Interleukin-10: A Potential Pre-Cannulation Marker for Development of Acute Kidney Injury in Patients Receiving Veno-Arterial Extracorporeal Membrane Oxygenation

Introduction: Acute kidney injury (AKI) in patients treated with veno-arterial extracorporeal membrane oxygenation (VA-ECMO) is associated with high mortality. The objective of this study was to investigate whether cytokine levels before the initiation of ECMO treatment could predict AKI. We also aimed to investigate the impact of AKI on 30-day and 1-year mortality. Methods: Serum cytokine levels were analyzed in 100 consecutive VA-ECMO-treated patients at pre-cannulation, at 48 h post-cannulation, and at 8 days. Clinical data to establish the incidence and outcome of AKI after the start of ECMO was retrieved from the local ECMO registry. Setting: The study was conducted at tertiary care, university hospital. Participants included 100 patients treated with VA-ECMO. Interventions: The blood samples for cytokine analysis were collected before VA-ECMO treatment, at 48 h after VA-ECMO treatment was started, and at 8 days. Results: Pre-cannulation serum IL-10 levels were significantly higher in patients who developed AKI (212 [38.9, 620.7]) versus those who did not (49.0 [11.9, 102.2]; p = 0.007), and the development of AKI can be predicted by pre-cannulation IL-10 levels (p = 0.025, OR = 1.2 [1.02–1.32]). The development of AKI during ECMO treatment is associated with increased 30-day mortality (p = 0.049) compared to patients who did not develop AKI and had a pre-cannulation estimated glomerular filtration rate ≥ 45 mL/min. The 1-year survival rate for patients with AKI who survived the first 30 days of ECMO treatment is comparable to that of patients without AKI. Conclusion: Increased pre-cannulation IL-10 levels are associated with the development of AKI during VA-ECMO support. AKI is associated with increased 30-day mortality compared to patients with no AKI and better renal function. However, patients with AKI who survive the first 30 days have a 1-year survival rate similar to those without AKI.

treatment, at 48 h after VA-ECMO treatment was started, and at 8 days. Results: Pre-cannulation serum IL-10 levels were significantly higher in patients who developed AKI (212 [38.9, 620.7]) versus those who did not (49.0 [11.9, 102.2]; p = 0.007), and the development of AKI can be predicted by pre-cannulation IL-10 levels (p = 0.025, OR = 1.2 [1.02-1.32]). The development of AKI during ECMO treatment is associated with increased 30-day mortality (p = 0.049) compared to patients who did not develop AKI and had a pre-cannulation estimated glomerular filtration rate ≥ 45 mL/min. The 1-year survival rate for patients with AKI who survived the first 30 days of ECMO treatment is comparable to that of patients without AKI. Conclusion: Increased pre-cannulation IL-10 levels are associated with the development of AKI during VA-ECMO support. AKI is associated with increased 30-day mortality compared to patients with no AKI and better renal function. However, patients with AKI who survive the first 30 days have a 1-year survival rate similar to those without AKI.

Introduction
Patients treated with extracorporeal membrane oxygenation (ECMO) in general and veno-arterial ECMO specifically represent the sickest patient population in the intensive care unit. Despite a significant increase in the use of ECMO over the past decade and continuous technological improvements, in-hospital mortality remains high and ranges from 40 to 60% [1,2].
In hemodynamically unstable patients, kidneys are usually the first organ to fail, and AKI might already be established before the cannulation and start of ECMO treatment. AKI might also develop as a consequence of ECMO-related factors, such as ischemia-reperfusion, reduced pulsatility, augmented inflammatory response, and hemolysis [2,3]. Furthermore, decreased renal function at the start of ECMO is a known risk factor even without the development of AKI. There is no specific treatment for AKI, so early detection of worsening kidney function is crucial for timely interventions to avoid hypoperfusion, hypoxia, and nephrotoxic medications, e.g., vancomycin. However, early diagnosis of AKI is challenging as conventionally used markers such as serum creatinine and urinary output will not reflect a sudden deterioration of renal function; it normally takes 2-3 days from renal insult to creatinine maximum [4].
In recent years, numerous new biomarkers have been evaluated for their ability to predict AKI [5]. In cardiac surgery, increased cytokine levels, particularly IL-6 and IL-10, have been linked to postoperative AKI in patients exposed to extracorporeal circulation [6]. In addition, increased levels of IL-10 are also found in septic shock patients and patients with acute and chronic kidney diseases [7,8]. It is not known whether increased precannulation cytokine levels could be used as early markers of renal injury before the initiation of ECMO treatment in a hemodynamically severely compromised intensive care unit population.
The development of AKI during ECMO increases both mortality and morbidity. However, studies that have shown a relationship between AKI during ECMO and outcome were performed on a mixed population of both VV-ECMO and VA-ECMO, and follow-up was limited to 30 or 90 days [9].
Accordingly, the objective of this study was twofold: first, to investigate whether cytokine blood concentrations can predict AKI during VA-ECMO, and, second, to study the incidence of AKI during VA-ECMO and association between AKI and survival.

Trial Design and Study Population
One hundred consecutive patients requiring veno-arterial circulatory support treated at Meijer-Heart Center/Spectrum Health, Grand Rapids, MI, USA, were included. Patients were excluded from the analysis if they were treated with renal replacement therapy before the initiation of circulatory support, were ≤18 years of age, or were treated for autoimmune disease.
Blood samples for cytokine analysis were collected after consent was obtained from next to kin. The blood was collected immediately before patients were cannulated, at 48 h post-cannulation, and at 8 days post-cannulation.
All patients treated with ECMO at our institution are included in a local database registered at ClinicalTrials.gov (NCT02748668). Clinical data for the study were collected from this local ECMO database. The study protocol was reviewed and approved by the local Institutional Review Board at Spectrum Health and Van Andel Institute, Grand Rapids, MI, USA, approval number 2016-171.

Study Endpoints
The primary endpoint was correlation between precannulation cytokine levels and AKI after the start of VA-ECMO. We also studied the temporal pattern of serum concentrations of cytokines over time and the relation between AKI and all-cause mortality at 30 days and 1 year.
Definition of Renal Dysfunction AKI was defined according to KDIGO (Kidney Disease Improving Global Outcome) criteria and assessed within the time frame from ECMO start as day 1 to the end of ECMO treatment [10]. Urinary output criteria for AKI were not used, as the use of diuretics might influence urinary output, and increased urinary output in response to diuretics might not correspond to the actual stage of AKI [11,12]. All stages of AKI from 1 to 3, including the use of CRRT, were defined as AKI. Also, in an attempt to stratify the patients in two categories based on pre-cannulation renal function, two groups were created according to KDIGO Clinical Practice Guideline for the Evolution and Management of Chronic Kidney Disease (CKD): mildly to moderately decreased (estimated glomerular filtration [eGFR] ≥45 mL/min/ 1.73 m 2 ) or moderately to severely decreased (eGFR <45 mL/min/ 1.73 m 2 ) before the initiation of extracorporeal support [13].

Statistical Analysis
Normally distributed and continuous data were compared using Student's t test, while ordinal, or data with nonnormal distribution, were analyzed with Mann-Whitney's U test. Paired continuous variables were tested in a paired t test. Proportions were compared with Fisher's exact test. Normally distributed data are described as mean ± standard deviation and non-normally distributed data as median with interquartile range. We analyzed the time to mortality that occurs within the defined time frame, and subjects surviving were censored onward. Differences in mortality were compared between patients who developed AKI versus no-AKI with pre-cannulation eGFR ≥45 mL/min/1.73 m 2 and between patients who developed AKI versus no-AKI with precannulation eGFR <45 mL/min/1.73 m 2 . Results of the analysis were presented by Kaplan-Meier survival curves. A log-rank test was used to determine the statistical significance of differences between the groups. The second set of analyses was done similarly but with instances of AKI during VA-ECMO, as defined above, as the outcome. Patients who were decannulated for any reason without developing AKI were censored after decannulation.
The predictive value of IL-10 sampled before VA-ECMO start compared to other common predictors of AKI (age, eGFR, and inotrope score at start) at pre-cannulation was tested in a multiple logistic regression model with AKI developed on the VA-ECMO circuit as the outcome. The analysis was initiated with a univariate analysis, and the criteria for selecting variables were set at p < 0.1. The selective power to identify the patients who developed AKI on the circuit from those who did not was investigated with a ROC analysis. All p values <0.05 were considered significant. Table 1 Out of 100 patients included in the study, 42% (n = 42) developed AKI during VA-ECMO treatment. The patient's demographics, baseline characteristics, and indications for ECMO treatment are presented in Table 1. The severity of illness as estimated by the SOFA (Sequential Organ Failure Assessment) score (AKI 9.95 ± 2.71 vs. no AKI 10.04 ± 2.76; p = 0.88) and pre-cannulation eGFR (AKI 58.8 ± 24.3 mL/min/ 1.73 m 2 vs. no AKI 57.6 ± 29.7 mL/min/1.73 m 2 ; p = 0.71) showed no difference between groups. Sixty-four percent of AKI patients received continuous renal replacement therapy.

Patient Characteristics
Pre-Cannulation IL-10 Concentrations Predict AKI IL-10 levels were higher in patients who developed AKI compared to those who did not (212 [38.9, 620.7] pg/ mL vs. 49.05 [25.8, 102.2] pg/mL; p = 0.007) (shown in Table 2 and Fig. 1a). A ROC analysis based on the prediction of AKI and the IL-10 levels showed an AUC of 0.71 (shown in Fig. 1b), with the best separation at 112 pg/mL. IL-10 levels normalized after 2 days on VA-ECMO (Fig. 2).

Independent Predictors of AKI in VA-ECMO-Treated Patients
In a linear regression, eGFR and IL-10 levels at cannulation did not correlate with each other (r = −0.08 [−0.27; 0.12]; p = 0.44) (shown in Fig. 3). In a multiple logistic regression model with AKI as an outcome, pre-cannulation IL-10 levels (OR 1.  Table 3.

Mortality at 30 Days and 1 Year in VA-ECMO AKI and No-AKI Patients
There were no differences in 30-day (p = 0.15) or 1-year mortality (p = 0.269) between the AKI and no-AKI patients (shown in Fig. 4a, b). However, when sub-groups of no-AKI patients (eGFR < or >45 mL/min) were compared to AKI patients, the mortality at 30-days was significantly increased (p < 0.049; shown in Fig. 4c) in patients who developed AKI versus those without AKI development and pre-cannulation eGFR ≥45 mL/min. Furthermore, in patients with decreased kidney function at pre-cannulation (eGFR <45 mL/min) who did not develop AKI, the survival at 30 days (p = 0.886; shown in Fig. 4c), 1-year (p = 0.826; shown in Fig. 4d) was similar to the patients who developed AKI.
The highest mortality was observed within 30 days of VA-ECMO treatment. Patients with AKI who survived 30 days had a 1-year survival similar to those without AKI (shown in Fig. 4b, d).

Discussion
In this clinical observational study, we found that increased anti-inflammatory cytokine IL-10 precedes AKI development in VA ECMO patients. Moreover, IL-10 levels decreased within 48 h after the initiation of circulatory support. We also found that the development of AKI during VA-ECMO treatment significantly increased 30-day mortality as compared to patients who did not develop AKI and had better renal function before the start of the treatment. However, AKI patients with 30day survival showed a 1-year survival similar to patients without AKI.
The clinical decision to initiate VA-ECMO treatment is complex, and timing is crucial. Earlier initiation of mechanical circulatory support will decrease the state of hypoperfusion and the risk of irreparable organ damage. On the other hand, the treatment is associated with a risk of severe complications, and a clear benefit of VA ECMO treatment has not been established in randomized clinical studies [14]. Therefore, it should only be used judiciously when other treatment options have failed. The kidney is one of the organs that requires the most cardiac output and is usually the first organ to fail in hemodynamically unstable patients. The conventional markers of kidney function, including serum creatinine and cystatin C, do not reflect a sudden deterioration in renal function. Therefore, early markers of ongoing renal hypoperfusion could assist in the clinical decision to initiate VA-ECMO support earlier.
In recent years, numerous new biomarkers have been evaluated for their ability to predict AKI [5], and three randomized controlled trials have confirmed the  beneficial role of early implementation of an AKI care bundle in biomarker-positive patients [15][16][17]. While conventional biomarkers cystatin C, creatinine, and urinary output are markers of renal function, the new biomarkers can reveal kidney damage. These biomarkers can be detected either in plasma or urine, as these substances are filtered or released from the nephron and can thus be detected in the urine. For example, neutrophil gelatinase-associated lipocalin has been used as an early biomarker for cardiac surgery-associated AKI. It is released by activated neutrophils and injured kidneys and can be detected both in plasma and urine [18]. Implementation of a KDIGO-derived treatment bundle in cardiac surgery patients with increased urinary levels of metalloproteinases-2 [TIMP-2] and insulin growth factor-binding protein 7 [IGFBP7] resulted in decreased incidence of moderate to severe AKI [17]. However, these biomarkers have not yet found their way into clinical praxis. Cytokines are attractive as biomarkers as they are easy to sample (blood or urine), relatively easy to analyze, and have a relatively short half-life, opening the possibility of an early and fast diagnosis of AKI. Increased plasma levels of cytokines during ECMO are associated with poor prognosis in adult and pediatric ECMO patients [19,20]. Furthermore, ratios of anti-inflammatory/ pro-inflammatory markers in blood and urine have a predictive utility in cardiac-surgery associated AKI [21].
However, no studies have investigated whether increased pre-cannulation cytokine levels in VA-ECMO patients could be used as a biomarker to predict AKI development.
Cytokines are messengers in the immune system and are functionally divided into pro-inflammatory and antiinflammatory. T lymphocytes expressing CD4 receptors are further divided into Th-1 and Th-2, and the cytokines they produce are classified as Th1 and Th2-type cytokines. IL-10 is an anti-inflammatory Th-2 cytokine, discovered initially for inhibiting the synthesis of IL-2 and IFN-γ [8,22].
In our study, before the start of ECMO, patients who developed AKI on the ECMO circuit had more than 4 times higher pre-cannulation levels of IL-10 and 3 times higher levels of IFN-γ, compared to no-AKI patients (Table 2). However, in the multiple logistic regression model, only pre-canulation serum concentrations of IL-10 were an independent marker of AKI risk with an area under the curve in a ROC of 0.72, which is considered acceptable discrimination (Fig. 1). IFN-γ levels were not associated with AKI neither in univariate nor multivariable regression analysis (Table 3). In addition to IL-10, we also found plasma glucose, sodium, and calcium to be predictors for AKI, as shown in Table 3. We cannot rule out a negative effect on the kidneys from higher glucose levels [23]. Higher S-Na levels could also indicate dehydration, which in turn can induce renal damage. However, the numerical increase in serum levels for these traditional biomarkers was modest compared to the more than fourfold increase in IL-10. So, although statistically significant, we interpret these findings as probably not clinically significant. On the other hand, with a more than fourfold increase in the AKI group, IL-10 seems to be a potential single marker to identify patients who will develop AKI on ECMO. Among healthy adults, normal serum levels of IL-10 range from 4.9 to 9.8 pg/mL [24]. The relation between increased IL-10 and AKI is complex. IL-10 is a pleomorphic cytokine with diverse functions and exerts both stimulatory and inhibitory roles on pro-inflammatory cytokines and inhibits lymphocytes, dendritic cells, NK cells, and macrophages [25]. The principal function of IL-10 is suggested to be the control of inflammation, and activation of the IL-10 receptor complex down-regulates the secretion of pro-inflammatory Th-2 cytokines such as TNF-alfa, IFN-γ, and IL-6. Altered expression of IL-10 is found in numerous clinical conditions, including inflammatory diseases, sepsis, autoimmune diseases, and cancer. In addition, serum concentrations of IL-10 can be affected by various therapeutic agents such as steroids, epinephrine, norepinephrine, heparin, etc. [26,27]. In the multicenter study by Payen et al. of patients with septic shock, the serum IL-10 levels on admission were associated with an increased risk of AKI and mortality [7].
Correlation between increased IL-10 levels and various kidney diseases has been described earlier, and as an anti-inflammatory cytokine, IL-10 plays an important role in the development and progression of acute and chronic kidney diseases [8]. In the kidneys, mesangial cells are the major local source of IL-10. Although it has been shown to aggravate renal injury, under some conditions, IL-10 is also demonstrated to have a protective role, including reducing kidney injury [8,28]. The therapeutic potential of IL-10 in AKI derives from its capacity to inhibit the macrophages, and experimental treatment with IL-10 was shown to ameliorate renal tubular injury in a murine model of AKI [29]. Some clinical studies have evaluated the therapeutic potential for IL-10 in autoimmune diseases and cancer, and although there is a scientific rationale for using recombinant IL-10 as a therapy, clinical experience has not shown a clear clinical benefit [30].
The fact that an IL-10 increase precedes creatinine makes IL-10 attractive as a possible early biomarker of AKI. In that case, patients with elevated levels of IL-10 might benefit from earlier circulatory support since IL-10 decreased to normal within 48 h (shown in Fig. 2). However, it remains unknown if earlier intervention guided by an increase in IL-10 would improve the prognosis.
We found no difference in mortality between the AKI and no-AKI groups, but the clinical picture is complex. To try to better understand the impact of kidney function on the prognosis of survival in VA-ECMO-treated patients, we compared patients with eGFR ≥45 mL/min, those with eGFR <45 mL/min, and patients who developed AKI during VA-ECMO treatment. The highest mortality was observed early in the period, and the development of AKI was associated with increased 30-day mortality but only compared to non-AKI patients with better kidney function (eGFR >45 mL/min) (shown in Fig. 4c). This highlights the importance of preserving renal function before and during VA-ECMO treatment and avoiding factors contributing to AKI, such as hypovolemia, nephrotoxic medications, and ECMO circuit-related hemolysis.
The data on long-term survival in VA-ECMO patients who developed AKI are limited as most studies report inhospital, 30-day, or 90-day mortality [31]. Moreover, previous studies included heterogeneous patient populations, including VV-, VA-ECMO, or mix, as well as pediatric and adult patients [32][33][34]. In our cohort of 100 VA-ECMO-treated patients who survived the first 30 days, the survival at 1 year was no different in AKI patients compared to non-AKI patients, regardless of kidney function at cannulation (shown in Fig. 4d).
This study has both strength and limitations. To our knowledge, this is the first study to investigate the association between pre-cannulation cytokine levels and AKI development in VA-ECMO-treated patients. Among the limitations are the single-center design and limited sample size.
In conclusion, AKI development during VA-ECMO may be predicted by elevated IL-10 levels before the initiation of circulatory support. The development of AKI during VA-ECMO significantly increased 30-day mortality compared to patients with better renal function (eGFR ≥45 mL/min before cannulation) and no AKI. Whether more aggressive intervention or protection of the kidneys in patients demonstrating elevated IL-10 levels will decrease the incidence of AKI, or even improve survival, requires further exploration in prospective interventional trials.

Statement of Ethics
The study was performed according to Declaration of Helsinki and its later amendments. The study protocol was reviewed and approved by the Local Institutional Review Board at Spectrum Health and Van Andel Institute, Grand Rapids, MI, USA, approval number 2016-171. Informed written consent was obtained from next to kin of all individual participants included in the study.