Fibrinogen–Albumin-Ratio is an independent predictor of thromboembolic complications in patients undergoing VA-ECMO

Veno-arterial extracorporeal membrane oxygenation (VA-ECMO) supports patients suffering from refractory cardiogenic shock. Thromboembolic complications (TeC) are common in VA-ECMO patients and are associated with increased morbidity and mortality. Valid markers to predict TeC in VA-ECMO patients are lacking. The present study investigated the predictive value of baseline Fibrinogen–Albumin-Ratio (FAR) for in-hospital TeC in patients undergoing VA-ECMO. This retrospective cohort study included patients who underwent VA-ECMO therapy due to cardiogenic shock at the University Hospital Duesseldorf, Germany between 2011 and 2018. Main exposure was baseline FAR measured at initiation of VA-ECMO therapy. The primary endpoint was the in-hospital incidence of TeC. In total, 344 patients were included into analysis (74.7% male, mean age 59 ± 14 years). The in-hospital incidence of TeC was 34%. Receiver operating characteristics (ROC) curve of FAR for in-hospital TeC revealed an area under the curve of 0.67 [95% confidence interval (CI) 0.61–0.74]. Youden index determined a cutoff of 130 for baseline FAR. Multivariate logistic regression revealed an adjusted odds-ratio of 3.72 [95% CI 2.26–6.14] for the association between FAR and TeC. Baseline FAR is independently associated with in-hospital TeC in patients undergoing VA-ECMO. Thus, FAR might contribute to the prediction of TeC in this cohort.

Definition and assessment of main exposure. Main exposure was FAR on the day of initiation of VA-ECMO therapy. Measurements of Fibrinogen and Albumin values were performed in the central laboratory of the University Hospital Duesseldorf. The FAR was calculated by dividing Fibrinogen to Albumin 12,17 . Furthermore, baseline Fibrinogen and baseline Albumin were investigated alone to see if the ratio of both values is superior to the single values. In an additional analysis, FAR on day five was analysed as alternative exposure. The decision to choose these two time points for analysis was based on the rationale that VA-ECMO initiation itself might be associated with TeC, for example due to cannula associated thrombosis. Hence, the predictive value of FAR before or after start of VA-ECMO therapy might be different. Day five was chosen based on local experiences suggesting that this might be a typical point of time where clinicians often have to decide about prognosis and whether therapy should be continued or not.
Outcome assessment. The primary endpoint of this study was the incidence of in-hospital TeC. TeC were defined as a composite of non-fatal arterial thrombosis or embolism, non-fatal venous thrombosis or embolism, non-fatal ischemic stroke, non-fatal myocardial infarction, or thromboembolic vascular mortality 18 . Arterial and venous thromboembolisms were defined as any new and symptomatic non-cardiac and non-cerebral arterial or venous thrombosis or embolism not causing death. Non-fatal myocardial infarction was defined according to the fourth universal definition of myocardial infarction 19 . Non-fatal ischemic stroke was defined according to the guidelines by the American Stroke Association 20 . Thromboembolic vascular mortality was defined as any TeC causing death. Data on TeC were extracted from electronic medical charts by personnel trained in the study definitions. TeC was confirmed when there was a clearly documented diagnosis that was approved by a physician specialized in intensive care medicine. Plausibility checks were done whenever further source documents were available. Sample size. Due to the nature of this retrospective exploratory data analysis, we did not conduct formal sample size calculation. However, based on the current literature, we expected TeC in approximately 30% of patients 8 . With an estimated study sample of 350 patients, we expected approximately 105 thromboembolic events. This allowed to include up to 10 predefined co-variables for multivariable adjustment (see "Statistical analysis"). As 117 events could be observed in this study, we were able to add two further covariates (= 12 covariates in total) to a separate analysis that was conducted post factum during review process. Statistical analysis. Patient characteristics are presented as absolute values with corresponding percentages for categorical data or as mean ± standard deviation for continuous data, as appropriate. Shapiro-Wilks test was used to test for normal distribution of data. Fisher exact test and t-tests were used to test for differences between categorical and dichotomous data. Discrimination of baseline FAR for in-hospital TeC was analyzed by receiver operating characteristics (ROC) curve and the "area under the curve" (AUC). ROC analysis was also done for Fibrinogen and Albumin alone. De Long-Test was performed to compare ROC curves. A cutoff value for FAR was determined by Youden Index. Multivariable logistic regression analysis was used to assess the independent association (Odds ratio (OR); 95% confidence interval (CI)) between elevated FAR and in-hospital TeC after adjustment by the following predefined covariables (forced entry): age, sex, chronic coronary syndrome, history of ischemic stroke, history of pulmonary embolism, arterial hypertension, diabetes mellitus, days of VA-ECMO therapy, continuous veno-venous hemodialysis treatment during hospitalization. The choice of covariates was based on literature research 2,3,21-23 and/or clinical experiences so that covariates were included if an association with TeC seemed possible. Baseline quick and baseline activated partial thromboplastin time (aPTT) could be added to an additional post factum logistic regression model. Model calibration was assessed using Hosmer-Lemeshow-Test. Net reclassification index (NRI) and integrated discrimination index (IDI) were calculated for FAR and for Fibrinogen and Albumin alone. For all statistical tests, p < 0.05 was considered significant. Analyses were performed with IBM SPSS Statistics version 26 (IBM, Armonk, New York, United States) and GraphPad-Prism © statistical software version 6 (GraphPad software Inc, San Diego, California, United States).

Discussion
The main finding of the present study is that baseline FAR is independently associated with in-hospital TeC in patients requiring VA-ECMO due to refractory cardiogenic shock. Furthermore, this study identified a cutoff of 130 for baseline FAR, which was related to a higher likelihood of TeC. The independent association between FAR and TeC was also present when FAR was measured on day 5 of VA-ECMO therapy.
Prediction and prevention of TeC in VA-ECMO patients. One of the most important issues in terms of treating VA-ECMO patients is to understand which patients could really benefit from VA-ECMO. This decision has to be faced prior to the initiation of VA-ECMO. Once initiated, another important question in terms of prognosis and risk stratification is to decide whether VA-ECMO therapy should be continued or limited, for example if patients suffer from severe complications. In the last decade, several scores such as the Survival after Veno-Arterial ECMO (SAVE) score 21 have been suggested to help clinicians with these issues, but data focused on the prediction of TeC are rare. In a small retrospective cohort study with 62 patients, Trudzinski and colleagues tried to find predictors for TeC in patients Table 1. Baseline characteristics of the whole cohort and in patients without and with TeC. Data are presented as mean ± standard deviation or as absolute values with percentages, as appropriate. TEC thromboembolic complication, AKI acute kidney injury, aPTT activated partial thromboplastin time, CVVHD continuous Veno-venous hemodialysis, VA-ECMO veno-arterial extracorporeal membrane oxygenation, FAR Fibrinogen-Albumin ratio, MI myocardial infarction. a p value of Chi-square test or two-tailed unpaired t-test after Levene's test for equality of variances.  22 27 . This prospective cohort study included 465 patients and found out that Argatroban was non-inferior to Heaprin regarding bleeding and thrombosis. In summary, data regarding anticoagulation and monitoring of coagulation parameters are still inconclusive.

FAR to predict TeC in VA-ECMO patients.
Regarding pathophysiologic mechanisms behind the association of FAR and TeC, Albumin is an essential plasma protein that has been proposed to be related to inflammatory and hemostatic processes 11 . Moreover, Albumin plays a role in the inhibition of platelet activation 15 . Fibrinogen on the other hand is an indicator of a procoagulatory status and contributes to inflammation at diverse levels 12,14 . The combination of these characteristics served as a basis to hypothesize that the ratio of Fibrinogen and Albumin may predict TeC in VA-ECMO patients as this cohort is at high risk for TeC. To date-to our best knowledge-there is only one study by Acharya and colleagues that investigated the predictive value of FAR in patients undergoing VA-ECMO 17 . In a retrospective single-center cohort study, this study analysed 157 patients regarding FAR measured within the first 24 h of VA-ECMO therapy and determined its prognostic value for the incidence of in-hospital ischemic stroke 17 . This study showed a significant association between an elevated FAR (> 125) and in-hospital ischemic stroke. Our results add to these data by not only investigating the association between FAR and ischemic stroke, but with TeC in general. In addition, our data reveal a very similar cutoff for FAR (= 130) so that this cutoff seems to be suitable. Finally, our study had a larger sample size (344 vs. 157).
Referring to patient characteristics of our study, overall in-hospital mortality rate was 58.1%. This is in line with previously published data 2 . Interestingly, mortality was not influenced by TeC (see Table 1). However, it is important to mention that similar mortality rates do not automatically mean that there was no life impact. Unfortunately, we cannot provide data on more patient-centered outcomes such as "days alive and out of hospital" 28 . Another remarkable finding was that duration of VA-ECMO was significantly different between patients with or without TeC (9.3 ± 6.9 vs. 6.7 ± 5.2). This aspect is underlined by the results of multivariate analysis, which also showed that the length of VA-ECMO therapy was independently associated with TeC (OR 1.08 [95% CI 1.04-1.13]. Thus, taken together, days of VA-ECMO seem to be a relevant risk factor to develop TeC.

Strengths and limitations
This study has several strengths. One strength of this study is that there was a high number of events (TeC: 117/344 = 34%) so that we could adjust for ten covariables in our multivariate logistic regression model (12 covariables post factum). Another strength is that all included patients were registered in a dedicated database, which ensured higher quality of our data. This study also has limitations: first, this is a retrospective, single-center cohort study. However, baseline characteristics and in-hospital mortality rate (58.1%) were in line with previously published data 2 , suggesting that both VA-ECMO indication and outcome in our centre might be representative for larger practice. Importantly, before drawing final conclusions, the predictive value of FAR for TeC should be investigated in prospective multicenter studies. Second, no cox regression could be performed as the exact time point of TeCs was not always documented in patients' medical records. Third, our database did not include reasons for initiation of VA-ECMO therapy. Although we know that all patients had refractory cardiogenic shock, we do not have information regarding the reason that led to cardiogenic shock.

Conclusions
In conclusion, this study shows that baseline FAR is independently associated with in-hospital TeC in patients undergoing VA-ECMO. Therefore, FAR might be used to support the prediction of TeC in this cohort. Future studies should validate these findings with a prospective design.

Data availability
The datasets generated during and/or analysed during the current study are available from the first author on reasonable request.