Diagnostic prediction of urinary [TIMP-2] x [IGFBP7] for acute kidney injury: A meta-analysis exploring detection time and cutoff levels

Acute kidney injury (AKI) most commonly occurs in critically ill and postoperative patients. Tissue inhibitor of metalloproteinases-2 (TIMP-2) and insulin-like growth factor-binding protein 7 (IGFBP7) are two newly-identified urinary biomarkers that can help to detect early AKI, yet their predictive accuracies range widely. Here, we conduct a systematic meta-analysis to evaluate the diagnostic values of [TIMP-2] x [IGFBP7] for AKI at different detection times and cutoff levels. Ten studies were meta-analyzed on 1606 patients. Overall, urinary [TIMP-2] x [IGFBP7] had a pooled sensitivity of 58% and specificity of 79%. Subgroup analysis showed that the sensitivity and specificity were 0.72 and 0.58 with a cutoff value of 0.3 (ng/mL)2/1000, and 0.38 and 0.94 with a cutoff value of 2.0 (ng/mL)2/1000, respectively. Moreover, when 0.3 was chosen as the cutoff value, restricting analysis to patients who were tested within 4 hours showed a sensitivity of 0.71 and specificity of 0.73, with the AUROC of 0.75. When 2.0 was chosen as the cutoff value, the sensitivity and specificity were 0.43 and 0.93, respectively in patients who were tested within 24 hours, with the AUROC of 0.70. In summary, urinary [TIMP-2] x [IGFBP7] can predict the occurrence of AKI with moderate diagnostic accuracy. In the earlier administrative periods (less than 4 hours), 0.3 (ng/mL)2/1000 is recommended to be used; whereas for patients who were administrated more than 24 hours, 2.0 (ng/mL)2/1000 is more appropriate.


INTRODUCTION
Acute kidney injury (AKI) is a chief public health concern worldwide, and it is projected to affect 5% of patients admitted to the hospital and up to 50% of patients in intensive care units (ICU) [1]. Clinically, dehydration and sepsis combined with nephrotoxic drugs, especially following major surgeries, were considered the most common causes of AKI [2]. Currently, a growing concern has been paid to AKI, as it can precipitate the development of chronic kidney disease (CKD) and end-stage renal disease, and is associated with prolonged hospitalization and an increased mortality [3].
AKI ranks as one of the most expensive conditions, and in the United States the aggregative costs reached nearly $5-10 billion per year [1,4]. Given that AKI is generally reversible and several therapies in animal models were proposed [5], to extend anti-AKI therapies from bench

Meta-Analysis
to bedside still has a long way to go. In view of the high prevalence of AKI and its deleterious consequences, it is clinically practical to identify certain accurate and reliable biomarkers to predict the occurrence of AKI.
Over the past decade, several urinary and blood biomarkers have been postulated for the early detection of AKI, including neutrophil gelatinase associated lipocalin (NGAL), kidney injury marker 1 (KIM-1) and interleukin-18 , with inherent limitations [5,6]. In pursuit of ideal biomarkers for AKI diagnosis, two cell-cycle arrest proteins, tissue inhibitor of metalloproteinases-2 (TIMP-2) and insulin-like growth factor-binding protein 7 (IGFBP7) have come to our sight. TIMP-2 and IGFBP7 were firstly identified in 2013 by Kashani et al. [7] and subsequently validated in the United States and European countries [8]. Afterwards, a growing number of studies have showed that the product of urinary TIMP-2 and IGFBP7, termed as  x [IGFBP7] is a promising early indicator of AKI [9,10]. However, when to detect this biomarker and how to select cutoff values thus far remains confused. To clear up this confusion, we attempted to evaluate the diagnostic value of urinary [TIMP-2] × [IGFBP7] for AKI through a metaanalysis based on recent clinical investigations. Figure 1 is a flow diagram that schematizes the selection of qualified articles in this meta-analysis. The initial literature search found 103 potentially relevant articles, and among them, 24 articles that seemingly met our inclusion and exclusion criteria were downloaded for further perusal. Finally, a total of 10 articles written in English and published between 2014 and 2016 were considered in this meta-analysis [11][12][13][14][15][16][17][18][19][20].

Characteristics of included studies
The basic characteristics of 10 qualified studies with 1648 participants are shown in Table 1. There were three multicenter trials and five clinical trials that recruited more than 100 participants. Nine of ten studies (n = 1597) were conducted in adult participants, and one study (n = 51) in children [14]. Six studies (n = 350) were conducted to evaluate the diagnostic value of urinary [TIMP-2] x [IGFBP7] for AKI after major surgeries (5 cardiac surgery [13,14,[16][17][18] and 1 non-cardiac surgery [15]), and four studies (n = 1298) were conducted among ICU patients [11,12,19,20]. The definition of AKI in eight studies accorded with the Kidney Disease: Improving Global Outcomes classification (KDIGO) criteria. There was one study defining AKI according to the Risk, Injury, Failure, Loss of kidney function, and End-stage kidney disease (RIFLE) criteria [14] and the U-score [20], respectively.

Comparison of diagnostic accuracies of different cutoff values and detection time
The diagnostic accuracy of urinary [TIMP-2] × [IGFBP7] for AKI was also assessed across different subgroups.  Figure 2). It is clearly shown that 0.3 (ng/mL) 2 /1000 had a relatively higher sensitivity and a lower specificity in diagnosing AKI, whereas, 2.0 (ng/mL) 2 /1000 had a lower sensitivity and a relatively higher specificity.
In addition, with prolonged sampling time, the sensitivity increased yet the specificity decreased. When 0.3 (ng/mL) 2 /1000 was chosen as the cutoff value of urinary [TIMP-2] × [IGFBP7] to predict AKI, the sensitivity and specificity were respectively 0.76 and 0.72 and the AUROC was 0.74 (95% CI, 0.70-0.78) in patients who were tested within 4 hours post-operatively (Table 4). When 2.0 (ng/mL) 2 /1000 was chosen as the cutoff value, the sensitivity and specificity were respectively 0.49 and 0.93 and the AUROC was 0.70 (95% CI, 0.51-0.89) in patients who were tested within 24 hours post-operatively (Table 4).

Diagnostic accuracy of urinary [TIMP-2] × [IGFBP7] for adverse kidney events
Adverse kidney events, including death, need for RRT, or persistence of renal dysfunction at 30 days were reported by only three studies [15,21,22]. In spite of inconsistent cutoff values in each study, we did not perform meta-analysis but only list their diagnostic accuracies of ( Table 4). Data from those three studies www.impactjournals.com/oncotarget showed that when detected urinary [TIMP-2] × [IGFBP7] on ICU admission, the lower cutoff values (less than 0.6) can be selected, with a high sensitivity (mean: 100%) and moderate specificity (mean: 61%).

DISCUSSION
AKI is a common, complex disorder featured by a high mortality and costly complications. Previously, serum creatinine (Scr) and urine output are regarded as two "gold-standard" functional markers for the early detection of kidney injury, yet their diagnoses were later proven to be insensitive. In addition, Scr and urine output were also subject to confounding impact of muscle volume, diets and diuretics usage, which limited their further clinical application. To look for more sensitive and specific markers for renal injury, several clinical trials on AKI have been conducted and suggested some promising biomarkers, including NGAL, KIM-1, IL-18, liver-type fatty acidbinding protein (L-FABP) and cystatin C [6]. There is compelling evidence that NGAL was affected by age, renal function and severity of AKI, and NGAL and IL-18 can stimulate inflammation response [23]. Considering the fact that TIMP-2 and IGFBP7 are cellcycle arrest proteins expressed in renal tubular cells during the periods of cellular stress or injury [8], it is tempting to speculate that urinary TIMP-2 and IGFBP7 are involved in the early pathological changes of AKI and carry the diagnostic probability of early renal injury. Now, several AKI cohort studies have evaluated the diagnostic value of the product of urinary TIMP-2 and IGFBP7, termed as [TIMP-2] × [IGFBP7] for AKI, whereas their detection times and cutoff values differed considerably. To fill this void in knowledge, we attempted to conduct a meta-analysis to quantify the appropriate detection time and cutoff value of urinary [TIMP-2] × [IGFBP7] for the early detection of AKI.
It is a general practice to determine the optimal cutoff value of a continuous predictive marker. There are two widely used cutoff points for urinary [TIMP-2] × [IGFBP7] in the literature, viz. 0.3 and 2.0 (ng/mL) 2 /1000 in the assessment of AKI. In the Gunnerson's study,       as the time went on, the secretion level of TIMP-2 and/ or IGFBP7 increased gradually. Several possible limitations in the present study merit serious consideration. First, with the purpose of avoiding low-quality studies, only published English articles were retrieved and articles in other languages was not covered, publication bias might be possible. Second, the types of participants in each study were various. Among of them, six studies were conducted among postoperative patients (5 studies cardiac surgery and 1 noncardiac surgery), and four studies were conducted among ICU patients. It might cause significant heterogeneity between the selected studies. Therefore, the jury must refrain from drawing a conclusion until large, multi-center and well-performed clinical trials confirm or refuse our findings.
In summary, this meta-analysis provided evidence that urinary [TIMP-2] × [IGFBP7] can predict the occurrence of AKI with moderate diagnostic accuracy. In the earlier administrative periods (less than 4 hours), 0.3 (ng/mL) 2 /1000 is recommended to be used; whereas for patients who were administrated more than 24 hours, 2.0 (ng/mL) 2 /1000 is more appropriate. Nonetheless, it still remains an open question to determine the optimal time for urinary TIMP-2 and IGFBP7 measurement and the optimal cutoff value of [TIMP-2] × [IGFBP7] for the diagnosis of AKI. We hope further clinical trials with larger sample sizes and high-quality evidence are designed to clear away the clouds of these controversial issues convincingly.

Data sources and search strategy
Two investigators (Ting Lin and Kai Qu) systematically and independently searched PubMed, EMBASE, Scopus and Web of Sciences databases for articles published before October 1, 2016 that provided data on the diagnostic accuracy of urinary [TIMP-2] × [IGFBP7] on the early identification of AKI in ICU patients. The conduct of this meta-analysis accorded with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (Supplementary PRISMA checklist). Subject terms used for literature search embraced ["TIMP-2" OR "tissue inhibitor of metalloproteinase-2"] AND ["IGFBP7" OR "insulin-like growth factor binding protein 7"] AND ["acute kidney injury" OR "AKI"] AND ["diagnosis" or diagnostic"]. Search spectrum was also extended to the reference lists of retrieved original and review articles. Only studies with a prospective design and articles published in the English language were retained for analysis.

Data extraction and synthesis
Two investigators (Ting Lin and Kai Qu) independently evaluated the study eligibility and quality according to the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) score system. The data of interest were extracted on the first author, year of publication, population, study design, clinical setting, age, gender, history of CKD, AKI definition, time of measurement, urine sample storage and detection method. Diagnostic accuracy estimates included TP, FN, FN, TN. In addition, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for each reported test threshold were evaluated accordingly.

Assessment of diagnostic test accuracy
The diagnostic test accuracy of the [TIMP-2] × [IGFBP7] for AKI diagnosis was quantified by the area under the summary receiver operating characteristic (SROC), summary diagnostic odds ratios (DORs) and summary sensitivities/specificities, respectively. Different SROC curves for the [TIMP-2] × [IGFBP7] were depicted and differences in SROC values were justified. The SROC curve delineates the relationship of the sensitivity against the specificity of different studies. Further summary DORs using a Der-Simonian and Laird random-effects model were computed, as well as between-study heterogeneity.

Assessment of heterogeneity and publication bias
Between-study heterogeneity was represented by the Q-I 2 statistic, which denotes the percentage of varied estimates accruing from heterogeneity rather than from sample errors. Significant heterogeneity was reported if the Q-I 2 statistic is 50% or over. Potential contributing factors responsible for significant heterogeneity in the diagnostic accuracy of the [TIMP-2] × [IGFBP7] for AKI were sought by a meta-regression analysis. Also, heterogeneity was explored further by subgroup analyses across different clinical settings. In addition, the possibility of publication bias was visually inspected by the asymmetry of a Deek's funnel plot in the prediction of urinary [TIMP-2] × [IGFBP7] for AKI.
Pooled sensitivity and specificity, DORs, and relevant 95% CIs were calculated on the basis of a bivariate normal model with log-transformed sensitivities and specificities. In the case of multiple cutoff points for www.impactjournals.com/oncotarget the [TIMP-2] × [IGFBP7] provided in a single study, the point with the maximum overall accuracy entered into the overall analysis.
All statistical tests were two sided, and P < 0.05 was considered significant unless otherwise indicated. Above statistical analyses were completed with Stata 12.0.

Author contributions
Ting Lin and Kai Qu: Designed the research and rafted the manuscript; Wenquan Niu and Chang Liu: Revised the paper; Zhenzhu Song, Ting Lin and Kai Qu: Performed meta-analysis; Sinan Liu: Participated in literature search and study selection; Ting Lin and Kai Qu: Analyzed data and constructed figures.

CONFLICTS OF INTEREST
There is no competing financial interest among the authors.