Serum Galectin-3 Predicts Mortality in Venoarterial Extracorporeal Membrane Oxygenation Patients

Objective We investigated the potential use of galectin-3 (Gal-3) as a prognostic indicator for patients with cardiogenic shock and developed a predictive mortality model for venoarterial extracorporeal membrane oxygenation (VA-ECMO). Methods We prospectively studied patients (survivors and nonsurvivors) who received VA-ECMO for cardiogenic shock from 2019 to 2021. We recorded baseline data, Gal-3, and B-type natriuretic peptide (BNP) before ECMO and 24–72 h after ECMO. We used multivariable logistic regression to analyze significant risk factors and construct a VA-ECMO death prediction model. Receiver operating characteristic (ROC) curves were plotted to assess the predictive efficacy of the model. Results We enrolled 73 patients with cardiogenic shock who received VA-ECMO support; 38 (52.05%) died in hospital. The median age was 57 years (interquartile range (IQR): 48–67 years); the median duration of ECMO therapy was 5.8 days (IQR: 4.62–7.57 days); and the median intensive care unit stay was 19.04 days (IQR: 13.92–26.15 days). Compared with the nonsurvivors, survivors had lower acute physiology and chronic health evaluation (APACHE) II scores (p < 0.001), increased left ventricular ejection fraction (p < 0.05), lower Gal-3 levels at 24 and 72 h (both p = 0.001), lower BNP levels at 24 and 72 h (both p = 0.001), and higher platelet counts (p = 0.009). Further multivariable analysis showed that APACHE II score, BNP-T72, and Gal-3-T72 were independent risk factors for death in VA-ECMO patients. Gal-3 and BNP were positively correlated (p < 0.05) and decreased significantly during ECMO treatment. The areas under the ROC curve (AUC) for APACHE II score, Gal-3-T72, and BNP-T72 were 0.687, 0.799, and 0.723, respectively. We constructed a combined prediction model with an AUC of 0.884 (p < 0.01). Conclusion Gal-3 may serve as a prognostic indicator for patients receiving VA-ECMO for cardiogenic shock. The combined early warning score is a simple and effective tool for predicting mortality in VA-ECMO patients.


Introduction
Cardiogenic shock (CS) is a complex clinical syndrome characterized by impaired cardiac function and end-organ hypoperfusion [1,2].Studies have shown that cardiogenic shock develops in 5-13% of patients with acute myocardial infarction (AMI) [3].CS can be caused by various factors, such as AMI, myocarditis, cardiomyopathy, pericardial tamponade, or postcardiac surgery.In Europe, approximately 60,000-70,000 patients experience AMI each year.Despite signifcant advances in interventional treatment and critical care, mortality from infarct-related CS remains high, reaching 50% within the frst 30 days of hospitalization [4].
Extracorporeal membrane oxygenation (ECMO) has become the primary rescue therapy for patients with severe CS.By providing adequate blood fow and oxygen supply through venoarterial extracorporeal membrane oxygenation (VA-ECMO), the damaged heart can rest and recover, leading to improved symptoms and prognosis in patients with CS.
Galectin-3 (Gal-3) is a multifunctional protein that plays a crucial role in many physiological and pathological processes, such as cell growth, diferentiation, apoptosis, cell adhesion, angiogenesis, infammation, fbrogenesis, and tumor progression.Recent studies have shown that Gal-3 is the most overexpressed protein in heart failure (HF) [5].
It can predict the progression of HF and guide its treatment.Sharma et al. found that Gal-3 is the most diferentially regulated gene associated with HF [6].Gal-3 induces the proliferation of fbroblasts and causes heterogeneous deposition of diferent types of collagen, ultimately leading to loss of cardiac function.Terefore, Gal-3 may be a novel biomarker for predicting circulatory failure.
Cardiovascular disease is a specialty of our hospital, and more than half of our ECMO patients sufer from this condition.Terefore, developing a predictive model for patients with CS is crucial.Te prognostic value of Gal-3 for patients with CS undergoing VA-ECMO has not yet been established.In this study, we aimed to investigate the association between Gal-3 levels and the clinical outcomes in patients receiving VA-ECMO.

Methods
Tis study was a prospective trial and all procedures followed ethical guidelines.All study subjects or their family members gave informed consent for the treatment protocol.Te study enrolled patients with cardiogenic shock who received VA-ECMO at Ganzhou People's Hospital from January 2019 to December 2021.All included subjects had class IV cardiac function according to the New York Heart Association (NYHA) classifcation.Our study estimated the Gal-3 levels before ECMO treatment and at 72 h for surviving and deceased patients.Based on the expected mean diference, standard deviation, and quantile values, we calculated that a total sample size of 58 was needed for our randomized controlled trial.To improve the reliability and stability of the study results, our team enrolled 73 patients.Patients were categorized into a survivor group and a nonsurvivor group based on their outcome during hospitalization.
Te inclusion criteria for this study were as follows: for patients who met the diagnostic criteria for CS and received VA-ECMO at our institution: (1) patients aged over 10 years, as our hospital currently lacks a pediatric ECMO circuit and could not provide ECMO support for younger patients; (2) presence of severe underlying heart disease (such as extensive myocardial infarction, myocarditis, arrhythmia, mechanical valve failure, or postcardiac surgery); (3) presented with typical clinical signs of shock, including persistent hypotension, oliguria, altered consciousness, and other symptoms; (4) after active fuid resuscitation, hypotension and clinical signs did not improve or worsened; or (5) hemodynamic parameters meet the following typical characteristics: (i) systolic blood pressure (SBP) ≤ 90 mmHg or mean arterial pressure drop ≥30 mmHg, or a decrease of 60 mmHg from the baseline systolic blood pressure in hypertension patients for at least 30 min; (ii) central venous pressure is normal or high; (iii) elevated left ventricular end-diastolic pressure or pulmonary capillary wedge pressure; and (iv) reduced cardiac output, and a cardiac index (CI) ≤ 2.2 [L/min]/m 2 ; and (6) informed consent from the patient or their family members for VA-ECMO treatment.
2.1.Data Collection.Basic information including gender, age, underlying disease, cause of cardiogenic shock, body mass index (BMI), and the acute physiology and chronic health evaluation (APACHE) II score were recorded for all patients before VA-ECMO.Venous blood samples were collected at three time points (before ECMO initiation, 24 h after ECMO initiation, and 72 h after ECMO initiation) to measure Gal-3, B-type natriuretic peptide (BNP), platelet, and lactate levels.Te durations of ECMO support and intensive care unit (ICU) stay were recorded, and patients were followed up by telephone 28 days after ECMO discontinuation.

VA-ECMO Terapy.
Before ECMO initiation, we assessed the patient's vascular status by ultrasound to select an appropriate puncture site and catheter size.An arterial and a venous cannula were inserted under ultrasound guidance by a specialist, and the cannula positions were verifed by ultrasound or X-ray.We used a standard oxygenator (MAQUET, Germany) that was primed with saline and connected in a sterile manner after successful cannulation of the artery and the vein.ECMO blood fow and oxygen sweep gas fow were adjusted according to the mean arterial pressure and peripheral oxygen saturation to maintain a mean arterial pressure above 65 mmHg.For systemic anticoagulation, the initial bolus dose of heparin sodium was 50 U/kg, and the maintenance infusion dose was 5-15 [U/kg]/h.Te target range of the activated partial thromboplastin time (APTT) was 55-65 s.

Statistical Method.
Statistical analysis was performed using SPSS 25.0 software.Continuous variables with normal distribution were expressed as mean ± standard deviation (mean ± SD) and were compared by the independent-samples t-test between groups.Continuous variables with non-normal distribution were expressed as median (interquartile range) and were compared by the Mann-Whitney U-test between groups.Categorical variables were tested using the chi-square test or Fisher's exact test, as appropriate.Backward stepwise logistic regression was used for mortality risk factor analysis and predictive modeling.Te receiver operating characteristic (ROC) curve was used to determine the optimal cut-of values of the risk factors, and the accuracy of each risk factor and prediction model in predicting mortality in ECMO patients was evaluated by calculating the area under the curve (AUC).A p value <0.05 was considered statistically signifcant.

Demographics and
Tere was no statistically signifcant diference in age and sex between the survivor and nonsurvivor groups of ECMO patients.Tere was no diference in Gal-3 and BNP levels between the two groups before ECMO initiation.However, after ECMO treatment, Gal-3 and BNP levels at 24 and 72 h were lower in the survivor group than in the nonsurvivor group, which was statistically signifcant (p < 0.05).Platelet counts were below normal in both groups but were higher in the survivor group than in the nonsurvivor group, which was statistically signifcant (p < 0.05).Tere was no statistical signifcance between the two groups in terms of lactate levels at 72 h after VA-ECMO initiation, duration of ECMO support, and length of ICU stay (p > 0.05) (as shown in Table 1).

Effects of VA-ECMO on BNP and Gal-3
Te Friedman test for repeated measures was used to compare Gal-3 and BNP levels at three time points in the study subjects.Tere were signifcant diferences among Gal-3 levels at baseline, 24 h, and 72 h (χ 2 � 134.33, p < 0.01), with mean ranks of 2.95, 2.03, and 1.03, respectively.Tere were also signifcant diferences among BNP levels at baseline, 24 h, and 72 h (p < 0.01), with mean ranks of 3.00, 1.52, and 1.48, respectively.Post hoc analysis with the Wilcoxon signed-rank test showed that BNP levels decreased signifcantly from baseline to 24 h and from baseline to 72 h, but not from 24 h to 72 h (p > 0.05), as shown in Table 2.

The Relationship between Gal-3 and BNP
Pearson correlation analysis was used to examine the relationship between Gal-3 and BNP at three time points in the 73 patients.Tere was a signifcant positive correlation between Gal-3 and BNP at each time point (p < 0.05), which is consistent with the average trend results of Gal-3 and BNP (as shown in Figure 2(d)).Specifcally, Gal-3-T0 (baseline) was positively correlated with BNP-T0 (r � 0.373; p � 0.01) (as shown in Figure 2(a)), Gal-3-T24 (24 h) was positively correlated with BNP-T24 (r � 0.510; p < 0.01) (as shown in

Analysis of the Risk Factors Afecting the Prognosis of ECMO Patients.
Variables with p < 0.05 in the univariate analysis were included in a binary logistic regression analysis using the backward stepwise method, as shown in Table 3. Te results showed that APACHE II score, Gal-3-T72, and BNP-T72 were independent risk factors for death in patients undergoing VA-ECMO for CS (p < 0.05), while platelets were not statistically signifcant (p > 0.05).

Te Predictive Value of ROC Curve Analysis of Risk Factors on the Prognosis of VA-ECMO Patients.
Te independent risk factors, APACHE II score, Gal-3-T72, and BNP-T72, identifed by the multivariate analysis in Table 3 are plotted as ROC curves, and the area under the curve (AUC) was calculated by software.Te results are shown in Figure 3, where all three risk factors had signifcant predictive value for mortality (p < 0.05).Te AUCs of the APACHE II score, Gal-3-T72, and BNP-T72 were 0.687, 0.799, and 0.723, respectively.Te optimal cut-of point was determined by the maximum Youden index, which corresponded to the largest AUC for Gal-3-T72.Te sensitivity and specifcity of Gal-3-T72 > 7.165 ng/ml were 0.895 and 0.429, respectively (as shown in Table 4).

Construction of a Prediction Model for Death in Patients Undergoing VA-ECMO for CS and Evaluation of Its Value.
Te APACHE II score, Gal-3-T72, and BNP-T72 were assigned based on the optimal threshold points, as detailed in Table 5. Te death prediction score was constructed as APACHE II score + Gal-3-T72 + BNP-T72, with a total score ranging from 0 to 3. Te ROC curve was plotted after assigning the raw data of the APACHE II score, Gal-3-T72, and BNP-T72 and calculating the death prediction score (as shown in Figure 4).Te optimal threshold point was 2. 5

Discussion
Tis study found that Gal-3 can be used to determine the prognosis of patients with CS undergoing VA-ECMO.Te main fndings of this study were as follows: (1) BNP and Gal-3 levels decreased signifcantly in patients with CS after VA-ECMO initiation, indicating that the ECMO technique is efective for treating patients with CS; (2) serum Gal-3 Normal distribution of measurement data is expressed as mean ± standard deviation (mean ± SD).Non-normal distribution is represented by the median, interquartile spacing (M (QL, QU)).Gal-3-T0 and BNP-T0 showed blood samples before ECMO, Gal-3-T24 and BNP-T24 at 24 h after ECMO, and Gal-3-T72 and BNP-T72 at 72 h after ECMO.T ECMO refers to the duration of ECMO treatment patients, and T ICU refers to the duration of the ICU stay.
During VA-ECMO treatment, we observed that VA-ECMO provided arterial pressure support to the patient, which altered the original physiological circulatory structure and changed the preload and afterload of the heart [7].Tis made it challenging to evaluate cardiac function using conventional examination methods.Qin et al. [8] studied the relationship between serum BNP and endothelin-1 (ET-1) levels and left atrial pump function in HF patients and found that BNP and ET-1 were negatively correlated with the left atrial ejection fraction, left atrial passive emptying fraction, and left atrial active emptying fraction (p < 0.05).Similarly, we also assessed cardiac function by BNP levels.Our study found that BNP levels at 24 h after ECMO treatment decreased signifcantly compared to before ECMO treatment, and the diference was statistically signifcant, indicating that VA-ECMO was benefcial for cardiac treatment in patients with CS [9].ECMO can provide rapid circulatory support and protect organ perfusion, allowing time for cardiac recovery and preventing multiple organ failure [10].However, the extent of BNP reduction in the late stage of ECMO depends on the patient's baseline cardiac function and recovery.In our sample, we observed that Gal-3 levels decreased signifcantly during ECMO treatment, and we hypothesized that the changes in Gal-3 might be related to the improvement of cardiac function in the patients.We further examined the relationship between BNP and Gal-3 using Pearson correlation analysis and found a strong correlation between BNP and Gal-3 at all three time points, with the strongest correlation between BNP and Gal-3 at 72 h (r � 0.523, p < 0.01), implying that serum Gal-3 could predict cardiac function.Tese results are consistent with Mahmoud U. Sani's observation [11].Interestingly, despite peripheral arterial cannulation being the exclusive method  for ECMO cannulation via the peripheral arteries, a study conducted by Fausto Biancari et al. examined the cannulation methods of 1269 patients requiring postcardiotomy VA-ECMO, which found that central arterial cannulation was associated with a higher risk of in-hospital mortality compared to peripheral arterial cannulation [12].Our ECMO team will pay attention to and collect cases with both types of cannulation in the future.
We then analyzed the basic characteristics between the survivor and nonsurvivor groups, and the results showed intergroup diferences in the APACHE II score, Gal-3, BNP, and platelets.In a study by Franziska Kaestner et al. [13], they analyzed 51 ECMO cases (including 38 VV-ECMO, 11 VAV-ECMO, and 2 VA-ECMO cases) and found a statistically signifcant diference in lactate levels between the survivor and nonsurvivor groups.Te mean value for the survivor group was 2.0 (range: 0.4-13.5)and for the nonsurvivor group it was 4.0 (range: 1-21).However, our data showed no signifcant diference in lactate levels at 72 h between the two groups.We believe that the diference in results is related to the study subjects we included.Our 73 cases were all patients with cardiac failure without severe pulmonary failure symptoms, and VA-ECMO can quickly relieve patients' ischemic-hypoxic symptoms with a signifcant treatment efect.On the other hand, the main study subjects included by Franziska Kaestner's team were administered VV-ECMO treatment for pulmonary failure.VV-ECMO poorly relieves ischemic-hypoxic symptoms in such patients compared with patients with CS, which may have led to inconsistent results.
In the multivariate model, we selected Gal-3-T72 and BNP-T72 as risk factors for analysis.Te results showed that the APACHE II score, Gal-3-T72, and BNP-T72 were independent risk factors for death in patients undergoing VA-ECMO for CS.Te ROC analysis showed that all three factors had signifcant predictive power for VA-ECMO patients, and the AUC of Gal-3-T72 was the largest (0.799).Patients with CS are characterized by organ ischemia, hypoxia, infammatory factor disorders, hemodynamic changes, and volume and pressure changes in the heart    Cardiology Research and Practice [14][15][16][17].Cardiac failure during the state of CS promotes the release of serum infammatory factors, macrophage activation, and migration to cardiac tissue, leading to a significant increase in Gal-3 that promotes cardiomyocyte fbrosis [18,19].When patients are given VA-ECMO support therapy, ECMO shares the workload done by the heart, resulting in the alleviation of cardiac failure symptoms, a decrease in the release of infammatory factors, and a gradual decrease in Gal-3 levels.
Tere are limited models for predicting ECMO outcomes, and the predictive variables difer for diferent populations.Because more than half of the patients in our ECMO cohort had cardiovascular disease, we sought to establish a prediction model for patients CS.AUC of the model comprising the APACHE II score, Gal-3-T72, and BNP-T72 was 0.884, and the model showed higher predictive accuracy than each risk factor alone.Patients with CS on VA-ECMO who scored >2 were categorized as high-risk for mortality, and those who scored ≤2 were categorized as low-risk for mortality.Tis study indicates that the combination of the APACHE II score, Gal-3-T72, and BNP-T72 can be applied to predict the outcomes of VA-ECMO and is feasible.In contrast, Franziska Kaestner included too many types of ECMO in the construction of their ECMO prediction model, including VV-ECMO, VAV-ECMO, and VA-ECMO [13]; thus, their model is not well suited for patients with CS.
Schmidt et al. developed the SAVE score to assess the prognosis of refractory CS, based on factors including the cause of CS, age, organ function, and blood pressure.Tey found that the SAVE score performed better than several standard ICU severity scores [20].Mohamed Laimoud et al. found that increasing the ∆1 SOFA score (odds ratio (OR) � 2.506, 95% CI: 1.681-3.735,p < 0.001) and increasing the blood lactate level (OR � 1.388, 95% CI: 1.015-1.898,p � 0.04) were signifcantly associated with hospital mortality after VA-ECMO support for adults with CS [21].
Te scoring system we used in our study is relatively simple and includes a limited number of indicators.However, it has proven to be more valuable in predicting mortality in ECMO patients than single indicators.Terefore, this scoring system may have clinical signifcance for patient evaluation.In future research, we hope to further investigate this by incorporating measures such as the SAVE score, SOFA score, and peak lactate levels.
Tere are several limitations to this study.First, our team failed to register the clinical trial, potentially leading to biased patient selection and jeopardizing the accuracy and applicability of our research outcomes.We will further improve our research process to ensure that our future studies comply with relevant regulations and ethical requirements.Second, the small sample size may cause possible bias in the statistical analysis and requires further multicenter and large sample studies to validate the results.Tird, most of the models constructed by our team are based on patients with cardiovascular disease and may not be suitable for use in VV-ECMO or VAV-ECMO patients.Further studies are needed to investigate the applicability of our model in these patient populations.Unfortunately, peak lactate levels were not included as an analytical measure.
Peak lactate levels are a more sensitive indicator for assessing tissue perfusion and can more accurately predict mortality in patients with VA-ECMO.Mohamed Laimoud and Mosleh Alanazi's study also supports this viewpoint; their research has shown that peak lactate levels had better performance (AUC � 0.889; 95% CI: 0.825-0.953;p < 0.001), and that lactate levels after 24 h of ECMO initiation performed best in terms of sensitivity and specifcity for discriminating mortality (AUC 0.93; 95% CI: 0.878-0.983;p < 0.001) [10].

Conclusion
Gal-3 may be used as a predictor of prognosis in patients on VA-ECMO for CS.Te VA-ECMO early warning score, which includes the APACHE II score, Gal-3-T72, and BNP-T72, has greater predictive value for mortality in ECMO patients than single indicators.Tis scoring system has the potential to provide valuable prognostic information for VA-ECMO patients.However, further validation of its generalizability is required with a larger sample size and broader population.

Figure 3 :
Figure 3: Receiver operating characteristic (ROC) curve of the predictive model and single index in predicting the prognosis of patients with venoarterial extracorporeal membrane oxygenation (VA-EMCO).

Figure 4 :
Figure 4: Receiver operating characteristic (ROC) curve of the death warning score in predicting the prognosis of patients with cardiogenic shock.

6
Characteristics.A total of 78 patients with CS undergoing VA-ECMO were enrolled in this study, with one automatic dropout, two deaths within 3 days, and two cases transferred from VA-ECMO to VAV-ECMO treatment.Terefore, the fnal analysis included 73 cases (shown in Figure1), with 35 (47.95%) in the survivor group and 38 (52.05%) in the nonsurvivor group.Te median age of the patients was 57 (48-67) years, the median ECMO treatment time was 5.8 (4.62-7.57)days,and the median ICU stay was 19.04(13.92-26.15)days.Te top three disease categories for performing VA-ECMO were AMI (32.88%), heart valve disease (28.77%), and aortic coarctation type A (21.92%).Among them, 33 (45.21%) cases were after open-heart surgery and 17 (23.29%)were percutaneous coronary intervention patients (as shown in Table . A (CI): 0.803-0.965;p < 0.05) with 81.6% positive predictive value, 88.5% negative predictive value, and 84.9% accuracy.Terefore, patients with CS undergoing VA-ECMO with a score >2 were classifed as high-risk for the nonsurvivor group, while those with a score ≤2 were classifed as low-risk for the nonsurvivor group.

Table 3 :
Binary logistic analysis of risk factors for patients with extracorporeal membrane oxygenation.

Table 5 :
Death early warning rating table.