ORIGINAL CLINICAL REPORT Serial Urinary C-C Motif Chemokine Ligand 14 and Risk of Persistent Severe Acute Kidney Injury John R. Prowle, MD1 OBJECTIVES: To assess the added prognostic value of serial monitoring of uri- Antonio Artigas, MD2 nary C-C motif chemokine ligand 14 (uCCL14) over that of single measurements, Sean M. Bagshaw, MD3 which have been shown to be prognostic for development of persistent severe acute kidney injury (AKI) in critically ill patients. Lui G. Forni, PhD4 Michael Heung, MD5 DESIGN: Retrospective observational study. Eric Hoste, PhD6 SETTING: Data derived from two multinational ICU studies (Ruby and Sapphire). Ostermann Marlies, PhD7 PATIENTS: Critically ill patients with early stage 2–3 AKI. Jay L. Koyner, MD8 INTERVENTIONS: None. Lakmir Chawla, MD9 MEASUREMENTS AND MAIN RESULTS: We analyzed three consecutive J. Patrick Kampf, PhD10 uCCL14 measurements at 12-hour intervals after diagnosis of stage 2–3 AKI by Thomas Kwan, PhD10 Kidney Disease Improving Global Outcomes criteria. Primary outcome was persis- Paul McPherson, PhD10 tent severe AKI, defined as 72 consecutive hours of stage 3 AKI, death, or receipt John A. Kellum, MD11 of dialysis prior to 72 hours. uCCL14 was measured using the NEPHROCLEAR uCCL14 Test on the Astute 140 Meter (Astute Medical, San Diego, CA). Based for the Sapphire and Ruby on predefined, validated cutoffs, we categorized uCCL14 as: low (≤ 1.3 ng/mL), Investigators medium (> 1.3 to ≤ 13 ng/mL), or high (> 13 ng/mL). Seventy-five of 417 patients with three consecutive uCCL14 measurements developed persistent severe AKI. Initial uCCL14 category strongly correlated with primary endpoint and, in most cases (66%), uCCL14 category was unchanged over the first 24 hours. Compared with no change and accounting for baseline category, decrease in cat- egory was associated with decreased odds of persistent severe AKI (odds ratio [OR], 0.20; 95% CI, 0.08–0.45; p < 0.001) and an increase in category with increased odds (OR, 4.04; 95% CI, 1.75–9.46; p = 0.001). CONCLUSIONS: In one-third of patients with moderate to severe AKI uCCL14 risk category altered over three serial measurements and such changes were as- sociated with altered risk for persistent severe AKI. Serial CCL-14 measurement may detect progression or resolution of underlying kidney pathology and help refine AKI prognosis. Copyright © 2023 The Authors. KEY WORDS: acute kidney injury; biomarkers; C-C motif chemokine ligand 14; Published by Wolters Kluwer Health, critical illness; prognosis Inc. on behalf of the Society of Critical Care Medicine. This is an A open-access article distributed under cute kidney injury (AKI) affects more than one in four hospitalized the terms of the Creative Commons patients and one in two admitted to intensive care (1). In the short Attribution-Non Commercial-No Derivatives License 4.0 (CCBY- term, AKI is associated with increased risk of death, length, and cost NC-ND), where it is permissible to of hospitalization and, in survivors, greater longer-term morbidity and mor- download and share the work pro- tality, often associated with the development of chronic kidney disease (CKD) vided it is properly cited. The work (2–5). In particular, persistent severe AKI (PS-AKI), defined as Kidney Disease cannot be changed in any way or Improving Global Outcomes (6) (KDIGO) AKI stage 3, lasting 3 or more days, used commercially without permis- has been strongly associated with adverse outcomes across a range of clinical sion from the journal. settings (7–9). However, predicting the clinical course of AKI is difficult, in part DOI: 10.1097/CCE.0000000000000870 Critical Care Explorations www.ccejournal.org     1 Prowle et al repair, serial measurement of uCCL14 could be infor- mative about clinical course of AKI. Accordingly, in a KEY POINTS preplanned analysis of two observational studies, we sought to characterize changes in categorized uCCL14 Question: Does serial measurement of urinary values over time and their association with clinical C-C motif chemokine ligand 14 (uCCL14) pro- outcome. vide complementary prognostic information over a single measurement early after the diagnosis of stage 2–3 acute kidney injury (AKI) complicating METHODS critical illness. Research Background Findings: In one-third of patients with moderate to severe AKI uCCL14 risk category altered over We conducted a secondary analysis of serial meas- three serial measurements and such changes urement of uCCL14 early after development of mod- were associated with altered risk for persistent se- erate to severe AKI in a mixed population of critically vere AKI. ill adults. The analytic cohort is composed of pooled Meaning: Serial C-C motif chemokine ligand 14 patients from the previously reported Ruby and measurement may reveal dynamic pathology and Sapphire studies (10, 14). The Ruby study (Identification help refine prognosis. and Validation of Biomarkers of Acute Kidney Injury Recovery) and the Sapphire study (Evaluation of Novel Biomarkers From Acutely Ill Patients at Risk for Acute Kidney Injury) were approved by the Western because changes in serum creatinine may significantly Institutional Review Board (Puyallup, WA). Sapphire lag changes in underlying glomerular filtration rate (Approval Number: 20101176; Date: August 27, 2010) while urine output may be well preserved even in the and Ruby (Approval Number: 20130523; Date: April 8, face of significant tubular injury. Thus, a patient with 2013). Study procedures followed the ethical standards newly detected stage 2 AKI could already be recover- of the Review Board and the Helsinki Declaration of ing, could have an intermediate level of kidney dam- 1975. All patients or their legally authorized repre- age which may respond to targeted intervention or sentatives provided written informed consent prior to may already have severe parenchymal damage that will enrollment. progress toward need for renal replacement therapy (RRT). Accordingly, better tools for identifying those Subjects at highest risk for PS-AKI are required to enable pre- dictive enrichment of populations for AKI-targeted The Ruby study, we enrolled critically ill adults with intervention. established KDIGO stage 2–3 AKI across 21 sites in Previously the Ruby study (10) described urinary Europe and the United States. We excluded patients C-C motif chemokine ligand 14 (uCCL14) as a novel with a prior kidney transplant, those receiving or in im- predictor of future PS-AKI, in patients shortly after minent need of RRT, those receiving comfort measures first diagnosis of stage 2–3 AKI. The ability of uCCL14 only, and those with known HIV infection or active to forecast PS-AKI has subsequently been verified in viral hepatitis. The Sapphire study enrolled critically ill two independent cohorts (11, 12). More recently in adults with cardiac or respiratory dysfunction without a preplanned secondary analysis, two cutoff values of known stage 2–3 AKI at the time of enrollment from uCCL14 have been described identifying groups at low, 35 sites in Europe and North America. Samples from moderate, and higher risk of PS-AKI (13). However, a subset of Sapphire patients who developed stage 2–3 previous analyses have been based only on the initial AKI within 1-week of enrollment were included in this values of uCCL14, close to first diagnosis of moderate present analysis. For this purpose, in the Sapphire co- to severe AKI and the relationship of serial measures of hort, serial biomarker measurements were included uCCL14 to the development of PS-AKI have not been from the point at which patients met stage 2–3 AKI described. We hypothesized that as a biomarker of un- criteria, mirroring the time course of enrollment and derlying kidney inflammation during tissue injury and biomarker measurement in the Ruby. We conducted a 2     www.ccejournal.org March 2023 • Volume 5 • Number 3 Original Clinical Report complete case analysis, where in both studies, patients Statistical Analysis without uCCL14 concentrations for all three samples or who received RRT or died before the third sample We stratified the uCCL14 concentrations in the first was taken were excluded. Patients were recruited based three urine collections into three levels: low (≤ 1.3 ng/ on AKI assessment using creatinine and urine output mL), medium (> 1.3 to ≤ 13 ng/mL), and high (> 13 ng/ criteria by local site investigators and all patients mL), using previously validated clinical risk cutoffs enrolled were included on an “intention to diagnose” (13). The high sensitivity cutoff at 1.3 ng/mL uCCL14 basis (10, 14). was derived to identify the majority of patients at risk of PS-AKI and the high specificity cutoff at 13 ng/mL to identify the highest risk patients where PS-AKI is Sample Collection and Testing most likely. To assess the validity of using a pooled In the Ruby study, urine samples were collected bid for dataset, we applied the Cochran Q test to assess het- 3 days from enrollment and then once daily for 4 days. erogeneity of relative risk (RR) between individual In the Sapphire study, urine samples were collected bid time points across Ruby and Sapphire. Patients in the for 4 days from enrollment and then once daily for 3 pooled analysis cohort were grouped by the pattern days. In both studies, urine samples were centrifuged, of uCCL14 levels across the three samples; these were flash frozen, stored at or below –70°C, and thawed prior then compared based on proportion of patients pos- to sample testing. The primary analysis used uCCL14 itive for PS-AKI within each group. Continuous and concentrations from the first three scheduled urine categorical variables were analyzed using the Wilcoxon collections in Ruby and the first three urine collections rank-sum and Fisher exact tests, respectively. Two- after onset of KDIGO stage 2 or 3 AKI in Sapphire sided p values of less than 0.05 were considered statis- for each patient. Technicians who were blinded to the tically significant. All analyses were performed using R clinical data measured uCCL14 concentrations in the 4.0.2 (R Foundation for Statistical Computing. Vienna, samples using the NEPHROCLEAR uCCL14 Test on Austria). the Astute 140 Meter (Astute Medical, San Diego, CA). Sensitivity Analyses Clinical Endpoint In addition to primary complete-case analysis, we per- The primary endpoint for the present analysis was the formed a sensitivity analysis including all individuals development of PS-AKI, as described (10, 12). In brief, with one or more uCCL14 measurements after enroll- patients who developed 72 consecutive hours of stage ment to assess consistency of our findings in those with 3 AKI, commencing within 48 hours of first sample missing values for clinical reasons including discharge, collection, or those who died following stage 3 AKI or death, and anuria. Finally, to test uCCL14 in the con- received RRT prior to 48 hours from first sample col- text of expected clinical use, our primary analysis uti- lection or within 72 consecutive hours of stage 3 AKI, lized previously derived clinical cutoffs to categorize were considered endpoint positive. For the purposes uCCL14 levels and changes. As an alternative, we per- of determining the primary endpoint, in Ruby, each formed a category-free analysis of absolute uCCL14 patient’s baseline serum creatinine was retrospectively levels and any trend in the first 24 hours. adjudicated by a physician blinded to biomarker data based on serum creatinine data collected from the Extended Measurement of uCCL14 medical record from 6 months prior to 3 months after enrollment (9). In Sapphire, the reference values for CCL14 measurement in urine samples collected at least serum creatinine were obtained as follows: if at least daily allowed assessment of change in category of se- five values were available, the median of all values rial uCCL14 measurements over a longer time and its available from 6 months to 6 days prior to enrollment relationship to the occurrence of PS-AKI on day 7. We was used. Otherwise, the lowest value in the 5 days then performed rolling assessment of the PS-AKI com- prior to enrollment was used. If no pre-enrollment cre- posite endpoint over time, that is, at each time point of atinine was available, the creatinine value at the time of measurement, the occurrence of 72 hours of sustained enrollment was used (10, 14). stage 3 AKI commencing within 48 hours or death or Critical Care Explorations www.ccejournal.org     3 Prowle et al RRT with 48 hours of that measurement was assessed. the primary endpoint of PS-AKI, while in the sensitivity For this exploratory analysis, we treated uCCL14 con- analysis, 135 (26%) reached the endpoint, often reflecting centration as a continuous variable and explored dis- the effect of early death or anuria preventing collection of crimination of PS-AKI over time. Ability of uCCL14 a full set of study samples. Of the 111 patients in the sen- concentration at any time point to discriminate the sitivity analysis but not the analysis cohort, 33 (30%) were rolling endpoint was assessed as area under the receiver receiving RRT, 18 (16%) were anuric, and 8 (7%) died. operating characteristic curve (ROC-AUC) with 95% In the Sapphire cohort, only 7.4% of patients reached the CI determined by the DeLong method (15). primary endpoint compared with 24% in the Ruby study (Table 1), reflecting lower overall severity of AKI in the RESULTS Sapphire cohort, AKI stage 3 at time of first sample collec- tion 4% versus 30%. Conversely, the Sapphire cohort had Patients higher nonrenal Acute Physiology and Chronic Health The Ruby study recruited 364 patients within 36 hours Evaluation II score (62 vs 52) and use of mechanical ven- of clinical diagnosis of stage 2–3 AKI, while of the 723 tilation (85 vs 54%) (Tables S1 and S2, http://links.lww. patients critically ill patients without AKI stage 2 or com/CCX/B147). In the pooled primary dataset, the greater in SAPPPHIRE analysis cohort, 212 then devel- median age was 65 years and 59% were male (Table 1). oped stage 2–3 AKI and were eligible for inclusion in this Baseline kidney function, history of CKD, demographics, pooled analysis from that time point. After further exclu- illness severity, and organ support did not significantly sion of patients without three consecutive measurements differ between patients with PS-AKI endpoint and those of uCCL14 after meeting enrollment criteria, a total of without. However, fluid balance on the first day after en- 417 (268 patients from the Ruby study and 149 patients rollment was significantly more positive in those who from Sapphire) were included in our complete-case anal- developed PS-AKI (2.3 vs 3.4 L; p = 0.04). Importantly, de- ysis (Fig. 1). In addition, a further 111 patients had one spite the differences between the studies, the RRs analysis or two valid uCCL14 measurements within 36 hours of showed that they behaved similarly across the uCCL14 enrollment and were included in a sensitivity analysis of strata at each time point, justifying the pooled cohort 528 patients. Of the pooled population, 75 (18%) reached analysis (Table S3, http://links.lww.com/CCX/B147). Figure 1. Patient flow diagram. AKI = acute kidney injury, CCL14 = C-C motif chemokine ligand 14, RRT = renal replacement therapy. 4     www.ccejournal.org March 2023 • Volume 5 • Number 3 Original Clinical Report TABLE 1. Baseline Characteristics of Analysis Cohort PS-AKI Endpoint PS-AKI Endpoint Characteristics All Patients Negative Positive p n 417 342 75 Male, n (%) 248 (59) 201 (59) 47 (63) 0.604 Agea (yr) 65 (55–73) 65 (55–74) 64 (56–69) 0.107 Body mass index (kg/m ) a 2 30 (26–36) 31 (26–37) 28 (25–34) 0.138 Race, n (%)  Black or African American 50 (12) 43 (13) 7 (9) 0.615  Other/unknown 31 (7) 24 (7) 7 (9)  White or Caucasian 336 (81) 275 (80) 61 (81) Chronic comorbidities, n (%)  Chronic kidney disease 58 (14) 42 (12) 16 (21) 0.063  Diabetes mellitus 155 (37) 128 (37) 27 (36) 0.895  Heart failure 99 (24) 82 (24) 17 (23) 0.882  Coronary artery disease 144 (35) 121 (35) 23 (31) 0.503  Hypertension 289 (69) 239 (70) 50 (67) 0.583  Chronic obstructive pulmonary disease 81 (19) 66 (19) 15 (20) 0.873  Cancer 108 (26) 89 (26) 19 (25) 1.000 Reason for ICU admission, n (%)  Respiratory 150 (36) 121 (35) 29 (39) 0.597  Surgery 136 (33) 117 (34) 19 (25) 0.173  Cardiovascular 176 (42) 136 (40) 40 (53) 0.039  Sepsis 98 (24) 77 (23) 21 (28) 0.367  Neurologic 27 (6) 24 (7) 3 (4) 0.443  Trauma 11 (3) 11 (3) 0 (0) 0.226  Other 107 (26) 89 (26) 18 (24) 0.772 Baseline serum creatinine (mg/dL) a 1.0 (0.7–1.2) 1.0 (0.7–1.2) 1.0 (0.8–1.4) 0.137 Baseline estimated glomerular filtration rate 76 (54–95) 76 (55–95) 76 (53–96) 0.319 Nonrenal Acute Physiology and Chronic 55 (43–74) 56 (42–73) 55 (45–79) 0.379 Health Evaluation III score at enrollment Nonrenal Sequential Organ Failure 6 (4–9) 6 (4–9) 6 (5–10) 0.204 Assessment score on study day 1 Time difference (in hr) between second and 11 (9–14) 11 (9–14) 10 (6–13) 0.011 first samplesa Time difference (in hr) between third and 24 (21–26) 24 (22–26) 22 (19–25) 0.005 first samplesa Time difference (in hr) between third and 12 (11–13) 12 (11–13) 12 (11–13) 0.564 second samplesa Fluid balancea (mL) on study day 1 2,426 (745–5,067) 2,300 (623–4,799) 3,373 (1,219–6,552) 0.040 Diuretic use on study day 1, n (%) 162 (39) 131 (38) 31 (41) 0.695 Vasopressor use on study day 1, n (%) 199 (48) 165 (48) 34 (45) 0.702 (Continued) Critical Care Explorations www.ccejournal.org     5 Prowle et al TABLE 1. (Continued) Baseline Characteristics of Analysis Cohort PS-AKI Endpoint PS-AKI Endpoint Characteristics All Patients Negative Positive p Mechanical ventilation use on study day 1, 273 (65) 228 (67) 45 (60) 0.285 n (%) Kidney Disease Improving Global Outcomes stage at first sample collectionb, n (%)  No acute kidney injury 51 (12) 51 (15) 0 (0) < 0.001  Stage 1 49 (12) 49 (14) 0 (0)  Stage 2 231 (55) 204 (60) 27 (36)  Stage 3 86 (21) 38 (11) 48 (64) RRT within 7 d, n (%) 43 (10) 14 (4) 29 (39) < 0.001 Death within 7 d, n (%) 29 (7) 19 (6) 10 (13) 0.024 RRT or death within 7 d, n (%) 69 (17) 31 (9) 38 (51) < 0.001 Ruby patients, n (%) 268 (64) 204 (60) 64 (85) < 0.001 PS-AKI = persistent severe acute kidney injury, RRT = renal replacement therapy. a Median (interquartile range). b Kidney Disease Improving Global Outcomes stage after retrospective readjudication of acute kidney injury (AKI) status. Study PS-AKI endpoint was development of persistent stage 3 AKI lasting at least 72 hr commencing within 48 hr of first sample collection, or those who died following stage 3 AKI or those who received RRT prior to 48 hr from first sample collection or within 72 consecutive hr of stage 3 AKI. Initial uCCL14 levels were low (≤ 1.3 ng/mL) in 196 assessed the additional information provided by changes (47%), medium (> 1.3 to ≤ 13 ng/mL) in 180 (43%), in uCCL14 level, accounting for the initial value in lo- and high (> 13 ng/mL) in 41 patients (9.8%) (Table 2). gistic regression. Accounting for initial uCCL14 cate- As previous described (13), the initial uCCL14 category gory, a decrease in uCCL14 category in the first 24 hours was strongly associated with development of PS-AKI was associated with decreased risk of PS-AKI (odds ratio (Table 2). Importantly, the association between uCCL14 [OR], 0.2; 95% CI, 0.08–0.45; p < 0.001). An increase in category and PS-AKI outcome was essentially unchanged category was associated with increased risk (OR, 4.04; at the 12 and 24 hours time points (Table 2). In line with 95% CI, 1.75–9.46; p = 0.001). When tested, there was no this observation, in the majority of cases (66%) uCCL14 significant interaction between initial status and change category was unchanged from 0- to 24-hour time points (p = 0.36 for interaction term) (Table 3). and in only one patient did a two-level change occur (Fig. 2; and Table S4, http://links.lww.com/CCX/B147). Sensitivity Analysis General consistency of uCCL14 values over the first three measurements is visually demonstrated in Figure Change in uCCL14 remained statistically significant S1 (http://links.lww.com/CCX/B147). Where changes when considered as a continuous variable in logistic occurred, they were generally consistent in direction regression including initial uCCL14 as a continuous with only 24 of 417 (6%) showing a fluctuating course in log-transformed variable with comparable discrimina- uCCL14 category (Table S4, http://links.lww.com/CCX/ tion compared with our model based on category and B147). In patients experiencing a change in uCCL14 cat- change in category supporting the additional informa- egory, we classified patients as decreasing or increasing tion provided by change in uCCL14 over serial meas- uCCL14 category between 0 and 24 hours and stratified urement (Table S5, http://links.lww.com/CCX/B147). these by the initial uCCL14 levels. Across all initial cate- In sensitivity analysis including patients with miss- gories, a change in uCCL14 category was associated with ing uCCL14 values, the association between change a corresponding change in risk of PS-AKI (Fig. 2). We in uCCL14 stratified by baseline was essentially 6     www.ccejournal.org March 2023 • Volume 5 • Number 3 Original Clinical Report TABLE 2. C-C Motif Chemokine Ligand 14 Concentrations at Three Time Points in the First 24 Hours, Categorized by Predefined C-C Motif Chemokine Ligand 14 Cutoff and the Study Persistent Severe Acute Kidney Injury Endpoint PS-AKI Endpoint PS-AKI Endpoint Sampling Time Points All Patients Negative, n = 342 Positive, n = 75 p First sample (0 hr)  Median of all results (IQR) 1.44 (0.60–4.53) 1.12 (0.52–2.79) 7.29 (4.03–24.49) < 0.001 ≤ 1.3 ng/mL, n (%) 196 (47) 186 (54) 10 (13) < 0.001  > 1.3 and ≤ 13 ng/mL, n (%) 180 (43) 143 (42) 37 (49)  > 13 ng/mL, n (%) 41 (10) 13 (4) 28 (37) Second sample (12 hr)  Median of all results (IQR) 1.05 (0.47–3.94) 0.82 (0.40–2.30) 5.97 (3.10–21.38) < 0.001 ≤ 1.3 ng/mL, n (%) 228 (55) 221 (65) 7 (9) < 0.001  > 1.3 and ≤ 13 ng/mL, n (%) 148 (35) 106 (31) 42 (56)  > 13 ng/mL, n (%) 41 (10) 15 (4) 26 (35) Third sample (24 hr)  Median of all results (IQR) 0.99 (0.44–3.12) 0.78 (0.38–1.82) 7.85 (2.31–19.64) < 0.001 ≤ 1.3 ng/mL, n (%) 231 (55) 221 (65) 10 (13) < 0.001  > 1.3 and ≤ 13 ng/mL, n (%) 143 (34) 107 (31) 36 (48)  > 13 ng/mL, n (%) 43 (10) 14 (4) 29 (39) IQR = interquartile range, PS-AKI = persistent severe acute kidney injury. unchanged apart from overall worse outcomes for “un- (Fig. 3). Overall, 75% of those who had died or received changed” values reflecting inclusion of patients where RRT by day 7 were in the medium or high category of measurement was truncated by death or need for RRT uCCL14 at their last available measurement (Fig. 3). (Fig. S2, http://links.lww.com/CCX/B147), with a Rolling assessment of ability to predict the PS-AKI similar pattern of association in the logistic regression composite endpoint commencing within 48 hours (Table S6, http://links.lww.com/CCX/B147). demonstrated sustained ability to well-discriminate the endpoint with ROC-AUC greater than 0.8 through 7 Extended Measurement of uCCL14 days (Fig. S4, http://links.lww.com/CCX/B147). uCCL14 measurements were conducted at least daily DISCUSSION up to 7 days after enrollment in surviving patients re- maining in ICU with urinary catheter and without In the current study, we demonstrate important clin- RRT. At 36 hours, 355 patients of the original 417 had ical features of the previously described and vali- measurements available, which progressively declined dated uCCL14 assay that will both guide clinicians to 148 by day 7. Temporal changes in uCCL14 cate- in its interpretation and provide additional confi- gorization and outcomes are presented in Figure 3. dence to its interpretation in real-world clinical set- Compared with the original value over time, uCCL14 tings. Importantly, we used a standardized uCCL14 values tended to decrease more than increase (Fig. 3 assay that has been developed and approved for rou- and Fig. S3, http://links.lww.com/CCX/B147) and 67% tine clinical use in Europe (16). We demonstrate that of patients were in the low category of uCCL14 at their in two-thirds of patients with moderate to severe AKI, last measurement (Fig. 3). Conversely, those with a last uCCL14 category remained unchanged over a 24-hour measured uCCL14 that remained in the high category period, and where changes occurred, this indicated a were more likely to have died or received RRT by day 7 corresponding change in risk level for PS-AKI. Critical Care Explorations www.ccejournal.org     7 Prowle et al result is likely to provide good prognostication of clinical course of AKI as that provided by the orig- inal clinical investigations that derived and validated the assay. This suggests uCCL14 is an indication of a relatively persistent biological process within injured kidneys. Second, the serial measurement of uCCL14 may modify the assessment of risk over time. Notably, although uCCL14 category overall was generally unchanged, 51% of those with an initial medium value experienced Figure 2. Urinary C-C motif chemokine ligand 14 (uCCL14) category changes stratified by initial a shift in category within uCCL14 category after the diagnosis of moderate to severe acute kidney injury (AKI) complete the first 24 hours. Thus, case analysis. Overall, 66% of patients had no change in C-C motif chemokine ligand 14 (CCL14) for patients with an initial category in the first 24 hr after enrollment. measurement in the me- There are two important clinical implications of dium risk category, serial measurement provides an our findings. First, even if measurement of uCCL14 opportunity to update risk assessment with relatively is delayed after first diagnosis of stage 2–3 AKI, the frequent and clinically important changes possible. Conversely, for patients with an initial uCCL14 in the high-risk category, a change to a lower category, while TABLE 3. being associated with a clear reduction in risk, was Logistic Regression Model for Study still associated with a greater than 50% occurrence of Persistent Severe Acute Kidney Injury the primary endpoint. Finally, it appears that uCCL14 Endpoint With Initial C-C Motif Chemokine retains ability to discriminate the composite endpoint Ligand 14 Category and Change in C-C up to 7 days after initial diagnosis of moderate to se- Motif Chemokine Ligand 14 Category vere AKI, suggesting uCCL14 is potentially a flexible Over the First Day As Predictor Variables— Primary Analysis Cohort tool that could be used to assess underlying kidney pathophysiology through the time course of AKI. This Variable OR (95% CI) p work, therefore, represents an important contribution to defining the clinical utility and implementation of Initial category: medium 9.91 (4.64–23.3) < 0.001 the uCCL14 assay for the identification of those at risk Initial category: high 107 (39.0–326) < 0.001 for PS-AKI, giving clinicians confidence in its inter- Change in category: 0.20 (0.08–0.45) < 0.001 pretation over time. decrease The ability to categorize patients as a high or low Change in category: increase 4.04 (1.75–9.46) 0.001 risk of PS-AKI, RRT, or death has the potential to in- OR = odds ratio. form clinical decision-making and refine patient care. Reference levels: initial category: Low and change in category: Guideline-based AKI care is often variably applied unchanged. owing to insufficient prognostic information (17–19). Interaction term was eliminated from the model as there was no statistically significant interaction between initial category and Uncertainty regarding risk-benefit of interventions change in category. may explain the failure of many interventional trials 8     www.ccejournal.org March 2023 • Volume 5 • Number 3 Original Clinical Report Figure 3. Alluvial plot demonstrating initial urinary C-C motif chemokine ligand 14 (uCCL14) category, first change in category (if any), final category, and outcome within 7 d for uCCL14 categories assessed serially over up to 7 d. AKI = acute kidney injury, CCL14 = C-C motif chemokine ligand 14, RRT = renal replacement therapy. in AKI to improve outcomes (19–21). In this context, creatinine, some patients in Ruby had less severe AKI, the ability of uCCL14 to provide consistent prognostic similarly in Sapphire, by the time of next biosampling, information over time could facilitate targeted trial in- some patients no longer met AKI-2+. However, we clusion in strategies for the treatment or prevention of believe our design represents a pragmatic assessment PS-AKI. of a prognostic biomarker in a real-world clinical set- Our study has both strengths and limitations. We ting. Finally, observational studies cannot assess the studied a pooled cohort derived from two temporally benefit of uCCL14 measurement in directing clinical separated studies of international mixed (medical/ treatment decisions. Prospective clinical trials will be surgical) ICU populations, and as such, our find- needed to examine the effect of differential treatment ings are likely to be widely generalizable across crit- choices based on uCCL14 levels. ically ill populations and AKI etiologies. We studied In conclusion, we have demonstrated the consist- patient-centered outcomes, death, need for RRT, and ency of uCCL14 for the identification of patients at PS-AKI, which have been strongly linked to longer- high risk for the development of PS-AKI across serial term adverse patient-centered outcomes. We used the sampling and shown where changes occur these are standardized assay implemented on an established, paralleled in clinical outcomes. clinically available testing platform, which has now re- ceived Conformité Européene -marking for marketing 1 William Harvey Research Institute, Barts and the London within Europe. While the use of patients derived from School of Medicine and Dentistry, Queen Mary University of two independent studies is a strength, there were much London, London, United Kingdom. fewer cases of PS-AKI with Sapphire than Ruby, how- 2 Critical Care Department, Corporacion Sanitaria Universitaria Parc Tauli, CIBER Enfermedades Respiratorias, ever, pooling appeared statistically appropriate (Table Autonomous University of Barcelona, Sabadell, Spain. S3, http://links.lww.com/CCX/B147). Importantly, 3 Department of Critical Care Medicine, Faculty of Medicine while patients were included based on clinical AKI and Dentistry, University of Alberta and Alberta Health stage 2 or greater, after expert adjudication of baseline Services, Edmonton, AB, Canada. Critical Care Explorations www.ccejournal.org     9 Prowle et al 4 Department of Clinical & Experimental Medicine, University REFERENCES of Surrey and Royal Surrey County Hospital NHS Foundation Trust, Guildford, United Kingdom. 1. Ronco C, Bellomo R, Kellum JA: Acute kidney injury. Lancet 5 Division of Nephrology, Department of Medicine, University 2019; 394:1949–1964 of Michigan, Ann Arbor, MI. 2. Hobson C, Ozrazgat-Baslanti T, Kuxhausen A, et al: Cost and mortality associated with postoperative acute kidney injury. 6 Department of Intensive Care Medicine, Ghent University Ann Surg 2015; 261:1207–1214 Hospital, Ghent University, Ghent, Belgium; Research Foundation-Flanders (FWO), Brussels, Belgium. 3. Churpek MM, Carey KA, Edelson DP, et al: Internal and ex- ternal validation of a machine learning risk score for acute 7 Department of Critical Care, King’s College London, Guy’s kidney injury. JAMA Netw Open 2020; 3:e2012892 & St Thomas’ Hospital, London, United Kingdom. 4. Chertow GM, Burdick E, Honour M, et al: Acute kidney injury, 8 Section of Nephrology, Department of Medicine, University mortality, length of stay, and costs in hospitalized patients. J of Chicago, Chicago, IL. Am Soc Nephrol 2005; 16:3365–3370 9 Veterans Affairs Medical Center, San Diego, CA. 5. Hsu CY, Chinchilli VM, Coca S, et al: Post-acute kidney injury 10 Astute Medical, Inc. (a bioMérieux company), San Diego, proteinuria and subsequent kidney disease progression: The CA. Assessment, Serial Evaluation, and Subsequent Sequelae in 11 Center for Critical Care Nephrology, Department of Critical Acute Kidney Injury (ASSESS-AKI) study. JAMA Intern Med Care Medicine, University of Pittsburgh, Pittsburgh, PA. 2020; 180:402–410 Supplemental digital content is available for this article. Direct 6. Kidney Disease: Improving Global Outcomes (KDIGO) Acute URL citations appear in the printed text and are provided in the Kidney Injury Work Group: KDIGO Clinical Practice Guideline HTML and PDF versions of this article on the journal’s website for Acute Kidney Injury. Kidney Int Suppl 2012; 2:1–138 (http://journals.lww.com/ccejournal). 7. Coca SG, Nadkarni GN, Garg AX, et al: First post-operative Drs. Prowle, Kampf, and Kellum devised the study developed urinary kidney injury biomarkers and association with the the analysis plan and interpreted the results. Dr. Kwan performed duration of AKI in the TRIBE-AKI cohort. PLoS One 2016; the statistical analysis and developed the figures and tables with 11:e0161098 critical input from Drs. Prowle and Kampf. Dr. Prowle wrote the 8. Peerapornratana S, Priyanka P, Wang S, et al: Sepsis-associated article. Drs. Prowle, Kampf, McPherson, and Kellum reviewed acute kidney disease. Kidney Int Rep 2020; 5:839–850 and developed the initial draft article. All authors reviewed the 9. Ozrazgat-Baslanti T, Loftus TJ, Mohandas R, et al; Sepsis, analysis plan and read and critiqued the article before final edits Critical Illness Research Center Investigators: Clinical trajec- and approved the final submission. tories of acute kidney injury in surgical sepsis: A prospective The Ruby study (official title “Identification and Validation of observational study. Ann Surg 2022; 275:1184–1193 Biomarkers of Acute Kidney Injury Recovery”) and the Sapphire 10. Hoste E, Bihorac A, Al-Khafaji A, et al: Identification and val- study (official title “Evaluation of Novel Biomarkers From Acutely idation of biomarkers of persistent acute kidney injury: The Ill Patients at Risk for Acute Kidney Injury”) were funded by RUBY study. Intensive Care Med 2020; 46:943–953 Astute Medical (Astute Medical) a bioMérieux company. 11. Massoth C, Küllmar M, Enders D, et al; Progressive AKI Group: Drs. Kampf, Kwan, and McPherson are employees of Astute Comparison of C-C motif chemokine ligand 14 with other Medical. The remaining authors have disclosed that they do not biomarkers for adverse kidney events after cardiac surgery. J have any potential conflicts of interest. Thorac Cardiovasc Surg 2023; 165:199–207.e2 For information regarding this article, E-mail: j.prowle@qmul. 12. Bagshaw SM, Al-Khafaji A, Artigas A, et al: External validation ac.uk of urinary C-C motif chemokine ligand 14 (CCL14) for predic- Full listing of Sapphire and Ruby investigators is provided in the tion of persistent acute kidney injury. Crit Care 2021; 25:185 Supplementary Material (http://links.lww.com/CCX/B147). 13. Koyner JL, Chawla LS, Bihorac A, et al: Performance of a The Ruby study (official title “Identification and Validation of standardized clinical assay for urinary C-C motif chemokine Biomarkers of Acute Kidney Injury Recovery”) and the Sapphire ligand 14 (CCL14) for persistent severe acute kidney injury. study (official title “Evaluation of Novel Biomarkers From Acutely Kidney360 2022; 3:1158–1168 Ill Patients at Risk for Acute Kidney Injury”) were approved by the 14. Kashani K, Al-Khafaji A, Ardiles T, et al: Discovery and vali- Western Institutional Review Board (Puyallup, WA): Sapphire dation of cell cycle arrest biomarkers in human acute kidney (Approval Number: 20101176; Date: August 27, 2010) and injury. Crit Care 2013; 17:R25 Ruby (Approval Number: 20130523; Date: April 8, 2013). All 15. DeLong ER, DeLong DM, Clarke-Pearson DL: Comparing the study procedures were followed in accordance with the ethical areas under two or more correlated receiver operating charac- standards of the Review Board and with the Helsinki Declaration teristic curves: A nonparametric approach. Biometrics 1988; of 1975. 44:837–845 Informed consent was obtained from all individual participants 16. bioMérieux: Baxter and bioMérieux Announce CE Mark for included in the study or their legally authorized representatives. NEPHROCLEAR™ CCL14 Test to Predict Persistent Severe Ruby study: https://clinicaltrials.gov/ct2/show/NCT01868724; Acute Kidney Injury. 2022. Available at: https://www.biomer- Sapphire study: https://clinicaltrials.gov/ct2/show/ ieux.com/corp/en/journalists/press-releases/ce-mark-neph- NCT01209169. rocleartm-ccl14-test.html. Accessed July 22, 2022 10     www.ccejournal.org March 2023 • Volume 5 • Number 3 Original Clinical Report 17. Küllmar M, Weiß R, Ostermann M, et al: A multinational obser- the KDIGO guidelines in high-risk patients identified by bio- vational study exploring adherence with the kidney disease: markers: The PrevAKI-multicenter randomized controlled trial. Improving global outcomes recommendations for prevention Anesth Analg 2021; 133:292–302 of acute kidney injury after cardiac surgery. Anesth Analg 20. Hodgson LE, Roderick PJ, Venn RM, et al: The ICE-AKI study: 2020; 130:910–916 Impact analysis of a Clinical prediction rule and Electronic 18. Moledina DG, Belliveau O, Yamamoto Y, et al: Variation in best AKI alert in general medical patients. PLoS One 2018; practice measures in patients with severe hospital-acquired 13:e0200584 acute kidney injury: A multicenter study. Am J Kidney Dis 21. Mutter M, Martin M, Yamamoto Y, et al: Electronic Alerts for 2021; 77:547–549 Acute Kidney Injury Amelioration (ELAIA-1): A completely 19. Zarbock A, Küllmar M, Ostermann M, et al: Prevention of car- electronic, multicentre, randomised controlled trial: Design and diac surgery-associated acute kidney injury by implementing rationale. BMJ Open 2019; 9:e025117 Critical Care Explorations www.ccejournal.org     11