the remaining 50% representing matched adult participants without AKI. Scheduled follow-up visits for each participant will occur during the subsequent four years. Informed consent will be obtained in all patients in accordance with the principles of the Declaration of Helsinki.

Clinical Research Center Networks

Participating Clinical Research Center Networks have active research programs related to AKI and the consortium is leveraging their existing research expertise, resources and infrastructure to establish the prospective, multi-center ASSESS-AKI cohort.

Kaiser Permanente Northern California Kaiser Permanente of Northern California in collaboration with the University of California, San Francisco has been conducting a series of NIDDK-sponsored (U01DK060902, R01DK067126, R01DK058411) longitudinal studies characterizing the epidemiology and outcomes of acute, chronic, and end-stage renal disease within Kaiser Permanente's large and diverse community-based population in the San Francisco and greater Bay area. Kaiser Permanente is one of the largest integrated health care delivery systems in the U.S and provides comprehensive care for >3.2 million members that are ethnically and socioeconomically diverse and highly representative of the northern California and statewide population. Kaiser delivers comprehensive inpatient and outpatient care to its members through 18 hospitals and >60 additional ambulatory medical offices and captures many aspects of its care through the use of its comprehensive clinical (e.g., inpatient and outpatient laboratory tests) and administrative (e.g., diagnoses, procedures, mortality) databases, which will be leveraged for ASSESS-AKI. The Kaiser Permanente Division of Research will lead recruitment and enrollment of participants hospitalized in medical and surgical wards as well as intensive care units (ICU) at three Kaiser Permanente Medical Centers located in Oakland, San Francisco and Walnut Creek.

Vanderbilt University Vanderbilt University Medical Center (VUMC) is a large tertiary referral center serving the middle Tennessee area and surrounding region and provides comprehensive acute and critical care services. ASSESS-AKI investigators are primarily leveraging the ongoing National Heart, Lung and Blood Institute (NHLBI)-funded (U01HL081332)-sponsored Validation of Acute Lung Injury Biomarkers for Diagnosis (VALID) Study. VALID is a single-center, prospective study whose purpose is to develop and validate a panel of diagnostic and prognostic plasma and/or urine biomarkers in a diverse cohort of 2550 critically ill patients at high risk for developing ALI/ARDS as well as AKI.[15] All adult (≥18 years) patients admitted to one of four ICUs at VUMC who remained in the ICU at day 2 were eligible for enrollment. Patients are excluded if they experienced a cardiac arrest before enrollment, had transfer orders written or anticipated within 4 hours, died or were discharged within 48 hours of ICU admission, were admitted for uncomplicated overdose, were in the ICU for >3 days before enrollment, or who had chronic lung disease requiring oxygen supplementation or pulmonary fibrosis. Per parent study protocol, blood and urine samples are currently collected at study enrollment on ICU day 2 and subsequent sampling on ICU day 4. In addition to VALID subjects, VUMC investigators are recruiting additional subjects from the same VALID ICUs as well as neurologic and burn ICUs who meet inclusion criteria for VALID.

TRIBE-AKI TRIBE-AKI is an ongoing prospective cohort study of more than 1800 adults and children sponsored by the NHLBI (R01HL085757) whose goal is to validate selected biomarkers for the diagnosis and risk stratification of AKI after cardiac surgery (coronary artery bypass and/or valvular repair or replacement). Patients are excluded if they had any of the following: pre-operative AKI (≥0.5 mg/dL increase in serum creatinine concentration from preadmission to initiation of cardiac surgery); pre-operative end-stage renal disease (serum creatinine level ≥4.5 mg/dL [400 μmol/L] receiving chronic dialysis or prior renal transplant), prior cardiac transplant or insertion of left ventricular assist device; receipt of nephrotoxic agents within 48 hours preceding cardiac surgery, or acute infective endocarditis. Pediatric participants are limited to those requiring cardiopulmonary bypass. TRIBE-AKI participants have comprehensive clinical data, blood and urine samples collected pre-operatively as well as from the first five post-operative days. ASSESS-AKI will include a subset of TRIBE-AKI sites: Yale University, London Health Sciences Center (Ontario), University of Cincinnati Children's Hospital, and Montreal Children's Hospital (Quebec).

Cohort Participants

The ASSESS-AKI Study will enroll a diverse group of adults (age 18 to 89 years) and children (age one month to 18 years) with and without AKI from participating Clinical Research Center networks.

Baseline Kidney Function To participate, all patients must have an available baseline pre-admission serum creatinine value, which is then used to estimate glomerular filtration rate (eGFR). A baseline serum creatinine value is defined as the outpatient, non-emergency department test result nearest to the index hospitalization. For the Kaiser Permanente and VALID Clinical Research Centers, the nearest value between 7 and 365 days before admission will be used, while for the TRIBE-AKI Clinical Research Center, the baseline serum creatinine can be present between 1 and 365 days before surgery, provided the patient is undergoing elective surgery. The rationale for this approach is based on preliminary data from the three Clinical Research Center networks demonstrating that the vast majority of potentially eligible participants do not have more than two to three pre-admission serum creatinine values during this time frame and that the most recent value is more likely to reflect the subject's "baseline" kidney function before the index hospitalization. All serum creatinine results must be performed using an isotope dilution mass spectrometry (IDMS)-traceable serum creatinine assay.

Exclusion Criteria Exclusion criteria were selected to balance the goal of maximizing representativeness of the cohort, clinical accuracy of AKI, and feasibility in achieving the project goals. These criteria are detailed in Table 1. The data sources used to ascertain information on these criteria include electronic and paper medical records, other electronic databases, and patient interviews.

Definition of AKI We recognize the limitations of the most recently proposed definitions for AKI (i.e., RIFLE[16] and Acute Kidney Injury Network [AKIN][17]), which are based only on changes in serum creatinine concentration and/or urine output. However, despite enthusiasm for potentially more sensitive and specific novel serum and urine biomarkers,[18] to date, none have been sufficiently validated as better measures of AKI or of subsequent prognosis than serum creatinine-based AKI criteria. A major goal of ASSESS-AKI is to provide key insights into the prognostic value of novel AKI biomarkers. Therefore, AKI will be operationalized as follows which is anticipated to capture a broad spectrum of kidney injury. For adult participants, AKI will be defined as ≥50% relative increase and/or absolute increase ≥0.3 mg/dL (26 μmol/L) in peak inpatient serum creatinine compared with baseline outpatient serum creatinine. For pediatric participants, AKI will be defined as ≥50% relative increase in peak inpatient serum creatinine compared with baseline serum creatinine. We did not incorporate urine output criteria from the AKIN classification scheme because of concern about the systematic availability and quality of data about urine output, especially in non-ICU patients who are unlikely to have indwelling urinary catheters. Furthermore, incorporating the urine output criteria for AKIN might overly enrich our cohort for patients with pre-renal azotemia.

To enhance the likelihood of enrolling an adequate number of adult participants with more severe AKI, we have set an enrollment target of at least one third of AKI participants having ≥100% relative increase in serum creatinine. To increase the probability of having an adequate number of adult participants with AKI due to causes other than rapidly reversible pre-renal azotemia, we have set an enrollment target of at least one third of AKI participants who meet AKI criteria lasting ≥48 hours. These additional enrollment targets are not mutually exclusive, and we anticipate significant overlap in these pre-specified subgroups. Furthermore, study nephrologists at each participating site will review selected index hospitalization information to classify each enrolled AKI case into one of the following presumptive categories: acute tubular necrosis (ATN), pre-renal azotemia, and other/unknown.

Definition of Absence of AKI

Subjects will be considered not to have AKI if they meet the following criteria. For adult participants, non-AKI status will be defined as having both <20% relative increase and an absolute increase ≤0.2 mg/dL (18 μmol/L) in peak inpatient serum creatinine compared with baseline outpatient serum creatinine. For pediatric participants, non-AKI status will be defined as <50% relative increase in peak inpatient serum creatinine compared with baseline serum creatinine.

Initial Screening and Enrollment of Participants

Subject recruitment will vary by participating site and among adult versus pediatric participants, in accordance with requirements of local institutional review boards' guidelines and the requirements of sites within each Clinical Research Center network. As described above, all pediatric patients undergoing cardiac surgery requiring cardiopulmonary bypass at the two pediatric TRIBE-AKI sites will be screened for enrollment into ASSESS-AKI. Enrollment of ASSESS-AKI pediatric participants will occur post-operatively during the index hospitalization. During this first inpatient visit of enrolled children, baseline clinical data and quality of life questionnaires will be administered and 1 blood and urine specimen obtained within the first four post-operative days will be stored for future biomarker testing.

Among adult participants, we will enroll a parallel matched cohort of patients with and without AKI (Figure 2). Adult patients with AKI will be identified during the index hospitalization and screened for initial eligibility. During this inpatient visit, enrolled AKI patients will undergo urinalysis with microscopy through the hospital clinical laboratory and provide at least one sample of blood and urine for future biomarker testing within 96 hours of the episode of AKI.

Figure 2.

Summary of identification and enrollment approach for AKI and Non-AKI participants. The figure applies to adult participants only. Any eligible pediatric subject undergoing cardiopulmonary bypass-requiring cardiac surgery is approached for participation during the inpatient phase.

Given the matched parallel cohort design, we will identify and enroll a sample of hospitalized adult patients who did not appear to suffer an AKI episode and who are matched in a minimum 1:1 AKI:non-AKI ratio, with each non-AKI subject individually matched to their corresponding AKI subject on the following set of key confounding characteristics: Clinical Research Center and presence of baseline chronic kidney disease using an CKD-EPI[19] equation-estimated GFR threshold of <60 ml/min/1.73 m2. In addition, we will further attempt to match on the presence or absence of clinical cardiovascular disease, presence or absence of diabetes mellitus, category of baseline eGFR (15–29, 30–44, 45–59, 60–89, 90–150 ml/min/1.73 m2), adult age category (18–39, 40–49, 50–59, 60–69, 70–79, 80–89 years), and hospital location where AKI episode occurred (ICU versus non-ICU).

Follow-up Visits and Retention Strategies

The follow-up visit and contact schedule are summarized in Table 2. All recruited participants will be invited to an in-person baseline study visit at 3 months after the AKI episode for AKI participants or 3 months after hospital discharge for non-AKI participants. Participants will return annually for in-person follow-up visits. Participants will be contacted by telephone at the 6-month intervals between clinic visits to obtain information on study events, or updates on general health and contact information.

Consistent with other cohort studies (e.g., Atherosclerosis Risk in Communities Study,[20] Cardiovascular Health Study[21]), we project that approximately 3 to 5% of participants may be lost to follow-up annually. Multiple approaches will be used to prevent participant dropout. The National Death Index will be searched periodically for all participants lost-to-follow-up to ensure complete vital status information. We will implement previously tested retention strategies to promote a high level of long-term participation. These will include free medical testing, semi-annual contact with participants via telephone calls, along with newsletters containing study updates and information about kidney disease, personalized mailings, and reimbursement of time and travel expenses.

Collection of Study Data

At the outpatient visit at 3 months following the index hospitalization, adult participants will be screened again for eligibility and eligible persons will have consent obtained for long-term follow-up (Table 2). Pediatric participants will have had consent for long-term follow-up visits obtained during the index hospitalization. Information will be collected on detailed sociodemographic information, lifestyle habits, medical and family history, quality of life, current medication use, quality of life and functional status (SF-12v2™ Health Survey[22] in adults, PedsQL Generic and Cardiac modules in children),[23] cognitive function (Modified Mini-Mental Status Examination,[3MS][24] and Trails B[25]), anthropometric measures (weight, height), and resting blood pressure and heart rate. In addition, blood specimens for DNA, sera and plasma as well as a random urine sample will be obtained for local urine dipstick testing (CLINITEK Status® Analyzer, Siemens, New York, NY). Sera, plasma and urine samples will be collected annually and stored for subsequent measurement of renal and cardiovascular-related biomarkers related to multiple pathways involving early AKI. The list of all the different biological specimen types that are being collected at each visit and will be stored in the NIDDK biorepository are given in Table 2.


A major goal of ASSESS-AKI is to evaluate the utility of urine and blood biomarkers for improving the diagnosis and risk stratification after a hospitalized episode of AKI. Given rapid and ongoing advances in the discovery of putative novel biomarkers for AKI,[18] the biomarkers to be evaluated within ASSESS-AKI will be prioritized based on the currently available evidence at the time of testing. The initial preliminary set of biomarkers includes those with the strongest clinical evidence base as markers of early AKI and will be measured in all study participants. Given the current data supporting the use of these markers for the detection of AKI, it will be important to know whether these markers predict short or long term outcomes. This set will include urine biomarkers (IL-18,[26] NGAL,[27] KIM-1,[28] cystatin C,[29] L-FABP[30] and NAG[31]) and blood biomarkers (serum cystatin C,[32] serum NGAL[33] and plasma IL-6[34]).

Renal Outcomes

Kidney Function Measurement The primary renal outcome is the change in kidney function during follow-up. Kidney function will be defined before and after an AKI episode (as well as among those without AKI) using outpatient serum creatinine concentration measurements. Given the known limitations of using serum creatinine alone as a measure of kidney function,[35] except for its use in defining an episode of AKI per the criteria described previously, we will use the CKD-EPI equation to estimate GFR using an IDMS-traceable serum creatinine assay among all adult study participants. For pediatric participants, we will estimate GFR using the recommended Schwartz formula[36] based on serum creatinine values measured in local laboratories using the same assays for baseline and follow-up measurements within those sites. Urine albuminuria will be measured using a spot albumin-to-creatinine ratio.[37]

Incident CKD Among participants without pre-existing CKD at the index hospitalization, we will examine time to development of incident CKD with significant loss of renal function defined as experiencing both a minimum 25% reduction in level of eGFR compared with baseline and achieving CKD Stage 3 or worse[37] during follow-up.

Progression of CKD Among participants with pre-existing CKD at the index hospitalization (defined as an eGFR <60 ml/min/1.73 m2), we will examine time to progression of CKD, defined as experiencing at least a 50% reduction in level of eGFR compared with baseline or progressing to CKD Stage 5.[37]

Development of ESRD Development of ESRD after the 3 month follow-up visit will be defined as any of the following: (1) peritoneal dialysis or hemodialysis treatment at least once a week for at least 12 consecutive weeks, (2) receipt of a kidney transplant and/or (3) death while receiving dialysis.

Incident or Recurrent Episodes of AKI We will attempt to ascertain incident and recurrent episodes of AKI. Based on available data collected during follow-up, we will use the same criteria described previously to define incident (among non-AKI participants) or recurrent (among AKI participants) episodes of AKI. However, we recognize that some study participants may be hospitalized at non-Clinical Research Center network facilities where complete medical and laboratory records may not be readily available (and where non-IDMS serum creatinine assays may be used), hindering accurate determination of whether observed changes in serum creatinine reflect progression of kidney dysfunction or a new episode of AKI. In such cases, we will pursue whether the hospitalizations included administrative diagnostic codes for AKI and acute dialysis treatments which are typically coded administratively and likely reflect more severe AKI episodes.

Cardiovascular Outcomes To maximize future collaborations with other studies focused on kidney disease among adult populations, we have modelled our definitions after those used in the CRIC Study[38] and various longitudinal studies (Cardiovascular Health Study,[21] Atherosclerosis Risk in Communities[20] and Antihypertensive and Lipid Lowering Treatment to Prevent Heart Attack Trial[39]). The general approach will be to obtain self-reported and/or site-specific database information on potential outcome events and subsequently obtain information on qualifying International Classification of Diseases, Ninth and Tenth Editions (ICD-9 and ICD-10) codes at each site which will facilitate review of relevant medical records to adjudicate. The following outcomes are only relevant for the subgroup of adult participants given that they are extremely rare among children.

Coronary Heart Disease Standard definitions will be used to classify a coronary heart disease event. This includes acute coronary syndromes such as unstable angina, non-ST-elevation myocardial infarction and ST-elevation myocardial infarction.[40,41] Myocardial infarction will be further classified according to recent international consensus guideline.[42] Sudden cardiac death will be obtained by mortality files and subject proxy contacts. It will be defined as either an unwitnessed death without another obvious cause or death occurring within one hour of the onset of ischemic symptoms per a proxy.[43] Silent myocardial infarction will be defined as new, pathologic Q waves on serial electrocardiograms (ECG)[44] among the subgroup of enrolled adult participants with the event date assigned as the mid-point between the relevant annual visits. Coronary artery revascularization will include either percutaneous coronary intervention with or without intracoronary stenting or coronary artery bypass surgery of one or more coronary blood vessels.

Heart Failure Heart failure will be based on hospitalizations for a clinical heart failure syndrome using relevant discharge diagnosis codes and confirmed based on Framingham Heart Study clinical criteria ascertained from medical records.[45] We will not require evidence of systolic dysfunction (e.g., left ventricular ejection fraction <40%) or diastolic dysfunction on echocardiography.[46]

Cardiac Arrhythmias and Electrocardiographic Abnormalities Arrhythmias and other ECG abnormalities will be based on serial ECGs using Minnesota Code definitions,[47] which have been used in epidemiological studies and have direct clinical applicability. These include development of atrial fibrillation, atrial flutter, left and right bundle branch block, atrioventricular conduction defects, and left ventricular hypertrophy among others.

Cerebrovascular Outcomes

Pertinent cerebrovascular disease outcomes include ischemic stroke and intracranial hemorrhage, and carotid endarterectomy. Ischemic stroke will be defined as acute development of a neurological deficit fitting a vascular distribution, lasting ≥24 hours, and no other evident etiology such as intracranial hemorrhage, vasculitis, tumor, or trauma.[48,49] Intracranial hemorrhage will require validation by brain imaging or pathologic evidence, and should have a documented history consistent with a stroke syndrome, diminished consciousness, or headache.[50] Carotid endarterectomy will include both surgical endarterectomy and balloon angioplasty with or without carotid stent placement.

Peripheral Arterial Disease Outcomes

Outcomes will include aortic aneurysm and lower extremity arterial revascularization or amputation for refractory ischemia. Lower extremity revascularization will include both percutaneous peripheral artery angioplasty and surgical arterial bypass procedures, and lower extremity amputation will include procedures performed for refractory ischemia. Hospitalizations for thoracic or abdominal aortic aneurysm dissection, rupture or repair (using percutaneous or surgical procedures) will be included.


Deaths will be identified primarily through surveys of subjects or their proxy contacts and review of medical records or death certificates, if available. We will also seek to obtain information on social security number from participants to conduct probability matches with Social Security Administration vital status files[51] and National Death Index[52] among the subset set of participants who are lost to follow-up. All-cause mortality will be the preferred outcome given known significant errors in assigning etiology.[53]

Statistical Approach and Power

The primary outcomes for the study are time-to-event outcomes, such as time to death, a renal event or a clinical cardiovascular event. Some of these events will be known exactly on a continuum and could be right-censored. Exact dates for the occurrence of some of the events, however, may only be known to occur within a specific time interval tk, k = 1, 2, 3, 4, 5 (t1 = 0–3 months, t2 = 3–12 months, t3 = 12–24 months, t4 = 24–36 months, t5 = 36–48 months). The endpoints of these five time intervals correspond to the planned in-person study visits. Therefore, the statistical analyses in these situations will invoke continuous time-to-event models that account for right-censored and interval-censored data.

The hazard function for a continuous time-to-event outcome is of the form


  1. λijk(t; xij) is the hazard function at time t with covariate vector xijk for the kth member of the jth matched set within the ith site, j = 1, 2,…,ni, and k = 0 (non-AKI), 1 (AKI),

  2. λ0(t) is the baseline hazard function,

  3. β is an unknown parameter vector, and

  4. xijk is a vector of regressors of interest.

The regressors that will appear in xijk for the primary statistical analyses are as follows using as examples different variables: binary indicator variables at month 0 for non-AKI/AKI status, CKD status, gender, Hispanic ethnicity, cardiovascular disease status, diabetes status and sepsis status; ordinal variables at month 0 for racial group (0 = white/European, 1 = Black/African American, 3 = Asian/Pacific Islander, 4 = Native American, 5 = Other/Admixed), eGFR (0 = 15 to 29, 1 = 30 to 44, 2 = 45 to 59, 3 = 60 to 89, 4 = 90–150) and albumin-to-creatinine ratio (0 = <0.15, 1 = 0.15 to 0.5, 2 = >0.5 to 1.0, 3 = >1.0 to 3.0, and 4 = >3.0), age (0 = 1 to 17, 1 = 18 to 39, 2 = 40 to 49, 3 = 50 to 59, 4 = 60 to 69, 5 = 70 to 79, 6 = 80 to 89).

The hazard model described above, however, is not the final form of the hazard model that will be applied in this study. Instead, the hazard model needs to account for (1) the dependency between an AKI subject and a non-AKI subject within a matched pair and (2) informative censoring. The occurrence of some of the renal and cardiovascular events may not be independent of the censoring event of death. For example, individuals who are censored because of death may have been at higher risk for renal and cardiovascular events. Therefore, a bivariate hazard function for the simultaneous modeling of the renal (or cardiovascular) event and death is invoked.[54,55] Let (Tijk, Dijk) denote the continuous time of a renal or cardiovascular event and the time of death, respectively, for the kth member of the jth pair within the ith site, j = 1, 2,…,ni, and k = 0 (non-AKI), 1 (AKI). The bivariate hazard model is

where the [eTij eDij]'s are independent and identically distributed according to a bivariate normal distribution with null mean vector and positive-definite variance matrix. The covariance between eTij and eDij represents the magnitude of the relationship between the occurrence of the renal (or cardiovascular) event and the censoring event of death. An estimated value of the covariance near zero indicates non-informative censoring, whereas an estimated value of the covariance distant from zero indicates informative censoring.

The sample size for the study is (1) 550 adult AKI subjects and 550 matched adult non-AKI subjects and (2) 50 pediatric AKI subjects and 50 pediatric non-AKI subjects, at the 3-month visit. For a two-sided, 0.05 significance level test of the relative risk equaling 1.0, the sample size of 1,200 yields greater than 80% statistical power, while allowing for a 15% withdrawal rate, for detecting a relative risk between 1.35 (when the event rate for AKI subjects is 30%) and 1.90 (when the event rate for AKI subjects is 10%).


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.