Role of Tissue Inhibitor of Metalloproteinases-2 and Insulin Like Growth Factor Binding Protein 7 for Early Recognition of Acute Kidney Injury in Critically Ill COVID-19 Patients

Background: A high proportion of critically ill patients with COVID-19 develop acute kidney injury (AKI) and die. Early recognition of subclinical AKI could contribute to AKI prevention. Therefore, this study was aimed at exploring the role of the urinary biomarkers NGAL and [TIMP-2]•[IGFBP7] for early detection of AKI in this population. Methods: This prospective, longitudinal cohort study included critically ill COVID-19 patients without AKI at study entry. Urine samples were collected on admission to critical care areas for determination of NGAL and [TIMP-2]•[IGFBP7] concentrations. Demographic information, comorbidities, clinical and laboratory data were recorded. The study outcomes were development of AKI and mortality during hospitalization. Comparisons of individuals who developed AKI during hospitalization vs. those without AKI were made using chi-squared test for categorical variables and Mann-Whitney U for continuous variables. Urinary biomarkers and their cutoff values were selected based on the highest sensitivity, specicity and area under the receiver-operating characteristics curve with 95% condence intervals for prediction of AKI. Selected biomarkers and cutoffs were used in the Kaplan-Meier survival analyses for the time to AKI. Logistic regression analysis was used to identify the association between relevant covariates with AKI and mortality. For all analyses, two-sided P values £0.05 were considered statistically signicant. Results: Of the 51 individuals studied, 25 developed AKI during hospitalization (49%). The risk factors for AKI were male gender (HR=7.57, 95% CI: 1.28-44.8; p=0.026) and [TIMP-2]•[IGFBP7] ³ 0.2 (ng/ml) 2 /1000 (HR=7.23 , 95% CI: 0.99-52.4; p=0.050). Mortality during hospitalization was signicantly higher in the group with AKI than in the group without AKI (p=0.004). Persistent AKI was a risk factor for mortality (HR=7.42, 95% CI: 1.04-53.04; in use of and AKI, acute kidney injury; AUC, area under the receiver-operating characteristics curve; CI, condence interval; PPV, positive predictive value; NPV, negative predictive value; NGAL; neutrophil gelatinase-associated lipocalin; TIMP-2, tissue inhibitor of metalloproteinases-2; IGFBP7, insulin like growth factor binding protein 7.

have been identi ed as possible AKI biomarkers, given that both are released following ischemic or in ammatory processes in the kidney, resulting in G1 cell cycle arrest for a short period [6,7,8]. Another biomarker of early AKI is the urinary neutrophil gelatinase-associated lipocalin (NGAL), as intrarenal concentration of this protein is abruptly up-regulated soon after ischemic or nephrotoxic kidney injury [9]. In view of the need to identify early signs of kidney involvement, this study was aimed at exploring the role of the urinary biomarkers NGAL, TIMP-2 and IGFBP7 for early detection of AKI in critically ill patients with COVID-19.

Study Population
This prospective, longitudinal cohort study was conducted at the National Institute of Respiratory Diseases (INER), the largest third-level institution designated by the Mexican Government for COVID-19 care. The Institutional Review Board approved the study  and written informed consent was obtained from all participants. We included individuals admitted to the ICU with diagnosis of severe pneumonia caused by SARS-CoV-2, who were 18 years of age or older; without AKI when urine sample was collected; with no history of chronic kidney disease (CKD) as indicated by interrogation of patients about CKD medical history and by an estimated glomerular ltration rate (eGFR) greater than 60 ml/min/1.73m 2 using the CKD-EPI equation [10].
SARS-CoV-2 severe pneumonia was de ned by clinical data of respiratory distress, bilateral alveolar opacities in 2 or more lobes, a ratio of partial arterial oxygen pressure/inspired oxygen fraction (PaO 2 /FiO 2 ) < 300 mm Hg and a positive result for SARS-CoV-2-real-time reverse transcription-polymerase chain reaction (rRT-PCR) assay in nasopharyngeal swab [11]. The primary outcome was the development of AKI during hospitalization. The secondary outcome was mortality during hospitalization in the group with AKI and the group without AKI.
Recorded variables included demographic and anthropometric variables, symptoms, comorbidities, treatments, critical care variables, blood chemistry, blood count, starting and termination dates of invasive mechanical ventilation (IMV), days in hospital, initial mechanical-ventilator settings, use of vasoactive drugs and outcomes.
Pregnant women were not included in the study. Patients with incomplete clinical records were excluded.
De nition of acute kidney injury AKI staging was based on serum creatinine (sCr) levels. The urine output criterion was not used for diagnosis of AKI since nursing records were out of reach, in COVID-19 areas. The baseline sCr level was de ned as the minimum inpatient value during the rst 7 days of admission [12]. Diagnosis of AKI was based on the Kidney Disease Improving Global Outcomes (KDIGO) criteria [13]. AKI stage 1 corresponded to an increase in sCr by ≥0.3 mg/dl within 48 hours or increase in sCr 1.5 to 1.9 times baseline within the prior 7 days; AKI stage 2 corresponded to an increase in sCr of 2.0-2.9 times baseline; and AKI stage 3 corresponded to an increase in sCr of ≥3 times baseline or initiation of renal replacement therapy. Persistent AKI was de ned by the continuance of AKI by serum creatinine or urine output criteria (as de ned by KDIGO) beyond 48 h from AKI onset. Transient AKI was de ned by complete reversal of AKI by KDIGO criteria within 48 h of AKI onset [14].

Biomarker determinations
Urine samples were collected on admission to critical care areas (day 1). Urine was frozen at -80 C within the rst 30 minutes after sample collection. Urinary concentrations of TIMP-2 and IGFBP7 were determined using commercially available ELISA kits (Human TIMP-2 Quantikine ELISA Kit, R&D, Minneapolis, Minnesota; Human IGFBP7 ELISA Kit, Abcam, Cambridge, UK) following manual instructions. ELISA plates were read at O.D. of 450 and calculations were done according to the signal given by the standard curve of each kit. NGAL determinations were done using the NGAL kit (Abbott, Chicago, Illinois) according to the manual instructions and using the Abbott™ ARCHITECT™ Analyzer.

Statistical Analysis
We performed descriptive statistics including means and standard deviations for normally distributed continuous variables, medians and interquartile ranges for non-parametric distributions, and proportions for categorical variables. Comparisons of individuals who developed AKI during hospitalization vs. those without AKI were made using chi-squared test for categorical variables and Mann-Whitney U for continuous variables.
For each biomarker, the area under the receiver-operating characteristics curve (AUC) with 95% con dence intervals was calculated, as well as the sensitivity, speci city, positive predictive value (PPV), negative predictive value (NPV) and accuracy at 3 different cutoff values using urine samples collected upon hospital admission. We considered that the prevalence of AKI for patients with SARS-CoV-2 infection in the ICU was 40% [15]. Cutoffs for each biomarker were selected based on the highest AUC, speci city and accuracy for prediction of AKI. Combinations of the top biomarkers were also explored. When combinations had no signi cant added value, individual biomarkers were preferred. For all analyses, two-sided P values ≤0.05 were considered statistically signi cant. The selected biomarkers and cutoff values were used in the Kaplan-Meier survival analyses for the time to AKI.
Logistic regression analysis was used to identify the association between relevant covariates with AKI and mortality. We obtained age-strati ed estimates considering 60 years and older as vulnerable population. Variables were entered into the models when the alpha level of risk factor was < 0.20 in the univariate analysis. Age and gender were entered into the models regardless of the alpha level. All statistical tests were two-sided, and twosided P values ≤0.05 were considered statistically signi cant. The analysis was conducted using RStudio 1.4.1717.

Characteristics of study population
During the period between May and August 2020, a total of 420 individuals were admitted to critical areas of the INER. Of those, 69 were negative for SARS-CoV-2 infection; in 44 the infection could not be con rmed; 60 remained in the emergency room due to hospital saturation and 20 died there. Informed consent could not be obtained for 196 patients. We thus included the 51 patients who provided informed consent for participating in the study (Fig. 1). Of those, 30 were male (58.8%); the median age was 53 years (IQR, 40-61); 14 had hypertension (27.5%); 16 had diabetes (31.4%); and 21 were obese 41.2% (Table 1). Of the 51 individuals studied, 25 developed AKI during hospitalization (the AKI group, 49.0%) and 26 did not develop AKI (the non-AKI group, 51.0%). Eleven individuals had AKI stage 1 (21.5%); 8 had AKI stage 2 (15.6%); and 6 had AKI stage 3 (11.7%).
As expected, mortality was higher in the group with AKI than in the non-AKI group (36.0% vs. 3.8%; p = 0.004).

Performance of biomarkers as AKI predictors
Based on the highest AUC, speci city and accuracy values, the biomarker with best performance for AKI prediction during the whole hospitalization period was NGAL at a cutoff of 45 ng/ml. Considering that most patients developed AKI during the rst 7 days at the hospital, we also determined the performance of biomarkers for AKI prediction on day 7 ( Table 2). The performance of NGAL was signi cantly better on day 7 than during the   Variables were entered into the model when the alpha level of risk factor was less than 0.15. Age and gender were added into the model regardless of the alpha level.

Mortality was higher in individuals with AKI
We constructed a Kaplan-Meyer curve for mortality comparing the group of 25 patients who developed AKI during follow-up, with the group of 26 individuals without AKI. The mortality of individuals who developed AKI at any time during hospitalization was signi cantly higher than in those who never had AKI, p = 0.019 (Fig. 4).   [17]; AKI in patients after major surgery [18]; imminent risk of AKI in critically ill patients [7], and AKI in platinum-treated patients at the ICU [19]. The mechanism proposed is that after initial damage, IGFBP7 and TIMP-2 are expressed in tubular cells. IGFBP7 directly increases the expression of p53 and p21, and TIMP-2 stimulates p27 expression, leading to transitory G1 cell cycle arrest, preventing division of damaged cells [5]. Thus, since G1 cell cycle arrest is a common response to tubular damage, these biomarkers may better re ect damage regardless of etiology.
The combination of [TIMP-2]•[IGFBP7] had the best performance for AKI prediction at values above 0.2 (ng/ml) 2 /1000. This cutoff was based on overall behavior of the biomarkers in the patients studied here.
However, different cutoffs for these biomarkers have been reported in other studies, so speci c groups of patients may require identi cation of optimal cutoff values, based on their respective values of AUC, sensitivity, speci city, PPV, NPV and accuracy. Cutoff values may be affected by the severity of AKI. That is, higher cutoffs may be found in patients with AKI stages 2 and 3; and lower cutoffs may be found in patients with AKI stage 1 or subclinical AKI. Moreover, AKI is a complex syndrome, involving a series of complex cellular and molecular pathways, and the different cutoffs may re ect mechanistic differences between various etiologies of AKI [5]. In our cohort, the time to AKI was signi cantly shorter in individuals with NGAL ≥45 ng/ml than in those with < 45 ng/ml, but NGAL was not a risk factor for AKI during hospitalization. The fact that performance of NGAL was signi cantly better on day 7 than during the whole hospitalization period, suggests that NGAL has a narrow predictive time window for AKI. In addition, NGAL has proved less discriminating in the development of septicassociated or adult cardiac-surgery-associated AKI than in other types of AKI, possibly because neutrophils themselves may be a source of NGAL in the setting of systemic in ammation [21].
Contrary to our ndings, a recent cohort study found that urinary NGAL > 150 ng/ml predicted diagnosis, duration, and severity of AKI and acute tubular injury, as well as hospital stay, dialysis, shock, and death in patients with acute COVID-19 [22]. Contrasting results may be explained by the fact that some patients in that study probably had AKI when urinary samples were collected, while we only included patients without AKI at the time of urine sample collection. Therefore, the median value of NGAL in the AKI group (50.2 ng/ml) and the selected cutoff (45 ng/ml), were far below in our patients since they had subclinical AKI. In addition, it is unclear if a higher proportion of their patients had AKI stage 2 and stage 3, while most of our patients developed AKI stage 1 on subsequent days. This is relevant because that study also reported a correlation between urinary NGAL levels and AKI severity. In another recent study, NGAL was also found as an independent risk factor for AKI in patients with COVID-19, but that study also included some patients who already had AKI when urine samples were collected [23]. Thus, we suggest that in patients with COVID-19, higher NGAL cutoff values seem to be useful in predicting AKI progression but not AKI onset. In contrast with our ndings, urinary NGAL but not [TIMP-2]•[IGFBP7], independently predicted AKI in a cohort of decompensated cirrhotic patients, suggesting that different biomarkers should be used in different patient groups [24].
The survival analysis indicated that mortality was more frequent in patients who developed AKI during hospitalization, and mortality was attributed to persistent AKI because it was a risk factor for mortality. The concept that time should also be considered in the description of AKI and not only severity, was demonstrated in a study reporting that duration of AKI following surgery was independently associated with hospital mortality after adjusting for severity of illness [25]. Transient AKI may re ect a temporary reduction in renal function without structural damage, whereas persistent AKI would re ect structural tubular damage [26]. Based on these observations, persistent AKI has become a relevant endpoint in subsequent studies, and it has consistently been associated with mortality [27]. avoidance of nephrotoxic drugs, and preventing hyperglycemia, resulted in an absolute risk reduction of 16.6% in the incidence of AKI compared with the standard care [29].
An important limitation of our study was the small sample size. Another study limitation was that patients with incomplete clinical les or those who were transferred to other hospitals due to local saturation were not included in the study, and this may represent a selection bias. Considering that standardized de nitions of AKI are based on sCr and urine output [30], then inaccessibility to nursing records restricted to COVID-19 areas represents an important study limitation because urine output was not used for diagnosis of AKI, and sCr was not adjusted for uid-balance. The lack of pre-hospital baseline sCr measurements was also a study limitation because baseline sCr values were an estimation. One additional study limitation was that our study was conducted at a national referral center for respiratory diseases receiving disproportionately more patients with severe COVID-19, and this represents a potential source of referral bias.   Cumulative AKI events according to urinary NGAL concentrations. AKI in individuals with urinary NGAL ≥45 ng/ml (blue line) vs. AKI in individuals with urinary NGAL <45 ng/ml (purple line) during hospitalization (p=0.028).

Conclusions
Time 0 corresponded to hospital admission.