Bagshaw et al. Crit Care (2021) 25:185 https://doi.org/10.1186/s13054-021-03618-1 RESEARCH Open Access External validation of urinary C–C motif chemokine ligand 14 (CCL14) for prediction of persistent acute kidney injury Sean M. Bagshaw1* , Ali Al‑Khafaji2, Antonio Artigas3, Danielle Davison4, Michael Haase5, Matthew Lissauer6, Kai Zacharowski7, Lakhmir S. Chawla8, Thomas Kwan9, J. Patrick Kampf9, Paul McPherson9 and John A. Kellum2 Abstract Background: Persistent acute kidney injury (AKI) portends worse clinical outcomes and remains a therapeutic challenge for clinicians. A recent study found that urinary C–C motif chemokine ligand 14 (CCL14) can predict the development of persistent AKI. We aimed to externally validate urinary CCL14 for the prediction of persistent AKI in critically ill patients. Methods: This was a secondary analysis of the prospective multi-center SAPPHIRE study. We evaluated critically ill patients with cardiac and/or respiratory dysfunction who developed Kidney Disease: Improving Global Outcomes (KDIGO) stage 2–3 AKI within one week of enrollment. The main exposure was the urinary concentration of CCL14 measured at the onset of AKI stage 2–3. The primary endpoint was the development of persistent severe AKI, defined as ≥ 72 h of KDIGO stage 3 AKI or death or renal-replacement therapy (RRT) prior to 72 h. The secondary endpoint was a composite of RRT and/or death by 90 days. We used receiver operating characteristic (ROC) curve analysis to assess discriminative ability of urinary CCL14 for the development of persistent severe AKI and multivariate analysis to compare tertiles of urinary CCL14 and outcomes. Results: We included 195 patients who developed KDIGO stage 2–3 AKI. Of these, 28 (14%) developed persistent severe AKI, of whom 15 had AKI ≥ 72 h, 12 received RRT and 1 died prior to ≥ 72 h of KDIGO stage 3 AKI. Persistent severe AKI was associated with chronic kidney disease, diabetes mellitus, higher non-renal APACHE III score, greater fluid balance, vasopressor use, and greater change in baseline serum creatinine. The AUC for urinary CCL14 to predict persistent severe AKI was 0.81 (95% CI, 0.72–0.89). The risk of persistent severe AKI increased with higher values of urinary CCL14. RRT and/or death at 90 days increased within tertiles of urinary CCL14 concentration. Conclusions: This secondary analysis externally validates urinary CCL14 to predict persistent severe AKI in critically ill patients. Keywords: Acute kidney injury, Prediction, Validation, Renal replacement therapy, Mortality Background Acute kidney injury (AKI) is a common and vexing clini- cal challenge [1]. For most patients, kidney dysfunction with AKI either resolves rapidly or improves along with *Correspondence: bagshaw@ualberta.ca 1 Department of Critical Care Medicine, Faculty of Medicine and Dentistry, resolution of their acute illness [2]; however, in up to University of Alberta, 2‑124 Clinical Sciences Building, 8440‑112 ST NW, one-third of patients, AKI fails to resolve [3]. This persis- Edmonton, AB T6G 2B7, Canada tent AKI has prognostic importance and portends high Full list of author information is available at the end of the article © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. The Creative Commons Public Domain Dedication waiver (http://​creat​iveco​ mmons.​org/​publi​cdoma​in/​zero/1.​0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Bagshaw et al. Crit Care (2021) 25:185 Page 2 of 8 risk of long-term sequelae, including incident chronic enrolled adult critically ill patients with cardiac or res- kidney disease (CKD) and reduced survival [2]. Unfor- piratory dysfunction without known stage 2–3 AKI at tunately, there are no interventions proven to modify the the time of enrollment. The subset of SAPPHIRE patients clinical course for patients with persistent AKI [4]. This who developed stage 2–3 AKI within one week of enroll- is likely attributable to the inability to predict those at ment were included in the present analysis. Aligned with high risk for development of persistent AKI, a factor that the eligibility in the RUBY study, no patients in this SAP- has undoubtedly hampered suitable patient selection for PHIRE cohort had received a kidney transplant, had clinical trials and contributed to therapeutic uncertainty prior receipt of RRT or had active HIV or hepatitis infec- for clinicians [5]. tion [6]. Serial serum creatinine and urine output data, Recently, the RUBY study described the discovery of mortality, and RRT utilization were abstracted from the urinary C–C motif chemokine ligand-14 (CCL14) to pre- medical record for ascertainment of AKI stage, using the dict the development of persistent severe AKI, defined as full KDIGO classification criteria (serum creatinine and Kidney Disease: Improving Global Outcomes (KDIGO) urine output) and for endpoint determination. The ref- stage 3 AKI for 72 h or greater, among 331 critically ill erence (baseline) serum creatinine was obtained by pri- patients with KDIGO stage 2 or greater AKI (AUC 0.83) ority in the following sequence: (1) if at least five values [6]. In this study, urinary CCL14 was found to outper- were available, the median of all values available from form several other AKI biomarkers to predict persistent six months to six days prior to enrollment was used; (2) severe AKI. CCL14 is a member of the chemokine fam- if not available, the lowest value in the five days prior to ily recognized for its role in macrophage trafficking and enrollment was used; and (3) if no pre-enrollment cre- may contribute to the development of persistent kidney atinine was available, the creatinine value at the time of damage, maladaptive repair, and risk for non-recovery of enrollment was used [9]. kidney function [6, 7]. Knowledge of the risk for patients to develop persistent Main exposure AKI would have important implications for bedside care, The main exposure was the urinary concentration of including decision making on the utilization of medica- CCL14, measured within 36 h of the onset of when stage tions with potentially direct toxic or accumulative effects 2 or greater AKI was first identified. and the informed application of renal-replacement ther- apy (RRT) [8]. Moreover, such knowledge would provide Clinical endpoints insights into the pathobiology and kinetics of kidney The primary endpoint was the development of persistent repair and recovery and further enable predictive enrich- severe AKI, using both serum creatinine and urine out- ment in randomized trials of potential therapeutic tar- put criteria, as previously defined in the RUBY study [6]. gets [5]. Accordingly, we aimed to externally validate the Briefly, patients who developed 72 consecutive hours of performance of the novel biomarker urinary CCL14 for stage 3 AKI or who died or received RRT prior to achiev- the prediction of persistent severe AKI among an estab- ing 72 consecutive hours of stage 3 AKI were considered lished, heterogenous cohort of critically ill patients. endpoint positive. Patients with stage 2 AKI at the time of sample collection for biomarker testing were consid- Methods ered endpoint positive if persistent stage 3 AKI started The SAPPHIRE study protocol was approved by the within 48 h of sample collection. This aligns with recent Western Institutional Review Board (Olympia, Washing- randomized trials evaluating timing of RRT initiation ton, USA) and by the investigational review boards/ethics that integrated persistent AKI at 72 h as indications for committees as required by each participating institution starting RRT [8]. The secondary endpoint was a compos- [9]. All enrolled patients or their legally authorized repre- ite of RRT and/or death by 90 days. sentatives provided written informed consent. Sample, data collection and measurement Design We collected urine samples twice daily for 4 days from This was a secondary analysis of the previously described enrollment and then once daily for 3 days. Supernatants SAPPHIRE study, a multi-national prospective observa- from centrifuged urine samples were flash frozen, stored tional cohort study of heterogeneous critically ill patients at or below − 70 °C, and thawed prior to analysis. To at risk for development of AKI [9]. align with the inclusion criteria for the RUBY study, all urine samples within 36 h of onset of the first instance Patients of stage 2 AKI (or stage 3 AKI if not preceded by stage Patients enrolled in the SAPPHIRE study served as the 2) and prior to initiation of RRT were analyzed. An aver- source population for the current study [9]. SAPPHIRE age of 2.5 urine samples per patient were included in the Bagshaw et al. Crit Care (2021) 25:185 Page 3 of 8 analysis. Technicians blinded to the clinical data meas- significant. We performed all analyses using R 4.0.2 (R immunoassay using the NEPHROCLEAR™ CCL14 Test ured CCL14 concentrations in the urine samples by Foundation for Statistical Computing. Vienna, Austria). on the Astute140® Meter (Astute Medical, San Diego, Results CA). Of 723 patients enrolled in the SAPPHIRE study, 195 developed stage 2–3 AKI and were included in the analy- Statistical analysis sis (Fig. 1). Of these, 28 (14%) developed the primary We used receiver operating characteristic (ROC) curve endpoint of persistent severe AKI, including 15 who had analysis to assess discriminative ability of urinary CCL14 AKI ≥ 72 h, 12 who received RRT and 1 who died prior to for the development of persistent severe AKI. The empir- achieving 72 consecutive hours in stage 3 AKI. ical area under the ROC curve (AUC) was calculated, and corresponding confidence interval (CI) was determined Clinical characteristics using bootstrap sampling at the patient level to account Patients who developed persistent severe AKI had greater for multiple samples analyzed per patient. The AUC from prevalence of chronic kidney disease (CKD) and diabe- the SAPPHIRE analysis was compared to the AUC previ- tes mellitus (DM) at baseline (Table 1). Persistent severe ously reported for the same endpoint in the RUBY study AKI was associated with higher non-renal APACHE III using a permutation test [6]. We assessed the improve- scores, greater positive fluid balance, and more vasopres- ment of risk prediction when adding urinary CCL14 to a sor use at enrollment. Persistent severe AKI patients had reference model consisting of baseline serum creatinine higher baseline serum creatinine values, greater relative and non-renal Acute Physiology and Chronic Health change and differences in serum creatinine compared Evaluation (APACHE) III score by integrated discrimi- with urine output and were more likely to have stage 3 nation improvement (IDI), category-free net reclassifi- AKI at enrollment compared to those with no persistent cation improvement (cfNRI), and AUC difference. Risk severe AKI (Additional file 1: S1 and S2). estimates for both the new and reference models were calculated using Generalized Estimating Equations (GEE) CCL14 and persistent severe AKI implemented in the geepack R package [10]. To assess the Urinary CCL14 showed good prediction for the occur- relationship between risk of persistent severe AKI and rence of persistent severe AKI with an AUC 0.81 (95% urinary CCL14 concentration, we plotted risk as a func- CI, 0.72–0.89) (Fig. 2). The AUC was not significantly tion of the measured urinary CCL14 concentration. We different from the AUC in the RUBY study [6]. The risk defined risk as the proportion of patient samples that for persistent severe AKI increased with greater values of were positive for persistent severe AKI among all patient urinary CCL14 (Fig. 3). This increased risk was consist- samples with a urinary CCL14 concentration above the ent across simulated curves that modified the baseline measured value. Risk curves were simulated for differ- prevalence of AKI in the SAPPHIRE population. Uri- ent prevalence levels of persistent severe AKI based on nary CCL14 improved the predictive ability of clinical the established relationship between positive predictive value and disease prevalence [11]. We calculated hazard ratios for the composite endpoint of RRT and/or death within 90 days, both unadjusted and adjusted for base- line serum creatinine and non-renal APACHE III score, using Cox proportional hazards analysis with CCL14 concentrations as a time-dependent covariate. Probabil- ity (1–survival) of the composite of RRT and/or death within 90 days was plotted by CCL14 concentration of the first urine sample collected after onset of stage 2–3 AKI stratified into tertiles. We compared curves across tertiles using the log-rank test. Continuous and categori- cal baseline variables were compared between patients negative and positive for the primary endpoint using the Mann–Whitney U test and Fisher exact test, respec- tively. Continuous predictors in GEE and Cox regressions were standardized by subtracting the group mean and Fig. 1 Patient cohort in SAPPHIRE study. Patients with at least one normalizing by the standard deviation. For all analyses, sample that met the persistent severe AKI criteria were considered endpoint positive two-sided p values < 0.05 were considered statistically Bagshaw et al. Crit Care (2021) 25:185 Page 4 of 8 Table 1 Baseline characteristics for all patients and those who met and did not meet the primary endpoint of persistent severe AKI All patients No persistent severe AKI Persistent severe AKI p ­valuec Patients 195 167 28 Male 113 (58%) 95 (57%) 18 (64%) 0.538 Agea (years) 66 (56–75) 66 (56–75) 68.5 (58–76) 0.592 Race White or Caucasian 148 (76%) 126 (75%) 22 (79%) Black or African American 25 (13%) 23 (14%) 2 (7%) Other/unknown 22 (11%) 18 (11%) 4 (14%) 0.607 Chronic comorbidities Chronic kidney disease 21 (11%) 12 (7%) 9 (32%) < 0.001 Diabetes mellitus 77 (39%) 59 (35%) 18 (64%) 0.006 Heart failure 43 (22%) 37 (22%) 6 (21%) 1.000 Coronary artery disease 70 (36%) 55 (33%) 15 (54%) 0.054 Hypertension 141 (72%) 118 (71%) 23 (82%) 0.258 COPD 47 (24%) 42 (25%) 5 (18%) 0.482 Cancer 54 (28%) 48 (29%) 6 (21%) 0.500 Reason for ICU admission Respiratory 90 (46%) 76 (46%) 14 (50%) 0.687 Surgery 67 (34%) 58 (35%) 9 (32%) 0.834 Cardiovascular 67 (34%) 54 (32%) 13 (46%) 0.196 Sepsis 42 (22%) 36 (22%) 6 (21%) 1.000 Neurological 18 (9%) 18 (11%) 0 (0%) 0.081 Trauma 7 (4%) 6 (4%) 1 (4%) 1.000 Other 30 (15%) 24 (14%) 6 (21%) 0.394 Baseline serum creatinine (mg/dL) 0.9 (0.7–1.2) 0.9 (0.7–1.2) 1.2 (0.8–2.2) 0.006 Maximum KDIGO stage before sample collection Stage 2 172 (88%) 160 (96%) 12 (43%) Stage 3 23 (12%) 7 (4%) 16 (57%) < 0.001 eGFR < 60 mL/min 88 (45%) 68 (42%) 20 (74%) 0.003 Diuretics 75 (38%) 63 (38%) 12 (43%) 0.676 Vasopressor 101 (52%) 80 (48%) 21 (75%) 0.008 Mechanical ventilation 164 (84%) 138 (83%) 26 (93%) 0.263 Fluid balance (mL)a 2044 (248–3362) 1879 (180–3154) 3226 (695–4554) 0.025 Non-renal APACHE III score 62 (43–86) 60 (42–85) 80 (58–104) 0.002 Time from ICU admission to enrollment (h) 14 (6–20) 14 (6–20) 14 (4–19) 0.711 Maximum serum c­ reatinineb (mg/dL) between enrollment and time 1.3 (0.9–1.8) 1.2 (0.8–1.6) 2.4 (1.4–3.2) < 0.001 of first sample collection after onset of AKI Stage 2 or 3 Time between enrollment and first urinary CCL14 sample collection 29 (15–60) 33 (15–60) 21 (0–34) 0.057 (h) a Fluid balance was defined as the difference in total fluid intake minus output from the day prior to through the day of study enrollment b Specifically, this is the greatest (maximum) of all serum creatinine concentrations (mg/dL) collected as standard of care between the time of study enrollment and the time of the first urine sample collection for CCL14 testing after the onset of AKI Stage 2 or 3. This is expressed as a median (interquartile range) c P values for age, baseline serum creatinine, non-renal APACHE III score, time from ICU admission to enrollment, serum creatinine at time of first sample collection after onset of AKI Stage 2 or 3, and fluid balance were computed by the Mann–Whitney U test. All others were computed by Fisher’s Exact Test. P values less than 0.05 are in bold variables evaluated with GEE. The addition of urinary alone (Table 2). This was similarly shown in sensitivity CCL14 to a baseline clinical model that included serum analysis adjusting instead for trajectory in serum creati- creatinine and non-renal APACHE III score for the nine (Additional file 1: S3). In a further sensitivity analy- development of persistent severe AKI provides incre- sis, we explored the effect of oliguria on the performance mental predictive information beyond clinical variables of urinary CCL14 to predict persistent severe AKI and Bagshaw et al. Crit Care (2021) 25:185 Page 5 of 8 well as following adjustment for baseline serum creati- nine and non-renal APACHE III score when comparing CCL14 values from tertile 3 to tertile 1 (Table 3). Discussion We performed a secondary analysis using the multi- national prospective observational SAPPHIRE study to externally validate the ability of the novel biomarker uri- nary CCL14 to predict persistent severe AKI, defined as at least 72 h of stage 3 AKI, receipt of RRT or death, among a heterogeneous cohort of critically ill patients with stage 2–3 AKI. We found that urinary CCL14 had good predictive performance for the development of per- sistent severe AKI (AUC 0.81), was not modified by olig- uria, and provided incremental value beyond commonly available clinical variables. We also found urinary CCL14 correlated with death and RRT use in unadjusted and in Fig. 2 Receiver operating characteristic curves for prediction of adjusted analysis when comparing CCL14 from the high- persistent severe AKI with urinary CCL14 in the RUBY (solid gray) [6] and SAPPHIRE (solid black) [9] studies. Dashed gray reference est and lowest tertiles. Our analysis largely replicated the line shows no discrimination between persistent severe AKI and no performance for urinary CCL14 with the findings of the persistent severe AKI RUBY study (AUC 0.83) and now provides added knowl- edge to extend our confidence in the validity and poten- tial role urinary CCL14 can have in clinical practice [6]. Furthermore, we have extended the findings of the RUBY study by validating the clinical assay using a commer- cially available analyzer and we have further investigated the relationship between CCL14 and persistent severe AKI by simulating different baseline prevalence rates. We submit that our findings provide a further foundation for evaluation of the application and use of urinary CCL14 in clinical practice. CCL14 is hypothesized to be an important mediator of kidney tissue damage and may represent a novel indicator of the risk for maladaptive repair and kidney non-recov- ery [6]. Prior work has shown CCL14 is one of a series of important “kidney risk inflammatory signature [KRIS]” Fig. 3 Risk for persistent severe AKI versus CCL14 concentration inflammatory proteins circulating in patients with DM observed in SAPPHIRE (solid black) and simulated at different levels which may mediate the risk of progressing to end-stage of persistent severe AKI: 20% (solid dark gray), 30% (solid gray), 40% kidney disease (ESKD) [7]. The mechanisms by which (dashed gray), and 50% (dotted gray) CCL14 mediates persistent severe AKI are not yet com- pletely understood. In AKI, CCL14 may be released from found no evidence of significant interaction (Additional the injured tubular epithelial cells through the activation file 1: S4). of inflammatory mediators (e.g., via stimulation of TNF-a receptor activity). However, we submit that further trans- lational work is needed to better understand the role of CCL14 and mortality and RRT​ CCL14 in potentially disrupting intrinsic kidney repair RRT use, death and a composite of RRT and/or death mechanisms following an episode of AKI, how this may occurred in 29 (15%), 63 (32%) and 74 (38%) patients contribute to persistent severe AKI and downstream risk within 90 days, respectively (Additional file 1: S5). The of new or worsening CKD and ESKD. probability of RRT and/or death within 90 days increased Our external validation of urinary CCL14 to pre- across tertiles of urinary CCL14 concentration (Fig. 4). dict the occurrence of persistent severe AKI has This was shown in the unadjusted models for urinary several noteworthy implications. From a research CCL14 both as a continuous variable and by tertiles; as standpoint, urinary CCL14 may represent an innovative Bagshaw et al. Crit Care (2021) 25:185 Page 6 of 8 Table 2 Integrated discrimination improvement (IDI), category-free net reclassification improvement (cfNCI), and AUC difference with the addition of CCL14 to a reference clinical model with baseline serum creatinine and non-renal APACHE III score Value 95% CI P value IDI 0.065 0.005–0.172 < 0.001 IDI: event 0.048 0.003–0.141 < 0.001 IDI: non-event 0.017 0.002–0.035 < 0.001 cfNRI 0.864 0.389–1.276 < 0.001 cfNRI: event 0.322 0.051–0.633 0.004 cfNRI: non-event 0.542 0.272–0.713 < 0.001 AUC ref model 0.768 0.662–0.872 < 0.001 AUC new model 0.834 0.738–0.918 < 0.001 AUC difference 0.065 0.007–0.143 < 0.001 Model Variable OR 95% CI P value Reference Non-renal APACHE 1.85 1.25–2.73 0.002 Baseline serum creatinine 2.26 1.42–3.62   0.001 New Non-renal APACHE 1.63 1.12–2.39 0.012 Baseline serum creatinine 2.09 1.24–3.50 0.005 CCL14 1.63 1.21–2.19 0.001 Event is persistent severe AKI. Odds ratio (OR), 95% confidence interval (CI), and P value for each variable in the models are shown in the lower table inhibition in clinical trials. In addition, urinary CCL14 can further be used to discriminate eligibility thresh- olds to predictively enrich the target population in clin- ical trials that aim to evaluate therapeutic strategies in patients at high risk for experiencing persistent severe AKI and its sequelae [14]. From a clinical perspective, urinary CCL14 can pro- vide incremental and valuable information for clini- cians that can translate into improved quality of care at the bedside. For example, a positive urinary CCL14 test can better advise clinician’s expectations about the clin- ical course in a given patient and directly inform care processes, including triaging to a suitable level of moni- toring (e.g., ICU or HDU; serial serum creatinine and Fig. 4 Cumulative incidence probability of RRT or death within urine output), augmented surveillance for and avoid- 90 days of the first sample collection after onset of stage 2 or 3 AKI ance of complications (e.g., withholding potassium stratified by CCL14 concentration tertiles. Tertile 1 (light gray), Tertile supplementation, modifications to enteric nutrition, 2 (dark gray), and Tertile 3 (black) correspond to concentrations at or fluid overload, etc.), and dose-modification of or overt below the 33rd percentile, between the 33rd and 67th percentile, and above the 67th percentile of all CCL14 concentrations. P value = 0.014 avoidance of selected medications with potential toxic for the log rank test or cumulative adverse effects. Alternatively, a negative urinary CCL14 could provide reassurance about de- intensification of monitoring. Similarly, knowledge of greater risk for persistent mechanistic surrogate endpoint used to evaluate novel severe AKI should provoke nephrology consultation therapeutic targets for the prevention of persistent and planning for the potential receipt of RRT for those severe AKI and its sequelae [5]. For example, CCL14 with further deterioration. Conversely, a negative uri- may be a critical chemokine for monocyte/macrophage nary CCL14 could reinforce a “watch and wait strategy” recruitment and activation and may mediate kidney and prompt reconsideration of early start to RRT unless injury and accelerate kidney fibrosis [12, 13]. In this absolute indications are present that are medically context, CCL14 may also be a target for modulation or refractory and unavoidable [8]. Bagshaw et al. Crit Care (2021) 25:185 Page 7 of 8 Table 3 Hazard ratio for (A) continuous CCL14 and (B) CCL14 tertiles, unadjusted and adjusted for baseline serum creatinine and non- renal APACHE III score in a Cox regression model with CCL14 as a time-dependent covariate Model HR (95% CI) P value A. Urinary CCL14 concentration as a continuous variable in model Unadjusted (CCL14 only) 1.67 (1.36–2.06) < 0.001 Adjusted for baseline serum creatinine 1.55 (1.24–1.93) < 0.001 Adjusted for baseline non-renal APACHE 1.51 (1.22–1.88) < 0.001 Adjusted for baseline serum creatinine and non-renal APACHE 1.41 (1.13–1.75) 0.003 Model Tertile 2 versus 1 Tertile 3 versus 1 HR (95% CI) P value HR (95% CI) P value B. Urinary CCL14 concentration divided into tertiles and used as a categorical variable in model Unadjusted (CCL14 only) 1.14 (0.61–2.13) 0.671 2.56 (1.46–4.50) 0.001 Adjusted for baseline serum creatinine 1.09 (0.58–2.02) 0.795 2.16 (1.20–3.88) 0.010 Adjusted for baseline non-renal APACHE 1.15 (0.62–2.14) 0.662 2.10 (1.18–3.75) 0.012 Adjusted for baseline serum creatinine and non-renal 1.15 (0.62–2.15) 0.651 1.83 (1.02–3.29) 0.043 APACHE Model is for time to RRT and/or death within 90 days Finally, a positive CCL14 should prompt considera- Conclusions tion for the long-term implications of incomplete kid- Our study externally validated urinary CCL14 to predict ney recovery and risk of incident CKD, progression to persistent severe AKI. The findings of an elevated urinary ESKD and other complications associated with AKI CCL14 in critically ill patients with AKI have relevance [15–19]. Our data now reinforce existing recommen- for clinical care and represent a mechanism of predictive dations in clinical practice guidelines that survivors of enrichment for clinical trials. AKI have a routine assessment of kidney function at three months and as needed thereafter, whereby such Abbreviations risk and complications can be identified and potentially AKI: Acute kidney injury; APACHE: Acute Physiology and Chronic Health modified [20]. Evaluation; AUC​: Area under receiver operating characteristic curve; CCL14: Our study has strengths. This study leveraged a high- C–C motif chemokine ligand 14; COPD: Chronic obstructive pulmonary disease; cfNRI: Category-free net reclassification improvement; CKD: Chronic fidelity international prospective observational cohort kidney disease; DM: Diabetes mellitus; ESKD: End-stage kidney disease; GFR: for external validation. The study integrated current Glomerular filtration rate; HR: Hazard ratio; ICU: Intensive care unit; IDI: Inte‑ consensus recommendations for the diagnosis and grated discrimination improvement; KDIGO: Kidney Disease: Improving Global Outcomes; RCT​: Randomized controlled trial; RRT​: Renal replacement therapy. staging of AKI [20]. The main study exposure, urinary CCL14, was blinded to clinicians during the study and Supplementary Information was evaluated from high-quality stored samples. How- The online version contains supplementary material available at https://​doi.​ ever, our study also has limitations. The number of org/​10.​1186/​s13054-​021-​03618-1. patients experiencing the primary outcome in our vali- dation study was small (14%); and as such, the study is Additional file 1. Supplementary data presentation and analysis to sup‑ certain to have limited statistical power. This is particu- port the main analysis. larly true for analyses of secondary endpoints, which should be viewed as exploratory and hypothesis gener- Acknowledgements ating. Further, we recognize that the SAPPHIRE study SMB is supported by a Canada Research Chair in Critical Care Nephrology. was performed several years ago and acknowledge that Authors’ contributions temporal changes in clinical practice (e.g., indications SMB, TK, JPK, PM and JAK designed the study. SMB, TK, JPK, PM and JAK and timing for RRT) may have evolved in response to performed analysis and interpreted data. SMB, JPK and JAK drafted the manu‑ new data [8, 21–23]. Additionally, the study did not script. All authors reviewed, contributed to critical revision and approved the final manuscript. All authors read and approved the final manuscript. further classify AKI events into sub-phenotypes which also may have important prognostic implications and Funding consideration for case-mix selection for future clinical Astute Medical/ bioMérieux. trials. Bagshaw et al. Crit Care (2021) 25:185 Page 8 of 8 Availability of data materials renal recovery: consensus report of the Acute Disease Quality Initiative Requests for access to the SAPPHIRE dataset can be made to Astute Medical. (ADQI) 16 Workgroup. Nat Rev Nephrol. 2017;13(4):241–57. 5. Billings FT, Shaw AD. Clinical trial endpoints in acute kidney injury. Nephron Clin Pract. 2014;127(1–4):89–93. Declarations 6. Hoste E, Bihorac A, Al-Khafaji A, Ortega LM, Ostermann M, Haase M, Zacharowski K, Wunderink R, Heung M, Lissauer M, et al. Identification Ethical approval and consent to participate. and validation of biomarkers of persistent acute kidney injury: the RUBY The SAPPHIRE study protocol was approved by the Western Institutional study. Intensive Care Med. 2020;46(5):943–53. Review Board (Olympia, Washington, USA) and by each investigational review/ 7. Niewczas MA, Pavkov ME, Skupien J, Smiles A, Md Dom ZI, Wilson JM, Park ethics boards/committees at each participating institution, as required. All J, Nair V, Schlafly A, Saulnier PJ, et al. A signature of circulating inflamma‑ enrolled patients or their legally authorized representatives provided written tory proteins and development of end-stage renal disease in diabetes. informed consent. Nat Med. 2019;25(5):805–13. 8. STARRT-AKI Investigators, Canadian Critical Care Trials Group tA, New Consent for publication Zealand Intensive Care Society Clinical Trials Group tUKCCRGtCNTN, the Not applicable. Irish Critical Care Trials G, Bagshaw SM, Wald R, Adhikari NKJ, Bellomo R, da Costa BR, Dreyfuss D et al. Timing of initiation of renal-replacement Competing interests therapy in acute kidney injury. N Engl J Med. 2020; 383(3):240–51. SMB discloses consulting fees and grant support from Baxter and consulting 9. Kashani K, Al-Khafaji A, Ardiles T, Artigas A, Bagshaw SM, Bell M, Bihorac fees from CNA Diagnostics. AAK, AA and DD have nothing to disclose. MH dis‑ A, Birkhahn R, Cely CM, Chawla LS, et al. Discovery and validation of closes consulting fees from Abbott outside the present analysis. ML discloses cell cycle arrest biomarkers in human acute kidney injury. Crit Care. research grant funding from Prescient Healthcare, LLC as well as grant support 2013;17(1):R25. from Astute Medical/ bioMérieux outside of the present analysis. KZ received 10. Hojsgaard S, Halekoh U, Yan J. The R package geepack for generalized grants from B. Braun Melsungen, CSL Behring, Fresenius Kabi, and Vifor Pharma estimating equations. J Stat Softw. 2006;15(2):1–11. for the implementation of Frankfurt’s Patient Blood Management program 11. Altman DG, Bland JM. Diagnostic tests 2: predictive values. BMJ. and honoraria for scientific lectures from B. Braun Melsungen, Vifor Pharma, 1994;309(6947):102. Ferring, CSL Behring, and Pharmacosmos outside the present analysis. LSC 12. Meng XM, Mak TS, Lan HY. Macrophages in renal fibrosis. Adv Exp Med reports receiving consulting fees from Astute Medical/ bioMérieux. TK, JPK and Biol. 2019;1165:285–303. PM are employees of Astute Medical/ bioMérieux. JAK discloses consulting 13. Meng XM, Tang PM, Li J, Lan HY. Macrophage phenotype in kidney injury fees and grant support from Astute Medical/bioMérieux outside the present and repair. Kidney Dis (Basel). 2015;1(2):138–46. analysis. 14. Stanski NL, Wong HR. Prognostic and predictive enrichment in sepsis. Nat Rev Nephrol. 2020;16(1):20–31. Author details 15. Wang WJ, Chao CT, Huang YC, Wang CY, Chang CH, Huang TM, Lai 1 Department of Critical Care Medicine, Faculty of Medicine and Dentistry, CF, Huang HY, Shiao CC, Chu TS, et al. The impact of acute kidney University of Alberta, 2‑124 Clinical Sciences Building, 8440‑112 ST NW, injury with temporary dialysis on the risk of fracture. J Bone Miner Res. Edmonton, AB T6G 2B7, Canada. 2 Department of Critical Care Medicine, 2014;29(3):676–84. Center for Critical Care Nephrology, University of Pittsburgh, 3550 Terrace St., 16. Wu VC, Wang CY, Shiao CC, Chang CH, Huang HY, Huang TM, Lai CF, Lin Scaife Hall, Suite 600, Pittsburgh, PA 15213, USA. 3 Critical Care Department, MC, Ko WJ, Wu KD, et al. Increased risk of active tuberculosis following Corporacion Sanitaria Universitaria Parc Tauli, CIBER Enfermedades Respirato‑ acute kidney injury: a nationwide, population-based study. PLoS ONE. rias, Autonomous University of Barcelona, Parc Tauli 1, 08208 Sabadell, Spain. 2013;8(7):e69556. 4 Department of Anesthesiology and Critical Care Medicine, School of Medi‑ 17. Wu VC, Wu PC, Wu CH, Huang TM, Chang CH, Tsai PR, Ko WJ, Chen L, cine and Health Sciences, George Washington University, 900 23rd St. NW, Wang CY, Chu TS et al. The impact of acute kidney injury on the long- Washington, DC 20037, USA. 5 Diaverum Renal Care Center, 14469 Potsdam, term risk of stroke. J Am Heart Assoc. 2014; 3(4). Germany and Medical Faculty, Otto Von-Guericke-University Magdeburg, 18. Yang WS, Hu FC, Chen MK, Ko WJ, Chen L, Wu KD, Wu VC. High risk of Leipziger Str. 44, 39120 Magdeburg, Germany. 6 Division of Acute Care Surgery, herpes zoster among patients with advance acute kidney injury—a Department of Surgery, Rutgers-Robert Wood Johnson Medical School, 125 population-based study. Sci Rep. 2015;5:13747. Patterson Street, New Brunswick, NJ 07746, USA. 7 Department of Anesthesiol‑ 19. Chawla LS, Amdur RL, Shaw AD, Faselis C, Palant CE, Kimmel PL. Associa‑ ogy, Intensive Care Medicine and Pain Therapy, University Hospital Frankfurt, tion between AKI and long-term renal and cardiovascular outcomes in Goethe University, Theodor‑Stern‑Kai 7, 60590 Frankfurt am Main, Germany. United States veterans. Clin J Am Soc Nephrol. 2014;9(3):448–56. 8 Veterans Affairs Medical Center, 3350 La Jolla Village Dr, San Diego, CA 92161, 20. KDIGO Clinical Practice Guideline for Acute Kidney Injury. https://​kdigo.​ USA. 9 Astute Medical, Inc. (a bioMérieux company), 3550 General Atomics Ct, org/​guide​lines/​acute-​kidney-​injury/ San Diego, CA 92121, USA. 21. Barbar SD, Gaudry S, Dreyfuss D, Quenot JP. Renal replacement therapy: time to give up on early initiation? Yes Anaesth Crit Care Pain Med. Received: 9 March 2021 Accepted: 25 May 2021 2018;37(6):501–3. 22. Gaudry S, Hajage D, Schortgen F, Martin-Lefevre L, Pons B, Boulet E, Boyer A, Chevrel G, Lerolle N, Carpentier D, et al. Initiation strategies for renal-replacement therapy in the intensive care unit. N Engl J Med. 2016;375(2):122–33. References 23. Zarbock A, Kellum JA, Schmidt C, Van Aken H, Wempe C, Pavenstadt 1. Hoste EA, Bagshaw SM, Bellomo R, Cely CM, Colman R, Cruz DN, Edipidis H, Boanta A, Gerss J, Meersch M. Effect of early vs delayed initiation K, Forni LG, Gomersall CD, Govil D, et al. Epidemiology of acute kidney of renal replacement therapy on mortality in critically ill patients injury in critically ill patients: the multinational AKI-EPI study. Intensive with acute kidney injury: the ELAIN randomized clinical trial. JAMA. Care Med. 2015;41(8):1411–23. 2016;315(20):2190–9. 2. Kellum JA, Sileanu FE, Bihorac A, Hoste EA, Chawla LS. Recovery after acute kidney injury. Am J Respir Crit Care Med. 2017;195(6):784–91. 3. Perinel S, Vincent F, Lautrette A, Dellamonica J, Mariat C, Zeni F, Cohen Y, Publisher’s Note Tardy B, Souweine B, Darmon M. Transient and persistent acute kidney Springer Nature remains neutral with regard to jurisdictional claims in pub‑ injury and the risk of hospital mortality in critically ill patients: results of a lished maps and institutional affiliations. multicenter cohort study. Crit Care Med. 2015;43(8):e269-275. 4. Chawla LS, Bellomo R, Bihorac A, Goldstein SL, Siew ED, Bagshaw SM, Bittleman D, Cruz D, Endre Z, Fitzgerald RL, et al. Acute kidney disease and