Introduction

Acute kidney injury (AKI) is a common and severe clinical syndrome that occurs in approximately 0.15 of hospitalized patients, and its morbidity has been reported in more than half of patients in the intensive care unit (ICU), with a high mortality rate [1,2,3]. Currently, the diagnosis of AKI is still based on the rapid increase in serum (or plasma) creatinine, a decrease in urine volume, or both [4,5,6]. However, the concentration of creatinine is affected by multiple clinical variables, such as hydration, nutritional status, muscle metabolism, or drug effects, and serum creatinine does not alter until the glomerular filtration rate (GFR) is reduced by about 0.50 [7, 8]. Owing to these limitations, creatinine remains an insufficient predictor of AKI. As a result, there is an urgent need for an accurate and timely biomarker to predict the occurrence or progression of AKI.

Soluble urokinase plasminogen activator receptor (suPAR) is a novel biomarker. SuPAR molecules are derived from the proteolytic cleavage of membrane-bound urokinase plasminogen activator receptor (uPAR, cd87), and can be detected in different body fluids, such as blood, urine, peritoneal fluid, and cerebrospinal fluid. Moreover, suPAR is highly stable, and the concentration of suPAR in serum is not affected by diet, drugs, inflammation, and time of collection throughout the day [9]. SuPAR is a signal transduction glycoprotein that is believed to be involved in the pathogenesis of kidney diseases [10]. Previous studies have shown that the systemic inflammatory biomarker suPAR is an important biomarker for the early identification of AKI [11, 12]. Hayek et al. have recently found that elevated suPAR is associated with an increased risk of AKI in patients undergoing coronary angiography or cardiac surgery and in those hospitalized in intensive care units [10]. However, there is still no evidence that suPAR can be used as a marker to predict the occurrence of AKI, or that high suPAR levels are associated with AKI. Therefore, we conducted this systematic review and meta-analysis to evaluate the predictive value of suPAR for AKI.

Materials and methods

Literature retrieval

The search databases for this study were EMBASE, Cochrane Library, PubMed, and Web of Science, and the search time was as of December 2021. The search strategy was a combination of subject headings and free words, and minor adjustments might be made to each database. The search terms in PubMed included “Acute Kidney Injury” [Mesh], and “Receptors, Urokinase Plasminogen Activator [Mesh]”. The detailed retrieval strategy is shown in Appendix 1. The present study has been registered on PROSPERO (Registration No. CRD42022324978).

Literature inclusion and exclusion criteria

Inclusion criteria: (1) The literature type was cohort study, cross-sectional study, or diagnostic trial study; (2) English literature; (3) Studies containing the gold standard for the definite diagnosis of AKI and non-AKI cases; (4) Studies in which the four-grid diagnostic table could be directly or indirectly extracted from outcome indicators.

Exclusion criteria: (1) Non-clinical studies, such as reviews, medical records, and conference abstracts; (2) In vitro or animal experiments; (3) Studies from which the diagnostic four-grid table could not be directly or indirectly extracted.

Literature screening and data extraction

The EndNote software was used for literature management, and the literature was screened according to the inclusion and exclusion criteria. After the final included studies were determined, information extraction was performed, including the first author, country, sample size, AKI occurrence background, number of true positives, number of false positives, number of true negatives, number of false negatives, sensitivity, specificity, and ROC curve.

The literature screening and information extraction were carried out independently by two researchers (Y.H. and S.C.H.), and cross-examination was conducted after completion. If there was any disagreement, a third researcher (X.Y.Z.) was invited to assist in adjudication.

Quality evaluation

In the present study, two investigators (Y.H. and S.C.H.) independently used QUADAS-2 to assess the methodological quality of included studies. They cross-checked the results after the evaluation was completed. If there was any disagreement, a third investigator (X.Y.Z.) assisted in adjudication. The application of the QUADAS-2 has four phases: summarize the review question, tailor the tool and produce review-specific guidance, construct a flow diagram for the primary study, and judge bias and applicability.

Statistical analysis

A complete meta-analysis was performed using Stata 15.0 (Stata Corporation, College Station, TX), and a bivariate mixed model was used to pool effect sizes. The combined sensitivity, specificity, positive likelihood ratio, negative likelihood ratio, diagnostic odds ratio with 95% confidence interval (95%CI) were analyzed. A sensitivity analysis was performed to evaluate the stability of meta-analysis results, and Deek’s funnel plot asymmetry test was adopted to assess publication bias. Owing to the small number of included studies, we did not perform meta-regression. In this study, p < 0.05 indicated that the difference was statistically significant.

Results

Retrieval results

A total of 215 relevant studies were retrieved, 157 of which were left after removing duplicate or irrelevant literature. According to the title and abstract, 125 studies that did not meet the criteria were excluded. After a full-text review of the remaining 32 studies, 7 studies were finally included in this meta-analysis. The literature screening process is shown in Figure 1.

Fig.1
figure 1

Flow chart of literature screening

Full size image

Basic information

The seven included studies were published between 2017 and 2021, with patients from four countries and multiple ethnicities, including two from Denmark [13, 14], three from Germany [12, 15, 16], one from China [17], and one from the United States [18]. The study design was mainly retrospective. A total of 2319 patients were included, comprising 635 AKI patients and 1684 non-AKI patients, and most of them were middle-aged and elderly. Most trials were done on critically ill patients, including those who had undergone cardiac surgery (extracorporeal circulation), elderly patients in the emergency department, patients in the ICU after colostomy or with sepsis, and hospitalized adults with COVID-19. The basic characteristics of all the included literature are presented in Table 1.

Table 1. Basic characteristics of the included literature
Full size table

Methodological quality evaluation

According to the evaluation items of QUADAS-2, the risk of bias was assessed in four domains: case selection, trial to be evaluated, reference standard, and flow and timing. The clinical applicability of the first three domains was evaluated at the same time. The risk of bias assessment showed that the quality of all included studies was relatively good, and the high risk items were mainly in the case selection domain, as shown in Figure 2.

Fig. 2
figure 2

Literature quality evaluation (Red indicates a high risk of bias; Yellow indicates insufficient information for risk assessment; Green indicates a low risk of bias)

Full size image

Meta-analysis results

The pooled sensitivity of suPAR in predicting AKI was 0.77 (95%CI 0.67–0.84); the pooled specificity was 0.64 (95%CI 0.53–0.75); the pooled predictive odds ratio was 6 (95%CI 3–10); the pooled positive likelihood ratio was 2.2 (95% CI 1.6–2.9); the pooled negative likelihood ratio was 0.36 (95% CI 0.26–0.52), and the area under the summary receiver-operating characteristic (SROC) curve was 0.77 (95% CI 0.12–0.99). The sensitivity, specificity, and SROC are presented in Figure 3.

Fig. 3
figure 3

Sensitivity, Specificity and Pooled SROC (A Forest plot of sensitivity and specificity; B: SROC curve of the meta-analysis)

Full size image

In this meta-analysis, the heterogeneity I2 was 0.93 (95% CI 88–99), and the box plot revealed that the studies of Anne Byriel Walls [13] and Yuhan Qin [17] might be the main sources of heterogeneity. In addition, Deek’s funnel plot showed no apparent publication bias among the studies (p = 0.06). The bivariate boxplot for heterogeneity is shown in Figure 4A, and Deek’s funnel plot is shown in Figure 4B.

Fig. 4
figure 4

Heterogeneity and Deek’s Funnel plot (A bivariate boxplot for heterogeneity. Studies outside the shadow may introduce major heterogeneity. B Deek’s Funnel plot for publication bias)

Full size image

Fagan’s plot was employed to reflect the clinical applicability of suPAR in predicting AKI. The incidence of AKI varies significantly in different contexts and even at different income levels. It is estimated to range from 0.01 to 0.66, and exceeds 0.50 in the ICU [1, 19]. Therefore, assuming a prior probability of 0.27, the probability of the diagnosis of AKI is 0.46 in the case of a positive likelihood ratio of 2, and the probability of the diagnosis of no AKI is 0.12 in the case of a negative likelihood ratio of 0.35. The clinical application is shown in Figure 5.

Fig. 5
figure 5

Clinical application

Full size image

Discussion

This meta-analysis showed that suPAR could be used as a predictor of AKI, with a comprehensive sensitivity of 0.77 and a comprehensive specificity of 0.64. There was no significant publication bias among the included studies. The prediction effect of suPAR seemed to be different in various AKI occurrence backgrounds. Among the seven included articles, the sensitivity of serum suPAR in the diagnosis of AKI ranged from 0.63 to 0.90, and the specificity varied from 0.40 to 0.83, with significant difference. This may be attributed to the differences in the research design and experimental methods in response to various population levels or diseases, resulting in different detection thresholds. The testing instruments and diagnostic reagents used in different countries were not identical, and differences also existed at the operator level. The box plot for heterogeneity reflected that the studies of Anne Byriel Walls [13] and Yuhan Qin [17] might be the main sources of heterogeneity.

In the recent years, the discovery and application of new biomarkers for early diagnosis of AKI have become one of the hot spots in kidney disease research. A previous study showed that plasma suPAR was an incidental phenomenon associated with the inflammatory shedding of receptors on neutrophils, monocytes, and macrophages [20]. A mechanistic study on models of kidney disease showed that circulating suPAR (derived from inflammatory cells) interacted with αvβ3 integrins on podocytes, and elevated plasma suPAR levels could predict the occurrence of nephropathy in seemingly healthy individuals and those at risk of chronic kidney disease [21]. This suggested that suPAR played a role in the pathogenesis of AKI. Targeting suPAR has therapeutic promise as it is pathogenic. In experimental mouse models, the use of anti-suPAR MABs eliminated the adverse effects of suPAR on the kidney, indicating that suPAR is a promising therapeutic target for alleviating AKI [10, 22, 23]. Therefore, eliminating suPAR from circulation or neutralizing its biological effects may be a reasonable strategy to reduce the incidence, morbidity, and mortality in AKI. In clinical trials, biomarkers such as suPAR have the advantage of identifying AKI at an early stage and predicting AKI [24].

In this study, Fagan’s plot was used to reflect the clinical applicability of suPAR in predicting AKI. Assuming that the probability before diagnosis is 0.50, the probability of the diagnosis of AKI is 0.68 in the case of a positive likelihood ratio of 2, and the probability of the diagnosis of no AKI is 0.27 in the case of a negative likelihood ratio of 0.36. Most importantly, if AKI can be predicted, there will be more testable therapeutics, since the majority of AKI interventions identified in preclinical studies are effective only when implemented before injury[25]. The current biomarkers under study, such as neutrophil gelatinase-associated lipid calin (NGAL), kidney injury molecule 1, urinary IL 18, and plasma cystatin C, are also early markers of AKI [26]. Studies have found that ROC analysis of suPAR and NGAL yielded an area under curve (AUC) of 0.69 and 0.78, respectively, without a significant difference (p = 0.117). The AUC of suPAR combined with NGAL was 0.80, significantly higher than that of suPAR alone (p = 0.032) [13]. Furthermore, tissue inhibitor of metalloproteinases-2 (TIMP-2) and insulin-like growth factor-binding protein 7 (IGFBP7) were found to be promising biomarkers for risk stratification in infectious AKI patients requiring renal replacement therapy (RRT) and predictors of AKI in out-of-hospital cardiac arrest survivors [27]. Compared with [TIMP-2] × [IGFBP7], baseline suPAR values have already been reported to have the ability to predict the demand for RRT with good diagnostic accuracy. Urinary [TIMP-2] × [IGFBP7] can be used as an early predictor of moderate and severe AKI as well as a potential tool to monitor the treatment of Kidney-oriented sepsis [28, 29]. Therefore, combining multiple markers can more effectively predict AKI and provide more accurate predictive value for clinical practice. Several studies have demonstrated that using these biomarkers in combination with recommended AKI care practices (e.g., avoidance of nephrotoxins and optimization of hemodynamics) can improve patients’ prognosis [28, 30].

This study has the following advantages. First, it is the first to discuss the predictive value of suPAR for AKI. The analysis results show that suPAR or its related indicators should be considered in the subsequent development or optimization of AKI prediction tools/scoring systems. Second, the quality of the included literature is ideal, and there is no bias in the results based on the data. Meanwhile, this study also has the following limitations. On the one hand, although we conducted a comprehensive and systematic search, there are still few studies included in this meta-analysis. On the other hand, with a limited number of included studies, there is no sufficient evidence to support our discussion on the predictive effect of AKI under different occurrence backgrounds.

Conclusion

SuPAR may be a valuable biomarker for predicting AKI, suggesting that suPAR or its related indicators should be considered in subsequent development or optimization of predictive tools/scoring systems for AKI. Although a comprehensive search was performed, the number of included articles is relatively small. Large-scale studies are desired to verify our findings in the future.