The value of urinary interleukin-18 in predicting acute kidney injury: a systematic review and meta-analysis

Abstract Aims The aim of this study was to systematically review relevant studies to evaluate the value of urinary interleukin-18 (uIL-18) in predicting acute kidney injury (AKI). Methods A comprehensive search of PubMed, Medline, Embase, and Cochrane Library was conducted for literature published up to 1 August 2022. Quality Assessment Tool for Diagnostic Accuracy Studies-2 (QUADAS-2) was applied to assess the literature quality. Then, relevant data were extracted from each eligible study and a random-effects regression model was utilized to pool sensitivity, specificity, and construct summary receiver operating characteristic (SROC) and area under curve (AUC). Results Twenty-six studies with 7183 patients were enrolled and relevant information was extracted. The estimated sensitivity and specificity of uIL-18 in the diagnosis of AKI were 0.64 (95% confidence interval (CI): 0.54–0.73) and 0.77 (95%CI: 0.71–0.83), respectively. The pooled diagnostic odds ratio (DOR) was 6.08 (95%CI: 3.63–10.18), and the AUC of uIL-18 in predicting AKI was 0.78 (95%CI: 0.74–0.81). Subgroup analysis showed that uIL-18 in pediatric patients was more effective in predicting AKI than in adults (DOR: 7.33 versus 5.75; AUC: 0.81 versus 0.77). Conclusions Urinary IL-18 could be a relatively good biomarker with moderate predictive value for AKI, especially in pediatric patients. However, further research and clinical settings are still needed to validate our findings.


Introduction
Acute kidney injury (AKI) is mainly manifested by the rapid decline of renal function in a short period of time and the accumulation of metabolic waste with a high morbidity and mortality [1,2]. It is reported that the incidence of AKI was 21.6% and the mortality rate related to AKI was 23.9% in adults [3]. A multicenter epidemiological study showed that the prevalence of AKI was 26.9% in critically ill children [4]. The early diagnosis and intervention for AKI not only provides better treatment options, but also improves patient prognosis [5].
At present, increased serum creatinine (SCr) and decreased urine volume are the main clinical indicators of AKI. However, due to the influence of muscle mass, diuretics and other factors, the changes of SCr and urine volume were unstable and lagged [6,7]. Both SCr and urine volume have limitations in the timely and accurate identification of decreased renal function, which may delay the diagnosis of AKI thus leading to the poor prognosis. The increase of SCr is mainly caused by the decrease of glomerular filtration rate (GFR) and usually could be observed until 2-3 days after the AKI occurrence [8]. All these facts demonstrated the urgent need for more effective predicting of AKI, and an increasing number of studies have been conducted in order to provide more information about this topic [9]. Several new biomarkers in serum and urine have been reported to show the potential for predicting AKI, including cystatin C (Cys C), neutrophil gelatinase-associated lipocalin (NGAL), interleukin-18 (IL-18), kidney injury molecular-1 (KIM-1), and so on [10,11]. Among them, IL-18 has aroused widespread attention. IL-18 is a member of the IL-1 family of cytokines, which is synthesized as an inactive 23 kDa precursor by monocytes, macrophages, and proximal renal tubular epithelial cells. In animal models, IL-18 could be activated by caspase-1, thus inducing ischemic injury and inflammation in the proximal renal tubule, then excreted into urine [12,13]. Previous studies about urinary IL-18  in predicting AKI suggested that uIL-18 may act as a better indicator of decreased renal function compared with SCr [14][15][16][17]. Its expression was upregulated in the early stage of AKI and could be detected well earlier than SCr, thus showing its potential in predicting AKI [18,19]. Although an increasing number of related studies have been conducted recently, additional clinical research and trials are still needed to validate the application of uIL-18 in different clinical settings [20].
With the accumulation of evidence, conflicting results have raised concerns about the predictive value of uIL-18 for AKI. To further clarify the performance and clinical detection value of uIL-18, we conducted a systematic review and meta-analysis based on 24 original articles, which help to evaluate its role in the early detection of AKI.

Data Source and search strategy
This meta-analysis was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [21]. A comprehensive search of PubMed, Medline, Embase, and Cochrane Library was conducted to identify relevant articles published up to 1 August 2022. The search strategy was applied to identify all literatures with following keywords: 'IL-18' or 'interleukin 18' plus 'acute kidney injury' or 'acute renal failure' or 'AKI' or 'ARF'. The details of search methods have been provided in Supplemental Item 1. In addition, we also checked the references of relevant studies and reviews manually to identify other potentially eligible studies. The search was preformed independently by two investigators (ZQ and HL).

Study selection
We encompassed all articles and conference papers retrieved without language or sample size restrictions. Studies were assessed and selected based on titles and abstracts by two independent reviewers (ZQ and HL) using the Endnote bibliography manager blinded to the authors and institute of studies and then retrieved and rescreened full-text articles. Further, the citations of reviewed full-text articles were also checked to avoid missing additional relevant studies. Any conflicts were resolved by a third reviewer (JG, blinded to the authors and institute of studies). The exclusion criteria were as follows: (1) review articles or duplicate papers. (2) Animal or in vitro based studies. (3) Studies could not provide the diagnosis value of uIL-18. Although there were no language restrictions initially, for the full-text review and data extraction, only articles in English language were included for the further analysis.

Data extraction and quality assessment
After a detailed full-text review of each study, two independent reviewers (ZQ and JP) utilized a standardized form to extract data from the retained studies. Disagreements were resolved by discussion and assessed by a third reviewer (PJ) until a consensus was reached. Extracted data were as follows: (1) research information: first author, year of publication, original country, sample size, study design, and population setting; (2) patient characteristics: age, sex, and baseline IL-18; (3) information of AKI: definition of AKI and number of patients who developed AKI; (4) information of IL-18: the time of obtaining specimen, assay method, and sample storage; (5) cutoff value for urinary IL-18, specificity, sensitivity, and area under the ROC curve (AUC) with 95% confidence interval (CI). The true positive (TP), true negative (TN), false positive (FP), and false negative (FN) results were calculated. If there is more than one cutoff value in a study, the cutoff value with the highest product of specificity and sensitivity was used.
Quality assessment was performed by two independent reviews (ZQ and HL) using the Quality Assessment Tool for Diagnostic Accuracy Studies-2 (QUADAS-2) [22]. Any discrepancies were resolved by a third reviewer (PJ).

Statistical analysis
All statistical analysis was performed by STATA version 14.0 (Stata Corp., College Station, TX) using the MIDAS module [23]. The kappa statistic was utilized to evaluate the agreement between two investigators in study selection. A bivariate random-effects regression model was used to calculate the pooled sensitivity (SEN), pooled specificity (SPE), the positive likelihood ratio (PLR), the negative likelihood ratio (NLR), and the diagnostic odds ratio (DOR) with their 95% CIs based on DerSimonian-Laird method [24]. A hierarchical summary ROC curve was constructed and the area under the curve was calculated to assess the overall diagnostic accuracy [25]. Deek's funnel plots were constructed to evaluate the publication bias and p < 0.1 was believed representative of statistically significant publication bias. The I 2 index was calculated to detect the heterogeneity between studies. The value of I 2 lied between 0% and 100%, a value of 0% suggested no observed heterogeneity, and I 2 values above 50% were regarded as indicative of substantial heterogeneity. Meta-regression was used to explore the source of heterogeneity. In addition, we conducted a subgroup analysis according to patients age (adults or pediatrics), predictive time ( 12 h or >12 h), the time of obtaining specimen (admission or other times) and the population setting (cardiac surgery patients or others), publication date (before or after 2009), AKI definition (using standardized definition or not) and population continents (Asia, North America, and others) to explore the potential sources of heterogeneity between included studies.

Search results and study characteristics
A total of 759 publications from different databases were retrieved upon the initial search. First, 176 studies were excluded after duplicates were identified. Then, we screened the titles and abstracts of the remaining 583 articles. Five hundred and twenty-seven of them were excluded because they are conference abstract, animal research, review, or non-English article. There were 56 studies selected for full-text review, and 30 studies were removed due to the incomplete data extraction. The kappa statistic following abstract review was 0.82 (p < 0.05), indicating substantial agreement between reviewers. The details of data for kappa statistic are shown in Supplemental Table 1. Finally, 26 studies with 7183 patients were included in this metaanalysis [20, (Figure 1).
The characteristics of included studies are shown in Table 1. All these studies were published from 2005 to 2021, varied in country, study design, sample size (from 40 to 1493), and population settings. It is worth noting that these studies defined AKI differently. In these 26 studies, 21 studies focused on adults, and the left five studies focused on the pediatric patients (age <18 years old). Twelve studies were conducted on patients who underwent cardiac surgery, other population settings included intense care unit patients, burn patients, non-septic critically ill neonates, and so on. All of studies determined IL-18 concentration by enzymelinked immunosorbent assay (ELISA) and the samples were stored at À80 C.

Quality assessment and publication bias
The methodological quality of the studies according to the QUADAS-2 is shown in Table 2. All of the domains were considered to be at low or unclear risk. Deek's funnel plots are shown in Figure 2 with a p value of

Data synthesis
Data in the 26 eligible studies were extracted and shown in Table 3, including TP, FN, FP, TN, the time of obtaining specimen, the assess time, the optimal cutoff value for urinary IL-18 with their sensitivity, specificity, and AUROC (95%CI). The estimated sensitivity and specificity values of uIL-18 in the diagnosis of AKI were 0.64 (95%CI: 0.54-0.73) and 0.77 (95%CI: 0.71-0.83), respectively ( Figure 3). There was a substantial heterogeneity both in sensitivity and specificity, according to the I 2 value of 86.47% and 93.64%, respectively. DOR was 6.08 (95%CI: 3.63-10.18) and shown in Figure 4; PLR  Deek's funnel plot asymmetry test.

Subgroup analysis
We used meta-regression to explore the source of heterogeneity (Supplemental Table 2). Age, predictive time, obtaining specimen, population settings, and AKI definition contributed to the heterogeneity of specificity while we failed exploring the source of heterogeneity of sensitivity. Subgroup analysis based on the patients age, predictive time, the time of obtaining specimen, population setting, publication date, AKI definition and population continents was conducted. As shown in Table 4, the diagnostic value of urinary IL-18 was higher in pediatrics group than in adult group, with a DOR of 7. 33

Discussion
Early prediction and diagnosis of AKI is of great importance to improve prognosis of patients. The existing diagnostic indicators including decreased urine volume In this diagnosis meta-analysis, we conducted a comprehensive search to identify articles that evaluated the diagnostic performance of uIL-18 in predicting AKI. After screen and full-text review, a total of 26 studies with 7183 patients were included and relevant information was extracted. We found using uIL-18 in diagnosis AKI was considerable, with the estimated sensitivity, specificity, DOR, and AUC of 0.64 (95%CI: 0.59-0.73), 0.77 (95%CI: 0.71-0.83), 6.08 (95%CI: 3.63-10.18), and 0.78 (95%CI: 0.74-0.81), respectively. When DOR is greater than 1, a higher DOR indicates a better test performance [51]. Although this topic has been studied before by Lin et al. [52], it has been almost 10 years since the last publication, and they only enrolled 11 studies. Our study updated the included studies and demonstrated that uIL-18 showed a moderate diagnostic accuracy for AKI, while its sensitivity and specificity were poor, which is similar with previous results. In addition, we found that uIL-18 in pediatric patients was more effective in predicting AKI than in adults, reminding us the clinical potential in different population settings.  A potential explanation of this predictive value of uIL-18 for AKI maybe the relationship between IL-18 and inflammation. The expression of IL-18 could be upregulated as a part of cell inflammation damage by activating macrophages, promoting T cell differentiation and stimulating NK/T cell in releasing c-interferon. The expression of IL-18 also could be observed in other cell lineages besides the renal tubule cells, including macrophages, lymphocytes, intestinal epithelial cells, and fibroblasts. In the acute phase of disease or in the state of systemic inflammatory response, the systemic expression of IL-18 could be upregulated, which may affect the relationship between urinary IL-18 and AKI, thus affecting the accuracy of diagnosis for AKI in sensitivity and specificity [53][54][55]. AKI is a complex clinical condition that may cannot be completely predicted by a single biomarker. It was believed that the combination of biomarker and clinical risk stratification for the diagnosis of AKI would be the trend of future research [56,57].
There was a significant heterogeneity among the included studies, that suggested the application of uIL-18 in predicting AKI might be limited. To fully evaluate the diagnostic value of uIL-18 for AKI, we conducted subgroup analysis based on the patients age, predictive time, the time of obtaining specimen, population setting, publication date, and AKI definition. The diagnostic value of urinary IL-18 was higher in pediatrics group than in adult group, with a DOR of 7.33 (95%CI: 2.86-18.78) compared to 5.75 (95%CI: 3.17-10.42). We hypothesized that this may be related to the significant comorbidities in adults, such as hypertension, atherosclerotic disease, diabetes mellitus, etc. These complex comorbidities were much more common in adults and may affect the concentrations of urinary IL-18 through inflammation pathway. In addition, due to the children and adolescents exhibit different physiological states in renal mutations, their kidneys undergo a process of growth and maturation, the risk factors associated with AKI and the timing of kidney injury remain unclear [58]. Both the metabolic capacity and compensatory ability of children and adolescents also varied from those of adults, which may contribute to the different kidney injury condition and urinary IL-18 levels as well [59,60]. Similar results also have been reported before, which reminds us the need of a broad panel of biomarkers [61]. In addition, there was no significant reduction of heterogeneity in these two subgroups, which may be due to the different cutoff values, different AKI definitions, different clinical setting, and varied patient number among the included studies. As for the predictive time, the predictive time 12 h group was much more sensitive but less specific than predictive time >12 h group, and there was no significant difference in its predictive value. Obtaining specimen at non-admission time showed higher diagnostic value than obtaining specimen at admission, with a DOR of 7. The results of this study suggested that the diagnostic accuracy of uIL-18 in predicting AKI was moderate, but we could not determine the ideal cutoff value for uIL-18 because the range of cutoff values included in the studies varied widely from 6.39 pg/ml to 125 pg/ml.
In addition, there were three studies used IL-18 concentration standardized by urinary creatinine or absolute concentration. Given the factors we mentioned before, it is still difficult for us to determine a suggested cutoff point of uIL-18. At present, the consensus threshold of IL-18 for risk stratification for AKI has not been reached yet, while some investigators suggested a cutoff range of 100-500 pg/mg Cr [62]. Ralib et al. found that absolute concentrations showed the best predictive value for AKI at admission, while standardized concentrations performed a better prediction of subsequent development of AKI and death [63]. However, Waikar et al. suggested that normalized levels of biomarkers reflecting tubular injury could be influenced by dynamic changes in the UCr excretion rate when the GFR changed. Thus, the standardized concentrations might lead to an underestimation or overestimation [64]. Therefore, the recommended range of uIL-18 in predicting AKI remains controversial and further studies are needed.
Many biomarkers have been studied to evaluate their predictive value for AKI. An international conference proceeding including the AKI definition, staging and the potential biomarkers has been published before, and recommended 14 biomarkers for further research [65]. It was noting that there is only one FDAapproved biomarker product commercially available in the United States, which is tissue inhibitor of metalloproteinase-2 (TIMP-2) and insulin-like growth factorbinding protein 7 (IGFBP7). It has been promoted to predict AKI and presented the best performance over all the other biomarkers. In 2013, Kashani [70]. However, one of the potential limitations of this biomarker is the inability to distinguish between transient and persistent AKI. In addition, although there is evidence that using [TIMP-2] Â [IGFBP7] in combination with routine testing, including SCr and urine volume, within 12 h of patient evaluation in the general surgical ICU helps predict the progression of AKI, there is little substantive evidence that its use improves clinically important patient outcomes [71]. Neutrophil gelatinase-related lipid carrier protein (NGAL) is a member of the lipid carrier protein superfamily, which is expressed by neutrophils and various epithelial cells. It has been regarded as a proximal tubular injury-related protein and been widely studied for the early prediction of AKI. Zhang et al. reported NGAL was a valuable predictor of sepsis-related AKI and associated with the mortality [72]. Xie et al. further explored the different diagnostic performance of urinary and plasma NGAL, they found that urinary NGAL showed higher diagnostic value than plasma NGAL in predicting AKI, especially in predicting sepsis related AKI (plasma NGAL: AUC ¼ 0.86, urinary NGAL: AUC ¼ 0.90) [73]. Kidney injury molecule-1 (KIM-1) is a type I transmembrane glycoprotein with immunoglobulin and mucin domains, which is not expressed in normal kidney tissue or urine, but highly expressed after ischemic or toxic renal injury. Its urine excretion could increase 2-8 h after injury significantly, thus helping the early diagnosis of AKI [74]. In addition, elevated urinary KIM-1 has the highest specificity in the recognition of renal injury, which can distinguish prerenal AKI from acute tubular necrosis [75]. In this analysis, we found the diagnosis of AKI with urinary IL-18 showed a sensitivity of 0.64 (95%CI: 0.59-0.73) and a specificity of 0.77 (95%CI: 0.71-0.83), indicating a moderate predictive value. Since IL-18 is a proinflammatory cytokine and its measurement may also be influenced by many coexisting variables including endotoxemia, inflammation, and autoimmune diseases, its sensitivity and specificity could be affected. In another meta-analysis comparing different biomarkers for sepsis-associated AKI, Xie et al. found the rank of urinary KIM-1 > urinary NGAL > plasma NGAL > urinary IL-18 according to the SROC curve area, suggesting that urinary KIM-1 has the highest diagnostic value. However, since the small simple size in Xie et al.'s study, studies with larger sample size are still needed in clinical practice [73]. With the addition of biomarkers to routine AKI monitoring parameters, further studies are warranted to identify those most likely to benefit among critically ill patients.
The limitation of this study cannot be ignored. First, the number of included studies was small. Second, there was a significant heterogeneity among the included studies in terms of different population setting, age, predictive time, AKI definition, the time of obtaining specimen and cutoff value. Moreover, even though both meta-regression and subgroup analysis were conducted to explore the source of heterogeneity, the marked heterogeneity still existed. Another limitation was the publication bias, which is suggested by Deek's funnel plots and cannot be altered through statistical methods. In addition, all the included studies used SCr or urine volume as the diagnostic criteria for AKI, which are not ideal indicators for the diagnosis of early kidney injury [76]. Radio-labeled tracer clearance or evidence of damage from renal biopsy maybe a better method to diagnosis AKI. But they are time-consuming, invasive, or radioactive, thus limiting their commonly application in routine clinical practice. Additionally, we did not register this study online before. Lastly, the predictive value of uIL-18 in the progression and prognosis of AKI is of great clinical significance as well, but it has not been discussed in this study. To sum up, current analysis based on very different clinical situation and patient populations, heterogeneity is inevitable with respect to AKI definition, AKI setting, time of specimen acquisition, and the predictive value of urinary IL-18 assessed by experimental groups. We tried to explore the sources of heterogeneity using meta-regression but failed, and the subgroup analysis did not reduce the heterogeneity significantly. Our findings are hypothesis generating but heterogeneity and publication bias limit use, thus, further studies are still needed.

Conclusions
Urinary IL-18 could be a relatively good biomarker for predicting AKI, especially in pediatric patients. However, further research and clinical settings are still needed to validate the potential application.