Cardiac dysfunction biomarkers are associated with potential for successful separation from extracorporeal membrane oxygenation in children

Abstract

Biomarkers of cardiac dysfunction may aid in decision making about organ recovery and optimal timing of separation from extracorporeal membrane oxygenation (ECMO). We conducted a prospective observational study of children 0 to <18 years who underwent ECMO between 7/2010 and 6/2015 in a single center. In this pilot study, we aimed to determine whether ST2, NT-proBNP, galectin-3 and endostatin were associated with ability to separate from ECMO. Fifty neonatal and pediatric participants supported on venoarterial ECMO were included (median age 13 days, 50% male). Twelve (24%) participants were unable to separate from extracorporeal support. Plasma ST2 concentrations at cannulation were higher in children who were ultimately unable to separate vs those who successfully separated from ECMO (median 395.3 ng/mL versus 207.4 ng/mL, p=0.012). ST2 and NT-proBNP concentrations decreased significantly from the first to the last ECMO day in patients successfully separated from ECMO (p<0.0001 and p=0.017, respectively). Endostatin concentrations increased significantly from the first to the last ECMO day in both groups. Galectin-3 concentrations were not associated with ability to separate from ECMO. Cardiac dysfunction biomarkers, particularly ST2, may aid in decannulation decision making in pediatric ECMO patients. These results should be validated with a larger study.

INTRODUCTION

Extracorporeal membrane oxygenation (ECMO) is increasingly utilized in pediatric patients to treat severe, refractory cardiopulmonary failure. Pediatric patients are cannulated onto ECMO in the setting of cardiac or respiratory failure, severe sepsis, or cardiac arrest.^1-5 Within the last decade, ECMO utilization in pediatrics has approximately doubled.^1,2,4 Survival to successful separation from ECMO is reported between 56% and 82%, depending on age and ECMO indication,² and has not changed significantly over time despite increasing medical complexity of pediatric ECMO patients.^1,2,4

Survival following ECMO support is difficult to predict. To date, studies have identified a wide range of variables that are associated with lower survival on ECMO ranging from pre-cannulation data,^4,6 medical comorbidity,^4,7,8 and complications while on ECMO.^5-8 Additionally, predictive models encompassing pre-ECMO data have been validated to risk-adjust patients’ likelihood of survival.^9,10 Despite this body of research regarding ECMO survival, models to predict optimal timing of decannulation do not exist. The decision to separate from ECMO is complicated and relies on clinical, hemodynamic, laboratory, and echocardiogram data. Early separation from ECMO may increase risk of mortality or need to re-establish extracorporeal support, while delayed separation may increase risk of ECMO-related complications.

Plasma biomarkers are noninvasive, rapid, and objective measures of disease processes that are increasingly studied in pediatric populations. They have the potential to assist in decision making regarding optimal timing of separation from ECMO. Suppression of tumorigenicity 2 (ST2), N-terminal pro–B-type natriuretic peptide (NT-proBNP), galectin-3, and endostatin have been amply studied and found to be associated with outcomes in patients with heart failure. They are secreted in response to volume or pressure overload and cardiomyocyte stress and are associated with the development of cardiac hypertrophy, fibrosis, and dysfunction.^11-15 The 2017 American College of Cardiology, American Heart Association, and Heart Failure Society of America (ACC/AHA/HFSA) guidelines recommend measuring NT-proBNP (level of evidence A), ST2, and galectin-3 (level of evidence B-NR) for diagnosis and prognostication in adults with heart failure.^13,16-27 Endostatin is newly recognized in heart failure prognostics.²⁸

To our knowledge, biomarkers of cardiac dysfunction have not been studied in children with severe cardiopulmonary failure requiring ECMO support. In this exploratory pilot study we aimed to determine if concentrations of widely recognized biomarkers of cardiac dysfunction in adults, including ST2, NT-proBNP, galectin-3 and endostatin, are associated with ability to separate from ECMO. We hypothesized that concentrations of biomarkers associated with cardiac failure would be elevated at cannulation, and remain elevated in children unable to successfully separate from ECMO.

MATERIALS AND METHODS

Study Design and Participants

This is a secondary analysis of a prospective observational study of neonatal and pediatric patients <18 years who were cannulated onto ECMO for any indication within an academic, urban, quaternary care pediatric intensive care unit between July 2010 and June 2015. Details on the parent two-center prospective cohort have been published.⁷ Briefly, written informed consent was obtained from guardians within 24 hours of cannulation. We excluded patients cannulated at an outside hospital, patients on ECMO for less than 6 hours, those with limitation of care discussions within 24 hours of cannulation, non-index ECMO cannulations during the same hospitalization, and patients in foster care or wards of the state. For this secondary analysis, we further excluded those on venovenous ECMO support, those with no echocardiogram studies around the time of cannulation and separation from ECMO, and those with ECMO courses <48 hours and only one plasma sample collected. This study was approved by the Johns Hopkins Institutional Review Board.

Demographic and clinical data were collected prospectively.⁷ We reviewed echocardiograms obtained for clinical purposes that were closest to cannulation and time of separation from ECMO. If a patient had multiple echocardiograms within 24 hours of cannulation or separation from ECMO, the echocardiogram indicating the most severe dysfunction was included in the analysis. We reviewed clinical echocardiogram reports and two coauthors (MG and MN) further reviewed images to achieve consensus for qualitative assessment of right ventricular or left ventricular function (mild, moderate, or severe dysfunction).

The primary outcome was inability to separate from ECMO defined as mortality on ECMO, transition to ventricular assist device, or cardiac transplant. The secondary outcome was presence of any left or right cardiac dysfunction at the end of the ECMO course.

The ECMO circuits during the study period consisted of custom-packed 1/4- or 3/8-inch flexible PVC tubing (Medtronic, Minneapolis, MN) with a silicone reservoir, a bladderbox (Johns Hopkins Hospital, Baltimore, MD), a 0.8 m² to 4.5 m² membrane oxygenator (Medtronic), a heat exchanger (Medtronic), and a roller pump (Sorin Cardiovascular USA, Arvada, CO) up to January 2011, and the Quadrox-ID oxygenator (Maquet Cardiopulmonary, Rastatt, Germany), the Better Bladder (Coastal Life Systems Inc, San Antonio, TX), and a roller pump (for infants <10 kg) or centrifugal pump (for children ≥10 kg) (Sorin), thereafter.

Biomarker Samples

We assayed banked plasma samples at two time points: on the first and on the last ECMO day. Biomarker specimens were assayed for cardiac dysfunction-related proteins from remaining fractions of plasma from daily complete blood cell count tests (EDTA-treated tubes) and 3.2% sodium citrate samples obtained for study purposes. Platelet-poor plasma was separated by centrifugation and stored at −80°C. A multiplex ELISA was developed to measure ST2, galectin-3, NT-proBNP, and endostatin simultaneously using robotically spotted capture antibodies on a 96-well plate format.

Statistical Analysis

We conducted descriptive data analysis to summarize patient and ECMO course data. Continuous variables are presented in medians with interquartile ranges (IQRs), and categorical variable with counts and frequencies. Between-group comparisons were done using Mann-Whitney U tests for continuous variables, and Fisher’s exact or Chi-squared tests for categorical variables. Within-person biomarker changes from the first to the last ECMO day were assessed using Wilcoxon matched-pairs signed rank tests. All hypotheses were 2-sided. A p-value of 0.05 was considered significant. Multivariable logistic regression analyses were performed, controlling for potential confounding factors (age and primary indication for ECMO), to assess the association between biomarkers concentrations on the first ECMO day and ultimate ability to separate from ECMO support. Odds ratios (OR) and 95% confidence intervals (CI) were reported for factors that were independently associated with the outcome of interest. Analysis was conducted with Stata 16.1 (StataCorp, College Station, TX), and GraphPad Prism (v9.1.0, GraphPad Software Inc., La Jolla, CA).

RESULTS

Sixty-four children were enrolled at one center in the parent cohort. Fourteen were excluded from this secondary analysis for: venovenous ECMO support (n=5), no or only one echocardiogram around the time of cannulation and separation from ECMO (n=6), or ECMO course <48 hours and only one plasma sample collected (n=3). The final cohort included 50 study participants.

Median age was 13 days (IQR, 2-305 days) and 50% were male. Twenty (40%) participants were cannulated for cardiac indication, 17/50 (34%) for primary respiratory indication, 7/50 (14%) for septic shock, and 6/50 (12%) for extracorporeal cardiopulmonary resuscitation (ECPR). Patient and ECMO course characteristics are presented in Table 1.

Table 1.

Characteristics of the Population for Study of Cardiac Dysfunction Biomarkers in Pediatric ECMO

Variable	All ECMO (n=50)	Successfully Separated from ECMO (n=38)	Unable to Separate from Extracorporeal Support (n=12)	p-value
Age, days	13.3 [2.0, 305.4]	10.3 [2.3, 305.4]	29.7 [1.9, 701.9]	0.90
Male	25 (50)	19 (50)	6 (50)	> 0.99
Race				0.76
Caucasian	31 (62)	24 (63)	7 (58)
African American	13 (26)	9 (24)	4 (33)
Other	6 (12)	5 (13)	1 (8)
Hispanic	2 (4)	2 (5)	0	> 0.99
Weight, kilograms	3.8 [3.1, 11.2]	3.9 [3.1, 11.1]	3.8 [3.1, 10.8]	0.89
Primary Indication for ECMO				0.037
Respiratory	17 (34)	15 (39)	2 (17)
Cardiac	20 (40)	11 (29)	9 (75)
Sepsis	7 (14)	6 (16)	1 (8)
ECPR	6 (12)	6 (16)	0
Vasoactive-Inotropic Score, pre-ECMO	15 [6.4, 30.0]	20 [7.7, 34.4]	10 [3.8, 17.5]	0.067
ECMO Duration, days	5.7 [3.8, 13.3]	4.9 [3.7, 11.1]	14.0 [6.4, 25.6]	0.015
Reason for Decannulation				n/a
Improved Organ Function	38 (76)	38 (100)	0
Death/VAD/Heart Transplant	12 (24)	0	12 (100)
RV dysfunction by echocardiography
Cannulation (n=49)	38 (78)	28 (26)	10 (91)	0.41
Decannulation (n=44)	15 (32)	9 (26)	6 (67)	0.044
LV dysfunction by echocardiography
Cannulation (n=47)	24 (51)	19 (51)	5 (50)	> 0.99
Decannulation (n=47)	11 (23)	7 (10)	4 (40)	0.21
Hospital mortality	18 (360)	8 (21)	10 (83)	< 0.001
Hospital length of stay, days	45.5 [23.5, 101.3]	54 [31, 115]	25 [13.8, 36.5]	0.015

Median [interquartile range] or n (%)

ECMO, extracorporeal membrane oxygenation; RV, right ventricle; LV left ventricle; VAD, ventricular assist device

Twelve (24%) participants could not separate from ECMO: ten died on ECMO, one transitioned to a ventricular assist device (VAD), and one received a heart transplant. Among those who died, all continued to demonstrate cardiopulmonary failure. Seven of these 10 participants exhibited systolic right and/or left ventricular dysfunction on their final echocardiogram, one exhibited low normal function but was unable to hemodynamically tolerate circuit clamp due to pulmonary hypertension, one child had normal function on their final echocardiogram 4 days prior to developing a non-survivable tension pneumothorax, and one exhibited normal left ventricular function, but right ventricular function was unable to be assessed due to technical issues. Among the 38 children who separated from ECMO, 8 died prior to hospital discharge.

All 50 participants had biomarkers measured on the first and the last ECMO day, within 24 hours of ECMO initiation and ECMO discontinuation, respectively.

Plasma Cardiac Dysfunction Biomarkers on the First ECMO Day and Ability to Separate from ECMO

ST2 plasma concentrations on the first ECMO day were significantly higher in patients unable to separate compared to those successfully separated from ECMO (395.3 ng/mL, IQR, 271.9-487.7, versus 207.4, IQR, 98.7-384.5, p=0.012). Plasma concentrations of NT-proBNP, galectin-3, and endostatin on the first ECMO day did not differ significantly by ability to separate from ECMO (Table 2 and Fig. 1).

Table 2.

Cardiac Dysfunction Biomarker Concentrations on the First ECMO Day, Overall and by Ability to Separate from ECMO (ng/mL)

Biomarker	All ECMO (n=50)	Successfully Separated from ECMO (n=38)	Unable to Separate from Extracorporeal Support (n=12)	p-value
ST2	222.1 [116.9, 424.1]	207.4 [98.7, 384.5]	395.3 [271.9, 487.7]	0.012
NT-proBNP	22.7 [7.8, 36.9]	22.9 [7.8, 39.2]	22.7 [10.6, 28.6]	0.83
Galectin-3	16.4 [12.1, 27.7]	16.4 [13.1, 26.7]	17.4 [11.9, 49.1]	0.72
Endostatin	52.5 [28.7, 87.0]	55.2 [30.6, 88.9]	42.1 [26.3, 65.2]	0.52

All values are reported as median [interquartile range]

ST2, suppression of tumorgenicity-2; NT-proBNP, N-terminal-proBNP

Figure 1. Plasma cardiac dysfunction biomarkers on the first and last ECMO day stratified by ability to separate from ECMO.

Successful separation from ECMO group (black circles); unable to separate from ECMO group (gray triangles)

ST2, Suppression of tumorigenicity 2; NT-proBNP, N terminal pro-BNP

p-values for between-patient comparison are based on Mann-Whitney U test and p-values for within-person comparisons are based on Wilcoxon matched-pairs signed rank tests (* indicates p < 0.05, *** indicates p < 0.001, **** indicates p < 0.0001)

The area under the receiver operator characteristic curve for inability to separate from ECMO for ST2 was 0.74 (95% CI, 0.60-0.88), with the optimal cutoff for plasma ST2 concentration (i.e., value providing the maximal sum of sensitivity and specificity) found to be 294.5 ng/mL. After adjusting for age (neonatal vs pediatric) and primary ECMO indication, the adjusted odds ratio for inability to separate from ECMO was 24.4 (95% CI, 1.96-304.3) for ST2 on the first ECMO day, using a cutoff of 294.5 ng/mL. NT-proBNP, galectin-3 and endostatin were not significantly associated with the outcome of ability to separate from ECMO after adjusting for age and primary ECMO indication.

Within-person Changes in Plasma Cardiac Dysfunction Biomarkers from the First to the Last ECMO Day

Among all patients, within-person ST2 plasma concentrations decreased significantly from the first to the last ECMO day. Endostatin concentrations increased significantly, while NT-proBNP and galectin-3 concentrations were not significantly different on the last vs the first ECMO day.

Within-person ST2 plasma concentrations decreased significantly from the first to the last ECMO day in those who were separated from ECMO, median 207.4 ng/mL (IQR, 98.7-384.5) to 82.7 ng/mL (IQR 40.5-197.7), p<0.0001, with no significant decline among those unable to successfully separate from ECMO, median 395.3 ng/mL (IQR, 271.9-487.7) to 297.8 ng/mL (IQR, 105.9-418.5), p=0.15.

Endostatin concentrations increased significantly from the first to the last ECMO day in both groups, from a median 42.1 ng/mL (IQR, 26.3-65.2) to 109.9 ng/mL (IQR, 89.4-134.5), p<0.001, in those unable to separate from ECMO, and from a median 55.2 ng/mL (IQR, 30.6-88.9) to 79.9 ng/mL (IQR, 52.5-120.9), p=0.001, in patients who were successfully separated from ECMO, respectively.

NT-proBNP decreased significantly from the first to the last ECMO day within those separated from ECMO successfully, median 22.9 ng/mL (IQR, 7.8-39.2) to median 8.7 ng/mL (IQR, 3.7-27.3), p=0.017, with no significant changes in those unable to separate from ECMO. There were no significant within-person changes in galectin-3 concentrations in either group (Table 3 and Fig. 1).

Table 3.

Within-person Change in Cardiac Dysfunction Biomarker Concentrations from the First to the Last ECMO Day, Overall and by Ability to Separate from ECMO (ng/mL)

Biomarker	All ECMO (n=50)	p-value	Successfully Separated from ECMO (n=38)	p-value	Unable to Separate from Extracorporeal Support (n=12)	p-value
ST2		<0.0001		<0.0001		0.15
Cannulation	222.1 [116.9, 424.1]		207.4 [98.7, 384.5]		395.3 [271.9, 487.7]
Decannulation	102.3 [43.2, 269.4]		82.7 [40.5, 197.7]		297.8 [105.9, 418.5]
NT-proBNP		0.11		0.017		0.28
Cannulation	22.7 [7.8, 36.9]		22.9 [7.8, 39.2]		22.7 [10.6, 28.6]
Decannulation	16.7 [4.2, 32.1]		8.7 [3.7, 27.3]		27.1 [19.2, 34.5]
Galectin-3		0.21		0.10		0.97
Cannulation	16.4 [12.1, 27.7]		16.4 [13.1, 26.7]		17.4 [11.9, 49.1]
Decannulation	15.9 [10.1, 25.3]		15.6 [7.1, 19.9]		22.9 [11.2, 43.3]
Endostatin		<0.0001		0.001		<0.001
Cannulation	52.5 [28.7, 87.0]		55.2 [30.6, 88.9]		42.1 [26.3, 65.2]
Decannulation	91.3 [56.7, 124.4]		79.9 [52.5, 120.9]		109.9 [89.4, 134.5]

All values are reported as median [interquartile range]

ST2, suppression of tumorgenicity-2; NT-proBNP, N-terminal-proBNP

Cardiac Dysfunction Biomarker Concentration and Ventricular Dysfunction by Echocardiography

All 50 patients had echocardiograms performed around the time of ECMO initiation and of separation from ECMO. The median time between ECMO initiation and echocardiogram was 2.6 hours (IQR, 0.8-11.7 hours) and the median time between separation from ECMO and echocardiogram was 7.44 hours (IQR, 2.7-39.2 hours). Overall, there were no significant differences in plasma concentrations of biomarkers on the first or last ECMO days by presence vs absence of ventricular dysfunction by echocardiography (Table 4).

Table 4.

Plasma Cardiac Dysfunction Biomarkers by Presence of Cardiac Dysfunction on Echocardiogram (ng/mL)

Biomarker	Cardiac Dysfunction Present	Cardiac Dysfunction Absent	p-value
First ECMO day	n=41	n=9
ST2	228.1 [116.9, 424.1]	158.2 [138.0, 236.8]	0.32
NT-proBNP	25.0 [7.8, 36.2]	18.7 [2.5, 39.2]	0.54
Galectin-3	15.0 [11.8, 27.7]	24.6 [18.3, 26.7]	0.09
Endostatin	52.2 [29.8, 86.7]	71.4 [28.7, 88.9]	0.70
Last ECMO day	n=19	n=31
ST2	205.7 [46.3, 369.9]	89.7 [36.4, 197.7]	0.05
NT-proBNP	27.1 [12.2, 33.4]	8.4 [3.7, 31.6]	0.08
Galectin-3	18.8 [11.1, 33.6]	13.0 [7.0, 21.3]	0.16
Endostatin	86.1 [51.8, 124.4]	92.0 [56.7, 125.6]	0.91

All values are reported as median [interquartile range]

ST2, suppression of tumorgenicity-2; NT-proBNP, N-terminal-proBNP

DISCUSSION

This prospective observational pilot study is the first examining cardiac dysfunction biomarkers in a pediatric ECMO population. Among the four biomarkers studied, ST2 was significantly associated with ability to separate from ECMO in unadjusted and adjusted analyses, while NT-proBNP, galectin-3 and endostatin were not.

ST2 was associated with potential to separate from ECMO. Plasma levels on the first ECMO day were significantly higher among patients ultimately unable to separate compared to those successfully separated from ECMO. This association remained significant after adjusting for age and primary ECMO indication. In addition, within-person analyses showed a significant decline in ST2 levels from the first to the last ECMO day among children who were successfully separated from ECMO, not seen within those unable to separate from extracorporeal support. Our results reflect findings described in adult heart failure populations where ST2 has been associated with mortality, and is utilized for prognostication and risk stratification, including in the acute decompensated state.^{16,22,23,25,29} Our results also indicate that change in ST2 concentration over time may be as important as absolute value. Boiset et al. measured ST2 concentrations at 6 timepoints during inpatient acute decompensated heart failure admissions and found that a larger percent decrease in ST2 concentration from admission to discharge was associated with survival, and that this was superior to initial ST2 concentration in predicting outcome.²⁵ Persistent ST2 elevations and increases over time have also been associated with mortality.^30,31 Overall, these pilot data provide support to pursue ST2 as a prognostic biomarker in critically ill pediatric patients on ECMO in future studies.

NT-proBNP concentrations on the first ECMO day were not associated with outcomes. NT-proBNP decreased significantly within patients who were ultimately separated from ECMO, and rose, though not significantly, among those unable to separate from extracorporeal support. Changes in NT-proBNP levels have been superior to initial levels in predicting mortality in some acutely decompensated or reduced ejection fraction heart failure studies in adult populations.^17,21,32 Kubler et al. showed that a lack of decrease in NT-proBNP in acute decompensated heart failure patients during the period of clinical stabilization was associated with higher mortality at one year.³² Other studies have shown that discharge NT-proBNP levels are more useful prognosticators than admission levels.^33,34 In our study, ST2 had a more significant association with outcome than NT-proBNP. While ST2 may better predict mortality,^23,31 NT-proBNP is the most widely recognized biomarker of heart failure and has been validated in diagnosis and prognostication.^16-18,27,35 Our findings may reflect a disproportionate influence of other variables known to affect NT-proBNP, such as renal function, body mass index, gender, and age.^23,35 Validating findings using a larger sample size could control for these potential confounders.

Galectin-3 showed no significant association with ability to separate from ECMO. Evidence to support prognostic utility of galectin-3 in heart failure is mixed. Current heart failure guidelines recommend measuring galectin-3 for additive risk stratification in chronic heart failure (level IIb).¹⁶ Among sicker heart failure populations (acute decompensated or those with reduced ejection fraction), this biomarker is less useful for prognostication,^36-38 and change in galectin-3 concentration over time has been found not to assist with prognostication.³⁷ This may be because galectin-3 is associated with the development of fibrosis, an irreversible, slower process, and not dynamic ventricular stretch, and so concentrations may not reflect clinical change during the rapid progression of cardiopulmonary failure supported with ECMO.

Endostatin concentrations in both groups increased significantly over time from the first to the last ECMO day. There is conflicting evidence regarding potential beneficial³⁹ versus harmful²⁸ roles of endostatin in the pathophysiology of heart failure. In a rat model, endostatin was shown to increase in cardiomyocytes in the setting of hypoxia. Treating these rats with anti-endostatin antibodies was associated with adverse left ventricular remodeling and fibrosis.³⁹ In the current study, it is possible that endostatin concentrations increased over the duration of ECMO as a protective mechanism in all patients, regardless of ultimate outcome. Additional research is necessary to elucidate the pathophysiology underlying this change. Endostatin concentrations on the first ECMO day were not significantly different between the two outcome groups. Similarly, in a large study of chronic heart failure, endostatin concentrations did not predict adverse outcome.⁴⁰

In pediatric patients requiring ECMO, prognostic biomarkers are particularly relevant. Readiness to separate from ECMO can be difficult to discern, and practitioners rely on a combination of echocardiogram and clinical data to make subjective assessments. In adults with heart failure, biomarkers are utilized to reclassify risk for outcomes, including death, over traditional clinical data such as ejection fraction and volume overload.^33,27 Biomarkers are beneficial because they are objective and non-invasively measured. Our pilot study demonstrates that ST2, and perhaps NT-proBNP are associated with ECMO outcome, and have potential to facilitate decision making regarding recovery potential and readiness to separate from ECMO. A larger sample size is needed to confirm the validity of these results, and assess cut off points associated with successful separation from ECMO. Additionally, biomarker samples in our study were collected on the first and the last days of ECMO support. Optimal collection timepoints for prognostication are unknown. It is possible that more frequent measurements would yield improved information regarding readiness for separation from ECMO. We found that ST2 and NT-proBNP decreased significantly from the first to the last day of ECMO in those successfully separated from extracorporeal support. Trending serial biomarker concentrations after ECMO initiation may provide earlier evidence of cardiac recovery, or could help determine optimal ECMO flow. During weaning of ECMO support, cardiac output depends increasingly on intrinsic myocardial function, and changes in biomarker concentrations could assist with prognostication: stable biomarker concentrations would be expected to reflect true cardiac recovery and thus decannulation readiness, while increasing concentrations during this period may indicate that ECMO was adequately supporting cardiac function but that the patient may not yet be ready to successfully separate from extracorporeal support. Taken together, ST2 and possibly NT-proBNP have potential to assist with risk stratification and clinical decision making among some of the sickest pediatric patients, when determining optimal timing for separation from ECMO.

This study has several limitations. In a patient population with multiple organ involvement and a relatively small sample size, our analysis was limited in its ability to control for variables that may confound biomarker concentration results. NT-proBNP and galectin-3 concentrations, in particular, are affected by renal clearance,^35,36 and renal replacement therapy is often utilized during ECMO. Elevations in biomarker concentrations may be accentuated by kidney injury. Future larger studies are needed to control for renal dysfunction and to determine whether elevations in some biomarkers may represent processes affecting both the heart and the kidneys vs cardiovascular disease leading to renal dysfunction. ST2’s prognostic value is not thought to be strongly influenced by renal dysfunction. ST2 has the most promise as a prognostic biomarker in ECMO patients, and its validity should be confirmed in a larger population evaluating its performance characteristics as a marker of cardiac recovery. Second, there was no association between biomarker concentration and echocardiographic measures of dysfunction. The reason is likely multifactorial and may be related to technical aspects of our study. Echocardiogram data were clinically collected, only qualitative measures of ventricular function were available, and right and left ventricular function were grouped together due to the small sample size. Prior work regarding ST2, for example, found an association with some (e.g. left ventricular ejection fraction, right ventricular systolic pressure, among others), but not all indices of function (e.g. no other right ventricular parameters were significant in multivariate analysis).^23,24 Furthermore, because ECMO inherently affects cardiac hemodynamics, these biomarkers may not maintain the same association with cardiac function as has been found in non-ECMO populations.^23,24,28 Future research should include echocardiograms with quantitative measures of dysfunction. Third, these data were measured in a heterogenous ECMO population consisting of cardiac and respiratory indications for ECMO support. Future research, focused specifically on cardiac indications for ECMO cannulation may strengthen the findings. Last, it would be interesting to collect pre-ECMO samples in all study participants. Many of these biomarkers are released in response to cardiomyocyte stress, and so could be altered by the cannulation process or the ECMO support itself. It would also be of interest to determine whether biomarker concentrations are associated with need for ECMO support in the first place, and to extend the biomarker panel to include novel biomarkers indicative of different pathophysiological processes involved in cardiac dysfunction and cardiogenic shock.⁴¹

CONCLUSIONS

In conclusion, in this prospective observational pilot study of pediatric patients on ECMO, lower ST2 concentrations on the first day of ECMO, and a decrease in ST2 concentrations over time were associated with successful separation from ECMO. NT-proBNP levels also decreased significantly over time in those who were successfully separated from ECMO support. While both of these biomarkers are well studied in adults with myocardial dysfunction, they are less well studied in pediatrics, and this is the first study to our knowledge assessing these biomarkers in pediatric ECMO patients. Readiness to separate from ECMO can be difficult to determine and relies on subjective decision making based on clinical information. These biomarkers of cardiac dysfunction have potential to provide additional objective data to impact these decisions.