Biomarker-Based Prediction of Survival and Recovery of Kidney Function in Acute Kidney Injury

The diagnosis of acute kidney injury (AKI) is still made according to the 2012 published “KDIGO clinical practice guidelines for AKI” [1]. Criterions 1 and 2 are fulfilled if dynamic changes in serum creatinine exceed certain thresholds (an increase of at least 0.3 mg/dL within 48 h or a 1.5-fold increase within 7 days). A reduction in urine output to under 0.5 mL/kg/h for at least 6 h defines criterion 3. The latter can only be applied if urine production is closely monitored, e.g., at intensive care units (ICUs). AKI affects increasing numbers of in-hospital patients [2]. It has a substantial prognostic impact concerning the short- and long-term mortality. Also, every individual AKI episode increases the risk for chronic kidney disease (CKD) during follow-up [3]. Serum creatinine is by no means an optimal AKI biomarker; therefore, new markers have been identified and evaluated clinically under various circumstances [4, 5]. The majority of these molecules have been tested for their diagnostic ability. Fewer data are however available about the biomarker-based prediction of AKI survival and of post-AKI recovery of kidney function (ROKF).

AKI remains a challenging problem for physicians in hospitals worldwide. In 2018, Hoste and colleagues [6] summarized epidemiological data on AKI incidence and outcome from a global perspective. The in-hospital incidence of the syndrome was reported with 19.3% in Northern Europe, with ∼31% in South America, and with ∼31% in Southeast Asia. The AKI in-hospital mortality has been reported with 23.8% for all affected individuals [7] but may for instance dramatically increase in sepsis (>50% [6, 8, 9]). Mortality of ∼100% has been documented in patients with non-solid tumors, sepsis and severe AKI [10]. The AKI risk also depends on age (higher incidences in older subjects [11]) but most likely not on gender. In addition to lowering the chance of survival in the short term, AKI also affects the prognosis of surviving individuals in the long term. In 2011, Wu and colleagues [12] published respective data (>7 years) of in-hospital AKI patients and showed a progressive reduction of the survival probability from “non-CKD-non-AKI” to “CKD-AKI” and “end-stage renal disease.” Subjects with “CKD-AKI” had an almost 50% higher risk of death overtime as compared to “non-CKD-non-AKI” individuals. In addition, the prevalence of long-term dialysis dependence increased from “non-CKD-non-AKI” to “non-CKD-failure” (failure stage of the RIFLE criteria [13]) and was even higher in “CKD-AKI.” Thus, AKI increases the CKD risk, particularly in the elderly [3]. Even mild AKI episodes, defined as subclinical AKI [14], must be considered as serious disorders concerning the long-term outcome.

Serum creatinine is by no means an optimal biomarker; the most significant disadvantage is the delayed increase of the metabolite [15]. Also, it lacks specificity since it may increase in various situations such as acute heart failure, sepsis, hypovolemia, obstruction, and intrinsic types of AKI (e.g., glomerulonephritis, acute interstitial nephritis). In AKI, the dynamic change in glomerular filtration is not adequately reflected by serum creatinine [16]. Furthermore, the marker does not indicate tissue damage that is not accompanied by a reduction in the glomerular filtration rate. Finally, its prognostic value is limited; the absolute increase has more or less no predictive value regarding the chance of renal recovery. The limitations of serum creatinine are widely known [4, 15, 16].

The field of biomarker research has significantly evolved over the last 10 years, not only in nephrology. The term biomarker must be used in a broader sense. It encompasses biological molecules such as hormones, cells, cellular metabolites, and even RNA molecules or genes that reflect biological states. It is fair to say that biomarkers are increasingly being used in almost all medical disciplines. It is nevertheless impossible to even name the most important candidates used in cardiology or neurology or hematology. By applying the search terms “RNA biomarker” AND “hematology” for instance, more than 34,000 references are provided. This finding only indicates the collective effort that has been invested in recent years. In general, AKI biomarkers can be distinguished by their ability to provide diagnostic and/or predictive/prognostic information. The literature on AKI biomarkers is vast. Principally, three AKI biomarker groups are distinguished: markers of (i) impaired function, markers of (ii) tissue damage, and markers of (iii) stress. The first group is currently represented by cystatin C [17] and proenkephalin A [18, 19]. The second group encompasses numerous proteins of heterogenous origin and function (unordered selection): neutrophil gelatinase-associated lipocalin (NGAL) [4, 20‒22], kidney injury molecule-1 (KIM-1) [23], osteopontin [24], netrin-1 [25], calprotectin [26], liver-type fatty acid-binding protein (L-FABP) [27], N-acetyl-β-D-glucosaminidase [28], and others. For a detailed summary of damage markers, we recommend the article by Schrezenmeier and colleagues [4] and the 2020 published “Consensus Statement on Acute Kidney Injury Biomarkers” by Ostermann et al. [29]. The third group is represented by two markers so far, the protein dickkopf-3 (DKK-3) and the product of urinary tissue metalloproteinase-2 and insulin-like growth factor-binding protein-7 ([TIMP-2]_U • [IGFBP7]_U). DKK-3 was evaluated in patients that received cardiac surgery [30]. Pre-surgery urinary DKK-3 concentrations were independently predictive for post-surgery AKI. The urinary product of TIMP-2 and IGFBP7 on the other hand has meanwhile been studied under numerous clinical circumstances. More than 160 references appear if the following search terms are applied: “TIMP-2” AND “IGFBP7”. A commercially available kit for routine use has been available for longer than 6 years now [31]. Regarding the robust limitations of serum creatinine, one may ask about the role of selected biomarkers in clinical practice. Ostermann and colleagues [29] published “Recommendations on Acute Kidney Injury Biomarkers’ (Acute Disease Quality Initiative Consensus Conference), which provided 11 consensus statements including a new proposal for AKI definition. The latter differentiates between 7 stages (1 S, 1/2/3 A or B). The numbers indicate the relative rise in serum creatinine and/or the reduction in urine output, whereas the letters stand for either negative (A) or positive (B) damage biomarker findings. Which types of damage markers have to be preferred remains debatable. Some candidates will most likely be incorporated in future KDIGO definition criteria [3], although final decisions are pending. At the moment, serum creatinine cannot completely be replaced by either one single marker or an individual combination of markers. Table 1 summarizes the most important characteristics of selected functional/damage/stress biomarkers.

Early AKI diagnosis is important, but two additional AKI-related outcome parameters are of fundamental importance also: in-hospital survival and ROKF. The current article is intended to provide information about biomarker-based prediction of in-hospital survival and kidney recovery. In the following two sections, references will be placed according to the release date (earlier > later). The discussion of every study will close with a short paragraph on availability and potential impact on daily clinical practice, respectively.

The following databases were screened: PubMed, Web of Science, Cochrane Library, Scopus. The period lasted from 2000 until 2022. The following terms were applied: “AKI” AND “biomarker” AND “survival” OR “mortality” OR “recovery of kidney function” OR “renal recovery” OR “kidney recovery”. Also, the following terms were used for additional literature search: “TIMP-2” AND “IGFBP7” and “RNA biomarker” AND “hematology”. Finally, only those studies were selected that addressed in-hospital mortality.

The following paragraphs are related to certain groups of biomarkers, respectively. The two aspects survival and ROKF will be discussed in each paragraph. A common definition of “ROKF” is not available yet. The criteria used in each of the following studies will be prepended.

A 2010 published study by Ferguson and colleagues [27] measured urinary L-FABP in 92 patients with established AKI and in 68 controls. The control group consisted of n = 29 healthy volunteers and n = 13 critically ill non-AKI subjects. It also included 26 patients that received cardiac catheterization for diagnostic purposes. AKI subjects showed the highest urinary L-FABP of all groups. The protein was predictive concerning the composite endpoint of death and KRT (kidney replacement therapy) (p = 0.03) but not regarding in-hospital death alone (p = 0.26). The authors also compared L-FABP with other markers (NGAL, KIM-1, interleukin-18 [IL-18]) and found a diagnostic performance level of the protein comparable to that of NGAL and KIM-1. Availability and potential impact on daily clinical practice: the study indicates a potential role of L-FABP in the prediction of in-hospital death/KRT. However, assays for L-FABP quantification have not been established in most hospitals yet. The same applies for the majority of markers discussed in the article.

In 2012, Nickolas and colleagues [32] published the results of a prospective cohort study, which analyzed the predictive ability of 5 urinary biomarkers: NGAL, KIM-1, L-FABP, IL-18, and cystatin C. More than 1,600 individuals were included, and urine was collected and analyzed at the time of emergency department admission. Three study sites participated (one center in Berlin, Germany; two centers in New York City, USA). The ethnic composition varied between study sites. All included individuals from Berlin were of Caucasian ethnicity. The following distributions were reported from the two other sites: Staten Island University Hospital – 7% Hispanic, 9% black, 79% white, 5% other; Allen Hospital of New York-Presbyterian Hospital – 52% Hispanic, 21% black, 27% white. Finally, 1,234 (75.5%) individuals were assigned to 1 out of 4 groups: normal kidney function, stable CKD, pre-renal AKI, and intrinsic AKI. Urinary NGAL showed the highest AUC-ROC regarding the discrimination between intrinsic AKI and all other AKI types. Logistic regression analysis identified both NGAL and KIM-1 as predictive for dialysis/in-hospital death, particularly if they were considered in combination with serum creatinine. Availability and potential impact on daily clinical practice: the most important finding was the ability to distinguish patients with low from those with intermediate risk of in-hospital death/KRT by the analysis of serum creatinine and either urine NGAL or urine KIM-1 combined. Thus, NGAL and KIM-1 potentially serve as survival predictors in AKI. At least NGAL analyzes are increasingly provided by clinical laboratories.

Primary endpoint in the RUBY study [33] was the development of persistent AKI stage 3 (KDIGO), lasting for 72 h or longer. The trial was designed in a multicenter, prospective, and observational manner; it included ICU-treated subjects with AKI stages 2 and 3. Finally, 331 out of 364 subjects underwent endpoint analysis, and 33% reached the primary endpoint. Serum and urine samples collected at study inclusion were employed for the quantification of certain biomarker molecules such as NGAL (urine), KIM-1 (urine), cystatin C (plasma), C-C motif chemokine ligand 14 (CCL14 – urine), and others. Elevated urine CCL14 was the strongest endpoint predictor of all tested proteins, even superior to KIM-1, cystatin C, and NGAL, respectively. Availability and potential impact on daily clinical practice: the data indicate a substantial role of CCL14 in post-AKI recovery prediction at least under intensive care conditions. Routine test is missing yet.

A common hallmark of all presented studies so far was the definition of “in-hospital survival” as distinct outcome variable. It must however be mentioned that other biomarker-related studies evaluated survival rates even after longer time periods (months to years after AKI onset): Coca and colleagues [34] for instance published a prospective long-term follow-up study already in 2014. Almost 1,200 individuals that received cardiac surgery were included; urinary NGAL, IL-18, KIM-1, and L-FABP were quantified at days 1–3 post-surgery. The concentrations of all four molecules were independently associated with increased risk of mortality in AKI subjects. Survival rates were documented until year 3 after discharge.

In 2020, Fiorentino and colleagues [35] published an investigation performed in patients with sepsis/septic shock (secondary analysis from the ProCESS trial cohort [36]). Urinary TIMP-2 • IGFBP7 was measured before and after (6 h) resucitation from septic shock; 688 individuals were included. The mean age was 62 years (50–74 IQR), 367 were females. The following criteria defined the primary (composite) endpoint: AKI stage 3 according to KDIGO [1], KRT, death within 7 days. Finally, 113 patients reached the endpoint. Patients with initially negative TIMP-2 • IGFBP7, who showed an increase of the product of more than 0.3 U after treatment, were at three-times higher risk in comparison to patients who remained negative. Positive patients after resucitation had worse clinical outcomes whether they were diagnosed with AKI or not. Availability and potential impact on daily clinical practice: a commercially available kit for TIMP-2 • IGFBP7 routine measurement has been available for longer than 6 years now [31]. The urinary product of TIMP-2 and IGFBP7 is potentially helpful not only for earlier AKI recognition but also for risk prediction in sepsis 7 septic shock.

The predictive power of urinary TIMP-2 • IGFBP7 was also studied by Koyner et al. [37], this time regarding the long-term (9 months) outcome of critically ill subjects. The primary endpoint was a composite of all-cause mortality and the need for KRT. 382 out of 692 included patients developed AKI stage 1 according to KDIGO (6). In the univariate analysis, a TIMP-2 • IGFBP7 of >2.0 was associated with an increased risk of reaching the composite endpoint. Multivariate analysis however revealed a TIMP-2 • IGFBP7 of >0.3 to be predictive of death or RRT exclusively in subjects who suffered from AKI. It needs to be noted that all urinary TIMP-2 • IGFBP7 were measured under baseline conditions.

A 2021 published study by Xia and colleagues [38] analyzed the predictive role of the fibrinogen level to the albumin level ratio (FAR) in over 5,000 critically ill patients with AKI. The study group consisted of 2,066 females and 2,935 males. The mean age was 63 ± 16 years. Patients belonging to the highest FAR quartile more often presented a history of atrial fibrillation, renal disease, pneumonia, malignancy, diabetes, and others. Also, they showed higher values of leukocytes, blood urea nitrogen, a higher anion gab, higher platelet counts, and a higher risk of mortality. Earlier, the FAR had been analyzed in other diseases such as malignant disorders [39‒41] and myocardial infarction [42]. The study was based on data extracted from the “Multiparameter Intelligent Monitoring in Intensive Care Database III”; it aimed to evaluate the association between FAR and the in-hospital mortality in a nonlinear manner. Higher FAR was identified as an independent mortality predictor. The findings were intriguing so far that all FAR were employed from the time of ICU admission. In addition, both proteins, fibrinogen, and albumin are measured on an almost regular basis in ICU-treated subjects since they are usually needed for other clinical decisions also. Availability and potential impact on daily clinical practice: the study findings are intriguing in so far that both proteins are measured in the majority of patients at ICUs; the costs are almost neglectable. The FAR may be used as additional marker for a more sophisticated risk profiling of AKI subjects at the ICU.

Sun and colleagues [43] performed so-called liquid chromatography coupled to mass spectrometry (LC-MS) analyses in critically ill patients, recruited from “Veteran’s Affairs/National Institutes of Health Acute Renal Failure Trial Network study.” This particular trial widely known as acute tubular necrosis (ATN) study was published in 2008 [44]. It included more than 1,100 critically ill AKI patients that suffered from failure of one or more nonrenal organs or from sepsis. Participants either received so-called intensive or less-intensive KRT. Finally, intensive KRT did not improve endpoints in a significant manner. The study by Sun et al. selected a total number of 202 individuals from the ATN study cohort. Blood samples from days 1 and 8 were analyzed by LC-MS [39]. The procedure allows the detection of nonvolatile and complex molecules. The study revealed mortality predictive models, based on differences between four serum metabolites (1-arachidonoyl-lysoPC, 1-eicosatetraenoyl-lysoPC, tyrosine, 5′-methylthioadenosine) at day 1 and between 11 metabolites at day 8 (regarding survival at days 8 and 28, respectively). As opposed to the FAR, LC-MS procedures have not been established for daily clinical routine yet. Availability and potential impact on daily clinical practice: as mentioned, metabolomics are predominantly used under experimental conditions and not in clinical routine. However, “-omics” studies in general allow the identification of potentially promising new diagnostic and predictive markers in AKI and in numerous other conditions.

Recently, our group submitted a study, performed on subjects with in-hospital diagnosed AKI [45]. Individuals were not exclusively recruited from the ICU. All patients underwent serum analysis at the time of AKI diagnosis, which was made if KDIGO criteria 1 or 2 [1] were fulfilled (in-hospital AKI alert system). In total, 151 subjects were included in the study (females 62, males 89). Their mean age was 74.9 +/−13.4 years. The study cohort was heterogenous in terms of AKI etiology (sepsis – 23.6%; pre-renal – 23.6%; cardiorenal – 20.1%; contrast-associated – 12.5%; obstruction – 0.7%; and others). In-hospital death occurred in 16.6%, KRT became mandatory in 39.7%, and ROKF was diagnosed in 72.2%. Since previous data indicated a diagnostic role of serum IL-33 in glomerular diseases [46], both serum IL-33 and the circulating isoform of its receptor (sST2) were quantified. While IL-33 quantification did not reliably succeed in many subjects, sST2 in contrast significantly differed between AKI subjects with versus without survival (lower in survivors). In addition, sST2 was higher in patients that required transient ICU therapy/ventilatory treatment/vasopressor infusion. Thus, sST2 may serve as AKI survival predictor in the future. Interestingly, all analyzes were performed with blood samples drawn at the time of AKI onset. Availability and potential impact on daily clinical practice: sST2 can potentially help to identify AKI patients at risk of a more severe disease course under acute circumstances. sST2 quantification has of course not been implemented in daily routine diagnostics. Also, additional data are needed, particularly from patients exclusively treated at the ICU.

Another long-term study focused on two pro-inflammatory cytokines, interleukin-6 and -10 [47]. Included patients also received cardiac surgery; survival was also documented over several post-operative years (median: 3 years). While pre-operative cytokine concentrations were not associated with increased AKI or mortality risks, post-surgery levels were indeed. For instance, patients belonging to the second tertile (176–323 pg/mL) of peak IL-6 were at significantly lower risk of death. The study did however not compare the mortality-predictive roles of IL-6/-10 in AKI subjects as compared to non-AKI patients.

The “Kidney in Sepsis and Septic Shock” (Kid-SSS) study from 2018 [19] was designed as a substudy of the “Adrenomedullin and Outcome in Severe Sepsis and Septic Shock (AdrenOSS)” investigation [48]. A validation cohort was part of the FROG-ICU trial [49]. The following definition of persistent AKI (no ROKF) was provided: persistent elevation of serum creatinine by > 1.5-fold or by ≥0.3 mg/dL (26.5 mmol/L) at day 7 (or discharge, if earlier), the ongoing need or RRT at day 7, or death within 7 days after ICU admission. The principal aim of the Kid-SSS trial was to evaluate the predictive power of proenkephalin A 119–159 (penKid) in septic patients. PenKid serves as a marker of glomerular function; samples for analyses were drawn within 24 h after ICU admission. It was finally significantly higher in subjects with persistent AKI. Availability and potential impact on daily clinical practice: penKid is already provided for routine diagnostics by some in-hospital laboratories. Its potential role in recovery prediction may not be underestimated, especially since study participants received penKid analysis early during the ICU treatment course.

In a 2021 published, prospective, multicenter study [50], ROKF was defined as discontinuation of KRT at day 28 after study inclusion. The ATN study focused on critically ill patients with severe AKI (defined as ATN including failure of one or more nonrenal organs). A total number of 1,124 patients was recruited, 827 provided at least one serum sample. On day 28, 626 individuals were alive, and 343 did not require any further KRT. Serum samples from both groups, dialysis-dependent (n = 34) and -independent (n = 38) patients, were employed for proteomic analysis. The latter encompassed over 1,300 proteins. Patients with ROKF showed higher levels of the following proteins: CXCL11, CXCL2/CXCL3, CD86, Wnt-7a, BTK, c-Myc, TIMP-3, CCL5, ghrelin, PDGF-C, survivin, CA2, IL-9, EGF, and neuregulin-1 and lower levels of CXCL16, IL1RL1, stanniocalcin-1, IL-6, and FGF-23. In summary, the study helped to identify several new candidates for ROKF prediction in RRT-requiring AKI. Availability and potential impact on daily clinical practice: almost all tested biomarkers are not available for routine diagnostics. However, the study shows that decisions about complex processes, such as ROKF, most likely require simultaneous analyses of several markers or marker panels instead of one individual protein or peptide.

Finally, it needs to be mentioned that “conventional parameters” such as proteinuria/albuminuria have been identified as AKI risk factors also. Grams et al. [51] analyzed more than 1.2 million subjects (8 distinct cohorts) and showed associations between (reduced) estimated glomerular filtration rate and (higher) albumin-to-creatinine ratio and AKI risk. Comparable observations were reported by Park et al. [52] and Yang and colleagues [53] who identified pre-operative dipstick albuminuria as AKI risk factor. The study by Park et al. also showed higher dipstick albuminuria as risk factor of death during the following year. Regarding the fact that every individual AKI episode increases the risk of death in the short and in the long term and also the risk of CKD in subsequent years, albuminuria must be classified as AKI survival predictor. However, none of the studies evaluated associations between albuminuria and in-hospital survival. Table 2 and Figure 1 summarize the presented studies.

Until now, most studies that investigated the role/usability of new AKI biomarker molecules were primarily designed to identify diagnostic parameters. On the other hand, AKI risk assessment is comparably important. Which patients are at higher risk for death? Which patients will most likely require KRT? Particularly this aspect, which has not specifically been addressed in the current article, has intensively been discussed by nephrologists all over the world for decades. Finally, an early prediction of kidney recovery would help to identify patients that need the utmost attention of nephrologists during follow-up. Persistent AKI, officially defined as acute kidney disease if renal function is impaired for longer than 7 days after an acute event [54], may increase the CKD risk even further. Currently, a clear definition of “ROKF” is still missing. In summary, more biomarker-related studies should address (i) the issue of in-hospital mortality prediction and (ii) follow-up or outcome characteristics in AKI.

The authors declare that they have no conflicts of interest.

No funding was provided.

Daniel Patschan wrote the article. Stefan Erfurt, Stefanie Oess, Martin Lauxmann, Susann Patschan, and Oliver Ritter searched for literature and assisted in writing. Meike Hoffmeister designed the article, searched for literature, and assisted in writing. All authors agreed to publish the final version of the manuscript.