Investigating biomarkers associated with mortality in patients receiving VA-ECMO for cardiogenic shock: a systematic review

Methods

This systematic review was conducted utilizing the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA) 2020 guidelines [10]. Analogous search strings were developed to implement a comprehensive literature search across multiple databases including Embase, Scopus, PubMed, and Cumulative Index to Nursing and Allied Health Literature (CINAHL). The literature review commenced in October 2024 with the aid of the library staff at the Richard A. Henson Research Institute. The search string for all databases include: ([(mortality OR prognosis) AND cardiogenic shock AND (VA ECMO OR VV ECMO OR ECMO OR extracorporeal membrane oxygenation) AND (risk factors OR biomarkers OR physiological markers)] AND y_5[Filter] AND English[Filter]). The search limits include data restriction of 1/1/2019 to 10/22/2024 and in the English language. The inclusion criteria were all genders aged 18 and older being treated for cardiogenic shock with VA-ECMO and intra-ECMO biomarker data with corresponding mortality data. Specific biomarkers included in the criteria were: lactate, creatinine, bilirubin, aspartate transaminase (AST), alanine transaminase (ALT), troponin, CRP, and white blood cell (WBC) count. These markers were chosen based on existing data for predictive usage in ECMO and relate to common causes of cardiogenic shock. Lactate, creatinine, bilirubin, AST, and ALT are utilized in the SAVE score for predicting survival in cardiogenic shock patients considering ECMO [11]. Troponin is an established marker of cardiomyocyte perfusion. A leading cause of cardiogenic shock is myocarditis, which could lead to CRP elevation and WBCs for infection-related cardiogenic shock such as endocarditis.

The search yielded 1,033 studies, with 650 studies remaining after the removal of duplicates. Then 538 studies were eliminated during the title and abstract phase. The titles that did not mention “VA-ECMO” or “cardiogenic shock” were eliminated from further review. Alternatively, if the title mentioned “VV-ECMO” or “pediatric population” or any other pathology not leading into cardiogenic shock, it was eliminated from further review. If the abstract did not mention biomarkers (lactate, creatinine, bilirubin, AST, ALT, troponin, CRP, and WBCs) or if the aim of the study was not explicitly looking at mortality, it was eliminated from further review. A total of 112 studies remained for the full-text review; common reasons for exclusion included articles with conference abstracts only, no mention of specific biomarkers, and wrong comparison of treatment modalities (Figure 1). Reviewer consensus was required when making decisions on which articles to include and exclude. When consensus was not met initially, adjudication was carried out by the five authors participating in the screening process (J.R., T.A., T.H., D.P., and A.C.); this was performed via a group meeting and having discussion among all group members as to why the article should or should not be included. Consensus as a group was the deciding factor if the study met the exclusion criteria, thereby being eliminated from further review or incorporation into the studies included in this systematic review. Five studies were included in the data extraction phase. The methodological quality of the included studies was evaluated utilizing the Downs and Black checklist, which assesses reporting, external validity, internal validity (bias and confounding), and study power. The Downs and Black tool was selected due to its applicability to both randomized and non-randomized studies, aligning with the design of the included research. Scores were categorized as high (≥21), moderate (14–20), or low (≤13) quality [12].

Figure 1:

Preferred reporting items for systematic reviews and meta-analyses (PRISMA) diagram.

Results

Data abstracted from four retrospective and one prospective cohort study reported a total of 589 patients supported by VA-ECMO following cardiogenic shock. Sex or gender was reported for all studies, with two studies utilizing convenience categories of male (n=230, or 39.0 %), two studies reporting female (n=50, or 8.5 %), and one study reported 88 (14.9 %) males and 64 (10.9 %) females. The combined cohort consisted of 399 (67.7 %) males and 190 (32.3 %) females, with an age range between 45 and 77 years old. When the reason for cardiogenic shock was defined, cardiac surgery (coronary artery bypass surgery [CABG], valve repair, or a combination) was the most common antecedent. Common comorbidities include diabetes mellitus, hypertension, and vascular disease. Overall mortality was 59.9 % (353/589) based on survival to 30-day post-ECMO initiation or hospital discharge.

Ding et al. [13] performed a retrospective cohort analysis on postcardiogenic shock patients supported with VA-ECMO at a tertiary care hospital. A total of 283 patients were included in the study; an in-hospital mortality rate of 140 (49.5 %) was reported. Lactate (mmol/L) at ECMO initiation for survivors and non-survivors was 9.63 ± 6.04 and 11.90 ± 5.82, (p=0.001) respectively. Univariate analysis demonstrated an odds ratio (OR)=1.067 (1.024–1.111, 95 % confidence interval [CI]) and p=0.002. Multivariate analysis showed an OR=1.078 (0.952–1.221, 95 % CI) and p=0.234. They also reported serum creatinine (Scr) (mg/dL) 48-h post-ECMO initiation. Survivors had a Scr of 1.32 ± 0.73, whereas non-survivors demonstrated a Scr of 1.75 ± 0.96 (p=0.0). Univariate analysis demonstrated an OR=1.863 (1.360–2.551, 95 % CI) and p=0.0; multivariate analysis with an OR=0.497 (0.082–3.020 95 % CI) and p=0.448.

Djordjevic et al. [14] conducted a retrospective analysis of 64 patients at a single institution who underwent VA-ECMO for postcardiotomy cardiogenic shock (PCCS) from fulminant right ventricular (RV) failure. The study reported a 30-day mortality rate of 88 % (56 out of 64 patients). The authors investigated several physiological markers to identify potential predictors of mortality, including lactate, platelets, hemoglobin, CRP, WBC count, creatine kinase-myocardial band (CK-MB), creatinine, urea, bilirubin, AST, and ALT. Only significant data were reported between survivors and non-survivors, which occurred in baseline hemoglobin levels, peak CK-MB, and average lactate at 24 h after ECMO initiation. The mean lactate levels were significantly lower in survivors (2.8 ± 1.4 mmol/L) compared to non-survivors (5.6 ± 4.7 mmol/L), with a p<0.01.

Hu et al. [15] performed a retrospective cohort study by querying a national database for cardiac surgery patients in Australia and New Zealand supported with VA-ECMO for PCCS. A total of 61 patients were identified, of which 27 (44 %) survived to hospital discharge or 30 days post-ECMO initiation. The authors compared peak troponin levels, peak ALT, peak lactate, and nadir lactate both at 24 and 48 h. Peak troponin values were 1,045 ng/L (standard deviation [SD]=1852) and 2,834 ng/L (SD=3,235) at 24 and 48 h respectively in survivors, and 2,248 ng/L (SD=4,204) and 9,409 ng/L (SD=27,707) in non-survivors (p=0.878 at 24 h, 0.831 at 48 h). Peak ALT in survivors was 692 U/L (SD=1,450) at 24 h and 675 U/L (SD=1,513) at 48 h compared to 394 U/L (SD=589) and 399 U/L (SD=605) for non-survivors (p=0.946 and 0.924 respectively). The peak lactate level was 10.1 mmol/L (SD=4.9) and 10.1 mmol/L (SD=4.9) at 24 and 48 h, compared to 10.6 mmol/L (SD=5.2) and 9.7 mmol/L (4.5) for non-survivors (p=0.784 at 24 h and 0.872 at 48 h). Finally, nadir lactate values had a significant difference between survivors with 2.3 mmol/L (SD=2.1) at 24 h and 1.4 mmol/L (SD=0.7) at 48 h, compared to 3.9 mmol/L (SD=2.4) at 24 h and 2.3 mmol/L (SD=1.3) for non-survivors (p=0.001 for 24 and 48 h).

Laimoud et al. [16] conducted a retrospective cohort analysis on adults (age >18) requiring VA-ECMO for PCCS. A total of 152 patients were included with a mortality rate of 104/152 (68.4 %). Data were reported as a single median number as well as 25th and 75th interquartile ranges.

At ECMO initiation (median value of biomarker and interquartile ranges for survivors and non-survivors): median lactate (mmol/L) and interquartile ranges for survivors and non-survivors were 5.8 (4.8–8.3) and 9.75 (6.55–13.4) (p<0.001) respectively. Scr (μmol/L) 88 (58–121) and 115 (72–156) (p=0.006). Bilirubin (μmol/L) 17.7 (13.4–32) and 43.5 (23.65–90.65) (p<0.001). AST (units/L) 82.4 (47.9–194) and 129.35 (61.1–293.65) (p=0.027). ALT (units/L) 28 (20.7–80) and 45.85 (23.4–171.85) (p=0.079).

During ECMO support (median value of biomarker and interquartile ranges for survivors and non-survivors): lactate (12 h) 4.1 (2.8–6.4) and 11.25 (7.3–18.9) (p<0.001); lactate at 24 h 1.9 (1.4–3.1) and 6.55 (4.05–20) (p<0.001); peak lactate 10.8 (9.2–14.3) and 19.05 (14.7–30) (p<0.001); Sct (24 h) 117 (87.4–186.3) and 198 (102–231.2) (p=0.023); bilirubin (24 h) 65.4 (23.7–83) and 113.6 (73.4–189.27) (p<0.001); AST (24 h) 127 (82.3–316) and 231.4 (162.7–438.2) (p=0.032); and ALT (24 h) 79 (51.3–173) and 121.6 (93.2–232.7) (p=0.046).

Porto et al. [17] performed a prospective analysis of 29 adult patients at two institutions who underwent extracorporeal life support (ECLS) for refractory cardiogenic shock; mortality rate 65.5 % (19/29). Lactate (mmol/L) for survivors and non-survivors was measured on admission, at the start of ECLS, and 24 h (24 h) after ECLS support, and was reported as the median. Respective values for survivors and non-survivors were: Admission=4.3 and 11.2; at start of ECLS=5.4 (1.8–6.4) and 10.3 (7.8–15); and at 24 h ECLS=2.5 and 5.7. All measurements of lactate had a p<0.001. WBCs (/mm³) were also reported: 27,375 ± 4,612 and 16,120 ± 7,545 (p=0.6). Table 1 conveys the key results from each study; the results are summarized in table format (Table 2).

All five studies demonstrated moderate methodological quality based on the Downs and Black checklist (scores 16–17/27). Overall, reporting was adequate, with clearly stated objectives, well-defined outcomes, and appropriate statistical analyses. However, there were several common limitations that were identified across all studies. Most did not clearly describe the distribution of key confounders, lacked blinding of investigators or outcome assessors, and provided limited information regarding losses to follow-up or adjustment for variable follow-up durations. Further, none of the studies were randomized, and allocation concealment was not applicable. Although most studies performed some form of multivariable adjustment, one (Porto et al. [17]) did not, which only further increased the susceptibility to confounding bias. These methodological constraints suggest that the findings should be interpreted with caution and regarded as moderate-quality of evidence due to the potential for residual confounding and selection bias.

Discussion

Currently, there are no prognostic indicators of mortality when utilizing ECMO (regardless of the mode being utilized). This systematic review examined the association between serum biomarkers and mortality risk in patients receiving VA-ECMO for cardiogenic shock, with the goal of yielding a better understanding of the prognostic value of physiologic biomarkers as risk stratification tools in predicting mortality. The ultimate goal was to better assist clinicians in their decision-making regarding VA-ECMO initiation and to ultimately improve patient selection and outcomes in this high-acuity population.

Among the biomarkers studied, serum lactate emerged as the most consistently utilized and reliable predictor of mortality. While Scr, bilirubin, and liver enzymes (AST/ALT) also showed associations with outcomes in select studies, they were less frequently reported and require further investigation.

Serum lactate is already widely recognized as a marker of end-organ hypoperfusion. This explains why serum lactate was the marker most commonly utilized in previous studies. Elevated levels are correlated with worse outcomes in critically ill patients. Across all included studies, lower intra-ECMO lactate levels (<4 mmol/L) were associated with increased survival, whereas levels >5.5 mmol/L were consistently linked to higher mortality.

Interestingly, Scr, bilirubin, AST, and ALT were all reported to have statistical significance in association with mortality as well. This comes as a big surprise because these are not common biomarkers one would associate with mortality in the setting of cardiogenic shock. In the grand scheme of multiple organ system failure, we would see an increase in various biomarkers in association with its congruent organ. However, in the setting of cardiogenic shock – a primary heart issue – the statistical significance of other biomarkers having indicative value for mortality came as a surprise. The expectation would have been to see an increase in troponin, B-type natriuretic peptide (BNP), lactate, and to a certain degree Scr (due to the kidneys being the first organ system to suffer from hypoperfusion), but witnessing an increase in hepatobiliary markers was not expected.

Several limitations affected this review. First, there was limited variability in which biomarkers were reported, with four out of the five studies focusing on lactate. Two of the markers we were interested in studying, CRP and WBC count, did not appear in any studies that met our inclusion criteria. This reflects the prioritization of biomarkers more closely related to metabolism and perfusion, such as lactate and creatinine. In contrast, CRP and WBC count are more clinically relevant to inflammation or infection, which would perhaps make them less useful for prognosis in this setting. Consequently, this limited the ability to assess the prognostic value of these two markers.

Second, inconsistencies in the timing of biomarker measurement, with some studies reporting pre-ECMO values and others reporting intra-ECMOs, made it difficult to evaluate the impact of trends or dynamic changes. This heterogeneity in measurement windows limits direct comparison of prognostic thresholds and may partly account for differences in reported predictive performance. Biomarker kinetics, particularly for fluctuating biomarkers such as lactate, are time-sensitive and reflect evolving hemodynamic and metabolic states. Consequently, the prognostic value of a given biomarker depends not only on its absolute level but also on when it is measured relative to ECMO initiation or cardiac insult. Future studies should standardize biomarker sampling intervals or report time-adjusted analyses to enhance comparability and improve the clinical utility of these markers in ECMO populations.

The moderate methodological quality of the included studies, as indicated by the Downs and Black assessment tool, warrants cautious interpretation of the results. While the studies were consistent in reporting outcomes and describing patient populations, the absence of randomization, incomplete adjustment for confounders, and variability in follow-up reporting may limit the strength of causal inferences. These methodological gaps highlight the need for more rigorously designed prospective studies that apply standardized biomarker timing, blinding procedures, and comprehensive confounder analysis to strengthen evidence for biomarker-guided prognostication in VA-ECMO patient. Lastly, patient cohorts varied in terms of baseline characteristics and comorbidities, with limited control for potential confounders, further complicating comparisons across studies.

Conclusions

This systematic review was able to reveal that there are biomarkers available that can be utilized by clinicians to monitor their patients’ mortality prognosis when being treated with VA-ECMO for cardiogenic shock. It further demonstrates that lactate has strong potential as a clinical tool to predict poor outcomes even during ECMO perfusion. Lactate’s utility in guiding fluid resuscitation in severe sepsis has already been established. Based on this review, more research needs to be conducted on lactate’s potential to guide perfusion parameters with VA-ECMO flow or the addition of cardiac support. This finding aligns with its widespread use, broad availability across healthcare settings, and familiarity among clinicians. However, serum lactate should not be utilized in isolation due to its levels being increased in a multitude of other pathologies and disease states; it must be interpreted alongside other laboratory findings and within the context of overall clinical judgment. First, further research is needed to identify additional biomarkers that, when utilized in combination with lactate, may influence treatment decisions to prevent mortality in VA-ECMO patients with cardiogenic shock. Preliminary evidence suggests that creatinine, bilirubin, and liver enzymes (AST/ALT) hold promise as complementary biomarkers, although more robust data are required to confirm their clinical utility. Secondly, further research is needed to have more consistent timing of biomarker measurement to assess if mortality risk is associated with a crude biomarker value or if the rate of change has more mortality value. Lastly, further research is needed to investigate the minimal concentration of these biomarkers associated with mortality to allow clinicians a better predictor of a patient’s mortality on VA-ECMO and when more aggressive interventions may be needed in relation to the value of the respective biomarker.

Emphasis on early biomarker-based risk prediction supports the principles of holistic, preventive, and patient-centered care. Recognizing and addressing physiologic dysfunction early aligns with the osteopathic focus on maintaining the body’s natural capacity for self-regulation and healing. Integrating these markers into patient assessment may help clinicians anticipate complications, tailor interventions, and ultimately improve outcomes in this critically ill population.