https://orcid.org/0000-0003-3610-6942
Figures
Abstract
Importance
Extracellular matrix proteins and enzymes involved in degradation have been found to be associated with tissue fibrosis and ureteropelvic junction obstruction (UPJO). In this study we developed a promising urinary biomarker model which can identify reduced renal function in UPJ obstruction patients. This can potentially serve as a non-invasive way to enhance surgical decision making for patients and urologists.
Objective
We sought to develop a predictive model to identify UPJO patients at risk for reduced renal function.
Setting
Pre-operative urine samples were collected in a prospectively enrolled UPJO biomarker registry at our institution. Urinary MMP-2, MMP-7, TIMP-2, and NGAL were measured as well as clinical characteristics including hydronephrosis grade, differential renal function, t1/2, and UPJO etiology.
Main outcome measurement
Primary outcome was reduced renal function defined as MAG3 function <40%. Multivariable logistic regression was applied to identify the independent predictive biomarkers in the original Training cohort. Model validation and generalizability were evaluated in a new UPJO Testing cohort.
Results
We included 71 patients with UPJO in the original training cohort and 39 in the validation cohort. Median age was 3.3 years (70% male). By univariate analysis, reduced renal function was associated with higher MMP-2 (p = 0.064), MMP-7 (p = 0.047), NGAL (p = 0.001), and lower TIMP-2 (p = 0.033). Combining MMP-7 with TIMP-2, the multivariable logistic regression model predicted reduced renal function with good performance (AUC = 0.830; 95% CI: 0.722–0.938). The independent testing dataset validated the results with good predictive performance (AUC = 0.738).
Citation: Wang H-HS, Cho PS, Zhi H, Kostel SA, DiMartino S, Dagher AM, et al. (2022) Association between urinary biomarkers MMP-7/TIMP-2 and reduced renal function in children with ureteropelvic junction obstruction. PLoS ONE 17(7): e0270018. https://doi.org/10.1371/journal.pone.0270018
Editor: Israel Franco, Yale University School of Medicine, UNITED STATES
Received: May 9, 2022; Accepted: June 2, 2022; Published: July 14, 2022
Copyright: © 2022 Wang et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data files are within the paper and its Supporting Information files.
Funding: NIH/NIDDK DK096238 (RSL); https://www.niddk.nih.gov/; The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Ureteropelvic junction obstruction (UPJO) is one of the most common anatomical anomalies presented as hydronephrosis. The concern with hydronephrosis is renal obstruction and subsequent loss of kidney function. Early diagnosis of UPJO and timely surgical intervention can prevent progressive and permanent renal damage. While some children with hydronephrosis resolve spontaneously with conservative management, some develop progressive UPJO and associated renal damage [1]. The unpredictable nature of hydronephrosis makes management challenging.
Current standard hydronephrosis follow-up modalities including renal ultrasound (RUS) and radioisotope diuretic mercaptoacetyltriglycine (MAG3) renography suffer from significant limitations. RUS detects hydronephrosis but cannot determine renal function and degree of obstruction directly. While MAG3 assesses renal function and drainage, it is invasive requiring IV administration and urethral catheter placement and involves radiation exposure [2]. Since MAG3 relies on relative values, changes in differential renal function may also be problematic, reflecting contralateral compensation rather than true decline in the affected kidney. Promising results have been reported in magnetic resonance imaging urography [3]. However, this modality is experimental in nature with the potential need of sedation or general anesthesia in children and associated steep cost [4, 5].
A recent trend investigating the urinary proteome could potentially provide an optimal non-invasive solution in managing hydronephrosis patients. Significant UPJO may result in the release of urinary proteins signifying a renal response [6, 7]. Proteins involved in or affected by renal tubular or interstitial injury hold promise to reflect a real-time change in renal status, as levels have been detected in blood and urine in the setting of renal injury [8, 9]. Previous reports have demonstrated increased urinary NGAL in UPJO children undergoing surgery compared to healthy controls which may reflect acquired renal damage [10, 11]. Other potential candidate proteins, such as matrix metalloproteinases (MMPs) and their endogenous regulators, tissue inhibitors of metalloproteinases (TIMPs), have also been associated with various diseases including renal injury and have shown promise in both animal and clinical studies [12–14]. MMPs are zinc-dependent endopeptidases involved in the remodeling of extracellular matrix and basement membrane during normal kidney morphogenesis [15]. MMPs and TIMPs maintain a balance of ECM synthesis and degradation. When levels are altered, this can lead to interstitial fibrosis and glomerulosclerosis in various clinical scenarios including urinary obstruction [16].
We therefore hypothesized that a combination of previously identified urinary biomarkers expressions can differentiate renal function in UPJO patients. Levels may indicate active response to stress from obstruction leading to renal injury and signaling a need for functional imaging or surgical intervention. In this study, we investigated urinary levels of MMP-2, MMP-7, TIMP-1, TIMP-2 and NGAL in patients with UPJO who underwent pyeloplasty before or at the time of surgery. We examined their value as urinary biomarkers in children with UPJO with varying degrees of renal dysfunction. With DRF as the primary outcome, we aimed to develop a prediction model utilizing clinical characteristics and the urinary biomarkers to predict impaired DRF < 40% in UPJO patients. These markers could then potentially be used in combination with screening renal US to determine when a child could be observed (low risk) versus undergo a functional study evaluation (high risk). These markers could hold greater utility in combination with other markers in future prospective studies to distinguish benign self-resolving hydronephrosis from persistent or progressive pathologic obstruction needing intervention.
Materials and methods
Sample source
Patient samples were obtained from the IRB-approved Boston Children’s Hospital Pediatric Urinary Proteome Program Initiative (PUPPI) sample repository. This repository includes samples obtained from patients who underwent surgery at our institution.
Patients with unilateral UPJO who underwent pyeloplasty were included. Patients with solitary kidney, as well as bilateral UPJO were excluded. The included cohort was randomly selected as UPJO training cohort. Exclusion criteria included history of vesicoureteral reflux or febrile urinary tract infection. All patients had a pre-operative renal bladder ultrasound and MAG3 diurectic renography. Demographics and clinical data including pre- and post-operative hydronephrosis grade based on Society for Fetal Urology (SFU) criteria, UPJO etiology (intrinsic versus extrinsic), MAG3 differential renal function and t1/2. All bladder urine samples were taken via urethral catheterization immediately prior to surgery. An independent validation testing cohort using the same inclusion and exclusion criteria as the training cohort was created from PUPPI repository to validate the prediction model.
Sample preparation and quantification
All urine samples were analyzed by urinalysis using a Siemens CLINITEK® status automated analyzer (Tarrytown, NY) to ensure no contamination or infection. All samples were centrifuged at 4000rpm/680g for 10 minutes to remove cellular debris, aliquoted, and frozen at -80C. Protein quantification was performed, as previously described [17, 18], using 100ul of thawed urine by Bradford assay (Biorad, Hercules, CA) read on a Beckman DU600 Spectrophotometer.
Human MMP2, MMP7, TIMP1, TIMP2, and NGAL Quantikine ELISAs were obtained from R&D Systems (Minneapolis, MN). Aliquots of 1ml were thawed and run in duplicate for each ELISA protein target according to the manufacturer’s protocol for urine. Optical density was measured at 450nm with a correction at 540nm using a FilterMax F3 Multi-Mode Microplate Reader (Molecular Devices, Sunnyvale, CA). Dilution series optimization was performed for each ELISA to ensure that sample readings would be within the upper and lower limits of detectability based on the standard curve. The following dilutions were utilized: MMP-2 and TIMP-1 undiluted, MMP-7 1:5, NGAL 1:10, and TIMP-2 1:2. Protein quantifications in ELISA were measured in concentration (ng/ml) and then normalized to total protein (pg/ug) as standard practice.
Statistical analysis
Primary outcome (reduced renal function) was defined as less than 40% function on MAG3. Biomarker levels for independent subgroups of 71 UPJO patients based on hydronephrosis grade (SFU 1–2 vs 3–4), etiology (intrinsic vs extrinsic) and parameters of MAG3 (t1/2 using 20 mins as cutoff; differential function using 40% as cutoff) were compared using Mann-Whitney U-test. Receiver operating characteristic (ROC) curve analysis with Youden’s J-index was applied to assess biomarkers in discriminating normal vs reduced renal function based on area under the curve (AUC) and 95% confidence intervals (CI) and to identify optimal cut-off values [19]. Backward selection was applied to select the significant independent biomarkers of UPJO to classify normal vs reduced renal function patients [20]. A two-tailed alpha p<0.05 and 95% confidence intervals (CI) were used as criteria for statistical significance. All analyses were performed using R 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria).
Results
Demographics of testing cohort
71 patients underwent surgical correction for UPJO. Median age was 3.3 years. 70% were males. Table 1 details the demographic and clinical characteristics of the model training cohort. 94% (67/71) of patients had SFU grade 3–4 hydronephrosis with the majority of the UPJO related to intrinsic (72%) obstruction and long t1/2 >20 minutes (59%).
Biomarkers and decreased renal function in UPJO patients
In the univariate analysis using the Mann-Whitney U-test, reduced renal function was significantly associated with higher MMP-7 (medium 167.9 vs 72.2 pg/ug, p = 0.047), higher NGAL (median 94.5 vs 24.7 pg/ug, p = 0.001), and lower TIMP-2 (median 81.6 vs 135.3 pg/ug, p = 0.033) as demonstrated in Table 2. With backward selection using reduced renal function as outcome (binary using 40% as cutoff), MMP-7 and TIMP-2 were selected to the final logistic regression model to predict reduced renal function with good biomarker predictive performance (AUC = 0.83; 95% CI: 0.722–0.938, as shown in Fig 1). The AUC in out-of-sample independent testing set was 0.738.
As demonstrating on the ROC curve, choice of biomarker cutoff needs to balance the sensitivity and specificity trade-offs. One can pick the cutoff and therefore the “best” sensitivity/specificity combination according to clinical need and context. Higher sensitivity is preferred as a screening test. On the other hand, higher specificity would be ideal for confirmatory test. We demonstrated several cutoff choice examples in Table 3 with the corresponding sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy in training and out-of-sample independent testing dataset. For example, if predicted probability cutoff of having reduced renal function was chosen at 0.35, this cut-off point corresponded to 94% sensitivity, 13% specificity, 77% accuracy, 81% PPV, 33% NPV in the independent testing dataset. If the predicted probability cutoff was lowered to 0.25, the performance would change to 48% sensitivity, 75% specificity, 53% accuracy, 88% PPV, 27% NPV in the independent testing dataset.
Model validation and sensitivity analysis
The independent validation testing cohort shared similar baseline characteristics as the original cohort with median age of 5.4 years (interquartile range 1.1–8.1 years), 67% male, and 95% preoperative high-grade hydronephrosis (S1 Appendix). The final multivariate model predicting reduced renal function performed reasonably well in the validation cohort with AUC of 0.73. The value of the biomarkers did not vary by age and gender, confirming robustness of the multivariable model.
Discussion
Our study demonstrates the value of predictive urinary biomarkers for risk stratification of children with UPJO. With the combination of MMP-7 and TIMP-2, we were able to perioperatively identify patients with clinically significant (with reduced renal function using 40% function on DMSA as the cutoff) UPJO in a non-invasive fashion. We specifically chose this cohort from the repository to focus on the clinically relevant question: among those who presented with suspected UPJO, can we identify those with reduced renal function by the urine biomarkers? Since not all UPJO patients needs to have pyeloplasty, the information of reduced renal function often plays an essential role in surgical decision. The study result would allow for selective management of UPJO patients using these markers as a non-invasive study in the follow-ups. For those children predicted to have a high risk of reduced renal function, providers would order a MAG3 to confirm and potentially proceed with surgical intervention. Conversely, for patients predicted to be at low risk for reduced renal function, continued observation with ultrasound and the deferral of invasive tests and intervention would be reasonable. Biomarker cutoff could be customized to be more sensitive or specific depending on the clinical use and goal.
Unlike other studies of urinary biomarkers of UPJO, we focused on using reduced renal function as the primary outcome of our model, as opposed to the subjective decision for surgery as an outcome. Encouragingly, the components of the model correlated well with existing knowledge of these urine biomarkers. Increased mRNA tissue expression of MMP-7 has been observed in embryological development of congenital renal dysplasia [21, 22], chronic renal allograft rejection [23, 24], as well as drug-induced nephropathy and obstructive uropathy via increased Wnt-beta catenin activity [25]. In our study, TIMP-2 levels were lower in UPJO patients with reduced renal function. Suggesting TIMP-2 may be involved in the affected renal unit’s active pathophysiology, as TIMP-2 inhibits the proteolytic activity of MMP-2 and mediates fibrosis [26]. Additionally, recent work has focused on TIMP-2 as part of the cell-cycle arrest proteins expressed in renal tubular cells during cellular stress or injury [27, 28].
Biomarkers such as MMP-2 and NGAL were also observed in obstructive renal pathologies. Increased tissue mRNA expression and urinary protein levels of MMP-2 were observed in glomerular damage in renal transplant rejection, FSGS, and glomerulosclerosis models [15, 29]. An in vivo model of renal ischemia demonstrated increased MMP-2 protein levels in renal tissue and increased active isoform of MMP-2 protein in urine in response to oxidative stress or hypoxia, which is highly relevant in urinary obstruction [16]. Prior human and rodent studies have reported increased renal mRNA expression and urinary protein levels of NGAL with UPJO, reflective of renal injury due to increased synthesis and decreased reabsorption [10]. Although, neither MMP-2 nor NGAL made it to the final multivariable prediction model despite the fact that they were strongly associated with reduced renal functions in UPJO patients. We hypothesized the overlap of biologic mechanism with MMP-7 and TIMP-2 contributed to their collinearity and therefore they were not retained in the final multivariable model.
The findings of our study should be viewed in the context of its limitations. Our cohort included mostly toilet-trained pre-teen children. The result might not be necessarily generalizable outside of this demographic range. We were optimistic of the model performance with the good validation performance of independent sample with different UPJO etiology and MAG3 t1/2 (as shown in S1 Appendix). A larger scale study is warranted to confirm the findings. Additionally, some may suggest the levels of urine biomarkers such of urine MMP-7 may peak prior to other conventional indices of AKI [30], our cohort may have already reached their maximum prior to the time point that we assayed. This presents an area for further study, as urinary MMP-7 may be an early indicator of ongoing renal damage. Furthermore, higher level of biomarkers found in more severe obstruction and reduced function might be counter-intuitive. Our hypothesis is that even with severe obstruction, high concentration of distinct markers can still pass through the UPJ to the bladder, or markers are being derived from the non-affected kidney in response to contralateral obstruction and enable the proper classification of phenotypes. Ultimately, utilizing urine biomarker from bladder only would yield much higher impact enabling non-invasive outpatient follow-ups.
One of the most significant challenges was the heterogeneity of the UPJO cohort. UPJO patients could have variable pathology and severity of obstruction. Therefore, to develop a predictive model for a “type” of UPJO patients is currently infeasible. Nonetheless, the results of our study showed a promising role of MMP-7 and TIMP-2 for patients with clinically significant UPJO who may be at risk for dysregulated renal tissue remodeling and continuous renal injury. Because the markers demonstrated the ability to distinguish diminished renal function in this heterogenous UPJO population, it implies that the molecular pathophysiology of renal obstruction irrespective of etiology or clinical severity may be similar. The promising results provide possibility for non-invasive testing. The sensitivity and specificity rely on the cutoff value of the biomarkers. One can pick the best sensitivity/specificity combination according to their need as shown in Fig 1. For example, if one is planning to use the biomarker as screening test, higher sensitivity is preferred. In that case one would pick the higher sensitivity/lower specificity combination. Vice versa for confirmation test, one would prefer a higher specificity/lower sensitivity combination. The AUC demonstrated very high performance- the sensitivity/specificity combinations would produce good predictions in variable cutoffs.
Future prospective longitudinal analysis of these and other markers in a larger cohort that is followed during observation of hydronephrosis and postoperatively in those who undergo surgery is necessary to determine the validity and clinical usefulness of these markers alone or in combination with other potential markers of kidney obstruction.
Conclusions
Our predictive model shows that MMP-7 and TIMP-2 can identify impaired renal function in patients with UPJO as thus useful for risk stratification. The trend or pattern of urine levels may be more helpful as a prognostic and diagnostic tool, as changes in baseline levels could signal the need and determine timing for additional imaging or surgical intervention.
Importantly, the results of this study are not associated with the potential subjective choice of performing surgery as the primary outcome was differential renal function. An objective UPJO risk stratification could potentially be used by clinicians and parents to determine the need for additional imaging after identifying possible UPJO on a screening ultrasound or serve as a follow-up tool.
Acknowledgments
Thank you to the generous support from Hilary and Geoffrey Grove and Family, and Sat and Sapna Shah
References
- 1. Fefer S, Ellsworth P. Prenatal hydronephrosis. Pediatr Clin North Am. 2006;53(3):429–47, vii. pmid:16716789.
- 2. Taylor A, Garcia EV, Binongo JN, Manatunga A, Halkar R, Folks RD, et al. Diagnostic performance of an expert system for interpretation of 99mTc MAG3 scans in suspected renal obstruction. J Nucl Med. 2008;49(2):216–24. pmid:18199609; PubMed Central PMCID: PMC3179983.
- 3. McDaniel BB, Jones RA, Scherz H, Kirsch AJ, Little SB, Grattan-Smith JD. Dynamic contrast-enhanced MR urography in the evaluation of pediatric hydronephrosis: Part 2, anatomic and functional assessment of ureteropelvic junction obstruction [corrected]. AJR Am J Roentgenol. 2005;185(6):1608–14. pmid:16304022.
- 4. Serafini G, Zadra N. Anaesthesia for MRI in the paediatric patient. Curr Opin Anaesthesiol. 2008;21(4):499–503. pmid:18660661.
- 5. Greenberg SB. Rebalancing the risks of Computed Tomography and Magnetic Resonance imaging. Pediatr Radiol. 2011;41(8):951–2. pmid:21626107.
- 6. Froehlich JW, Kostel SA, Cho PS, Briscoe AC, Steen H, Vaezzadeh AR, et al. Urinary Proteomics Yield Pathological Insights for Ureteropelvic Junction Obstruction. Mol Cell Proteomics. 2016;15(8):2607–15. pmid:27215552; PubMed Central PMCID: PMC4974339.
- 7. Siwy J, Zurbig P, Argiles A, Beige J, Haubitz M, Jankowski J, et al. Noninvasive diagnosis of chronic kidney diseases using urinary proteome analysis. Nephrol Dial Transplant. 2016. pmid:27984204.
- 8. Andreucci M, Faga T, Pisani A, Perticone M, Michael A. The ischemic/nephrotoxic Acute Kidney Injury and the use of renal biomarkers in clinical practice. Eur J Intern Med. 2016. pmid:28011057.
- 9. Alderson HV, Ritchie JP, Pagano S, Middleton RJ, Pruijm M, Vuilleumier N, et al. The Associations of Blood Kidney Injury Molecule-1 and Neutrophil Gelatinase-Associated Lipocalin with Progression from CKD to ESRD. Clin J Am Soc Nephrol. 2016;11(12):2141–9. pmid:27852662; PubMed Central PMCID: PMC5142061.
- 10. Cost NG, Noh PH, Devarajan P, Ivancic V, Reddy PP, Minevich E, et al. Urinary NGAL levels correlate with differential renal function in patients with ureteropelvic junction obstruction undergoing pyeloplasty. J Urol. 2013;190(4 Suppl):1462–7. pmid:23791906; PubMed Central PMCID: PMC4106144.
- 11. Noyan A, Parmaksiz G, Dursun H, Ezer SS, Anarat R, Cengiz N. Urinary NGAL, KIM-1 and L-FABP concentrations in antenatal hydronephrosis. J Pediatr Urol. 2015;11(5):249 e1–6. pmid:26096437.
- 12. Haase M, Bellomo R, Devarajan P, Schlattmann P, Haase-Fielitz A, Group NM-aI. Accuracy of neutrophil gelatinase-associated lipocalin (NGAL) in diagnosis and prognosis in acute kidney injury: a systematic review and meta-analysis. Am J Kidney Dis. 2009;54(6):1012–24. pmid:19850388.
- 13. Dagher A, Curatolo A, Sachdev M, Stephens AJ, Mullins C, Landis JR, et al. Identification of novel non-invasive biomarkers of urinary chronic pelvic pain syndrome: findings from the Multidisciplinary Approach to the Study of Chronic Pelvic Pain (MAPP) Research Network. BJU Int. 2017;120(1):130–42. pmid:28263447; PubMed Central PMCID: PMC5951631.
- 14. Roy R, Louis G, Loughlin KR, Wiederschain D, Kilroy SM, Lamb CC, et al. Tumor-specific urinary matrix metalloproteinase fingerprinting: identification of high molecular weight urinary matrix metalloproteinase species. Clin Cancer Res. 2008;14(20):6610–7. pmid:18927302; PubMed Central PMCID: PMC2879331.
- 15. Lenz O, Elliot SJ, Stetler-Stevenson WG. Matrix metalloproteinases in renal development and disease. J Am Soc Nephrol. 2000;11(3):574–81. pmid:10703682.
- 16. Novak KB, Le HD, Christison-Lagay ER, Nose V, Doiron RJ, Moses MA, et al. Effects of metalloproteinase inhibition in a murine model of renal ischemia-reperfusion injury. Pediatr Res. 2010;67(3):257–62. pmid:19915515; PubMed Central PMCID: PMC3366106.
- 17. Moses MA, Wiederschain D, Loughlin KR, Zurakowski D, Lamb CC, Freeman MR. Increased incidence of matrix metalloproteinases in urine of cancer patients. Cancer Res. 1998;58(7):1395–9. pmid:9537238.
- 18. Pories SE, Zurakowski D, Roy R, Lamb CC, Raza S, Exarhopoulos A, et al. Urinary metalloproteinases: noninvasive biomarkers for breast cancer risk assessment. Cancer Epidemiol Biomarkers Prev. 2008;17(5):1034–42. pmid:18483323.
- 19.
Zhou XH, Obuchowski NA, McClish DK. Statistical methods in diagnostic medicine. 2nd ed. New York: John Wiley & Sons; 2011.
- 20.
Vittinghoff EG D.V.; Shiboski S.C.; McCulloch C.E. Regression Methods in Biostatistics: Linear, Logistic, Survival, and Repeated Measures Models New York: Springer; 2005 January 4, 2005.
- 21. Jain S, Suarez AA, McGuire J, Liapis H. Expression profiles of congenital renal dysplasia reveal new insights into renal development and disease. Pediatr Nephrol. 2007;22(7):962–74. pmid:17450386.
- 22. McGuire JK, Harju-Baker S, Rims C, Sheen JH, Liapis H. Matrilysin (MMP-7) inhibition of BMP-7 induced renal tubular branching morphogenesis suggests a role in the pathogenesis of human renal dysplasia. J Histochem Cytochem. 2012;60(3):243–53. pmid:22215634; PubMed Central PMCID: PMC3351126.
- 23. Petra H, Eva H, Irena B, Petra H, Ondrej V. Molecular profiling of acute and chronic rejections of renal allografts. Clin Dev Immunol. 2013;2013:509259. pmid:24302958; PubMed Central PMCID: PMC3834978.
- 24. Han SS, Lee H, Oh YJ, Lee JP, Kim S, Ha J, et al. Identification of the effects of aging-related gene-matrix metalloproteinase on allograft outcomes in kidney transplantation. Transplant Proc. 2013;45(6):2158–64. pmid:23953525.
- 25. He W, Tan RJ, Li Y, Wang D, Nie J, Hou FF, et al. Matrix metalloproteinase-7 as a surrogate marker predicts renal Wnt/beta-catenin activity in CKD. J Am Soc Nephrol. 2012;23(2):294–304. pmid:22095947; PubMed Central PMCID: PMC3269179.
- 26. Bourboulia D, Stetler-Stevenson WG. Matrix metalloproteinases (MMPs) and tissue inhibitors of metalloproteinases (TIMPs): Positive and negative regulators in tumor cell adhesion. Semin Cancer Biol. 2010;20(3):161–8. pmid:20470890; PubMed Central PMCID: PMC2941566.
- 27. Vijayan A, Faubel S, Askenazi DJ, Cerda J, Fissell WH, Heung M, et al. Clinical Use of the Urine Biomarker [TIMP-2] x [IGFBP7] for Acute Kidney Injury Risk Assessment. Am J Kidney Dis. 2016;68(1):19–28. pmid:26948834; PubMed Central PMCID: PMC4921267.
- 28. Gocze I, Koch M, Renner P, Zeman F, Graf BM, Dahlke MH, et al. Urinary biomarkers TIMP-2 and IGFBP7 early predict acute kidney injury after major surgery. PLoS One. 2015;10(3):e0120863. pmid:25798585; PubMed Central PMCID: PMC4370650.
- 29. Mazanowska O, Zabinska M, Koscielska-Kasprzak K, Kaminska D, Krajewska M, Banasik M, et al. Increased plasma matrix metalloproteinase-2 (MMP-2), tissue inhibitor of proteinase-1 (TIMP-1), TIMP-2, and urine MMP-2 concentrations correlate with proteinuria in renal transplant recipients. Transplant Proc. 2014;46(8):2636–9. pmid:25380883.
- 30. Yang X, Chen C, Teng S, Fu X, Zha Y, Liu H, et al. Urinary Matrix Metalloproteinase-7 Predicts Severe AKI and Poor Outcomes after Cardiac Surgery. J Am Soc Nephrol. 2017;28(11):3373–82. pmid:28698269; PubMed Central PMCID: PMC5661292.