Use of Biomarkers to Identify Acute Kidney Injury to Help Detect Sepsis in Patients With Infection John A. Kellum, MD1 OBJECTIVES: Although early recognition of sepsis is vital to improv- Antonio Artigas, MD2 ing outcomes, the diagnosis may be missed or delayed in many patients. Kyle J. Gunnerson, MD3 Acute kidney injury is one of the most common organ failures in patients with sepsis but may not be apparent on presentation. Novel biomarkers for Patrick M. Honore, MD4 acute kidney injury might improve organ failure recognition and facilitate J. Patrick Kampf, PhD5 earlier sepsis care. Thomas Kwan, PhD5 DESIGN: Retrospective, international, Sapphire study. Paul McPherson, PhD5 SETTING: Academic Medical Center. H. Bryant Nguyen, MD6 Thomas Rimmelé, MD7 PATIENTS: Adults admitted to the ICU without evidence of acute kidney injury at time of enrollment. Nathan I. Shapiro, MD, MPH8 Jing Shi, PhD9 INTERVENTIONS: None. Jean-Louis Vincent, MD10 MEASUREMENTS AND MAIN RESULTS: We stratified patients Lakhmir S. Chawla, MD11 enrolled in the Sapphire study into three groups—those with a clinical di- agnosis of sepsis (n = 216), those with infection without sepsis (n = 120), for the Sapphire Investigators and those without infection (n = 387) at enrollment. We then examined 30-day mortality stratified by acute kidney injury within each group. Finally, we determined the operating characteristics for kidney stress markers (tis- sue inhibitor of metalloproteinases-2) × (insulin-like growth factor binding protein 7) for prediction of acute kidney injury as a sepsis-defining organ failure in patients with infection without a clinical diagnosis of sepsis at en- rollment. Combining all groups, 30-day mortality was 23% for patients who developed stage 2–3 acute kidney injury within the first 3 days compared with 14% without stage 2–3 acute kidney injury. However, this difference was greatest in the infection without sepsis group (34% vs 11%; odds ratio, 4.09; 95% CI, 1.53–11.12; p = 0.005). Using a (tissue inhibitor of metalloproteinases-2) × (insulin-like growth factor binding protein 7) cutoff Copyright © 2021 The Author(s). of 2.0 units, 14 patients (11.7%), in the infection/no sepsis group, tested Published by Wolters Kluwer Health, positive of which 10 (71.4%) developed stage 2–3 acute kidney injury. Inc. on behalf of the Society of Critical The positive test result occurred a median of 19 hours (interquartile range, Care Medicine and Wolters Kluwer 0.8–34.0 hr) before acute kidney injury manifested by serum creatinine or Health, Inc. This is an open-access ar- urine output. Similar results were obtained using a cutoff of 1.0 for any ticle distributed under the terms of the stage of acute kidney injury. Creative Commons Attribution-Non Commercial-No Derivatives License CONCLUSIONS: Use of the urinary (tissue inhibitor of metalloprotein- 4.0 (CCBY-NC-ND), where it is per- ases-2) × (insulin-like growth factor binding protein 7) test could identify missible to download and share the acute kidney injury in patients with infection, possibly helping to detect work provided it is properly cited. The sepsis, nearly a day before acute kidney injury is apparent by clinical criteria. work cannot be changed in any way or used commercially without permission KEY WORDS: acute kidney injury; insulin-like growth factor binding from the journal. protein 7; tissue inhibitor of metalloproteinases-2 DOI: 10.1097/CCM.0000000000004845 e360     www.ccmjournal.org April 2021 • Volume 49 • Number 4 Online Clinical Investigations E arly recognition of sepsis is a key to improv- METHODS ing outcomes. Timely antibiotics and source Study Design control, recognition and reversal of shock, and provision of organ support are cornerstones of sepsis We conducted our analysis using a previously published management. These actions are different from those cohort of critically ill patients (Sapphire) in which uri- taken in patients with infection without sepsis. For this nary (TIMP-2) × (IGFBP7) was measured and complete reason, sepsis care protocols have been developed to details are reported elsewhere (6). Briefly, Sapphire in- capture these cornerstones of treatment and are imple- cluded 728 adult subjects with critical illness defined as mented at many, if not most, hospitals. cardiac or respiratory failure and without evidence of A pragmatic definition of sepsis has been devel- AKI at enrollment. After excluding five patients enrolled oped in which infection-associated organ dysfunction under consent from a legally authorized representative is the central tenant (1). Implicit in this definition is who later died prior to reconsent, we analyzed 723 sub- the notion that clinicians can identify organ dysfunc- jects. The primary endpoint was moderate-to-severe tion as well as its relationship to infection. The three AKI (KDIGO stage 2–3) within 12 hours of sample most common organ dysfunctions in sepsis are car- collection. Data were collected by the investigators and diovascular (i.e., shock), renal (i.e., acute kidney in- analyzed by independent statisticians not directly affil- jury [AKI]), and respiratory, each occurring in 50% iated with the study. The study protocol was approved or more of patients (2). For previously healthy adults, by the Western Institutional Review Board (Olympia, shock and hypoxemia are usually easy to recognize, WA) and also by the institutional review board or and other organ failures such as mental status changes ethics committee of each study site, as required. All can produce obvious and dramatic symptoms as well. patients (or authorized representatives) provided writ- However, early detection of AKI can be difficult since ten informed consent. We defined infection and sepsis it does not cause symptoms and changes in serum cre- based on the clinical diagnosis assigned by the treating atinine, and urine output can be delayed relative to physicians at enrolling sites at the time of enrollment. organ damage. In some patients, occult (normoten- Antibiotics were prescribed for 98% of patients with sive) shock may occur, but arterial lactate levels can be infection without sepsis. AKI was defined and staged used to identify these patients. Similarly, oxygen satu- using full KDIGO criteria (both serum creatinine and rations or arterial blood gases can help detect respira- urine output) (7) based on interpolation (linear) across tory failure. Importantly, even in patients with septic each hour to provide precise timing for AKI manifesta- shock, AKI is associated with a substantial increased tion. The design, execution, and reporting of this study risk of death—27.7% for stage 2–3 AKI versus 6.2% for meet the Strengthening the Reporting of Observational patients without AKI (3). Studies in Epidemiology (8) and the Standards for Biomarkers to detect AKI prior to clinical/lab- Reporting of Diagnostic Accuracy criteria (9). oratory indicators of kidney dysfunction are now routinely available. The NephroCheck Test (Astute Measurements Medical, San Diego, CA) measures the product of two biomarkers, tissue inhibitor of metalloproteinases Paired urine and serum samples for biomarker and cre- (TIMP)–2 and insulin-like growth factor binding pro- atinine assessment, respectively, were obtained at the tein (IGFBP) 7. At a cutoff of 2.0 (ng/mL)2/1,000, the time of enrollment and within 24 hours of ICU admis- test has a specificity of 95% to predict Kidney disease sion, flash-frozen, stored at less than or equal to –70°C, improving global outcomes (KDIGO) stage 2–3 AKI and thawed immediately prior to analysis. TIMP-2 in the next 12 hours (4). We have previously shown and IGFBP7 concentrations were measured by immu- that the test performs well in patients with sepsis with noassay with the NephroCheck Test on the Astute140 an area under the receiver operating characteristics Meter (Astute Medical, San Diego, CA) by technicians curve of 0.84 (5). Thus, urinary (TIMP-2) × (IGFBP7) blinded to clinical data. The TIMP-2 and IGFBP7 con- might serve as a means to identify occult sepsis (i.e., centrations in ng/mL were multiplied and divided by not clinically apparent) by identifying AKI in patients 1,000 to obtain the composite (TIMP-2) × (IGFBP7) with infection. results in units of (ng/mL)2/1,000. We assessed Critical Care Medicine www.ccmjournal.org     e361 Kellum et al severity of illness and organ dysfunction/failure with Baseline characteristics for the cohort as a whole, as the Acute Physiology and Chronic Health Evaluation well as broken out by these three categories, can be (APACHE)–III (10) and Sequential Organ Failure found in Table 1 (with additional analysis provided Assessment (SOFA) (11) scores. Nonrenal APACHE- in Supplemental Table 1, http://links.lww.com/CCM/ III and SOFA scores were calculated by subtracting the G112). Compared with patients with a diagnosis of renal components from these scores. sepsis at enrollment, those with infection only had lower severity of illness by APACHE and SOFA but Statistics were otherwise comparable. We stratified patients into three groups—those with AKI Rates and Outcomes a clinical diagnosis of sepsis (n = 216), those with in- fection without a clinical diagnosis of sepsis (n = 120), Across the entire cohort, 175 patients (24.2%) devel- and those without infection (n = 387) at enrollment. oped stage 2–3 AKI within the first 3 days after enroll- All patients had either respiratory or cardiac dysfunc- ment. Mortality at 30 days was 23% for patients who tion or both; thus, in the infection without sepsis group, developed stage 2–3 AKI compared with 14% without organ failure was not attributed to infection. AKI status stage 2–3 AKI (p = 0.003). Comparison of AKI rates was determined within the first 3 days after enrollment, and outcomes across the three groups are shown in and mortality was assessed for up to 9 months from Table 2. Patients with infection with or without sepsis enrollment. We compared 30-day mortality rates be- had similar rates of stage 2–3 AKI (27.3% vs 26.7%) tween patients with or without AKI for each infection and 30-day mortality (20.4% vs 17.5%). However, group by using a logistic regression model. Rates are patients with infection without a diagnosis of sepsis expressed as proportions, and time to event is reported had a larger differential effect on mortality from stage as median with interquartile ranges (IQRs) among 2–3 AKI (34% vs 11%; odds ratio [OR] 4.09; 95% CI, patients with events. Mortality out to 9 months was 1.53–11.12; p = 0.005) compared with (27% vs 18%; plotted using Kaplan-Meier curves stratified by both OR, 1.71; 95% CI, 0.84–3.44; p = 0.13) for patients with urinary (TIMP-2) × (IGFBP7) and AKI. Differences sepsis and (17% vs 12%; OR, 1.44; 95% CI, 0.72–2.75; between strata were analyzed using the log-rank test. p = 0.29) for patients without infection (Fig. 1). Hazard ratios for 9-month mortality by urinary (TIMP- 2) × (IGFBP7) and AKI strata were determined using Biomarker Results Cox proportional hazards analysis. For comparisons Using a urinary (TIMP-2) × (IGFBP7) cutoff of 2.0, of baseline characteristics across the three infection 14 patients (11.7%) in the infection/no sepsis group groups or between pairs of groups, categorical variables tested positive, of which 10 (71.4%) developed stage were analyzed using Fisher exact test and continuous 2–3 AKI. The positive test result occurred a median of variables using Kruskal-Wallis test across three groups 19 hours (IQR, 0.8–34.0 hr) before stage 2–3 AKI man- and Wilcoxon rank-sum test for pairwise comparisons. ifested by serum creatinine or urine output. However, A p value of less than 0.05 was considered to indicate this timing is based on retrospective interpolation of statistical significance. All reported p values are two creatinine and urine output values such that the actual sided. All analyses were performed using R 3.4.4 (R time for clinical diagnosis was likely much later. To Foundation for Statistical Computing, Vienna, Austria). explore this further, we excluded three patients who would have already met clinical criteria for stage 2–3 RESULTS AKI at the time of enrollment based on these extrapo- lated values. For the remaining patients, the median Baseline Characteristics lead time was actually 31 hours (IQR, 19–40 hr). Of the 723 patients enrolled in Sapphire, 216 had a We also repeated this analysis using a 1.0 cutoff for clinical diagnosis of sepsis at enrollment, whereas 120 urinary (TIMP-2) × (IGFBP7) to detect any stage of had infection as a primary or secondary diagnosis but AKI (not just stage 2–3). We found that 24 patients without a clinical diagnosis of sepsis. The remainder (20%) tested positive, of which 18 (75%) developed (n = 387) had neither infection nor sepsis at enrollment. AKI (six were stage 1). e362     www.ccmjournal.org April 2021 • Volume 49 • Number 4 Online Clinical Investigations TABLE 1. Baseline Characteristics Infection Characteristics All Patients No Infection No Sepsis Sepsis pa n 723 387 120 216 Male, n (%) 444 (61.4) 256 (66.1) 74 (61.7) 114 (52.8) 0.005 Age, median (interquartile range) 64 (53–73) 64 (53–72) 61 (52–73) 66 (54–74) 0.207 Race, n (%) 0.006 White 568 (78.6) 312 (80.6) 92 (76.7) 164 (75.9) Black 87 (12) 50 (12.9) 18 (15) 19 (8.8) Other/unknown 68 (9.4) 25 (6.5) 10 (8.3) 33 (15.3) Medical history, n (%) Chronic kidney disease 65 (9) 41 (10.6) 9 (7.5) 15 (6.9) 0.288 Diabetes 210 (29) 116 (30) 29 (24.2) 65 (30.1) 0.445 Congestive heart failure 121 (17.3) 77 (20.6) 12 (10.3) 32 (15.3) 0.023 Coronary artery disease 215 (30.3) 146 (38.6) 24 (20.2) 45 (21.2) < 0.001 Hypertension 429 (59.3) 239 (61.8) 56 (46.7) 134 (62) 0.009 Chronic obstructive 146 (20.2) 81 (20.9) 27 (22.5) 38 (17.6) 0.482 pulmonary disease Cancer 187 (26.2) 88 (23) 30 (25.4) 69 (32.2) 0.050 Reason for ICU admission, n (%) Cardiovascular 239 (33.1) 154 (39.8) 33 (27.5) 52 (24.1) < 0.001 Cerebrovascular 70 (9.7) 46 (11.9) 13 (10.8) 11 (5.1) 0.017 Sepsis 135 (18.7) 0 (0) 0 (0) 135 (62.5) < 0.001 Respiratory 308 (42.6) 116 (30) 73 (60.8) 119 (55.1) < 0.001 Trauma 55 (7.6) 49 (12.7) 4 (3.3) 2 (0.9) < 0.001 Surgery 247 (34.2) 180 (46.5) 18 (15) 49 (22.7) < 0.001 Other 126 (17.4) 47 (12.1) 33 (27.5) 46 (21.3) < 0.001 Enrollment serum creatinine (mg/dL), 0.8 (0.7–1.1) 0.9 (0.7–1.1) 0.8 (0.6–1.1) 0.8 (0.6–1.1) 0.147 median (interquartile range) Nonrenal APACHE-III, 60 (43–82) 57 (41–80) 59 (42–73) 63 (48–87) 0.008 median (interquartile range) APACHE-III, 69 (51–91) 67 (50–90) 66 (51–82) 75 (55–96) 0.006 median (interquartile range) Nonrenal SOFA, 7 (5–10) 7 (5–10) 7 (4–9.25) 8 (5.75–10) 0.035 median (interquartile range) SOFA, median (interquartile range) 8 (5–10) 8 (5–10) 7 (4.75–10) 8 (6–11) 0.033 APACHE = Acute Physiology and Chronic Health Evaluation, SOFA = Sequential Organ Failure Assessment. a Comparisons were made across the three subgroups. Critical Care Medicine www.ccmjournal.org     e363 Kellum et al TABLE 2. Outcomes All Patients, No Infection, Infection No Sepsis, Outcomes n (%) n (%) Sepsis, n (%) n (%) pa n 723 387 120 216 Any AKI within 3 d 413 (57.1) 218 (56.3) 67 (55.8) 128 (59.3) 0.742 Stage 2–3 AKI within 3 d 175 (24.2) 84 (21.7) 32 (26.7) 59 (27.3) 0.229 Renal replacement therapy within 30 d 48 (6.6) 20 (5.2) 8 (6.7) 20 (9.3) 0.158 Death within 30 d 116 (16) 51 (13.2) 21 (17.5) 44 (20.4) 0.060 Major adverse kidney events at 30 d 155 (21.4) 68 (17.6) 27 (22.5) 60 (27.8) 0.014 AKI = acute kidney injury. a Comparisons were made across the three subgroups. Long-Term Survival by Biomarkers failure (1). AKI is one of the most common organ fail- and AKI Status ures and yet can be difficult to identify early. These dif- ficulties stem from both delayed clinical manifestations Finally, we examined mortality out to 9 months strat- (serum creatinine and urine output changes) as well ified by both urinary (TIMP-2) × (IGFBP7) and AKI. as from challenges to recognition (e.g., lack of baseline Figure 2A shows results for the 2.0 cutoff and stage creatinine, incomplete urine output data). Importantly, 2–3 AKI, and Figure 2B shows the 1.0 cutoff and stage AKI is strongly associated with sepsis mortality. In the 1–3 AKI. Higher mortality was seen for patients with Protocolized Care for Early Septic Shock (ProCESS) biomarker-positive AKI compared with biomarker- trial (14), hospital mortality truncated at 60 days was negative AKI in both analyses. The hazard ratio (95% 6.2% for patients without AKI, whereas it was 16.8% CI) for 9-month mortality was 2.32 (1.46–3.67) for for patients with maximum AKI stage 1, and 27.7% patients with stage 2–3 AKI and (TIMP-2) × (IGFBP7) for stage 2–3 AKI (p = 0.0001) (3). We observed sim- greater than 2 relative to patients with no or stage 1 ilar results here in that 30-day mortality was 27% in AKI and (TIMP-2) × (IGFBP7) less than or equal patients with sepsis and stage 2–3 AKI. Surprisingly, to 2. Similarly, the hazard ratio (95% CI) was 2.44 in patients with infection and no clinical diagnosis (1.68–3.54) for patients with any AKI and (TIMP-2) of sepsis, 30-day mortality was 34% for patients with × (IGFBP7) greater than 1 relative to patients with no- stage 2–3 AKI compared with 11% without stage AKI and (TIMP-2) × (IGFBP7) less than or equal to 1 2–3 AKI. Of course, these patients likely “had” sepsis (ng/mL)2/1,000. (infection plus AKI) even if it could not be identified at enrollment. DISCUSSION Many authors have lamented the difficulty asso- Despite no specific therapy, sepsis mortality rates have ciated with screening for sepsis among hospitalized steadily decreased in recent years (12). Presumably, this patients. Pooled results for 23 studies including more is due to better recognition and prompt therapy with than 150,000 non-ICU patients illustrate that quick antibiotics and reversal of shock. Delay of appropriate SOFA has poor sensitivity for sepsis mortality (51%) therapy, even by a few hours, can dramatically increase and organ failure (47%) (15). Numerous studies have mortality (13). As such many hospitals have instituted shown that urinary (TIMP-2) × (IGFBP7) is sensitive sepsis protocols that are designed to deliver broader and specific for AKI including for patients with sepsis spectrum antibiotics quickly, screen for shock (lactate), (5, 16–18). Many patients with infection are admitted and admit to the ICU for close monitoring for organ to hospital, and, in the absence of shock or respira- dysfunction. The current definition of sepsis, Sepsis-3, tory failure, few are admitted to ICU. However, AKI is pragmatic requiring infection with related organ can have significant consequences in such patients. e364     www.ccmjournal.org April 2021 • Volume 49 • Number 4 Online Clinical Investigations included broader spectrum, faster initiation of IV anti- biotics, greater emphasis on source control (e.g., more diagnostic imaging), more use of functional hemody- namic monitoring, etc. The downside to this approach is that for every four positive tests, one would be a false positive. These patients would receive more expensive care, including more ICU admissions and more invasive monitoring. They might also receive antibiotics generally reserved for patients with sepsis. These measures would increase costs and risks for these patients, even if only slightly. However, these risks need to be balanced against miss- ing sepsis in nearly one of 10 cases. Importantly, we can envision clinical implementa- tion of these findings in a number of settings. First, sepsis can be challenging to diagnose in the presence of organ failures that appear to be due to the primary disease or condition (e.g., shock in a patient following cardiac surgery or hypoxia in a patient with an exac- erbation of chronic lung disease). Thus, patients with infection who are in the ICU for another reason are at particularly high risk for sepsis, and screening for AKI might be very useful. Patients like this are rep- resented in our study. Second, patients being admit- ted to the hospital outside the ICU are not as closely monitored. Kidney function in particular is not well tracked outside the ICU where Foley catheters are actively discouraged, and serum creatinine is rarely measured more than once daily. In these patients, measures of AKI biomarkers might also be helpful. However, we did not include these types of patients in Figure 1. Thirty-day mortality by acute kidney injury (AKI) status. our study, and because the risk for AKI will be lower, A, Shows mortality for stage 2–3 AKI versus no-AKI, and stage false-positive rates will increase. Thus, we recommend 1 stratified by no infection, infection only, and sepsis. B, Shows further study in patients outside the ICU or similar mortality for stage 1–3 AKI versus no-AKI stratified by the same high-risk groups. A third category of patients might three patient groups. *p < 0.05. be those being evaluated in the emergency depart- ment with infection and under consideration for ad- Murugan et al (19) found that AKI occurred in a third mission. Diagnosing organ failure in this population of patients admitted to hospital with community- would significantly alter the care plan. However, this acquired pneumonia (only 16% to the ICU), and yet, group might include very low-risk patients, and thus mortality at 1 year was three-fold higher in patients care needs to be taken to select patients for testing. with stage 2–3 AKI. Our patients in this study were all A recent study in patients not restricted to infection admitted to ICU, and yet, an additional 8% of patients (20) used three criteria to assess high risk for AKI: with infection could be diagnosed with sepsis without 1) critical illness, 2) two or more systemic inflamma- waiting for AKI to manifest by changes in urine output tory response syndrome criteria (21), or 3) triaged for or serum creatinine. These patients could therefore immediate or highly urgent treatment based on the have been treated for sepsis much earlier, potentially manchester triage system (22). Using these criteria, improving outcome. These treatments could have the event rate for stage 2–3 AKI among patients with Critical Care Medicine www.ccmjournal.org     e365 Kellum et al urinary (TIMP-2) × (IGFBP7) greater than 0.3 was more than 30%. Finally, across our entire cohort, we found important differences in 9-month survival when patients with AKI tested positive with urinary (TIMP-2) × (IGFBP7) compared with when AKI occurred without a positive biomarker. Both patients with stage 1 with (TIMP-2) × (IGFBP7) greater than 1 and stage 2–3 with (TIMP-2) × (IGFBP7) greater than 2.0 exhibited worse survival compared with biomarker-negative AKI at each re- spective stage. These results are consistent with those reported recently by Joannidis et al (23) in which biomarker-positive stage 1 AKI behaved like stage 2–3 AKI in terms of outcomes such as death, dialysis, and per- sistent stage 3. The primary study was not designed to test the hypotheses we sought to address with this secondary analysis. As such, there are important limitations. For example, patients in the “infection only” group may have developed AKI for reasons other than sepsis—such as from a nephrotoxic drug. Although this is possible, it seems unlikely given the strong association with mortality. Second, we do not know when or if patients in the infection group were diagnosed with sepsis, only that it was not recorded in the medical record at the time of enrollment. The clinical team may have made the diag- nosis based on other criteria, or they may never have diagnosed it. Similarly, we did not collect detailed information on the care Figure 2. Mortality curves out to 9 mo stratified by the both urinary tissue these patients received such as the use of inhibitor of metalloproteinases-2 (TIMP-2) × insulin-like growth factor binding functional hemodynamic monitoring, so protein 7 (IGFBP7) and acute kidney injury (AKI). A, Shows results for the 2.0 we cannot comment on whether care would cutoff and stage 2–3 AKI. Using no-AKI or stage 1 and (TIMP-2) × (IGFBP7) less than or equal to 2 as reference, the individual hazard ratios (95% CI) for have changed if the diagnosis of sepsis was each group were as follows: 1.27 (0.64–2.49), p = 0.49 for no-AKI/stage 1 and made sooner. Nevertheless, we can say that (TIMP-2) × (IGFBP7) greater than 2; 1.61 (1.16–2.24), p = 0.005 for stage 2–3 for these patients, the biomarker correctly AKI and (TIMP-2) × (IGFBP7) less than or equal to 2; and 2.32 (1.46–3.67), identified AKI several hours before it could p < 0.001 stage 2–3 AKI and (TIMP-2) × (IGFBP7) greater than 2 (ng/mL)2/ be diagnosed using standard criteria. 1,000. B, Shows the 1.0 cutoff and stage 1–3 AKI. Using no-AKI and (TIMP- In conclusion, urinary (TIMP-2) × (IGFBP7) 2) × (IGFBP7) less than or equal to 1 as reference, the individual hazard ratios could identify AKI as a sepsis-defining organ (95% CI) for each group were as follows: 0.56 (0.20–1.54), p = 0.26 for no-AKI and (TIMP-2) × (IGFBP7) greater than 1; 1.55 (1.12–2.14), p = 0.008 for AKI failure in patients with infection several hours and (TIMP-2) × (IGFBP7) less than or equal to 1; and 2.44 (1.68–3.54), before clinical diagnosis with less than 30% of p < 0.001 for AKI and (TIMP-2) × (IGFBP7) greater than 1 (ng/mL)2/1,000. positive test results being false positives. e366     www.ccmjournal.org April 2021 • Volume 49 • Number 4 Online Clinical Investigations 1 Department of Critical Care Medicine, Center for Critical he received funding from Grifols, Fisher & Paykel. Drs. Kampf, Care Nephrology, University of Pittsburgh School of Kwan, and McPherson received funding from Astute Medical Medicine, Pittsburgh, PA. (employee). Drs. Nguyen’s and Shapiro’s institutions received 2 Critical Care Center, Corporacion Sanitaria Parc Tauli, funding from Astute Medical. The remaining authors have dis- CIBER Enfermedades Respiratorias, Autonomous University closed that they do not have any potential conflicts of interest. of Barcelona, Sabadell, Spain. For information regarding this article, E-mail: kellum@pitt.edu 3 Departments of Emergency Medicine/Critical Care, Anesthesiology, Internal Medicine, Michigan Center for Integrative Research in Critical Care (MCIRCC), University of Michigan, Ann Arbor, MI. REFERENCES 4 Department of Intensive Care Medicine, Brugmann 1. Seymour CW, Liu VX, Iwashyna TJ, et al: Assessment of clinical University Hospital, Brussels, Belgium. criteria for sepsis: For the third international consensus defi- 5 Astute Medical, Inc. (a bioMérieux company), San Diego, nitions for sepsis and septic shock (Sepsis-3). JAMA 2016; CA. 315:762–774 6 Division of Pulmonary and Critical Care Medicine, 2. Vincent JL, Sakr Y, Sprung CL, et al; Sepsis Occurrence in Department of Medicine, Loma Linda University, Loma Acutely Ill Patients Investigators: Sepsis in European intensive Linda, CA. care units: Results of the SOAP study. Crit Care Med 2006; 34:344–353 7 Department of Anesthesiology and Critical Care Medicine, 3. Kellum JA, Chawla LS, Keener C, et al; ProCESS and Edouard Herriot Hospital, Hospices Civils de Lyon, Lyon, ProGReSS-AKI Investigators: The effects of alternative resus- France. citation strategies on acute kidney injury in patients with septic 8 Department of Emergency Medicine, Beth Israel Deaconess shock. Am J Respir Crit Care Med 2016; 193:281–287 Medical Center, Boston, MA. 4. Bihorac A, Chawla LS, Shaw AD, et al: Validation of cell-cycle 9 Walker BioSciences, Carlsbad, CA. arrest biomarkers for acute kidney injury using clinical adjudi- 10 Department of Intensive Care, Erasme University Hospital, cation. Am J Respir Crit Care Med 2014; 189:932–939 Université Libre de Bruxelles, Brussels, Belgium. 5. Honore PM, Nguyen HB, Gong M, et al; Sapphire and Topaz 11 Department of Medicine, Veterans Affairs Medical Center, Investigators: Urinary tissue inhibitor of metalloproteinase-2 San Diego, CA. and insulin-like growth factor-binding protein 7 for risk strati- fication of acute kidney injury in patients with sepsis. Crit Care A complete list of Sapphire Investigators is provided in the Med 2016; 44:1851–1860 Appendix (Supplemental Digital Content 1, http://links.lww. 6. Kashani K, Al-Khafaji A, Ardiles T, et al: Discovery and vali- com/CCM/G111). dation of cell cycle arrest biomarkers in human acute kidney Clinical Trial Registration: NCT01209169. injury. Crit Care 2013; 17:R25 The parent study was designed by Drs. Kellum, Kampf, 7. Kidney Disease: Improving Global Outcomes (KDIGO) Work McPherson and Chawla and was conducted by Drs. Artigas, Group: KDIGO clinical practice guildeline for acute kidney in- Gunnerson, Honore, Nguyen, Rimmel,o Shapiro, and Vincent. Dr. jury. Kidney Int Suppl 2012; 2:1–138 Kellum conceived of the current study and drafted the article. Dr. 8. von Elm E, Altman DG, Egger M, et al; STROBE Initiative: Shi performed the statistical analysis. Drs. Kellum, Kampf, Kwan, The strengthening the reporting of observational studies and McPherson interpreted the results, and all authors reviewed in epidemiology (STROBE) statement: Guidelines for re- and revised the article. Dr. Kellum takes responsibility for the con- porting observational studies. Ann Intern Med 2007; 147: tent of the article, including the data and analysis. 573–577 Supplemental digital content is available for this article. Direct 9. Bossuyt PM, Reitsma JB, Bruns DE, et al; STARD Group: URL citations appear in the printed text and are provided in the STARD 2015: An updated list of essential items for reporting HTML and PDF versions of this article on the journal’s website diagnostic accuracy studies. BMJ 2015; 351:h5527 (http://journals.lww.com/ccmjournal). 10. Knaus WA, Wagner DP, Draper EA, et al: The APACHE III The study sponsor (Astute Medical) was primarily responsible for prognostic system. Risk prediction of hospital mortality for crit- design of the parent study and data collection; the current study ically ill hospitalized adults. Chest 1991; 100:1619–1636 was designed and interpreted by the authors. 11. Vincent JL, Moreno R, Takala J, et al: The SOFA (sepsis-related Dr. Kellum’s institution received funding from Astute Medical organ failure assessment) score to describe organ dysfunc- and bioMerieux. Drs. Kellum, Rimmelé, Shi, and Chawla re- tion/failure. On behalf of the working group on sepsis-related ceived funding from Astute Medical. Drs. Kampf, Kwan, and problems of the European society of intensive care medicine. McPherson are employees of Astute Medical/bioMerieux; Drs. Intensive Care Med 1996; 22:707–710 Kellum, Rimmelé, Shi, and Chawla are paid consultants of Astute 12. Landoni G, Bove T, Székely A, et al: Reducing mortality in Medical/bioMerieux; and Dr. Kellum has received grant support acute kidney injury patients: Systematic review and interna- from Astute Medical/ bioMerieux apart from the current work. Dr. tional web-based survey. J Cardiothorac Vasc Anesth 2013; Artigas’s institution received funding from Lilly Foundation, and 27:1384–1398 Critical Care Medicine www.ccmjournal.org     e367 Kellum et al 13. Kumar A, Roberts D, Wood KE, et al: Duration of hypotension short-term outcomes in septic and non-septic critically ill before initiation of effective antimicrobial therapy is the critical patients. Ann Intensive Care 2020; 10:46 determinant of survival in human septic shock. Crit Care Med 19. Murugan R, Karajala-Subramanyam V, Lee M, et al; Genetic 2006; 34:1589–1596 and Inflammatory Markers of Sepsis (GenIMS) Investigators: 14. Yealy DM, Kellum JA, Huang DT, et al: A randomized trial of Acute kidney injury in non-severe pneumonia is associated protocol-based care for early septic shock. N Engl J Med with an increased immune response and lower survival. Kidney 2014; 370:1683–1693 Int 2010; 77:527–535 15. Song JU, Sin CK, Park HK, et al: Performance of the quick se- 20. Schanz M, Wasser C, Allgaeuer S, et al: Urinary [TIMP- quential (sepsis-related) organ failure assessment score as a 2]·[IGFBP7]-guided randomized controlled intervention trial prognostic tool in infected patients outside the intensive care to prevent acute kidney injury in the emergency department. unit: A systematic review and meta-analysis. Crit Care 2018; Nephrol Dial Transplant 2019; 34:1902–1909 22:28 21. Bone RC, Balk RA, Cerra FB, et al: Definitions for sepsis and 16. Maizel J, Daubin D, Vong LV, et al: Urinary TIMP2 and IGFBP7 organ failure and guidelines for the use of innovative therapies identifies high risk patients of short-term progression from in sepsis. The ACCP/SCCM consensus conference commit- mild and moderate to severe acute kidney injury during tee. American College of chest physicians/society of critical septic shock: A prospective cohort study. Dis Markers 2019; care medicine. Chest 1992; 101:1644–1655 2019:3471215 22. Mackway-Jones K, Marsden J, Windle J: Emergency Triage. 17. Cuartero M, Ballús J, Sabater J, et al: Cell-cycle arrest bio- Hoboken, NJ, John Wiley & Sons, 2014 markers in urine to predict acute kidney injury in septic and 23. Joannidis M, Forni LG, Haase M, et al; Sapphire Investigators: non-septic critically ill patients. Ann Intensive Care 2017; 7:92 Use of cell cycle arrest biomarkers in conjunction with clas- 18. Godi I, De Rosa S, Martino F, et al: Urinary [TIMP-2] × [IGFBP7] sical markers of acute kidney injury. Crit Care Med 2019; and serum procalcitonin to predict and assess the risk for 47:e820–e826 e368     www.ccmjournal.org April 2021 • Volume 49 • Number 4